Breaking Up Firms

By John Kwoka & Tommaso Valletti¹

I. Introduction

Many features of antitrust policy continue to reflect strongly held priors rather than good economics and evidence. For a recent example of this, one need look no further than the question of whether policy should view breaking up consummated mergers and dominant firms as a legitimate tool. We are routinely told that corporate breakups are impossible, or too costly and difficult, or counterproductive, and certainly unwise for competition agencies even to consider.

One might expect these arguments to be followed by examples of breakups that proved these points, that is, where breakups were unworkable or disastrous and ex post generally seen as unwise efforts by competition authorities. Paradoxically, the most common references in these discussions are to antitrust cases such as Standard Oil and AT&T. These are truly odd citations because both are considered successful breakups — not perfect, to be sure, but hardly the disasters that breakup opponents forewarn.

Moreover, the historical record includes many other examples of successful breakups by competition agencies. Some of these have been dominant firms, while others have involved consummated mergers. Regulators, too, in various countries have broken up hundreds of firms in major industries, including telecom, electricity, and rail, among others. Beyond that, each year companies themselves initiate almost as many divestitures as they do mergers, further demonstrating the feasibility of breakups. Notably, Johnson & Johnson, General Electric, and Toshiba have recently announced their plans to break up or spin off major units, following Exelon, Pfizer, and others earlier this year.

Breakups are, in short, not infeasible, not uncommon, and certainly not doomed to failure. To be clear, we are not advocating that any particular firm be broken up, let alone all firms that are large or have certain characteristics. Rather, we argue that competition agencies should view breakups as a viable and useful policy option under certain circumstances.

In this note we will explain the relevant framework for determining the right circumstances for a breakup to be an effective tool. We then evaluate breakups’ actual use in several different settings — the AT&T case among others as relevant examples — and draw from these experiences some lessons and principles that should guide their future use.

II. Designing Policy Toward Consummated Mergers and Dominant Firms

Although there has been considerable discussion about breaking up both dominant firms as well as consummated mergers, our discussion begins with the latter. One reason is that a focus on mergers highlights several key analytical issues raised by breakups as a policy tool. These include ex ante vs. ex post policy actions, prediction accuracy, and costs. A second reason is that a great deal of the tech companies’ size has in fact resulted from their merger activity. The five major tech companies — Amazon, Apple, Facebook, Google, and Microsoft — have collectively acquired more than 1,000 companies over the past twenty years,² and many of the competitive concerns surrounding these companies stem from conduct related to these acquisitions. We shall return to the issue of breaking up dominant firms later, but for now, we study merger breakups.

A. Consummated Mergers

Merger enforcement has almost always been an exercise in prediction. The competition agency, alerted to a merger of some consequence, uses various quantitative and qualitative tools to predict the likely competitive effects of the merger if consummated. But evidence indicates that economic predictions of the outcomes of mergers are too often incorrect, erring systematically toward permissiveness.³ The reasons for the error rate lie with the limitations of economic analysis, which applies theoretical and empirical understanding to each new merger that may well differ from models or past experiences. The reasons for the systematic error favoring “clearance” (approval) of mergers are caution on the part of the competition agency together with a high judicial standard of proof that a merger is likely to be anticompetitive before prohibiting it.

But if ex ante predictions suffer from these unavoidable flaws and limitations, good policy design requires consideration of ex post intervention into those approved mergers that prove to be anticompetitive. Otherwise, erroneously permissive ex ante predictions amount to a permanent license to harm competition. Keys to any ex post policy are the incremental gains in information from observation of post-merger conduct and the relative costs associated with ex post policy action. In most cases, the incremental gains in information are self-evident. At some point, the merged firm will act in a manner that either confirms the original competitive concerns or makes clear that approval was the appropriate policy. This determination can be facilitated by a requirement that approved mergers file basic information on the key variables of concern with the competition agency for a reasonable period after the merger.

The other factor to be considered is the incremental cost of any ex post breakup. Some assert that challenge and prohibition to a proposed merger typically imposes few costs relative to breakup of a consummated merger, which is routinely said or at least implied to be prohibitive, and by itself a reason to avoid breaking up merged firms. There is scant evidence on this question, but what exists does not support that view. One of the few systematic investigations was completed by Bain & Company.⁴ It studied some forty self-initiated breakups and found one-time breakup costs averaging about 1 percent of revenues, mostly incurred in the year of the breakup and involving administrative expenses. Perhaps more to the point, Turnstall offers an insider’s account of the massive, court-mandated AT&T spinoff of its long-distance operation and breakup of its local service into seven geographical divisions.⁵ He concludes, “divestiture was accomplished in minimal time, with the least possible impact of the corporation’s constituencies, and with no major disruption in the nation’s telephone service.”⁶

These studies and experiences underscore two important points. First and foremost, though breakups do have costs, they obviously can be successfully managed. One database of U.S. enforcement actions against consummated mergers reports 41 post-closing divestitures since 2006.⁷ While most divestitures were completed shortly after consummation and perhaps therefore involved little additional cost, eleven occurred at least three years after closing and more likely after at least some integration. Second, as illustrated by the AT&T breakup, divestiture is facilitated by the existence of natural “fault lines” between a company’s stages and divisions. Mergers often leave similar fault lines between the merging entities. These represent obvious starting points for a competition agency to consider the scope of post-merger divestitures and breakups.

