1 Draft, August 29, 2016 Is EU Antitrust Enforcement a Tool for Protectionism? An Empirical Analysis Anu Bradford, Robert J. Jackson, Jr., and Jonathon Zytnick * Abstract Over the last two decades, the European Commission has often used its power to intervene in high-profile mergers and acquisitions involving foreign companies, giving rise to concerns that the Commission is a protectionist regulator that has converted its competition authority into a powerful tool for industrial policy. Among other things, the Commission has been accused of deliberately targeting foreign, and especially American, acquirers, while facilitating the creation of European national champions. These concerns, however, have long rested on a few famous anecdotes, rather than the kind of systematic empirical analysis that could test those intuitions. In this Article, we introduce a unique dataset that allows us, for the first time, to rigorously examine whether the Commission has protectionist tendencies. The data include information on all of the over 5,000 mergers reported to the European Commission between 1990 and 2014, detailing the transaction value, market size, market concentration, nationality, and the Commission s decision on each transaction. Our study offers two principal findings. First, we find evidence that the Commission has not systematically used its authority to intervene more frequently or more extensively in transactions involving a foreign firm s acquisition of an EU-based firm, or transactions involving a firm based in the United States. Second, we find support for the hypothesis that a Commission challenge is more likely where all of the parties to a transaction are headquartered in the same EU member nation. This is consistent with theories that suggest that the Commission uses its antitrust powers as a tool for market integration rather than an instrument to fend off foreign competition. * Henry L. Moses Professor of Law and International Organization, Columbia Law School; Professor of Law, Columbia Law School; and Economics PhD candidate, Columbia University. We are grateful to the European Commission for sharing data on its merger decisions, and to Jo Seldeslachts and his coauthors for sharing data on measures of industry concentration. We also benefited from helpful discussions with Jeffrey Gordon, Amit Khandelwal, and Joshua Mitts, and from the feedback received at the American Law and Economics Conference, Comparative and International Administrative Law and Politics Workshop at UC Berkeley, the Conference for Empirical Legal studies in Europe, as well as the Faculty Workshop at Columbia Law School. We are grateful for excellent research assistance by Julian Beach, An Duong, Kevin Hennecken, Abraham Lowenstein, Sarah Mac Dougall, Gregory Swartz, Kate Thompson, Sara Thomson, Eliana Torrado Franco, and Claudie Tirefort, and for the support from the Philippe P. Dauman Faculty Research Fund. 1
2 Introduction The European Commission s willingness to shape global M&A activity though its merger control review power has been a subject of controversy among lawmakers and commentators for the past two decades. That is true, in part, because the Commission has often used its power to intervene in high-profile mergers and acquisitions involving foreign companies. The Commission s 2001 decision to ban the $42 billion proposed acquisition involving two US companies, General Electric and Honeywell, is perhaps the most famous example of a case where the EU has overridden a decision by US authorities. But the GE/Honeywell decision does not stand alone. The Commission has intervened in other proposed mergers involving well-known American firms, including Boeing/McDonnell Douglas, MCI WorldCom/Sprint, and EMI/Time Warner. These prominent cases against foreign companies have earned the EU a reputation of the world s antitrust cop but also aroused suspicions of its motivations. Critics portray the Commission as a protectionist regulator that has converted its competition authority into a tool for industrial policy. 2 The Commission has been accused of deliberately targeting US technology giants and aggressively pursuing foreign bidders while facilitating the creation of European national champions. Fears that the EU is deploying its antitrust policy as a tool for economic nationalism often rest on a few famous anecdotes. Yet its 2 For examples, see Tom Fairless, France Feeds New European Economic Nationalism: Experts See Anti- American Sentiment in Mood Swing, WSJ (June 27, 2014); Tom Fairless, EU Displaces U.S. as Top Antitrust Cop, WSJ (September 3, 2015); Joe Nocera, Europe s Google Problem, NYT (April 28, 2015); Michael Orey, M&A: Behind the Heat on Global Deals, BusinessWeek (March 25, 2009); Interview with U.S. Senator Herb Kohl [D-WI], Chairman, Antitrust Subcommittee, SPG Antitrust (2007); European Takeovers: To the Barricades, The Economist (Mar 2, 2006); Jeff Compo, U.S. Senators Warn EU Against Protectionism When Reviewing Mergers, International Securities Outlook Online Archive (2000).
