The Oxford Handbook of International Antitrust Economics 9780199388592, 0199388598

Citation preview

T H E O X F O R D HA N D B O O K O F

I N T E R NAT IONA L A N T I T RU ST E C ON OM IC S Volume 2

CONSULTING EDITORS Michael Szenberg

Lubin School of Business, Pace University Lall Ramrattan

University of California, Berkeley Extension

THE OXFORD HANDBOOK OF

INTERNATIONAL ANTITRUST ECONOMICS Volume 2 Edited by

ROGER D. BLAIR and

D. DANIEL SOKOL

1

3 Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford New  York Auckland  Cape Town  Dar es Salaam  Hong Kong  Karachi Kuala Lumpur Madrid Melbourne Mexico City Nairobi New Delhi Shanghai Taipei Toronto With offices in Argentina Austria Brazil Chile Czech Republic France Greece Guatemala Hungary Italy Japan Poland Portugal Singapore South Korea Switzerland Thailand Turkey Ukraine Vietnam Oxford is a registered trademark of Oxford University Press in the UK and certain other countries. Published in the United States of America by Oxford University Press 198 Madison Avenue, New  York, NY 10016 © Oxford University Press 2015 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by license, or under terms agreed with the appropriate reproduction rights organization. Inquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above. You must not circulate this work in any other form and you must impose this same condition on any acquirer. CIP data is on file at the Library of Congress ISBN 978–0–19–938859–2

1 3 5 7 9 8 6 4 2 Printed in the United States of America on acid-free paper

Contents

List of Contributors Introduction

ix xiii

PA RT I   M ON OP OL I Z AT ION :  C ON DU C T 1. A Framework for the Economic Analysis of Exclusionary Conduct B. Douglas Bernheim and Randal Heeb

3

2. Predatory Pricing Kenneth G. Elzinga and David E. Mills

40

3. Raising Rivals’ Costs David T. Scheffman and Richard S. Higgins

62

4. Predatory Buying John E. Lopatka

72

5. Competitive Discounts and Antitrust Policy Kevin M. Murphy, Edward A. Snyder, and Robert H. Topel

89

6. Squeeze Claims: Refusals to Deal, Essentials Facilities, and Price Squeezes Barak Orbach and Raphael Avraham 7. Innovation and Antitrust Policy Thomas F. Cotter 8. Continental Drift in the Treatment of Dominant Firms: Article 102 TFEU in Contrast to Section 2 Sherman Act Pierre Larouche and Maarten Pieter Schinkel 9. Treatments of Monopolization in Japan and China Ping Lin and Hiroshi Ohashi

120 132

153 188

vi   Contents

10. Monopolization in Developing Countries Alberto Heimler and Kirtikumar Mehta

234

11. Business Strategy and Antitrust Policy Michael J. Mazzeo and Ryan C. McDevitt

253

PA RT I I   V E RT IC A L I N T E G R AT ION A N D C ON T R AC T UA L E Q U I VA L E N T S 12. Resale Price Maintenance of Online Retailing Benjamin Klein

277

13. Exclusive Dealing Howard Marvel

304

14. Tying Arrangements Erik Hovenkamp and Herbert Hovenkamp

329

15. Vertical Restraints across Jurisdictions Edward M. Iacobucci and Ralph A. Winter

351

16. Franchising and Exclusive Distribution: Adaptation and Antitrust Francine Lafontaine and Margaret E. Slade

387

PA RT I I I   C OL LU SION A M ON G O S T E N SI B L E C OM P E T I TOR S 17. Cartels and Collusion: Economic Theory and Experimental Economics Jay Pil Choi and Heiko Gerlach

415

18. Cartels and Collusion: Empirical Evidence Margaret C. Levenstein and Valerie Y. Suslow

442

19. Tacit Collusion in Oligopoly Edward J. Green, Robert C. Marshall, and Leslie M. Marx

464

20. Auctions and Bid Rigging 498 Ken Hendricks, R. Preston McAfee, and Michael A. Williams

Contents   vii

21. Screening for Collusion as a Problem of Inference Michael J. Doane, Luke M. Froeb, David S. Sibley, and Brijesh P. Pinto

523

22. Competition Policy for Industry Standards Richard Gilbert

554

23. Antitrust Corporate Governance and Compliance Rosa M. Abrantes-Metz and D. Daniel Sokol

586

Index

619

List of Contributors

Rosa M.  Abrantes-Metz, Adjunct Associate Professor, Stern School of Business, New York University and Principal, Global Economics Group Raphael Avraham, Associate, Snell & Wilmer B. Douglas Bernheim, Edward Ames Edmunds Professor of Economics, Stanford University and Partner, Bates White Economic Consulting Roger D.  Blair, Walter J.  Matherly Professor and Chair, Department of Economics, Warrington College of Business, University of Florida Jay Pil Choi, Scientia Professor, School of Economics, Australian School of Business, University of New South Wales and Professor of Economics, Yonsei University Thomas F. Cotter, Briggs and Morgan Professor of Law, University of Minnesota Law School Michael J. Doane, Director, Competition Economics LLC Kenneth G. Elzinga, Robert C. Taylor Professor of Economics, University of Virginia Luke M. Froeb, William C. Oehmig Chair of Free Enterprise and Entrepreneurship, Owen Graduate School of Management, Vanderbilt University Heiko Gerlach, Associate Professor, School of Economics, University of Queensland Richard Gilbert, Emeritus Professor of Economics and Professor in the Graduate School, University of California, Berkeley Edward J. Green, Professor of Economics, Penn State University Randal Heeb, Partner, Bates White Economic Consulting Alberto Heimler, Professor of Economics, Italian School of Public Administration Ken Hendricks, Professor of Economics, University of Wisconsin Richard S. Higgins, Director, Berkeley Research Group Erik Hovenkamp, JD/PhD Candidate, Graduate School, Economics, Northwestern University Herbert Hovenkamp, Ben V. & Dorothy Willie Professor of Law, University of Iowa

x   List of Contributors Edward M. Iacobucci, Professor of Law, Osler Chair in Business Law, and Associate Dean of Research, University of Toronto Law School Benjamin Klein, Professor Emeritus of Economics, University of California, Los Angeles Francine Lafontaine, William Davidson Professor of Business Economics and Public Policy, Professor of Economics, Ross School of Business, University of Michigan Pierre Larouche, Professor of Competition Law, Tilburg Law and Economics Center (TILEC), Tilburg Law School, Tilburg University Margaret C. Levenstein, Executive Director, Michigan Census Research Data Center; Research Scientist, Survey Research Center, Institute for Social Research; Adjunct Professor of Business Economics and Public Policy, Ross School of Business, the University of Michigan Ping Lin, Professor, Department of Economics, Lingnan University of Hong Kong John E. Lopatka, A. Robert Noll Distinguished Professor of Law, Dickinson School of Law, the Pennsylvania State University Robert C.  Marshall, Department Head, Professor, Department of Economics, Pennsylvania State University Howard Marvel, Professor Emeritus of Economics, The Ohio State University Leslie M.  Marx, Robert A.  Bandeen Professor of Economics, The Fuqua School of Business, Duke University Michael J. Mazzeo, Associate Professor, Department of Management and Strategy and Faculty Associate, Institute for Policy Research, Northwestern University R. Preston McAfee, Research Leader, Google Ryan C. McDevitt, Assistant Professor of Economics, The Fuqua School of Business, Duke University Kirtikumar Mehta, former Cartels Director, Directorate General for Competition, European Commission David E. Mills, Professor, Department of Economics, University of Virginia Kevin M.  Murphy, George J.  Stigler Distinguished Service Professor of Economics, University of Chicago Booth School of Business Hiroshi Ohashi, Professor of Economics, University of Tokyo Barak Orbach, Professor of Law and the Director of the Business Law Program, University of Arizona College of Law Brijesh P. Pinto, Senior Consultant, Competition Economics LLC David T. Scheffman, Director, Berkeley Research Group

List Of Contributors   xi

Maarten Pieter Schinkel, Professor of Economics, Amsterdam Center for Law & Economics (ACLE), University of Amsterdam David S. Sibley, Professor, Department of Economics, University of Texas at Austin Margaret E. Slade, Professor Emeritus, Vancouver School of Economics, The University of British Columbia Edward A.  Snyder, Dean & William S.  Beinecke Professor of Economics and Management, Yale School of Management D. Daniel Sokol, Professor, Levin College of Law, University of Florida and Senior Research Fellow, George Washington University Law School Competition Law Center Valerie Y.  Suslow, Associate Dean of Graduate Programs, Professor of Business Economics and Public Policy, Louis and Myrtle Moskowitz Research Professor of Business and Law, Ross School of Business, University of Michigan Robert H. Topel, Isidore Brown and Gladys J. Brown Distinguished Service Professor in Urban and Labor Economics, The University of Chicago Booth School of Business Michael A. Williams, Director, Competition Economics LLC Ralph A. Winter, Canada Research Chair, Business Economics and Policy and Professor, Strategy and Business Economics Division, Sauder School of Business, University of British Columbia

Introduction RO GER D. BLAIR AND D. DANIEL SOKOL

Antitrust economics is a subset of industrial organization economics. What makes antitrust economics rather unique is the centrality of economic analysis to the development of antitrust law and policy. In the United States antitrust economics guides all antitrust analysis by government enforcers (at the federal level the Department of Justice Antitrust Division and the Federal Trade Commission) and courts. In other systems, the centrality of antitrust economics to antitrust law (typically called competition law) and policy has not been established. Instead, cutting edge antitrust economic analysis competes with non-antitrust economics goals. Nevertheless, across the major non-US jurisdictions, antitrust economics is far more utilized now than previously. With global mergers and various types of conduct, increased coordination across agencies, practitioner lawyers and economists around the world trained in the latest theories of antitrust economics, and a rise of economic analysis in decision-making by adjudicators, the increasing role of international antitrust economics seems somewhat inevitable. The desire to provide scholars and policy-makers across jurisdictions a reference tool to understand the most important developments in antitrust economics motivates this handbook. We have assembled many of the most important scholars in the field to provide overviews and analysis of the core issuers in antitrust economics. Although no handbook can be exhaustive, we have attempted to cover all of what we believe to be the major topics in the field. The developments in economic analysis across these areas that the handbook covers will shape policy and legal issues in the field for some time. We hope that the handbook will provide inspiration for new avenues of theoretical and empirical research in the field. Many people deserve thanks for this book. The project took a number of years to complete. Our editors at Oxford University Press deserve our gratitude for their patience and excellent editing. Coordinating production across so many chapters was not always easy. We particularly thank those authors who turned in their work in a timely manner.

T H E O X F O R D HA N D B O O K O F

I N T E R NAT IONA L A N T I T RU ST E C ON OM IC S Volume 2

PA R T I

M ON OP OL I Z AT ION : C ON DU C T

CHAPTER 1

A F R A M E WO R K F OR T H E E C O N OM IC A NA LYSI S OF E XC LU SIO NA RY C ON DU C T B. D OUGLAS BERNHEIM AND RANDAL HEEB

1.1. Introduction “Exclusionary conduct” is a phrase commonly used to describe practices that a firm might undertake to deny a rival access to a market, or some portion thereof. The appropriate antitrust treatment of such conduct is a matter of spirited debate among both economists and legal scholars. The courts have likewise struggled to articulate consistent standards governing the legal status of various practices potentially deemed exclusionary, and some commentators go so far as to pronounce a circuit split between the applicable precedential cases, especially noting the apparently conflicting legal and economic principles articulated in 3M v. LePage’s in the Third Circuit and Cascade Health Solutions v. PeaceHealth in the Ninth Circuit (see, e.g., Jaeckel 2010; Markus 2008; Hungar and Koopmans 2009). The lack of any clear consensus on principles heightens the risk that courts will inadvertently establish economically counterproductive precedents.1 Legal practitioners complain that the resulting ambiguity precludes them from counseling clients effectively, leaving companies uncertain as to whether any given mode of conduct is 1 

This lack of consensus and the resulting prescriptive confusion was perhaps most visibly on display following the DOJ’s issuance of the Single Firm Conduct report in September 2008. The report emerged from lengthy joint hearings held by the DOJ and FTC, but was issued independently by the DOJ and instantly disavowed by four commissioners of the FTC (http://www.ftc.gov/opa/2008/09/section2.​ shtm). The DOJ subsequently withdrew the report in May 2009.

4   B. Douglas Bernheim and Randal Heeb permissible or proscribed, and hence reluctant to pursue business strategies that might be procompetitive. Devising clear and broadly applicable principles for analyzing allegedly anticompetitive exclusionary conduct is challenging in large part because that heading subsumes a wide range of diverse practices, including exclusive dealing and other agreements that limit a customer’s or supplier’s ability to do business with a rival, predatory pricing, bundled pricing, tying, and loyalty discounts. While these practices are unquestionably related, each differs from the others in potentially substantive ways. Those differences contribute to the multiplicity of models and theories pertaining to exclusionary conduct found in the scholarly literature, as well as to the disparate conduct-specific legal precedents. As a result, the topic of exclusionary conduct is widely perceived as complex, confusing, and unsettled. Despite the manifest confusion surrounding these issues, we maintain that it is possible to distill from the extant body of scholarly thought a single, consistent, and tractable economic framework for analyzing the antitrust implications of a wide variety of exclusionary practices. Analytic unification is possible because all of the exclusionary practices listed above share an ability to activate a single economic mechanism. Briefly, the mechanism has the feature that exclusion from a market or some portion thereof weakens a rival, and thereby impairs its ability to compete for other business.2 Our preference for a single, unified framework—rather than a conceptually distinct approach for each pertinent practice—reflects our belief that this mechanism overwhelmingly provides the most commonly and persuasively alleged pathway through which exclusionary practices potentially harm consumers. From that premise, it follows that the core objective of any inquiry into the potential anticompetitive effects of exclusionary conduct should be to detect and evaluate the mechanism’s operation, a task that naturally calls for a unified approach. While we acknowledge that the scholarly literature has explored other economic mechanisms by which exclusion might achieve anticompetitive ends, we view these as considerably less important in practice, and better handled as special cases and/or exceptions to general principles (much as the Horizontal Merger Guidelines handle failing firms).3 With this perspective in mind, we describe a framework for systematically implementing a rule-of-reason inquiry into the competitive effects of exclusionary practices. In the first of three stages, one asks whether the conduct is in fact exclusionary. (To be absolutely clear from the outset, an affirmative answer to this first question does not signify that the conduct is necessarily anticompetitive.) For that purpose, we divide 2  The “other business” often involves future sales for which exclusion will render the rival less competitive, but it may also consist of contemporaneous sales to other customers, for instance in other geographic regions. 3  For example, a pattern of patent abuse that prevents an established rival from entering a market can potentially achieve anticompetitive ends through exclusion without impairing a rival’s ability to compete for other sales. Similarly, a network of vertical arrangements with a supplier might be used to coordinate and discipline collusive horizontal agreements among downstream firms, or to deter entry that might diminish overall industry profits. See Asker and Bar-Isaac (2014).

A Framework for the Economic Analysis of Exclusionary Conduct   5

conduct into two broad classes: practices that involve “exclusionary conditions” and those that do not. We define an exclusionary condition as any practice that renders aspects of transactions between a company and any one of its customers or suppliers effectively contingent upon that party’s dealings with the company’s rival(s).4 Practices involving exclusionary conditions include exclusive dealing (both total and partial exclusivity), loyalty discounts (with discounts tied to purchase shares),5 and myriad restrictions on customers’ dealings with rivals. Potentially exclusionary practices that do not involve exclusionary conditions include predatory pricing, simple bundling, volume discounts, and conventional tying.6 For the latter practices, the principle lever over the customer’s behavior is price, which the excluding firm links to the volume and/or composition of its own sales to the customer without conditioning on the customer’s purchases from rivals.7 For conduct involving exclusionary conditions, some degree of exclusion is generally explicit, in the sense that those conditions preclude or discourage incremental purchases from (or sales to) rivals, even if there is no change in purchases from (or sales to) the excluding firm. Consequently, establishing that such conduct is exclusionary (but not necessarily anticompetitive), as required in the first stage of the analysis, involves little more than documenting the conditionality. In contrast, when exclusionary conditions are not employed, determining whether a particular practice serves exclusionary and nonexclusionary objectives is more challenging, and practical compromises are unavoidable. For example, a firm may reduce prices either to weaken a rival by depriving it of sales or simply to win more business. In that case, because the conduct at issue closely resembles the most common form of procompetitive behavior, subjecting it to close scrutiny creates a substantial risk of chilling beneficial rivalry. Consequently, it is appropriate to set a high bar for establishing that low prices are exclusionary (but not necessarily anticompetitive), as required in the first stage of the analysis; applying a price-cost test represents a reasonable compromise between the conflicting public policy objectives. A more nuanced standard is 4 

Fiona Scott Morton (2012) coined the term “contracts referencing rivals” to describe agreements containing such terms, implicitly or explicitly. 5  We use the term “loyalty discounts” here to mean discounts that are conditional on the share of the customer’s business given to the discounting firm and, consequently, the share of business given to rivals, as opposed to volume discounts, which are conditional only on the volume but not the share purchased from the discounting firm. Some authors call this practice “fidelity discounts” or “market-share discounts.” A more descriptive term might be “purchase-share discounts” or “partial exclusivity discounts.” These terms are not used consistently by all commentators. Note that while nondiscriminatory volume discounts do not entail exclusionary conditions, they can achieve the same ends as loyalty discounts if they are tailored to the customer’s size or market potential. 6  Conventional tying (a “tie-in”) involves offering two or more products together when at least some of those products are not offered separately; it does not require exclusionary conditions. Negative tying (a “tie-out”) involves forbidding a customer from buying a particular product from a rival as a condition of sale for another product; it does involve exclusionary conditions. However, for reasons discussed later, that distinction turns out to be of relatively little practical significance within our framework. 7  A similar statement pertains to practices involving relationships with suppliers rather than with customers.

6   B. Douglas Bernheim and Randal Heeb appropriate when the conduct resembles modes of nonexclusionary competition that have ambiguous implications for consumer welfare to begin with, such as price discrimination, or that are relatively uncommon, so that the downside consequences of chilling legitimate rivalry through scrutiny are modest. In the second stage of the inquiry, one examines whether the exclusionary conduct has anticompetitive effects. Such an investigation is naturally structured around empirical manifestations of the main anticompetitive mechanism. Specifically, one asks whether, by virtue of exclusion, a rival’s ability to compete for other business is impaired, so that market power is enhanced and consumers are harmed. One must also conduct a factual investigation of the industry to determine whether the types of market failures that are prerequisites for anticompetitive exclusion are present. Upon reaching a determination that exclusionary conduct has anticompetitive effects, the inquiry proceeds to the third stage, in which one then asks whether the conduct has offsetting procompetitive effects. To demonstrate the presence of such effects, one must identify the market imperfections and contracting failures for which exclusion provides a potential remedy. The task of weighing anticompetitive and procompetitive effects against each other also requires one to document consumers’ gains from the use of exclusionary measures, rather than the next best nonexclusionary remedy, to address the contracting failures. The unified analytic framework set forth herein incorporates familiar principles and approaches and is at least arguably consistent with the legal precedents established in the guiding cases pertaining to exclusion. Importantly, the framework harmonizes the apparently disparate treatment of bundled pricing practices in LePage’s, Ortho, and PeaceHealth. The recent decision by the Third Circuit in ZF Meritor explicitly addresses the distinction between price-related conduct governed by predatory pricing case law, including PeaceHealth, and exclusionary conduct governed by LePage’s, in a way that fits nicely into the framework. As we explain, our approach is also compatible with the court’s treatment of foreclosure and exclusive dealing in Dentsply and Microsoft, as well as in predatory pricing cases such as Brooke Group and Concord Boat.8 This ­chapter 1 is organized as follows: Section 1.2 describes the anticompetitive mechanism of primary concern and explains how the various aforementioned practices can activate it. Section 1.3 details the three stages of an inquiry into the competitive effects of exclusionary practices and provides brief examples. Section 1.4 discusses how to identify procompetitive effects and to balance these against anticompetitive effects. Section 1.5 concludes.

8 

Even though each of these cases is typically cited as establishing precedent for one particular type of conduct, most if not all of them involve allegations of multiple exclusionary practices, or they at least discuss economic principles that are applicable to multiple practices. For a summary of the various elements present in key cases before 2005, see Kobayashi (2005).

A Framework for the Economic Analysis of Exclusionary Conduct   7

1.2.  Why Is Exclusionary Conduct Potentially Anticompetitive? Antitrust policy is primarily concerned with the possibility that exclusionary conduct may lessen competition to the detriment of consumers. To determine whether conduct has that effect in any given instance, we must first understand how the effect might arise. Looking to the scholarly literature for guidance, one can become somewhat discouraged by the multiplicity of models and theories pertaining to exclusionary conduct, and by the absence of a clear focus on any single channel through which exclusion might affect market outcomes.9 However, the fact that scholars have explored a variety of issues pertaining to exclusion (and that journals favor studies that present novel perspectives) does not mean that it is impossible to distill useful general principles. In this section, we identify the economic mechanism that we take to be the most important reason for concern about exclusionary conduct in practice, and we identify the modes of conduct that potentially implicate that mechanism. In section 1.3, we describe a framework for analysis that is designed to detect the operation of that mechanism and gauge its effectiveness. We do not mean to suggest that other mechanisms are never important, but they appear insufficiently common to serve as the central focus for a practical analytic framework.10 Throughout the following discussion, we focus on examples in which conduct excludes an upstream firm from a portion of the downstream market (e.g., a buyer agrees to make purchases from only one seller). Similar principles apply to cases in which conduct excludes a downstream firm from a portion of the upstream market (e.g., a seller agrees to make sales to only one buyer).

9  For instance, Aghion and Bolton (1987) show how exclusive arrangements can reduce the likelihood of entry by specifying a damage fee that the entrant must pay in order to make sales to the customers; Rasmusen, Ramseyer, and Wiley (1991) and Segal and Whinston (2000) highlight how exclusive arrangements can thwart entry when the incumbent monopolist can exploit the externalities between buyers; Bernheim and Whinston (1998) explain how exclusive dealing in one market can weaken the rival in another market. Salop and Scheffman (1983) explore mechanisms that achieve exclusionary outcomes by raising rivals’ costs. Asker and Bar-Isaac (2014) show how a monopolist can support exclusion by sharing a portion of its monopoly rents with downstream firms and facilitating collusion among them so that they can avoid competing away those rents. 10  A branch of the literature suggests that exclusive dealing may provide a solution to what is known as the “hold-up problem,” thereby allowing firms with market power to exercise that power more efficiently (Hart and Tirole 1988). We do not see that consideration as the central concern for antitrust policy (though it may play a role in particular contexts).

8   B. Douglas Bernheim and Randal Heeb

1.2.1.  The Mechanism The mechanism of concern is most easily illustrated in a market with the following features. A well-established firm (the “market leader”) competes with a smaller and less established rival (possibly a recent entrant) by selling a product to a group of customers. Entry barriers are high, so the market leader is not concerned about potential competition. The degree to which the rival will pose a future competitive threat to the market leader depends on the rival’s current success in the marketplace. (We discuss possible reasons for that dependence in section 1.3.) The market leader understands that dependence and exploits it to weaken the rival in the future. Specifically, through one or more of the exclusionary devices discussed in section 1.2.2, the market leader effectively “buys” a substantial chunk of the downstream demand so that customers will not purchase the product from the rival. While customers may benefit from improved terms in the short run, the future harm to competition leaves them worse off overall.11 For the purpose of our analysis, the central feature of this example is that one customer’s decision to enter into an exclusionary arrangement with the market leader reduces the benefits that other customers can expect to derive from vendor competition. In the language of economists, this effect is an example of a “negative contracting externality” (which we will abbreviate as NCE)—that is, an adverse effect that one party experiences due to the nature of a contract between other parties. Here, the NCE results from the conjunction of three conditions: first, an exclusionary arrangement between the market leader and a customer impairs the rival’s ability to compete for future sales; second, entry is difficult (which typically implies that the market leader has market power), so that a reduction in the rival’s competitive efficacy leads to an overall decline in competition among vendors; and third, vendors have many potential customers, so that the burden of reduced competition resulting from the exclusionary arrangement is borne, at least in part, by customers who were not part of that arrangement. In the absence of NCEs, vertical agreements that exclude rivals are generally procompetitive. To understand why that is the case, observe that any negotiation involving a vendor and a customer consists of two separable components, one encompassing all considerations that determine the size of the pie they expect to split and the other being the division of that pie. With respect to the second component, the vendor’s and customer’s interests are diametrically opposed; however, with respect to the first component, they are perfectly aligned, in the sense that both parties benefit from maximizing the pie’s size.12 Consequently, when a vendor and a customer enter into an exclusionary 11 

For analyses that share the essential features of this example, see Bernheim and Whinston (1998, section IV); Rasmusen, Ramseyer, and Wiley (1991); and Segal and Whinston (2000). 12  If an agreement between a vendor and a customer did not maximize the sum of their economic benefits, then one could propose an alternative that would make both better off, and the other would accept. This principle presupposes that vendors and customers can freely divide up the benefits they expect to derive from an agreement without otherwise altering the substance of the agreement—a condition that rarely fails to hold.

A Framework for the Economic Analysis of Exclusionary Conduct   9

arrangement, one can infer that exclusion maximizes the sum of their economic benefits. What about the economic benefits enjoyed by the customer? There is no reason to think that the mere possibility of agreeing to an exclusionary contract would improve the vendor’s negotiating position with respect to the customer and thereby allow it to extract a larger share of the pie. Consequently, if exclusion maximizes the sum of the economic benefits enjoyed by the vendor and the customer, then it will also typically maximize the customer’s benefits. This principle applies even when the seller allegedly possesses the power to “compel” the buyer. Unless exclusion increases the sum of the economic benefits received by the vendor and the customer, any bargaining power that is used to compel the customer’s participation in an exclusionary arrangement could be deployed more profitably to obtain greater financial consideration without that arrangement.13 Robert Bork (1978) correctly recognized this point when commenting on Standard Fashion: “Standard can extract in the price that it charges all that its line is worth. It cannot charge the retailer that full amount in money and then charge it again in exclusivity that the retailer does not wish to grant.” From the principles in the two preceding paragraphs, it follows that, in the absence of NCEs, exclusionary agreements will emerge only if they are both socially beneficial, in the sense that they maximize total economic benefits to all members of society, and beneficial to customers. That is why (as noted above), unless NCEs are present, one can reasonably infer that such agreements are procompetitive whenever they are used. An alternative way to state this principle is that, in the absence of NCEs, a market leader cannot profitably engage in anticompetitive exclusion. Assuming that the exclusion of a rival has no procompetitive effects, it generally reduces the total economic value shared by the market leader and the customer.14 In that case, although a market leader could induce the customer to accept an exclusionary arrangement and thereby impair a rival, the terms required to secure the consent of a customer who anticipates the full impact of the rival’s exclusion would leave the market leader with lower profits than it could achieve by negotiating a nonexclusive relationship. While there are exceptions to this principle,15 it applies in a wide range of circumstances.

13  This observation presupposes that the buyer and seller can write contracts that include volume-insensitive payments (e.g., a fixed fee that is imposed on top of per-unit charges). If one assumes instead that the buyer can only charge a fixed per-unit price, then Bork’s argument (as well as the more general one given in the text) breaks down, and there is greater scope for anticompetitive exclusionary conduct (Matthewson and Winter 1987). As a practical matter, however, we believe that this alternative assumption is typically (though not always) incorrect. See O’Brien and Shaffer (1992) for criticisms of that assumption. 14  For example, if the established firm and the rival offer differentiated products, then the weakening of the rival deprives the buyer of value that the established firm cannot fully replace. 15  For example, as noted by Whinston (1990), commitments to certain types of exclusionary practices can alter strategic incentives in ways that favor the established firm, even with only one (current and future) buyer. Also, in some instances, customers may not fully anticipate the effects of exclusive arrangements with the vendor on future competition.

10   B. Douglas Bernheim and Randal Heeb In contrast, when NCEs are present, the chain of reasoning that rules out the possibility of anticompetitive exclusion can break down. Specifically, an exclusionary arrangement can maximize the sum of the economic benefits enjoyed by the market leader and a given customer precisely because the resulting impairment of the rival allows the market leader to extract greater economic value from other customers.16 The seller can then secure each customer’s consent to an exclusive arrangement for consideration that is potentially of much smaller value than the harm to all customers collectively.17 Accordingly, the market leader may well find anticompetitive exclusionary conduct profitable, even though it is highly inefficient.18 A simple numerical version of our example helps to illustrate this point. Suppose there are ten customers. When the market leader and any one customer enter into an exclusionary agreement, each customer loses $10 due to reduced future competition (given that the rival is incrementally weakened), while the market leader gains $80. Because all customers are adversely affected, NCEs are present. We have chosen these numbers so that exclusion is inefficient: the total loss to all customers ($100) exceeds the gains to the market leader ($80). Now suppose the market leader offers one of the customers $11 to enter into an exclusive relationship. That customer will accept the deal because it comes out $1 ahead. After paying for exclusion, the market leader comes out $69 ahead. But every other customer comes out $10 behind, and customers as a whole come out $89 behind. From a collective perspective, the problem here is that each customer’s willingness to enter into an exclusionary relationship reflects the balance of competing considerations that determine its own economic costs and benefits, rather than the costs and benefits for all customers. By signing up all customers to exclusive relationships for $11 each, the market leader can achieve a total profit of $690, leaving each buyer worse off by $89, or $890 in total. Thus, due to the presence of contracting externalities, the market leader profits from anticompetitive exclusion. Our example shows that exclusionary agreements can be anticompetitive when they generate NCEs that are borne by the market leader’s other customers. What if those externalities are instead borne by other parties? The fact that another party suffers from an NCE does not by itself provide an incentive to engage in anticompetitive exclusion.19 For such 16 

Farrell (2005) provides an intuitive discussion of this mechanism. In fact, when negative contracting externalities are strong, the consideration may be extremely small (Segal and Whinston 2000). For a more general discussion of negative contracting externalities, see Segal (1999). 18  See Bernheim and Whinston (1998, section IV). This was essentially the government’s theory of harm in Regional Health (http://www.justice.gov/atr/cases/unitedregional.html). 19  For example, an exclusionary agreement between a vendor and a customer will typically impose an NCE on the vendor’s rival. Such an effect would not necessarily implicate the antitrust laws, which are designed to protect competition rather than competitors. Leaving legal issues aside, in standard economic models an NCE borne by a rival does not reflect more effective expropriation of economic benefits from the rival by the market leader, the customer, or both. On the contrary, it reflects a reduction in aggregate economic benefits, as a result of which the market leader and customer also jointly suffer (e.g., because the incremental benefits of the rival’s differentiated product are lost). Therefore, such NCEs cannot provide a motivation to engage in exclusion. In addition, the relationship between the rival and customer may effectively internalize any contracting externality experienced by the rival; see Bernheim and Whinston (1998). 17 

A Framework for the Economic Analysis of Exclusionary Conduct   11

an incentive to arise, the market leader and the customer must jointly benefit from that NCE; in other words, the NCE must exist because either the market leader, the customer, or both more effectively expropriate economic benefits from another party (just as in our example). For that to occur, the parties bearing the NCEs usually must lie downstream from the market leader—that is, they must either be the market leader’s customers, final consumers downstream from the market leader’s customers, or (in industries with long supply chains) companies situated between the two.20 When vendors engaging in exclusionary practices sell intermediate goods, it is indeed the case that parties downstream from the market leader’s customers, such as final consumers, often suffer from NCEs. For example, such NCEs will be present if exclusion allows the market leader to charge higher prices in the future by weakening the rival, and if customers pass some portion of the price increase downstream.21 As long as an exclusionary arrangement between the market leader and a customer imposes NCEs borne by (or passed through) the market leader’s other customers, NCEs borne by downstream parties as a consequence of pass-through can magnify the incentive to engage in exclusion for anticompetitive purposes.22 To illustrate this point, we will modify the last numerical example as follows. Suppose that when any one customer enters an exclusive deal with the market leader, all customers lose $5 in future benefits due to reduced competition (given that the rival is incrementally weakened), and downstream consumers lose $50, while the market leader again gains $80. The total loss to all customers and downstream consumers combined is $100, just as before, but now that loss is not absorbed entirely by the market leader’s customers. In this case, the market leader can lure a customer into an exclusive deal by offering only $6 rather than $11—the consenting customer ignores the benefits of an unimpaired rival not only to the market leader’s other customers but also to downstream consumers. After paying for exclusion, the market 20 

One can imagine exceptions. Suppose, for example, that the market leader and the rival purchase a critical input from an upstream supplier with market power. By weakening the rival, the market leader may be able to create offsetting monopsony power, and thereby obtain the input at lower cost. While the NCE suffered by the upstream supplier would provide an incentive for the market leader and the customer to enter into an exclusionary agreement, the agreement might well benefit consumers, and consequently might not qualify as an antitrust violation. 21  Pass-through is not a foregone conclusion. The market leader may have an incentive to exploit its market power by charging lump-sum fees, which the customer would tend to absorb. 22  See Abito and Wright (2008) and Simpson and Wickelgren (2007), which conclude that exclusion is more likely when buyers are downstream firms that compete with one another. A different conclusion (i.e., that exclusion is less likely when buyers are downstream competitors instead of final consumers) is reached by Fumagalli and Motta (2006), though their result depends crucially on some of their simplifying assumptions. For example, as discussed by Wright (2009), they assume that the downstream firms sell a homogenous good and have to pay a fixed cost to stay in business. As a consequence, a downstream firm that secures the input from the rival at a lower price can serve the entire downstream market without competitive constraint. The prospect of the resulting profit makes rejecting the dominant firm’s exclusive offer attractive; as a result, exclusion does not occur in equilibrium. However, with differentiated products and small fixed costs, the downstream firms that have entered into exclusive deals will remain in the market and continue to exert a competitive constraint on the deviating firm, making deviations less profitable.

12   B. Douglas Bernheim and Randal Heeb leader comes out $74 ahead. But every other customer comes out $5 behind, or $45 in total, while downstream consumers come out $50 behind. The market leader could induce all of them to sign the same exclusive deal, thereby profiting $740 in total, while leaving the customers all worse off to the tune of $44 individually, or $440 in total, and leaving downstream consumers worse off by $500. Because the NCEs borne by downstream consumers reduce the price of exclusion (from $11 in the previous numerical example to $6 in this one), anticompetitive exclusive dealing presents the market leader with an even greater profit opportunity in this example ($74 per customer, or $740 in total) than in the previous one ($69 per customer, or $690 in total). Downstream consumers may suffer from NCEs due to pass-through even in the absence of NCEs borne by (or passed through) other customers. However, in that case, firms do not generally have incentives to engage in exclusion to achieve purely anticompetitive ends. Consider the simplest case, in which the market leader and rival sell to a single customer, who in turn supplies a product to final consumers. Because exclusion of the rival does not increase the customer’s market power over the final consumers, there is no reason to think that it allows the customer to expropriate greater economic benefits from them. Indeed, the effect can be precisely the opposite: if, by weakening the rival, exclusion subsequently allows the market leader to charge the customer a higher price, then expropriation of economic benefits from final consumers may become less efficient (e.g., due to double marginalization).23 In that case, the NCE suffered by final consumers will be associated with a reduction in the joint benefits enjoyed by the market leader and the customer, rather than an increase, which attenuates the incentive to exclude. Thus, an exclusionary agreement between a market leader and a customer does not usually raise anticompetitive concerns unless it imposes NCEs borne by (or passed through) the market leader’s other customers, and unless those NCEs reflect greater expropriation of economic benefits by the market leader.24 It is worth emphasizing that, in the preceding examples (and more generally), exclusionary conditions are anticompetitive even though each customer voluntarily agrees to exclude the rival. Considered in isolation, each deal between the market leader and a customer is necessarily mutually beneficial, even accounting for its subsequent impact on competition. However, every such deal harms other parties (other customers and possibly their downstream consumers) who are not part of the deal. 23 

Spengler (1950) is credited with first articulating the inefficiencies of double marginalization. Notably, when NCEs are borne by (or passed through) the market leader’s other customers, exclusion can lead to greater expropriation of economic benefits from final consumers. In that case, the rival’s exclusion can mitigate the extent to which downstream competition limits the extraction of economic benefits from consumers: with a weakened rival, the market leader can raise the input price, thereby causing customers to increase the prices of their downstream offerings. Because competition would otherwise keep those prices below profit-maximizing levels, the sum of the economic benefits received by the market leader and the customer who enter into the exclusionary arrangement can rise, thereby making exclusion attractive for anticompetitive reasons. 24 

A Framework for the Economic Analysis of Exclusionary Conduct   13

Consequently, every customer would be better off if no deals were consummated; still, each customer has a strong individual incentive to consummate its own deal. It follows that coercion is not necessary to achieve exclusion. Indeed, recognizing the potential for mutual benefit, individual customers may actively seek exclusive arrangements with the market leader, but that possibility renders the arrangements no less anticompetitive. Certain features of our examples play critical roles in generating anticompetitive effects, while others do not. Distinguishing between those two sets of features is important, because a well-designed antitrust inquiry should focus on the first set and not the second. In tracing anticompetitive exclusion to particular types of NCEs, we have emphasized the importance of entry barriers (and thus the market leader’s market power), the rival’s vulnerability, and the presence of multiple customers. Many other features of our examples are either inessential or less critical. For instance, similar conclusions follow regardless of whether there is one rival or many and irrespective of whether the rival is active or merely a potential entrant. In the remainder of this section, we elaborate on the roles played by several other features of our examples. While our examples assume that exclusion impairs the rival’s future competitive efficacy, the same anticompetitive mechanism can operate when exclusion only limits the rival’s current ability to compete, for example through the lost economies of scale (see Rasmusen, Ramseyer, and Wiley 1991 and Segal and Whinston 2000). In the latter case, however, the mechanism may be more fragile. For example, it may operate less reliably when contract negotiations with customers are synchronized than when they are staggered (see Segal and Whinston 2000). Alternatively, if the source of the scale economies is a one-time entry cost, and if contract duration is short, then a new rival may well find it profitable to incur that cost even if the market leader has locked up a large fraction of the market in short-term exclusive arrangements; once the entry cost is sunk, the rival can operate efficiently at small scale, so that the exclusionary mechanism is subsequently disabled. There is a sense in which the mechanism highlighted in this section always involves both profit sacrifice and recoupment. Any provisions in an agreement between the market leader and a customer that restrict the customer’s freedom to do business with other vendors leaves the customer worse off, ceteris paribus. The customer will not enter into such an agreement voluntarily unless the market leader compensates it for that loss, thereby sacrificing profits. Similarly, the market leader will find the agreement unattractive unless it expects to recoup the compensation. It achieves recoupment by exercising greater market power over other parties. Thus, recoupment is the source of the NCEs at the heart of the mechanism. Courts have long recognized that profit sacrifice and recoupment are essential features of predatory pricing, and require proof of both in that context (see Concord Boat; Brooke Group). Some subsequent commentators have advocated the application of this principle to exclusionary practices more generally, and have in particular proposed that courts use evidence of profit sacrifice as a criterion for distinguishing between

14   B. Douglas Bernheim and Randal Heeb anticompetitive and procompetitive exclusionary conduct (see Melamed 2005; Werden 2006). Others have argued that exclusionary conduct does not necessarily require profit sacrifice, that the sacrifice may not be readily observable, and that it may not be closely associated with the underlying anticompetitive harm (see Nalebuff 2005a and 2005b; Salop 2006). Though forms of sacrifice and recoupment are always elements of the anticompetitive mechanism highlighted above, a broad requirement that plaintiffs provide direct evidence of sacrifice and/or recoupment is inadvisable. Predatory pricing is a rather special case, in that there is a clear temporal separation between the periods during which the market leader sacrifices profit on the one hand and receives recoupment on the other. For many other forms of exclusionary conduct, there is no such separation. For example, when exclusionary contracts with multiple customers impair the rival’s current competitive efficacy, each contract implicitly involves both profit sacrifice to secure that customer’s assent, and (partial) recoupment of the profits sacrificed to secure the assent of other customers. One cannot, however, measure those components individually; the terms of the contract reflect only their combined effects. Even when the sacrifice and recoupment associated with an exclusionary agreement are temporally separated (as in our examples), exclusionary conduct may be ongoing, in which case the current profit sacrifice may be obscured by recoupment associated with past conduct. Of course, for the market leader to engage willingly in anticompetitive exclusion, the anticipated recoupment must exceed the profit sacrifice, in which case the firm earns supracompetitive profits. To test that implication of the theory, one would not need to distinguish between sacrifice and recoupment. However, the use of such a test is also often inadvisable, given the difficulty of measuring economic profits, let alone ex ante profit expectations. Measurement problems aside, the firm may have other sources of supracompetitive profits or, alternatively, economic losses that offset the gains from anticompetitive exclusion. Thus, it is usually better to investigate whether economic conditions favor the operation of the anticompetitive mechanism. For example, supracompetitive profit (i.e., net recoupment) is possible only if the rival is weakened. This principle has been articulated most clearly in the context of predatory pricing: the predator cannot profitably exclude rivals by selling output at prices below cost unless, as a result, rivals collectively pose a reduced competitive threat in the future.25 Of course, focusing only on the weakening of a rival would not distinguish the effects of anticompetitive conduct from the natural demise of a less efficient competitor. Thus, this element is necessary but not sufficient for the conduct in question to be anticompetitive via the highlighted mechanism.

25  See Elzinga and Miles (1994). Matsushita v. Zenith Radio is the guiding case on the requirement for recoupment.

A Framework for the Economic Analysis of Exclusionary Conduct   15

1.2.2.  Modes of Conduct That Potentially Implicate the Mechanism So far, we have been intentionally vague concerning the nature of the conduct through which a company excludes a rival from all or part of the market. Modes of exclusionary conduct fall into two main categories: practices that rely on the use of exclusionary conditions (defined in the next section), and those that do not. The latter category includes various pricing strategies, including predatory pricing, bundling, volume discounts, and conventional tying. There are, of course, important substantive differences between the modes of conduct that fall within these categories, and the case law treats them differently.26 However, despite their differences, they have critical commonalities, chief among which is their ability to activate the anticompetitive mechanism described in the previous section. Accordingly, an analytic framework designed to detect the operation of that mechanism can be applied in a unified way to all modes of conduct within the two categories.

1.2.3.  Practices That Involve Exclusionary Conditions We use the phrase exclusionary conditions to denote practices that render aspects of transactions between a seller and one of its customers effectively contingent upon the customer’s dealings with the seller’s rival(s).27 Often, an exclusionary condition takes the form of a restriction, which the customer accepts in exchange for a fixed payment or some other additional consideration (such as a discounted price for goods or services). Alternatively, a company may reward or penalize a customer based on the volume that it purchases from a rival. The simplest and most readily recognizable type of exclusionary condition is a “100% exclusive deal,” in which the seller either pays the customer an agreed sum of money or offers a commensurate price reduction, to refrain from doing business with the rival. Alternatively, instead of 100% exclusivity, the seller may place some other explicit or implied limit on the customer’s dealings with the rival. For example, the seller may require the customer to restrict sales of its rival’s products to no more than a specified (usually small) percentage of total sales.28 Alternatively, it might require the customer to

26  For predatory pricing, see Concord Boat and Brooke Group; for tying, see Jefferson Parish, Microsoft, and Eastman Kodak Co.; for exclusive dealing, see Microsoft and Dentsply. 27  See Fiona Scott Morton (2012) for a characterization of such conditions. 28  Applicable case law includes Tampa Electric (exclusive dealing arrangements violate antitrust law if it is probable that they “will foreclose competition in a substantial share of the line of commerce affected”); LePage’s (exclusive dealing contracts with large customers); Microsoft (arrangements that closed to rivals “a substantial percentage” of distribution opportunities); SmithKline (conditioning rebates on market-share thresholds).

16   B. Douglas Bernheim and Randal Heeb limit sales of rivals’ products to secondary channels, forbid active marketing of rivals’ products, or insist upon inferior placement of rivals’ products in advertisements, store displays, or on customers’ websites.29 Provided that the scope of the resulting exclusion is sufficient to meaningfully handicap the rival and thereby impair its ability to serve other customers, such arrangements can activate the anticompetitive mechanism described in the preceding section. Significantly, exclusionary conditions involve more than merely winning all of a customer’s business through aggressive competition on the merits. Instead of focusing on the business that the customer transacts with the seller, these conditions place limits on the business that the customer can transact with the rival. This distinction is substantive. For instance, a requirement of exclusivity precludes the rival from selling additional units to the same customer beyond those the seller supplies, whereas merely winning a bid to sell the customer some number of units does not. The fundamental character of arrangements involving exclusionary conditions does not depend on the form of consideration received by the customer in exchange for accepting the exclusionary condition. In the simplest case, the quid pro quo consists of a monetary payment. Alternatively, it could take the form of a large rebate or additional marketing funds (see, e.g., LePage’s; Microsoft; SmithKline). It may also involve an in-kind payment, such as technological information, engineering support, preferential supply consideration, or other advantages.30 Any good or service received as an in-kind payment has an equivalent monetary value—either the price for which the good or service sells in the open market or the price at which the seller would provide it to the buyer separate from the broader agreement. Transferring the good or service is economically equivalent to transferring its monetary value, although it may be very difficult to determine precisely what that value is. When the seller’s customers compete with each other downstream, the compensation that a customer receives in exchange for exclusivity may entail a promise to supply the customer at terms that are more favorable than those given to other customers. Such a promise effectively amounts to a payment for exclusivity: the seller forgoes revenue by supplying the customer at preferential terms; the customer receives an opportunity to earn greater profits in competition with other downstream firms. Significantly, the quid pro quo for accepting an exclusionary condition may also take the form of a threat unexecuted. For example, the seller can “reward” the customer’s cooperation by refraining

29 

Applicable case law includes Grinnell; LePage’s; Conwood; General Industries. See, e.g., Microsoft; in its complaint against Intel, the Federal Trade Commission charged that “[o]‌n one hand, Intel threatened to and did increase prices, terminate product and technology collaborations, shut off supply, and reduce marketing support to OEMs that purchased too many products from Intel’s competitors. On the other hand, some OEMs that purchased 100% or nearly 100% of their requirements from Intel were favored with guarantees of supply during shortages, indemnification from intellectual property litigation, or extra monies to be used in bidding situations against OEMs offering a non-Intel product” (United States of America before the Federal Trade Commission, Complaint in the Matter of Intel (2009)). 30 

A Framework for the Economic Analysis of Exclusionary Conduct   17

from withholding products that are essential to the customer’s business and/or from supplying the customer’s competitors at favorable terms.31 An exclusionary condition need not be memorialized in a formal written contract. A shared understanding of the relationship between the extent of the customer’s dealings with the rival and the consideration received from the seller will suffice.32 Indeed, ambiguity in the terms of the relationship can be useful to the supplier, as it permits the supplier to interpret the exclusionary requirement broadly and flexibly in light of new developments.

1.2.4.  Practices That Do Not Involve Exclusionary Conditions Anticompetitive exclusion does not require the use of exclusionary conditions. Instead, a company that seeks to limit a customer’s purchases from a rival can simply “buy” the customer’s business by offering sufficiently attractive terms for its own products. The most straightforward approach, predatory pricing, is to charge a price so low that the rival is unable or at least unwilling to compete. Of course, that practice closely resembles the type of competitive conduct that the antitrust laws are designed to protect. Conceptually, the considerations that distinguish predatory pricing from ordinary competitive pricing are motives and their associated effects. When a company sets its prices to maximize profits without regard to the effect that sales will have on the future viability of rivals, it is acting competitively, even if the rival suffers as a result. However, when a company sets a lower price than it would otherwise have charged because it recognizes that limiting a rival’s sales will weaken the rival as a future competitor, it is acting anticompetitively.33 Thus, anticompetitive pricing can be defined as prices below those that would prevail if the firm ignored any resulting impairment of rivals. Significantly, costs do not enter into that definition. However, because motives and counterfactual profit maximizing prices are difficult to establish in practice, predatory pricing is usually

31 

See, e.g., LePage’s (“[T]‌he evidence in this case shows that Scotch brand tape is indispensable to any retailer in the transparent tape market”). If the conduct succeeds in inducing the buyer’s cooperation, then such a threat need never be carried out. Whether or not such threats can sustain an exclusionary equilibrium in a theoretical model or compel cooperation in the real world depends upon parties’ beliefs. See for example Nalebuff (2005b). Thus, evidence illuminating both the beliefs of the parties and observable efforts to influence those beliefs may be informative about whether or not (anticompetitive) exclusion is possible. 32 The Dentsply court noted that although the arrangements between monopolist and dealer were “technically only a series of independent sales” rather than “agreements,” “the economic elements involved . . . realistically make the arrangements here as effective as those in written contracts.” See also Tampa Electric (“[E]‌ven though a contract does not contain specific agreements not to use the [goods] of a competitor, if the practical effect . . . is to prevent such use, it comes within the condition of the section as to exclusivity” (internal quotation omitted)). 33  Ordover and Willig (1981) have proposed a definition of predatory pricing that is based on this distinction.

18   B. Douglas Bernheim and Randal Heeb defined as pricing below some cost-based threshold. We will return to its practical definition in section 1.3.1. Companies can accomplish the same end through more complicated pricing arrangements. One strategy, bundled pricing (or simply “bundling”), is to charge lower prices when goods are purchased in specified combinations than when they are bought separately. Antitrust concerns related to bundling tend to arise most often in practice when a seller offers a discount on a good over which it has monopoly power,34 conditional upon the customer purchasing a sufficient volume of a product for which rivals provide close substitutes. As in the case of predatory pricing, the seller limits a customer’s purchases from rivals by “buying” the customer’s contested business; in this case, it compensates the buyer through the contingent discount on the monopolized good rather than through an unconditional discount on the contested good.35 For the purpose of the mechanism described in section 1.2.1, it is not actually necessary for the seller to have significant market power over the good that bears the contingent discount,36 inasmuch as this discount is simply a vehicle for packaging a payment to the buyer. From the perspective of a company seeking to exclude a rival, attaching the discount to a monopolized good is cosmetically advantageous; in that case, the discounted price may be well above cost and, hence, less likely to raise suspicion of anticompetitive predatory pricing. As we explain in section 1.3.1, any given contingent discount is more likely to be anticompetitive if the seller monopolizes the good that bears the discount, than if that good is provided competitively. A seller can also use single-product pricing schemes (as opposed to the multiple-product pricing schemes employed in bundling arrangements) to reward buyers with lower prices and/or rebates when the volume purchased exceeds specified thresholds. As with predatory pricing and bundling, the seller limits a customer’s purchases from rivals by “buying” the customer’s contested business, but in this case, it compensates the buyer through a conditional discount on the contested good (e.g., with a volume discount or “loyalty” discount), rather than through either an unconditional discount or a conditional discount on some other good.37

34  Throughout, we use the phrase “monopoly power” to denote a high degree of market power, rather than that held by a textbook monopolist. 35  When the seller links the discount on the monopolized good to the buyer’s purchases of the contested good from a rival, the conduct also involves exclusionary conditions, because it renders aspects of transactions between a company and one of its customers or suppliers effectively contingent upon that party’s dealings with the company’s rival(s). 36  In contrast, for other mechanisms through which bundled pricing may have anticompetitive effects, a high degree of market power over the product that bears the contingent discount is essential (Whinston 1990). 37  Brooke Group addressed discriminatory volume discounts and market share “loyalty” discounts as predatory pricing and examined them under a price-cost test; see Kobayashi (2005) for a discussion; see also Greenlee and Reitman (2006); ZF Meritor addressed contracts with high share requirements and other conditions restricting dealing with rivals without requiring a price-cost test.

A Framework for the Economic Analysis of Exclusionary Conduct   19

Sellers can also exclude rivals through tying. Tying occurs when a company links a monopolized product to another potentially competitive product and either sells the products only as a package (conventional tying, or a “tie-in”) or requires that purchasers of the monopolized product refrain from purchasing the competitive products of rival firms (negative tying, or a “tie-out”). Negative tying involves an exclusionary condition, but conventional tying does not. Nevertheless, a seller can still exclude rivals through a conventional tie. Indeed, one can think of this practice as an extreme form of bundling, where the seller sets the prices of the products at prohibitive levels if they are purchased separately (so that the contingent discounts are extremely large). Even so, courts have treated the two practices differently. Any tie between a monopolized product and a competitive product is a per se antitrust violation, while a rule of reason is applied in cases that involve bundling.38 Although a rule-of-reason analysis is better suited conceptually for correctly sorting out the exclusionary implications of tying than a per se standard, there is indeed a good reason to treat both forms of tying more skeptically than practices involving less rigid links between goods, such as bundled pricing; see section 1.3.1. In addition, tying may implicate some special anticompetitive mechanisms that bundled pricing is less likely to activate (see, e.g., Carlton and Waldman 2002).

1.3.  Evaluating Exclusionary Conduct Having elaborated on the primary economic mechanism through which exclusionary conduct may achieve anticompetitive ends, we turn to the question of how one properly evaluates suspect conduct. It is helpful to divide the inquiry into three stages, each of which addresses a distinct question. First, we ask whether the conduct is, in fact, exclusionary. If we find that it is, we then move to the second stage and ask whether it has anticompetitive effects. To be clear, a finding that conduct is exclusionary does not by itself answer the second question; it merely establishes that further inquiry is warranted. If we find that anticompetitive effects are indeed present, we then move to the third stage, asking whether the conduct also has procompetitive effects and, if so, whether the company could achieve those benefits through less restrictive practices. If such effects are also present, we then weigh them against the harms to competition. 38  Jefferson Parish establishes that tying is a per se violation if the firm has sufficient market power to compel customers of the tying good to purchase a “second, unwanted product.” Whether this precondition, which would make the tie a per se violation, is met in practice would appear to require a rule-of-reason analysis; the court further limited the assumption of market power based on patent rights in Independent Ink, observing that “many tying arrangements, even those involving patents and requirements ties, are fully consistent with a free and competitive market”; some commentators predict that the Supreme Court is on the verge of revoking the per se status of tying arrangement entirely. See, e.g., Werden (2009).

20   B. Douglas Bernheim and Randal Heeb

1.3.1.  Determining Whether Conduct Is Exclusionary Antitrust policy seeks to limit anticompetitive conduct without chilling legitimate competitive activity. Where exclusion is concerned, the similarities between procompetitive and anticompetitive conduct render that objective especially challenging. In some sense, all competitive conduct involves exclusion:  the losing party is excluded from making the sales won by the prevailing party. Because exclusion of that form can motivate firms to better serve consumers’ interests by lowering prices and improving quality, it is essential to define the scope of suspect conduct much more narrowly. The type of exclusion that occurs with plain-vanilla competition has two features that render it innocuous. First, from the perspective of the prevailing firm, depriving the rival of sales is an incidental consequence of winning business; it does not provide additional motivation for prevailing over the rival beyond the benefits of making the sales. Second, the rival is free to make additional sales to the customer over and above those made by the prevailing firm. It is therefore appropriate to treat conduct as exclusionary (in the suspect sense) only when at least one of these features is absent; that is, either when there are good reasons to conclude that the excluding firm views the reduction in the rival’s sales as an added benefit over and above the increase in its own sales or when the conduct precludes or impedes the rival from making incremental sales to the customer. Based on these criteria, conduct involving exclusionary conditions is easily distinguished from plain-vanilla competition. If the objective is simply to win business, then, conditional on doing so, there is no reason in addition to require exclusion of the rival or otherwise seek to limit the rival’s dealings with the customer. The inclusion of such provisions, which are secured only at a cost to the excluding firm, necessarily implies that it views the reduction in the rival’s sales as an extra benefit. Exclusionary conditions also go beyond the units at stake in any given negotiation—they prevent the buyer from obtaining additional units from the rival, even when the excluding firm would not itself meet that incremental demand (e.g., because either it has limited inventories or the rival offers a differentiated product). To be clear, exclusionary conditions are not necessarily anticompetitive. In some settings, they may help firms resolve incentive problems and thereby establish more productive business relationships, in which case their effects are at least partially procompetitive. However, they do not routinely arise in the course of competition on the merits. The potentially chilling effects of subjecting exclusionary conditions to scrutiny under the antitrust laws is therefore narrowly circumscribed and can be minimized provided that the courts carefully examine the conduct’s causes and effects, recognizing the potential procompetitive uses of such conditions in particular circumstances. Certainly, there is no risk of chilling straightforward price and quality competition, which are the main engines of the competitive marketplace. Accordingly, evidence of exclusionary conditions adequately answers the threshold question—whether the conduct is exclusionary—and calls for a thorough investigation of competitive impact.

A Framework for the Economic Analysis of Exclusionary Conduct   21

Unfortunately, if the challenged conduct does not involve exclusionary conditions, then matters are not so clear-cut. The distinction between exclusionary and nonexclusionary conduct is necessarily more subtle when the instrument of exclusion is the price of the good in question. Conceptually, the most challenging case involves predatory pricing. Some economists have proposed classifying a price as anticompetitive if it falls below the level that would prevail if the seller maximized short-term profits, ignoring effects on a rival’s future competitiveness or on the number of rivals (Ordover and Willig 1981). While such definitions are conceptually appealing from the perspective of economic principles, they would potentially expose all forms of price cutting to routine challenge and would preclude procompetitive, dynamic rationales for such low prices, such as accelerated learning or penetration pricing. Such challenges could severely chill legitimate price competition to the detriment of consumers, contrary to the intent of the antitrust laws. To reduce the risk of broadly chilling price competition, exclusionary pricing can be defined more narrowly, ideally with reference to characteristics that, like exclusionary conditions, are more distinctively anticompetitive. One widely discussed possibility is to focus on prices that are below costs (Areeda and Hovenkamp 2008). The law and most commentators treat such prices as presumptively exclusionary (because they prevent equally efficient rivals from winning business profitably) (see Brooke Group and Advo), and they arise in the course of legitimate competition only in relatively limited and identifiable circumstances.39 Indeed, economic principles teach us that competition is desirable not because it drives prices downward, but rather because it pushes them toward costs (Bernheim and Whinston 2008). From the perspective of achieving a desirable allocation of society’s resources, prices that are too low can be just as problematic as prices that are too high, even if they benefit consumers in the short term. To be clear, evidence of below-cost pricing does not by itself establish that a seller’s conduct is anticompetitive, but it is another cause for potential concern and, like exclusionary conditions, calls for a thorough investigation of competitive impact. Consistent with this reasoning, the Supreme Court has ruled that below-cost pricing is a necessary (but not sufficient) condition for a finding of predation (see Brooke Group). It is critical to emphasize, however, that as a matter of economic theory, prices above costs can be exclusionary and anticompetitive (Ordover and Willig 1981). To illustrate, suppose a monopolist in a differentiated product market faces competition from an emerging rival but realizes that the rival will be weakened in subsequent rounds of competition if it is denied opportunities to benefit from learning-by-doing, to raise capital at attractive terms based on evidence of market success, and to build valuable relationships with customers. Taking these effects into account, the monopolist would set a lower price than it would have otherwise chosen. Its conduct is exclusionary (because its objective is not merely to win sales but also to deprive the rival of sales) and anticompetitive 39  E.g., prices below costs may be used procompetitively by a firm to promote product introductions, to gain positions of leadership in emerging industries, to benefit from learning-by-doing, or to generate positive externalities that increase the value of a firm’s offerings in a two-sided market.

22   B. Douglas Bernheim and Randal Heeb (because it reduces subsequent competition). And yet, there is no particular reason to think that the price it sets will be below cost. Thus, the Supreme Court’s requirement of a price-cost test in the context of predation is appropriate not because economic theory tells us that below-cost pricing is necessary for exclusion (which it is not), but rather because extreme caution is warranted to avoid chilling legitimate competition when price is the only instrument of exclusion. In principle, one could apply price-cost tests to exclusive deals (and other exclusionary conditions), by asking whether prices are below costs when adjusted to reflect an (appropriately attributed) share of the payment received in exchange for exclusivity.40 However, the courts have pointedly not required such a test for exclusive dealing or exclusionary conditions (see, e.g., Dentsply; Microsoft). The absence of such a requirement is appropriate. The extreme level of caution exercised in the context of predatory pricing is excessive in the context of conduct involving exclusionary conditions because, unlike low prices, such conduct is (1) clearly differentiated from plain-vanilla price competition, and (2) a relatively uncommon consequence of routine competition, so that the adverse effects of chilling competition by subjecting those conditions to scrutiny is relatively small. In addition, when the monopolist supplies an intermediate good, exclusive dealing or exclusionary conditions (again, unlike a low price) can harm downstream customers immediately by blocking current sales of the rival’s products, including those that are significantly differentiated from the monopolist’s offerings. For bundled pricing, an analysis of exclusion is neither as clear-cut as for exclusionary conditions, nor as challenging as for predatory pricing. A systematic pattern of linking discounts on monopolized products to competitive offerings in response to a rival’s marketing efforts is certainly more suspicious than simply charging a low price. As with exclusionary conditions, if the object is simply to win the business, then the monopolist could incorporate the discount into the price of the good facing competition; there is no obvious need to provide it in a more convoluted way through bundled pricing. Moreover, because bundled pricing is used far less often than straightforward price competition, subjecting it to scrutiny under the antitrust laws does not carry nearly the same risk of chilling important forms of competition as scrutinizing low prices, particularly inasmuch as some of its other uses (e.g., facilitating price discrimination) may also be contrary to consumers’ interests (Nalebuff 2005a). Even so, unlike exclusionary conditions, bundled pricing can arise when exclusion is not the objective; hence, the practice is not intrinsically exclusionary. Consequently, some additional consideration may be required to determine whether or not bundled pricing is exclusionary within the context of a particular factual setting. Courts considering particular cases involving different factual circumstances have reached different conclusions as to the showing required to establish exclusion for 40 

Other commentators in this volume and elsewhere have discussed appropriate attribution of discounts in the context of bundled pricing. Mechanically, the same principles could be employed to compute attribution in the context of a payment for exclusivity. However, as we explain in the text, that calculation does not provide an appropriate basis for evaluating exclusionary conditions.

A Framework for the Economic Analysis of Exclusionary Conduct   23

bundled pricing. Most notably, within a context where bundled pricing was intermingled with a broader pattern of exclusionary conduct by a monopolist (including exclusive dealing and other exclusionary conditions), the Third Circuit in LePage’s determined such pricing to be exclusionary, without need for a showing that effective prices were below cost.41 In Ortho and PeaceHealth, two other courts examined bundled pricing in contexts in which it constituted the principal basis for alleged exclusion. Both courts recognized the potential for exclusion but endorsed the use of a price-cost test to determine whether bundled pricing actually had that effect, given the facts of those cases. The apparently disparate precedents mentioned in the preceding paragraph are potentially reconcilable within our framework. It is far less likely that the objective of bundled pricing is nonexclusionary when a monopolist’s intent to exclude is apparent from other intrinsically exclusive practices, as in LePage’s, than when it is not, as in Ortho and PeaceHealth. Recalling that the argument for protecting bundled pricing is less compelling than that for protecting plain-vanilla price-cutting, one could reasonably conclude that the price-cost test is excessively demanding in the circumstances of LePage’s, but not necessarily in those of Ortho and PeaceHealth. In section 1.2.4, we noted that bundled pricing can in principle activate an anticompetitive mechanism even when the good that bears the discount is not monopolized. However, whether or not that good is monopolized affects the proper implementation of a price-cost test. The reason is that a properly performed price-cost test attributes contingent discounts to contested sales.42 Consider the following example: a company produces good A at a cost of $5 per unit and good B at a cost of $10 per unit, selling good A at a price of $10 per unit and good B at a price of $15 per unit. A customer purchases 1,000 units of good A and none of good B. Seeking to boost sales of good B, the company offers a discount of $1 per unit on all units of good A if the customer purchases at least 100 units of good B. Under this bundled pricing scheme, the customer buys the first 100 units of good B at a net cost of $500,43 or $5 per unit, which is below cost. If good A is monopolized, then the bundled pricing scheme fails the price-cost test: an equally efficient rival for good B cannot win business without losing money. In contrast, if good A is also contested, then the same scheme passes an appropriately formulated price-cost 41 See LePage’s (“3M raises various objections to the trial court’s decision but essentially its position is a legal one: it contends that a plaintiff cannot succeed in a § 2 monopolization case unless it shows that the conceded monopolist sold its product below cost. Because we conclude that exclusionary conduct, such as the exclusive dealing and bundled rebates proven here, can sustain a verdict under § 2 against a monopolist and because we find no other reversible error, we will affirm”). 42 See United Regional (“To accurately determine whether United Regional’s discounted prices are above cost, however, the entire discount should be attributed not to the entire volume of the ‘competitive product[s]‌,’ as suggested by the court in PeaceHealth, id. at 909, but rather to the patients that United Regional would actually be at risk of losing if an insurer were to choose non-exclusivity (the ‘contestable volume’)”), http://www.justice.gov/atr/cases/f267600/267653.pdf. 43  If the customer can buy 1,000 units of good A for $10,000 or 1,000 units of good A and 100 units of good B for $10,500 (i.e., $9,000 for 1,000 units of good A and $1,500 for 100 units of good B), the net difference is $500.

24   B. Douglas Bernheim and Randal Heeb test: an equally efficient rival can match and even beat the bundled price for 1,000 units of good A and 100 units of good B ($10,500) while still earning a profit (because the total cost of the bundle is $6,000). Indeed, if both goods are contested, then a bundled pricing scheme cannot fail an appropriate price-cost test unless the discounted price of at least one of the goods is below its cost, which typically is not the case. When the good that bears the discount is not only monopolized but also generates large amounts of revenue, the attributed discount for the contested good can be very large, even when the nominal discount on the monopolized good is small (as in our example). Hence, under those conditions, there is legitimate reason for heightened concern that bundled pricing schemes are exclusionary. Single-product volume discounts are widespread, likely for legitimate procompetitive reasons. Subjecting such discounts to scrutiny under the antitrust laws therefore runs a risk of chilling an important and largely beneficial form of competition. For that reason, the standard for evaluating exclusion should reflect a high level of caution, just as with aggressive price-cutting. Once again, a price-cost test emerges as a reasonable and practical compromise.44 Some legal scholars have expressed the view that pricing schemes involving either a single product or multiple products belonging to the same market do not raise any antitrust concerns apart from the possibility of predatory pricing (see, e.g., Areeda and Hovenkamp 2008). Their argument asserts that if only one product is involved, a rival can then defeat any putative exclusionary effects of the pricing scheme by competing to sell all of the volume (in other words, by offering a package to meet all of the buyer’s needs). That conclusion is warranted in some circumstances, but not in others, because it depends on critical and unstated assumptions: that the buyer is willing to purchase all units of the single product from the rival, and that the rival is able to provide all of the units that the buyer requires. These assumptions may be untenable. The first assumption is violated, for example, when the buyer uses the single product in two or more distinct applications and the perceived suitability of the rival’s offering is low for at least one important application.45 The second assumption is violated, for example, when the rival is capacity constrained. In such cases, because the rival cannot offer a competing package to meet all of the buyer’s needs, single-product bundled pricing schemes, including certain types of volume discounts, can raise the same concerns about exclusion as bundled pricing involving multiple products. For volume discounts, exactly as for bundled pricing, all contingent discounts are appropriately attributed to the contested units. To illustrate, let’s suppose that a company sells its product for $10 per unit, and that a customer buys 800 units. Seeking 44 

See notes 10, 21, 27 above. For example, with respect to allegations that Intel excluded AMD from portions of the market for microprocessors, the European Commission found that computer purchasers fell into two classes: those who insisted upon purchasing Intel-based computers, and those who were open to purchasing AMD-based alternatives; moreover, in order to remain viable, computer makers needed to carry Intel-based computers for the first group. Decision of the European Commission (Non-confidential version), (COMP/C-3 /37.990—Intel) at 870–71. 45 

A Framework for the Economic Analysis of Exclusionary Conduct   25

to boost sales, the company offers a discount of $2 per unit on the first 1,000 units if the consumer purchases at least 1,000 units; the customer responds by purchasing 1,000 units (paying $8,000 in total). If all units sold to the customer are contestable, an equally efficient rival could then conceivably secure the customer’s business by charging a total of $8,000 for 1,000 units, for an average of $8 per unit. Applying the same standard as for predatory pricing, the volume discount fails the price-cost test if the unit cost of production is greater than $8 and passes it if the unit cost is less than $8. Now let’s assume that the first 600 units of the product sold to the customer are not actually contestable. (Possibly the customer has downstream clients for whom the brand of the input is essentially nonnegotiable.) In that case, the rival can only contest the last 400 units. Under the volume discount scheme, the customer pays $6,000 for the first 600 units and $8,000 for 1,000; therefore, incrementally, the customer pays $2,000, or $5 per unit, when buying the last 400 units from the volume discounter. If the cost of production is, say, $7 per unit, then this pricing arrangement fails the appropriate price-cost test (because it precludes an equally efficient rival from winning the 400 contestable sales without losing money), even though it would pass the corresponding test when evaluated as conventional predatory pricing. One can think of conventional tying as an extreme form of bundling, where the seller sets the prices of the products at prohibitive levels if they are purchased separately (so that the contingent discounts are extremely large). Consequently, to determine whether it is exclusionary, we apply the same test as for bundling. To illustrate, suppose a seller ties two products, good A and good B, so that customers must buy an equal number of each. Our objective is to determine whether the tie excludes a rival producer of good B. For the customer, the incremental price of obtaining good B from the tying firm, conditional on purchasing good A, is zero, which is always below cost. Therefore, if the tying firm monopolizes good A, then the tie automatically fails the pertinent price-cost test. It follows that a tie between a monopolized good and a contested good is intrinsically exclusionary. To determine whether the exclusion is anticompetitive, one would proceed directly to the second stage of the analysis; however, because other uses of tying are not necessarily in consumers’ interests, a per se standard may well be justified. Thus, even though conventional tie-ins between contestable and monopolized goods need not involve exclusionary conditions, our framework effectively treats them as if they do: because such tying fails the pertinent price-cost test automatically, one moves immediately to the second stage of the analysis, just as if exclusionary conditions were present. In practice, that same treatment is also sometimes warranted because the tie is associated with implicit exclusionary conditions that make it costly or infeasible for the customer to deal with a rival. In contrast, if the tying firm does not have market power over good A, then a rival could contest the entire bundle. In that case, we simply treat the bundle as a single product: the inquiry into anticompetitive exclusion ends unless the bundle’s cost exceeds its price. The fact that there is a tie raises no additional issues.

26   B. Douglas Bernheim and Randal Heeb

1.3.2.  Determining If Exclusionary Conduct Has Anticompetitive Effects Exclusionary conduct is not necessarily anticompetitive; indeed, it sometimes has procompetitive effects. Consequently, even when conduct has been deemed exclusionary, further investigation is required to determine whether it is in fact problematic from an antitrust perspective. To determine whether exclusionary conduct generates anticompetitive effects through the mechanism discussed in section 1.2.1, one can apply a test consisting of the following four elements: lement 1:  Diminished ability to compete E Element 2:  Enhanced market power Element 3:  Harm to consumers Element 4:  Negative contracting externalities We will elaborate on each of these elements in turn.

1.3.2.1.  Element 1 (Diminished Ability to Compete) Exclusion from the portion of the market targeted by the excluding firm’s conduct must significantly impair the rival’s ability and/or incentive to compete effectively for business other than that which the excluding firm captures directly. In other words, the impact on the rival of the exclusionary conduct must extend beyond the loss of the sales covered by the agreement. It is not enough that the conduct simply deprives the rival of the sales that are captured by virtue of the conduct; after all, the customer forgoes the opportunity to purchase those units from the rival voluntarily and cannot be worse off as a consequence.46 However, if the rival’s ability to compete for other sales—for example, sales at later points in time or to other customers—is substantially impaired by conduct that precludes the rival from doing business with a customer, then that conduct is potentially anticompetitive. Notice that this first element of the test automatically protects firms that adopt exclusionary practices in contexts where there is no serious risk of meaningfully weakening rivals; see section 1.4.1 for an example involving a soft drink duopoly. Why might exclusion substantially impair a rival’s ability and/or incentive to compete effectively at later points in time? There are many possibilities. One effect of exclusion is that it reduces the rival’s cash flow by limiting its ability to make profitable sales. When a company cannot access external capital markets on attractive terms (which is often the case for the types of firms that are the victims of exclusionary practices), limited cash 46  Even if the customer is coerced by the threat of retaliation for purchasing from a rival, the choice to accede to the condition must be in the customer’s interest, given the threat, and the customer is no worse off than if the seller had chosen to exercise the same negotiating power in some other way.

A Framework for the Economic Analysis of Exclusionary Conduct   27

can constrain its investments in research and development, as well as in plant and equipment, thereby undermining its ability to offer competitive products in the future. Cash constraints can slow a company’s growth and prevent it from achieving an economically efficient scale.47 They can prevent the company from enjoying a “cushion” against hard times, bad luck, or even bad decisions of the sort that the monopolist, not similarly constrained, can weather. Often, a company with good prospects can turn to external investors to overcome its liquidity constraints. In practice, a greater need for external financing can increase a rival’s cost of capital (Myers 2003). External financing can be particularly expensive when insiders have much better information about the company’s prospects than outside investors, which is often the case for newly emerging rivals (see Jensen and Meckling 1976; Myers and Majluf 1984). The problem of asymmetric information may be so severe that that a company’s ability to raise capital essentially vanishes (Stiglitz and Weiss 1981). Significantly, when outside investors are unaware that the rival’s failure to generate profits results from anticompetitive exclusion (or are simply uncertain about the existence, continuation, and/or impact of anticompetitive conduct), they may conclude incorrectly that the company itself is at fault, that its business plan is ill-conceived, and that its prospects are therefore poor, even if it is, in fact, positioned to compete successfully on the merits.48 Accordingly, anticompetitive exclusion can leave a company cash-starved, dependent on costly sources of finance, and with limited ability to raise funds. A second effect of exclusion is that it can prevent the rival’s product from earning the degree of customer acceptance required to compete effectively. In some contexts, customer interest and acceptance depend on whether the product is widely used, for example because positive evaluations spread by word of mouth, because extensive use establishes reliability, or because adoption by certain market leaders, or in certain applications, confers validation. Limiting usage through exclusion (particularly selective exclusion from opinion leaders) deprives the rival of the opportunity to acquire those competitive advantages. In other circumstances, successful product development depends upon iterative feedback from customers, so hindering access to initial customers can inhibit future product innovation. A third effect of exclusion is that it can prevent the rival from achieving the scale and scope necessary to compete effectively. For example, in industries where firms 47  A leading treatise explains, “A set of strategically planned exclusive dealing contracts may slow the rival’s expansion by requiring it to develop alternative outlets for its products or rely at least temporarily on inferior or more expensive outlets. Consumer injury results from the delay that the dominant firm imposes on the smaller rival’s growth.” (Dentsply at 191, quoting Hovenkamp 2002). The Dentsply court found that the monopolist’s exclusionary conditions “help[ed] keep sales of competing teeth below the critical level necessary for any rival to pose a real threat to Dentsply’s market share,” making them “a solid pillar of harm to competition.” 48  In this way, anticompetitive exclusion can create a “signal jamming” problem. See Fudenberg and Tirole (1986); see also Dentsply (recognizing that the monopolist’s conduct could create the impression that the rivals were ineffective: “The apparent lack of aggressiveness by competitors is not a matter of apathy, but a reflection of the effectiveness of [the monopolist’s] exclusionary policy”).

28   B. Douglas Bernheim and Randal Heeb achieve significant cost reductions and/or insights into product development through learning-by-doing, exclusion can significantly reduce the future competitive threat posed by a rival. A few clarifying remarks are in order concerning the extent of foreclosure required to satisfy the first element of the test. As the courts have recognized, exclusion may be problematic from an antitrust perspective even if the rival is not foreclosed from the entire market.49 The pertinent question is whether the overall scope of exclusion is material, in the sense that it is sufficiently widespread to meaningfully weaken the rival or otherwise prevent the rival from effectively competing for other business. In assessing materiality, the following considerations come into play. First, the scope of exclusion may be material even if the conduct does not entirely prevent the rival from selling to any particular buyer. All else equal, a condition that explicitly or effectively excludes a rival from all of the business served by a customer who purchases 90 units may have the same impact on the rival, and, hence, on competition and consumers, as an exclusionary agreement covering 90% of the business served by a customer who purchases 100 units, or one covering half of the business served by a customer who purchases 180 units. Complete exclusion from any particular buyer may not be needed to reduce the rival’s ability to compete (see, e.g., Dentsply; LePage’s; Microsoft). Second, one must judge the materiality of a company’s exclusionary conduct as a whole, not episode by episode.50 Neither is it sufficient to judge the effect of all episodes of a particular type of conduct or exclusionary condition without taking into account all of the other forms of exclusionary conduct. In the extreme, individual episodes or collections of episodes may each inflict only a little damage on the rival, and yet the overall pattern of conduct may have a substantial effect on the rival’s ability to compete. Any analysis that attempts to evaluate component parts of the conduct independently, rather than assess the totality of the effect, is inevitably biased against a finding of anticompetitive impact, which necessarily depends upon all facets of an exclusionary agenda. Third, a restriction on the channels through which a rival’s product is distributed, including exclusion from important or uniquely positioned downstream firms, can be material in and of itself, even if other routes to the customer remain, and even if the scope of that exclusion is somewhat limited relative to the entire market. For example, the adoption and promotion of a product by an influential set of downstream firms may validate a rival’s product in the minds of consumers; consequently, exclusion from one or more of those firms can materially weaken the rival by depriving its products of validation. As the Third Circuit 49 

“The test is not total foreclosure, but whether the challenged practices bar a substantial number of rivals or severely restrict the market’s ambit” (Dentsply). 50  See, e.g., Continental Ore (“The character and effect of a conspiracy are not to be judged by dismembering it and viewing its separate parts, but only by looking at it as a whole; and in a case like the one before us, the duty of the jury was to look at the whole picture and not merely at the individual figures in it” (internal citations and quotation omitted)); LePage’s (“The relevant inquiry is the anticompetitive effect of [the defendant’s] exclusionary practices considered together”); Anaheim (“[I]‌t would not be proper to focus on specific individual acts of an accused monopolist while refusing to consider their overall combined effect”).

A Framework for the Economic Analysis of Exclusionary Conduct   29

concluded in Dentsply, the “realities of the marketplace” may make the foreclosed channel much more valuable and significant than those channels that remain available.51 Fourth, one must also judge the materiality of a company’s exclusionary conduct relative to the portion of the market that is, in principle, open to the rival at a given point in time, absent artificial barriers. For example, customers may differ in their willingness to purchase products that are less familiar to them or less familiar to those on whom they rely for advice. To earn potential acceptance with such customers, an upstart rival might first have to make significant inroads among those who are less wedded to established products. Artificial exclusion from 20% of the total market might not be sufficiently material to meaningfully weaken the rival if it implies that the rival can legitimately compete for the other 80% of a sufficiently large market. However, the same foreclosure looms much larger if the rival’s product is not yet sufficiently well established to gain acceptance with a large portion of the market. Suppose, for example, that 50% of consumers in the overall market are committed to purchasing the more familiar brand and will remain so until the rival is much better established. In that case, the same exclusion from 20% of the total market limits the rival to competing for a residual 30% slice of the market, rather than 80%, in which case the exclusion may well be material. To put the matter another way, in this example, exclusion from 20% of the market leaves the rival effectively excluded from 70% of total sales. If those customers from whom the rival is excluded are particularly influential in winning over other customers (e.g., because they are opinion leaders), then the effect of the exclusion can be compounded. Fifth, the duration of the exclusionary conduct also bears on materiality. If the conduct is brief, then the effect on the rival will presumably be small. However, in evaluating duration, it is important to avoid confusing the duration of the conduct with the contractual duration of any particular exclusionary agreement. As we noted in section 1.2.1, the latter consideration does not necessarily play an essential role in what we take to be the anticompetitive mechanism of greatest practical concern.52 As long as negative contracting externalities exist at short durations (Element 4, discussed later), the same analysis potentially applies, and exclusion over long time periods can be achieved by stringing together successive short-term agreements. Consistent with this economic logic, agreements imposing exclusivity can be anticompetitive even if the contracts are terminable at will.53 51 The Dentsply court rejected the argument that vendors of artificial teeth had a “viable” method of

distribution in the form of direct sales when they were foreclosed from the dealer network. 52  The contractual duration of exclusionary agreements may, however, be relevant in other settings, where other anticompetitive mechanisms are implicated. E.g., some theories of exclusive dealing require that exclusion directly limits the achievable future sales the rival can make as a result of current investments. Either the agreements must span a period of time that starts before the investment and includes the resulting sales, or exclusive contracts with different customers must cover overlapping time periods that collectively bridge the same time period. See, e.g., Rasmusen, Ramseyer, and Wiley (1991); Segal and Whinston (2000). 53 See Dentsply (“Although the parties to the sales transactions consider the exclusionary arrangements to be agreements, they are technically only a series of independent sales. Dentsply sells teeth to the dealers on an individual transaction basis and essentially the arrangement is ‘at-will.’ Nevertheless, the economic elements involved—the large share of the market held by Dentsply and its conduct excluding competing manufacturers—realistically make the arrangements here as effective as those in written contracts”).

30   B. Douglas Bernheim and Randal Heeb

1.3.2.2.  Element 2 (Enhanced Market Power) The conduct must increase, extend, or maintain market power. The excluding firm must start out with market power and, once the rival is weakened, have greater market power than it would otherwise possess. The purpose of this element of the test is to distinguish harm to competition from mere harm to the rival. Antitrust laws protect competition, not competitors. If, for example, a rival harmed by exclusionary conduct is simply replaced by yet another equally capable competitor, then competition is not diminished. The existence of significant entry barriers is, therefore, generally a necessary but not sufficient condition for enhanced market power. Notably, this element automatically and appropriately protects conduct by firms that neither have nor verge upon monopoly power.54 Even among those firms with substantial market power, only conduct that demonstrably increases or maintains that power raises concerns.

1.3.2.3.  Element 3 (Harm to Consumers) The conduct must cause harm to consumers, whom the antitrust laws are designed to protect. When the conduct enhances the excluding firm’s market power (as Element 2 of the test requires), harm to consumers is usually ensured: consumers will pay higher prices and potentially forgo other benefits of competition, such as improved variety and innovation. However, because it is often necessary to weigh the anticompetitive effects of exclusionary conduct against procompetitive effects, it is not enough simply to establish the existence of harm; it is also important to assess its magnitude. Measuring the degree to which exclusionary practices impair a rival, reduce competition, and raise prices can prove challenging. An alternative approach is to evaluate the magnitude of gains that consumers have derived from a rival’s past competitive activities. Estimates of those gains provide indicators of the economic benefits that the conduct places at risk. When measuring harm to consumers, it is also important to recall that some forms of exclusionary conduct (specifically, exclusionary conditions) are designed to suppress sales beyond those that the excluding firm captures. That outcome is particularly likely in industries with differentiated products. When weighing anticompetitive and procompetitive effects, one should include the lost economic value associated with those suppressed sales. Measuring those losses can be more straightforward than quantifying the lost consumer benefits associated with reductions in future competition. As long as the first three elements are present, exclusion of the rival from a portion of the market during one period of time reduces competition and allows the excluding firm to extract greater rents from buyers in other markets or at future points in time. However, unless a fourth condition is also present, the exclusion is not necessarily anticompetitive. 54  Thus, it is appropriate to treat the absence of market power as a “safe harbor” for the use of vertical practices with potentially exclusionary effects. That said, it is also important to bear in mind that firms may collectively wield significant market power through explicit or tacit collusion.

A Framework for the Economic Analysis of Exclusionary Conduct   31

1.3.2.4.  Element 4 (Negative Contracting Externalities) By diminishing the rival’s ability to compete, the exclusionary conduct must give rise to negative contracting externalities, reflecting more effective expropriation of economic benefits from other parties, from which the parties to an exclusive arrangement can jointly benefit. As we explained in section 1.2.1, such externalities are typically present when the rival potentially serves many customers, either in the same market or different markets,55 especially when a portion of the benefits of competition would be passed along to downstream parties, such as final customers. Thus, evaluating this fourth element of the test is usually straightforward. However, as we emphasized in section 1.2.1, the issue involves some subtleties. For example, negative contracting externalities borne only by the rival, or by downstream consumers served by a single customer, typically do not sustain anticompetitive exclusionary conduct, because they usually do not reflect more effective expropriation of economic benefits by the parties to the exclusionary agreement. It is worth emphasizing that, in the presence of negative contracting externalities, evidence that certain customers eagerly agreed to or even sought out an exclusionary relationship with the seller sheds no light on the question of whether the conduct is anticompetitive. Considered in isolation, each deal between the seller and a buyer is necessarily mutually beneficial, even accounting for its subsequent impact on competition. However, such deals may harm other parties (especially other buyers and/or downstream consumers) who are not part of the deal. Consequently, every buyer might be better off if no deals were consummated; nevertheless, recognizing the potential for mutual bilateral benefit, each buyer has a strong individual incentive to enter an exclusionary arrangement, and even to seek it out. This four-part test is generally consistent with the approach to predatory pricing adopted by the US Supreme Court. Specifically, a competitor may be held liable under the antitrust laws for setting prices below costs if there is a sufficient probability of recouping, through subsequent monopoly profits, more than the losses sustained (see Brooke Group; Matsushita). Elements 1–3 ensure that the seller benefits from greater monopoly profits as a consequence of exclusion. Those benefits will exceed the cost of securing the participation of customers only if Element 4 is also present. While it is possible that exclusionary conduct could be anticompetitive without meeting this four-part test through some other mechanism, when considering such possibilities, it is important to avoid common fallacies. For example, it is sometimes alleged that loyalty discounts permit a dominant firm facing limited competition to earn monopoly profits instantly by denying rivals access to customers.56 The following simple example illustrates the idea. A single customer buys five units of a good from a monopolist at a price of $200 per unit. A new entrant arrives on the scene, possessing an ability (in the

55  Anticompetitive effects can arise if the excluding firm and the rival compete in “noncoincident markets.” See Bernheim and Whinston (1999). 56  See, e.g., Jacobsen (2010), who quotes testimony to this effect by Einer Elhauge.

32   B. Douglas Bernheim and Randal Heeb near term) to supply one unit. Both produce the good at a cost of $100 per unit. To defeat the entrant’s threat, the monopolist increases its price to $250 per unit but offers a discount of $50 per unit if the customer buys five. Faced with a choice between buying four units from the monopolist at a total cost of $1,000 or four units from the monopolist at a total cost of $1,000 plus a fifth unit from the rival at any positive price, the customer will clearly buy all five units from the monopolist. According to the argument, the entrant is foreclosed, and the dominant firm continues to charge the monopoly price for all five units. There is, however, a fly in the ointment, because the argument ignores an important possibility: facing these terms, the customer might choose to buy one unit from the rival and nothing from the dominant firm. Implicitly, the example assumes that the customer’s marginal benefit from consuming the good is $200 per unit for the first five units, and less thereafter.57 Let’s suppose that the rival sets the price of a single unit equal to its cost. Then the consumer’s net benefit from buying five units under the loyalty discount program is zero, while his net benefit from buying a single unit from the rival is $100. The second option is clearly better than the first. Because the dominant firm must leave the customer with a net benefit of at least $100 (lest the customer decide to deal only with the rival), it cannot improve on the profits it receives when offering the first four units at a price of $200 and a fifth unit at a price of $100, matching (rather than excluding) the rival.

1.4.  Determining If Exclusionary Conditions Are Procompetitive If conduct is found to be both exclusionary in the first stage of the inquiry and to have anticompetitive effects in the second, we then move to the third stage:  determining whether the conduct also has procompetitive effects, and weighing those effects against the costs of reduced competition. The scholarly literature identifies a number of potential procompetitive rationales for exclusionary practices (see, e.g., Marvel 1982; Bernheim and Whinston 1998, section V). Generally, these rationales proceed from the premise that, because written contracts are imperfect, the conflicting interests of any given buyer and seller can cause their relationship to operate inefficiently. In some

57 

Because the example involves loyalty discounts, it plainly assumes that the monopolist can employ nonlinear price schedules. If the consumer’s marginal benefit differed over the first five units, then a monopolist would generally earn higher profits by using a nonlinear price schedule, rather than by setting a fixed price. See Oi (1971); Maskin and Riley (1984). Thus, the premise that the monopolist charges a fixed price of $200 in the rival’s absence implies that the consumer’s marginal benefit is the same over the first five units. That benefit cannot be less than $200 per unit or the consumer would buy nothing; it cannot be greater than $200 per unit or the monopolist would raise its price.

A Framework for the Economic Analysis of Exclusionary Conduct   33

circumstances, those inefficiencies may be reduced when the relationship excludes other sellers, other buyers, or both. As an example, suppose that a manufacturer’s sales depend heavily on its reputation with customers, and that the point of contact with customers is a sales-and-service organization (which we will call the “rep” for short). If the relationship between the manufacturer and the rep is nonexclusive, the rep can then potentially expropriate some of the benefits generated by the manufacturer’s investments in product promotion and/ or quality. First, the rep can divert customers at the point of sale to other products from which it receives greater profits, even if the customers come to the rep seeking the manufacturer’s product because of the latter’s investments. Second, if the manufacturer invests in quality improvements, customers may then misattribute part of the incremental value they receive to the quality of service and repairs provided by the rep, which again allows the rep to benefit by selling competing products. Anticipating these outcomes, the manufacturer may underinvest or choose not to make any investments in the first place. Designing a nonexclusive contract that overcomes these problems can prove difficult. For example, as a matter of principle, one potential solution is to shift responsibility for the pertinent investments to the rep. But in practice, an agreement cannot call upon the rep to make investments in product quality on the behalf of a manufacturer, and in any event, conflicting incentivization by multiple manufacturers can produce inefficient outcomes (see Bernheim and Whinston 1998, section V). Thus, exclusivity can emerge as a more efficient solution. Automobile dealerships and prestige goods retailers are often cited as exemplifying this motive for exclusivity. The procompetitive and anticompetitive rationales for exclusion share a common characteristic: they all presuppose the existence of market imperfections that prevent collections of parties from achieving mutually efficient outcomes through contracts. In a Coasian world with perfectly efficient contracts, there would be no contracting externalities. Any exclusionary relationship that inefficiently reduced a market’s total contribution to economic value by increasing some party’s market power would be avoided; the parties would instead collectively opt for a more efficient alternative, along with a distribution of benefits that would leave all of them better off. Consequently, anticompetitive exclusion would not occur. Similarly, the agreements reached by buyers and sellers would be sufficiently comprehensive to preclude opportunism. Because there would be no incentive problems to remedy, procompetitive exclusion would not be necessary. Significantly, procompetitive and anticompetitive exclusion reflect different types of contracting failures. Procompetitive exclusion occurs when contracting imperfections afflict the relationship between a buyer and a seller; in that case, exclusion provides a partial solution to the contracting problem. In contrast, anticompetitive exclusion occurs when contracting imperfections afflict the relationships between multiple parties on the same side of the exclusionary relationship (e.g., several buyers, and potentially customers who are downstream from those buyers). In that case, a party on the opposite side of the relationship (e.g., a seller) takes advantage of the afflicted parties’ lack of coordination and conjures its business relationships to extract a larger share of the total pie. Thus, determining whether exclusionary conduct is pro- or anticompetitive always

34   B. Douglas Bernheim and Randal Heeb requires an examination of the types of market imperfections and associated contracting failures that lead to exclusion. When evaluating procompetitive explanations for exclusionary conduct, it is important to be wary of ex post rationalizations. The mere fact that an economist can concoct a logically coherent rationalization consistent with the details of a case does not mean that the rationalization is correct. Unlike a company seeking anticompetitive ends, one engaging in procompetitive exclusion has no reason to disguise its objectives. Thus, when the objective is procompetitive, one expects to find contemporaneous documents that describe the problem and characterize the conduct as an attractive solution.58 Exchanges between the buyer and seller may be particularly informative concerning the nature of the contracting problem (if any).59 However, it is also important to keep in mind that a strategically savvy company can defensively conjure a helpful paper trail. Therefore, courts should not automatically lend credence to such evidence when it is uncovered. When conduct yields both anticompetitive and procompetitive effects, one must weigh one against the other. In the previous subsection, we emphasized the importance of quantifying the harm borne by consumers; the need to quantify procompetitive benefits is no less important, and the same standard of proof should apply. Consistent with the antitrust laws, the focus should once again be on benefits to consumers. Furthermore, it is generally not appropriate to measure those benefits by comparing market outcomes with and without the conduct. Rather, the proper approach is to compare the market outcome with the conduct to the outcome with the most efficient nonexclusionary alternative solution to the contracting problem that the conduct addresses. Consider the following simple example: without the conduct, consumers would gain $100 million in value from increased competition but lose $150 million in value due to the consequences of worsened incentive problems between buyers and sellers; however, were buyers and sellers to adopt a less efficient, nonexclusive solution for the incentive problems, the losses would be $50 million rather than $150 million. In that case, it is appropriate to compare the $100  million gain with the $50  million loss, not the $150 million loss, and to conclude that the conduct is, on balance, anticompetitive.

1.4.1.  Some Illustrative Examples To illustrate the application of our framework, we will briefly describe two contrasting examples. In the first, a dominant supplier of false teeth demands (and receives) exclusive deals with most of its large distributors. The resulting exclusion prevents a 58  That said, if the genesis of the conduct predates the available records, then the absence of more recent documents that continue to reference the problem may or may not be surprising, depending on the facts of the case. 59  The absence of such discussion does not imply that the restrictions are anticompetitive, as their necessity may be obvious to both parties, or at least to the party proposing the contract.

A Framework for the Economic Analysis of Exclusionary Conduct   35

smaller rival from gaining access to a substantial and important portion of the market. As a result, the rival loses not only the immediate sales denied to it by the exclusionary conduct but also future sales that it would gain from customers who, but for the exclusivity, might have made introductory or exploratory purchases leading to larger future contracts. The rival is denied both the opportunity to establish its reputation with those end customers and also the scale that would support a more extensive and effective marketing and distribution operation. As a consequence, the rival is weaker in the future, and the dominant firm gains not only from its current sales but also from reduced competition for future sales. There is no contracting failure that calls for exclusivity between the distributor and supplier. The harms resulting from reduced future competition are divided among distributors (both those that entered into exclusive deals and those that did not), as well as consumers, who cannot efficiently coordinate their actions to forestall the dominant firm’s opportunism. Here, the conduct is plainly anticompetitive. In the second example, the owner of a sports stadium or a fast-food franchise has a limited degree of monopsony power with respect to soft drinks that it resells on its premises.60 Practical considerations, such as space constraints at food-vending stations, rule out offering highly similar products (i.e., both Coke and Pepsi, both Sprite and 7UP, and so forth). The establishment owner therefore requires the Coca-Cola Company and PepsiCo to compete for the right to be the establishment’s exclusive supplier of soft drinks. It is difficult to imagine that the Coca-Cola Company would noticeably impair the future competitiveness of PepsiCo by entering into such an agreement with the establishment owner or that PepsiCo would noticeably impair the Coca-Cola Company.61 Each enters into exclusive arrangements with similar types of establishments, and both distribute their products through a wide array of unaffected channels, such as grocery stores, convenience stores, and vending machines. Accordingly, in this example, the conduct is procompetitive.

1.5.  Concluding Remarks The framework for analyzing the competitive effects of exclusionary conduct proposed in this chapter does not aspire to an unachievable ideal. Rather, it reflects a 60 

While this hypothetical example involves exclusivity at the venue or chain, another case, PepsiCo, Inc. v. Coca-Cola Co., addressed a similar question involving soft drink exclusivity at the distribution level. There, too, the court found no evidence of an anticompetitive effect. 61  Some customers might have sufficiently strong preferences to go without a soft drink entirely rather than purchase the exclusively provided alternative, but in the present example, this would not result in a material reduction in the rival’s ability to compete for other sales. Competition in the soft drink market remains robust, even under exclusivity. More generally, however, such arrangements could adversely affect competition, depending on the fact pattern. E.g., if exclusive arrangements are sufficiently widespread, then they could adversely affect the viability of an upstart soft drink manufacturer.

36   B. Douglas Bernheim and Randal Heeb combination of both sound economic principles and reasonable, practical compromises. We have made those compromises, where possible, to favor simplicity, clarity (in terms of what is and is not permitted), and predictability (with respect to the outcome of an informed inquiry). It is also worth emphasizing that the framework is reasonably conservative. For example, the requirements for establishing the presence of anticompetitive effects in stage 2 of the inquiry resemble those that are currently applied in cases of predatory pricing (once below-cost pricing is established); to our knowledge, relatively few commentators have claimed that the standards for establishing predation are too lax. A conservative approach is, in our view, preferable to one that favors a wider range of challenges and, thereby, risks chilling beneficial competition. Contrary to our objective of achieving clarity, we have intentionally left portions of the analytic framework somewhat vague—for example, the appropriate measure of cost to apply when evaluating whether bundled pricing is exclusionary, given any particular fact pattern. In such instances, the appropriate details depend on subjective judgments concerning social costs and benefits, and consequently are more appropriately left to policymakers and the courts; our framework usefully clarifies the pertinent tradeoffs.

References Abito, Jose Miguel, and Julian Wright. 2008. Exclusive Dealing with Imperfect Downstream Competition. International Journal of Industrial Organization 26: 227–46. Aghion, Philippe, and Patrick Bolton. 1987. Contracts as a Barrier to Entry. American Economic Review 77: 388–401. Areeda, Phillip E., and Herbert Hovenkamp. 2008. Fundamentals of Antitrust Law. New York: Aspen Publishers (3rd ed.). Asker, John, and Heski Bar-Isaac. 2014. Raising Retailers’ Profits: On Vertical Practices and the Exclusion of Rivals. American Economic Review 104: 672–86. Bernheim, B. Douglas, and Michael D. Whinston. 1998. Exclusive Dealing. Journal of Political Economy 106: 64–103. Bernheim, B.  Douglas, and Michael D.  Whinston. 1999. Anti-competitive Exclusion and Foreclosure through Vertical Agreements. CORE Lecture Series, Center for Operations Research and Econometrics, Louvain-La-Neuve, Belgium, Université Catholique de Louvain. Bernheim, B. Douglas, and Michael D. Whinston. 2008. Microeconomics. New York: McGrawHill/Irwin. Bork, Robert. 1978. The Antitrust Paradox: A Policy at War with Itself. New York: Basic Books. Carlton, Dennis W., and Michael Waldman. 2002. The Strategic Use of Tying to Preserve and Create Market Power in Evolving Industries. Rand Journal of Economics 33: 194–220. Elhauge, Einer. 2011. United States Antitrust Law and Economics. New York: Foundation Press (2nd ed.). Elzinga, Kenneth G., and David W. Mills. 1994. Trumping the Areeda Turner Test: Recoupment Standard in Brooke Group. Antitrust Law Journal 62: 559–84.

A Framework for the Economic Analysis of Exclusionary Conduct   37

Farrell, Joseph. 2005. Deconstructing Chicago on Exclusive Dealing. University of California, Berkeley Competition Policy Center Working Paper No. CPC05-053. Fudenberg, Drew, and Jean Tirole. 1986. A “Signal-Jamming” Theory of Predation. Rand Journal of Economics 17: 366–76. Fumagalli, Chiara, and Massimo Motta. 2006. Exclusive Dealing and Entry:  When Buyers Compete. American Economic Review 96: 785–95. Greenlee, Patrick, and David S.  Reitman. 2006. Competing with Loyalty Discounts. US Department of Justice EAG Discussion Paper 04-2. Hart, Oliver D., and Jean Tirole. 1988. Contract Renegotiation and Coasian Dynamics. Review of Economics Studies 55: 509–40. Hungar, Thomas G., and Ryan G.  Koopmans. 2009. Appellate Advocacy in Antitrust Cases: Lessons from the Supreme Court. Antitrust 23: 53–59. Jaeckel, Jeffrey A. 2010. “LePage’s”, “Cascade Health Solutions” and a Bundle of Confusion: What Is a Discounter to Do? Antitrust 24: 46–51. Jensen, Michael C., and William H. Meckling. 1976. Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure. Journal of Financial Economics 3: 305–60. Kobayashi, Bruce. 2005. The Economics of Loyalty Discounts and Antitrust Law in the United States. Law and Economics Series, George Mason University, Working Paper No. 05-26. Markus, Blake I. 2008. The Ninth Circuit and Third Circuit Are on Separate “LePage’s”. Missouri Law Review 73: 907–20. Marvel, Howard. 1982. Exclusive Dealing. Journal of Law and Economics 25: 1–25. Maskin, Eric, and John Riley. 1984. Monopoly with Incomplete Information. Rand Journal of Economics 15: 171–96. Matthewson, Frank, and Ralph Winter. 1987. The Competitive Effects of Vertical Agreements: Comment. American Economic Review 77: 1057–62. Melamed, A.  Douglas. 2005. Exclusionary Conduct under the Antitrust Laws:  Balancing, Sacrifice, and Refusals to Deal. Berkeley Technology Law Journal 20: 1247–67. Myers, Stewart C. 2003. Financing for Corporations. In George M.  Constantinides, Milton Harris, and René M.  Stulz, eds., Handbook of the Economics of Finance 1A:  215–53. Amsterdam: Elsevier. Myers, Stewart C., and Nicholas S. Majluf. 1984. Corporate Financing and Investment Decisions When Firms Have Information That Investors Do Not Have. Journal of Financial Economics 13: 187–221. Nalebuff, Barry. 2005a. Exclusionary Bundling. Antitrust Bulletin 50: 321–70. Nalebuff, Barry. 2005b. Tied and True Exclusion: Comment on Jean Tirole’s “The Analysis of Tying Cases: A Primer.” Competition Policy International 1: 41–53. O’Brien, Daniel P., and Greg Shaffer. 1992. Vertical Control with Bilateral Contracts. Rand Journal of Economics 23: 299–308. Oi, Walter Y. 1971. A Disneyland Dilemma: Two-Part Tariffs for a Mickey Mouse Monopoly. Quarterly Journal of Economics 85: 77–96. Ordover, Janusz A., and Robert D. Willig. 1981. An Economic Definition of Predation: Pricing and Product Innovation. Yale Law Journal 91: 8–53. Rasmusen, Eric B., J. Mark Ramseyer, and John S. Wiley. 1991. Naked Exclusion. American Economic Review 81: 1137–45. Salop, Steven C. 2006. Exclusionary Conduct, Effect on Consumers and the Flawed Profit-Sacrifice Standard. Antitrust Law Journal 73: 311–74.

38   B. Douglas Bernheim and Randal Heeb Salop, Steven C., and David T.  Scheffman. 1983. Raising Rivals’ Costs. American Economic Review 73: 267–71. Scott Morton, Fiona. 2012. Contracts That Reference Rivals. Transcript, Georgetown University Law Center Antitrust Seminar. Retrieved from http://www.justice.gov/atr/public/​ speeches/281965.pdf. Segal, Ilya. 1999. Contracting with Externalities. Quarterly Journal of Economics 114: 337–88. Segal, Ilya, and Michael D. Whinston. 2000. Naked Exclusion: Comment. American Economic Review 90: 296–309. Simpson, John, and Abraham L.  Wickelgren. 2007. Naked Exclusion, Efficient Breach, and Downstream Competition. American Economic Review 97: 1305–20. Spengler, Joseph J. 1950. Vertical Integration and Antitrust Policy. Journal of Political Economy 58: 347–52. Stiglitz, Joseph E., and Andrew Weiss. 1981. Credit Rationing in Markets with Imperfect Information. American Economic Review 71: 393–410. US Department of Justice. 2008. Competition and Monopoly:  Single-Firm Conduct under Section 2 of the Sherman Act. Werden, Gregory J. 2006. Identifying Exclusionary Conduct under Section 2: The “No Economic Sense” Test. Antitrust Law Journal 73: 413–33. Werden, Gregory J. 2009. Next Steps in the Evolution of Antitrust Law: What to Expect from the Roberts Court. Journal of Competition Law & Economics 5: 49–74. Whinston, Michael D. 1990. Tying, Foreclosure, and Exclusion. American Economic Review 80: 837–59. Wright, Julian. 2009. Exclusive Dealing and Entry: When Buyers Compete: Comment. American Economic Review 99: 1070–81.

Cases Advo, Inc. v. Phila. Newspapers, Inc., 51 F.3d 1191 (3d Cir. 1995). Anaheim v. S. Cal. Edison Co., 955 F.2d 1373 (9th Cir. 1992). Brooke Grp. v. Brown & Williamson Tobacco Corp., 509 U.S. 209 (1993). Cascade Health Solutions v. PeaceHealth, 515 F.3d 883 (9th Cir. 2008). Concord Boat Corp. v. Brunswick Corp., 34 F.Supp. 2d 1125 (E.D. Ark. 1998). Continental Ore Co. v. Union Carbide & Carbon Corp., 370 U.S. 690 (1962). Conwood Co. v. U.S. Tobacco Co., 290 F.3d 768 (6th Cir. 2002). Eastman Kodak Co. v. Image Technical Serv., Inc., 112. S. Ct. 2072 (1992). General Indus. Corp. v. Hartz Mountain Corp., 810 F.2d 795 (9th Cir. 1987). Indep. Ink, Inc. v. Ill. Tool Works, Inc., 396 F.3d 1342 (9th Cir. 2005). Jefferson Parish Hosp. Dist. No. 2 v. Hyde, 466 U.S. 2 (1984). LePage’s, Inc. v. 3M, 324 F.3d 141 (3d Cir. 2003). Matsushita Elec. Indus. Co. v. Zenith Radio Corp., 475 U.S. 574 (1986). Ortho Diagnostic Sys. v. Abbott Lab., 920 F.Supp. 455 (S.D.N.Y. 1996). PepsiCo, Inc. v. Coca-Cola Co., 315 F.3d. 101 (2d Cir. 2002). SmithKline Corp. v. Eli Lilly & Co., 575 F.2d 1056 (3d Cir. 1978). Tampa Elec. Co. v. Nashville Coal Co., 365 U.S. 320 (1961). United States v. Dentsply Int’l, Inc., 399 F.3d 181 (2005).

A Framework for the Economic Analysis of Exclusionary Conduct   39

United States v. Grinnell Corp., 384 U.S. 563 (1966). United States v. Microsoft Corp., 253 F.3d 34 (D.C. Cir. 2001). United States v. United Reg’l Health Care Sys., No. 7:11-cv-00030-O (N.D. Tex. 2011). ZF Meritor LLC v. Eaton Corp., 696 F.3d 254 (3d Cir. 2012). Summary of Commission Decision of 13 May 2009 relating to a proceeding under Article 82 of the EC Treaty and Article 54 of the EEA Agreement (Case COMP/C-3/37.990—Intel), OJ C227/13, 22.9.2009.

CHAPTER 2

P R E DATO RY P R IC I N G KENNETH G. ELZINGA AND DAVID E. MILLS

Given the enormous stake that antitrust has in low prices, and our extraordinary difficulties assessing predation claims, the best course is to develop predation rules that are both simple and somewhat underdeterrent. Herbert Hovenkamp (2005, p. 161)

2.1. Introduction Predatory pricing occurs when a dominant firm, motivated by the prospect of charging high prices in the future, uses temporary, low prices to drive its competitors out of business. Economists have long recognized that “underselling for predatory purposes” can harm competition and reduce consumer welfare (Giddings, 1887, p. 77). Indeed antitrust enforcement in the United States cut its teeth on claims that predatory pricing violated the Sherman Act.1 It was not until McGee (1958) mounted a serious challenge to the canonical predatory pricing story that economists and legal scholars began to examine this pricing phenomenon carefully. McGee disputed whether a profit-seeking dominant firm would ever launch a predatory pricing campaign against a rival. His challenge provoked two important reactions. The first was judicial skepticism about the plausibility of predatory pricing claims that earlier courts did not have. This development spawned a debate about the proper test or criterion to apply in predatory pricing cases, a debate that is not yet resolved to everybody’s satisfaction. The second reaction was a flood of activity on the part of economic theorists who were skeptical about McGee’s hypothesis that predatory pricing would 1  The US Supreme Court’s landmark opinions are Standard Oil and American Tobacco. Allegations of predatory pricing actually precede the US antitrust laws. Mogul Steamship is an English tort law case that involved claims of below-cost pricing to destroy ocean-shipping competition.

Predatory Pricing   41

rarely be a successful monopolizing strategy. This research program used game theory to identify conditions that would circumvent McGee’s objections and proposed internally consistent economic theories of predatory pricing. The search for a compelling antitrust test to evaluate predatory pricing claims is an exercise in balancing false positive against false negative judicial outcomes. How this balance is made depends in large part on whether one thinks aggressive low-price episodes typically are about ejecting rivals or generally reflect vigorous competition among rivals.2 These priors are shaped and informed (an economist hopes) by what economic theory has to say about the plausibility of predatory pricing. This chapter summarizes both of these post-McGee endeavors: the search for credible economic theories of predatory pricing that can identify those pricing episodes that warrant antitrust sanction, and the search for an optimal predatory pricing rule for use in antitrust enforcement. In its 1993 decision in Brooke Group, the US Supreme Court developed standards for establishing liability in predatory pricing cases. The Antitrust Modernization Commission commended these standards for being “clear and predictable in application and administrable” (2007, p. 89). While Brooke Group has proved too cautious for some economists and antitrust scholars who favor more judicial intervention, the caution built into the Brooke Group standards reflects the Court’s concern with false positives and shows the Court’s determination not to let antitrust get in the way of aggressive price competition.3 This chapter shares the Court’s caution and concurs with its embrace of the Brooke Group standards for conventional predatory pricing claims.

2.2.  The Economics of Predatory Pricing The practice of predatory pricing traditionally is identified with pricing below cost. In her influential history of the Standard Oil Company, Tarbell wrote that a firm threatens “predatory competition” when it sets out ruthlessly and persistently “to sell at cost or less, until the rival is worn out” (1904, p. 60). Not long after, Clark declared that “predatory competition differ[ed] from ordinary competition in that producers who have the . . . most influence on the market . . . do not stop lowering prices at a point which covers all costs, . . . but go below this level” (1926, p. 131). The presumed intention of predatory pricing is to eliminate rivals, whether by forcing them out of the market or by acquiring them on advantageous terms, and thereby to 2 

A false positive concludes that predation has occurred when it did not. A false negative concludes that there is no predatory pricing when in fact there is. 3  The foundation for this caution was laid earlier in the Court’s opinion in Matsushita: “predatory pricing schemes are rarely tried, and even more rarely successful” (p. 589).

42   Kenneth G. Elzinga and David E. Mills establish or maintain monopoly power that will yield supracompetitive profits for the predator. Ordover and Willig (1981) defined a predatory practice as one that is profitable only because it induces the exit of a competitor. While the elimination of rivals was the first recognized consequence of predatory behavior, economists have since classified pricing strategies with other goals such as deterring entry or disciplining rivals to be predatory as well.4 Strategies that deter entry or discipline rivals have adverse effects on consumer welfare, of course, and are the proper concern of antitrust policy. But the focus in this chapter is on pricing behavior that reduces consumer welfare because it induces exit as a monopolizing strategy. In the most common scenario, a predator charges prices below cost and willingly sacrifices profits so it can inflict an unsustainable loss on a competitor, often a new entrant. The predator supposedly is able to withstand the losses that accompany below-cost pricing better than its target because it is much larger and has a “long purse.”5 Once the competitor exits the market, the predator raises its prices and recovers the lost profits. The general notion that a dominant firm might establish or strengthen a monopoly position by waging a price war against a smaller competitor was widely accepted at the time the US antitrust laws were enacted. It was not until McGee (1958) revisited Standard Oil, and raised objections to the prevailing interpretation of that important case, that the strategy of predatory pricing was submitted to careful economic analysis. McGee concluded that Standard Oil did not use predatory pricing to drive out its competitors, as was widely assumed, and in the course of his argument offered several reasons for doubting the plausibility of predatory pricing in general.6 He argued that, as compared to a small competitor, a predator with a large market share would suffer a disproportionate loss from below-cost prices—a loss that the small competitor would be able to amplify by curtailing its own unremunerative sales. Realizing this, the competitor would remain in the market and wait for the predator to capitulate. Anticipating this futile sequence of events, the dominant firm would never resort to predatory pricing. McGee also questioned whether a “long purse” (or what is sometimes called a “deep pocket”) would confer an advantage to a prospective predator. He reasoned that a small firm with limited financial resources would have access to creditors willing to sustain the firm through a siege of low prices because the creditors would understand that this access would convince a predator to abandon the siege. Even if a predator succeeded in driving a small competitor out of the market, the availability of durable equipment and

4  For instance, Joskow and Klevorick write that “[p]‌redatory pricing behavior involves a reduction of price in the short run so as to drive competing firms out of the market or to discourage entry of new firms in an effort to gain larger profits via higher prices in the long run” (1979, pp. 219–20). Milgrom and Roberts define predatory pricing as “the temporary charging of particularly low prices in order to improve long-run profitability by inducing exit, deterring entry, or ‘disciplining’ rivals into accepting relatively small market shares” (1990, p. 112). 5  When a large firm “finds itself matching expenditures or losses, dollar for dollar, with a substantially smaller firm, the length of its purse assures it of victory” (Edwards, 1955, p. 334). 6  While acknowledging the “extraordinary influence” of McGee’s analysis of Standard Oil, Dalton and Esposito offer an alternative interpretation of events (2011, p. 245).

Predatory Pricing   43

specialized human capital released by the failed competitor would facilitate subsequent entry or reentry. Consequently, the predator’s gains from driving the competitor from the market would be short-lived. This, according to McGee, would make predatory pricing a costly and futile undertaking. Finally, McGee argued it would be more profitable for a prospective predator to achieve a monopoly position by acquiring its bothersome competitors instead of waging a costly price war to drive them out of business.7 One consequence of McGee’s spadework in the Standard Oil record was to raise suspicion about the merits of other alleged instances of predatory pricing. Elzinga (1970) examined the record of an early case in which US explosives manufacturers were accused of using predatory pricing tactics, and found no evidence that they had. Koller (1971) examined twenty-three cases in which defendants were convicted of predatory pricing and found little evidence of below-cost pricing and even less evidence that below-cost pricing tactics were directed at the elimination of a rival. When Adelman deconstructed the antitrust case against A&P, he found that “[t]‌here is not a single instance in the record of the sequence: lower prices—fewer competitors—higher prices again,” notwithstanding the government’s successful prosecution of the once prominent food retailer (1959, p. 373).8 Another consequence of McGee’s paper was to arouse interest in the economic theory of predatory pricing and stimulate critical commentary on the thesis that successful predatory pricing would be rare and unlikely. These criticisms have led to qualifications and refinements of the theory of predatory pricing.9 For instance, McGee’s argument about the threat of reentry requires that the fixed costs of entry or reentry are small, or if they are large they must not be sunk. Yamey (1972) countered McGee’s reentry argument by contending that episodic predatory pricing might enable a predator to discourage reentry by cultivating a “predatory” reputation, an insight that frequently reappears in subsequent analyses of predatory pricing. To counter the fault McGee found with the long-purse story, it was necessary to explain why a small competitor might not have access to sufficient financial resources to withstand a siege of below-cost pricing by a predator with substantial financial resources (Telser, 1966). One such explanation holds that the smaller competitor’s access to capital markets is limited by asymmetric information prevalent in these markets. McGee’s claim that predatory pricing is a less profitable monopolization strategy than acquiring one’s competitors does not mean that a dominant firm bent on eliminating its competitors would always reject predatory pricing. Acquiring competitors became 7 

Two years before McGee’s seminal article, Leeman expressed skepticism that Rockefeller used predatory pricing as a monopolizing tactic and argued that “localized price cutting” by a large incumbent was unlikely to thwart new entry (1956, p. 332). He considered whether a dominant firm might use predatory pricing to “soften up” smaller rivals in order to acquire them at fire-sale prices, but concluded that price wars are “a very costly method of preserving a dominant position and that probably in most cases in the long run the costs are prohibitive” (1956, p. 330). 8  Adelman’s extensive analysis of the A&P case actually preceded McGee’s article. 9 Ordover and Saloner (1989) and, more recently, Motta (2004) provide summaries of the counterarguments.

44   Kenneth G. Elzinga and David E. Mills more difficult in the years after McGee made this claim because federal antimerger enforcement became more stringent. Also, predatory pricing and horizontal mergers are not necessarily alternative monopolizing strategies. A  dominant firm might use predatory pricing to discourage a competitor and thereby reduce the firm’s acquisition price (Telser, 1966; Yamey, 1972). Burns’s (1986) investigation of American Tobacco’s acquisition of over forty small, geographically separated competitors at the turn of the 20th century found that the terms of sale for these buyouts were depressed by price wars instigated by American Tobacco.10 Countering McGee’s skepticism about predatory pricing launched a search for more complete and internally consistent theories of predation, a project that coincided with the ascension of game theory within the economics profession. This search led away from case study narratives about predation and to the identification of theoretical conditions that would uphold predatory pricing as an equilibrium strategy. A characteristic generally shared by the reformulated theories of predatory pricing is that incomplete information handicaps firms that are the victims of predatory pricing. A predator exploits this handicap to raise the competitor’s doubts about the rationale behind the predator’s low prices and hence raise doubts about the competitor’s prospects for success in the market. The predator raises these doubts to persuade the victim to withdraw from the market (or persuade potential entrants to refrain from entering). Ordover and Saloner (1989) summarized the theories of predatory pricing that emerged in reaction to McGee’s critique, and assigned them to three categories:  multiple-market reputation theories, signaling theories, and reconstituted long-purse theories. Exploring the theory of predatory pricing attracted a talented group of economic theorists, and an expansive literature on the subject appeared following Selten’s (1978) influential demonstration of the chain store paradox. This is not the place to give a full account of these theories, but a few prominent examples are indicative of the literature. McGee argued that the losses a predator would experience in a predatory pricing episode would be too great to recoup even if the targeted competitor is driven out of the market. But if the predator operates in multiple markets in which it faces the prospect of new entry, the payoff from driving a new entrant out of one market with below-cost prices may come from other markets where new entry is deterred. If new or prospective entrants have incomplete information about the predator’s costs, they are uncertain whether the predator’s low, postentry prices are below the predator’s own costs (and hence unsustainable for the predator in the long run) or above the predator’s costs due to the firm’s efficiency (and hence sustainable over time). Kreps and Wilson (1982) showed that a multimarket monopolist facing poorly informed potential entrants in a succession of markets may find it profitable to respond to one or several early entrants with 10  Genesove and Mullin examined the pricing behavior of the American Sugar Refining Company at roughly the same period and concluded that the firm used below-cost pricing “to lower the acquisition price of entrants and small incumbents” (2006, p. 67).

Predatory Pricing   45

below-cost prices to establish a reputation for being more efficient than it really is. The payoff to the predator then comes from the entry-deterring effect this reputation has on subsequent prospective entrants. The loss the predator incurs in a “demonstration” market becomes an investment that pays off in other markets. Signaling theories of predatory pricing incorporate a similar informational asymmetry but can apply to a single market in which an incumbent’s goal is to deter entry (Milgrom and Roberts, 1982; Fudenberg and Tirole, 1986). To understand this scenario, assume that a potential entrant is uncertain about its own postentry prospects because it is not fully informed about the incumbent firm’s costs, the size of the market, or other factors that would affect the potential entrant’s profits. The entrant’s best signal of its own prospects may be the pre-entry prices of the supposedly better-informed incumbent. By charging low prices, the incumbent can manipulate this indicator to magnify the entrant’s doubts about whether it would be efficient enough to survive competing with the incumbent. The low price signal thereby deters the entrant. For the long-purse theory to trump McGee’s criticism, there must be a reason why a new entrant would lack sufficient external funding to remain in the market with a predatory incumbent gorged with liquidity. One such explanation holds that the new entrant’s access to the necessary resources is limited by asymmetric information of a kind that is prevalent in capital markets (Bolton and Scharfstein, 1990). The key insight in this theory is that lenders would be reluctant to continue funding an entrant whose initial financial performance is poor. This reluctance is due to a lender’s inability to tell whether the entrant’s poor financial performance is due to the incumbent’s predatory pricing or the entrant’s inefficiency. As the price war continues, the entrant exhausts its own financial resources and exits the market. Victory follows for the predator.

2.3.  Predatory Pricing and Antitrust Enforcement In the United States, predatory pricing allegations have been brought under both the Sherman Act and the Clayton Act.11 The US Supreme Court has defined predatory pricing as “pricing below an appropriate measure of cost for the purpose of eliminating

11  Section 2 of the Sherman Act prohibits monopolization. Section 2 of the Clayton Act singles out price discrimination for special scrutiny, distinguishing primary from secondary line price discrimination. Primary line price discrimination is when the alleged effect of the discriminatory pricing is at the level of the discriminating firm. Secondary line price discrimination takes place when the alleged effect of the discriminatory pricing is at the level of the buyers of the discriminating firm. Predatory pricing need not involve price discrimination.

46   Kenneth G. Elzinga and David E. Mills competitors in the short run and reducing competition in the long run.”12 In the EU, predatory pricing allegations are brought as an abuse of a dominant position under Article 102 of the Treaty on the Functioning of the European Union. The European Court of Justice held in Tetra Pak II that an abuse of dominance is established when a dominant firm charges prices below average variable cost (AVC), or when prices are below average total cost (ATC) if the intention to eliminate a competitor is established. Detecting predatory pricing is a greater challenge than detecting many other antitrust offenses because the offense involves charging low prices. Low prices, in themselves, generally confer benefits on consumers. But low, predatory prices are problematic because they trigger a sequence of events that leads to high prices. From a consumer’s vantage point, predatory pricing (at least initially) looks a lot like vigorous competition. This is why antitrust requires an error-cost framework to evaluate predatory pricing allegations in order to weigh the likelihood and consequences of false positive outcomes against false negative outcomes. In the years before Areeda and Turner (1975) proposed a specific test for distinguishing whether prices were predatory, predatory pricing in the United States was inferred mainly from ad hoc demonstrations of predatory intent. Demonstrating predatory intent usually involved an assessment of the defendant’s conduct, and this could range from a pattern of declining prices to an assessment of the language used by a firm in its internal strategy documents. If an enthusiastic sales manager gave instructions to “kill the competition,” this might be construed as predatory intent. There was, to say the least, no clear standard for identifying predatory intent, and it is not hard to see the difficulty of distinguishing a subjective state of mind seemingly bent on the destruction of a rival from a subjective state of mind that is set on winning business and letting the rival’s chips fall where they may. In Utah Pie the Supreme Court deduced predatory intent from a “drastically declining price structure” when the defendant cut prices to establish itself in a new sales territory, and from the firm’s internal documents. In an effort to bring economic analysis and a measure of predictability to the task of proving predatory intent, Areeda and Turner proposed a concrete test for determining whether a defendant’s low prices were predatory. This test asked whether those prices were below the firm’s reasonably anticipated average variable cost AVC, where AVC was a proxy for marginal cost (MC).13 Areeda and Turner rationalized that if the defendant’s prices were below its AVC, the low prices probably revealed predatory intent. There are two reasons for drawing this inference; both would be familiar to any economics student. One reason is that prices this low generally are not remunerative for the defendant in the short run and hence require an explanation other than short-run profit maximization. One such explanation is predatory pricing. The other reason for the inference of predatory intent is that with prices below AVC, an equally efficient competitor would rather close down its operations than match 12 

Cargill, Inc. v. Monfort of Colo., Inc., 479 U.S. 104, 117 (1986). Areeda and Turner believed that MC was the proper benchmark, but that AVC was more easily measured. 13 

Predatory Pricing   47

the defendant’s low prices and continue producing. Areeda and Turner proposed this test to protect otherwise efficient firms from being forced out of the market. But the Areeda-Turner test does not encumber defendants with an obligation to protect the viability of an inefficient competitor. Effective competition provides no such protection for inefficient competitors, so it would be perverse for competition policy to impose this obligation. Posner (2001) proposes that the equally efficient competitor standard should be applied for all claims that involve exclusionary practices. The Areeda-Turner test was designed to assign the burden of proof in predatory pricing cases. If those prices were above AVC, then they are presumed to be nonpredatory because prices above AVC generally do not signal a departure from short-run profit maximization. Areeda and Turner proposed that such prices should be a safe harbor for defendant firms. If a defendant’s prices were below AVC, the defendant must prove that its prices were not motivated by predatory intent. This might be done, for instance, by showing the low prices were an introductory offer or a temporary promotion. The Areeda-Turner test is perceived as a cautious test and in some corners it is viewed as being too permissive. Early dissatisfaction arose because the test does nothing to detect strategic price-cutting behavior that threatens the viability of competitors when, if ever, the defendant’s prices are above AVC but below ATC. Scherer (1976) disapproved of evaluating an allegation of predatory pricing by applying a single price-cost test, and favored a more open ended rule-of-reason analysis as had been customary in the past. Others proposed rules that placed less reliance on the price-cost relationship than the Areeda-Turner standard. Every alternative that has been proposed has its own set of difficulties in its application.14 Recognizing that a predatory incumbent would raise prices once a new entrant is driven from the market by the incumbent’s low prices, Baumol (1979) proposed a rule that would require an incumbent who cuts prices by any amount in response to entry to maintain those low prices for an extended period. The rationale for this approach was to deter predation by making it more difficult for an incumbent to recoup any losses that stem from price-cutting. Williamson (1977) proposed a rule that would limit the aggressiveness of an incumbent’s response to a new entrant. Williamson’s approach would prohibit any increase in an incumbent’s output level in response to entry. Edlin (2002) advocated a rule that would prohibit an incumbent from reducing prices substantially in response to entry. The goal of Edlin’s “price freeze” proposal, which combined elements

14  Detecting predatory pricing by a state-owned enterprise presents a different set of problems because the firm’s objectives are various and multidimensional and are not limited to the pursuit of private profit (Lott, 1990). This does not mean that state-owned enterprises are less likely to acquire or preserve a dominant position by means of below-cost pricing. For instance, Sappington and Sidak show “how the diverse goals that a public enterprise faces may lead it to act more aggressively toward its rivals than a private enterprise” (2003, p. 199). In any case, the non-profit-seeking conduct of a public enterprise makes a comparison of the firm’s prices and costs a flawed indicator of predatory intent. See generally Sokol (2009).

48   Kenneth G. Elzinga and David E. Mills of the Baumol and Williamson rules, was to discourage dominant firms from charging high prices before an entrant appears. Joskow and Klevorick (1979) favored a two-tier test. The first tier would examine structural conditions in the market, such as market shares, entry conditions, and profit histories, to assess whether a defendant has sufficient market power to make it plausible that predatory pricing could be successful. If a defendant in a predatory pricing case does not have significant market power, there is no need to inquire further. But if the first-tier examination indicates that the defendant has enough market power to be a predatory threat, the analysis would proceed to the second tier, in which the firm’s prices and costs would be compared. For an incumbent firm that satisfies their first-tier criterion, Joskow and Klevorick proposed that prices below average total costs should establish a presumption of predation. Motta advocates using a more lenient variation of Joskow and Klevorick’s second-tier test. He proposes that a price below AVC “should be presumed unlawful, with the burden of proving the opposite on the defendant”; a price above AVC but below ATC “should be presumed lawful, with the burden of proving the opposite on the plaintiff,” and a price above ATC “should definitely be considered lawful, without exceptions” (2004, p. 442). The principal reservation some economists have about relying on the Areeda-Turner test is that AVC or even MC may not be the “correct price floor: whether some potentially higher price floor would not be more appropriate from the welfare standpoint given the strategic nature of predatory conduct” (Ordover and Saloner, 1989, p. 582). This reservation springs from the post-McGee game-theoretic theories of predation that incorporate incomplete information. In these theories, strategic behavior by a dominant firm seeking to induce a competitor’s exit often does not involve prices below cost, regardless of how costs are measured. If these theories are relevant for antitrust policy, the relationship between a defendant’s prices and its marginal costs is not a reliable indicator of predatory intent. Charging prices below cost is a feature of some strategies for eliminating a competitor, but it is not an essential feature of other strategies. The Areeda-Turner test does not distinguish whether a defendant’s pricing conduct conforms to the predictions of any above-cost theory of predatory pricing. But this shortcoming does not matter in the great majority of predatory pricing claims (going all the way back to Standard Oil) that explicitly allege below-cost pricing. The Areeda-Turner test provides vital guidance for assigning the burden of proof in these predatory pricing cases. Fisher (2007) disapproves of reliance on any one “bright-line” test for distinguishing predatory pricing from competitive pricing. But Fisher acknowledges that the Areeda-Turner test is a useful diagnostic tool, given the difficulty of drawing the correct inference about a defendant’s intent from the firm’s conduct, and recognizing that a test for predatory intent is not the same thing as a definition of predation. The circumstance that generates the most serious reservation about reliance on the Areeda-Turner test is when firms have high fixed costs and very low MC or AVC. In markets with this characteristic, a defendant has more leeway than elsewhere to cut prices and impose losses on a competitor (vis-à-vis ATC) without running afoul of the

Predatory Pricing   49

Areeda-Turner test. The most frequently cited examples are airlines and firms with valuable intellectual property.15 Notwithstanding the scrutiny and the criticism it has received, the Areeda-Turner test has altered the way courts evaluate predatory pricing claims. Phlips (1995) proposes that the reason the Areeda-Turner test became so influential is the test’s conceptual simplicity. One thing is clear: an important consequence of the Areeda-Turner test was to raise the burden for plaintiffs in predatory pricing cases. The number of such cases fell sharply in the years that followed (Salop and White, 1988).

2.4.  Brooke Group The most important predatory pricing opinion issued by the Supreme Court since Areeda and Turner proposed their test is Brooke Group (1993).16 This case involved two cigarette manufacturers and focused on the aggressive pricing of newly introduced discount cigarettes. In its opinion,17 the Court adopted a two-pronged approach for proving predatory pricing: “A plaintiff must prove (1) that the prices complained of are below an appropriate measure of its rival’s costs and (2) that the competitor had a reasonable prospect of recouping its investment in below cost prices” (p. 210). If a plaintiff demonstrates both prongs, this establishes a rebuttable presumption of predatory intent. In the price-below-cost prong, the Court did not specify which measure of the defendant’s costs is appropriate, but Brooke Group is widely read as an endorsement of the Areeda-Turner test (Baker, 1994). A defendant’s marginal or average variable cost has been adopted as the appropriate cost measure in the First, Second, Fifth, and Sixth federal circuits. Prices above that benchmark but below the firm’s average total cost will not necessarily exonerate a defendant in the Eighth and Ninth circuits, and in the latter, prices above the defendant’s average total cost may still be considered predatory. The Seventh Circuit’s benchmark currently is a defendant’s long-run incremental cost. The Third, Fourth, and Tenth circuits have not specified what they consider an appropriate measure of costs for the purpose of applying the Brooke Group standard (ABA Section of Antitrust Law, 2012, pp. 278ff.).

15  In the US Department of Justice’s lawsuit against American Airlines, the airline successfully defended its practice of dramatically reducing fares on routes served by new entrants. In this case, the court did not accept the government’s inclusion of the opportunity cost of fixed assets (such as airplanes) in its AVC calculations with the result that the airline’s fares remained above AVC. However, the court in a subsequent predatory pricing case against another airline was more receptive to including these opportunity costs. See Elzinga and Mills (2009). 16  The authors were consultants to the defendant in the course of the litigation. See Elzinga and Mills (1994). 17 Also, Brooke Group established a single standard for proving predatory pricing under both the Robinson-Patman Act and Section 2 of the Sherman Act.

50   Kenneth G. Elzinga and David E. Mills The price-below-cost prong is similar to, but more permissive than, a profit-sacrifice test based on Ordover and Willig’s definition of predation (Salop, 2006). Instead of establishing predatory intent by showing that a defendant’s prices constitute a short-run profit sacrifice (i.e., an avoidable reduction in profit), the Brooke Group standard requires a defendant’s prices to be low enough to create an actual short-run loss. Substituting a negative profit test for a profit sacrifice test avoids the difficulty of identifying an alleged predator’s most profitable alternative pricing strategy—an exercise that would bring its own set of problems (Motta, 2004). Although some ambiguity remains about which cost measure is appropriate, Brooke Group’s below-cost pricing requirement creates a less problematic safe harbor for the defendant than a strict profit-sacrifice test. The more striking feature of the Court’s opinion in Brooke Group is the recoupment prong: “The plaintiff must demonstrate that there is a likelihood that the scheme alleged would cause a rise in prices above a competitive level sufficient to compensate for the amounts expended on the predation” (p. 210). Here the Court recognizes that predatory pricing has the characteristics of an investment. The recoupment test assesses the likelihood that the defendant could charge monopoly prices for a long enough period after disposing of the target competitor to recoup its investment in low prices. The prospect of recoupment is not a consideration that materialized in US law with Brooke Group. The ability of the defendant to recoup allegedly predatory losses figured prominently in preceding decisions in Cargill (1986), Matsushita (1986) and A. A. Poultry Farms (1989). In William Inglis (1982), the Ninth Circuit Court of Appeals held that “Predation exists when the justification for . . . [low] . . . prices is based . . . on their tendency to eliminate rivals and create a market structure enabling the seller to recoup his losses. This is the ultimate standard, and not rigid adherence to a particular cost-based rule” (p. 1035). An inquiry about the likelihood of recoupment shifts the focus in predation cases away from the narrow issue of the defendant’s short-run losses to the larger issue of the economic rationality of the firm’s alleged predatory scheme in its entirety. An evaluation of a defendant’s prospects for recoupment involves both the firm’s conduct and structural conditions in a properly defined relevant market: “The determination requires an estimate of the alleged predation’s cost and a close analysis of both the scheme alleged and the relevant market’s structure and conditions” (p. 210). Structural conditions conducive to recoupment are present if the defendant’s market share is large, its rivals are small and limited in their ability to discipline the defendant’s prices, and if there are significant barriers to entry in the market. This inquiry harkens back to the first-tier market power component of Joskow and Klevorick’s two-tier approach. A  necessary condition for successful recoupment is a defendant’s possession of, or a reasonable prospect of acquiring, significant market power. In keeping with the screening role Joskow and Klevorick assigned to the first tier of their approach, the Antitrust Modernization Commission (2007, p. 89) noted that the recoupment test “enhances administrability for the courts by allowing summary disposition of claims where market circumstances—such as easy entry—preclude the possibility of recoupment.”

Predatory Pricing   51

The recoupment prong of the Court’s approach creates a safeguard or “reality check” (Katz, 2006, p.  6) for courts against reaching false positive conclusions, especially because there are legitimate, procompetitive reasons why the price-cost analysis in the first step should not be dispositive.18 The Court’s concern with false positive outcomes in predatory pricing cases was a prominent factor in its earlier Matsushita and Cargill opinions as well as in Brooke Group.19 In European law, an abuse of dominance is an objective notion that stands apart from the defendant’s subjective intent. Accordingly, in the EU it is not necessary to show that a defendant had predatory intent in order to prove that the firm’s low prices constituted an abuse of dominance. The standard for proving an abuse of dominance such as predatory pricing in the EU does not include a recoupment test. Vickers (2008, p. 419) suggests that a separate recoupment test is unnecessary in the EU because “dominance—without which there can be no abuse in European law—implies ability to recoup.” The two-pronged Brooke Group formula, and for that matter the two prongs themselves, are not without their detractors. Depending on one’s priors about how frequently dominant firms employ exclusionary pricing practices, and about how detectable those practices are, the Brooke Group formula is susceptible of both false negative and false positive errors. Because of the infrequency of “positive” findings since Brooke Group, most of the dissatisfaction with the Brooke Group formula has been expressed by the commentators who find the formula, and especially its price-below-cost provision, too permissive. Post-McGee theories of predatory pricing provide a variety of theoretical scenarios in which a defendant might reduce its prices to impair the profitability and threaten the survival of a rival without resorting to prices so low that the defendant runs afoul of a price-below-cost test. In fact some theoretical predatory scenarios do not even require sacrifice, let alone prices below cost. The fact that the Brooke Group formula might excuse a hypothetical defendant engaged in what Edlin (2002) calls “above cost predatory pricing” is a source of concern for economists and antitrust scholars who are more sanguine than the Court about the ease of separating procompetitive from anticompetitive pricing. Recognizing that the competitive discipline brought to bear against a dominant firm by even a less-efficient rival may create consumer benefits,20 these critics of Brooke Group favor the use of more stringent rules or elaborate rule-of-reason analyses to reduce such firms’ exposure to the aggressive pricing of a dominant firm. But it would not be a trivial matter to ferret out by means of a 18  Areeda and Turner discuss temporary “promotional pricing” and “meeting competition” as two examples. Learning effects, whereby costs fall over time with greater experience and cumulative output, are another circumstance where prices below an otherwise appropriate measure of cost in the short run would not necessarily be predatory. 19 In Matsushita the Court held that “we must be concerned lest a rule or precedent that authorizes a search for a particular type of undesirable pricing behavior end up by discouraging legitimate price competition” (p. 594). 20  For example, a duopoly with two firms—one larger and somewhat more efficient than the other— will charge consumers a lower price than the large, efficient firm would if it were a monopolist. In this instance, consumer welfare would be greater if the small, inefficient firm remains in the market. Total welfare could be greater or less if the inefficient firm remains.

52   Kenneth G. Elzinga and David E. Mills rule-of-reason analysis which vulnerable, inefficient, or disgruntled rivals should be preserved without putting spirited head-to-head price competition at risk. Elhauge, while recognizing the potential for false positive outcomes, takes the position that “efforts to restrict above-cost reactive price cuts are likely to be futile and harmful” because many such price cuts are procompetitive and distinguishing the exceptions would be difficult (2003, p. 826). Other commentators are more optimistic about the ability of the courts to undertake analyses of a predatory pricing defendant’s prices applying methods that are more elaborate and nuanced than the Areeda-Turner test. Salop, for instance, proposes that courts should deemphasize formulaic price-cost comparisons and apply a “consumer welfare effect standard” to “evaluate whether the consumer benefits from lower prices achieved during the predatory period outweigh the likely harm during the recoupment period. . . . [This] standard could be more or less permissive than . . . Brooke Group . . . depending upon the exact facts” (2006, p. 337–38). The informational requirements of the kind of inquiry Salop proposes could be demanding, and the result is unlikely to be as predictable as applying the Brooke Group standard. The Court’s approach in Brooke Group is not a blueprint for detecting every kind of strategy with exclusionary designs and effects. Identifying exclusionary behavior without jeopardizing a broad spectrum of procompetitive conduct that is unwelcome to rivals is a formidable undertaking. In its review of antitrust enforcement in the United States several years ago, the Antitrust Modernization Commission concluded that evaluating firm conduct under Section 2 of the Sherman Act “poses among the most difficult questions in antitrust law. Appropriate antitrust enforcement must distinguish aggressive competition that benefits consumers, such as most price discounting, from conduct that tends to destroy competition itself, and thus maintains, or facilitates acquiring, monopoly power” (2007, p. 81). Given the inherent difficulty of discerning in an adversarial proceeding the intent that animates a firm’s conduct, and given the cost of deterring conduct that ultimately is procompetitive in an overly zealous attempt to protect vulnerable competitors and the status quo, the judicial modesty and restraint of the Brooke Group formula is warranted. While it is true that positive findings of predation since Brooke Group are few, a plausible explanation could be that the Court’s establishing a clear, administrable standard deters predatory pricing in the same way that radar-enforced speed limits deter speeding. Bolton, Brodley, and Riordan have another explanation as to why “no predatory pricing plaintiff has prevailed on the merits in the federal courts” since Brooke Group: “ the judicial belief that predatory pricing is extremely rare, if not economically irrational conduct” (2000, p. 2241–42). They maintain that this belief is in tension with the “the consensus view in modern economics that predatory pricing can be a successful and fully rational business strategy.”21 21  Milgrom and Roberts observed in the early 1980s that “a large fraction of the economics profession would argue that . . . predation is an irrational strategy for attempting to gain or maintain a monopoly position and that it is, therefore, unlikely to be adopted in practice” (1982, p. 280). Bolton, Brodley, and Riordan’s assessment of economists’ views twenty years after Milgrom and Roberts’s reflects the influence of the predatory pricing theories that appeared during the intervening years.

Predatory Pricing   53

It is one thing for economists, or even the courts, to acknowledge that predatory pricing is a theoretical possibility in certain circumstances. But until the courts are confronted with unambiguous evidence in actual cases, there is little reason for this acknowledgment to reduce judicial skepticism about the likelihood of successful predatory pricing.22 The courts have found multimarket predatory pricing theories that involve acquiring an intimidating reputation to be particularly unpersuasive: “The courts have to date rejected efforts to establish recoupment of predatory losses in one group of markets by demonstrating that the predation will facilitate the exercise of monopoly power in other markets” (ABA Section of Antitrust Law, 2012, p.  285). Easterbrook (1981) makes the observation that a strong piece of evidence supporting the courts’ presumption that predatory pricing is rare is because so few plaintiffs have prevailed under any theory of predatory pricing. The Brooke Group formula does not prevent a plaintiff from demonstrating that a defendant’s conduct fits the mold of a reputation-based predatory pricing theory, or any other theory that involves below-cost pricing. It is just that the plaintiff must assume the burden of proving such theories if the defendant passes both prongs of Brooke Group. Some commentators recognize so many difficulties with the enterprise of detecting predatory pricing that they pose the impish question: whether the entire enterprise should be abandoned? Writing in this vein, Milgrom and Roberts question whether “it may be best simply to give up on attempts to control predation, even if one believes that it can and does occur” (1982, p. 134). Hovenkamp speculates (without agreeing) that “the world would very likely be better off with no law of predatory pricing at all” (2005, p. 160–61). There is nothing impish about Easterbrook’s conclusion: “If there is any room in antitrust law for rules of per se legality, one should be created to encompass predatory conduct. The antitrust offense of predation should be forgotten” (1981, p. 337; footnote omitted).

2.5. Implementing Brooke Group Competing price-cost comparisons have been the focal point of many predatory pricing cases in the years since the Areeda-Turner article appeared. The analysis of prices and costs in these disputes often is complex and highly contextual. Applications of the recoupment prong generally are not as complex as price-cost comparisons, but they are equally fact-specific. Although there are idiosyncratic considerations to applying both prongs of the Brooke Group standard in any market, there are some principles that apply generally. 22  Bolton, Brodley, and Riordan (2000, 2001) believe that several prominent predatory pricing plaintiffs present fact situations that are similar to those of recognized economic theories of predatory pricing, but see Elzinga and Mills (2001) for a dissenting view.

54   Kenneth G. Elzinga and David E. Mills

2.5.1.  Price-Cost Comparisons Implementing the price-cost comparison prong of Brooke Group requires an answer to two fundamental questions: Which prices (e.g., which sales?) and which costs? Litigants often dispute the scope of a defendant’s products and customers over which the firm’s prices and costs should be compared. A plaintiff may argue that the proper test involves only those sales of the defendant—those specific products and customers—that the plaintiff actively contested. Under this definition of the proper scope, a defendant’s prices and costs due to sales of the same product to different customers, or of similar products to the same or different customers, are irrelevant to the analysis. The proper scope of sales for a price-cost test may, in some instances, be narrow, but narrowing the scope of sales to match the extent of the plaintiff ’s endeavors as a matter of principle is inappropriate. The reasoning behind this contention verges on equating predation with the drawing away of a current or prospective customer of the plaintiff by means of price competition. Where there are as-yet-unexploited prospective sales of the same or similar products within the reach of a plaintiff—sales that are not off limits because of formidable barriers—a predator reasonably could not expect to drive a competitor out of business without extending its below-cost pricing to these prospective sales. Predatory pricing is not a practice carried out on a transaction-bytransaction basis. Courts usually have approved extending the scope of sales to the entire relevant product and geographic market that contains the contested sales. This is because predatory pricing in one part of the relevant market leaves a refuge from low prices in other parts (Denger and Herfort, 1994, p. 551). If one purpose of a price-cost test is to detect whether the defendant intends to make it unprofitable for a competitor to remain anywhere in the market, the test should be applied everywhere in the market. Consequently, comparing prices and costs throughout the relevant market is consistent with the goal of detecting predatory intent. Similarly, if the other purpose of a price-cost test is to detect whether the defendant intends to loses money on its sales in the relevant market, the test should be applied market-wide. The procompetitive reasons why a defendant might be willing to lose money in the short run on sales of a narrow set of products to a narrow set of customers, such as promotional pricing or meeting competition, are more numerous and more plausible than reasons why the firm would be willing to lose money on sales throughout the relevant market. Calculating an appropriate measure of a defendant’s costs is often the most nuanced element of a Brooke Group price-cost comparison. In designating their cost benchmark as AVC, Areeda and Turner drew on the distinction between fixed and variable costs as described in economics textbooks. While this is a clear and unambiguous distinction in the textbooks, the distinction is not always easy to operationalize when implementing the Areeda-Turner test in real markets. The circumstances and time frame in which a cost is variable rather than fixed are different for different costs. The costs that are “variable” with respect to a very small increment of output (e.g., labor and

Predatory Pricing   55

materials) are not necessarily the same as the costs that are “variable” with respect to a larger increment of output (e.g., an additional production line). And the costs that are incremental with respect to an increase in output are not the same as the costs that are decremental with respect to a decrease in output (e.g., an advertising campaign when opening a new sales territory). Baumol (1996) made the relevant distinction clearer when he substituted the twin notions of incremental and avoidable costs for variable costs. The relevant costs for implementing a price-cost test depends on which question the test is meant to answer. While most commentators interpret such tests to emphasize whether the defendant lost money because it charged such low prices on the contested sales, Baumol puts more emphasis on the second question: whether the defendant’s prices were so low that it would be unremunerative for an equally efficient competitor to match. If the question is whether the defendant lost money on the contested sales, Baumol argued that the test should compare the defendant’s average prices on those sales to the average incremental costs (AIC) the defendant incurs to make those sales. These are the costs that the defendant must incur to increase output when it captures the contested sales. If the question is whether it would be unremunerative for the plaintiff to match the defendant’s prices, Baumol argued, the test should examine decremental rather than incremental costs. Specifically, the relevant test compares the defendant’s average prices to the plaintiff ’s average avoidable costs (AAC)—those costs the plaintiff could avert by forfeiting the contested sales. These two cost measures are not always the same even where the plaintiff and the defendant are equally efficient (when similarly positioned with respect to the contested sales). For instance, the costs a plaintiff could avoid by surrendering certain sales that it made in the past, or would have made but for the defendant’s challenged pricing, are not necessarily the same as the costs the defendant would incur by increasing its output to take away those sales. Baumol reasons that certain costs incurred by a firm when it expands output are, like a ratchet, not as quickly and easily averted when the firm reduces output. This means that AIC usually is greater than AAC. Thus, if a defendant passes a test that uses its AIC, the firm would almost certainly pass a test based on AAC. Several economists have stressed that the proper price-cost comparison must calculate costs in ways that are consistent with the opportunity cost principle. When a multimarket defendant cuts prices in one market, allegedly to drive a competitor out of that market, it may deploy assets whose costs are fixed in character to increase output. If those assets otherwise would be deployed in different markets, their cost should be treated as incremental in the relevant market because of the opportunity cost principle. For instance, an incumbent airline that deploys a plane and crew in one (city-pair) market to increase the number of passengers it can carry in conjunction with a steep fare reduction has alternative uses for the plane and crew in other markets. This means that the relevant measure of AIC should include these costs. The time span a plaintiff proposes for comparing a defendant’s prices and costs is another factor that comes into play in a Brooke Group price-cost comparison. When the low prices that form the basis of a plaintiff ’s complaint are of short duration, they may

56   Kenneth G. Elzinga and David E. Mills not last long enough to credibly threaten the plaintiff ’s survival. Low prices over a time span that is too short to put the plaintiff ’s survival in jeopardy are an insufficient basis for inferring predatory intent. In addition, a short time span, just like a narrow set of contested sales, increases the number and plausibility of alternative explanations for a defendant’s low prices.

2.5.2. Recoupment The Court’s opinion in Brooke Group held that the prospect of recoupment is a necessary component of a predatory pricing claim: “Evidence of below cost pricing is not alone sufficient to permit an inference of probable recoupment and injury to competition. The determination requires an estimate of the alleged predation’s cost and a close analysis of both the scheme alleged and the relevant market’s structure and conditions” (p. 210). Consequently, under Brooke Group, a plaintiff ’s claim must pass the recoupment test even if the price-cost test creates a rebuttable presumption that the defendant’s prices were predatory. Proof that the defendant, looking forward from the time the firm’s low pricing began, could expect to recoup lost profits once the competitor expired buttresses the presumption established by the price-cost test. However, if a plaintiff proves that the defendant’s prices were below an appropriate measure of costs, but cannot prove that the scheme alleged held out the prospect of recoupment, the defendant is under no obligation to explain its low prices. Although the recoupment requirement in Brooke Group creates an additional hurdle for predatory pricing plaintiffs, this burden is warranted by three considerations. These are the risk of price-cost miscalculations, the potential welfare loss from disrupting legitimate price competition, and the infrequency of low-price episodes that have proved, on close examination, to be predatory. The recoupment test examines whether structural conditions in the market and the defendant’s market power, once the plaintiff is out of the picture, make recoupment plausible.23 There are several structural conditions that would be favorable to recoupment. A low price elasticity of demand is favorable because it limits a predator’s losses when prices are low (e.g., low prices do not elicit a large increase in unprofitable sales), and enhances the firm’s gains when prices subsequently rise. Indications of growing demand are favorable, because recoupment is easier in an expanding market, all else equal. Factors that make the target competitor’s withdrawal more likely or quicker are favorable to recoupment, as are barriers to entry or reentry. A large postpredation market share for the predator and any factors that limit the supply response of the predator’s remaining competitors when prices rise also are favorable. Neither the Court’s Brooke Group opinion nor any subsequent opinions have specified exactly how to assess a defendant’s prospects for recoupment. Hemphill proposed that the recoupment test should be seen as a filter that would weigh relevant structural 23 

See generally Elzinga and Mills (1989).

Predatory Pricing   57

factors, such as those listed above, but also including factors suggested by post-McGee theories of predatory pricing: “information asymmetries and linkages across markets [that] might make recoupment more likely, by allowing a predator to build a reputation for predation and/or send misleading signals about demand and cost in a market” (2001, p. 1589). A structural analysis to evaluate a defendant’s prospects for recoupment is similar to an analysis of the firm’s market power such as was proposed by Joskow and Klevorick, but there is a difference. When assessing the defendant’s prospect for recoupment, the question is not whether the firm possesses market power in some substantial degree. The question is whether conditions in the market are such that the firm reasonably could expect to accomplish specifically what plaintiff ’s predatory pricing claim alleges. In most instances, because such claims are idiosyncratic and fact intensive, this is not a question amenable to a generic answer. The goal of the recoupment test in Brooke Group is to gauge whether the defendant had, at the outset, a “dangerous probability, of recouping its investment in below-cost prices” (p. 224). This calls for estimates and a comparison of the defendant’s likely discounted gains and losses, viewed prospectively.24 Calculating such gains and losses would require evaluating the structural factors mentioned above and making some assumptions about the defendant’s expectations. The proper benchmark for calculating and comparing gains and losses is the profit stream the defendant would expect if the firm followed its most profitable alternative course of action. In most instances this means that the relevant losses are greater, and the gains less, than if the benchmark were a short-run break-even scenario. Comparing the defendant’s expected gains and losses can provide answers to questions that bear directly on the plausibility of recoupment. For example, how quickly must the defendant expect the competitor to capitulate in order to recoup? Or in order to recoup, how much time after the competitor capitulates would the defendant require before entry or reentry is likely? Or, what rate of return would the defendant expect to earn on its investment in predatory prices? In answering these questions, particular attention must be given to how sensitive the answers are to the assumptions made about the defendant’s expectations.

2.6. Conclusion The Brooke Group standard for proving predatory pricing claims sets the bar high to avoid false negative conclusions. A further reason for maintaining a high bar is 24 

In this connection, Denger and Herfort (1994, p. 543) propose that ex post knowledge that the target competitor exited the market during the course of the alleged predation should be irrelevant. Motta (2004, p. 450) proposes that maintaining a prospective point of view precludes a defense based on ex post evidence that the defendant “did not manage to recoup losses.”

58   Kenneth G. Elzinga and David E. Mills that the market power of dominant firms, left alone, tends to erode. If one thinks of US firms with high market shares who, in the past, have been prosecuted under Section 2 of the Sherman Act, it often has been market forces more than antitrust enforcement that marked their decline in prominence: witness American Tobacco, U.S. Steel, Alcoa, A&P, DuPont, and Kodak, among other once purportedly dominant firms. Applying the Brooke Group standard in all predatory pricing cases, rather than adopting a portfolio of tests or launching an open-ended rule-of-reason inquiry, has other virtues. Brooke Group provides criteria that are sufficiently predictable that a prudent incumbent or dominant firm can avoid price cuts that would incur antitrust liability. In addition, the Brooke Group standard is sufficiently administrable to provide meaningful guidance for courts that face the task of deciding whether a dominant firm’s low prices threaten competition—or instead only threaten competitors. Economic theories of predation are differentiated in so many ways that trying to embrace the particulars of every theory in a coherent, predictable, and administrable rule is unworkable. In Barry Wright, a well-known predatory pricing opinion he authored before being elevated to the Supreme Court, Justice Breyer (who once taught antitrust law at Harvard Law School) expressed his concern with trying to embed every theory of predation into the enforcement of the antitrust laws in the following manner: “Rules that seek to embody every economic complexity and qualification may well, through the vagaries of administration, prove counterproductive, undercutting the very economic ends they seek to serve” (p. 234).

References ABA Section of Antitrust Law. 2012. Antitrust Law Developments. 7th ed. Chicago: American Bar Association. Antitrust Modernization Commission. 2007. Report and Recommendations. Viewed at http://​ govinfo.library.unt.edu/amc/report_recommendation/toc.htm. Areeda, Philip, and Donald F. Turner. 1975. Predatory Pricing and Related Practices under Section 2 of the Sherman Act. Harvard Law Review 88:697–733. Baker, Jonathan B. 1994. Predatory Pricing after Brooke Group:  An Economic Perspective. Antitrust Law Journal 62:585–604. Baumol, William J. 1979. Quasi-Permanence of Price Reductions:  A  Policy for Prevention of Predatory Pricing. Yale Law Journal 89:1–26. Baumol, William J. 1996. Predation and the Logic of the Average Variable Cost Test. Journal of Law & Economics 39:49–72. Bolton, Patrick, Joseph F. Brodley, and Michael H. Riordan. 2000. Predatory Pricing: Strategic Theory and Legal Policy. Georgetown Law Journal 88: 2239–330. Bolton, Patrick, Joseph F. Brodley, and Michael H. Riordan. 2001. Predatory Pricing: Response to Critique and Further Elaboration. Georgetown Law Journal 89:2495–530. Bolton, Patrick, and David S. Scharfstein. 1990. A Theory of Predation Based on Agency Problems in Financial Contracting. American Economic Review 80:93–106.

Predatory Pricing   59

Burns, Malcolm R. 1986. Predatory Pricing and the Acquisition Costs of Competitors. Journal of Political Economy 94:266–96. Clark, John M. 1926. The Social Control of Business. New York: McGraw-Hill. Dalton, James A., and Louis Esposito. 2011. Standard Oil and Predatory Pricing: Myth Paralleling Fact. Review of Industrial Organization 38:245–66. Denger, Michael L., and John A. Herfort. 1994. Predatory Pricing Claims after Brooke Group. Antitrust Law Journal 62:541–58. Easterbrook, Frank H. 1981. Predatory Strategies and Counterstrategies. University of Chicago Law Review 48:263–337. Edlin, Aaron S. 2002. Stopping Above-Cost Predatory Pricing. Yale Law Journal 111:941–91. Edwards, Corwin D. 1955. Conglomerate Bigness as a Source of Power. In National Bureau of Economic Research, Business Concentration and Price Policy. Princeton, NJ:  Princeton University Press. Elzinga, Kenneth G. 1970. Predatory Pricing: The Case of the Gunpowder Trust. Journal of Law & Economics 13:223–40. Elzinga, Kenneth G., and David E. Mills. 1989. Testing for Predation: Is Recoupment Feasible? Antitrust Bulletin 34:869–93. Elzinga, Kenneth G., and David E. Mills. 1994. Trumping the Areeda-Turner Test: The Recoupment Standard in Brooke Group. Antitrust Law Journal 62:559–84. Elzinga, Kenneth G., and David E.  Mills. 2001. Predatory Pricing and Strategic Theory. Georgetown Law Journal 89:2475–94. Elzinga, Kenneth G., and David E. Mills. 2009. Predatory Pricing in the Airline Industry: Spirit Airlines v. Northwest Airlines. In John E. Kwoka, Jr. and Lawrence J. White, The Antitrust Revolution. 5th ed. New York: Oxford University Press. Fudenberg, Drew and Jean Tirole. 1986. A Signal-Jamming Theory of Predation. Rand Journal of Economics 17:173–90. Genesove, David, and Wallace P. Mullin. 2006. Predation and Its Rate of Return: The Sugar Industry, 1887–1914. Rand Journal of Economics 37:47–69. Giddings, F. H. 1887. The Persistence of Competition. Political Science Quarterly 2:62–78. Hemphill, C. Scott. 2001. The Role of Recoupment in Predatory Pricing Analyses. Stanford Law Review 53:1581–612. Hovenkamp, Herbert. 2005. The Antitrust Enterprise. Cambridge, MA: Harvard University Press. Joskow, Paul, L. and Alvin K. Klevorick. 1979. A Framework for Analyzing Predatory Pricing Policy. Yale Law Journal 89:213–70. Katz, Michael. 2006. The Current State of Economics Underlying Section 2: Comments of Michael Katz. Viewed at www.antitrustsource.com. Koller, Roland H. 1971. The Myth of Predatory Pricing: An Empirical Study. Antitrust Law and Economics Review 4:105–23. Kreps, David, and Robert Wilson. 1982. Reputation and Imperfect Information. Journal of Economic Theory 27:253–79. Leeman, Wayne A. 1956. The Limitations of Local Price-cutting as a Barrier to Entry. Journal of Political Economy 64:329–34. Lott, John. 1990. Predation by Public Enterprises. Journal of Public Economics 43:237–51. McGee, John S. 1958. Predatory Price Cutting: the Standard Oil (NJ) Case. Journal of Law & Economics 1:137–69. Milgrom, Paul, and John Roberts. 1982. Predation, Reputation and Entry Deterrence. Journal of Economic Theory 27:280–312.

60   Kenneth G. Elzinga and David E. Mills Milgrom, Paul, and John Roberts. 1990. New Theories of Predatory Pricing. In Giacomo Bonanno and Dario Brandolini, Industrial Structure in the New Industrial Economics. Oxford: Clarendon Press. Motta, Massimo. 2004. Competition Policy:  Theory and Practice. Cambridge:  Cambridge University Press. Ordover, Janusz A., and Garth Saloner. 1989. Predation, Monopolization and Antitrust. In Richard Schmalensee and Robert D. Willig, Handbook of Industrial Organization, vol. 1. Amsterdam: North-Holland. Ordover, Janusz A., and Robert D. Willig. 1981. An Economic Definition of Predation: Pricing and Product Innovation. Yale Law Journal 91:8–53. Phlips, Louis. 1995. Competition Policy: A Game-Theoretic Perspective. Cambridge: Cambridge University Press. Posner, Richard A. 2001. Antitrust Law. 2nd ed. Chicago: University of Chicago Press. Salop, Steven C. 2006. Exclusionary Conduct, Effect on Consumers, and the Flawed Profit-Sacrifice Standard. Antitrust Law Journal 73:311–74. Salop, Steven C., and Lawrence J. White. 1988. Private Antitrust Litigation: An Introduction and Framework. In Lawrence J. White, Private Antitrust Litigation: New Evidence, New Learning. Cambridge, MA: MIT Press. Sappington, David E. M., and Gregory Sidak. 2003. Incentives for Anticompetitive Behavior by Public Enterprises. Review of Industrial Organization 22:183–206. Scherer, F. M. 1976. Predatory Pricing and the Sherman Act: A Comment. Harvard Law Review 89:869–903. Selten, Reinhard. 1978. The Chain Store Paradox. Theory and Decision 9:127–59. Sokol, D.  Daniel. 2009. Competition Policy and Comparative Corporate Governance of State-Owned Enterprises. Brigham Young University Law Review 2009:1713–812. Tarbell, Ida M. 1904. The History of the Standard Oil Company. New York: McClure, Phillips & Co. Telser, Lester G. 1966. Cutthroat Competition and the Long Purse. Journal of Law & Economics 9:259–77. Vickers, John. 2008. Abuse of Market Power. In Paolo Buccirossi, Handbook of Antitrust Economics. Cambridge, MA: MIT Press. Williamson, Oliver E. 1977. Predatory Pricing: A Strategic and Welfare Analysis. Yale Law Journal 87:284–340. Yamey, Basil. 1972. Predatory Price Cutting: Notes and Comments. Journal of Law & Economics 15:129–42.

Cases Cited A. A. Poultry Farms, Inc., v. Rose Acre Farms, Inc., 881 F.2d 1396 (7th Cir 1989). Brooke Group Ltd. v. Brown & Williamson Tobacco Corporation, 509 U.S. 209 (1993). Cargill, Inc. v. Monfort of Colo., Inc., 479 U.S. 104 (1986). Matsushita Electric Industrial Co., Ltd, et al. v. Zenith Radio Corp. et al. 475 U.S. 574 (1986). Mogul Steamship Co Ltd v McGregor, Gow & Co [1892] AC 25. Standard Oil Co. of New Jersey v. United States, 221 U.S. 1 (1911). Tetra Pak International SA v. Commission [1996] ECR I-1365, Case C-333/96P.

Predatory Pricing   61

United States v. American Tobacco Company, 221 U.S. 106 (1911). United States v. AMR Corp., 335 F.3d 1109 (10th Cir. 2003). United States v. New York Great Atlantic & Pacific Tea Co., 67 F.Supp. 626 (E.D. Ill., September 21, 1946). Utah Pie Company v. Continental Baking company et al. 386 U.S. 685 (1967). William Inglis v. ITT Continental Baking Co., 688 F.2d 1014 (9th Cir. 1982).

CHAPTER 3

R A I SI N G R I VA L S’ C O ST S DAVID T. SCHEFFMAN AND RICHARD S. HIGGINS

3.1.  Overview of Theory The theory of “raising rivals’ costs” (RRC) deals with actions that might be taken by a firm with market power (“dominant firm”) to harm its rivals even if those actions may also harm the dominant firm.1 The sorts of actions dealt with in the theory are ones that, in various ways, raise the costs of rivals and of the dominant firm. Originally, a focus of the theory was a type of “vertical squeeze” theory, that is, theories in which the dominant firm and its rivals purchased an input, and the dominant firm “overpurchased” that input in order to raise the price of the input to its rivals (Scheffman and Higgins 2003). By doing so, however, the dominant firm also raised its own costs by purchasing at higher prices than without the “overpurchasing.” The fundamental insight of the theory is that increases in rivals’ marginal costs will lead the rivals to reduce their output relative to an initial equilibrium level. Other things equal, that will cause their prices to rise, which, alone, is a benefit to the dominant firm. However, recall that in this theory, the dominant firm also experiences an increase in its marginal costs, which, other things equal, reduces its profits. The net effect on the dominant firm’s profits is the combination of lower output / higher prices of rivals and the higher input costs of the dominant firm. It is straightforward to derive a condition for which raising rivals’ costs is profitable for the dominant firm. Suppose the cost-raising strategy involves raising the price 1  The original unique contribution of the original RRC papers was that they showed that it could pay for the dominant firm to take actions that raised the costs of its rivals, even though, as a consequence, the dominant firm raised its own costs.

Raising Rivals’ Costs   63

of an input X that is used by both the rivals and the dominant firm. Denote the price PX. Suppose further that all competitors produce an identical homogeneous product. Finally, assume that the rivals act as perfect competitors. The increase in the rivals’ costs arising from the increase in PX causes them to restrict their output at the original output price, PQ. That results in an increase in PQ, keeping sales of the dominant firm constant (lower total output moves up the industry demand curve and raises the output price). The actual amount of increase in PQ depends on the amount of increase in PX, the amount by which the rivals’ marginal costs increase due to the increase in PX, the resulting amount by which rivals restrict their outputs, and the effect on output price, PQ, which is determined by the demand. Recall that in this example, since the rivals are assumed to act as perfect competitors, they take prices as given, and produce where their marginal costs equal the market price. Suppose, for example, that the rivals’ only cost of production is the use of X, and they need one unit of X to produce one unit of Q for any level of output. This means that if we assume that the dominant firm output is kept fixed, the market price will increase by the amount of the increase in the rivals’ (aggregate) marginal costs. Then the increase in their marginal costs is equal to the increase in PX, and that is also the amount by which PQ will increase, keeping the dominant firm’s output constant. If the dominant firm uses less than one unit of X to produce one unit of Q, the dominant firm’s average cost (and marginal cost) goes up by less than the change in PX. In this situation the dominant firm’s profits have gone up—for constant output for the dominant firm, price goes up more than the dominant firm’s average cost. Because in this example we have assumed that the fringe’s marginal costs are constant for different levels of output, the change in the fringe’s marginal cost equals the change in the market price of Q. That greatly simplifies things. Consider all the moving pieces in a more complex situation. First, to increase the price of X, the dominant firm must purchase more X. Therefore, one relevant factor is the quantitative relationship between purchases of X and the price of X—which determines how much the dominant firm’s purchases of X must increase to lead to a given increase in PX for rivals. The second relevant factor is how an increase in PX affects the marginal costs of the fringe. The final factor is how an increase in the fringe’s marginal costs affects fringe total sales (or, in the case of differentiated products, their prices)— and therefore the size of the movement up the demand curve for Q (or, in the case of differentiated products, their prices). With differentiated products, another factor is the effects of rivals’ increases in prices on the dominant firm. The general condition that is sufficient (but not necessary) for the dominant firm to gain from raising PX is that the resulting increase in price PQ (dominant firm’s output constant) is larger than the increase in the dominant firm’s average costs (Salop and Scheffman 1983). This condition holds for quite general models of competition and actions by which the dominant firm can increase its rivals’ costs, while at the same time increasing its own costs.

64   David T. Scheffman and Richard S. Higgins

3.1.1.  The Theory Is More General Than the Model of “Overpurchasing” an Input Although the modeling can be more complex, the raising rivals’ costs theory covers any situation in which a dominant firm takes actions that increase the costs of rivals. For example, we would view the basic Microsoft case as one in which Microsoft was alleged to have taken actions that had the purpose and effect of making it difficult for various rivals to compete or potentially compete in the future, and that those actions had no significant precompetitive benefits. As will be discussed below, one common application of RRC theory is in the context of vertical mergers.

3.2.  Demand Effects and Efficiencies Care must be taken in assessing the potential net effects of actions that increase costs to take into account the totality of the costs and benefits of the increase in costs. For example, if a firm incurs costs to improve its product and in order to compete effectively rivals must also incur costs to improve their products, quality-adjusted output rises. That, of course, is competition—not anticompetitive cost raising. Similarly, if a firm takes actions that improve its production processes, resulting in lowering its costs relative to its rivals, that generally would be competitive.

3.3. Counterstrategies Another factor that must be taken into account is that rivals my engage in counterstrategies that would thwart otherwise potential anticompetitive cost-increasing strategies. Suppose, for example, the dominant firm incurs costs to lobby for regulatory changes that would disproportionately affect its rivals—leading to the rivals raising their prices to the benefit of the dominant firm. In response, the rivals may instead incur costs for lobbying to thwart those regulatory changes. Thus, in assessing potential cost-raising strategies by a dominant firm against its rivals, we must assess the ability of rivals to thwart those efforts.

Raising Rivals’ Costs   65

3.4. Assessment It probably cannot be stressed enough that raising rivals’ costs or exclusion is not necessarily anticompetitive. In concentrated markets, competition is necessarily rivalry—that is, competitors are well aware that their rivals’ actions affect them and vice versa. Particularly in mature industries, significant increases in sales for one firm generally reduce the sales of rivals. Price changes by one firm also have effects on rivals. Thus in concentrated markets, particularly mature ones, competition is “about” taking actions that directly or indirectly take sales from rivals. This is the case in markets in which no firm has significant market power. As stated by Krattenmaker and Salop, “A firm that raises its rivals’ costs has not necessarily gained anything. It may have harmed one or more of its competitors, but has it harmed competition? Competition is harmed only if the firm purchasing the exclusionary right can, as a result, raise its price above the competitive level” (1986, 262). In fact, much of “competition on the merits” in concentrated industries involves strategies and tactics that in some way disadvantage rivals, but is not anticompetitive (Holt and Scheffman 1989).

3.4.1.  Strengths of the RRC Analysis The RRC framework has a number of virtues. The analysis is pretty straightforward, so that you do not need to be an economic theorist to grasp the basic logic. In some senses the results are theoretically powerful. RRC makes clear that cost-raising and exclusionary strategies are generally, if not always, going to be superior (for an instigating dominant firm) to predatory pricing or other strategies that require recoupment, since a RRC strategy will often be profitable nearly from the outset. Put differently, the RRC analyses (and the literature on predatory pricing) make clear that cost-raising and exclusionary strategies should be the predominant antitrust concern about a dominant firm’s behavior. A further strength, at least as a matter of theory, of some of the RRC literature, particularly Salop and Scheffman (1987), Salop, Scheffman, and Schwartz (1984), and Ordover and Saloner (1989), is that empirically “testable” conditions are derived. These conditions are probably too cryptic for most lawyers, but they are conditions amenable, at least in principle, to application by economists in specific fact situations.

3.4.2.  Limitations of the RRC Analysis Both the original RRC article and the later articles2 make clear that as a matter of economic theory, the net welfare effects of cost-raising strategies are ambiguous. That is, 2 

Particularly Salop and Scheffman (1987).

66   David T. Scheffman and Richard S. Higgins that a dominant firm engages in cost-raising strategies, does not, itself prove, as a matter of economic theory, that such strategies are anticompetitive—that is, harm consumers. It is important to understand this theoretical ambiguity, particularly in light of the many papers written since the RRC paper that focus on the potential for vertical mergers to be anticompetitive (a topic we will discuss in more detail below). The analyses in the RRC papers largely focus on a situation with a dominant firm that is assumed to have significant market power, independent of any cost-raising strategies. Although it appears pretty simple, the model is actually quite complex in generating general results, as are all general models that involve market power short of monopoly. Thus, as is pointed out clearly in the Salop and Scheffman (1987) paper, as a matter of theory, cost-raising strategies by a dominant firm may raise or lower prices, raise or lower total welfare, and even raise or lower the profits of the “victims.” The ambiguity arises from a number of sources. The most straightforward reason is that in the models, the dominant firm prices according to the elasticity of demand that it faces. A cost-raising strategy shifts out the demand faced by the dominant firm, but it is possible that it also makes the demand more elastic—sufficiently more elastic that the profit maximizing price falls. In addition, by restricting its output and raising price, a dominant firm increases the incentives for less efficient competitors to produce more than they otherwise would. Under some circumstances a cost-raising strategy can increase total welfare by shifting a product to a more efficient dominant firm. Again, RRC theory lays out, in principle, testable conditions under which in a specific situation cost-raising strategies are likely, from an economic perspective, to be anticompetitive. A more serious limitation of the RRC analysis is that it does not provide simple, practical guidance on how to distinguish cost-raising strategies from “competition on the merits,” or procompetitive strategies that shift business from rivals.3 As a matter of simple theoretical modeling, in principle this could be tackled, in part, in the RRC models by having the cost-raising strategy also impact market demand and/or the production costs of the dominant firm (to incorporate the possibility the strategy that increases rivals’ costs makes the dominant firm more efficient). Needless to say, such changes greatly increase the ambiguity of the competitive effects of cost-raising strategies. As a matter of policy, conduct by a dominant firm that clearly increases market demand should not be challenged absent a showing that clearly “separable and unnecessary” portions of the conduct were anticompetitive. As we will discuss below, one problem in some monopolization cases is that what at best is highly ambiguous conduct is labeled monopolization, and a case with “bad” conduct can sweep in what at best is highly ambiguous conduct.

3 

This is discussed in some detail by Holt and Scheffman (1989).

Raising Rivals’ Costs   67

3.5.  Policy Implications of RRC Many commentators have been very skeptical about the viability of RRC-type cases. For example, Granitz and Klein (1996, 39) state, “our analysis provides no support for a new antitrust policy which would condemn a vertical relationship without the presence of a horizontal conspiracy.” This leaves unclear their view of vertical relationships for which at least one party has substantial market power and is able to anticompetitively exclude rivals or entrants. Judge Easterbrook states, “My recommendation is that for the foreseeable future we leave raising rivals’ costs to the academy.”4 We are not as skeptical. To begin, we suspect that both Klein and Easterbrook would support going after anticompetitive exclusion through manipulation of governmental or legal process. However, we advocate enlarging the focus of potential governmental or legal process abuse cases. In the modern economy, barriers to entry or to effective competition increasingly do not arise from bricks and mortar, economies of scale, and so on. In the modern economy, the traditional sources of competitive advantage have often been eroded by globalization and technological advances. Competitive advantage increasingly involves intangibles such as intellectual property. Such intangibles are often more amenable to manipulation than are bricks and mortar. Thus, we would argue that the sound policy basis for potential concern with nonprice predation by a dominant firm has increased over time. Certainly, manipulating the government and the patent system is a fruitful area of concern with potentially anticompetitive conduct. Two Federal Trade Commission cases, Rambus5 and Unocal,6 are examples. More difficult to reach are what appear to be anticonsumer (but may be more difficult to reach as anticompetitive) activities of “patent vultures,” and the use of patent thickets, sometimes combined with high stakes (for the defendant) actions at the International Trade Commission. Of course, we do not mean that conduct involving patents should be challenged because patents sometimes “create” market power. Rather, we think it is appropriate to be aggressive about patent “misuse,” involving patents “inappropriately” obtained, and misused, for example, in standard-setting contexts. FTC/DOJ hearings on intellectual property7 highlighted widespread concern with “patent quality,” that is, concern that patents may too frequently be “inappropriately” granted. Of course the solution to this problem, if it exists, lies with the Patent and Trademark Office, and perhaps legislation in intellectual property (IP) law. But in limited circumstances, the antitrust (and perhaps other) laws can, and should, attack anticompetitive use of patents. In recent years the FTC has been very active in pursuing matters involving alleged anticompetitive use of 4 

Easterbrook (2003: 345). See, also Easterbrook (1986, 972). Available at http://www.ftc.gov/os/caselist/d9302.htm. 6  Available at http://www.ftc.gov/os/caselist/d9305.htm. 7  FTC/DOJ hearings on intellectual property: available at http://www.ftc.gov/opp/intellect/index.​ htm. 5 

68   David T. Scheffman and Richard S. Higgins patents. This includes pharmaceutical cases in involving “Hatch-Waxman” issues, using patents in connection standard setting, and patent litigation before the International Trade Commission.8 In any event there is little reason to use many enforcement resources to search for suitable and significant potentially anticompetitive RRC cases. The “beauty” of RRC is that it is likely to leave its fingerprints on a disadvantaged rival, whether or not the conduct is anticompetitive, and disadvantaged rivals are not shy about suing and/or complaining to enforcement agencies. The problem is sorting through what are mostly complaints about competition and competitive advantage to find the relatively few “nuggets” of harm to competition. More empirical research is needed to develop reliable empirical analyses that facilitate the evaluation of the nuggets.9 Of course the great weakness in trying to apply RRC is that there are so many false positives. Competition on the merits often injures rivals and potential rivals. It cannot be stressed enough that allegations of injury to rivals and potential rivals should not “pass Go” unless there is a credible concern that the result is anticompetitive. We agree with Tim Muris that credible anticompetitive effects must be required for any viable RRC-theory case (Muris 2000). This is another area in which economic research could be helpful.

3.6.  Vertical Mergers and RRC One of the major areas of RRC application is vertical mergers.10 The context is competitor A purchases supplier X that was supplying B, a competitor of A. If A has significant market power, there is a possibility that the merger will be a vehicle of raising B’s cost, à la RRC. However, the RRC analysis is considerably more complicated because of the vertical merger. This is because a vertical merger may entail significant efficiencies. One such efficiency may be avoiding “double marginalization”—which occurs when X was selling at a price above its marginal costs. Other efficiencies can arise from more efficient coordination of various kinds between X and A. Thus, for example, A’s acquisition of X may lead to increase in X’s price to B, but lower A’s costs sufficiently so that A’s price falls. The complexity of the analysis of vertical mergers is illustrated here by recounting DGComp’s analysis of the vertical merger of Tom Tom and TeleAtlas.11 8 

See, for example, http://online.wsj.com/article/BT-CO-20120531-712251.html. Easterbrook (2003) stresses the inherent difficulties in evaluating situations that appear to be procompetitive in the short run but potentially anticompetitive in the longer run. We agree that we are a long way from having any research that could provide significant general assistance in complex cases involving product innovation, etc. 10  A fuller treatment can be found in Higgins and Perelman (2013). 11  The US antitrust agencies have also evaluated the competitive effects of several vertical mergers recently, most notably, the acquisition of Double Click by Google. However, the EU provides more transparency into its merger investigations and, thus, we have a more informed basis for applying the theory of RRC to vertical merger in Europe. 9 

Raising Rivals’ Costs   69

The EU Non-Horizontal Merger Guidelines express concern that vertical merger will lead to foreclosure, in particular, in our example, input foreclosure. At the same time, the Guidelines recognize the potential efficiencies associated with elimination of double marginalization. In TomTom/TeleAtlas, the European Commission undertook an econometric analysis of the up- and downstream markets before permitting the combination.12 Before the acquisition, TeleAtlas and NavTeq competed as duopolists in the market for digital maps, which were sold downstream to Tom Tom and to Nokia, two prominent rival manufacturers of portable navigation devices (“PNDs”). Vertical foreclosure would be of concern if, after merger, TeleAtlas had significant incentives to refuse to sell digital maps externally. The Commission correctly noted that foreclosure would be profitable if the upstream loss incurred by TeleAtlas by refusing to deal with nonaffiliated PNDs were more than offset by the increased margin Tom Tom realized on the sale of PNDs. Ultimately, the Commission’s economists estimated that TeleAtlas’s input foreclosure would have to increase the price of digital maps by as much as 200% to be profitable, which price increase would most certainly have elicited competitive entry.13 The analysis proceeded as follows: econometric and quantitative estimates were developed for downstream own- and cross-price demand elasticities, upstream and downstream contribution margins, market shares, and the cost of digital maps as a percentage of the unit cost of supplying PND services. The hypothetical withdrawal of TeleAtlas’s merchant market sales would cause the merged entity to lose the margin on these sales and, at the same time, cause the price of digital maps to rise, since NavTeq would be the sole remaining merchant market supplier. The magnitude of the merchant market input price increase depends on the derived demand for the input from Nokia and possibly others, and its elasticity. Based on this merchant market input price increase, Nokia’s PND price will rise to an extent that depends on the relative importance of digital maps as an input into PND services. The higher price of Nokia PNDs will cause an increase in the demand for Tom Tom PNDs the magnitude of which depends on the diversion ratio from Nokia PNDs to Tom Tom PNDs—that is, on the cross elasticity of downstream demand. Finally, Tom Tom’s contribution margin—assuming marginal cost pricing of digital maps—when multiplied by the estimated increase in unit sales provides a measure of the expected gains from foreclosure. The Commission apparently found that the downstream gain to TomTom/TeleAtlas from RRC would just equal the cost of withdrawing from the merchant input market only if foreclosure would cause the price of digital maps to increase an extraordinary amount. The implausibility of a nontransitory price increase of this magnitude apparently convinced the Commission not to object to the proposed merger. According to 12  The description of the antitrust analysis of the TomTom/TeleAtlas acquisition is based on the paper by Rafael de Coninck (2008). 13  In large part this is because Tom Tom’s closest rival, Garmin, purchased digital maps from NavTeq subject to a long-term contract that prevented an input price increase for one of the major PND manufacturers.

70   David T. Scheffman and Richard S. Higgins Coninck, the critical factor was the relative insignificance of digital maps as an input to the supply of PND services.14 If foreclosure had been found to be likely, the Commission would have had the additional task of determining the consumer welfare effects of the vertical merger. As indicated above in the theory section, enhanced competition may accompany foreclosure, including complete foreclosure. According to the account in Coninck, the Commission was spared this burden.15 Were the additional analysis necessary, the first-order effects cited above could be used to predict the overall output effects of the proposed vertical merger. In light of the second-order and feedback effects associated with RRC, a full merger simulation may be necessary to predict accurately the welfare effects of vertical merger.

3.7. Conclusion RRC is now a widely accepted mode of analysis for addressing allegations of a dominant firm taking actions that may injure, anticompetitively, its rivals. Proper use of RRC requires demonstration of the alleged dominance (substantial market power within a properly defined relevant market) and economic RRC analysis based on the relevant facts.

References Coninck, Rafael de. 2008. Economic Analysis in Vertical Mergers. Competition Policy Newsletter 3: 48–52. Easterbrook, Frank H. 1986. On Identifying Exclusionary Conduct. Notre Dame Law Review 61: 972–980. Easterbrook, Frank H. 2003. When Is It Worthwhile to Use the Courts to Search for Exclusionary Conduct? Columbia Business Law Review 345–358. Granitz, Elizabeth, and Benjamin Klein. 1996. Monopolization by “Raising Rivals’ Costs”: The Standard Oil Case. Journal of Law & Economics 39: 1–47. Higgins, Richard and M. Perelman. 2013. Vertical Merger. In William Shughart and Christopher Thomas, Oxford Handbook of Managerial Economics. New York: Oxford University Press 424–448.

14 

This fact seems unlikely on its face. The Commission’s decision may have had more to do with the fact that during the course of analyzing the merger of Tom Tom and TeleAtlas, NavTeq and Nokia also proposed to merge, together with the protection from elevated digital map prices enjoyed by Garmin due to its long-run supply contract with NavTeq. Whatever the critical factors involved, both vertical mergers were allowed to proceed without opposition. 15  See Coninck (2008). Apparently, the Commission also considered the likelihood of partial foreclosure and found no significant cause for concern.

Raising Rivals’ Costs   71

Holt, Charles, and David Scheffman. 1989. Strategic Business Behavior and Antitrust. In R. Larner and J. Meehan, Economics and Antitrust Policy. New York: Quorum Books, 39–82. Krattenmaker, Thomas, and Steven Salop. 1986. Anticompetitive Exclusion: Raising Rivals’ Costs to Achieve Power over Price. Yale Law Journal 96: 215–291. Muris, Timothy J. 2000. The FTC and the Law of Monopolization. Antitrust Law Journal 67: 693–723. Ordover, Janusz, and Garth Saloner. 1989. Predation, Monopolization, and Antitrust, in R. Schmalensee and R. Willig, Handbook of Industrial Organization, vol. 1. New York: North Holland, 538–592. Salop, Steven C., and David T.  Scheffman. 1983. Raising Rival’s Costs. American Economic Review 73: 267–271. Salop, Steven C., and David T. Scheffman. 1987. Cost-Raising Strategies. Journal of Industrial Economics 36: 19–34. Salop, Steven C., David T. Scheffman, and W. Schwartz. 1984. A Bidding Analysis of Special Interest Regulation:  Raising Rivals’ Costs in a Rent Seeking Society. In B. Yandle and R. Rogowsky, The Political Economy of Regulation: Private Interests in the Regulatory Process. Washington, DC: Federal Trade Commission, 102–127. Scheffman, David T., and Richard S. Higgins. 2003. 20 Years of Raising Rivals’ Costs. George Mason Law Review 12: 371–387.

CHAPTER 4

P R E DAT O RY BU Y I N G JOHN E. LOPATKA

Monopsony is the negative image of monopoly. Narrowly defined, it is the structural condition in which a well-specified good or service has only one buyer (Blair and Harrison 2010). In a broader sense, monopsony power is the ability to pay less for a product by reducing the quantity purchased,1 and it can be possessed by a single firm or a group of firms acting jointly. Technically, a buyer has monopsony power if it confronts in its purchasing decisions a positively sloped supply curve; by contrast, a buyer in a perfectly competitive market faces a horizontal supply curve at the competitive price. A firm can be a monopsonist, therefore, as the term is commonly used, despite the presence of a fringe of competing buyers.2 A firm can have monopsony power for competitively legitimate reasons. For instance, a firm might become the only substantial employer in a region by foresight or natural attrition. In general, a firm does not violate antitrust law unless it acquires, maintains, or increases monopsony power through anticompetitive means (Bauer and Page 2002, p. 329). Predatory buying—the rough counterpart of predatory pricing—is one method a firm might use to increase its monopsony power, though the term can be understood more broadly. It can mean the anticompetitive practice of increasing the amount of an input purchased and hence the price paid for it either to reduce competition in the input market or to impose costs on competitors in the input market and thereby raise price in the output market. The idea that overbuying can generate profits either by lessening competition in the input market alone or by causing an anticompetitive effect in

1  In unusual situations, a monopsonist may be able to reduce the price paid for a good without reducing the quantity purchased. For instance, if a firm can make a credible threat that it will buy nothing unless suppliers sell to it at the lower price offered, the firm can force suppliers onto their all-or-none supply curve, and they will sell to the monopsonist the amount they would have sold in a competitive market, but at a subcompetitive price (Blair and Harrison 2010). 2  Although a monopsonist could be an ultimate consumer, in most monopsony cases of concern to competition policy the firm is an intermediary, purchasing the monopsonized good as an input or for resale. This discussion will assume that a monopsonist is an intermediary.

Predatory Buying   73

the output market too distinguishes it from predatory pricing, which when successful impairs competition only in one market.

4.1.  The Economics of Monopsony An understanding of predatory buying starts with the economic rudiments of monopsony. If supply of an input is infinitely elastic, no buyer can have monopsony power. But a buyer facing an upward-sloping supply curve recognizes that by increasing its purchases of the input, it will ordinarily have to pay the higher price not only for the last unit purchased but for all other units purchased as well.3 As a result, for every possible quantity above 1, the marginal factor cost, or marginal outlay, of the buyer exceeds the price paid. The monopsonist confronts a marginal factor cost curve that lies above and rises more sharply than the supply curve, which represents the schedule of prices that must be paid to suppliers for different quantities of the input. The relationship between the supply curve and the marginal factor cost curve is fundamentally the same as the relationship between a negatively sloped demand curve and the associated marginal revenue curve. In a competitive input market, a single buyer’s demand represents so small a share of total demand that the buyer cannot affect the price of the input by varying the quantity purchased. The buyer perceives a horizontal supply curve at the competitive price, and the competitive price is determined by the intersection of the market supply curve and the aggregate value of the input for all purchasers in the market. The firm buys the quantity at which market price equals the value of the marginal product. The value of the marginal product, or marginal revenue product, equals the output price multiplied by the marginal physical product, which is the additional output attributable to another unit of input. Because marginal physical product normally declines, the value of the marginal product declines as output increases. The monopsonist, like the perfect competitor, calculates the value of its marginal product based on the demand for its output. But unlike a competitive firm, the monopsonist purchases the quantity of input such that the marginal factor cost equals the value of its marginal product. The price it pays for that quantity is determined by the supply curve. Therefore the monopsonist pays a price that is below the value of its marginal product. Monopsony power can be measured using an adapted form of the Lerner Index (Lerner 1934; Blair and Lopatka 2008). It is the percentage deviation from the competitive result. Thus, a monopsonist’s power (I) equals the value of the marginal product (VMP) minus input price (w) divided by input price I = (VMP − w ) / (w ), all measured at the profit-maximizing price. Equivalently, this index of monopsony power equals the reciprocal of the elasticity of supply.

3 

The exception is a buyer that can price discriminate, a practice that is usually infeasible.

74   John E. Lopatka Price and Cost MFC

a

S b

w1 w2

c

VMP

0

Q2 FIGURE  4.1  Competitive

Q1

Quantity

and Monopsony Conditions

Figure 4.1 illustrates the position of the monopsonist in the input market. A firm in a competitive market would purchase the quantity of input supplied where price equals the value of the marginal product. The firm would purchase Q1 units at a unit price of w1. A monopsonist would restrict purchases to the amount at which marginal factor cost equals the value of the marginal product. It would purchase Q2 units at a price of w2. Monopsony behavior is considered exploitation of input suppliers because the monopsonist pays a price for the input that is less than the value of the marginal product to the monopsonist. In most cases, an exercise of monopsony power causes a distortion in the input market and therefore reduces total welfare relative to the competitive outcome (Blair and Harrison 2010, p. 44).4 In the short run, the lower price paid by the monopsonist for the units it purchases results in a wealth transfer from input suppliers to the buyer. But the input not purchased by the monopsonist results in a deadweight loss. In figure 4.1, the deadweight loss is depicted as area abc. The input is released for use elsewhere in the economy, but alternative uses are less valuable than the use the monopsonist would make of it; both suppliers and the monopsonist lose surplus. Further, in the long run input suppliers may shift investments to the production of less valuable resources. In a simple exercise of monopsony power, therefore, input suppliers suffer a loss and the monopsonist enjoys a gain, for the value of its marginal product exceeds the input price. 4  The exceptions are situations in which the monopsonist engages in perfect price discrimination or forces input suppliers onto the all-or-none supply curve.

Predatory Buying   75

Price and Cost

mfc

Sr

S

w* Ddf

Df 0

Q*f

Q*df

FIGURE  4.2 Dominant

Q*

Quantity

Buyer and Competitive Fringe

When a firm is a dominant buyer competing for the purchase of an input, the analysis is substantially the same (Blair and Lopatka 2008, p. 427). The collective demand of the competitive fringe lies below the value of the marginal product of the dominant buyer, either because the dominant buyer is more efficient than its competitors or because the dominant buyer sells in a market that places a higher value on its output than the value different markets place on the output sold by its competitors. The dominant buyer confronts a residual supply curve that equals market supply minus the demand of the competitive fringe at every output. The residual supply curve implies an associated marginal outlay curve. The dominant buyer maximizes profits by purchasing the quantity supplied at the price where its marginal outlay equals the value of the marginal product. The competitive fringe collectively purchases the quantity determined by fringe demand at that price. The total quantity purchased in the market is greater and the price higher than they would be if a buyer was a pure monopsonist, but quantity and price are lower than if the buyer was a perfect competitor. The monopsony power of a dominant buyer can be measured as the reciprocal of the elasticity of supply of the dominant buyer, which depends on the market elasticity of supply, the elasticity of demand of the fringe, and the dominant buyer’s market share. The welfare and distributional effects of dominant buying behavior are comparable though smaller than in the case of pure monopsony. Figure 4.2 illustrates an input market in which a dominant buyer confronts a fringe of competitors. The demand curve of the dominant firm (Ddf ), or the value of the marginal product, lies above the demand curve of the competitive fringe (Df ). The dominant buyer confronts a residual supply curve (Sr) equal to the market supply (S) minus the

76   John E. Lopatka demand of the competitive fringe. The dominant buyer purchases the quantity determined by the intersection of its demand curve and the residual marginal factor cost (mfc). The firm buys quantity Q*df at a price of w*. At that price, the competitive fringe purchases quantity Q*f. The total quantity purchased is the sum of Q*df and Q*f, or Q*. All of the effects described so far occur in the input market. As a firm with monopsony power reduces input purchases, it will almost certainly reduce output.5 The effects of the output reduction depend upon conditions in the output market (Blair and Harrison 2010, p. 47). At one extreme, the monopsonist might be a perfect competitor in an output market in which supply is infinitely elastic. In that case, any reduction in the output of the monopsonist will be completely offset by an increase in production by other suppliers. Price and total quantity are unchanged, and consumers suffer no loss. Nevertheless, the monopsonist earns profits. The monopsonist sells the quantity at which marginal cost equals market price, just as a competitive firm would do, but the marginal cost of the monopsonist is higher than that of a competitive firm. And at the output the monopsonist produces, its average cost is lower than its marginal cost (Blair and Durrance 2008). Monopsony profits, therefore, are profits earned on the sale of output because of the exercise of market power in the input market. The perfectly competitive output market from the perspective of an input competitor and a monopsonist is depicted in figure 4.3. The firm can sell any amount it chooses at a price of p. If the firm is a perfect competitor in the input market, it will operate along marginal cost curve MC. It will sell q1 units, where price equals marginal cost, and in equilibrium price will also equal average cost. At the competitive quantity, the firm’s marginal cost curve and average cost curve (AC) intersect, and the firm’s revenues equal its costs. The monopsonist’s marginal cost curve (MCm) lies above a perfect input competitor’s marginal cost curve. The monopsonist equates price and marginal cost and sells q2 units. The firm earns profits because its average cost at q2 is less than its marginal cost. The monopsonist’s average cost equals price at the quantity a perfect input competitor would sell, q1. Thus, the monopsonist’s average cost curve (ACm) intersects with the marginal cost and average cost curves of a perfect input competitor at the competitive quantity. The monopsonist sacrifices profits if it expands output beyond q2; it does not suffer a loss unless it expands output beyond q1. In a more realistic setting, the monopsonist has no monopoly power, but output supply is not perfectly elastic. A reduction in output by the monopsonist will result in a reduction in total output. But the proportionate reduction in market output will be minute, and any associated price increase negligible, if even detectable. At the other extreme, the monopsonist may have monopoly power. In that case, the firm will set output quantity where its marginal cost equals its marginal revenue. It will sell less than if it only had monopsony power. An exercise of monopsony power in these circumstances will cause measurable welfare losses in both the input and output markets. 5  A monopsonist could substitute other inputs and continue to produce the same quantity of output. But the input substitution would be inefficient, and the welfare effects would be analogous.

Predatory Buying   77

Price and Cost

ACm AC

MCm

MC

p

0

q2

FIGURE  4.3  Output

q1

Quantity

Market for Input Competitor and Monopsonist

The welfare effects of monopsony are legally important. If competition law is interpreted to promote consumer welfare in a narrow sense, then an exercise of pure monopsony power—an exercise that has no appreciable effect on output price—is unproblematic, and the acquisition of that power is unobjectionable. If, however, the law is interpreted to promote total welfare or to recognize a value in the welfare of input suppliers, an exercise of pure monopsony power is a matter of concern, and the acquisition of such power by anticompetitive means is objectionable.6 Under any prevailing interpretation of competition law, the anticompetitive acquisition and exercise of monopsony power that causes a significant reduction in output quantity and a consequent increase in output price would be condemned.

4.2.  Predatory Buying Judicial decisions addressing claims of predatory buying are rare, either because the practice is rare or because the practice is rarely challenged. The most notable case is the 6  For a sample of differing views on whether antitrust law should be interpreted to protect input suppliers when output consumers are not injured, see Werden (2007), Rosch (2007), Salop (2005), Kirkwood (2005), Zerbe (2005), Grimes (2005), Hylton (2008), and Jones (1989).

78   John E. Lopatka US Supreme Court’s decision in Weyerhaeuser Co. v. Ross-Simmons Hardwood Lumber Co. (2007). Nevertheless, the economic logic of increasing purchases in the input market as an anticompetitive strategy can be examined, and it depends upon the setting in which the conduct takes place. A number of settings are possible, three of which will be addressed here:  (1)  the monopsonist competes with completely different firms in the input and output markets, and the monopsonist has no monopoly power; (2) the monopsonist competes with different firms in the two markets but has monopoly power; and (3) the monopsonist competes with at least some of the same firms in the input and output market.

4.2.1.  Completely Different Rivals and No Monopoly Power In the simplest setting for predatory buying, the predator is a dominant firm in the input market; its input competitors do not compete with it in the output market; and it has no monopoly power in the output market. These are the conditions the US Supreme Court assumed prevailed in Weyerhaeuser (2007). In that case, the Court rejected a claim that a lumber company that purchased about 65% of the alder sawlogs in a market but had no monopoly power in the hardwood lumber market excluded a competitor by paying timberland owners and loggers high prices for alder sawlogs; sawlogs accounted for as much 75% of sawmill’s cost of producing lumber. As a technical matter, the predator need not have monopsony power before it initiates predation. But as a practical matter, a buyer without some pre-predation monopsony power is unlikely to succeed in a predatory strategy. That strategy depends upon at least making a credible threat to bid up the price of an input by increasing the quantity purchased, and increasing purchases by the necessary amount will be implausible unless the firm has monopsony power. In this setting, any increase in profits brought about by predatory buying does not depend on price effects in the output market. Rather, the buyer can increase profits only by increasing its monopsony power, or increasing the gap between its average cost and the market price of the output. The conventional theory of predatory buying—which itself is adapted from the conventional theory of predatory pricing—assumes that the predator and its rivals have perfect information. Predation involves a two-period strategy: during the first, or predatory, period, the predator increases the quantity of input it purchases, forcing the price up, until rivals exit the market or permanently shrink their capacity; during the second, or recoupment, period, the predator reduces input purchases below pre-predation levels and pays a correspondingly lower price. The losses sustained during the predatory period represent an investment that pays off during the recoupment period. Predation is a rational strategy if the net effect is positive. Notably, the net effect must take into account the time-use value of money, for the firm sustains losses sooner than it enjoys gains, and gains therefore must be discounted to present value (Posner 2001, p. 210). Successful predation will reduce total welfare. The predator is able to exclude competitors because its marginal revenue product is greater than their collective demand. In a

Predatory Buying   79

Price and Cost MFC

S

w1

VMP Df 0

Q1 FIGURE  4.4  Predatory

Quantity

Buying with No Allocative Distortion

broad sense, the predator is more efficient than its competitors—its costs are lower or the value of its output is higher than its competitors’. Just how high the predator must increase price during the predatory period to exclude its competitors depends primarily on the relative efficiency of the predator and its rivals. As the monopsonist increases input purchases, it necessarily sacrifices profits, for the assumption is that it maximizes profits at a level that permits a competitive fringe to operate. Putting aside any potential legal liability, the predator will likely convert that additional input into output in the ordinary production cycle, because once the input is purchased its use in production of the output is normally its highest valued use; if the firm stores the extra input for future use, it will do so at a positive cost, reducing the expected net profits of the strategy. The dominant buyer may be able to exclude its rivals at an input price equal to the value of its marginal product. That is the input price that would prevail in a competitive market, and if competitors are excluded at this price, total welfare necessarily increases during the predatory period. The situation is depicted in figure 4.4. At an input price of w1, the competitive fringe purchases no input. The dominant buyer purchases Q1 units of input, and the input price equals the value of the marginal product to the dominant buyer. If the dominant buyer were instead a perfect competitor, it would also purchase Q1 units at a price of w1. The dominant buyer, however, may have to increase price beyond this level to exclude its competitor, and an input price above the competitive level implies an allocative distortion. The dominant buyer in figure 4.5 must purchase Q′ units at a price of w′ to exclude competitors. The deadweight loss is area def. The distortions caused by paying

80   John E. Lopatka Price MFC

S

e w′ w1 d

f

VMP

Df

0

Q1 FIGURE  4.5  Predatory

Q′

Quantity

Buying with Allocative Distortion

a supracompetitive price for an input and by paying a monopsony price have opposite directions, and normally a distortion caused by a buyer paying more than necessary for a good is not a matter of antitrust concern, for the buyer incurs a market penalty. But if paying a supracompetitive price is part of an exclusionary strategy, it is not necessarily self-deterring. If the distortion is greater than the distortion caused by the dominant firm’s pre-predation monopsony pricing, total welfare declines during the predatory period. