The Beesley Lectures: Lectures on Regulation
European Union merger control is ten years old this year. An infant sixteen years in the gestation has grown to maturity in far less time. Such an anniversary inevitably invites an assessment, but it is not always obvious what questions one may reasonably ask. The birth of the infant was attended by considerable scepticism that it could cope with even the simplest practicalities of life: would the Commission's resources prove adequate to the workload, would the European system really ensure speed, a one-stop shop and the legal certainty that the Commission's procedures in other areas of competition policy had so conspicuously failed to provide? So it is tempting to confine my assessment to an expression of considerable admiration that the infant has survived at all (not to mention displaying a nimbleness and agility one might not have anticipated from an observation of its parenthood), while noting the various remaining areas of procedural contention regarding its future role in life.
In fact, although I shall devote a section of this lecture to noting some of these areas of contention, procedural questions will not be my main theme. This is not because such questions are dull or of secondary importance: indeed, some of them raise profound constitutional issues regarding the distribution of power between the European Union and its member states2, the proper role of philosophical divergences between member states in their regulation of market transactions, and the place of sheltered technocracies in a modern democratic society. Rather it is because such procedural questions have dominated discussion of EU merger control to date, at the expense of a number of other questions whose importance I should like to affirm. Briefly, in discussing ten years of EU merger control I should like to point to some intriguing questions about the nature of mergers without ignoring more familiar questions about the nature of control.
The European Commission has been highly active, and the pace of its activity has been accelerating. Table 1 divides the ten years into two periods, one comprising nearly three-quarters of that time (from September 1990 to December 1997) and the second barely over a quarter (from January 1998 to August 2000). There have been slightly more cases notified in the second period than the first, so the acceleration has been remarkable. Although only thirteen mergers have been prohibited in ten years, many more have seen their approval subject to conditions, either after the four-month Phase 2 investigation or to an increasing extent after the one-month Phase 1. A further sixty cases have been withdrawn after notification, most of these because of fears that the Commission would prohibit them or because modifications were demanded that the parties were unable to accept. Chiefly because of the increase in Phase 1 conditions, the proportion of cases that can be said to be directly influenced by the operation of the Merger Regulation has risen from a little over 10% in the first period to nearly 15% in the second3. Since the cases under review involve all the largest corporate transactions in the European Union, this must count by any standards as a lot of influence. In terms of the Commission's lasting impact on the operation of European markets, merger control dwarfs the remainder of its activity in all other areas of competition policy combined4.
How has this influence been exercised? In particular, can we say anything about the role mergers are playing in the overall portfolio of strategies available to Europe's businesses, and the way in which EU merger control interacts with their choice of strategy? The last ten years have taught us a good deal, theoretically and empirically, about the point of mergers in a modern economy, and I want to suggest that these lessons can fruitfully inform our understanding of the point of merger control.
The outline of this lecture is as follows. In section 2 I shall describe some basic facts about the Commission's activities and about their setting, the evolution of merger activity within the EU. I shall mention, without giving them more than superficial comment, some of the remaining difficult questions about the Commission's procedure and their constitutional place in the institutional structure of the European Union. Then in section 3 I shall outline two of the last decade's major advances in our understanding of the role mergers play in a modern economy: the contribution of the theory of incomplete contracts to the theory of the firm, and the empirical analysis of productivity change using census panel data. In section 4 I shall conclude by saying why I think these advances help us pose sharper questions about the future of European merger control. I shall not in this lecture presume to try to answer them.
The Merger Regulation and EU merger activity
On September 21, 1990, Council Regulation 4046/89 of December 21, 1989, on the control of concentrations between undertakings (the "Merger Control Regulation")5 came into force, introducing into Community law a legal framework for the systematic review of mergers and other forms of concentration. From the outset, the Commission has considered the Merger Control Regulation a "vital additional instrument made available Â… to ensure a system of undistorted competition in the Community."6 Over the intervening decade, the Commission has dedicated significant resources and energy to ensuring its effective application, improving its jurisdictional, substantive, and procedural rules, and winning the approval of Member States and the business community for its four fundamental principles: the exclusive competence of the Commission to review concentrations of Community dimension; the mandatory notification of such concentrations; the application of market-oriented, competition-based criteria; and the provision of legal certainty through rapid decision-making.