3 decision-making has not yet been subjected to the kind of systematic empirical analysis that could test those intuitions. In this Article, we introduce a unique dataset that allows us, for the first time, to systematically examine the allegations on the Commission s protectionist use of its authority. The data include information on all of the mergers reported to the Commission between 1990 and August 2014 more than 5,000 proposed transactions. The dataset also includes, for each transaction, variables describing the extent of Commission intervention, if any, as well as industry, market size, market concentration, transaction value, and nationality associated with each transaction. Our study offers two principal findings regarding the Commission s use of its antitrust authority. First, and contrary to prior studies with much smaller sample sizes, we find evidence that the Commission has not systematically used its authority to intervene more frequently or more extensively in transactions involving a foreign or US-based firm s acquisition of an EU-based firm. Second, we also find that the Commission appears to act in a manner to promote crossborder integration. In particular, a Commission challenge is more likely where all of the parties to a transaction are headquartered in the same EU member nation, even compared to other mergers involving only European parties, suggesting that the Commission is inclined to support mergers that integrate the European economies and inclined to oppose
4 those that increase local concentration and reinforce national boundaries between EU member states. Given the extent of the Commission s authority, the stakes in this debate are high; in Europe alone, the value of mergers in acquisitions in 2014 was some $901 billion. 3 Although the Commission does not investigate every transaction, it does investigate every significant transaction that has an effect on the European market. In 2013 alone, for example, the Commission examined 274 mergers, 58% percent of which involved at least one foreign company. Only rarely, if ever, would a significant global M&A deal escape the EU s regulatory review. This makes the EU an important gatekeeper with an unparalleled ability to shape global M&A activity with its regulatory review. The Article proceeds as follows. We first describe the legal and institutional background of Commission merger review. We then examine potential determinants of the Commission s use of that authority as well as the (relatively scant) prior work on this question. Next we describe our data, and results. Finally, we briefly conclude by describing future work we intend to conduct using the unique dataset introduced here. Legal and Institutional Background The EU s merger regulation was adopted in December 1989 and entered into force in September Pursuant to that regulation, the EU s main executive body, the European Commission, reviews every merger that exceeds the revenue threshold set in the EU Treaty. Parties are obligated to notify their transaction to the Commission 3 See
5 whenever (i) the parties combined annual worldwide revenue exceeds 5 billion and each of at least two of the parties annual EU-wide revenue exceeds 250 million, or (ii) the parties combined annual worldwide revenue exceeds 2.5 billion and each of at least two of the parties annual EU-wide revenue exceeds 100 million. 4 If these thresholds are not met, national competition authorities in individual EU member states may still have the power to review the merger under national merger regulations. The Commission s jurisdiction is tied to the merger having an effect on EU market. Every time a merger affects trade between member states, the Commission has the power to intervene. The nationality of the merging party is irrelevant for the purpose of exercising jurisdiction. The 2004 Merger Regulation directs the Commission to oppose the transaction in cases where [the] concentration would significantly impede effective competition, in particular by the creation or strengthening of a dominant position, in the common market or in a substantial part of it. Such mergers are considered incompatible with the common market and hence prohibited. The old EU Merger Regulation, adopted in 1990, prohibited mergers that create or strengthen a dominant position as a result of which effective competition would be significantly impeded. After the merging parties notify their transaction to the Commission, the Commission has 25 days to reach a decision. After this initial Phase I review, the Commission can either clear the merger without subjecting it into any conditions or approve it subject to 4 A few additional conditions, not relevant to our analysis, must also be met in order for a particular merger to be reportable to the Commission.
6 conditions. 5 The Commission can also decide that more information is needed and open an in-depth investigation (Phase II). This Phase II investigation extends the review by an additional 90 days and ends with an unconditional clearance, conditional clearance or, in the most extreme cases, a decision to prohibit the merger. At times, parties withdraw the merger after its notification either in Phase I or Phase II. This could be because the economic climate has changed, the rationale for the merger no longer exists, or to preempt a negative decision. As a matter of law, the merger-review process has a suspending effect: the parties are not permitted to close the transaction before obtaining a clearance from the Commission. The Commission enjoys vast powers to review mergers. Unlike the U.S. Department of Justice, the EU Commission does not need to go to court to enjoin a merger. Instead, it has the power to reach this decision without involving the European judiciary. Many critics have noted that this procedure essentially renders the Commission the prosecutor, jury and judge of merger review in Europe. And, although the parties have the right to appeal the decision to the EU s General Court and, ultimately, the European Court of Justice, strikingly few Commission decisions in this area are challenged in the courts. Indeed, to our knowledge the Commission has been defeated in court only seven times. Although the few losses the Commission suffers in court are hard-felt, these numbers suggest that in almost all cases over 99% of mergers reviewed the Commission has the final say. Yet, as we explain below, relatively little is known about the motivations behind the Commission s action when reviewing those cases. 5 The conditions can be behavioral, such as an obligation to issue a license to a third party, or structural, such as an obligation that the combined firm divest part of its operations.
7 Theoretical Motivation and Prior Work In this Part, we consider potential theoretical reasons why or why not protectionism or market integration might motivate Commission decisions, as well as previous work related to those questions. As explained below, our dataset permits us for the first time to evaluate these theories systematically across thousands of Commission decisions over more than two decades. Protectionism Governments innate tendency to engage in trade protectionism is a well-accepted feature of international political economy. Traditionally, governments have utilized tariffs as a way to shield their industries from foreign competition. However, subsequent rounds of trade liberalization have dramatically diminished the governments ability to supply these traditional instruments of protection. Protectionism did not, however, end with the removal of tariffs. Instead, governments have increasingly turned to alternative ways to protect their domestic interests from foreign competition, including adopting various nontariff barriers. Today, protectionism is often obscured and hence harder to detect (Kono 2006). It is exercised through domestic laws and regulations that are either discriminatory in their nature or, while facially neutral, are enforced in a way that results in the suppression of foreign competition. Critics have suggested that antitrust policy is increasingly used as a tool to offset the gains from trade liberalization (Guzman 1998; Horn and Levinsohn 2001; Iacobucci ; Richardson 1999; Williams and Rodriguez 1995). Protectionist antitrust
8 policy can manifest itself, for example, in the form of a biased enforcement strategy that applies different standards to domestic and foreign firms, with antitrust agencies applying more stringent standards to foreign acquirers than domestic ones. And there are reasons to think that governments would be motivated to pursue protectionism through merger control. For example, mergers are frequently associated with employment losses, leading labor unions and local politicians to oppose mergers with an adverse effect on domestic labor conditions. In addition, antitrust agencies might have reason to be responsive to public demand to protect domestic brands. Just as many Americans may be unwilling to contemplate Coca Cola as a foreign-owned company, Europeans might have strong opposition to allowing Siemens or Mercedes-Benz be associated with anything but European industrial might. But there are compelling reasons to be skeptical that the Commission would use its merger authority to achieve protectionist aims. For one thing, a systematic bias against foreign acquirers would generate significant collateral damage, undermining the interests of many European firms as well as those of individual (European) shareholders and employees. For another, if the Commission attempted to pursue protectionist ends by way of antitrust enforcement for example, by disproportionately targeting industries where the EU has a trade deficit doing so would harm European firms that rely on production of imported goods as inputs or raw materials. 6 6 For a more extensive theoretical debate on whether countries set their antitrust policies in response to their trade flows i.e., net importers having overly stringent antitrust laws while net exporters having overly lenient antitrust laws see Andrew Guzman, The Case for International Antitrust, in COMPETITION LAWS IN CONFLICT: ANTITRUST JURISDICTION IN THE GLOBAL ECONOMY 99 (Richard A. Epstein R. and Michael S. Greve eds., 2004) and Anu Bradford, International Antitrust Negotiations and the False Hope of the WTO, 48 HARV. INT L L. J. 383 (2007).
9 The Commission would also have difficulty hiding any protectionist tendencies given the high degree of transparency underlying its merger enforcement. Unlike some antitrust agencies that publish only scant reasoning for their decisions, the Commission decisions are detailed and available for close scrutiny by the public and, in case of an appeal, by the European courts. Unlike many other antitrust authorities, including those in the U.S., the Commission is also obliged to justify any decision not to challenge a merger, extending transparency to cases it decides not to pursue and thereby making it impossible for domestic anti-competitive mergers to proceed quietly underneath the public radar. Finally, the effects doctrine underlying most antitrust laws, including EU merger control, limits the usefulness of antitrust policy for protectionist purposes, in Europe and elsewhere. The effects doctrine holds that every country may claim antitrust jurisdiction over any company as long as the company's activities have an effect on the domestic market of that country. Thus, even if European Commission let an anticompetitive mergers involving European companies proceed, the merger could face antitrust scrutiny in another jurisdiction in which merging parties have sales or assets, and hence presumably an effect on that foreign market. The Commission s ability to shield European firms from merger control is hence diluted by the prospect of concurrent and compensating jurisdiction exercised by other impacted states.
