Jump to Content Jump to Main Navigation

7 Mergers

Gunnar Niels, Helen Jenkins, James Kavanagh

From: Economics for Competition Lawyers (2nd Edition)

Dr Gunnar Niels, Dr Helen Jenkins, James Kavanagh

From: Oxford Competition Law (http://oxcat.ouplaw.com). (c) Oxford University Press, 2015. All Rights Reserved. Subscriber: null; date: 27 October 2020

Oligopoly — Remedies in merger cases — Conglomerate effects — Horizontal mergers — Substantial lessening of competition test (SLC) — Unilateral effects — Vertical effects

(p. 297) Mergers

(p. 298) 7.1  Mergers Under Scrutiny

7.1.1  What’s the competition concern?

7.01  Mergers and acquisitions routinely make the financial headlines. They take place around the world and across different industries, and often involve big names and big numbers: Facebook acquired WhatsApp for $22 billion in 2014; Berkshire Hathaway and 3G Capital acquired H.J. Heinz Company for $28 billion in 2013; Inbev merged with Anheuser-Busch in a $52 billion transaction in 2008; Glaxo Wellcome merged with SmithKline Beecham in a $76 billion deal in 2000. Several of the world’s largest law firms—including Dentons, Norton Rose Fulbright, Freshfields Bruckhaus Deringer, and Hogan Lovells—have been through one merger or more to reach their current size. Mergers and acquisitions are part and parcel of a market economy. Companies are forced to consolidate or reposition when an industry goes through fundamental changes on the supply side or the demand side—for example, market liberalization, technological developments, changes in consumer preferences, or a recession. Mergers are often aimed at improving efficiency through economies of scale and synergies. Ambitious companies (and their managers) see mergers and acquisitions as a means to achieve rapid expansion or to enter new markets more quickly than through organic growth. Some mergers are defensive, protecting a market position in the face of new entry. Some are designed outright to eliminate competition and create market power.

7.02  Whatever the rationale for the merger, deal-makers nowadays are well aware that they may require clearance from the competition authorities. Over the years many high-profile deals have been scuppered on competition law grounds. For example, AT&T abandoned its plans to buy T-Mobile USA for $39 billion in 2011 when the DOJ challenged the deal.1 The European Commission blocked the $21 billion acquisition of Honeywell by General Electric in 2001 (even though the US authorities had cleared it), and the $9.5 billion merger between Deutsche Börse and NYSE Euronext in 2012.2 Around 130 jurisdictions now have a competition regime, typically including rules on merger control, from Albania and Barbados to Yemen and Zimbabwe.3 The need for regulatory clearance is now often built into the deal preparations and negotiations, sometimes with explicit provisions for what happens if competition authorities object to the merger. The more far-sighted among deal-makers also consider at an early stage which divestments or other remedies may need to be offered to the authorities.

7.03  Why are competition authorities concerned? Because a horizontal merger may eliminate competition between the merging parties, create market power for the merged entity, or dampen competition between the remaining suppliers in the market. A vertical merger may give a company control over bottlenecks in the supply chain. Across many jurisdictions, these concerns are now codified as a test asking whether the merger will lead to a substantial lessening of competition (SLC). The term SLC is used in the United States, Australia, New Zealand, and the United Kingdom. Canada and South Africa refer to a substantial prevention or (p. 299) lessening of competition (SPLC), and in the EU the test is phrased as significant impediment to effective competition (SIEC). In China the test is whether a merger eliminates or restricts competition. Yet the substantive test is similar across jurisdictions. How do you determine whether a merger will result in an SLC? One challenge is that the analysis is necessarily forward-looking. Competition authorities must assess how the market will evolve with and without the merger—that is, a combination of forecasting and counterfactual analysis. This is where economics comes in.

7.1.2  Other policy questions

7.04  Before turning to the economics of merger analysis, it is worth exploring a number of other policy questions in relation to merger control. First, why not simply prohibit all mergers? Companies that want to grow would have to do so organically, earning success on their own account. Such a policy would stimulate healthy competitive effort, and save enforcement costs. Yet outright prohibitions of mergers are relatively rare. Competition authorities recognize that mergers and acquisitions are an inherent part of how markets work and often have efficiency rationales. Instead of a prohibition, what often happens is that mergers that risk reducing competition are cleared with remedies to address the competition concerns, often in the form of divestments.

7.05  A second policy question is whether there should be a limit on how much companies can grow through mergers. Increasing consolidation has long been of concern. In 1966, during a wave of conglomerate mergers in the United States, humorist Art Buchwald wrote the following in a newspaper column (which was reproduced that same year in a Supreme Court judgment concerning a brewery merger):

It is 1978 and by this time every company west of the Mississippi will have merged into one giant corporation known as Samson Securities. Every company east of the Mississippi will have merged under an umbrella corporation known as the Delilah Co. It was inevitable that one day the chairman of the board of Samson and the president of Delilah would meet and discuss merging their two companies . . .

The Antitrust Division of the Justice Department studied the merger for months. Finally the Attorney General made his ruling. ‘While we find some drawbacks to only one company being left in the United States, we feel the advantages to the public far outweigh the disadvantages. Therefore, we’re making an exception in this case and allowing Samson and Delilah to merge.’4

7.06  Consolidation has not turned out to be as inevitable as feared in past decades. Companies have discovered that growth through mergers has its limitations, conglomerates have become less popular, and divestments of businesses are now almost as common as acquisitions. In any event, it has been generally accepted that competition authorities should not judge mergers and acquisitions by their size, but rather by reference to their effects on competition.

7.07  A third policy question is whether competition authorities should take into account the fact that mergers often fail. Received wisdom in the business literature seems to be that merger (p. 300) and acquisition activity has an overall success rate of at most 50 per cent. Several economics and finance studies have shown that mergers on average destroy more value for shareholders than they create.5 Case studies abound of post-merger integration failing due to a range of strategic and cultural differences (famous examples are Daimler Benz/Chrysler, AOL/Time Warner, and Mattel/The Learning Company). From this perspective, by prohibiting a merger the competition authority might in fact be doing the merging parties and their shareholders a favour. Yet the merger rules are not normally applied with such paternalistic considerations in mind. While the question of whether a merger is likely to succeed or fail is in principle not considered during the competition assessment, to the extent that efficiencies from the merger become an important part of the assessment, the authority will need to judge how realistic it is that the claimed efficiencies will be achieved. The prevalence of merger failures may have generated some scepticism among competition authorities about such claims (we turn to the assessment of merger efficiencies later in this chapter).

7.1.3  The remainder of this chapter

7.08  If you got to this chapter having read Chapters 2 and 3 you will already have a good understanding of many of the economic concepts used in merger cases: market definition, market concentration, barriers to entry, and countervailing buyer power. In Chapter 8 we deal with the design of merger remedies. In this chapter we explain the additional economic principles and analyses that are of relevance to merger control. We start in section 7.2 by exploring the principles behind the SLC test and its variants—which types of competition concern are addressed, and should the focus be on consumer welfare or total welfare? We go on in section 7.3 to discuss counterfactual analysis, which is used to assess what would happen in the absence of the merger. This includes situations where one of the merging parties is a failing firm. In section 7.4 we discuss unilateral effects, in particular the analysis of closeness of competition. We explain the diversion ratio as an indicator of closeness of competition, and describe best practice in designing consumer surveys to obtain diversion ratios. In section 7.5 we continue with unilateral effects and discuss the various techniques for simulating price effects from mergers. These range from simple illustrative price rise analysis to full merger simulation. Section 7.6 deals with co-ordinated effects. We explain static and dynamic oligopoly theory and the circumstances in which tacit collusion can arise. Section 7.7 explores the theories of harm arising in non-horizontal mergers, in particular vertical mergers, but also those that are diagonal or have portfolio effects. In section 7.8 we discuss the assessment of minority shareholdings, which sometimes raise competition concerns. Finally, in section 7.9 we deal with merger efficiencies: how they arise, how they can be measured, and in which situations they benefit consumers sufficiently to offset any anti-competitive effects of the merger.

7.2  The Substantive Test: SLC, SIEC, and Other Variants

7.2.1  Dominance, and unilateral and co-ordinated effects

7.09  The SLC, SIEC, and other substantive tests have evolved over time. There have been extensive debates about the scope of these tests. US antitrust law established one of the earliest (p. 301) frameworks for merger control. Section 7 of the Clayton Act of 1914 prohibits mergers where ‘the effect of such acquisition may be substantially to lessen competition, or to tend to create a monopoly’. Thus there are two criteria for judging whether a merger is anti-competitive: it gives the merged entity market power, or it substantially lessens competition. US merger control has traditionally placed great importance on concerns about tacit collusion or co-ordinated effects between the remaining competitors in the market. Such co-ordinated effects are difficult to address directly under competition law (in the absence of explicit collusion; see Chapter 5). Merger control is a tool to prevent market structures that are prone to co-ordination. In contrast, EU merger control originally focused on the first criterion, that is, the creation or strengthening of a dominant position. This led to some confusion during the 1990s as to whether situations of tacit collusion were also covered by the EU merger rules. The answer was yes, and in a number of cases the EU courts confirmed that concerns about tacit collusion could be made part of the dominance test by calling it collective (or joint) dominance.6 The SIEC test that was adopted under the 2004 EU Merger Regulation and accompanying Commission Guidelines captures co-ordinated effects explicitly.7

7.10  A dominance or monopoly test focuses on the market position of the merged entity. The SLC and SIEC tests allow for a wider range of competition distortions to be considered, covering both co-ordinated and unilateral effects. The distinction between co-ordinated and unilateral effects is not clear-cut, either in theory or in practice. In merger control it is usually interpreted as follows. Co-ordinated effects involve both the merged entity and the remaining competitors, and refer to situations in which the post-merger market structure is such that these companies can more easily engage in tacit collusion. Unilateral effects, in contrast, result directly from the loss of competition between the merging parties. With this direct rivalry removed, the merged entity can profitably raise its own prices, regardless of the response of its remaining competitors. In reality such price increases are not purely unilateral, as suppliers do not normally ignore their competitors’ responses. Yet it can be useful to make the distinction between co-ordinated and unilateral effects, as it requires the competition authority to be explicit about the theory of harm from the merger—is it co-ordination between the remaining suppliers, or a unilateral price increase by the merged entity? The theoretical distinction—which is also not pure—is that unilateral effects are based on static oligopoly theory, where prices and outputs change from one competitive outcome to the next as the number of suppliers is reduced. Co-ordinated effects relate to dynamic oligopoly theory, where suppliers seek to avoid competitive outcomes by tacitly colluding.

7.11  The creation of dominance through a merger is one type of unilateral effect, but the concept is broader. Unilateral effects may arise, for example, if the merger is between the second- and third-largest competitors in the market and these two competed more vigorously with each other than with the market leader. This type of unilateral effect arose in the Heinz/Beech-Nut case in 2001, which we discuss in section 7.4.8 Such mergers were (p. 302) believed to have been outside the scope of the old dominance test in EU merger control (this was termed the ‘unilateral effects gap’), but now come under the remit of the SIEC test. Another type of merger that is caught under the SLC and SIEC tests but not necessarily the dominance test is where a large company acquires a small but very aggressive competitor. An example is the acquisition of tele.ring by T-Mobile Austria in 2006.9 This deal combined the second- and fourth-largest mobile operators in Austria. At 30–40 per cent, their post-merger market share would be smaller than that of the largest provider, Mobilkom, which had 35–45 per cent. tele.ring itself had a market share of 10–20 per cent. According to the HHI (explained in Chapter 3), the market was concentrated and the increase in the HHI by 500–600 points would be enough to raise concerns under the SIEC test. The European Commission found that the merged entity would not be dominant because of Mobilkom’s presence, and nor would the nature of the mobile telephony market be conducive to tacit collusion between T-Mobile and Mobilkom. The Commission did, however, establish an SIEC on the basis of tele.ring’s competitive behaviour before the merger. It noted that ‘tele.ring, as a maverick, has a much greater influence on the competitive process in this market than its market share would suggest’.10 Using an aggressive entry strategy, tele.ring had cut prices and doubled in size in just a few years. The merger would reduce the intensity of competition by integrating the maverick competitor into one of the larger incumbents. The deal was eventually cleared, with remedies in the form of divestments of spectrum and mast sites. The aim of these remedies was to allow H3G, one of the other smaller operators, to grow, and thereby to reproduce the pre-merger competitive dynamics.

7.2.2  Consumer or total welfare?

7.12  Another aspect of the substantive merger test is whether it should focus on consumer welfare or total welfare. As we discussed in Chapter 1, there can be a tension between protecting consumer welfare and using a total welfare standard that encompasses producers and consumers. This is most clearly seen in the treatment of merger efficiencies. If weight is attached to producer surplus as well as consumer surplus, merger efficiencies are regarded as beneficial in and of themselves. With a consumer welfare standard, only those efficiencies that are passed on to consumers are considered relevant. This can make an important difference. Figure 7.1 presents a (somewhat hypothetical) situation of a merger to monopoly from a starting point of perfect competition.

Figure 7.1  Efficiencies versus market power in mergers

7.13  The price P1 equals marginal costs in the starting situation. The merger generates efficiencies: marginal costs decrease. Yet because the merger also gives market power to the new entity, the price increases in the end, from P1 to P2, which is the new profit-maximizing price. From a total welfare perspective, only the triangle marked ‘welfare loss’ would be of concern in this merger—recall from Chapters 1 and 2 that this is the deadweight welfare loss resulting from market power. You can undertake a simple welfare analysis based on this figure (as first introduced by Williamson, 1968). The welfare loss must be balanced against the efficiencies rectangle marked ‘welfare gain’. If the area of the welfare loss triangle is greater than the welfare gain rectangle, the merger harms total welfare. If the efficiencies rectangle is larger, total welfare increases. However, the result is different if the authority (p. 303) applies the consumer welfare standard. The concern would be that the price increases post-merger despite the cost efficiencies, and that there is a welfare transfer from consumers to the merging companies—the upper rectangle in Figure 7.1. In this framework the fall in marginal cost would have to be very large for there to be no net price increase after the merger. Otherwise, the efficiency gains would be considered insufficient, even if they compensated for the deadweight welfare loss in the triangle. This means that the consumer welfare standard can result in the blocking of mergers that would lead to higher prices but that would also ultimately enhance total welfare.

7.14  Most competition authorities place greater emphasis on consumer welfare than on total welfare. Canada is an exception. Under the Competition Act 1985, two standards can be applied. The first is the total surplus standard, which involves quantifying the deadweight welfare loss (the triangle in Figure 7.1). Equal weight is given to producer and consumer surplus. Any welfare transfers from producers to consumers are considered to be neutral. The second is the balancing weights standard, under which the effect on consumer surplus may be given greater weight (in the extreme, this test becomes the same as the consumer welfare test). The burden of quantifying the anti-competitive effects lies with the authority. In 2015 the Canadian Supreme Court ruled on a merger between three of the four secure landfill sites for hazardous waste in Northeastern British Colombia.11 The Competition Bureau, Competition Tribunal, and Federal Court of Appeal all found this merger to be anti-competitive. However, the Supreme Court determined that the Bureau had not properly quantified the anti-competitive effects of the merger—a tentative estimate had suggested a 10 per cent price increase, but was not carried out in a robust and verifiable manner. The court determined that the anti-competitive effects should therefore not be given any weight in the welfare trade-off. This meant that the efficiencies that had been quantified by the merging parties outweighed the anti-competitive effects. The original decision to block the merger was therefore overturned.

(p. 304) 7.3  The Counterfactual: Current Market, Entry, or Failing Firm?

7.3.1  What would happen without the merger?

7.15  Counterfactual analysis plays an important role in different areas of competition law—in horizontal and vertical agreements to assess whether there is a restriction of competition (see Chapters 5 and 6); and in the quantification of damages (Chapter 9). In merger cases, the idea of a counterfactual is to isolate and assess the specific effects of the transaction. You need to compare two hypothetical situations: the market outcome that is likely to result from the merger (which requires forecasting), and the market situation as it would be in the absence of the merger (the counterfactual). The difference between the two tells you whether there is an SLC post-merger. Fortunately, in most cases you can take as the counterfactual the conditions of competition prevailing before the merger. In these cases the SLC assessment boils down to considering the likely impact of the merger on the market as it currently stands.

7.16  However, the current situation is not always a good guide for what would happen in the absence of the merger. There can be potential imminent changes in the market structure. The EU Horizontal Merger Guidelines recognize this:

In assessing the competitive effects of a merger, the Commission compares the competitive conditions that would result from the notified merger with the conditions that would have prevailed without the merger. In most cases the competitive conditions existing at the time of the merger constitute the relevant comparison for evaluating the effects of a merger. However, in some circumstances, the Commission may take into account future changes to the market that can reasonably be predicted. It may, in particular, take account of the likely entry or exit of firms if the merger did not take place when considering what constitutes the relevant comparison.12

7.17  We consider two situations where future competitive conditions without the merger differ from the prevailing conditions: one in which new entry is expected, and one in which the target may exit (i.e. it is a ‘failing firm’).

7.3.2  Entry in the absence of the merger: Anti-viral drugs, steel drums, and concert tickets

7.18  The following cases illustrate how the counterfactual may differ from the pre-merger situation as a result of changing market dynamics and entry. In the Glaxo Wellcome/SmithKline Beecham merger in 2000, the European Commission explored whether the counterfactual should include products that were expected to enter the market in the short term regardless of the merger:

In the pharmaceuticals industry, a full assessment of the competitive situation requires examination of the products which are not yet on the market but which are at an advanced stage of development (normally after large sums of money have been invested). These products are called pipeline products.13

(p. 305) 7.19  This new entry in the counterfactual is subtly different from the general entry analysis that you carry out in the market power stage (see Chapter 3), since it is entry that does not depend on, or is not triggered by, the merger. In Glaxo Wellcome/SmithKline Beecham this issue arose in relation to anti-viral medicines. Although there was a pipeline of products from other companies, the Commission did not consider this to be sufficiently strong to constrain the merged entity. Indeed, there were concerns that the merger would make it more difficult for the pipeline products to achieve their market potential. In the end the parties agreed to make divestments in this product category.

7.20  In 2007 the CC investigated the completed acquisition by Greif—the largest manufacturer of industrial steel containers in the United Kingdom—of the steel drum and closures business of Blagden Packaging Group.14 The merged entity’s market share in large steel drums was 85 per cent. The CC found that, of all the competitors in the relevant market pre-merger, Blagden had posed the most significant competitive constraint on Greif. The CC’s analysis of switching showed that Greif and Blagden had lost more customers to each other than to any other competitor. Furthermore, there was limited competitive constraint from smaller producers or imports. However, this pre-merger market situation did not tell the whole story. The CC found that the merged entity would be constrained in future by a new steel drum manufacturing line run by Schütz Group, a competitor, at its new industrial packaging facility in the Netherlands. Schütz Group stated that the manufacturing line would have been installed in the absence of the merger—i.e. in the counterfactual—but that it brought forward the installation by eighteen months because of the merger. This new rival plant was capable of supplying large orders from UK customers, and the CC’s analysis showed that imports from Schütz Group could be competitive for some types of steel drum and for some UK customers. The merger was therefore cleared without remedies.

7.21  The third example is the 2010 merger between Ticketmaster Entertainment, a ticketing agent, and Live Nation, a venue owner and promoter. The deal was scrutinized in both the United States and United Kingdom. The CC cleared the deal unconditionally, but had initially raised concerns in relation to horizontal and vertical effects.15 This was predominantly a vertical merger in the supply chain for live music events. One theory of harm specific to the United Kingdom was that the merger would have a horizontal effect through a loss of potential competition in the supply of tickets for live music events. Before the merger, Live Nation had entered into an agreement with another ticketing agent, CTS Eventim, a strong player in the German market but new to the United Kingdom. The potential horizontal harm was that the merger could prevent CTS’s planned entry, as Live Nation might use only Ticketmaster as a partner going forward. Hence the counterfactual was not the pre-merger situation but the future market with competition from CTS, in partnership with Live Nation. The CC carried out a detailed analysis of this counterfactual, assessing whether the contract with Live Nation was a necessary stepping stone for CTS to enter the UK market. It also explored whether the merger would remove the rationale for that contract (both Live Nation and CTS publicly expressed their intention to honour the contract irrespective of the former’s merger with Ticketmaster). Live Nation’s ultimate commercial aim was to sell tickets directly to consumers using its own website, powered by CTS’s ticketing (p. 306) platform. As part of the agreement, CTS would be allocated a proportion of the tickets to events promoted by Live Nation. However, this in itself would not be sufficient to make CTS capable of constraining the two larger companies, Ticketmaster and See Tickets. The CC noted that CTS becoming a large-scale ticket retailer in the United Kingdom would depend mainly on its own efforts and abilities. The agreement between Live Nation and CTS was neither necessary nor sufficient to ensure the successful large-scale entry of CTS, and therefore there was no SLC from the merger in this regard.