In contrast to the Commission's broad interpretation of Article 81, where it has considered a large number of agreements to be anti-competitive, the Merger Control Regulation is based on the premise that major concentrations are "in line with the requirements of dynamic competition and capable of increasing the competitiveness of European industry, improving the conditions of growth and raising the standard of living in the Community."7 In the decade since its adoption, the Merger Control Regulation has evolved into an integral part of Community antitrust practice. Unlike other areas of Community competition law, where few formal decisions have been adopted,8 the Merger Control Regulation has produced a rich and extensive jurisprudence that provides invaluable guidance to a variety of issues, including market definition and the substantive assessment of transactions affecting a broad array of markets. Of the operations notified, by far the largest proportion has involved joint ventures (around 48% of all notified transactions), with the remainder comprising acquisitions (38%), takeover bids (8%), and other forms of concentration (8%).9 As to the market sectors examined, over 260 cases (representing more than 20% of all notified operations) have involved oil, gas, mining, and chemical markets; more than 250 operations (around 20% of all notifications) have concerned consumer products; over 150 transactions (equivalent to 10-15% of all notified operations) have concerned financial and insurance markets; and over 125 transactions (around 10% of all notifications) have involved the wholesale and retail trade. The single largest other major categories have been telecommunications, media, and transportation.
The Commission has adopted a pragmatic and comparatively informal approach to the Merger Control Regulation's application. It has made extensive use of pre-notification meetings to clarify threshold legal issues, discuss the scope of any filing required, obtain a preliminary understanding of the relevant markets, and consider procedural questions. Such meetings have undoubtedly made the procedure more business-friendly, with both the advantages and the risks that implies (see Neven et.al., chapters 5-7). The Commission has also acted swiftly to address shortcomings in the original form of Merger Control Regulation adopted in 1989, abandoning the distinction between "concentrative" and "co-operative" joint ventures, correcting the lack of explicit authority to accept undertakings at the end of Phase I, clarifying the ambiguity surrounding the Merger Control Regulation's application to situations of collective dominance, formulating clear rules on ancillary restraints, and introducing a "short-form" notification for unproblematic transactions.
Since the Merger Control Regulation came into force, the number of transactions notified each year has grown significantly: from 60 in 1991, to 110 in 1995, to 292 in 1999. The growing complexity of many transactions, increased expectations as to the detail and sophistication of Commission investigations, and changes in Commission practice have also resulted in increasingly extensive and detailed decisions. For perspective, the first decision rendered after an in-depth investigation, Alcatel/Telettra,10 was 45 paragraphs long and filled seven pages of the Official Journal; in 2000, the Commission decision in Volvo/Scania11 ran to almost 400 paragraphs and extended over 100 pages.
What about the merger activity the Commission is seeking to control? Table 2 shows the total number of M&A transactions (drawn from the AMDATA database) involving EU firms: it increased nearly threefold between 1987 and the end of the 1990s, though by 1998 the figures were not much above their peak in the previous merger boom in 1990. A different source (Securities Data Company) reports the annual number of transactions worldwide to have risen from 11,300 in 1990 to 26,200 in 1998, so transactions involving EU firms would appear to have declined as a proportion of the international total, from over half to around 30% unless there are incompatibilities in the data. What is indisputable is an increase in transactions values, particularly in the late 1990s: Table 3 shows KPMG data reporting an increase in total world cross-border M&A from $237bn in 1995 to $797bn in 1999, and a fourfold increase in average deal size over the same period. These are large numbers, and it is worth bearing in mind that even in an era of globalization, cross-border transactions are in a minority, even weighted by value: on SDS figures cross-border transactions rose from around 24% to only around 33% of all transactions. Table 4 shows the industries in which such activity has been principally concentrated: telecoms, oil and gas, banking, the automotive, food and publishing industries. As I shall discuss in section 3 below, these are all industries in which intangible assets play a large part in the creation of value, whether through research and development and its associated intellectual property rights or through advertising and the creation of brand loyalty.
Table 5 shows the total notifications of merger transactions to competition authorities (national and supra-national) in the EU for the period between March 1998 and December 1999. Interpolating from Table 2, around 15,000 transactions are likely to have taken place in the EU during this period, so something under 30% would have been notified to competition authorities. Of the latter, only some 12% were notified to the European Commission. A further 8% were notified to two or more national competition authorities, indicating that the "one-stop shop" principle is honoured two-thirds as much in the breach as in the observance.