10 The above reasons combined suggest to us that antitrust would be a blunt and highly compromised instrument to deploy in service of protectionist goals. Market Integration Even in the absence of protectionism, it is plausible that the Commission uses its merger control to accomplish policy goals that go beyond a narrow focus on market efficiency. Bradford (2013) observed that across the range of regulatory policies, the Commission has followed its tendency to build and enhance the functioning of a single European market. This has equipped the Commission with an inward focus dedicated to eradicating barriers to intra-community trade as opposed to worrying about fending off foreign competition as its primary goal. An alternative motivation for Commission merger enforcement may hence be its desire to build and strengthen the common European market. This hypothesis is rooted in the historical origins of The European Union, which was created to serve both economic and political goals in an effort to guarantee peace and prosperity in post-world War II Europe. An integrated Europe was seen as allowing for a stronger Europe to emerge, capable of delivering higher rates of economic growth and better competition in global markets while tying the economic destinies of European countries together to make waging another war in Europe prohibitively costly. The removal of trade barriers among individual member states was the primary objective in the creation of the European single market. However, European leaders feared that, once public barriers to trade were removed, they would simply be replaced by private barriers
11 to trade. 7 Common European antitrust policy was therefore enacted to ensure that the gains from trade liberalization would not be offset by anticompetitive practices by private companies. Trade liberalization and antitrust policy were viewed as complements: antitrust laws serve the goal of promoting European integration, and an integrated European market promotes competition within the EU. 8 Commission decisions and European courts case law reveal that market integration was historically a central concern in European antitrust enforcement. In its early days, the Commission and the European courts explicitly referred to market integration as one of the central goals of EU antitrust law (alongside consumer welfare). 9 These references to the establishment of the single market as a key enforcement goal were common until the end of the 1990s, after which the Commission began to typically refer to the importance of pursuing consumer welfare through antitrust enforcement. 10 Market integration was never explicitly rejected as a goal of EU antitrust policy, but the Commission is typically thought of as having gradually shifted further away from that concern. 7 Certain types of anticompetitive practices would be particularly prone to compromise the goals of European integration. Vertical restraints and vertical mergers, for example, have always been a concern for the architects of an integrated Europe. For example, even without trade barriers between France and Germany, there would be no common market between the two if German companies were allowed to impose territorial restrictions that prevented French car dealers to compete with German car dealers, or if German manufacturers only acquired German distributors in an effort to bring distribution activities inhouse. The borders between member states also form natural territorial entities for market-dividing cartels as German manufacturers can agree to focus their efforts on competing domestically while letting French companies dominate the French market. 8 Steps towards a deeper economic integration: the internal market in the 21st century - Fabienne Ilzkovitz, Adriaan Dierx, Viktoria Kovacs, Nuno Sousa, see 9 Établissements Consten S.à.R.L. and Grundig-Verkaufs-GmbH v Commission of the European Economic Community, Joined cases 56 and  ECR Compare e.g., the Commission 1999 Report on Competition Policy with Commission 2006 and 2010 Report on Competition Policy.
12 In addition to these historical reasons, additional considerations may explain the Commission s persistent focus on integration even today. The Commission may take a harder line on same-country mergers within the EU because of its desire to offset member state protectionism. Individual member states are often prone to build national champions by favoring domestic acquirers of domestic targets. 11 This insular instinct cuts directly against the idea of building European-wide companies that operate across the EU and enhance the single market and the welfare of European consumers. This argument is consistent with Dinc and Erel (2013), who portray the Commission as the guardian of neutral competition policy, exercising its powers to rein in member states tendencies to favor domestic companies. It is not surprising that the Commission would consistently seek to eradicate any nationalist bias in an effort to safeguard the common market. Of course, it is possible that the Commission acts as a bulwark against individual member state protectionism yet engages in protectionism when the integrity of the single market is not under threat but where the EU interests are depicted against non-eu interests. However, an alternative view suggests that the fears of member state protectionism provide a strong incentive for the Commission to defend (and model) the importance of neutral antitrust enforcement that has no room for industrial policy. 11 Several authors make this claim without conducting any empirical analysis, For example, Jones and Davies (2014) argue that member states seek to prevent attempts by foreign firms to acquire domestic entities while supporting domestic bidders efforts to build national champions. This type of economic patriotism responds to pressures from national business, political and interest groups and stems from the desire to protect employment, industry and security. Nourry and Jung (2012) make similar arguments. Neither paper examines their claims empirically.
13 Thus, the reasons discussed above lead us to assume that, rather than protecting the European market from non-european competitors, the Commission uses its merger policy to counter member state protectionism and to facilitate pan-european mergers. Existing Empirical Research Previous empirical work on these questions has generated mixed results. Bergman et al. (2005), relying on a sample of 96 mergers notified to the Commission between 1990 and 2002, find that political variables like the nationality of the merging firms have no significant effect on the probability of an adverse ruling. Similarly, Lindsay et al. (2003) examining 245 Commission merger decisions between 2000 and 2002, do not find statistical significance on the nationality of the bidder. Aktas et al. (2004, 2007, 2012) have published a series of papers seeking to establish whether Commission merger review harbors a pro-eu bias. In their initial 2004 study, the authors found that investors anticipate higher costs to merging parties when the Commission intervened in a case involving a foreign bidder. In a 2007 follow-up piece, the authors examined a sample of 290 Commission merger decisions between 1990 and 2000, finding that the Commission is more likely to oppose a merger when the bidder is a foreign national and when the merger adversely affects European competitors. 12 But in 2012, Aktas et al reevaluated their protectionism finding. Based on an updated sample, they found that, between 2001 and 2007, the Commission no longer evinced bias against foreign bidders. By contrast, Ozden (2005) studies the 209 largest mergers between 1995 and 1999 involving at least 12 Importantly, the nationality of the bidder alone was not statistically significant in Aktas et al (2007). Instead, a bias in Commission decisionmaking materialized only when two conditions were met: first, when the bidder was a foreign firm, and second, when local competition is harmed (as measured by negative competitor returns at the time the merger is announced).
14 one US firm. That study finds that more extensive merger review is more likely if, among other things, the target is European or all US firms in the industry have high market share. Ozden concludes that the higher likelihood of merger review in cases involving a European target reveals a political and economic tendency to protect European firms. 13 We are not aware of any studies that have tested the integration hypothesis. A 2013 study by Dinc and Erel on intra-eu protectionism examines the 25 largest merger in each EU country in 1997 to 2006, finding that that individual member state governments harbor nationalist tendencies with respect to merger enforcement. Intra-EU protectionism by member states, as we have suggested, could provide a rationale for the Commission to encounter such tendencies with a vigorous policy to promote market integration at the EU level. Data and Summary Statistics Data As the above discussion indicates, empirical research on the drivers of the European Commission merger policy has been limited. While these studies shed some light on Commission s decision patterns and motivations, they have important shortcomings. Previous work, for example, has been limited to a small sample of cases, ranging from Other empirical work on European merger control has focused on establishing the decisions error rate (Duso et al 2003) or effectiveness (Duso et al. 2011). These studies are often structured as event studies, comparing stock market reactions at the time of the transaction s announcement date to their stock market reaction when the decision is reached. Other papers have considered the relative influence of various factors, such as market share or entry barriers, on Commission merger decisions (Plagnet 2005; Lindsay 2003). More recently, Clougherty and Seldeslachts (2012) have considered the effects of Commission policy on merger deterrence. Finally, other work, such as Dinc and Erel (2013), has focused on enforcement at the EU member-state level. That work often assumes the Commission to be the guardian of neutral competition policy, exercising its powers to rein in member-state tendencies to favor domestic companies in antitrust enforcement.
15 to 295 decisions. Several of the studies also suffer from significant data and design limitations. 14 In this Article, we introduce a novel dataset that includes every merger reported to the Commission between 1990 the first year of the current EU merger-control regime and August We merge these data with separately obtained information on the transaction value, market concentration, and market size. The dataset is drawn from more than seven different sources. The Commission provided data on each transaction reported to the EU, including the parties to the transaction, the role of each party (such as seller or acquirer), the nationality of each party (as reflected by the location of each firm s headquarters), and the industry or industries (identified by NACE code 16 ) for each proposed transaction. 17 We collected by hand transaction values for each transaction from the FactSet Financial Services Dataset, Factiva, Zephyr, and other sources For example, the previous work described above obtains many of its independent variables, such as the market share of the parties, from the Commission s own decisions. Because those decisions are likely written in a fashion designed to best support the Commission s outcomes, it is far from clear that such data are a reliable basis for determining the underlying drivers of Commission decisionmaking. 15 We obtained the dataset from the Commission itself, and again express our deep gratitude to the Commission for sharing these data with us. 16 NACE, which refers to nomenclature statistique des activités économiques dans la Communauté européenne, is the statistical classification of economy activity used by the European Union. It builds on the United Nations industry classification, known as ISIC, and is comparable to the SIC codes used in the United States for similar purposes. In our dataset, we record NACE industry at the six-digit level but, for simplicity, conduct our analysis at the two-digit or four-digit level. 17 Occasionally, the dataset provided to us by the Commission was missing observations or information; in those cases, we supplemented the data by hand. 18 Through the combination of these datasets as well as additional research, we were able to locate transaction values for more than one half of our transactions.
16 Next, we created a dataset on European relative market size and market concentration using sales data from Amadeus, which has European sales data for 55 million companies. As our measure of market concentration, we construct the standard Herfindahl- Hirschman index (HHI), in which market concentrations range from 0 to 1. As our measure of relative industry market size, we use industry sales divided by the total sales in that year across industries. We then constructed several additional variables of interest. Over the period we consider, firms from 97 countries have been involved in transactions reviewed by the Commission. We identified and created dummy variables signifying whether a party s nationality is among the member nations of the EU at the time the Commission received notice of the transaction that is, whether the party is an EU party or foreign party. We separately identified and created dummy variables signifying whether a party is U.S.-based. 19 We identified the role of each party to the transaction (for example, acquirer or target) using a combination of the EU Commission s designations and independent research. For our variables of interest related to the nationalities of the parties, we defined a merger as having a foreign acquirer if at least one acquirer-side party was foreign and an EU seller if at least one seller-side party had EU nationality. 20 For robustness, we include extensive alternative variable definitions in our regressions in the Appendix. 19 The dataset includes nearly 1,500 U.S. entities, many among the world s best-known brands These companies range in size from the largest in the world by gross revenue to small regional firms, and ranging in industry from airlines (Delta) to television and film (LucasFilm and Warner Brothers) to the financial sector (Goldman Sachs). 20 Throughout, we use acquirer to mean any acquirer-side party, and seller to mean any seller-side party (generally the target).