7.3.3  The failing-firm defence

7.22  The failing-firm (or exiting-firm) defence can be relevant when the target firm is in distress. If in the counterfactual it would go out of business and therefore no longer be a competitor to the acquirer, you may conclude that the merger changes little about the competitive landscape. However, the fact that a target company would be unlikely to survive is not in itself sufficient. Competition authorities also consider what would happen to its productive assets. The failing-firm defence applies if the proposed merger is the only, or least anti-competitive, way of maintaining the assets of the failing firm in the market. The US Merger Guidelines highlight the importance of the final whereabouts of the failing firm’s assets:

[A] merger is not likely to enhance market power if imminent failure . . . of one of the merging firms would cause the assets of that firm to exit the relevant market. This is an extreme instance of the more general circumstance in which the competitive significance of one of the merging firms is declining: the projected market share and significance of the exiting firm is zero. If the relevant assets would otherwise exit the market, customers are not worse off after the merger than they would have been had the merger been enjoined.16

7.23  In the absence of the merger, would the target firm really exit the market? Would its assets be bought by another firm? Would it not recover and become a viable business again? As noted by one authority, failing-firm counterfactuals ‘are easily the subject of self-speculation—relatively easily alleged but difficult, given the informational asymmetries, to verify independently’.17 Not surprisingly, therefore, the requirements for qualifying as a failing firm are stringent. Most jurisdictions explicitly spell out the necessary conditions for a successful application of the failing-firm defence. While the detailed conditions may differ, they can broadly be grouped under the following two criteria: exit of the target business and its assets is inevitable in the near future; and there is no realistic and less anti-competitive alternative outcome than the proposed merger.

7.24  The first criterion—exit is inevitable in the near future—is usually the result of businesses in financial difficulties, although it can also be due to a change in corporate strategy (e.g. an otherwise healthy company pulling out of a particular market). Evidence on a company’s viability should be evaluated in the context of the prevailing economic and market conditions—for example, a recession affects both the financial health of the target and the availability of alternative purchasers. Furthermore, it should be shown that the assets of the target business will exit the market as well, and that the financial difficulties cannot be easily rectified by restructuring the business. This will involve examining the company’s (p. 307) accounts, board minutes, and strategic plans. Another way of looking at this is that the evidence must show that there is no investor willing to provide the necessary capital for the business to remain a going concern.

7.25  This analysis can be based on a number of well-established financial metrics and ratios. A common metric is the estimated likelihood of default as reflected in the credit rating of the business. Credit rating agencies use financial tests to assess the risk of creditors not receiving the interest and principal in a timely manner, as these factors typically lead to default. As an example, in one survey AA-rated debt had a default rate of only 0.03 per cent, while for CCC-/C-rated companies this was 25.7 per cent (Altman and Narayanan, 1997). Other metrics relate to profitability, liquidity, and solvency. The relevant profitability measure would reflect a firm’s ability to generate economic value by realizing the returns required by debt and equity investors. For liquidity, the relevant financial ratios indicate a firm’s ability to meet its short-term obligations, such as interest payments, from current cash flows. An illiquid but potentially solvent firm (i.e. one that cannot obtain funding for short-term liabilities, but whose overall liabilities do not outweigh its assets) might not be able to survive as a going concern if investors are unwilling to commit additional funding. For solvency, the relevant measures would capture a firm’s ability to repay its fixed financial obligations, including the principal on its debt (the credit rating is relevant here).

7.26  As noted above, the rationale for the failing-firm defence is that the proposed merger allows productive assets to be retained in the market in a way that does not substantially impair competition. The second overall criterion requires that there is no realistic and substantially less anti-competitive alternative means of achieving this. If in the absence of the merger the target business would have been sold to one or more other purchasers (even if potentially for a lower price), this would represent a realistic counterfactual. In some cases there will have been a formal bidding process to sell the failing business, which the authorities can use to assess the most likely counterfactual. These other purchasers could be existing competitors or new entrants. In the case of a sale to a new entrant, the counterfactual is likely to be similar to the pre-merger competitive conditions (the assets in question simply change hands, with no change in the number of competitors). If instead the assets are dispersed across several existing firms, this could suggest a counterfactual with one less competitor, but with each remaining competitor getting slightly stronger (with more assets). To satisfy this second overall criterion, the merging parties need to show either that these alternatives are not feasible, or that the competitive situation will not be better in these scenarios.

7.3.4  Failing-firm examples: From solvents to Greek airlines

7.27  A European Commission case where the failing-firm defence was successful is the acquisition in 2001 by BASF of Eurodiol and Pantochim.18 These were two Belgian chemicals producers owned by the SISAS Group from Italy, and both the subsidiaries and parent company were in the process of liquidation. Even though this merger enhanced BASF’s strong position in a number of solvents markets, the Commission considered that it met the conditions of the failing-firm defence. The target companies were going to exit the market and, despite the best efforts by the Tribunal de Commerce of Charleroi to find suitable buyers under the Belgian pre-bankruptcy regime, only BASF had made an offer. The (p. 308) criterion regarding the exit of the assets of the target business was also satisfied since the two plants operated as a single unit and could not have been sold independently. Hence the Commission concluded that these assets would have exited the market had BASF not made the offer to acquire Eurodiol and Pantochim.

7.28  A high-profile case where the failing-firm defence was ultimately rejected is the acquisition by Lloyds TSB of HBOS in 2008.19 Lloyds TSB was one of the large UK banks. HBOS also had a significant presence in retail banking, itself the product of an earlier merger between a large building society, Halifax, and the Bank of Scotland. HBOS was in distress as a result of liquidity issues and exposure to toxic assets that reached a critical point at the start of the global financial crisis. Aware of the potential systemic risks (one bank failure leading to others), the Secretary of State for Business intervened in the public interest, taking control of the ultimate decision of whether to approve the merger. He asked the OFT for a report on the competitive effects of the transaction. The OFT found that the HBOS acquisition resulted in an SLC that was not easily remediable. Central to this conclusion was the chosen counterfactual. The OFT agreed that using pre-merger competitive conditions was not appropriate given the unprecedented volatility in the financial markets. However, it considered that in the counterfactual the competitive constraints from HBOS would not be completely eliminated. Rather, the OFT judged the appropriate counterfactual to be that the government would step in to ensure that HBOS did not collapse (as indeed the government did soon thereafter with Lloyds TSB itself). In the medium term, when that support was no longer required, the government could sell HBOS to another bank with no existing strong position in the United Kingdom, or return it to viability. Thus the OFT did not consider this case to meet the requirements for treating HBOS as a failing firm with no alternative, less anti-competitive options. Nevertheless, the acquisition was cleared by the Secretary of State on the basis of its benefits to the financial stability of HBOS and the UK financial system as a whole.

7.29  In a bus merger in the United Kingdom in 2009, the failing-firm argument was acknowledged but not considered to be the relevant counterfactual because it was alleged that the target was failing due to predatory actions by the acquirer.20 The CC found that there had been an SLC from Stagecoach’s completed acquisition of Preston Bus, a rival bus operator in Preston in the north of England. The authority required Stagecoach to divest a reconfigured Preston Bus business. The decision was appealed to the CAT, which upheld the CC’s finding of an SLC, but questioned the counterfactual.21 The background to the case was that Stagecoach first approached Preston Bus in 2006 with an offer to purchase the business. The offer was rejected. Following this, Stagecoach developed a plan for expansion in the Preston area, and launched a number of local bus services in direct competition with Preston Bus. Both Stagecoach and Preston Bus incurred losses as a result of this competition. After a year of unprofitable services, Preston Bus suffered financial difficulties, and decided to sell. It approached a number of major bus companies. Stagecoach made an (p. 309) offer in October 2008 which Preston Bus accepted. Stagecoach submitted that Preston Bus was a failing firm and that, on liquidation, it was highly unlikely that any other company would purchase its assets to provide local bus services in Preston. The CC concluded that the counterfactual should be based on the most recent period of ‘normal’ competition, that is, before Stagecoach entered. Therefore, the counterfactual would be that of Preston Bus profitably running the local services and Stagecoach running inter-urban services outside Preston—in essence, the situation that existed about eighteen months before the merger. The CAT rejected the CC’s decision to ignore the events of the June 2007–September 2008 period for the purposes of constructing the counterfactual. The CAT agreed with the CC that the counterfactual would not have been a complete exit by Preston Bus. However, it concluded that the alternative counterfactual put forward by the CC—the competition prevailing before Stagecoach’s entry—was incorrect. As a result the CAT found the CC’s remedy to be disproportionate.

7.30  A failing-firm case that did get accepted by the UK authorities was Optimax/Ultralase in 2013, involving providers of refractive eye surgery.22 The merging parties were the second- and third-largest competitors in the market, with Optical Express being the market leader with national presence, and Optegra and Accuvision as other notable providers. Refractive eye surgery is often considered a luxury purchase, and hence demand for such services fell considerably during the recession. This drop in demand, coupled with the significant installed laser eye treatment capacity in the United Kingdom, meant that many clinics were under-utilized. There was ample evidence that Ultralase was expected to run into funding problems towards the end of 2012, and that its previous restructuring programmes had been unsuccessful. The sale process of Ultralase started in April 2012, and a number of potential buyers expressed interest. However, Ultralase received only one offer in the end, from Optimax. Faced with the alternative prospect of going into administration, the company accepted this offer. Economic modelling of Ultralase’s likely future profitability in a range of scenarios showed that a private equity or other non-trade purchaser would not be able to achieve a sufficient return to justify the initial investment. The CC confirmed that there was no credible alternative purchaser that would have acquired Ultralase as a going concern, or its assets on a piecemeal basis. As to other possible alternatives, the CC carried out a consumer survey which found high diversion ratios from Ultralase to both Optical Express, the largest supplier, and Optimax, the acquirer. As a result, it concluded that Ultralase’s exit would not have led to a substantially less anti-competitive outcome than the merger, as Optical Express and Optimax would have picked up most of Ultralase’s customers. Note the somewhat counterintuitive implication of this last part of the test: high diversion ratios between the acquirer (Optimax) and the target (Ultralase) are normally of concern as they indicate closeness of competition, but here they were actually favourable to the merger as they indicated that exit by Ultralase would not result in a much more competitive situation than the merger itself.

7.31  Prevailing economic conditions also played a role in the European Commission’s approval in 2013 of the acquisition of Olympic Air by its main domestic rival in Greece, Aegean Airlines, on failing-firm grounds.23 This was less than three years after the Commission had (p. 310) blocked the same deal on competition grounds, rejecting the failing-firm defence.24 The merger would lead to a quasi-monopoly on the main domestic routes in Greece. In the 2011 decision the Commission did not accept that Olympic would inevitably exit the market, and considered that some form of restructuring or less anti-competitive sale was still possible. However, in the 2013 decision the Commission approved the renewed bid by Aegean without conditions. Olympic was facing even greater financial difficulties, as was the Greek economy as a whole. Domestic air travel had fallen by 26 per cent, several routes had been closed, and there seemed to be no other credible purchaser of the assets. Accordingly, the Commission accepted the failing-firm defence this time round.

7.4  Unilateral Effects: Assessing Closeness of Competition

7.4.1  Homogeneous versus differentiated products

7.32  If the products of the two merging companies are reasonably homogeneous, adding up their market shares can give a useful indication of the strength of the merged entity. The assessment of mergers in homogeneous product markets focuses on market definition, market shares, and entry barriers, in line with the principles set out in Chapters 2 and 3. The greater the concentration resulting from the merger and the higher the entry barriers, the more likely it is that there will be a competition case to answer. This traditional framework for merger analysis is still widely used internationally.

7.33  In most markets there is a degree of product differentiation. We discussed in Chapter 2 that market definition works best as an intermediate tool for merger analysis when products are homogeneous. You can test whether one group of products competes with another (e.g. apples with bananas, or cellophane with other wrapping materials). With product differentiation there is a spectrum of products that are close but imperfect substitutes. Drawing the line between products that are inside and outside the relevant market, and adding up market shares, may be less informative in these circumstances. Yet as we also noted in Chapter 2, market definition can still be useful in differentiated product markets. In many of these markets you can find meaningful groups of products that are reasonably similar, such that you can group them together and hypothetically monopolize them to test the competitive constraints from other groups—for example, top-of-the-range sports cars, high-quality chocolates, or children’s breakfast cereals. An element of judgement will be required. If the degree of product differentiation is very high you may be better off focusing directly on the competitive effects of the merger and place less emphasis on market definition. This is where unilateral effects analysis comes in.

7.34  Unilateral effects analysis involves an assessment of expected post-merger prices. Would the merger give the new entity additional incentives and ability to raise prices? This depends on whether the merging parties are close competitors. The closer they are to each other on the spectrum of differentiated products or geographic areas—e.g. they produce very similar top-of-the-range sports cars, or their stores are located very near to each other—the more intense the competition between them, and hence the higher the likelihood of a significant price increase after the merger. The key measure of closeness of competition used in merger analysis is the diversion ratio. We saw the diversion ratio in Chapter 2 in the context of (p. 311) market definition, where it is used to rank substitutes for inclusion in the hypothetical monopolist test. In this section we explain the main economic features of the diversion ratio, and how to estimate it in practice.

7.35  First, however, we show two examples of merger cases where the closeness of competition turned out to be more important than the position of the merged entity in the overall market. One is Heinz’s proposed acquisition of Beech-Nut in the market for baby food in the United States.25 Heinz and Beech-Nut were the second- and third-largest suppliers in the US baby food market, far behind market leader Gerber. The Court of Appeals upheld the FTC’s injunction to halt the transaction (thereby rejecting an earlier district court judgment which had overruled the FTC). Although the merging parties were relatively small—Heinz had a 17.4 per cent market share; Beech-Nut 15.4 per cent—the FTC found that each was the other’s closest competitor. Competition between the parties mainly took place in vying for shelf space at retail outlets, including offering lower prices and promotions. Most retailers stocked two baby food product ranges—Gerber, plus either Heinz or Beech-Nut. Gerber’s products were sold in over 90 per cent of all supermarkets in the United States, while for Heinz this was only 40 per cent and for Beech-Nut 45 per cent. The competition between the two smaller rivals led to pressure on Gerber’s pricing as well. The FTC was concerned that this competition would be removed after the merger. In addition, entry barriers were found to be high. The court concluded that:

the merger will eliminate competition at the wholesale level between the only two competitors for the ‘second shelf’ position. Competition between Heinz and Beech-Nut to gain accounts at the wholesale level is fierce with each contest concluding in a winner-take-all result …

Heinz’s own documents recognize the wholesale competition and anticipate that the merger will end it. Indeed, those documents disclose that Heinz considered three options to end the vigorous wholesale competition with Beech-Nut: two involved innovative measures while the third entailed the acquisition of Beech-Nut. Heinz chose the third, and least pro-competitive, of the options.26

7.36  The issue of the closest competing products also arose in LOVEFiLM’s acquisition of Amazon’s online DVD rental service in the United Kingdom.27 Online DVD rental involved customers paying a fixed monthly fee entitling them to select and receive DVDs via the post. This was a relatively new type of business model at the time (although streaming of films would overtake this model within a few years). The merged entity would have a share of online DVD rentals of 92 per cent. Despite this strong position, the OFT approved the merger unconditionally. Survey evidence showed that customers saw other methods of accessing DVDs as close substitutes, in particular renting DVDs from bricks-and-mortar shops (now virtually extinct) and downloading films. Internal documents indicated that the merging parties regularly monitored activities relating to these other products, as opposed to monitoring each other, thus confirming the competitive constraint from these other products. This evidence was consistent both with online DVD rentals forming part of a wider relevant market, and with each merging party not being the other’s closest competitor. Either way, the conclusion was that no SLC would result from the merger.

(p. 312) 7.4.2  Diversion ratios as a measure of closeness of competition

7.37  Diversion ratios seek to capture a basic economic logic: if the price of a product is raised, some consumers will switch away from that product and spend their money elsewhere. The alternative product that captures most of this diverted expenditure will have the highest diversion ratio, and can be regarded as the closest substitute of the first product. In general, the higher the diversion ratio between two products, the greater the competition is between them, and the more concerned you should be if they merge. The diversion ratio in its own right gives useful information on closeness of competition. It is also the key parameter used in the simulation of price effects of mergers, as discussed in section 7.5.

7.38  There are various definitions of the diversion ratio. The differences between these definitions are subtle, and sometimes do not matter. We illustrate them through a numerical example, which we try to keep simple by using the ubiquitous apples and pears (based on ten Kate and Niels, 2014). As an indicator of the closeness of competition between products, any of the diversion ratios shown in this example will do a reasonable job. However, if the diversion ratio is used to simulate the price effects of a merger, not all these ratios are suitable, as we discuss later.

7.39  Say there are 10,000 customers in an area, buying on average 10 kg of apples per year, at €3 per kg. The total sales of apples amount to 100,000 kg, worth €300,000. Now the price of apples is increased by 10 per cent. This leads to 2,000 customers (20 per cent of all customers) ceasing to buy apples. 800 of them turn to pears, 600 to other fruits, and 600 give up consuming fruit altogether (i.e. they spend their money elsewhere in the economy). Assume that the 8,000 non-switching customers were buying 10.5 kg of apples on average before the price increase; the 800 customers switching to pears were buying 9 kg of apples on average; the 600 customers switching to other fruits 8 kg; and the 600 customers giving up fruit 6.67 kg. This leads to the situation shown in Table 7.1.

Table 7.1  Diversion ratios: an apples and pears example

Apple purchases

Before price increase

After price increase

Loss after price increase

Switching to pears

Switching to other fruits

Give up fruit








Initial apple sales to those customers (kg)







Initial apple sales to those customers (€)







New pear sales to those customers (kg)


New pear sales to those customers (€)


Switching-customers diversion ratio

800/2,000 = 40%

Lost-sales diversion ratio

7,200/16,000 = 45% (volume); 21,600/48,000 = 45% (value)

Volume capture ratio

6,400 kg of pears/16,000 kg of apples = 4 kg of pears/10 kg of apples

Value capture ratio

22,400/48,000 = 46.7%

(p. 313) 7.40  You may have worked out that the own-price elasticity of demand for apples is –1.6: the 10 per cent price increase resulted in a 16 per cent volume loss (16,000 out of 100,000 apples). However, here we are interested in diversion ratios. A frequently used definition illustrated in this example is the fraction of customers switching away from A that is diverted to B. Of the 2,000 apple customers who switch after the 10 per cent price increase, 800 turn to pears. So in this case the diversion ratio from apples to pears would be 800/2,000 = 40 per cent. We call this the switching-customers diversion ratio. Another way of defining the diversion ratio is in terms of loss of sales, either in volume or value terms. We call this the lost-sales diversion ratio. In volume terms this ratio is 7,200/16,000 = 45 per cent. The 800 apple customers switching to pears used to buy 7,200 kg of apples, while the total volume loss of apples is 16,000 kg. In value terms the lost-sales ratio is the same in this example, since apple prices are uniform at €3 per kg—that is, 21,600/48,000 = 45 per cent. The 800 switchers used to spend €21,600 on apples, while the total loss of apples sales is €48,000.

7.41  So far this example has considered diversion from the perspective of the apples. Now we turn to the pear side to define the diversion ratio. A more accurate term would be ‘capture ratio’, as it is about the pear side capturing a share of the lost apple sales. Assume that the 800 former apple customers buy on average 8 kg of pears at a price of €3.50 per kg (recall that they used to buy on average 9 kg of apples at €3 per kg). So they buy 6,400 kg of pears for a total amount of €22,400. If we define the diversion ratio as the volume capture ratio it is equal to 6,400/16,000 = 4 kg of pears/10 kg of apples. This cannot be expressed as 40 per cent, because (literally and technically) you cannot divide kilos of pears by kilos of apples. If we define the diversion ratio in value terms—the value capture ratio—it equals 22,400/48,000 = 46.7 per cent. The increased expenditure on pears is €22,400, while the total loss of sales of apples is €48,000. Altogether this example gives four different diversion ratios: a switching-customers ratio of 40 per cent, a lost-sales ratio of 45 per cent, a volume capture ratio of 4 kg pears/10 kg apples, and a value capture ratio of 46.7 per cent.