One might conclude that a procedure which is amending only 15% of the Commission's 12% of the notified 30% of all the deals that are taking place in the European Union cannot be having so great an influence after all: only one deal in 200 is visibly modified as a result of the process. Such a conclusion would obviously be misleading, if only because of a failure to weight for size: the one deal in 200 that is modified is certainly among the largest and most important. Its amendment may correspondingly have a value as a signal, to other firms and to other competition authorities, about the principles that are to govern merger control overall, and therefore to govern the regulation of the market for corporations in the EU as a whole. But - of course - this imposes a correspondingly daunting burden of accuracy: is the procedure selecting the one deal in 200 that truly needs amendment, and is it amending it in a defensible and constructive way? Even to begin to answer this question requires some hard thinking about what is motivating the other 199 deals that the procedure currently allows to proceed.
Before returning to this question in section 3, let me sketch some of the procedural issues that remain outstanding and will certainly be debated over the next months and years:
Table 5 is drawn from a report to the Council adopted by the European Commission12 and is used to argue that "too many mergers with cross-border effects still fail to meet the turnover thresholds set in the Merger Regulation as revised in 1997". Even if this judgment is correct, it need not follow that the appropriate response is to revise the thresholds downwards, though that is one possible solution. Others include the introduction of rules to allocate responsibility to a "lead" national authority, a procedure for triggering upward referrals to Brussels for deals with multiple filings (but without an explicit revision of thresholds), and a move towards more standardised filing and investigation procedures that would reduce the costs and uncertainty associated with multiple filings rather than reduce their number.
The increasing decentralisation of other areas of competition authority to member states has highlighted not only that national authorities cooperate with the Commission, but that national courts are expected to apply EU law in these areas. National authorities do not, however, apply EU merger law. This may or may not be an anomaly but it will certainly be subject to increasing scrutiny. It will also raise the question whether differences in regulatory philosophy between member states are a problem with the system or one of its sources of strength.
There has been discussion at various times about the desirability of making merger control a quasi-independent 13activity as it is in Germany and is increasingly becoming in the U.K. The analogy with monetary policy might be considered to militate in favour of such a development, growing concerns about the democratic accountability of the EU institutions might be considered to militate against.
A related issue concerns the relation between the Commission's investigative and decision-making roles, which are currently combined (albeit subject to judicial review) and which on some views would be better separated.
The growing willingness of the Commission to negotiate remedies to merger deals in the first month of investigation raises questions not only about the economic rationale for such remedies but also about their enforceability. The Commission does not have the resources to monitor remedies on a systematic basis.
Cooperation with competition authorities outside the EU has generally been good but tensions that notably surfaced over the Boeing/McDonnell-Douglas case could re-emerge. Enforceability is a major issue here: if two major US firms whose merger had been cleared by the US authorities refused to acknowledge the legitimacy of a blocking decision by the Commission the result might be very serious indeed.
The accession of new members to the EU in the next few years will require not only additional resources but additional thought: does merger evaluation require modification to be adequate to the circumstances of economies so poor and so recently under central planning?
I now turn to questions raised by the merger process itself.
What part do mergers play in a modern economy?
Mergers, acquisitions and the creation of joint ventures involve the transfer of ownership of corporate assets, so a reasonable way to ask what part these transactions play in a modern economy is first to ask what part ownership itself plays. The theory of incomplete contracts, whose pioneer was Oliver Williamson, which addressed itself to questions first posed by Ronald Coase, and which has more recently been formalised and developed by Oliver Hart, John Moore, Jean Tirole and others, has emphasised the value of seeing ownership as primarily about the allocation of residual rights of control of productive assets14. These rights are described as residual in the sense that they operate in circumstances whose details have not been and could not be foreseen to the extent of being made the basis of contractual enforcement of rights and responsibilities. When contingencies cannot be adequately foreseen, someone must exercise discretion over the asset, and that someone is ordinarily described as an owner15.
So much for ownership in general: what is special about ownership of corporate assets? Typically corporations are large combinations of assets that have to be used in combination and are not suitable for dividing up among individual owners (even if they have individual owners they are typically too large to be managed by any single individual). The owners therefore have to work together, and much of the value of the corporation consists not simply of the individual component assets but of their ability to work together efficiently. This ability could not be contractually enforceable otherwise ownership could be made an entirely individual matter (one individual might own the lathe, the other the maintenance equipment, the other the materials for processing and so on). This has a number of implications:
First, the rules governing the exercise of owners' discretion are more tightly drawn than for individual assets, typically because corporate owners they need certain safeguards (such as those governing the rights of minority shareholders) to give them the confidence to cooperate efficiently.
Secondly, we can consider any corporation as consisting of a combination of physical assets (machinery, buildings and so on) that could in principle be owned and managed individually, plus a combination of human assets that could not be so owned (since slavery is illegal), plus a combination of intangible assets which can be described as all the conditions that determine the way in which the physical and the human assets work together. These include the knowledge embodied in corporate rules and procedures, the corporate culture, the firm's reputation in the market place, and so on. The corporation itself formally owns only the physical assets and a very few of the intangible assets (those embodied in formal intellectual property rights such as patents and trade marks), but its value as a corporation consists primarily in the reasonable expectation that whoever owns the physical assets can benefit also from the human and intangible assets that are complementary to them and which may be called the non-tradeable assets.
Thirdly, when one corporation buys another (rather than directly buying its physical assets) it is really buying the tradeable assets in the hope that the non-tradeable assets will come too. If it wanted just the tradeable assets it could usually buy them directly; if it wanted just individual human assets it could hire them directly through the labour market. There are some exceptions due to indivisibilities: I cannot realistically buy half a mine, but if I buy the whole mine I typically buy a company and not just a site. But in most other contexts we can only understand the motivation for corporate mergers and acquisitions if we appreciate that the parties could just have exchanged tradeable assets but chose not to do so. So paradoxically the key to understanding this kind of trade is to see it as a contractual exchange of assets, in which the assets enumerated in the contract are not the primary motive for the transaction. This explains why the sectors in which M&A activity has been important are those for which intangible assets are an important component of overall value.
Fourth, buying and selling corporations is an intrinsically high-risk business, much more so than buying other assets. This is for reasons more fundamental than just the large amounts of money at stake; it is due instead to the fact that the principal contribution to value creation in the transaction comes from the non-traded assets. If I buy a house principally for the view, the contract entitles me to the house and not to the view. But still, when I take possession of the house the view usually comes too (there are exceptions, but they are relatively rare). But if I buy a firm for the brilliance of its management or an investment bank for the brilliance of its M&A team, the management or the M&A whizzkids may choose to leave. Mergers in the banking industry in particular are littered with stories of just this kind. Or the human assets may come but just be unable to integrate their culture with that of the acquiring firm. Likewise, I may buy a brand but I cannot be sure I am buying the brand's success in the market place: the cachet of Fortnum & Mason might not survive if it were bought by, say, Burger King, while the Virgin brand name has proved less spectacularly profitable for Richard Branson than it did for the Roman Catholic Church. In fact, the only point of my buying a corporation is that its non-tradeable assets should turn out to be more valuable to me than they have been to its existing owners (since it is this latter value that will determine the owners' reservation price). To put it another way, the deal creates value through asset-specificity. In the nature of things that is a gamble on untried circumstances.
Fifth, many corporate transactions are structured in ways that reflect requirements associated with creating additional value through the non-traded assets. No less than 48% of transactions notified under the Merger Regulation in the last decade have been joint ventures. It is inconceivable that all of these transactions have been motivated purely by a wish to escape a regulatory prohibition on full merger. On the contrary, most firms have chosen this form because, in spite of the coordination difficulties that attend all joint ventures, the form has compensating advantages. These consist principally in the ability to bundle non-traded assets together within the firm while protecting them from potentially damaging interaction with other parts of the business. A research department can be given autonomy to pursue its projects without a few initial successes turning it into a cash cow for the rest of the corporation. Staff and assets can be transferred to a production joint venture with a guarantee that the venture will end and they will be transferred back to the parent. A speculative venture can be held at arms length so that in the event of its failure the reputation of the rest of the firm is not damaged, and so on. Understanding the role of and the vulnerability of non-traded assets is crucial to understanding why such deals take place at all.
Loosely speaking, we can think of mergers and acquisitions as arrayed on a spectrum from those in which the prospective gains consist solely of improvements in the value of the assets in their existing combination (in short, where the target firm is an under-performer), to those in which the prospective gains are likely to be realised solely through combining the existing assets in new and imaginative ways with the assets of the acquiring firm16. The distinction between performance-correcting transactions and synergistic transactions is not a hard-and-fast one, of course: we are more likely to consider a particular firm an under-performer if an improvement in its performance could be achieved by many possible alternative management teams, and not just one with a particular flair. Conversely, if it would take the special know-how of one particular alternative owner to make the most of the acquired firm's current assets we are more likely to consider any improvement in performance as the consequence of a synergy rather than the correction of a failure.