17 We construct two dependent variables that reflect the Commission s decision in each individual case. In response to a merger notification, the Commission can proceed in a number of ways. For our main specifications, we categorize the decision outcomes in binary fashion as constituting or not constituting a challenge. We consider any decision that imposes costs or delays to the parties as a challenge. Since a withdrawal of a merger is often a sign of the parties anticipating a challenge, we treat withdrawals as challenges as well. In order to better capture the richness of the Commission s powers, however, we also create a numerical index, ranging from 1 through 6, assigning values to each class of merger decisions, as shown in the table below. These numbers correspond approximately to the deterrent effect of the Commission s activity on the progress of the merger or, in other words, the more granular degree of opposition between cleared and prohibited mergers. Each number corresponds to one or more decisional articles referenced as the legal basis for the Commission s eventual decision. Decision Outcome Assigned value Ruled beyond the scope of the Commission s authority. 1 Phase I clearance without conditions. 2 Phase I clearance with conditions and obligations and mergers 3 withdrawn at Phase I The Withdrawn (N/1) decision article was placed within Category 3 to account for the balance between unconditionally cleared Phase I mergers in Category 2 and Phase II mergers in Category 4. If the parties
18 Phase II clearance without conditions. 4 Phase II clearance with conditions and obligations and mergers 5 withdrawn at Phase II. Absolute prohibition. 6 Our main specifications use the subset of 2,711 mergers for which we have transaction value, HHI, and market size data. To address the possibility that analysis of this subset leads to sample bias, we include robustness checks in the appendix in which we use all 5,124 mergers for which the structure is identifiable (that is, either an identified acquirer and seller, equal merging parties, or a joint venture). Summary Statistics We begin with summary statistics describing the frequency of mergers, including those involving foreign or U.S. parties, over time throughout our data. Figure 1 below describes those data from 1990 through 2014: withdrew the merger after or during a Phase I evaluation, the merger presumably faced a degree of opposition beyond an automatic clearance. A parties notifies the Commission, the Commission may apply conditions or otherwise cause a delay in Phase I, causing the parties abandon the merger. From a temporal and degree of opposition point of view, this places mergers withdrawn at Phase I as most comparable to mergers where the Commission applied Art. 6(1)(b) with conditions and obligations. It's possible that parties withdraw the merger for purely economic reasons unrelated to regulatory intervention, but the Decision Articles don't permit us to distinguish the cause of withdrawal. Similarly, as Category 4 refers to Phase II mergers, placing mergers Withdrawn in Phase I would be inappropriate. The same reasoning applies to why Withdrawn (N/2) mergers have been placed within Category 5.
19 FIGURE 1. TOTAL MERGERS REPORTED TO EUROPEAN COMMISSION, As Figure 1 shows, the number of mergers reported to the Commission has fluctuated with the well-known merger waves of the late 20th and early 21st centuries. In general, the proportion of mergers including foreign- (that is, non-eu) or U.S.-headquartered parties has remained relatively steady over time. On average, 48.3% of the transactions had at least one foreign party while 25.9% of the transactions had at least one US party. 44.6% of mergers feature two parties from the same EU country. We also observe the fraction of mergers that are challenged by the Commission over time. Figure 2 below describes the proportion of mergers in our dataset, including mergers with foreign or U.S.-headquartered parties, that are challenged by the Commission:
20 FIGURE 2. PERCENTAGE OF NOTIFIED MERGERS CHALLENGED BY EUROPEAN COMMISSION, Overall, 10.08% of the noticed mergers in our dataset are challenged in some way by the Commission. The challenge rate has generally fallen over time: while 11.3% of mergers were challenged in 2001, just 5.9% were challenged in This graph also reveals that, across all years, mergers involving US and foreign parties are not on average challenged at a higher rate (before controlling for any aspects of individual transactions). Finally, it may be useful to consider summary statistics for those mergers that, on the one hand, feature foreign acquirers and EU-based sellers that is, those that may raise protectionist concerns and those that do not. Figure 1 below provides summary statistics for each group of mergers in our sample, respectively: [Insert Table 1 Here.]