7.4.3  Which diversion ratio?

7.42  Which of these ratios is the right one to use? That depends on the purpose of the diversion ratio. As mentioned before, any of the four ratios will do a reasonable job as an indicator of the closeness of competition between products. In the above example they had broadly similar values. However, if the diversion ratio is used to simulate the price effects of an apples-and-pears merger (or for the hypothetical monopolist test), only the two capture ratios on the pear side are strictly correct. This is because you need to assess the effect of the price increase on the merged entity’s profitability. This depends on the relationship between lost profits and captured profits—that is, how do the lost profits from the price increase in apples compare with the profits captured by pears following the switching from apples to pears? Neither of the ratios defined on the apple side—the switching-customers ratio and the lost-sales ratio—is capable of establishing such a relationship. The ratios defined on the pear side do perform the task, as discussed further in section 7.5. Hence, the ratios of 4 kg pears/10 kg apples (volume capture ratio) and 46.7 per cent (value capture ratio) are the most relevant diversion ratios from apples to pears when considering likely merger price effects.

7.43  Still, in practice the switching-customers ratio (here 40 per cent) and lost-sales ratios (45 per cent) are often used. In part this is because these ratios are intuitive—they represent the fraction of apple customers or apple sales that is diverted away to the other product. In part it is also because these ratios may be easier to estimate. In consumer surveys—as discussed below—it is straightforward to obtain the switching-customers ratio and the lost-sales ratio by first identifying the total base of customers who would switch away from the product (2,000 apple customers), and their total expenditure (€48,000), and then to ask where they would switch to (800 to pears, 600 to other fruits, and 600 giving up fruit). While theoretically incorrect for price rise analysis, this practical approach can still yield meaningful insight if a number of conditions hold. One is that all customers who switch must reduce their purchases of the base product by the same amount. This condition was met in the example above, as all switching apple customers were buying the same amount of apples before the price increase, and switched away from apples completely. Another condition is that all switching customers buy the same quantity of the new product as they previously bought of the base product. In some circumstances this is plausible. If switching is between near-identical products at different geographic locations—think of ice-cream sellers on the beach, or petrol stations—it is not unreasonable to assume that a switching customer buys the same quantity of the product at the other location. The same may hold for switching between different brands of a product that is otherwise nearly homogeneous—for example 2-litre bottles of soft drinks of different brands. In the above example this condition was not met, as the 800 switchers used to buy 9 kg of apples each but now buy only 8 kg of pears. These and other conditions must hold in order for the switching-customers and lost-sales ratios on the apple side to be the same as the capture ratios on the pear side. When designing a survey, you should ideally try to obtain the right diversion ratio. However, if that makes the survey too complicated (capture ratios are more difficult to identify in one question), (p. 314) it is still useful to obtain the switching-customer or lost-sales ratio. The presentation of the survey results should state explicitly that these ratios are used as approximations of the capture ratios, and discuss why this assumption is reasonable in the case at hand.

7.4.4  Consumer surveys to estimate demand responsiveness and diversion ratios

7.44  Consumer surveys are a tool that can be used to estimate demand elasticities and diversion ratios. They can be used to ask the SSNIP question directly—how would customers react if prices were raised by a small amount? And they can be used to obtain information on diversion ratios—if prices were raised, or the product were no longer available, which products would customers switch to? Surveys focused on diversion ratios are often preferred over surveys which try to gauge consumer reactions to a hypothetical price increase. Instead of asking what consumers would do if faced with a SSNIP, a diversion survey asks what their second choice would be if the first choice were no longer available. Forced-choice questions of this type require (somewhat) less hypothetical thinking by the respondents than the SSNIP question, and can be asked to all survey respondents rather than just to those who said they would switch after a price increase, thus generating a larger response base. This allows you to infer diversion ratios, which you can use to simulate price rise effects resulting from the merger.

7.45  Surveys provide information on consumers’ ‘stated preferences’—that is, what they say they would do after a price increase or if forced to select an alternative product. This contrasts with ‘revealed preferences’ reflecting what they actually did (or would do). Clearly this raises an important question about the use of surveys: can you rely on the results to show how consumers would actually respond when faced with a price increase or forced choice between options? There is a good deal of literature on how to undertake consumer research in a robust manner. Some common understanding on best practice in the use of surveys for (p. 315) competition cases has developed. A significant contribution in this regard was the guidance document issued by the UK authorities in 2011.28 Surveys that are suitably designed and executed can provide useful evidence complementing that from other sources, such as internal documents showing which competitors a company focuses on most, or data on where new customers came from and lost customers went. As noted in the 2010 US Horizontal Merger Guidelines:

The Agencies consider any reasonably available and reliable information to evaluate the extent of direct competition between the products sold by the merging firms. This includes documentary and testimonial evidence, win/loss reports and evidence from discount approval processes, customer switching patterns, and customer surveys.29

7.46  The UK guidance puts it as follows.

Amongst other things, consumer survey evidence may be used for market definition or for the assessment of the closeness of competition between firms. We welcome this type of evidence and believe that the use of statistically robust consumer survey research can help us reach informed decisions.30

7.47  We give some flavour of what good practice looks like in the next sub-section. First we return to the LOVEFiLM/Amazon case, in which a consumer survey was carried out focusing on both market definition and diversion.31 The transaction created a near-monopoly in online DVD rentals. Hence the question was whether LOVEFiLM and Amazon were each other’s closest competitors, which in this case (effectively a two-to-one merger) was similar to asking whether the relevant product market was limited to online DVD rentals. Internet-based surveys were commissioned on behalf of the acquiring party, canvassing the views of approximately 2,000 customers of online DVD rental services. First, questions about current behaviour and choice were asked, such as which provider they currently rent online DVDs from, and how much they pay per month. The survey then asked the SSNIP question: say that your online DVD provider decided to increase its prices, so that instead of paying £x per month, you had to pay 10 per cent more, that is, £yx + 10 per cent) per month to receive the same service, what would be your most likely response? The ‘x’ was shown as the price that the respondent actually paid for online DVD rentals, as answered in an earlier question. The online survey question also showed what £y was in money terms.

7.48  Finally, questions about switching and diversion were asked: were you to cancel your online rental contract, how would you meet your demand for films? The survey results showed that 30–40 per cent of respondents said they would switch after a 10 per cent price increase. Since the critical loss threshold of 20–30 per cent was below this actual loss, the relevant market was deemed to be wider than online DVD rental services. There was no clear nearest substitute. Instead, alternatives such as renting DVDs from bricks-and-mortar shops, (p. 316) downloading films, and watching fewer films, were all given frequent mentions by switchers. In the end, while there was some discussion about the interpretation of the survey results, the OFT cleared the merger because it found internal business and strategy documents confirming that the parties saw other products as their closest competitors, which was consistent with the market being broader than online DVD rentals.

7.49  Another case where diversion ratio evidence from consumer surveys played an important role is the Co-op/Somerfield (2008) supermarket merger, which was cleared at phase one by the OFT.32 The authority based its decision on what it described as ‘probably the largest consumer survey ever conducted in a merger case’.33 More than 40,000 customers were surveyed in over 400 Co-op and Somerfield stores. Customers were asked where they would do their grocery shopping if the store where the survey was conducted were to close (this was posed as a hypothetical closure, simply to get respondents’ views on what they considered to be their next-best alternative; Co-op did not actually intend to close one of the two stores after the merger). Higher diversion ratios between merging stores imply more intense competition between them pre-merger, and hence higher predicted price rises after the merger. On this basis, potential competition concerns were identified for 126 local markets, and divestment remedies were agreed for those markets.

7.4.5  Good practice in consumer survey design

7.50  The literature on marketing, psychology, and more recently, behavioural economics, provides useful lessons for the design of surveys.34 A survey used for the hypothetical monopolist test or diversion ratio analysis should normally be directed at existing customers of the product in question. Even so, there are various consumer biases that can raise doubts about the reliability of survey responses. If you have ever been asked to respond to a survey of this nature (by telephone, on the street, or online), you may agree that it helps if it is not too long and follows a logical structure. Framing the questions such that the real-world choice situation is approximated as closely as possible is of critical importance.

7.51  A survey must be representative of the relevant population in order to provide statistically meaningful results. Representativeness means that the sample has been selected randomly from the appropriate population. If you are looking for information on apple consumption, the whole population is probably reasonable. If it is underwired bras, you might sample randomly from women. You need similar proportions of geographic, income, and other demographic characteristics in your sample as exist in the underlying population. For example, data from a survey may be biased if it is conducted in a location that tends to have different types of people passing through it at different times. Carrying out a survey at a railway station on a weekday morning will yield a different sample of travellers (commuters) than if the same survey were carried out during the weekend (leisure travellers). Established statistical tests can help determine representativeness.

7.52  Having designed your sampling process to ensure a representative sample, the next task is to decide on the size of the sample so as to ensure that the results will be statistically valid. (p. 317) There are no hard and fast rules and there will always be a margin of error, but the larger the sample size, the narrower (and more acceptable) the confidence interval around the survey results.35 Rules of thumb have been developed from statistical theory, known as the law of large numbers, indicating that you ought to have at least thirty responses in the smallest group you wish to examine. To give an indication from the examples we saw earlier, in Co-op/Somerfield around 100 shoppers were surveyed at each of the 400 stores. The survey in LOVEFiLM/Amazon was among around 2,000 online DVD rental customers. This was considered sufficient to ensure at least thirty responses for most of the plausible diversions categories (e.g. pay-TV movies, bricks-and-mortar rental), while also allowing the results to be segmented by existing purchasing behaviour of the respondents (e.g. do they currently subscribe to pay-TV?).

7.53  Formulating the questions is probably the trickiest part of survey design. The language used should be clear, unambiguous and neutral. Complexity should be avoided—for example, many people have difficulties understanding percentage changes, and are more likely to understand questions about price increases if these are framed in terms of absolute amounts, or in both absolute and percentage terms. The survey should start by asking some ‘warm-up’ questions—about the respondents themselves, their current behaviour and habits, why the customer chose the given product, and whether alternative products were considered. These questions reveal certain characteristics of customers and help bring respondents towards the actual product choice situation that they would be in. Then you can ask the SSNIP or diversion questions. A SSNIP question for a customer who pays €3 for a kilo of apples could be phrased as follows: ‘Suppose that the price of apples increases by 10 per cent, from €3 per kilo to €3.30 per kilo, what would you do?’ In the first instance respondents could be given options capturing whether they would continue buying or switch (partially or fully), and the likelihood of doing so (e.g. a scale from very likely to continue buying to very likely to switch).

7.54  The next question is to ask the switchers which alternative they would buy instead—the diversion question. In merger cases where the focus is on unilateral effects rather than market definition, the diversion question is usually asked directly to all respondents, omitting the SSNIP question. It could be phrased in terms of forced diversion. In the supermarket example above, the wording was along the following lines: ‘If this store had not been available, which, if any, of these types of store would you have used instead?’ In some cases respondents could be shown a list of generic alternatives (e.g. a same-size supermarket, an out-of-town supermarket, and a convenience store), or a list of specific alternative stores in the area. The proportion of customers choosing a particular alternative provides the diversion ratio from the original store to that alternative. This is what we referred to as the switching-customers ratio, or, if weighed by the volumes or value of their purchases, the lost-sales ratio.

7.4.6  Conjoint analysis

7.55  Conjoint or discrete-choice surveys are a more sophisticated and potentially more robust type of survey. Instead of asking respondents what they would do in a hypothetical situation (p. 318) of a price increase or forced diversion, a conjoint survey asks them to choose among products with different characteristics, including price. This process more closely resembles the actual choice-making situation that customers may find themselves in. Given the time involved, conjoint analysis is not frequently used in the context of merger cases. It is more common in other types of competition inquiry. For completeness we briefly set out the basics of conjoint analysis here.36

7.56  A conjoint survey presents customers (or potential customers) with a menu of products that differ slightly from each other in their ‘attributes’, including price, functionality, and different aspects of quality. Two options are presented each time, and the respondent has to express a preference. By varying the product attributes in each option, a picture emerges of how customers trade off price and other attributes against each other. Econometric analysis can then be applied to the responses in order to estimate a price elasticity (or indeed to estimate an elasticity of demand with respect to any of the product attributes).

7.57  Conjoint analysis was used in Ofcom’s review of the pay-TV market in the United Kingdom.37 The regulator sought to understand whether channels containing premium content, such as Football Association Premier League (FAPL) matches, constituted a separate market. The conjoint survey sought to test the importance of different sports when deciding to subscribe to premium sports channels, and to evaluate the degree of substitution between sports. It covered 1,904 respondents who subscribed to pay-TV and watched sports at least once a week (85 per cent of this sample were male, which was apparently deemed representative of the sports viewing population). Respondents were presented with a series of pay-TV sports packages and in each case were asked which of two options they would prefer. Figure 7.2 shows a stylized example of one of these choice situations: two sports packages where the main differences are the price and whether they include FAPL matches. The survey, combined with other evidence, led Ofcom to (p. 319) conclude that channels containing premium football content did indeed constitute a separate market. Setanta Sports, then a new (but ultimately unsuccessful) entrant that had purchased a package of premium football rights, was found to be the closest competitor to Sky Sports, the incumbent channel provider. Channels without this premium football content were not seen as sufficiently close substitutes, even when offering lower prices and other attractive content.

Figure 7.2  Stylized example of the choices presented in a conjoint survey

Source: Based on the sports bundles conjoint survey in Ofcom (2008), ‘Pay TV second consultation—Access to premium content’, 30 September, Annex 10.

7.4.7  Other evidence on closeness of competition

7.58  There are other sources of information on closeness of competition besides consumer surveys. Internal strategy or board documents may give an indication of which of its rivals a company monitors or reacts to most. In some markets companies keep data on lost and gained customers, including where they came from or which competitor they switched to. Such data can be useful to construct a picture of which competitors target each other’s customers most. In bidding markets it can be useful to collect data on tender participation. How often was one of the merging parties a bidder in tenders won by the other?

7.59  The European Commission looked at several of these information sources in the Baxter/Gambro merger in 2013.38 The deal involved two producers of medical equipment, with an overlap in various renal replacement therapy products (used for kidney failures). At the European level, the parties would have a combined market share of 30–40 per cent in a group of products for haemodialysis (HD) treatments, and in some countries this share would be above 50 per cent. In another group of products, for continuous renal replacement therapy (CRRT), the parties would have 60–70 per cent at European level, and even more in some countries. The Commission focused on the closeness of competition, given that Baxter was a relatively small supplier and the parties faced at least one other strong competitor, Fresenius. One source of information was a questionnaire that the Commission sent to customers. In HD, a large majority of these customers named Fresenius as Gambro’s closest competitor, describing Baxter as weaker in terms of quality and product range. A questionnaire given to competitors revealed a similar picture for HD. In CRRT, customers and distributors saw Gambro, Baxter, and Fresenius all as close competitors. Such questionnaires undertaken by the Commission in merger inquiries are not usually as robustly designed as the consumer surveys described in the previous section, but can nonetheless give some indication of closeness of competition.

7.60  The Commission also obtained switching information from the questionnaires. In HD, there were twice as many instances of switching between one of the parties and Fresenius as there were between the parties themselves. In CRRT, switching was found to be less common. Finally, the Commission obtained bidding data at the national level in a number of countries. The overall picture that emerged for HD was similar, with Fresenius participating in most of the high-value tenders won by Gambro, and also winning most of the tenders in which Gambro participated but did not win. More limited bidding data was available for CRRT, but in many cases all three providers were invited to bid. In the end, the Commission cleared the merger on condition that Baxter divest its CRRT business.

(p. 320) 7.5  Unilateral Effects: Simulating Price Rises

7.5.1  From simple to complex merger simulation

7.61  Diversion ratios tell you whether products are close competitors. But the question remains: how close is too close? To answer this you need to consider diversion evidence within a framework of how prices are set before and after the merger. A range of tools is available for this kind of analysis, with varying levels of complexity. At one end are models with simplifying assumptions on company behaviour and demand; at the other are full-blown merger simulations based on complex estimates of demand systems and competitive interaction. In between are various tools which relax some of the assumptions of the simplest approaches, without being as comprehensive and time- and data-intensive as merger simulations.

7.62  We set out these approaches below, but first we go through the basic economic framework. This begins with the pricing logic of a profit-maximizing company after it has acquired a close, but differentiated, competitor. The thought process is very similar to that for the SSNIP test. Product differentiation gives the supplier a degree of monopoly power. In unilateral effects analysis you have the same demand system as for the hypothetical monopolist test, except that every product or brand is already a monopoly. You can therefore focus the analysis on competition between products or brands. With control over two products, the merged entity will have an incentive to raise the price of the new product portfolio (in the same way that the hypothetical monopolist sets prices in the second iteration of the SSNIP test, where it controls both the focal product and the nearest substitute). Pre-merger, if a company tried to raise the price of its existing product, it would earn higher profit margins on the remaining sales, but would forgo the margins on sales that are lost as a result of the price rise. Owning both products post-merger, the company will no longer be concerned about losing sales to the recently acquired product. Whether it is profitable to raise prices will therefore depend on the company’s margins on the existing and newly acquired products, and on the diversion between the two. If it is rational to raise prices post-merger in this framework, the merger may result in an SLC. There are no clear thresholds, but 5 per cent is often used for judging a price increase to be of concern.

7.5.2  Illustrative price rise analysis

7.63  Full merger simulation is often not practical. There are simpler models of competition that are less data-intensive and can be used to calculate indicative or illustrative price increases post-merger. An established approach is to take the Bertrand oligopoly model (see also Chapter 3 and section 7.6 below). In this model companies compete on price, and capacity is not constrained. With homogeneous products, the Bertrand oligopolists behave as if the market were perfectly competitive and prices are in line with marginal costs. With differentiated products, which is the relevant model here, raising the price of one product does not lead to the entire market switching to the cheapest supplier. Customers have differing preferences for the various product characteristics and this gives each company some market power. In this model, price will be above marginal cost in equilibrium—by how much will depend on the degree of differentiation and the number of competitors. To ‘simulate’ the effect of the merger, you put the merging products together under common ownership and you get a new equilibrium of prices and outputs. By comparing the pre- and post-merger outcomes, you get an indication of the price effect of the merger. All you need (p. 321) for the simplest version of this simulation exercise are estimates of the diversion ratios (e.g. from a survey) and data on existing price–cost margins.

7.64  How does this work in practice? Consider a merger between products A and B. Assume that they have the same price–cost margin (m). The diversion ratios (d) from A to B and from B to A are symmetric. A further assumption relates to the shape of the demand curve. Two common specifications are linear demand (as in most of the charts presented in this book) and isoelastic demand (where the own-price elasticity is equal along the whole of the demand curve). Based on all these assumptions, the differentiated Bertrand model generates a formula for the post-merger price increase which depends on two factors: m and d. For linear demand the predicted price increase equals md / 2(1 – d). For isoelastic demand it is md / (1 – md). Table 7.2 takes you through these formulae and shows the sensitivities of the simulated price rises to different assumptions on demand and different values for m and d.

Table 7.2  Illustrative price rises based on simple assumptions


Moderate margin and diversion

High margin

Low diversion

Margin (m)

40% = 0.4

60% = 0.6

40% = 0.4

Diversion ratio (d)

20% = 0.2

20% = 0.2

10% = 0.1

m × d




1 – d




1 – md




Price rise, linear demand: md / 2(1 – d)

0.05 = 5%

0.075 = 7.5%

0.022 = 2.2%

  • Price rise, isoelastic demand:

  • md / (1 – md)

0.2 = 20%

0.6 = 60%

0.08 = 8%

7.65  The results in Table 7.2 should be intuitive. In both formulae, the predicted price increase is larger the higher the margin and the higher the diversion ratio. With linear demand, the price increase is 5 per cent if the margin is 40 per cent and the diversion ratio 20 per cent (see the second column). The price increase is 7.5 per cent if the margin is 60 per cent, for the same level of diversion (third column). The higher the margin on product B, the greater the profit on the sales captured by product B from customers diverted away from product A, and hence the more attractive it becomes to raise the price of A after the merger. In the fourth column, with low diversion, the predicted price rise is lower (2.2 per cent with linear demand). Low diversion means that consumers do not see products A and B as close substitutes, and therefore the unilateral effect of the merger is weaker.

7.66  Table 7.2 also demonstrates the sensitivity with respect to the demand curve assumptions. Across all three examples, the difference between isoelastic and linear is marked. If the competition authority were to use a 5 per cent threshold for the price rise, in one of these scenarios (low diversion) the merger would be cleared if the linear specification is used, but it would be blocked under isoelastic demand. The linear specification reflects a self-correction mechanism in demand: as the price rises, demand becomes more elastic and therefore consumers are more likely to switch away (we explained this in Chapter 2). This makes the merged entity less keen to raise prices beyond some optimal point. By contrast, in the isoelastic formulation the responsiveness to prices does not change, regardless of the (p. 322) price level. This can result in very high, and rather unrealistic, predicted price increases. In the Somerfield/Morrisons supermarket merger in the United Kingdom, the CC found a price rise in one local area exceeding 1,900 per cent, based on the assumption of isoelastic demand.39 The diversion ratio between two stores in this area was 72 per cent, and the average profit margin 27 per cent (the corresponding price rise based on linear demand was 35 per cent, as you can work out from the formula).