Since performance-correcting transactions do not yield acquirer-specific gains, the principles of auction theory tell us that virtually all of the gains from the transaction should accrue to the owners of the under-performing firm. Gains to the acquiring firm depend on its ability to exploit informational imperfections which are not only likely to be small in relation to the overall assets but also counterbalanced by elements of the "winner's curse", the phenomenon whereby the likelihood of success in a common values auction increases in the degree of over-optimism of the bidder. This asymmetry between gains to the acquirer and gains to the acquired is indeed a perennial theme of the now large empirical literature on the benefits of merger activity17.
While agreeing that any benefits are asymmetric, that literature has had much more difficulty reaching a consensus on whether the benefits are positive overall18. The points just made help to explain why this uncertainty is unsurprising. First, since mergers and acquisitions are so highly risky we should expect empirical studies to be characterised by high standard errors on all the parameter estimates of interest; finding statistically significant results will be intrinsically hard. Second, even if, on balance, it were to appear that mean returns were significantly negative, this could itself be due to the high variance of returns in the presence of limited liability constraints. Limited liability would lead managers of acquiring firms might chance their arm at a number of transactions with negligible or even negative expected gains provided the upside potential were large enough: there is no need to invoke stupidity or testosterone to explain the phenomenon. Third, the difficulty of defining the appropriate counterfactual for comparison19 is compounded in this case by the fact that, in synergistic transactions, merging firms are likely to be different from their non-merging counterparts (which is why they merge at all): consequently the performance of their non-merging counterparts may be an inadequate control for how the merging firms would themselves have performed without the merger. Fourth, the high value of intangible assets at stake in M&A transactions, coupled with the very fragility of those assets, means that the acquisition process may involve a degree of turbulence that itself threatens the value of the very assets by which it is motivated. This creates a problem of endogeneity that makes evaluating such studies peculiarly difficult. Finally, distinguishing empirically between mergers that do and those that do not enhance market power will be made harder by the significance of intangible assets: a rise in prices after a merger could indeed indicate that market power had increased, but could instead be due to the improved value of the product or service that the new brand or service quality had created20.
It would at least be valuable if empirical studies gave us some kind of insight into the ways in which synergistic transactions (those based on new combinations of productive assets) could add value. This is important for policy, since we know that sometimes transactions add value to the firm while diminishing value to society, since they create market power (for example by concentrating non-reproducible productive assets into the hands of fewer producers). The more we know about the other benefits from synergistic transactions the more comfortably we can identify the particular asset combinations that create market power and evaluate them against any other benefits of the combination.
This is not the place to try to evaluate all the empirical material bearing on this large question. What I want to do instead is more limited: to look at some lessons from a very important literature that has sprung up almost entirely in the last decade, and which examines the causes of productivity growth on the basis of detailed census panel data about individual economic establishments. Once again I shall be illustrative, not comprehensive21.
I illustrate using findings by Jonathan Haskel and various co-authors, particularly from a paper by Disney, Haskel & Heden (2000). Table 6 shows the results of decomposing productivity growth in manufacturing into that which is due to growth within existing establishments (plants), that which is due to transfers of market share between plants (mostly though not always in the direction low-productivity to high-productivity plants on average), and that which is due to the net impact of entry and exit. Several important messages emerge from this table:
First, plant turnover (entry and exit) contributes a large share of aggregate productivity growth, comparable in magnitude to the share of within-plant productivity growth for all periods when labour productivity is considered.
When total factor productivity measures are considered, plant turnover and transfers between plants together contribute virtually everything to productivity growth over the period 1980-92 as a whole, while contributing a little over 60% during the period 1982-9. This corroborates evidence from other sources such as Geroski & Gregg (1997) that, contrary to some wishful thinking, recessions do not induce impressive productivity growth in plants that survive them.
Transfers of market share between plants contribute substantially to TFP though negligibly to labour productivity in all periods, suggesting that plants gaining market share were substantially those that made more efficient use of capital.
Table 7, from the same source, then tells us something very interesting about ownership. It compares the contribution to productivity growth of single-plant firms with those of multi-plant firms. The following findings stand out:
Multi-plant firms accounted for around 80% of both labour productivity and total factor productivity growth over the period as a whole.
This was partly because such multi-plant firms proved to have a consistently better record at gaining productivity from the closure of relatively unproductive plants and the opening of relatively productive ones.
It was partly because multi-plant firms were more successful than single-plant firms at raising labour productivity by increasing the capital-labour ratio.
And it was partly because multi-plant firms were more successful than single-plant firms at increasing the proportion of output produced by plants that used capital relatively efficiently.
Ownership (of multiple plants by a single firm) appears to have been yielding important benefits. Firms belonging to a larger group probably learn from other parts of the group: we know that the transfer of tacit knowledge is one of the most important functions of the modern corporation, but we know to date rather little about what determines the transfer of tacit knowledge between plants in the same firm rather than within a single plant22. In addition, the group appears to be playing the roulette wheel of plant opening and closure with a notably higher success rate than do single firms. Given the hazards of bankruptcy and the difficulties associated with start-up capital, it is perhaps unsurprising that the internal capital markets of larger groups are capable of smoothing some of these market imperfections: what is more surprising, as well as reassuring, is that they actually use this capacity well, rather than dissipate it in a flurry of internal rent-seeking (as the work of Meyer, Milgrom & Roberts, 1992, might have led one to expect).
More evidence on the latter is provided by the same authors' calculations of hazard probabilities of exit in which the year of entry appears as a dummy variable, along with various controls for size and age. Single plants that enter in boom years appear more likely to exit in subsequent years, though plants belonging to groups show no such bias. This suggests booms may encourage low-productivity entry among single plant firms, and corroborates the impression that multi-plant firms prosper better on average from the turnover gamble than do single-plant establishments.
So ownership by a group may well be conferring benefits (on average, it should be stressed) to individual plants that those plants are less likely to realise on their own. Is there independent evidence from similar sources of the benefits due to changes in ownership? The relevant analysis has not yet been performed on UK data, but Baldwin's major (1995) study on Canadian industry includes a chapter on the effects of ownership change on productivity, which comes up with a substantially more positive assessment of the aggregate gains from ownership change than is typical of earlier (non census-based) studies based on productivity data. He also stresses a number of other striking facts about ownership change, such as that acquisitions of plants from firms exiting the industry are more likely to lead to output reductions than transfers of plants between firms staying in the industry (if true this suggests that the failing-firm defence should perhaps be transformed into a failing-firm offenceÂ…). He supports the view that "mergers are more successful when they involve the transfer of intangible knowledge-based assets". He draws attention to the importance of distinguishing between barriers to entry at the level of the plant and barriers to entry at the level of the firm (in many industries it is hard to build new plants but not hard to challenge incumbents by taking over old plants). And above all he corroborates the importance of plant entry and exit for overall productivity change where it is able to occur.
So, to sum up, evidence from detailed micro studies, which has become available only relatively recently, tends to corroborate the importance of the transfer of non-tradeable assets in corporate transactions. It also highlights the immense amount of turnover that takes place continually in many industries, whether through entry and exit or the transfer of market shares between existing firms, and demonstrates the importance of that turnover for overall productivity change23. Finally, it emphasises the importance of distinguishing empirically between the levels of the firm and the establishment (or plant). Apparent stasis in the share of firms in an industry may be disguising considerable turbulence at the level of plants24; indeed, one of the achievements of firms appears to be that they facilitate more efficient and productivity-enhancing competition between individual plants.
In the final section I want to consider what all this might mean for European merger control.
Implications and questions
It is hard not to be impressed by the sheer productivity of the EU merger control procedure, if one measures its output in terms of cases processed, jurisprudence created and major judicial embarrassments avoided. In 1990 the establishment of the Merger Task Force represented the triumph of hope over bureaucratic experience, and for all the remaining questions about the future of the procedure, that hope has been amply vindicated. That it should now be reasonable to pose broader questions about the purpose of the whole process is, in one sense, the best compliment a young institution can receive.
These broader questions mainly concern how accurately the Commission's procedures for evaluating mergers enable it to identify the one case in 200 that, according to its own procedural statistics, really does pose a danger to competition. (By the way, this is not a question specific to the Commission; it concerns the procedures of all competition authorities, which have seen considerable convergence over the last decade.) What we have learned in recent years about ownership and its transfer, theoretically and empirically, implies that analysis of mergers should be more concerned than it has been to date with identifying the specific assets, many of them intangible, which are transferred in the merger process. Once these assets have been identified it is easier to determine the nature of the specificity that enables the merging parties to gain synergy from the reallocation of these assets. It should therefore be easier to see whether this specificity arises significantly from the exercise of monopoly power.
The two key stages of EU merger analysis (market definition and the analysis of dominance) make no explicit reference to the notion of asset specificity but rather allow it to creep in at the back door. In the procedure of market definition, if the merging parties gained control over certain productive assets that can easily be reproduced by others, the notion of "supply-side substitutability" allows the market definition to be widened in consequence. But this happens in a somewhat mechanical fashion, and can easily miss key features of the assets that are being exchanged25. Likewise, when dominance is being considered decisions sometimes identify asset specificity but do not do so in a systematic fashion. This is not to say that the Commission ignores the issue. Indeed, it is striking that a significant proportion of its prohibition decisions (as well as some in which it has imposed conditions) concern the broadcasting, media and information technology industries, where there may be important risks of certain parties gaining control over highly specific and non-reproducible combinations of productive assets (cable distribution systems, for example). The important point is that its identification of the nature of the assets exchanged needs to be examined openly and systematically rather than as a by-product of the more traditional stages of merger analysis.
This leads to an issue that is more particular to the Commission's own procedures. The Regulation, unlike the competition laws of some member states, does not give the Commission the explicit right to invoke an efficiency defence. Although there are some reasons to think that this self-denying ordinance is a valuable barrier against the worst kinds of lobbying and special pleading, it has one major cost. This is that the Commission is never obliged to set out explicitly its understanding of the business rationale for the transaction under investigation (it sometimes does so in practice, but always as an afterthought). Everything we know about the merger process suggests that a competition authority that has failed to understand why the parties wish to merge in the first place (as opposed to merely trading physical assets between them or undertaking some wholly different kind of market-based transaction) will have great difficulty understanding the competition hazards as well. This does not mean that competition investigators should play management consultants, merely that it would be good to see a section in every decision labelled something like: "Why this deal? Why not another?"
Finally, the empirical evidence on productivity change suggests that a great deal of action may be taking place underneath the umbrella of the firm, at the level of individual plants. To some extent competition occurs at that level too. Merger control procedures can be usefully informed by an understanding not just of the market shares of the firms but of the production shares of the merging parties' individual plants.
It may appear that I am calling for EU merger control to do even more than the already impressive amount it does, to take account of even more information and process even larger quantities of data. That is not so. So far at least I am posing questions rather than providing answers. I expect that our gradually deepening understanding of what drives mergers in a modern economy may make it possible to use more precise empirical diagnostic tools. Quite how that will be possible I do not know, but it provides an interesting challenge for the second decade of EU merger control.
1 I am grateful to David Currie and Colin Robinson for the invitation to deliver this lecture, as well as to Jennifer Halliday, Damien Neven and Pascal Reiss (my collaborators on the CEPR project on European merger control) for their valuable contributions to the material reported in section 2, to the CEPR and its Director Stephen Yeo for backing the project, as well as to Wendy Carlin and Jonathan Haskel for recent intensive discussions on the sources of productivity change. Helen Weeds, Paul Kattuman, Daniel Sgroi and Jon Stern gave me helpful comments on an earlier draft.
2 The Merger Regulation represents the most sophisticated practical application anywhere to date of the reasoning underlying the subsidiarity principle (see Neven et.al., 1993, ch.6).
3 It is possible, also, that a considerable number of other cases have been indirectly influenced by the very presence of EU merger control, in that the deals would not have taken the form they did in its absence. It is evidently hard to get evidence on this phenomenon. Neven et.al. (1993, esp. chapter 4) discuss this issue.
4 For perspective, of the 4,679 transactions notified in the United States during the fiscal year ending September 30, 1999, requests for additional information were issued in 113 (2.4%), and only 76 transactions (1.6%) resulted in enforcement actions: U.S. Department of Justice, Premerger Office. During the fiscal year ending March 31, 1999, the Canadian Competition Bureau received notification of 192 transactions (an additional 222 requests were made for advance ruling certificates). Of the examinations concluded during the year, all but 5 were approved outright (see the Annual Report of the Canadian Commissioner of Competition). The situation is similar at the Member State level. In the United Kingdom, the Office of Fair Trading examined 425 transactions in 1998, of which only 8 (less than 2%) were referred to the Monopolies and Mergers Commission (the "MMC") for further investigation. Undertakings were accepted in an additional three (less than 1%) in lieu of a reference to the MMC, although others may have been abandoned in response to confidential guidance. See International Competition Policy Advisory Committee Report (2000), 96.