The Bain study focused on breakups that firms themselves decided to undertake. Although some issues are therefore different, these experiences make clear that breakups are entirely feasible and not prohibitively costly. Indeed, one study found a total of 1611 self-initiated breakups by Fortune 100 companies in the 1990s, a number more than two-thirds the number of acquisitions by these same companies during the same period.⁸

B. Dominant Firms

Dominant firms that engage in anticompetitive actions raise issues that are in part similar and in part different from those above. As noted earlier, the major tech companies are in no small part constructs built up from large numbers of mergers and acquisitions, and many of the competitive concerns they raise are the result of these acquisitions. These competitive concerns involve matters such as self-preferencing, display bias, and misuse of competitive information. These could in principle be addressed by undoing the mergers using the same method as described above. That method involves identifying the fault lines between the businesses and uncoupling the pieces along those lines.⁹

Competition agencies have been reluctant to attempt to break up such firms. They have been wary of difficulties and hazards of doing so and of the consequences if a breakup proves unsuccessful. Some caution is appropriate, of course; in industries subject to tipping such as the tech industry, ex ante policy is all the more important in order to preserve any emerging competition. That said, optimal policy design makes clear why strict avoidance of breakups is both unwise and unnecessary. One reason already described is that there are many precedents and examples of successful breakups, while systematic contrary examples are rare. The other reason is that the alternatives commonly proposed to address anticompetitive activity by the dominant tech companies have proven themselves generally ineffective.

What are those alternatives? The basic alternative is a “rules and remedies” approach. This approach does not involve breakup or divestiture, or any structural change to the dominant company. Rather, it imposes rules on the company to prevent it from engaging in the specific anticompetitive conduct at issue. For example, it may instruct the firm to avoid biasing search results in its favor or require it to disclose information necessary to allow for viable competition. This approach, of course, is essentially the same as so-called conduct or behavioral remedies for anticompetitive mergers, with much the same rationale: if the specific offensive conduct can be prevented, structural reform is not necessary.

The problem with “rules and remedies” is that considerable experience now reveals the inherent flaws in this approach, and evidence makes clear its ineffectiveness. The fundamental problems are that (1) these rules and remedies instruct the firm to act against its own interests, that is, ordering it to engage in conduct that reduces its profit; (2) consequently, the firm is incentivized to seek to avoid, evade, or otherwise minimize the effect of the rule; and (3) the firm has advantages compared to the agency overseeing the remedy in terms of informational asymmetries, technology management, and pretextual advantages in succeeding in defeating the restraint.

Two examples serve to illustrate the point. The U.S. government settled its case against Microsoft with a conduct or behavioral remedy that sought to make it stop certain practices and to provide outside firms with certain protocols that would allow them to compete. With respect to the latter, the company immediately claimed it could not comply in the initial three-year time frame, then that it could not do so within five years. Each time, the court “blinked,” unable to assess the merits of Microsoft’s excuses, and extended the deadline for compliance. The necessary protocols were finally provided, albeit in not quite complete form, nearly a decade after the issues were decided and order put in place — and long after they were important for competition.¹⁰

In the EU, Google has long been under orders from the European Commission to open up its shopping site and, separately, to untie Google applications and services, such as its Play Store, distributed on Android devices with other Google applications such as its search engine.¹¹ It has repeatedly delayed or undermined compliance, despite billions of euros in fines. As asserted above, these types of orders or remedies generally are ineffective since they basically instruct the firm to act against its own interest—indeed, to weaken its own creation. Such mandates are fundamentally at odds with the incentives of the company, so the company will predictably minimize or evade the intended constraint. And the information, technology, and tools to do so are very much in their hands.¹²

Faced with these factors, regulatory agencies have often employed structural changes to resolve otherwise intractable competitive problems. Dozens of electricity companies in the United States have undergone both vertical and horizontal breakups. In other countries, dozens of telecoms, hundreds of electricity companies, as well as railroads and other integrated infrastructure companies have been broken up. Regulators often first attempt to address the problems with rules and remedies, only to discover — for the reasons we have described — the futility of those efforts.

In light of this considerable experience and evidence demonstrating the lack of a viable alternative, we conclude that where the essential competitive problem with a company is its structure — in the sense that its anticompetitive behavior flows inexorably from that structure and is otherwise difficult if not impossible to prevent — it follows that the necessary solution likely lies in altering that structure. Efforts to address the competitive problems with these companies through rules and remedies is a deeply flawed strategy, one that may sound appealing because it avoids confronting the hard reality of a breakup, but one that has not worked and does not work. We have run that experiment and know how it comes out.