7.5.3  Illustrative price rises with efficiencies and asymmetric diversion

7.67  The framework for illustrative price rises can be extended by incorporating other features and relaxing some of the assumptions. One extension is to account for cost reductions resulting from the merger. We discuss in section 7.9 how merger efficiencies in general can be analysed in merger cases. Here we discuss the narrower point of how to incorporate any variable cost reductions into the price rise analysis. As we discuss in Chapter 9 (in the context of the pass-on of cartel overcharges), if the variable or unit cost of production decreases, even a profit-maximizing monopolist will have an incentive to pass on part of this cost reduction. Thus, if the merger delivers such cost efficiencies (e.g. by giving the merged entity greater buyer power and therefore an ability to purchase inputs more cheaply), this will reduce the predicted price rises. If large enough, this efficiency improvement could cancel out the price rise or even reduce prices, as we also saw in Figure 7.1.

7.68  In the case of linear demand, there is a simple adjustment to the price-rise formula that takes into account such efficiencies. The price rise formula becomes equal to md / 2(1 – d) – Δm / 2. As above, m is the pre-merger margin and d is the diversion ratio. Δm reflects the increase in the product margin due to cost reductions from efficiencies. It is assumed that half of the cost savings will be passed on in the form of a lower price increase. As an illustration, using the final scenario in Table 7.2 (40 per cent margin, 10 per cent diversion ratio), if merger efficiencies are forecast to deliver a 3 per cent decrease in marginal costs, we would expect a 0.7 per cent price increase with linear demand. Without efficiencies this was 2.2 per cent. An example of where merger efficiencies were considered as part of the price-rise analysis was the South African Competition Tribunal assessment of the Masscash/Finro merger in the groceries wholesale sector.40 Assuming linear demand, the Tribunal accepted that a 1 per cent efficiency improvement would lead to a reduction in the predicted price rise by half a percentage point.

7.69  Another assumption that can be relaxed is that of symmetric diversion. In practice, one product may constrain the other product’s pricing more strongly than the other way round. For example, in a merger where location is key, it may be that one of the stores has a number of other competitors close to it, but that they are too far away to affect the other store. This is likely to result in different diversion ratios between the stores, since customers at one store have more choice than customers at the other. With asymmetry, you have to carry out the simulation for both products separately, and the resulting price rises will be different. The (p. 323) price-rise formula becomes a bit more complicated with asymmetric diversion. Rather than describing it here we show how the results in Table 7.2 may change. Looking at the second column, instead of a symmetric 20 per cent diversion between the two products, we assume that the diversion from A to B is 30 per cent and the diversion from B to A is 10 per cent. This means that B places more of a constraint on A than the other way round. With symmetric diversion, the simulated price rise was 5 per cent. With asymmetric diversion the model predicts that the price of A will rise by 6.7 per cent after the merger and the price of B will rise by only 3.3 per cent. The intuition is that product B is more of a constraint pre-merger on product A than A is on B, and therefore post-merger the price of A will rise more as the constraint from B is removed.

7.5.4  UPP and GUPPI

7.70  A variant of illustrative price rise analysis gained prominence at the time of the 2010 US Horizontal Merger Guidelines: the upward price pressure (UPP) test. A related concept introduced around the same time is the gross upward price pressure index (GUPPI). The economists proposing these tests certainly chose labels that caught on.41 The basic logic of UPP is captured in the Guidelines:

Adverse unilateral price effects can arise when the merger gives the merged entity an incentive to raise the price of a product previously sold by one merging firm and thereby divert sales to products previously sold by the other merging firm, boosting the profits on the latter products. Taking as given other prices and product offerings, that boost to profits is equal to the value to the merged firm of the sales diverted to those products. The value of sales diverted to a product is equal to the number of units diverted to that product multiplied by the margin between price and incremental cost on that product. In some cases, where sufficient information is available, the Agencies assess the value of diverted sales, which can serve as an indicator of the upward pricing pressure on the first product resulting from the merger.42

7.71  The UPP and GUPPI tests in fact sit somewhere in between just looking at diversion ratios and the illustrative price rise analysis discussed above. UPP starts with assessing the diversion from the original product to the acquired product. We saw that before. But it then assesses the profitability of those sales diverted to (or rather, captured by) the acquired product. The more profitable they are, the more attractive it is to raise price on the original product; hence, the greater the upward pricing pressure. In essence, the UPP test takes the diverted sales and multiplies these by the margin, so d ´ m in terms of the price rise formulae we saw earlier. In Table 7.2, d ´ m equaled 0.08, 0.12, and 0.04 in the three scenarios. It is important to be precise which d and which m we are talking about. Recall that in the illustrative price rise analysis in Table 7.2 we first assumed that the margins were the same for both products and diversion was symmetric in both directions (we then relaxed these assumptions). Applying the UPP test to product A, the relevant diversion ratio is that from product A to product B. So A increases price, and the diversion ratio reflects the lost sales in A that are captured by B. The relevant profit margin is that of product B. Multiplying the two then gives the profitability of the sales captured by B after the price increase in A. The higher this profitability, the greater the upward pricing (p. 324) pressure on product A. Applying the UPP test to product B, you use the diversion ratio from B to A, and the margin on A, to measure the profitability of the sales captured by A after the price increase in B.

7.72  The UPP measure still doesn’t tell you how high is too high. One proposed approach is to compare the UPP estimate with any cost reductions in product A that result from the merger. This comparison seeks to capture the net effect of two opposing forces following a merger: the elimination of competition pushes the price of product A up; a reduction in marginal costs of product A tends to drive its price down. A merger creates a net upward price pressure if the UPP effect (d ´ m) outweighs the cost efficiency effect. You should perform the same analysis from the perspective of product B (i.e. taking into account diversion from B to A, the margin on A, and cost savings on B).

7.73  Another approach is to take GUPPI rather than UPP. GUPPI is given by UPP (d ´ m) multiplied by the ratio of prices of the two products. This simply relates the UPP to the initial price levels (though GUPPI does not yet represent a percentage price change either). An advantage of both UPP and GUPPI is that they do not rely on assumptions regarding the shape of the demand curve (in contrast with the illustrative price rise approach). However, there are no clear thresholds for when a GUPPI or UPP is too high, as neither represents a price change as such. The GUPPI estimates are usually compared against some ‘tolerable’ threshold. To give an indication, under certain assumptions GUPPI can be shown to be twice the optimal SSNIP for a hypothetical monopolist, so a GUPPI threshold of 10 per cent is sometimes used as this may be comparable to a 5 per cent threshold for SSNIP. Thus, if the GUPPI measures are larger than, say, 10 per cent, the merger could raise competition concerns. Following this initial screening, consideration can be given to whether any upward pricing pressure might be offset by factors such as efficiencies, entry, innovation, or product repositioning by competitors.

7.5.5  GUPPI in practice: Petrol stations, cinemas, and channel crossings

7.74  The European Commission and the UK authorities have applied UPP and GUPPI analysis in a number of differentiated-product mergers, involving mobile telephony, airlines, cinema chains, petrol stations, supermarkets, and soft drinks, among other products. Two examples are the acquisition by Sainsbury’s of petrol stations from Rontec in 2012, and the merger between Cineworld and Picturehouse (City Screen) in 2013.43 In both cases diversion ratio and GUPPI analyses were carried out in specific local areas where few competitors would remain. Petrol stations are differentiated mainly geographically (though some differentiation occurs at the level of additional services, such as a grocery store in the case of Sainsbury’s, and car wash facilities). The two merging cinema chains were more differentiated in their product offering. Cineworld was a large operator with seventy-seven multiplex cinemas (defined as having five screens or more). Picturehouse had twenty-one mainly smaller cinemas located in city centres.

7.75  In the petrol stations case, diversion ratio evidence obtained from a consumer survey in each local area indicated that the merging parties were not each other’s closest competitors (p. 325) in many of these locations. In one area the diversion ratio from one of the petrol stations to another was 80 per cent, but diversion in the other direction was only 7 per cent (this was due to the specific location of these and competing petrol stations, and highlights the fact that one should not too readily assume that diversion ratios are symmetric). Yet even the 80 per cent diversion ratio did not result in a high GUPPI value. This was because the petrol stations concerned made very low profit margins (less than 5 per cent). The GUPPI was therefore less than 4 per cent in this area, and even lower for the other areas. The OFT cleared the merger unconditionally.

7.76  In the cinemas case the OFT found high GUPPI levels in the first phase and referred the matter to the CC. In one city the diversion ratio was 30–40 per cent and the profit margin 50–60 per cent, resulting in a GUPPI of 17.0 per cent. In other areas the GUPPI varied from 3.2 per cent to 13.6 per cent. The CC largely confirmed these findings on GUPPI, and complemented them with other analyses. It eventually required divestments in three local areas. In one of these, Cineworld had two multiplexes, Vue (a rival chain) had one, while Picturehouse operated a smaller cinema in the city centre. This Picturehouse cinema was clearly differentiated from the multiplexes. It received subsidies from local government in exchange for maintaining a diverse programme of films and events. The survey evidence showed high diversion from the Picturehouse cinema to Cineworld (40–50 per cent), resulting in a GUPPI of 30 per cent for the Picturehouse cinema. Diversion from the two Cineworld multiplexes to Picturehouse was lower (10–30 per cent), reflecting closer competition from the Vue multiplex. These low diversion rates, combined with a low profit margin at Picturehouse, generated modest GUPPI values for the Cineworld sites (5–9 per cent). The theory of harm in this local area was therefore that the merger would result in upward pricing pressure on the Picturehouse cinema, and not that the Cineworld multiplexes would substantially raise prices. The merging parties were required to divest one cinema in the area.

7.77  A perhaps more unusual application of GUPPI was in relation to the completed acquisition by Eurotunnel of certain assets of SeaFrance, a bankrupt ferry operator, in 2013.44 Crossing the Channel through the Eurotunnel is clearly different from a crossing by ferry, but to some extent they are substitutable options for cars and lorries. The question was how substitutable? The CC found that Eurotunnel’s main rationale for the deal had been to prevent the SeaFrance assets from being acquired by DFDS/LD, another ferry operator that had just entered the Dover–Calais route. The merger was considered to result in the exit of DFDS/LD, leaving only P&O as a competing ferry operator to Eurotunnel. The test was whether Eurotunnel would unilaterally increase prices after DFDS/LD’s exit, caring less about sales lost to its own new ferry operations. In the absence of direct evidence of diversion between tunnel and ferry crossings, the CC considered that lost Eurotunnel customers would be captured by the two ferry operators (Eurotunnel’s own and P&O) in proportion to their expected market shares. With ferry operator margins for both passengers and freight being in the range of 65–85 per cent, the CC concluded that the merger resulted in strong upward pricing pressure, as reflected in high values for GUPPI. As a remedy it prohibited Eurotunnel from operating its own ferry operations, thus effectively undoing the original acquisition.

(p. 326) 7.5.6  Full merger simulation

7.78  Full merger simulation is the most comprehensive approach to estimating price effects. It involves specifying an IO model that best reflects the nature of competition in the market. This can be the Bertrand model used in the approaches discussed above, but also any other model specifying how firms compete. Unlike the illustrative price rise analysis and UPP and GUPPI, merger simulation can capture price reactions of the remaining competitors. It requires estimating a demand function that reflects consumer preferences and price sensitivity. The model is then calibrated using actual market data (calibration in this context means assigning values to the key parameters of the model so that it fits the actual data). Some of the data used for this purpose, such as pre-merger prices and market shares, may be readily available (e.g. from industry reports or the merging parties’ own market intelligence), but parameters such as the elasticity of demand will usually need to be estimated. In the price rise framework we referred to linear and isoelastic demand, but full demand-system modelling often uses other, more flexible, demand specifications. One of these is the logit framework, which treats consumer demand decisions as a series of discrete choices—for example, consumers first decide whether to buy a chocolate bar or other type of snack, and then they choose between different chocolate bars. While potentially generating more robust results than the approaches discussed above, full merger simulation is not very common because of its complexity and data requirements.

7.79  Nevertheless there are cases where such models have been used in the SLC assessment. One of the earliest was United States v Interstate Bakeries and Continental Baking in 1995.45 This concerned a merger between two leading producers of white bread in a number of cities in Illinois and California. To complement its findings of high concentration and entry barriers in local markets, the DOJ carried out a merger simulation exercise. This was based on a Bertrand oligopoly model to reflect the brand-level competition in the market, and a logit specification to represent the demand conditions. It accounted for the competitive constraints imposed by the closest competitors of the merging parties in the two regions. Under this approach, the predicted price increases were around 5–10 per cent for the brands of the merging parties (which included Wonder, Weber’s, Butternut, Sunbeam and Mrs. Karl’s), and 3–6 per cent in the overall market. In the end, the merger simulation was not relied on in court as the DOJ settled with the parties, resulting in a number of divestments.

7.80  A similar model was considered by the European Commission in the Volvo/Scania merger in 2000 between two manufacturers of trucks and buses.46 The simulation employed a logit demand system and a Bertrand model of price competition, with the parameters estimated from data on prices, market shares and other variables over a two-year period. The model was applied to two types of truck for each of the major truck manufacturers, and in each relevant country. It predicted price increases in excess of 10 per cent in most markets. The model was criticized by the merging parties in relation to data measurement errors, and the mismatch between actual price–cost margins and those estimated by the model. In its final decision, the Commission did not rely on the results, recognizing the novelty of the approach and the level of disagreement relating to the model results. In any event the (p. 327) Commission prohibited the merger on the basis that it would create a dominant position in several national markets for heavy trucks, touring coaches, and buses.

7.81  The Commission did rely on a model of this kind in the Lagardère/Natexis/VUP merger in the publishing sector.47 Editis (formerly known as VUP) was the leading publisher of French-language books at the time. Lagardère owned Hachette, the number two, and wanted to acquire Editis. The merger simulation model focused on the market for general literature in paperback and hardcover format, and estimated demand using a nested logit framework. Data on prices and sales volume were obtained from a market research company. Consumers’ demand decisions were assumed to be hierarchical (nested): they first choose the genre of book—e.g. crime, romance or humour—and then the specific title. The simulation exercise predicted a price increase for the merging parties’ books of 4.5 per cent for paperback and hardcover titles combined. For paperbacks only, the predicted price rise was 5.5 per cent and for hardcover titles it was 1.6 per cent. There were other concerns, such as the strong access that the merged entity would have to key authors and sales channels. The Commission approved the merger on condition of a number of divestments by the parties.

7.82  A further example of the European Commission’s reliance on merger simulation is Kraft’s acquisition of Cadbury, approved (with conditions) in 2010.48 In this case a detailed merger simulation was conducted by the merging parties, based on econometric estimations of demand conditions. The chocolate sector was divided into three segments: countlines (chocolate bars), tablets, and pralines. In the United Kingdom and Ireland the merged entity would have market shares of 30–40 per cent in countlines and pralines. They would have 60–70 per cent in tablets, with a number of strong brands (Toblerone, Milka, and Côte D’Or for Kraft; Dairy Milk, Green & Black’s, and Bournville for Cadbury). The parties based their merger simulation on a differentiated-goods Bertrand model of price competition and employed a nested logit demand system. As in the publishing case, this allows for a two-stage (nested) structure of consumers’ choices: first, the choice between consuming a count-line, tablet, or praline product; and second, the choice of brand within the selected segment. The model predicted a price increase of less than 1 per cent in the UK and Irish markets, without accounting for efficiencies. The Commission requested further sensitivity analyses of these results, but accepted that the merger simulation exercise provided evidence that the proposed operation was unlikely to lead to significant price increases in tablets in the United Kingdom and Ireland. These were Cadbury’s main national markets, but Kraft’s brands were not as established. The Commission did require divestments in Poland and Romania, where the parties had a strong combined market position.

7.83  A very different simulation model, involving auctions, was considered by the European Commission in Oracle/PeopleSoft, a merger between the second- and third-largest vendors of service software products.49 The case involved ‘high-function’ enterprise software for human resource and financial planning functions. The parties faced strong competition from SAP, and a number of smaller software companies. In addition, the Commission found that customers are typically sophisticated when purchasing enterprise software, (p. 328) organizing competitive tenders. It analysed data on hundreds of tenders. A sealed-bid auction model was used to represent the competition in procurement, along with possible efficiency gains. The model predicted price increases of 6.8–30 per cent for various products. However, it was later disregarded by the Commission in light of new evidence, and because it was a model of only three suppliers, whereas market definition evidence had indicated that there was a larger number of potential bidders. Instead, based on analysis of bidding data, the Commission found that Oracle’s bidding behaviour was not particularly affected by the specific identity of the rival bidders in the final rounds of a given bidding contest. The presence of PeopleSoft as a rival did not give rise to more aggressive discounting by Oracle relative to bids in which it faced SAP or other rivals. The DOJ challenged the merger, having found price effects from a merger simulation exercise that used a different auction format in the model (an English auction with complete information). However, the District Court rejected this simulation on the basis of the uncertainty of the predictions, and the merger was allowed to proceed.50

7.5.7  Rivals’ reactions: Entry and repositioning

7.84  The merger simulation approaches discussed above—both the simple and more complex ones—assess what happens to prices in the market if two products or brands are joined together. They do so by taking the current degree of product differentiation as given—that is, products do not change the way they are positioned in the market; all that changes is that two previously independent products are brought under common ownership. In reality, however, other suppliers may reposition their products in response to the merger. Capturing such repositioning in merger simulation models is complicated, and to an extent speculative. The US Guidelines suggest that repositioning can be analysed in a similar way to entry

In some cases, non-merging firms may be able to reposition their products to offer close substitutes for the products offered by the merging firms. Repositioning is a supply-side response that is evaluated much like entry, with consideration given to timeliness, likelihood, and sufficiency.51

7.85  The likelihood of repositioning will depend on factors such as the sunk costs and length of time that any such response would take. Repositioning can mitigate the concerns about SLC. Even if each merging party is the other’s closest competitor, other rivals may enter their part of the product spectrum. Yet repositioning may also be done by the merged entity. One case that shows both effects is Heinz’s acquisition of HP Foods UK in 2006.52 This deal led to a potential overlap in a number of sauce products, baked beans, and other tinned food products. It was unconditionally cleared by the CC on the basis that the Heinz and HP brands were not close competitors in ketchup and baked beans, although barriers to entry were high for these products. For other sauce products, the CC found that there was a sufficient threat of entry. A 2009 review of past merger decisions by the UK authorities showed that products in this market were subsequently repositioned in a way that was not predicted by the CC (Deloitte, 2009). There was successful entry into the baked beans segment, with the new entrant capturing 10 per cent of the market (supporting the original decision to (p. 329) clear the Heinz/HP Foods merger). However, Heinz itself repositioned a number of its sauce brands and entered the brown sauce market, making it more difficult for entrants to fill this space in the product spectrum than envisaged at the time of the merger review.

7.5.8  Price–concentration analysis to infer price effects

7.86  An alternative to simulating price rises based on IO models is to assess directly the relationship between prices and changes in market concentration. This can be done by reference to past events, or to similar markets. The US Horizontal Merger Guidelines set this out as follows:

The Agencies look for historical events, or ‘natural experiments,’ that are informative regarding the competitive effects of the merger. For example, the Agencies may examine the impact of recent mergers, entry, expansion, or exit in the relevant market. Effects of analogous events in similar markets may also be informative.

The Agencies also look for reliable evidence based on variations among similar markets. For example, if the merging firms compete in some locales but not others, comparisons of prices charged in regions where they do and do not compete may be informative regarding post-merger prices.53

7.87  A well-known example of such price–concentration analysis across local markets is the Staples/Office Depot merger of 1997. The FTC looked at how the pricing of these two retail chains of large office supply stores differed pre-merger across local markets.54 It found evidence in internal documents that the two chains saw each other as direct competitors and that they generally set lower prices in those cities in which both chains had a presence than in cities where only one of them had a store. The FTC’s econometric analysis confirmed this, showing a statistically significant price difference of more than 5 per cent between cities with just one of these stores and cities with both (after controlling for other factors that may have contributed to the price difference). On this basis, the FTC concluded that the merger would lead to a price increase.55

7.88  The European Commission pursued this type of econometric evidence in analysing Ryanair’s first attempt to acquire Aer Lingus, in 2007.56 It identified significant overlaps between the two airlines in thirty-five point-to-point markets. In twenty-two markets the merger would create a monopoly and in the other thirteen markets it would lead to market shares above 60 per cent. The Commission highlighted that this was an unusual airline merger case since it involved two airlines based at the same airport (Dublin). Both the Commission and Ryanair carried out a price–concentration analysis. Ryanair undertook a study of 313 city pairs on which it operated, and systematically tested for any price differences between the routes depending on whether Aer Lingus was present. The Commission estimated the impact of the entry of Ryanair on routes operated by Aer Lingus. It considered that the effect on competition of Ryanair’s entry on a route in the past was a good natural (p. 330) experiment to understand the likely effect of the removal of this competitive constraint after the merger. The Commission used a ‘difference-in-differences’ regression technique (see Chapter 9 for more detail on this technique), which covers information both over time and across a number of routes. It found that Aer Lingus’s prices were 5–8 per cent lower when Ryanair was present on a route, and that the same effect was not observed when other carriers had entered a route. In contrast, Ryanair’s approach (which had comparisons only across routes, not over time) found there to be no significant difference in its pricing depending on the presence of Aer Lingus. The Commission rejected Ryanair’s approach on the grounds that it lacked robustness and was not a suitable natural experiment for judging the price effect. It concluded that the merger would create or strengthen a dominant position on the overlap routes.