5 Council Reg. 4064/89 on the control of concentrations between undertakings, 1990 O.J. L257/13; with amendments introduced by Council Reg. 1310/97, 1997 O.J. 1997 L180/1; corrigendum 1998 O.J. L40/17. See the consolidated version at the Commission's website: http://europa.eu.int/comm/competition/mergers/legislation/c406489_en.pdf .
6 XXth Report on Competition Policy (1990), para. 20. See also: "Mergers may be carried out in the interest of economic efficiency, permitting improved exploitation of economics of scale and the pooling of expertise, and may thus help Community industry adjust its structure to the challenge posed by the integration of the internal market and the internationalisation of the economy." XXIst Report on Competition Policy (1991), para. 5.
7 Recital 4, Merger Control Regulation. See also Joseph Gilchrist, "Procedures in Merger Cases,"  I.B.A. Seminar: "The Merger Regulation is user friendly. Unlike Articles 85, 86 or 92, it does not begin with an interdiction. Big is not necessarily bad. Although lawyers and business executives who have dealt with the Commission, particularly in second phase cases, may find this a rather hollow statement, the Merger Regulation does not operate in the kind of adversarial setting and system which is often considered to be the characteristic of operations under Articles 85 and 86."
8 For perspective, the Commission has rendered only around 40 decisions applying Article 82 since the EC Treaty came into force in 1965.
9 XXVIIIth Report on Competition Policy (1998), p. 89.
10 Case IV/M.042, Commission decision of April 12, 1991 (1991 O.J. L122/48-55).
11 Case COMP/M.1672, Commission decision of March 14, 2000.
12 The Report and more information concerning the review is available on line at http://europa.eu.int/comm/competition/mergers/review.
13 independent, that is, of the political control of the Commission.
14 Hart (1995) is an excellent introduction to this literature.
15 But may in other circumstances be a trustee or a representative.
16 Holmstrom & Kaplan (2000) suggest that the 1980s takeover wave in the US consisted primarily of the first type of transaction, while the wave of the 1990s has consisted mainly of the second.
17 See Fridolfsson & Stennek (2000, p.2), World Investment Report (2000, chapter 5).
18 This is particularly true because event studies have tended to find that mergers raise combined gains to acquirer and target shareholders (with acquirer shareholders breaking even), while firm accounting studies tend to find that mergers reduce combined profitability. See Fridolfsson & Stennek (2000) and Gugler et.al. (2000).
19 Fridolfsson & Stennek (2000) note that when mergers impose an externality on other firms then performance relative to a non-merging control group may not be an adequate measure of the absolute benefits of the transaction.
20 In effect the quality-adjusted ("hedonic") price would not have risen even though the crude price per unit had done so.
21 However, Caves (1998) and Bartelsman & Doms (2000) provide thorough reviews and discussions of this literature.
22 This is consistent with cross-sectional evidence (such as that of Baily et.al., 1992) to the effect that there is a positive correlation between the productivity of individual plants and that of other plants belonging to the same firm. In addition, Allen & Phillips (2000) have found evidence that part ownership of firms by other corporations leads to improved performance and investment on the part of the owned firms.
23Geroski (1992) stresses that entrants in many industries not only have high failure rates but may take many years to gain high market share. However, this does not mean that entry is of secondary importance for productivity, for several reasons. First, as he himself acknowledges, entrants may be importantly different from the rest of the industry, particularly at early points in the product life cycle. Second, it is entry of plants that matters rather than simply entry of firms. Third, entry and exit need to be understood together (each may facilitate the other) and it is the productivity of the whole process that is important.
24 On the broader role of turbulence in a market economy, see Carlin et.al. (2001).
25 In the De Haviland decision, for example, it is hard to believe that the market analysis would have taken the same course if the Commission had taken serious account of the fact that among the most important assets of an aerospace company are its development plans (its potential models) and not just its actual models. This is because many orders are advance orders and many development decisions are taken on the basis of such advance orders. The nature of competition in the industry is dependent as much upon whether there are technologically competent potential developers of new models as on whether models of a particular specification currently exist.
Allen, J. & G. Phillip