To be sure, the same structural remedy is probably not the solution for the core platform monopoly that lies at the heart of many tech companies. Fragmenting a platform natural monopoly will not likely produce competing platforms. Unless there is substantial differentiation among the resulting pieces (which can in fact happen), those pieces of the original platform will likely coalesce as network economies and tipping drive reconsolidation. Other measures instead of, or together with, a breakup are likely needed in this case.

III. Recommendations and Conclusions

The above analysis makes clear that firm breakups are quite realistic and, indeed, not unusual. In the case of consummated mergers, they represent an essential complement to ex ante policy based on inherently imperfect predictions. In the case of dominant tech companies, breakups address a crucial source of their growth and competitively problematic behavior. Experience with breakups of consummated mergers reveals a record of considerable success, whereas efforts to find alternatives to breakups in both cases — mergers and tech company conduct — have regularly failed.

In this setting breakups represent a viable and effective tool for competition agencies — not an entirely novel one, nor one that should be used automatically, but certainly not one to be avoided due to alleged but unsubstantiated concerns over costs or ineffectiveness.

Click here for a PDF version of the article

1 John Kwoka, Northeastern University; Tommaso Valletti, Imperial College London. This note reflects the ideas contained in our recently published academic article, John Kwoka & Tommaso Valletti, Unscrambling the Eggs: Breaking Up Consummated Mergers and Dominant Firms, INDUS. & CORP. CHANGE, Aug. 31, 2021, https://academic.oup.com/icc/advance-article/doi/10.1093/icc/dtab050/6360491.

2 Geoffrey Psrker, Georgios Petropoulos & Marshall Van Alstyne, Platforms Mergers and Antitrust, INDUS. & CORP. CHANGE, Sept. 7, 2021, https://academic.oup.com/icc/advance-article/doi/10.1093/icc/dtab048/6365871. Key to Facebook’s growth, for example, has been its acquisitions of WhatsApp and Instagram and perhaps now Oculus. Google’s dominance in digital advertising has in large part been the result of its acquiring DoubleClick, AdMob and other elements of the ad stack, with new extensions such as Fitbit. Amazon, Apple, and Microsoft have acquired numerous complementary, upstart or nascent competitors.

3 See John Kwoka, MERGERS, MERGER CONTROL, & REMEDIES (2015).

4 Andy Pasternak & Jim Wininger, Mastering the Good Breakup, BAIN & CO. (July 11, 2014), https://www.bain.com/insights/mastering-the-good-breakup/.

5 W. Brooke Turnstall, DISCONNECTING PARTIES: MANAGING THE BELL SYSTEM BREAK-UP, AN INSIDE VIEW (1985).

6 Id. at 160­–61. We note Fox & Baker’s characterization of the AT&T case as “unusual,” apparently because of the operational success of divestiture. See Eleanor M. Fox & Donald I. Baker, Antitrust and Big Tech Breakups, CPI ANTITRUST CHRON., Oct. 2021 (2), at 11. Elsewhere we have shown the numerous component parts of the tech companies, which provide in principle the same “fault lines” that could be used, with justification, for breakups. See Kwoka & Valletti, supra note 1. The Fox & Baker criticism of the AT&T case appears largely focused on trial management, which is universally agreed to have been very weak, Fox & Baker, supra note 6, at 21, especially in the face of the determined opposition of the company to the suit. That seems a different point, inviting a different solution than conceding to continued market power.

7 Consummated Mergers Antitrust Enforcement, PRACTICAL L. (2021), https://www.westlaw.com/4-525-8653?transitionType=Default&contextData=(sc.Default)&VR=3.0&RS=cblt1.0.

8 See Belen Villalonga & Anita M. McGahan, The Choice Among Acquisitions, Alliances, and Divestitures, 26 STRATEGIC MGMT. J. 1183 (2005).

9 Indeed, the Federal Trade Commission has embarked on precisely this strategy to address concerns with Facebook, by seeking to break it up along the fault lines of its acquisitions of Instagram and WhatsApp.

10 See generally Carl Shapiro, Microsoft: A Remedial Failure, 75 ANTITRUST L.J. 739 (2009) (discussing the remedial process of the case and the alleged failure of the Final Judgment to restore competition).

11 See, e.g. Simon Van Dorpe, Vestager’s Court Win Opens Way for More Google Cases, POLITICO (Nov. 10, 2021), https://www.politico.eu/article/eu-commission-margrethe-vestager-wins-google-shopping-case/ (“After the Shopping decision [including a fine of €2.4 billion], Vestager fined Google a record €4.34 billion in 2018 over its Android operating system; and in another case fined the company €1.49 billion over advertising on other websites in 2019.”).

12 In his review, Jenny states that “a limited number of cases brought by the European Commission against Microsoft, Google or Amazon have not had a tangible effect on their behavior.” Frederic Jenny, Competition Law and Digital Ecosystems: Learning to Walk Before We Run, INDUS. & CORP. CHANGE, Sept. 1, 2021, https://academic.oup.com/icc/advance-article/doi/10.1093/icc/dtab047/6360667.