7.6  Co-ordinated Effects

7.6.1  The two dead Frenchmen and their models

7.89  An SLC or SIEC can also arise in the form of co-ordinated effects. The theory of harm is that the merger results in a market structure that lends itself to tacit collusion between the remaining suppliers. They compete less vigorously with each other than before the merger. Co-ordinated effects require oligopoly markets with few competitors. In such markets, no one supplier can act fully independently, and they all know that their individual actions have an impact on the others. The basic models of duopoly (a market with two suppliers) were developed in the nineteenth century by two French mathematicians/economists, Antoine Augustin Cournot (1838) and Joseph Bertrand (1883). These models are easily extendable to a larger numbers of suppliers (oligopoly).

7.90  In the Cournot model suppliers set quantities rather than prices. The model assumes that there are capacity constraints in production. The two suppliers commit in advance to the production quantity and then place their products in the market. The market clears at a price where supply meets demand. Real-world industries that resemble this model include heavy industries where large capacity investments are decided on years in advance and cannot be altered quickly, and tour operators booking charter flights and hotel room capacity for the following holiday period (see the Airtours case below). In deciding on the level of output, each supplier recognizes that its profit will depend on the choice made by the other supplier. They both choose an optimal output level on the assumption that the other will not change its output once chosen. Each then effectively behaves like a monopolist on its residual demand, that is, the amount of the demand left after the other supplier’s production decision. Because of the capacity constraints, the two suppliers cannot adjust their choices once the behaviour of their rival is revealed.

7.91  Without going into detail on the mathematics of the model, the end result is intuitive: output, price, and profits in Cournot oligopoly lie at a certain point between perfect competition and monopoly. The essence of Cournot is that the two suppliers are still competing with one another but recognize their strategic interdependence. This means that they can achieve prices above the competitive level and exploit some market power. However, this market power falls short of what could be achieved if there were a monopoly. Neither supplier has the incentive to restrict its output to (half) the monopoly level because each knows that the other would then produce more and capture a greater market share. Achieving a (p. 331) jointly profit-maximizing output would require co-ordination (tacit or explicit) between the two—we return to this below. An appealing feature of the standard Cournot model is that as the number of competitors increases, the outcome moves further away from the monopoly outcome, increasing total output until it approaches the competitive level. In terms of the Lerner index of market power (see Chapter 3), the price–cost margin in the Cournot model decreases directly with the number of suppliers in the model, and increases with the HHI (the more concentrated the market, the higher the margin).

7.92  Bertrand altered the assumption that there are capacity constraints and in so doing dramatically changed the predictions of market outcomes under duopoly. He posited that instead of choosing output in advance, the duopolists choose their price. Without capacity constraints, price and output will be the same as in perfect competition. This is called the Bertrand paradox: even with only two horses you get a perfectly competitive race. To see why, imagine that one supplier chooses a price that is 5 per cent lower than its rival’s. With no capacity constraints, the cheaper supplier captures all the demand for the product. Anticipating this, the other supplier will seek to adjust its price below that of its rival in order to steal back the demand. This (anticipated) interaction continues until the price is at the level of costs and neither supplier can profitably undercut the other. As we saw in the discussion of merger simulation, economists now commonly use Bertrand as their base model, but with differentiated goods. This model is very distinct from the original homogeneous-goods Bertrand model. Product differentiation is one way out of the paradox: the more differentiated the products, the lower the intensity of competition, and the further prices can be raised above marginal cost.

7.93  There are many extensions of the Cournot and Bertrand models. One important variant relates to the assumption that suppliers see their rivals’ choices as fixed, either in quantities (Cournot) or in prices (Bertrand). This assumption can be changed—a feature known as the conjectural variation.57 In Cournot, rather than assuming that rivals do not react once output has been chosen as in the standard model (conjectural variation equals zero), it can be assumed that rivals either match or offset output decisions. At one extreme, the assumption can be that any quantity changes will be fully matched by rivals. This conjectural variation of 1 will result in a monopoly outcome—suppliers will be reluctant to increase output too far, since they know that the others would match this and prices would fall. At the other extreme, it can be assumed that any changes in output will be fully offset by rivals. If one supplier increases its quantity, it expects others to reduce their quantities proportionately. This conjectural variation of –1 leads to an outcome resembling perfect competition—each supplier thinks it can bully the others and raises output accordingly, but they all end up doing this. Hence, even in Cournot oligopoly you can theoretically obtain any outcome ranging between monopoly and perfect competition once conjectural variations are included.

7.94  The Cournot and Bertrand models are static oligopoly models. Like the standard monopoly and perfect competition models that we saw previously, they refer to only one time period (also described as one-shot games). The players simultaneously select their output or price, the market clears, and that’s it. Even the conjectural variations (which introduce a form of (p. 332) dynamics with predicted reactions to others’ decisions) still take place within the context of a one-shot game. These static models have provided significant insight into the outcomes that you can expect in markets that are neither monopolistic nor perfectly competitive. However, when looking at co-ordinated effects we turn to dynamic oligopoly models, based on game theory. These take the Cournot or (more commonly) differentiated Bertrand models as the starting point, but then play these games over and over again. This opens the possibility of signalling, reputation, and retaliation effects between competitors, and hence allows us to analyse tacit collusion. We turn to this topic now.

7.6.2  Oligopolists and the prisoner’s dilemma

7.95  Dynamic oligopoly theory provides a good understanding of tacit collusion. For the purposes of competition law, tacit collusion can also be referred to as collective or joint dominance, co-ordinated effects, or conscious parallelism—the economics is essentially the same. Oligopolists may naturally recognize their shared incentive to limit production and raise prices without any communication. Here we do not deal with explicit co-ordination, such as price-fixing and market-sharing cartels (this is covered in Chapter 5), but the economic principles discussed here apply to both tacit and overt collusion. So, if the two oligopolists recognize their interdependence, will they always find a cosy solution that benefits both? A situation where they don’t compete too hard but settle on a price and output that maximizes their joint profits? Unfortunately for the oligopolists—and fortunately for consumers and competition authorities—the answer is no. The mere recognition of their interdependence is insufficient. Effective co-ordination requires more than that. This is because oligopolists face a fundamental problem that they cannot easily overcome without some form of collusion: the attractiveness of cheating on each other.

7.96  To understand this point, it is useful to think of a well-known illustration from game theory known as the prisoner’s dilemma. Game theory is a branch of mathematics that looks at strategic interactions between players and predicts behaviour according to the structure of the game and its incentives. It is commonly used in economics to understand companies’ behaviour and lies at the heart of modern IO theory. Given the set-up of a game, players may have dominant strategies—that is, strategies that give a player its highest pay-off regardless of the strategic choice made by the other. If both players have a dominant strategy, the outcome is predictable and stable. This is a type of Nash equilibrium, after the pioneering work in this area by mathematician John Nash (1950a).

7.97  Figure 7.3 shows this game in the form of a pay-off matrix. The set of strategies available to the two duopolists, firm 1 and firm 2, is to price either ‘high’ or ‘low’. Both firms pick their move simultaneously (we are still in a one-shot game). There are four possible outcomes to this game. If firm 1 decides to price high and firm 2 prices low, firm 1 receives a pay-off of zero while firm 2 gets the whole market for itself and receives €15. We are in the bottom-left cell of the matrix. If it is firm 1 that undercuts firm 2’s attempt to price high, we are in the top-right cell and the pay-offs are reversed (both firms have exactly the same size and costs in this example). If both price low, they get €5 each (bottom-right cell). If both price high, they get €10 (top-left). You can see that this (10, 10) cell is the preferred outcome for both firms as combined profits are maximized. The bottom-right cell represents the competitive outcome (which can be Cournot or Bertrand; this does not really matter here). Will they reach the joint desired outcome of (10, 10)?

Figure 7.3  The oligopoly (prisoner’s) dilemma game: Pay-off matrix

(p. 333) 7.98  Consider firm 2’s preferred choice given each of firm 1’s possible strategies. If firm 1 prices high, firm 2 will prefer to price low, steal the market, and earn high returns (it gets €15, rather than just €10 if it also prices high). If firm 1 prices low, firm 2 will again prefer to price low, rather than have the market stolen from it and earn nothing if it prices high (the €5 thus obtained is better than zero). This means that firm 2’s dominant strategy is always to price low, regardless of what firm 1 does. The game is symmetric, so firm 1 has the same dominant strategy of pricing low. They therefore end up in the bottom-right of the matrix, each obtaining a profit of €5. Clearly not what they had hoped to achieve.

7.99  Both companies would like to reach the point where they price high and obtain profits of €10 each. However, this is not a stable (Nash) equilibrium because each could gain even more by cheating and pricing low while the other prices high. This incentive to compete and undercut one’s rival is present even in a duopoly, and it shows that tacit collusion is not necessarily the natural outcome in an oligopolistic market. The more suppliers there are in the market, the more likely it is that one of them will find the temptation to cheat irresistible. This fundamental oligopolist dilemma should give consumers and competition authorities some reassurance. The original prisoner’s dilemma has two prisoners who committed a crime jointly, but are interrogated separately. They get the choice between ‘confessing’ and ‘not confessing’. The game has the same pay-off structure, with confessing being like pricing low and not confessing like pricing high. Ideally they would both like not to confess and walk free, but the temptation for both to confess and get a reward while the other gets locked up proves too tempting, so both end up confessing. You may notice some similarity with the leniency policy for cartel infringements that has been introduced successfully in many competition regimes—the first company to blow the whistle on the cartel gets immunity from fines, and many cartelists have succumbed to the temptation. Leniency policy is another example of the prisoner’s dilemma in action (see Chapter 5).

7.6.3  Tacit collusion to escape the prisoner’s dilemma

7.100  Alas, oligopolists (and economists modelling their behaviour) have a number of ways of getting around the prisoner’s dilemma. In the real world, they do not play the Cournot or Bertrand game only once. Rather, oligopolists meet each other time and time again in (p. 334) the same market, or indeed in other markets. Dynamic oligopoly theory recognizes that rivalry between oligopolists is not limited to a single game in a single time period. Games are repeated over time, either over a finite number of periods or indefinitely. Dynamic interaction over time allows suppliers to learn about their rivals’ behaviour, and between them they could reach an implicit understanding about behaviour that is in their common interest (infinitely repeated games make this easier than games with a finite number of periods). Repeated interactions provide scope for co-ordination, signalling, and reputation building. Importantly, no explicit collusion is required; companies can anticipate the likely reactions of their rivals. For example, if firm 1 always prices low when firm 2 prices low, but responds to an increase in firm 2’s price by likewise increasing its price, firm 2 may note this behaviour and continue to price high. In this case the two firms have reached the desired joint profit optimum of (10, 10). But firm 2 will still have an incentive to cheat and diverge from this outcome to earn higher profits of €15 for a while. If firm 1’s response is then rapidly to price low as well, both firms gain only €5 in the next period and are unlikely to trust each other again.

7.101  Hence, members of the oligopoly need to devise an effective retaliation mechanism in order for ‘high’ pricing to be the economically sustainable outcome. A ‘low’ price reaction to another firm’s divergence is a form of retaliation or punishment, even though both firms suffer the consequences. In general, for sustained collusion the cost of the punishment to the deviator needs to exceed its gain from divergence. This will be influenced by the difference in payoffs between the two strategies. Continued co-ordination has a payoff of €10 for a large number of periods. Cheating gives a higher payoff of €15, but only until it is detected. Retaliation brings both firms back to the competitive payoffs of €5, and they may not trust each other to co-ordinate thereafter (a permanent breakdown in trust results from what is known in game theory as the ‘grim strategy’). So a key factor is the speed with which cheating is detected and acted upon. This determines how long the cheater can continue earning €15 before retaliation kicks in. Another factor that matters is the discount rate, that is, the degree to which companies value future profits less than current profits. The higher the discount rate, the more the cheater values its short-term gains from undercutting rivals, and the less it cares about future retaliation. Economists have combined these various economic factors—payoff from cheating versus co-ordination; speed of detection; payoff under retaliation; and discount rate—into what is known as the folk theorem or general feasibility theorem. This sets out the conditions for tacit collusion to form a stable Nash equilibrium in infinitely repeated oligopoly games. While dynamic oligopoly models have a range of possible outcomes (sustained tacit collusion being only one of these), they provide useful insight into the conditions required for tacit collusion to be feasible. These conditions have made their way into EU case law on collective dominance and co-ordinated effects—in particular the Airtours judgment, to which we turn now.

7.6.4  The Airtours conditions for co-ordinated effects

7.102  In Gencor (1999) the CFI defined collective dominance as the ‘relationship of interdependence’ between the parties of an oligopoly that ‘encourages them to align their conduct in such a way as to maximize joint profits’.58 The CFI’s ruling in Airtours (2002) further confirmed this approach to joint dominance.59 Under the 2004 EU Merger (p. 335) Regulation and accompanying guidelines the concept of joint dominance is now more commonly referred to as co-ordinated effects.60 The Airtours case was an appeal against the European Commission’s decision to block the proposed acquisition of First Choice, a rival tour operator.61 The Commission was concerned that the UK market for short-haul foreign package holidays was already concentrated, with Airtours, First Choice, Thomson, and Thomas Cook having a combined market share of more than 80 per cent. In coming to its prohibition decision, the Commission had largely relied on a theory of oligopolistic interdependence in a static sense, with tour operators behaving like Cournot oligopolists setting quantities. The CFI found this to be insufficient. It stated that collective dominance arises when the adoption of a long-lasting common policy by the members of an oligopoly is ‘possible, economically rational, and hence preferable’.62 In other words, joint dominance is about tacit collusion, not mere oligopolistic interdependence. The CFI translated this into guidance in the form of three conditions that need to be met to support a concern of collective dominance. First, the members of the ‘dominant oligopoly’ must have the ability to monitor the other members (transparency). This makes economic sense: the market must be sufficiently transparent such that members can co-ordinate on prices or output without communication, and can easily detect deviations so there is limited opportunity for cheating. Second, suitable retaliation mechanisms must exist should one of the oligopolists cheat. Third, the reaction of current and future competitors, as well as consumers, must be unable to destabilize the outcome of the common policy. This basically means that all major suppliers in the market must participate in the tacit agreement, and entry barriers must be high.

7.103  In assessing whether these conditions are met, an understanding of the nature of the market is necessary. As we saw earlier, a wide range of outcomes is possible within these oligopolistic market structures. A number of indicators are of relevance. To some extent taking a ‘checklist’ approach to assess each of these indicators is inevitable. One indicator is that there are very few competitors in the market (EU case law has called this ‘tight oligopoly’, a term that does not really exist in economic theory). Another is transparency, which makes it easier to co-ordinate prices and to detect cheating. A further factor is whether products are homogeneous. If they are, co-ordination may be easier as it needs to focus only on price, and the oligopolists will understand each other’s cost structures. Co-ordination is generally more difficult in markets with differentiated products, as there are more competitive dimensions than just price on which a common understanding must be reached. Stable demand and low levels of technological change also make tacit collusion easier.

7.104  Excess capacity is an indicator of the ability to retaliate. It allows oligopolists to increase production quickly and thereby retaliate effectively by lowering the market price in response to cheating. Multi-market contact is another indicator. Retaliation can occur in another market where the companies also compete. The speed with which prices can be adjusted matters too. The ability to retaliate will be restricted if prices or output choices are (p. 336) committed significantly in advance and cannot be easily altered. High entry barriers reduce the likelihood of outsiders undermining the collusive equilibrium. Note that high switching barriers for customers enhance unilateral market power, but may make retaliation less effective as customers cannot easily be taken away from the cheater by undercutting its price. Unilateral market power and co-ordinated effects often do not go hand-in-hand.

7.105  One of the challenges of analysing oligopoly markets is known as the ‘topsy-turvy principle’. Homogeneous goods and stable demand and technology all make the market potentially fiercely competitive, but at the same time make co-ordination easier. Indicators such as excess capacity can signal that the market will be very competitive, as suppliers will have a strong incentive to price low to capture market share and earn revenue to cover the fixed costs of operation. However, as set out above, excess capacity may facilitate retaliation. It is precisely the fact that fierce competition can result in very low prices (because of the excess capacity) that gives an extra incentive to collude and keep the prices high. In terms of Figure 7.3, if the market is fiercely competitive without co-ordination and the (low, low) outcome generates paltry payoffs, this in itself may induce suppliers to seek to achieve and sustain co-ordination. Cheating would be unattractive since retaliation means a return to those low payoffs.

7.6.5  Cases of co-ordination: Platinum, package holidays, and recorded music

7.106  Against this framework we can consider some cases to understand the assessment of tacit collusion. A deal in 1996 between two major mining companies, Gencor and Lonrho, to combine their activities in platinum group metals was blocked on the basis of collective dominance.63 Although this decision was taken before Airtours, it meets the criteria set out therein. The markets for platinum group metals (platinum, palladium, rhodium, iridium, ruthenium, and osmium) were found to be transparent. There was only moderate growth in demand, production technology was mature, and suppliers had very similar cost structures. It was also a ‘tight’ oligopoly with significant barriers to entry. The two parties, together with Amplats (owned by Anglo American) and Russia (as a national producer), had a combined share of 90 per cent of the global platinum industry. The main producers had financial links and multi-market contacts (geographically and across different metals), so detection of deviation would be relatively swift. Retaliation would be possible because of significant excess capacity and the homogeneous nature of platinum group metals.

7.107  We have set out the economic principles defined in Airtours, but what was the result of the application of these principles to the facts in that case? Airtours and First Choice were two of the four major package tour operators in the United Kingdom. The Commission’s disputed decision stated that a merger is to be blocked on the grounds of collective dominance when the degree of interdependence between the oligopolists is such that it is rational for them to restrict output.64 It did not consider the presence of a retaliation mechanism to be a condition for collective dominance. This conclusion was at the core of the CFI’s rejection of the prohibition decision: competing oligopolists will always have an incentive to restrict output, but an outcome of tacit collusion is not sustainable if the prisoner’s dilemma cannot be overcome. The CFI reassessed the facts and concluded that the merger would not result in a market structure that was amenable to co-ordinated effects. It did not consider the market to be transparent, since flight and accommodation commitments were negotiated on a confidential basis at least one year (p. 337) in advance. Rival tour operators would not necessarily know the full range of locations that others had developed until the brochures for the next holiday season were released. Thus it would be hard to co-ordinate tacitly on output choices. Importantly, retaliation was considered to be difficult. There was limited excess capacity for operators to use for punishment of a rival that deviated from the collusive outcome since capacity choices were made more than a year ahead. Adding extra capacity late in the annual process (once the brochures of rivals had been produced) was found to be expensive, difficult, and generally of lower quality. Hence two of the core criteria to maintain the collusive equilibrium were not met.

7.108  In another case that reached the European courts, Sony had sought to acquire the BMG music business, increasing concentration in the market for recorded music. The merged entity would have a market share of 20–25 per cent and the industry would move from five to four major international publishers. The Commission cleared the merger in 2004, rejecting the alleged risk of collective dominance post-merger.65 The General Court overturned this decision after an appeal by Impala, an association of independent music producers, forcing the Commission to re-examine the case.66 The court’s main reason was that this was indeed a market in which co-ordinated effects were likely to be a concern. The Commission had found that the music products were not homogeneous, as different publishers focused on different genres. While there was reasonable transparency of pricing, there was frequent discounting of list prices that undermined this transparency and would be likely to disrupt attempts to co-ordinate. There was a possible mechanism for retaliation—exclusion of a major’s recording artists from compilation albums—but there was little evidence that this had ever been used.

7.109  The court disagreed with the Commission’s assessment. It found that the market was in fact sufficiently transparent for effective monitoring by members of the ‘dominant oligopoly’ to take place. Even in the absence of direct evidence of transparency in the market, the court considered that the fact that prices were closely aligned and in excess of competitive levels was sufficient, given that there was no alternative reasonable explanation for these features. The General Court’s assessment of past pricing behaviour led it to conclude that tacit co-ordination had probably already taken place before the merger. Yet in 2007 the Commission reinstated its clearance, after undertaking further analysis of the Airtours criteria for tacit collusion.67 The Commission did not find online and offline markets for recorded music to be transparent. It also found no evidence of existing price alignment between record companies, either at the level of individual albums or as regards overall pricing policies.

7.110  This last case illustrates that, even when using the same framework and analysing the same facts, conclusions can differ markedly in terms of whether a market is considered conducive to tacit collusion. Since Airtours and Sony/BMG, few mergers have been blocked, or (p. 338) required remedies, on the basis of a theory of harm from co-ordinated effects. Prominent examples include ABF/GBI Business (2008), a merger concerning the production of yeast in France, Spain, and Portugal, reviewed by the European Commission, and the Anglo American/Lafarge joint venture (2012) in the bulk cement market, reviewed in the United Kingdom.68 It seems that competition authorities find the Airtours criteria difficult to meet—economically sound as these criteria are. It appears to be more common now to challenge four-to-three and three-to-two mergers under the banner of unilateral effects rather than co-ordinated effects.

7.7  Non-horizontal Mergers

7.7.1  A different concern

7.111  Non-horizontal mergers are between companies that do not compete directly with each other. They often involve complementary goods, in particular when they are vertical mergers between parties in the same supply chain. As we know from Chapter 2, bringing two complementary goods under the control of one company has the opposite effect of combining two substitutes: prices will fall rather than rise. In other words, there is inherent downward pricing pressure after the merger, as opposed to upward pricing pressure (the term DPP has not caught on like UPP has). The logic is that the merged entity is now aware that lowering prices will have a positive impact on both of its goods. The vertical merger removes the problem of double marginalization (explained in Chapter 6): rather than each producer setting its own profit margin, with higher prices and lower output at the end of the supply chain, the merged entity maximizes profits across the two layers of the chain. For this reason, non-horizontal mergers often result in lower prices for consumers. Another reason for welcoming the combination of complementary goods is that it can eliminate inefficiencies in investment decisions, such as co-ordination problems and ‘hold-ups’ (again, see Chapter 6).

7.112  The concerns about non-horizontal mergers are different. The theories of harm are similar to those in abuse of dominance cases: the ability and incentive to foreclose rivals in upstream or downstream markets, raising rivals’ costs, or weakening their offering through practices such as refusal to supply, bundling, and discrimination. In this sense, non-horizontal merger control is trying to prevent such exclusionary practices from materializing post-merger. This may to some extent require a policy choice between ex ante merger control and ex post competition enforcement: is the authority so concerned about the possibility of vertical leveraging that it blocks the merger or imposes behavioural remedies? Or does it allow the merger on the basis that any future attempt at leveraging market power may be caught under the abuse of dominance rules as and when it arises?

7.113  In this section we discuss three types of non-horizontal merger: vertical, diagonal, and mergers with portfolio effects. A vertical merger spans different layers of the supply chain, where one party uses an input from the other. A diagonal merger is between an upstream company and a downstream company that does not use that upstream company’s input, (p. 339) but competes with other downstream companies that do use the input. A merger with portfolio effects is one between complementary goods that may be bundled or tied together. All these non-horizontal combinations produce no immediate change in the level of concentration in any relevant market. They are therefore not caught by the usual market share or HHI thresholds. The theories of harm are based on concerns about foreclosure of rivals post-merger.

7.114  The key questions in these cases are: does the merged entity have the ability to foreclose rivals? Does it have the incentive to foreclose rivals? Is there a negative effect on competition? Only if all three conditions are met would there be serious concerns. This is in line with the approach set out in the 2008 EU Guidelines on non-horizontal mergers.69 The ability and incentive to exclude certain competitors does not always equate to the ability to harm competition to the detriment of consumers. The third question, whether competition will be harmed, takes into account efficiencies and the extent of foreclosure of competitors. Foreclosure is a matter of degree and only significant foreclosure would be of concern. Non-horizontal mergers are generally more likely to generate efficiencies than horizontal mergers since, as noted above, they may eliminate double marginalization and investment hold-up. In the legal assessment these efficiencies can be put forward as a ‘defence’, just as they are in horizontal mergers. Competition authorities must be careful not to use the very efficiencies created by a non-horizontal merger against the parties—we discuss this ‘efficiency offence’ in section 7.9.

7.7.2  Input foreclosure

7.115  The ability to foreclose depends on whether the merged company controls an input that is important to downstream rivals, such that a lack of access to it weakens their competitive position. This will depend on the cost of that input as a proportion of the total costs of producing the downstream product. If the input is a relatively small cost item, having to use a more expensive alternative will have little effect on downstream competition. It will also depend on what alternative inputs are available. If an input is essential and has no alternatives, the disadvantage to downstream rivals is absolute (but in this case the upstream company should be constrained by the rules on abuse of dominance, both before and after the merger).

7.116  The incentive to foreclose matters as much as the ability. If the merged company stops selling its upstream product to rivals, it will forgo profits on those lost sales. Hence, input foreclosure involves a profit sacrifice. Against this, it may generate additional profits downstream as rivals find it more difficult to compete without access to the upstream input. This depends on some degree of imperfect competition downstream. If the downstream market is fully competitive, input foreclosure makes little sense. Even an upstream monopolist that acquires a downstream company will not automatically wish to foreclose the other downstream rivals, since healthy competition downstream can contribute to expanded sales, which in turn enhances demand for the monopoly input and profits upstream. Thus, trying to monopolize the downstream market generates extra profits only if that downstream market is not fully competitive.

(p. 340) 7.117  The incentive to foreclose depends on the balance of profits upstream and downstream. This is illustrated in Figure 7.4. Upstream firm 1 and downstream firm A merge. The new entity has the option of cutting off downstream firms B and C from the input supplied by upstream firm 1. If it does, it will sacrifice profits upstream on those units formerly sold to B and C. However, at the same time it will earn higher profits downstream as B and C are weakened. These rivals struggle to compete as their offering becomes less varied or they are forced to rely on an inferior or more expensive input supplied by upstream firm 2. Figure 7.4 illustrates a case of total input foreclosure, where upstream firm 1 stops selling to B and C. A variant on this is partial input foreclosure, where upstream firm 1 reduces the quality (or increases the price) of the input sold to B and C, while keeping quality high (or price low) for its newly integrated downstream arm.

Figure 7.4  Vertical merger: Input foreclosure effect

7.7.3  Incentives to foreclose: The vertical arithmetic

7.118  The profitability of an input foreclosure strategy reflects a trade-off between profits lost upstream and profits gained downstream. You can analyse this by looking at the arithmetic of profit incentives. Working out the incentives can be a complex exercise since you need to know what drives profits upstream and downstream. For example, the ability to win significant profits downstream will depend on the profit margin downstream and the degree to which the merged company can win additional sales if its rivals are disadvantaged by lack of access to its upstream product. This in turn depends on the extent of the rivals’ disadvantage, the diversion from B’s and C’s products to that of A, and the price elasticity of demand for the downstream product. Table 7.3 looks at this vertical arithmetic in a simple example. (p. 341) Producing 1 downstream unit requires exactly 1 upstream unit. Pre-merger, the upstream business sold eight out of ten units in the upstream market (i.e. its market share was 80 per cent), with profits per unit of €1.50 and hence total profits of €12.00. The downstream business A sold five out of ten units in its market (i.e. a downstream market share of 50 per cent), with profits per unit of €1 and hence total profits of €5.00. What effect does the merger have on the incentive to foreclose?

7.119  Consider the situation where the merged company refuses to supply B and C with the input, such that these rivals must rely on an alternative supplier. Say that this rival supplier’s input is either inferior or more expensive. Accordingly, B and C become less competitive, and the merged company has an opportunity both to raise prices and to increase its market share (it can still undercut the price of B and C, which face higher costs). If the merged company’s downstream sales rise from five to seven (the middle column), the foreclosure strategy works: the merged firm forgoes €1.50 of profit upstream but gains €3.40 downstream. It does so partly by selling more units downstream and partly through increased margins. However, if the merged company’s sales rise only from five to six (the right column), the foreclosure strategy is unprofitable. It has lost two units of upstream sales, worth €3.00 in upstream profits, while gaining only €2.20 in downstream profits. In this example the parameter we varied is the extent of downstream switching to the merged firm. Equally, we could vary the profit margins, the market shares, or the pricing reactions. All these factors would influence the incentives outcome, and hence whether the merger is likely to lead to foreclosure.

Table 7.3  Input foreclosure incentive: Vertical arithmetic


High switching post-foreclosure

Low switching post-foreclosure

Upstream units




Upstream margin (€)




Upstream profits (€)


10.50 (down by €1.50)

9.00 (down by €3.00)

Downstream units




Downstream margin (€)




Downstream profits (€)


8.40 (up by €3.40)

7.20 (up by €2.20)

Total profit (€)


18.90 (up by €1.90)

16.20 (down by €0.80)

7.120  The European Commission undertook a similar modelling exercise in TomTom/Tele Atlas in 2008, a vertical merger between a sat nav manufacturer and a producer of digital maps.70 The main theory of harm in this case was input foreclosure: would the merged entity foreclose TomTom’s competitors in the sat nav market from access to Tele Atlas maps? In order to assess the profitability of such an input foreclosure strategy, the Commission estimated how many sales TomTom would be able to capture downstream (which is similar to working out whether we are in the high or low switching scenario in Table 7.3). It found that the sat nav sales captured by the merged entity downstream by raising its rivals’ costs would not be sufficient to compensate for the lost digital map sales upstream. An important factor in this was the remaining upstream competition from NAVTEQ, another producer of maps. The Commission also considered the efficiencies of the vertical integration between TomTom and Tele Atlas. Its model predicted a small decrease in average sat nav prices as a result of the elimination of double marginalization (i.e. downward pricing pressure). The overall price effect depended on the balance of pricing power and merger efficiency. Long-term, higher prices are likely if the foreclosure effect is substantial (and barriers to entry are high), and less likely if the merger leads to significant cost efficiencies. A potential outcome of these dynamics is that one vertical merger prompts another, since rivals that are disadvantaged by foreclosure might compensate by seeking their own vertical partnership. Indeed, around the time of the TomTom/Tele Atlas transaction, rival sat nav maker Garmin agreed a long-term contract with NAVTEQ, and NAVTEQ itself merged with Nokia (which incorporated sat nav services into its mobile devices).71 Within the space of two months, the Commission cleared both these vertical mergers unconditionally.

(p. 342) 7.7.4  Customer foreclosure

7.121  A different theory of harm from vertical mergers is that of customer foreclosure. This is illustrated in Figure 7.5, where upstream firm 2’s access to downstream firm A is cut off or degraded by the merged entity. The analysis of customer foreclosure is similar to that of input foreclosure, and the same vertical arithmetic logic can be applied. An important factor with customer foreclosure is the presence of economies of scale or scope at the upstream level. If downstream firm A is an important route to market for upstream firm 2, cutting off access may make upstream firm 2 a less effective competitor to upstream firm 1. This can arise if downstream firm A is such an important distributor or customer that upstream firm 2’s business now falls below the minimum efficient scale for production, such that it becomes uncompetitive. The merged entity will lose those downstream customers who still prefer upstream firm 2’s product, but could offset this loss by increased sales and market power upstream.

Figure 7.5  Vertical merger: Customer foreclosure effect

7.122  Customer foreclosure can be accompanied by input foreclosure—a double whammy. Imagine that, in addition to the customer foreclosure in Figure 7.5, the merged entity refuses to sell upstream firm 1’s product to downstream firms B and C. B and C can now buy only from upstream firm 2. If firm 2 is still below efficient scale at that point, the inputs bought by B and C would be inferior to those bought by firm A, and accordingly the merged entity gains market power downstream as well. For the merged entity to have this incentive, downstream firm A has to be a very important distributor, and upstream economies of scale or scope must be significant. Whether this harms consumers still depends on the net effect of merger efficiencies and market power accruing to the merged entity.

7.123  The European Commission assessed both customer and input foreclosure in 2015 when Telenet, a Belgian cable operator owned by Liberty Global, merged with De Vijver Media, producer of two major TV channels in Belgium.72 The theory of harm regarding customer foreclosure was that Telenet would no longer include rival TV channels in its retail TV offering, particularly those produced by Medialaan and VRT. The Commission first determined that Telenet would have the ability to engage in customer foreclosure, given its (p. 343) dominant position as a retail TV platform in its footprint (Flanders, part of Brussels, and one municipality in the Walloon Region). As to the incentive to foreclose, it was noted that Telenet would be reluctant to degrade the quality of its retail TV offering by removing popular Flemish TV channels (e.g. Medialaan’s VTM channel was the second most popular TV channel in Flanders, with a viewer share of close to 20 per cent). This would affect Telenet’s competitive position vis-à-vis Belgacom in the sale of retail multiplay packages (TV, internet, telephony). Vertical arithmetic analysis indicated that switching by only a small percentage of Telenet customers would already undermine a strategy of customer foreclosure. The channels produced by VRT, the Flemish public broadcaster, could in any event not be foreclosed due to their official must-carry status.

7.124  The Commission did find an incentive to engage in partial customer foreclosure. This would be in the form of degrading the viewer experience of rival TV channels by making them less easily accessible (e.g. through their position in the electronic programming guide), or by hindering their non-linear content offerings, such as video-on-demand. However, new carriage agreements between Telenet and Medialaan and VRT, put in place during the merger investigation, were deemed sufficient to prevent such partial customer foreclosure strategies. As regards input foreclosure, the Commission found that Telenet would have the ability and incentive to refuse to license the newly acquired channels (Vier and Vijf) to rival TV platforms, the largest of which was Belgacom. The Commission imposed a behavioural remedy, requiring Telenet to license Vier and Vijf to other TV platforms on FRAND (fair, reasonable, and non-discriminatory) terms for a period of seven years—see Chapter 8 on FRAND remedies.

7.7.5  Diagonal mergers

7.125  A diagonal merger is between an upstream company and a downstream competitor to a customer of that upstream company. See Figure 7.6, where downstream company A merges with upstream company 2. This type of merger lacks the co-ordination efficiencies that arise in a vertical merger, since company A does not actually use the products of upstream company 2. The theory of harm is that the merged entity could harm downstream company B by raising the prices it charges for the input used by company B.

(p. 344) 7.126  The usual analysis of ability, incentive, and effect applies. The merged company has a greater ability to foreclose company B if its upstream input is an important component of B’s product, and if B lacks a good alternative to that input. The classic (hypothetical) example of a diagonal merger creating a competition problem involves steel and brass as the downstream products, and iron and zinc as the upstream products. Steel is an alloy of iron and carbon, and brass of copper and zinc. For a number of applications, steel and brass compete in the same relevant market. What happens if the main producer of zinc is acquired by a producer of steel? Although there is no obvious horizontal overlap, a merger between the steel and the zinc producer can have anti-competitive effects if the merged entity can raise the price of zinc and this feeds through to a higher price for brass. Subsequently, customers will switch to steel from brass, benefiting the steel business of the merged company. Thus, a price rise for zinc becomes more profitable than it was before the merger of zinc and steel. This conclusion relies on three conditions. First, the zinc supplier must have a very significant market position, as otherwise brass manufacturers could simply source zinc from other suppliers in response to higher prices. Second, brass and steel must be close competitors, such that customers will rapidly switch to steel when the price of brass rises. Third, zinc has to be a significant input to the production of brass in terms of its share in total production costs. If not, the zinc price increase might be diluted in the total brass costs, and therefore not significantly affect competition between steel and brass.

Figure 7.6  Diagonal merger

7.127  This theory of diagonal mergers was invoked in the Google/DoubleClick merger in 2008.73 As this was a complex merger, we stylize the facts here to focus on the diagonal theory of harm. Google was at the time a supplier of text-based internet advertising (‘steel’), whereas DoubleClick provided services and technology used as an input (‘zinc’) for internet display advertising (‘brass’). Display advertising consisted of images and other audiovisual content, and competed with text-based advertising. DoubleClick’s technology was a necessary input for display advertising, but not for Google’s text-based advertising. The theory of harm was that the merged company might raise DoubleClick’s prices, which would result in higher input costs for display advertising and therefore ultimately generate switching towards Google’s text-based adverts. The European Commission cleared the merger, having concluded that the incentive for a DoubleClick price rise was not consistent with this theory of harm. It found that there were credible alternatives to DoubleClick’s technology and services. This would enable display advertisers to substitute away from DoubleClick in the event of the merged entity charging higher prices for DoubleClick. This is equivalent to our brass producers being able to switch to other options for their zinc. The Commission also investigated the dilution effect and found that the cost of DoubleClick’s technology represented only a small proportion of the total costs of providing display advertising. Hence, a price rise would have little impact on downstream competition between display advertising and Google’s text advertisements. (As it turned out, the DoubleClick acquisition allowed Google to grow in display advertising itself; in other words, it became a steel and brass producer.)

7.7.6  Mergers with portfolio effects

7.128  Portfolio effects—also called conglomerate or range effects—arise if a merged company can offer bundled products against which competitors offering a smaller portfolio cannot (p. 345) compete. The general idea is that a company that is active in various different, but related, markets can exercise market power even without necessarily being dominant in the individual markets. The theory of harm is that having a portfolio of activities in different markets gives the merged entity a degree of market power that is greater than the sum of its parts. The concept of portfolio power was applied by the European Commission in Guinness/Grand Metropolitan in 1997, a deal involving a wide range of alcoholic beverages:

The holder of a portfolio of leading spirit brands may enjoy a number of advantages. In particular, his position in relation to his customers is stronger since he is able to provide a range of products and will account for a greater proportion of their business, he will have greater flexibility to structure his prices, promotions and discounts, he will have greater potential for tying, and he will be able to realise economies of scale and scope in his sales and marketing activities. Finally the implicit (or explicit) threat of a refusal to supply is more potent.74

7.129  Is there really a concern here? A broad portfolio may allow a company to offer its customers product bundles or one-stop shopping, or to obtain economies of scope in production and distribution. All of this can be efficient and beneficial for consumers—we explained the efficiency benefits from bundling in Chapter 4. But imagine that the merged company produces two complementary products while all its rivals produce only one (i.e. the merged entity has the greater portfolio). The theory of harm is that the merged entity will increase the price of one product when sold on a stand-alone basis, while keeping constant or lowering the bundled price of the two products. This would give customers an incentive to buy the second product from the merged entity as well, and potentially allow it to exclude competition for that second product. This is the same theory of harm as for bundling and tying as an abuse of dominance. However, as with other non-horizontal mergers, the framework of ability, incentive, and effect applies when analysing portfolio effects. In particular, in the merger context, a portfolio effects theory of harm does not work unless customers have preferences for buying the bundle and there are significant economies of scope in supplying that bundle (which gives the ability to foreclose). Another precondition is that rivals in the second market can be kept at a disadvantage (which gives the incentive to foreclose). If rivals have counterstrategies, such as entering the second market themselves and offering both products, competition will be ‘bundle to bundle’ and the portfolio concern is diminished. We return to portfolio and range effects when discussing merger efficiencies in section 7.9.

7.8  Minority Shareholdings

7.8.1  Theories of harm from minority shareholdings

7.130  Merger control usually deals with transactions where control of a company changes: one party acquires a controlling stake in another, or the two parties create a new company with a new ownership structure. Competition authorities also increasingly deal with partial acquisitions and minority shareholdings under the merger rules (albeit that not all authorities have jurisdiction to do so). Such shareholdings are common in some countries. They can be pure financial investments, or strategic investments with an eye on future market developments. Economic theory shows that minority shareholdings can have negative effects on competition in certain circumstances. There are two main drivers for this. One is that the minority interest grants the acquirer a limited, non-controlling degree of influence on the company, which can nonetheless affect certain strategic decisions (e.g. through voting or veto rights). The other is that the minority shareholding normally entitles the acquirer to a proportionate share of the profit (through dividends), and this may give it an incentive to dampen competition so as to enhance the profitability of the target company. This gives rise to a number of possible theories of harm, which we discuss here. However, a preliminary observation (p. 346) is that any negative effects on competition from minority shareholdings will logically tend to be an order of magnitude smaller than those from full mergers and acquisitions.

7.131  One theory of harm is that of horizontal unilateral effects. When a company acquires a financial stake in a competitor, this usually brings with it an incentive for the acquiring company to raise its own price (or, more generally, to compete less aggressively). In doing so it may lose customers, but some of these will switch to the target company if the two are close competitors. This increases the target company’s sales and profits and, as a result, the value of the acquirer’s financial stake. The analysis of such unilateral effects is very similar to the price rise analysis we discussed previously. You can basically apply the same formulae for the illustrative price rise, UPP and GUPPI. The only difference is that the acquirer does not obtain 100 per cent of the profits of the acquired product, but a fraction that is proportionate to its shareholding. The resulting upward pricing pressure will therefore be smaller than for a full merger, but in some cases may still be sufficient to cause SLC concerns. In practice, the acquiring company may not always benefit directly from an increase in the target company’s profits. For example, the target may decide to invest its profits in additional production capacity, rather than paying dividends to shareholders. In this case the proportionate price rise analysis will overstate the true unilateral effects.

7.132  Another theory of harm is that of co-ordinated effects. Competitors owning minority shares in each other, or both holding minority shares in a third company, may provide another mechanism to co-ordinate actions, communicate intentions, or retaliate in the event of deviation. Minority or cross-shareholdings are therefore a relevant factor when assessing the criteria for co-ordinated effects. The European Commission raised concerns about co-ordinated effects through minority shareholdings in the VEBA/VIAG merger in 2000.75 This merger, together with another deal (RWE/VEW) that was reviewed in parallel by the German competition authority, would create a strong duopoly in the German wholesale electricity market. In addition, there was a complex web of minority shareholdings held by both VEBA/VIAG and RWE/VEW in various regional and local electricity suppliers. The Commission was concerned that these shareholdings would enhance the likelihood of co-ordination between the duopolists. Both deals were cleared subject to structural and behavioural remedies. These included extensive divestments by both groups of their minority shareholdings in regional and local suppliers. VEBA/VIAG and RWE/VEW were also required to sell the shares they each held in VEAG, another wholesale producer, enabling it to compete more effectively with the two groups.

7.133  A further theory of harm is that of vertical foreclosure. For diverse strategic and commercial reasons, companies sometimes take a non-controlling stake in a distributor (forward shareholding) or a supplier (backward shareholding). As in the case of outright vertical (p. 347) mergers, such shareholdings could in theory give rise to concerns about input or customer foreclosure. An example is the IPIC/MAN Ferrostaal merger in 2009.76 IPIC, an investment company with shares in various industrial companies, sought to acquire MAN Ferrostaal, a contractor company. There were no horizontal overlaps between the parties, but one vertical issue arose involving a minority shareholding. IPIC held a controlling stake in AMI, one of two leading producers of melamine (a chemical compound used across a range of industries). MAN Ferrostaal had a 30 per cent stake in Eurotecnica, which owned essential technology for the production of melamine and licensed this to producers such as AMI. The Commission was concerned that this minority shareholding, while not controlling, was sufficient to exercise influence over certain strategic decisions that required a ‘super majority’ according to Eurotecnica’s governance structure. This influence might be used to disadvantage AMI’s rivals. The Commission cleared the deal on the condition that IPIC divest its entire minority shareholding in Eurotecnica.

7.8.2  A minority saga: The Ryanair/Aer Lingus case

7.134  Ryanair tried to acquire Aer Lingus, the other major airline in Ireland, a number of times, but twice found the European Commission in its way—in 2007 and 2013.77 From 2006, as part of its (hostile) acquisition move, Ryanair had gradually increased its share in Aer Lingus, reaching 29.82 per cent in 2008. After the Commission’s prohibition of the full merger in 2013, the UK authorities challenged the minority shareholding itself (the European Commission had lacked the jurisdiction to do so). The CC found an SLC from the 29.82 per cent shareholding and ordered Ryanair to reduce it to a maximum 5 per cent.78

7.135  The CC was not very concerned about unilateral price effects. At the time of the analysis Ryanair had held the minority share in Aer Lingus for several years, so plenty of data was available on the effects of this shareholding on competition in the relevant markets. There was no evidence that Ryanair’s own pricing had been affected. The CC accepted that strong competition remained between the two airlines in the period of Ryanair’s shareholding (indeed, the evidence suggested that competition had intensified over the years). Moreover, the CC did not expect the shareholding to cause Aer Lingus’s management to compete less fiercely with Ryanair in order to avoid antagonizing its largest shareholder. It also rejected the theory of harm of co-ordinated effects between the two airlines, given the well-publicized animosity between them.

7.136  Instead, the CC’s concerns focused on potential ways in which Ryanair could affect Aer Lingus’s commercial policy and strategy, and thereby weaken it as a competitor. This mechanism would arise through Aer Lingus’s governance structure. In particular, the CC was concerned that the shareholding would enable Ryanair to prevent Aer Lingus from combining with another airline (through a merger, acquisition or joint venture), or from other corporate actions such as issuing new shares. Ryanair’s minority stake would provide it with sufficient votes to block special resolutions, which required approval by more than 75 per cent of shareholders at a general meeting. The CC was also concerned that having a (p. 348) competitor like Ryanair as a shareholder would in itself make Aer Lingus less attractive to other potential airline partners.

7.137  The CC concluded that its concerns could be addressed only through a (partial) divestment of Ryanair’s shares. In order to determine the minimum size of such divestment, it assessed Ryanair’s effective voting power—that is, the number of votes available to Ryanair relative to the total number of votes actually submitted at a shareholders’ meeting. The CC identified a range of potential voting scenarios that would increase Ryanair’s effective voting power beyond its actual share. One scenario was where the Irish government—the other major shareholder in Aer Lingus, with a stake of 25.1 per cent—abstained from voting at a shareholders’ meeting. Another scenario was one of limited turnout by shareholders other than Ryanair and the Irish government. The CC also considered it possible that some shareholders (‘allies’) might vote in line with Ryanair on a particular issue, and that the Irish government might sell its stake at some point in the future, with its shares being dispersed to the general public (thereby potentially reducing the turnout for those shares). A number of scenarios that combined these situations were considered for different levels of voter turnout at Aer Lingus’s shareholders’ meetings. This is shown in Table 7.4.

Table 7.4  Ryanair shareholding that would give it 25% effective voting power (%)

Scenario at the shareholder meeting

Turnout of shareholders other than Ryanair and the Irish government (%)

Historical low (23.4)

Historical average (37.2)

Historical high (41.4)

Full participation (100)

Ryanair without allies, Irish government votes





Ryanair with allies, Irish government votes





Ryanair without allies, Irish government stake dispersed





Ryanair with allies, Irish government stake dispersed





Ryanair without allies, Irish government abstains





Ryanair with allies, Irish government abstains





Source: Adapted from Competition Commission (2013), ‘Ryanair Holdings plc and Aer Lingus Group plc’, 28 August, Table 3.

7.138  The key threshold in this analysis is the 25 per cent of votes cast that would give Ryanair the ability to block special resolutions. The lower the overall voter turnout, the higher Ryanair’s share of total votes cast. For each of its scenarios, the CC calculated the minimum level of actual shareholding that would correspond to an effective 25 per cent blocking minority. Consider the cell top-right: here the assumption is that all shareholders attend the meeting (100 per cent participation). Ryanair would then need 25 per cent of the shares to get 25 per cent of the votes. Staying in the same row, assume that turnout of shareholders other than Ryanair and the Irish government is equal to the historical average of 37.2 per cent. In this (p. 349) case Ryanair would need only 15.7 per cent of the shares to block resolutions. You cannot reproduce this calculation immediately from the table. In this situation you have Ryanair (with 15.7 per cent), the Irish government (with 25.1 per cent), and 37.2 per cent of the other shareholders turning up at the meeting. These others represent 22.0 per cent of shares (37.2 per cent of 59.2 per cent, which is the proportion of shares not owned by Ryanair or the government in this situation). So 62.8 per cent of shares are represented at the meeting (15.7% + 25.1% + 22.0%), and Ryanair has 25 per cent of the vote. In other scenarios, Ryanair has some allies (defined by the CC as 3 per cent of total shares voting the same way as Ryanair), the Irish government abstains, or the government shares are dispersed.

7.139  You can see that the 5 per cent maximum shareholding determined by the CC in its divestment remedy is based on one of the more adverse scenarios for Ryanair. It lies in the bottom row of Table 7.4, in the situation where Ryanair has allies, the Irish government abstains from voting, and turnout among the other shareholders is somewhere between the historical average (where 4.7 per cent of shares is sufficient to get 25 per cent of the effective vote) and the historical high (where 5.6 per cent is sufficient). The CC rejected behavioural remedies proposed by Ryanair that would prevent the airline from voting against specific types of decision, thereby addressing the SLC concerns. This framework for analysing voting rights in the corporate governance structure can be applied in other cases. Following the same criteria as the CC would suggest that competition authorities do not look favourably on minority shareholdings in companies that are direct competitors, unless these shareholdings are too small to stand a chance of blocking corporate decisions that require special shareholder approval.

7.9  Merger Efficiencies

7.9.1  A balancing act

7.140  Mergers usually give rise to some efficiencies. At the very least they cut out some of the overheads (only one CEO and General Counsel required), but efficiencies are often are more substantive than that. They can potentially outweigh the adverse effects of reduced competition. We saw the basic logic of the trade-off between a lessening of competition and enhanced efficiencies in Figure 7.1. A merger will increase overall welfare if the efficiency gains to the merged entity exceed the deadweight loss from the price increase. We also saw that most competition authorities apply a stricter consumer welfare test: the efficiencies must be so large that they offset any upward pricing pressure. Another way of formulating this test is to require efficiencies to be passed on to consumers. Usually the burden of proof for showing efficiencies is on the merging parties, and the authority’s role is to verify these claimed efficiencies and to evaluate whether they are sufficient to offset any price increases. Most merger regimes follow a similar approach. The EU Horizontal Merger Guidelines require that efficiencies ‘benefit consumers, be merger specific and be verifiable’, with these three conditions being cumulative.79 Similarly, the US Horizontal Merger Guidelines state that the agencies will:

consider whether cognizable efficiencies likely would be sufficient to reverse the merger’s potential to harm customers . . .

(p. 350) credit only those efficiencies likely to be accomplished with the proposed merger and unlikely to be accomplished in the absence of either the proposed merger or another means having comparable anticompetitive effects . . .

verify by reasonable means the likelihood and magnitude of each asserted efficiency, how and when each would be achieved (and any costs of doing so), how each would enhance the merged firm’s ability and incentive to compete, and why each would be merger-specific.80

7.141  In most of the merger simulation techniques we discussed earlier, efficiencies came in the form of reduced marginal costs. In reality there any many types of efficiency.81 Yet not all of these would necessarily be of relevance to the merger review. An important distinction is that between efficiencies based on simple scale economies and genuine merger synergies (Farrell and Shapiro, 1990 and 2001). Scale economies can, at least in principle, also be achieved unilaterally through organic growth. They may therefore not be merger-specific in that sense. A merger synergy involves the combination of assets that were previously owned by different companies, allowing output or cost configurations that would not be feasible or practical without the merger. This still does not make the synergy merger-specific. There may be a means of combining the assets that is less restrictive. For example, there may be other merger candidates that have the required assets to generate efficiencies but trigger fewer competition concerns. In general, the more difficult it is to find the assets, the more likely it is that the synergies are merger-specific. Efficiencies that are not merger-specific tend to carry less weight in the analysis. In the merger between Ticketmaster and Live Nation that we saw before, the CC considered the possible vertical efficiencies between the ownership and operation of live music venues and the business of selling tickets.82 It concluded that such efficiencies could also be achieved through long-term contracts between Ticketmaster and Live Nation, and therefore were not specific to the merger.

7.142  Cost savings resulting from anti-competitive output reductions following a merger also do not carry much weight. In FTC v Cardinal Health (1998), two simultaneous mergers would have replaced competition between the four largest drug wholesalers in the United States with a duopoly controlling nearly 80 per cent of the market.83 The FTC was concerned that hospitals, pharmacies, and government purchasers would find themselves paying higher prices for drug wholesaling services. The merging parties claimed that the proposed acquisitions would result in significant efficiencies. Their principal argument was that cost savings would result from the consolidation and closing of distribution centres. However, the FTC argued that this was actually not a benefit but rather ‘the very anticompetitive effect flowing from the transaction’. It pointed out that the parties themselves had recognized that excess capacity drove down prices, and hence the elimination of this spare capacity after the merger would increase prices to consumers.

7.9.2  Sources of merger efficiencies

7.143  The most obvious merger efficiencies are on the supply side: fixed and variable cost savings. A merger can achieve cost reductions in a number of ways, one being economies of scale. As the scale of the production is increased, average costs per unit fall. Another is economies (p. 351) of scope—joint production or marketing of different products leading to lower variable or fixed costs. Cost reductions can also be achieved through rationalization of production processes, such as improved capacity utilization; lower transport costs by optimizing production locations or distribution networks; and the shifting of production to facilities with lower costs. As noted above, economies of scale as such are not merger-specific. However, a merger can speed up the process of realizing scale efficiencies, which is a relevant merger benefit. In the 2001 merger between AmeriSource Health and Bergen Brunswig, the third- and fourth-largest drug wholesalers in the United States (i.e. another merger case in this market after Cardinal Health), the FTC accepted the parties’ merger efficiency arguments, partly because the merger would speed up the process of achieving efficient scale relative to a counterfactual where each party sought scale independently:

Based on our review, the proposed transaction likely will give the merged firm sufficient scale so that it can become cost-competitive with the two leading firms and can invest in value-added services desired by customers. Furthermore, we believe that the combined firm will be able to initiate these improvements more rapidly than either could do individually, and that this timing advantage will be significant enough to constitute a cognizable merger-specific efficiency. The resulting firm, operating in a market increasingly characterized by value-added services, likely will provide customers with greater choices among suppliers and therefore will give customers sufficient leverage to obtain competitive prices.84

7.144  There can also be merger efficiencies directed at the demand side—for example, product repositioning post-merger resulting in greater variety and choice for consumers. In the 2008 merger of Global Radio and GCap, two commercial radio stations in the London area, product repositioning was seen by the OFT as a relevant benefit of the merger.85 The transaction combined largely complementary assets. In the counterfactual of no merger, the independent radio stations would each target a ‘middle-of-the road’ music mix in order to appeal to a wide audience. The merged company was better positioned to exploit the advantages of having more narrowly defined target audiences—e.g. younger versus older listeners—which could be complementary. Product repositioning by the radio stations would allow advertisers (the main paying customers of radio stations) to reach their target audiences more effectively. As the OFT explained:

Global will reposition its now commonly-owned stations to attract listeners, in a way designed to increase total audience size for all stations combined, and increase the demographic focus of the respective station audiences. While directly benefiting end-consumers—who are at no risk of price effects—advertisers also benefit: not only from the ability to reach a greater audience, but also to better target their advertising towards more focused demographics (because many product advertisements are targeted, to greater or lesser degree, towards certain age, gender and income groups), which means less wastage of the message and better value-for-money for the advertising customer. Both types of efficiencies, if realized, will improve the Global/GCap station offer to listeners and advertisers.86

7.145  Another example of demand-side merger efficiencies is ‘one-stop shopping’ for customers, which may reduce their transaction or search costs. This was also found in Global Radio/(p. 352) GCap. The OFT considered that bundles of radio airtime could be sold to advertisers more efficiently as a result of the merger, since advertisers could opt to purchase only one large bundle instead of several smaller bundles from independent radio stations. Portfolio effects where therefore regarded in a positive light in this case.

7.146  Competition authorities also take into account dynamic merger efficiencies in the form of quality and innovation. These are typically considered over a longer time period than efficiencies resulting from cost reductions. Dynamic merger efficiencies include the diffusion of know-how, more efficient use of IP, and increased R&D. For instance, mergers may help develop new products or reduce costs by combining certain assets and expertise that are not easily transferred between separate companies. A merger could also eliminate the duplication of R&D efforts, or facilitate obtaining finance for R&D projects. In 2010 the European Commission approved the acquisition by Microsoft of Yahoo’s internet search business.87 Both companies had lost significant ground to Google in the online search market, with a combined market share of less than 10 per cent in Europe. The Commission found that Yahoo! lacked the ability to compete effectively through innovation, and Microsoft lacked scale. Together they would achieve efficiencies and be better placed to compete with Google. Advertisers and users were expected to benefit from the improved innovation and services, and from reductions in fixed costs.

7.147  A merger between two hospitals in the Netherlands in 2009 illustrates efficiencies in the form of better service and quality.88 The transaction was approved by the Dutch competition authority despite the combined market share of the two parties being in excess of 80 per cent in the regional markets for clinical and non-clinical general hospital care. Both hospitals had experienced problems and inefficiencies pre-merger. They had difficulties filling staff vacancies, and neither hospital had ‘level-2’ accident and emergency facilities or adequate intensive-care units. These issues had led to patients in the region having to travel further afield for surgery or being treated in inadequate facilities. The problems were interrelated. On the one hand, the lack of intensive-care facilities limited the types of surgery that could be performed. On the other hand, their small catchment areas meant a relatively low demand for specialist procedures and intensive-care facilities. Specialists were not attracted to these hospitals since they offered limited specializations and low complexity of procedures. In essence, this was an economic problem—the merging parties were operating below the minimum efficient scale for general hospitals. The aim of the merger was to achieve the required scale of operation by combining the catchment areas. This would allow the merging parties to attract larger teams of medics with greater specialization and thereby improve the quality of service to patients. Thus, the efficiency argument in this case was in terms of quality and service rather than costs. The authority found that the efficiencies outweighed the loss of competition after the merger.

7.9.3  Measuring efficiencies: Data envelopment analysis

7.148  For all their importance in merger analysis, merger efficiencies are only infrequently quantified in a robust manner. Economists have developed various techniques to measure efficiency and productivity, which are used in a range of commercial, regulatory, and policy (p. 353) contexts. They can also be used for merger cases. Here we present the basic features of data envelopment analysis (DEA), one of the most widely used techniques for efficiency assessment. DEA measures efficiency by reference to an efficiency frontier, reflecting companies that produce the most output at the lowest cost. It is commonly used for measuring comparative efficiency when there are multiple inputs and outputs that cannot readily be captured in a single input or output measure. In commercial merger analysis, DEA is sometimes used as a planning tool—that is, to assess the potential efficiencies to be realized by merging two units or companies.89 In this same vein it can be used by competition authorities and merging parties to predict potential efficiency gains from a merger.

7.149  Figure 7.7 shows an example of the efficiency frontier and what happens after a merger. The frontier is constructed as linear combinations of efficient (best-practice) companies—that is, those producing the most output at the lowest cost. DEA assumes that two or more companies can be ‘combined’ to form a composite producer with optimally efficient costs and outputs—a ‘virtual company’. The actual companies are then compared with these virtual companies to see how efficient they are. In Figure 7.7 this is shown for two dimensions only: total costs and output. More sophisticated versions of DEA can capture multiple inputs and outputs.90 The figure shows where ten actual companies in the market—A to J—are located in terms of their total cost (the vertical axis) and their output (horizontal axis). You can see that the further a company is to the bottom or to the right in the chart—low costs, high output—the more efficient it is. The DEA efficiency frontier is drawn by joining together the points representing the most efficient companies: B, C, D, E, and F. These companies expend the least cost to produce a given level of output. An equivalent way of putting this is that they produce the maximum output given the level of cost. Company A is clearly not as efficient. It could achieve its current level of output with much lower costs, like virtual company V which is on the frontier. V is effectively a weighted average of the frontier companies B and C. Companies B and C are referred to as A’s peers, with B having a higher weighting than C (as its output level is closer to that of A). DEA gives the following insight to the management of A: you could improve your productive efficiency by adopting best practices from companies B and C.

Figure 7.7  An illustration of DEA and merger efficiencies

(p. 354) 7.150  Now let’s consider two mergers, between B and C and between D and E. DEA gives insight into the expected levels of synergy. If the businesses integrate without exploiting any synergies, the outcome would be the points (B + C) and (D + E), which simply add together their existing costs and outputs. DEA can be used to estimate the additional cost savings (or improvements in output) that should be feasible by merging two companies that are already on the efficiency frontier. The possibilities for improvement in the first merger are shown in the smaller shaded triangle, which is obtained by moving to the frontier from point (B + C) horizontally (equally efficient output) and vertically (equally efficient costs). If the merging parties are indicating efficiencies of this magnitude, DEA would confirm that this is consistent with two efficient companies remaining on the frontier. If the merging parties are suggesting significantly higher efficiency gains, this would take them beyond current best practice. A competition authority might thus require additional evidence of these claimed savings, since they would represent a shift in the industry efficiency frontier. As regards the merger between D and E, these companies are also already efficient, and if they achieve no further synergies they reach point (D + E), summing the outputs of D and E into one large company. This point in fact pushes out the efficiency frontier, reflecting economies of scale (there were previously no comparable companies of that size). This changes industry dynamics. Company F, previously the largest supplier and judged to be efficient, now has a larger peer (D + E) showing that additional output growth can be achieved at a lower unit cost (represented by the larger shaded triangle reaching out from point F to the new frontier).

7.151  DEA was used in the context of hospital mergers in Denmark, not under competition law but as part of the policy debates on consolidation. In 2007 the Danish government rolled out a substantial programme centralizing medical services in fewer hospitals, claiming that consolidation of hospitals would increase efficiency. Applying DEA, Kristensen et al. (2010) assessed the government’s claim that the consolidations would deliver efficiency gains in the hospital sector. The analysis followed a two-step approach: the cost frontier was identified using DEA analysis; and the efficiency levels of existing and virtual hospitals were measured, thus identifying the potential for efficiency gains. These gains were then decomposed into a number of sources, including learning effects (hospitals adopting best practice from the most efficient ones), scope effects (hospitals focusing on the services that they deliver most efficiently), and scale effects. Because some of these potential gains could be realized without a merger, decomposing the effects in this way allowed identification of those efficiency gains that were actually merger-specific. The results showed that sizeable cost reductions could be realized through learning effects and improved economies of scope. Economies of scale could arise where smaller hospitals merged, but diseconomies of scale were likely to arise if larger hospitals merged. DEA proved a useful tool to test the claimed efficiencies robustly.

7.152  A specific merger context where DEA is relevant is in the water industry in England and Wales. Water companies are regional monopolies but are regulated based on their comparative efficiency—that is, they engage in a form of benchmark competition through the regulatory framework. The regulator, Ofwat, assesses each company’s efficiency during its periodic price control reviews, using DEA and other efficiency techniques. It sets price caps and efficiency targets with reference to companies at the efficiency frontier. In recent years there have been several mergers between water companies, reducing their number from (p. 355) over twenty to seventeen at present. These mergers involve no direct competitive overlaps. Instead, the competition question—to be addressed by the CMA, and before it the CC—is whether the remaining number of companies is still sufficiently large in order for Ofwat to carry out a robust comparative efficiency analysis during its price control reviews, and to preserve effective benchmark competition in the industry.91

7.9.4  When do merger efficiencies benefit consumers?

7.153  From an economic perspective, efficiencies achieved through a merger are beneficial to overall welfare. Competition authorities weigh these efficiencies against the negative effects of the merger on competition. We saw that they often apply stringent conditions in doing so: the efficiencies must be merger-specific, verifiable, and passed on to consumers. Sometimes these conditions are too stringent, especially when a consumer welfare test is applied and the efficiencies must be so great that they offset any upward pricing pressure—that is, the efficiency benefits are passed on to consumers. Many types of efficiency cannot easily be assessed against the pass-on test. Moreover, from an economic perspective there will always be a degree of pass-on, even if the merger leads to a degree of market power. We discuss pass-on in Chapter 9 in the context of damages claims, but note here that even a profit-maximizing monopolist will pass on a proportion of any cost savings to lower prices (50 per cent if demand is linear and marginal costs are constant). As to the nature of the efficiencies, variable cost reductions are more likely to benefit consumers in the form of lower prices than fixed cost reductions, since in theory prices are set with reference to variable (marginal) costs. However, fixed cost reductions (e.g. resulting from innovation efforts) can also benefit consumers in the longer term.

7.154  The European Commission’s assessment in 2004 of the proposed alliance between Air France and Alitalia is one example of where the efficiency defence was dismissed because pass-on was not sufficiently established.92 Air France and Alitalia sought to interconnect their worldwide aviation networks by creating a European multi-hub system at Paris, Rome, and Milan, and by co-ordinating their passenger service operations, including code-sharing, scheduled passenger networks, and sales. The Commission acknowledged that the proposed alliance could generate significant efficiencies in terms of an extensive network that would offer customers more direct and indirect flights. There were also possible cost reductions due to an increase in traffic throughout the network, better planning of frequencies, and other operational efficiencies. The Commission recognized that these efficiencies were merger-specific. However, it was concerned that the expected cost reductions would not result in lower fares:

The Parties have not, however, shown how such cost savings and synergies would be passed on to the customer and were not able to identify precisely on which routes (trunk routes, other routes within the France-Italy bundle, other routes) price decreases would be applied as a result of the Alliance. If their co-operation results in the elimination of competition in certain markets, there will be no incentive for them to pass on these efficiencies to local passengers.93

(p. 356) 7.155  The Commission approved the alliance, but subject to remedies on particular routes where the parties faced little or no competition. Note that the Commission’s point about there being no incentives to pass on efficiencies is not consistent with economic theory; as we mentioned before, even a monopolist will normally pass on cost savings to some extent (for profit-maximizing, rather than altruistic, reasons).

7.156  Another example is the proposed acquisition of TNT Express by UPS in 2013, which the Commission ultimately prohibited.94 The deal would have reduced the number of competitors in international small-package deliveries from three to two in fifteen EU Member States. DHL was the only alternative in those countries. FedEx, the fourth of the global ‘integrators’ had a less well-established position in Europe. UPS presented an extensive analysis of the efficiencies resulting from the merger. Combining the delivery and logistics networks of the two integrators would generate significant economies of density and scope, and improve service quality. UPS estimated these at €400 million–€550 million (for comparison, UPS’s turnover in the relevant market was €4.6 billion in 2011, and that of TNT €2.2 billion). It did not have access to data from TNT, this process having started as a hostile takeover bid. The Commission acknowledged that the market for international small-package deliveries is characterized by significant economies of scale and density. However, it did not consider the analysis carried out by UPS to be sufficiently robust and verifiable. The Commission did accept the quantification of the expected cost savings in the European air network. The workings for these were better documented, and originated from UPS’s internal network planning model. To estimate pass-on of these cost savings, the Commission used the modelling results from the price–concentration analysis that was carried out to measure unilateral effects—this model had produced a coefficient reflecting the relationship between prices and total average costs, implying 60–70 per cent pass-on. The Commission then analysed the price effects of the merger in each country, and assessed whether the predicted price rise would be offset by the estimated cost savings. In some countries the Commission found that the merger efficiencies were sufficient to prevent an SIEC, but in many countries they fell short. This contributed to the Commission’s decision to block the merger.

7.9.5  The ‘efficiency offence’

7.157  In Chapter 1 we saw how US antitrust law in the 1960s was not very sympathetic towards large, efficient companies. We gave the example of FTC v Procter & Gamble (1967), a merger concerning the household bleach market.95 The Supreme Court established the ‘entrenchment doctrine’, whereby mergers could be prohibited if they strengthened the position of merging parties relative to rivals through efficiencies, broader product ranges, or greater financial resources. Becoming too efficient was deemed harmful to competition. This approach changed radically in the United States in the 1970s under the influence of the Chicago School, but not necessarily elsewhere.

7.158  The General Electric/Honeywell merger case of 2001 has been much discussed, largely because it illustrates the different approaches of the US and the EU competition authorities—while the former cleared the merger (subject to a number of divestments), the latter blocked (p. 357) it.96 One of the main differences in the assessment of this merger was how efficiencies were treated. General Electric (GE) was a major supplier of large aircraft engines, while Honeywell supplied avionics and other components for commercial aircraft. Thus, there were limited horizontal overlaps between the merging parties (mainly just in certain products, such as US military helicopter engines). This was predominantly a conglomerate merger. A rationale for such mergers can be improved efficiency, resulting from economies of scale and scope.

7.159  The case in itself was complex, but in essence the US authorities approved the merger (subject to some conditions in the markets where GE and Honeywell had horizontal overlaps) on the basis that it would enhance efficiency, and thus would foster competition. In contrast, the European Commission blocked the merger because it believed that the more efficient merged entity would have both the incentives and the market power to exclude competitors. The theory of harm in the EU was referred to as range effects, which are in essence similar to the portfolio effects discussed above. The concern was that the merged entity would have the ability and incentive to offer low-priced bundles of aircraft engines and avionics systems which its non-integrated competitors would be unable to match. Thus, the same efficiencies that were seen as desirable in the United States were seen as an unfair advantage in Europe—economies of scope yielding the ability to bundle products and sell them at a lower price. This case gave rise to extensive debate about whether EU merger control was seeking to protect competitors rather than competition, (Majoras, 2001; Kolasky, 2001; and Emch, 2004). These days the ‘efficiency offence’ does not play a major role in merger control, but that does not preclude the possibility that from time to time some competition authority may be tempted to use it.(p. 358)


1  United States v AT&T Inc., T-Mobile USA Inc., and Deutsche Telekom AG, Civil Action No. 11-01560, (ESH) (D.D.C. filed 30 September 2011).

2  General Electric/Honeywell (Case COMP/M.2220), Decision of 3 July 2001; and Deutsche Börse/NYSE Euronext (Case COMP/M.6166), Decision of 1 February 2012.

4  Buchwald (1966), as reproduced in United States v Pabst Brewing Co., 384 U.S. 546 (1966).

5  For example, Mitchell and Stafford (2000), Moeller et al. (2005), and Malmendier et al. (2012).

6  For example, Case T-102/96 Gencor v EC Commission [1999] 4 CMLR 971.

7  Council Regulation (EC) No 139/2004 of 20 January 2004 on the control of concentrations between undertakings (the EC Merger Regulation) (Text with EEA relevance); and European Commission (2004), ‘Guidelines on the Assessment of Horizontal Mergers Under the Council Regulation on the Control of Concentrations Between Undertakings’, [2004] OJ C31/03.

8  Federal Trade Commission v HJ Heinz Co 246 F 3d 708 (US Ct of Apps (District of Colombia Cir.), 2001).

9  T-Mobile Austria/Tele.ring (Case COMP/M.3916), Decision of 26 April 2006.

10  Ibid., at [129].

11  Tervita Corp v Canada (Commissioner of Competition), 2015 SCC 3 [SCC Decision].

12  European Commission (2004), ‘Guidelines on the Assessment of Horizontal Mergers Under the Council Regulation on the Control Of Concentrations Between Undertakings’, [2004] OJ C31/03, at [9].

13  Glaxo Wellcome/SmithKline Beecham (Case COMP/M.1846), Decision of 8 May 2000, at [70].

14  Competition Commission (2007), ‘Greif Inc and Blagden Packaging Group’, 17 August.

15  Competition Commission (2010), ‘Ticketmaster and Live Nation’, 7 May. We advised a third party in this inquiry.

16  Department of Justice and Federal Trade Commission (2010), ‘Horizontal Merger Guidelines’, August, p. 32.

17  Office of Fair Trading (2008), ‘Restatement of OFT’s Position Regarding Acquisitions of “Failing Firms”’, December, p. 3.

18  BASF/Eurodiol/Pantochim (Case COMP/M.2314), Decision of 11 July 2001.

19  Office of Fair Trading (2008), ‘Anticipated Acquisition by Lloyds TSB plc of HBOS plc’, 24 October; and Department for Business, Enterprise and Regulatory Reform (2008), ‘Decision by Lord Mandelson, the Secretary of State for Business, not to refer to the Competition Commission the merger between Lloyds TSB Group plc and HBOS plc under Section 45 of the Enterprise Act 2002’, 31 October.

20  Competition Commission (2009), ‘Stagecoach Group plc/Preston Bus Ltd’, 11 November. We advised the acquiring party in this case.

21  Stagecoach Group PLC v Competition Commission [2010] CAT 14, 21 May 2010.

22  Competition Commission (2013), ‘Optimax Clinics Limited and Ultralase Limited’, 20 November. We advised the merging parties in this case.

23  Aegean/Olympic II (Case Comp M.6796), Decision of 9 October 2013.

24  Olympic/Aegean Airlines (Case Comp M.5830), Decision of 26 January 2011.

25  Federal Trade Commission v HJ Heinz Co 246 F 3d 708 (US Ct of Apps (District of Colombia Cir.), 2001).

26  Ibid., at [23] and [15].

27  Office of Fair Trading (2008), ‘Anticipated Acquisition of the Online DVD Rental Subscription Business of Amazon Inc. by LOVEFiLM International Limited’, ME/3534/08, 8 May. We advised the acquiring party in this transaction. Through this deal Amazon became a minority shareholder in LOVEFiLM, and in 2011 it took full control of the company.

28  Competition Commission and the Office of Fair Trading (2011), ‘Good Practice in the Design and Presentation of Consumer Survey Evidence in Merger Inquiries’, March.

29  Department of Justice and Federal Trade Commission (2010), ‘Horizontal Merger Guidelines’, August, p. 20.

30  Competition Commission and the Office of Fair Trading (2011), ‘Good Practice in the Design and Presentation of Consumer Survey Evidence in Merger Inquiries’, March, at [1.2].

31  Office of Fair Trading (2008), ‘Anticipated Acquisition of the Online DVD Rental Subscription Business of Amazon Inc. by LOVEFiLM International Limited’, ME/3534/08, 8 May.

32  Office of Fair Trading (2008), ‘Anticipated Acquisition by Co-operative Group Limited of Somerfield Limited’, 17 November. We advised the acquiring party in this transaction.

33  Office of Fair Trading (2008), ‘OFT considers grocery store divestments in Co-op/Somerfield Merger’, press release, 20 October.

34  See, for example, Tourangeau et al. (2000), Bradburn et al. (2004), and Lucey (2005).

35  See, for example, Mazzocchi (2008).

36  Textbooks on conjoint analysis include Raghavarao et al. (2010) and Rao (2014).

37  Ofcom (2008), ‘Pay TV Second Consultation—Access to Premium Content’, 30 September.

38  Baxter International/Gambro (Case Comp/M.6851), Decision of 22 July 2013.

39  Competition Commission (2005), ‘Somerfield plc/Wm Morrison Supermarkets plc’, September, Appendix E.

40  Masscash Holdings (Pty) v Finro Enterprises (Pty) Ltd t/a Finro Cash and Carry (04/LM/Jan09) [2009] ZACT 66 (Competition Tribunal in South Africa). We assisted the South African Competition Commission which brought the case originally.

41  Leading proponents included Salop and Moresi (2009) and Farrell and Shapiro (2010).

42  Department of Justice and Federal Trade Commission (2010), ‘Horizontal Merger Guidelines’, 19 August, p. 21.

43  Office of Fair Trading (2012), ‘Proposed acquisition by J Sainsbury plc of 18 Petrol Stations from Rontec Investments LLP’, 7 June 2012. We advised the acquiring party on this transaction. Competition Commission (2013), ‘Cineworld Group plc and City Screen Limited’, 8 October.

44  Competition Commission (2013), ‘Groupe Eurotunnel S.A. and SeaFrance S.A. Merger Inquiry’, 6 June.

45  United States v Interstate Bakeries Corp and Continental Baking Co, No. 95 C 4194, 1995 WL 803559 (N.D. Ill. 1995), final judgment, 9 January 1996. See also Werden (2000).

46  Volvo/Scania (Case COMP/M.1672), Decision of 15 March 2000. See also Ivaldi and Verboven (2005).

47  Lagardère/Natexis/VUP (Case COMP/M.2978), Decision of 7 January 2004. See also Ivaldi (2005).

48  Kraft Foods/Cadbury (Case COMP/M.5644), Decision of 6 January 2010.

49  Oracle/PeopleSoft (Case COMP/M.3216), Decision of 26 October 2004.

50  United States of America et al. v Oracle Corporation 331 F 2d 1098 (ND Cal 2004).

51  Department of Justice and Federal Trade Commission (2010), ‘Horizontal Merger Guidelines’, August, p. 22.

52  Competition Commission (2006), ‘HJ Heinz and HP Foods’, 24 March.

53  Department of Justice and Federal Trade Commission (2010), ‘Horizontal Merger Guidelines’, August, p. 3.

54  Federal Trade Commission v Staples Inc 970 F Supp 1066 (DDC 1997). See also Baker (1999).

55  In 2015, 18 years after their previous attempt, Staples and Office Depot agreed another merger. At the time of writing the deal was still under review by the FTC.

56  Ryanair/Aer Lingus (Case COMP/M.4439), Decision of 27 June 2007. The Commission undertook similar analysis in the 2013 case, Ryanair/Aer Lingus III (Case COMP/M.6663), Decision of 27 February 2013. We advised Ryanair on the 2013 case.

57  The OFT issued an entire research report on conjectural variations and its use in competition policy. See RBB Economics (2011).

58  Case T-102/96 Gencor v EC Commission [1999] 4 CMLR 971, at [276].

59  Case T-342/99 Airtours plc v EC Commission [2002] 5 CMLR 7.

60  Council Regulation (EC) No 139/2004 of 20 January 2004 on the control of concentrations between undertakings (the EC Merger Regulation) (Text with EEA relevance); and European Commission (2004), ‘Guidelines on the Assessment of Horizontal Mergers under the Council Regulation on the Control of Concentrations Between Undertakings’, [2004] OJ C31/03.

61  Airtours/First Choice (Case IV/M.1524), Decision of 22 September 1999.

62  Case T-342/99 Airtours plc v EC Commission [2002] 5 CMLR 7, at [61].

63  Gencor/Lonrho (Case No IV/M.619), Decision of 24 April 1996; Case T-102/96 Gencor v EC Commission [1999] 4 CMLR 971.

64  Airtours/First Choice (Case IV/M.1524), Decision of 22 September 1999, at [54].

65  Sony/BMG (Case COMP/M.3333), Decision of 19 July 2004.

66  Case T-464/04 Independent Music Publishers and Labels Association (Impala association internationale) v Commission [2006] ECR II-2289. See also Pilsbury (2007).

67  Sony/BMG (Case COMP/M.3333), Decision of 3 October 2007.

68  ABF/GBI Business (Case COMP/M.4980), Decision of 23 September 2008; and Competition Commission (2012), ‘Anglo American PLC and Lafarge S.A.’, 1 May.

69  European Commission (2008), ‘Guidelines on the Assessment of Non-horizontal Mergers Under the Council Regulation on the Control of Concentrations Between Undertakings’ [2008] OJ C265/07, 18 October.

70  TomTom/Tele Atlas (Case COMP/M.4854), Decision of 14 May 2008.

71  Nokia/NAVTEQ (Case COMP/M.4942), Decision of 2 July 2008.

72  Liberty Global/Corelio/W&W/De Vijver Media (Case COMP/M.7194), Decision of 24 February 2015. We advised the acquirer in this case.

73  Google/DoubleClick (Case COMP/M.4731), Decision of 11 March 2008.

74  Guinness/Grand Metropolitan (Case IV/M. 938), Decision of 15 October 1997, at [40].

75  VEBA/VIAG (Case COMP/M.1673), Decision of 13 June 2000.

76  IPIC/MAN Ferrostaal AG (Case COMP/M.5406), Decision of 13 March 2009.

77  Ryanair/Aer Lingus (Case COMP/M.4439), Decision of 27 June 2007; and Ryanair/Aer Lingus III (Case COMP/M.6663), Decision of 27 February 2013. We advised Ryanair on the latter case.

78  Competition Commission (2013), ‘Ryanair Holdings plc and Aer Lingus Group plc’, 28 August. We advised Ryanair also on this case.

79  European Commission (2004), ‘Guidelines on the Assessment of Horizontal Mergers under the Council Regulation on the Control of Concentrations Between Undertakings’, [2004] OJ C31/03, at [78].

80  Department of Justice and Federal Trade Commission (2010), ‘Horizontal Merger Guidelines’, 19 August, p. 30.

81  See Organisation for Economic Co-operation and Development (2007 and 2013a).

82  Competition Commission (2010), ‘Ticketmaster and Live Nation’, 7 May.

83  Federal Trade Commission v Cardinal Health Inc and Others 12 F Supp 2d 34 (D.D.C. 1998).

84  ‘Statement of the Federal Trade Commission: AmeriSource Health Corporation/Bergen Brunswig Corporation’, File No 011-0122, 24 August 2001, pp. 2–3.

85  Office of Fair Trading (2008), ‘Completed Acquisition by Global Radio UK Limited of GCap Media plc’, 27 August.

86  Ibid., at [18].

87  Microsoft/Yahoo! Search business (Case COMP/M.5727), Decision of 18 February 2010.

88  NMa (2009), Walcheren Hospital—Oosterschelde Hospitals, Case 6424, 25 March.

89  See, for example, Lozano and Villa (2010).

90  See Thanassoulis (2001) and Kumbhakar et al. (2015).

91  See, for example, Competition Commission (2007), ‘South East Water Limited and Mid Kent Water Limited’, 1 May; and Competition Commission (2012), ‘South Staffordshire Plc/Cambridge Water PLC merger inquiry’, 31 May. We advised the merging parties in both cases.

92  Société Air France/Alitalia Linee Aeree Italiane SpA (Case COMP/38.284/D2), Decision of 7 April 2004.

93  Ibid., at [137].

94  UPS/TNT Express (Case COMP/M.6570), Decision of 30 January 2013.

95  Federal Trade Commission v Procter & Gamble 386 US 568 (1967).

96  Department of Justice (2001), ‘Justice Department requires divestitures in merger between General Electric and Honeywell’, press release, 2 May; Majoras (2001); and General Electric/Honeywell (Case COMP/M.2220), Decision of 3 July 2001.