21 As Table 1 shows, at a summary level we observe few systematic differences in the characteristics of mergers that appear, on the surface, to raise protectionist concerns and those that do not. In the Part that follows, however, we consider and examine the protectionism hypothesis more carefully. Results We begin by describing why we think the Commission s responses to merger notifications represent an especially appealing setting for empirical study of antitrust policy. Every proposed transaction that exceeds the notification thresholds in the EU is subjected to review and included in our dataset. Unlike in the United States in which the Department of Justice and Federal Trade Commission publish only decisions of mergers they oppose, the European Commission is required to publish all of its decisions, including a decision not to oppose the merger. Thus, unlike in the United States where it is far more difficult to observe the characteristics of mergers that the authorities do not contest the EU offers a complete universe of transactions and outcomes for analysis. Commission Decisionmaking Before proceeding to test the protectionist bias directly, we seek to explain the factors that play into the European Commission merger decisions. The binned scatterplot in Figure 3 shows that, as one would expect, the likelihood of Commission action increases with transaction value (along the x-axis), decreases for large markets (solid shapes, as compared to the small markets with hollow shapes), and increases for high HHI
22 industries (triangles, as compared to the low HHI industries in diamonds). 22 This is by far the most comprehensive evaluation of the factors of EU Commission decisionmaking, both in terms of number of observations and number of regressors, to date. FIGURE 3. PERCENTAGE OF NOTIFIED MERGERS CHALLENGED BY EUROPEAN COMMISSION BY TRANSACTION VALUE, MARKET SIZE AND HHI Industry-specific effects, too, are of independent interest as determinants of the Commission s exercise of its authority. Table 2 displays the ten two-digit industries with the highest and lowest t-statistics for industry effects (that is, the industries for which we are most certain there is a non-zero effect on regulatory challenges). 23 [Insert Table 2 Here.] Note that the industry fixed effect is distinct from and more informative than the raw challenge rate within the industry due to the inclusion of control variables. For instance, 22 Big markets and markets with large HHI, respectively, are defined as those above the median in those measures. For robustness, in the Appendix, Table shows similar results from a regression, which also includes various control variables, including industry fixed effects. 23 For purposes of this analysis, we limit our sample to industries with at least 10 transactions in our dataset.
23 an industry may have a high challenge rate because it has a high HHI and large transactions, but the industry would only have a high fixed effect if it has a high challenge rate even controlling for those factors. A fixed effect of.08 implies that the challenge rate on transactions within that industry increases by 8% as compared to a neutral industry. The industries that spur the most challenges are those that relate to transport or chemical manufacturing. The industries with the greatest downward effect on challenges appear to be those that are not critical components of the industrial supply chain, such as food and beverage service. Protectionism Antitrust protectionism is difficult to test rigorously, which may explain the paucity of the attempts to date to do so. We acknowledge the complexities involved in distinguishing cases that were driven by protectionist or other motivations. What we can test, however, is whether the nationality of merging firms is a statistically significant predictor of the Commission s merger challenge. Table 3 shows the results of a multivariate regression of Commission challenges on whether the merger had a foreign acquirer and EU seller: [Insert Table 3 Here.] Across all combinations of covariates and fixed effects, we find that the coefficient is negative, whereas we would expect to find a positive coefficient if there were
24 protectionism. Column (6) contains our primary specification, which includes all covariates, and shows a marginally significant negative coefficient (p-value.09). Table 4 includes 2-digit and 4-digit fixed effects with similar results. This means that, if anything, mergers with a foreign acquirer and EU seller are, if anything, less likely to be challenged. [Insert Table 4 Here.] For robustness, in the Appendix, in Table 10 we repeat this exercise, this time replacing the binary dependent variable with our numerical index outcome variable, which takes the values 1-6 depending on the degree of the challenge, with similar results. Table 11 in our Appendix performs a series of robustness checks on the analysis in Table 10. Column (1) uses the share of acquirers that are foreign and share of sellers that are EU-based rather than a simple binary dummy; column (2) redefines joint venture parties as sellerside parties; column (3) excludes all joint ventures from the dataset entirely; column (4) uses the entire set of 5,124 mergers, rather than limiting to those for which we have transaction value, HHI, and market size information; and column (5) uses year fixed effects instead of quarterly fixed effects. All have similar results and provide consistent evidence of the absence of protectionism as a variable explaining decision outcomes. Our finding of no protectionism is further bolstered by the first column of Appendix Table 17, which shows that a merger is less likely to be blocked if there is a foreign acquirer but not also a foreign seller.
25 We note that our finding of no protectionism is consistent across even the largest of mergers, as shown in Figure 4 below: FIGURE 4. PERCENTAGE OF NOTIFIED MERGERS CHALLENGED BY EUROPEAN COMMISSION BY TRANSACTION VALUE We offer two preliminary observations regarding our results. First, although we report standard errors and t-test results, they have limited relevance in this situation. Standard errors and significance tests envision a finite sample drawn from an infinitely sized population. In our case, the sample is more than half the size of the total relevant population, Commission decisions, meaning that any estimated coefficient is going to be a far more precise measure of the true coefficient than standard errors would suggest (barring sampling biases). 24 Moreover, the specification in the Appendix containing all 24 Consider, for example, an estimated sample coefficient of -1. For a sample drawn from an infinite population, the null hypothesis can be rejected if it is sufficiently unlikely that the sample coefficient would be so low if the true population coefficient were 0 or higher. For a sample that consists of half the population, however, the null hypothesis can be rejected if it is sufficiently unlikely that coefficient for the
26 transactions comprises the entire population and thus has a true standard error of 0 (aside from measurement error). Second, we wish to address the obvious concerns regarding endogeneity specifically, that mergers with a foreign acquirer and EU seller are different than those without, and that, if this unobserved difference were controlled for, we may see a positive coefficient and evidence of protectionism. Since the ideal experiment random assignment of nationalities to parties is impossible, we cannot fully address such concerns. We note, however, that any unobservable factor would have to be something that is not captured by number of parties, industry, time of transaction, transaction value, market size, or HHI; and which is correlated with party nationality. The R2 of our regressions is quite high for a microdata regression, reaching 31% when 4-digit industry fixed effects are included, leaving less room for additional explanatory variables not captured by the existing covariates. Such an omitted factor would have to change the coefficients quite substantially, since they are currently negative.25 And, as we show in Table 2, as additional covariates are added, the coefficient of interest remains consistently negative; any unobserved variable would have to behave substantially differently from the covariates we have already included. other half of the population would be 1 or higher, a substantially lower bar to clear for rejection of the null hypothesis. 25 If we were making an assertive claim of, say, a significantly negative coefficient, then the omitted factor would only need to reduce the coefficient enough to eliminate significance. We are making the far more limited claim that the true coefficient is not in the opposite direction of our estimated coefficient.
27 Anti-US Bias As noted above, the Commission has routinely been criticized for being especially interventionist when a firm based in the United States is a party to the transaction. We repeat our regression with the variable of interest a dummy variable for whether the merger has a US-based acquirer and an EU-based seller. Tables 5 and 6 and the robustness checks in Appendix Tables 12 and 13, mirror Tables 3 and 4 and Appendix Tables 10 and 11. Again, we find nothing but negative effects of US-based acquirer and EU-based seller on probability of Commission challenge, which suggest that, if anything, mergers with US acquirers and EU sellers are less likely to be challenged. This is further bolstered by the second column of Appendix Table 17, which shows that a merger is (marginally significantly) less likely to be blocked if there is a US acquirer but not also a US seller. [Insert Tables 5 and 6 Here.] Same-Country Mergers We proceed to test whether the Commission is more likely to oppose the merger when the merging parties come from the same EU member state. This, we argue, would suggest that the Commission is concerned of mergers that re-erect and reinforce national boundaries as opposed to facilitate the creation of pan-european companies that actively seek to expand their activities across EU member states. In the multivariate regressions show in Tables 7 and 8 below, we test the hypothesis that the Commission is more likely to challenge a merger in which any two parties are of the same European country nationality:
28 [Insert Tables 7 and 8 Here.] These results offer support for the hypothesis. We show that the probability of a challenge is economically and statistically significantly more likely for the transactions in our dataset involving firms from the same EU member state when controlling for transaction size. In Appendix Table 14, we show results of additional tests to ensure our results are in fact driven by same EU nation mergers. In the first and second column, we include as a control in column (1) whether there are two parties from the EU and in column (2) whether all parties are from the EU, and find that the coefficient on same-eu-nation remains marginally significant. In column (3), we test as a placebo mergers with parties from the same non-eu country, and find no effect. In Appendix Table 15, we conduct a series of robustness checks: the first column redefines the variable of interest to be those with an identified acquiring party and sell-side party from the same EU country; the second redefines the variable of interest to transactions that in which all parties are from the same EU country; and the third replaces our binary dependent variable with our numerical index dependent variable. All three tests show at least marginally significant results in the main specification, and we find highly significant results in unreported regressions which include 2- or 4-digit industry fixed effects. In Appendix Table 19, we test whether mergers with parties from multiple European countries are less likely to be blocked. The coefficients are negative, though not significant. However, we observe that mergers with multiple EU parties with no cross-
29 European parties are significantly less likely to be blocked, further reinforcing that European mergers with same-eu country parties are more likely to be blocked. We note that this finding is consistent with the intuition that the Commission s mandate is to encourage market integration in Europe and, thus, that the Commission should marginally disfavor economically significant mergers within a single member state. Of course, we acknowledge that the finding is consistent with alternative hypotheses. For example, if a market were contained within a country, then an intra-country merger would increase the market concentration where an inter-country merger would not. If market concentration were driving the Commission s differential treatment of intracountry and inter-country mergers, then we would expect to see a higher coefficient on mergers within small countries, where the concentration concerns from an intra-country merger would be highest. Appendix Table 18 shows that the higher challenge rate for same-eu-country mergers holds equally for large and small countries, which is evidence that market concentration (often inherent in small market size) is not driving the higher challenge rate for intra-country mergers. In future work, we intend to conduct additional analysis that would help us distinguish among these alternative explanations. Conclusion and Future Work In this Article, we have sought to introduce a unique dataset including all mergers reported to the European Commission between 1990, the first year of Union-wide antitrust enforcement policy, and August Our data allow us to examine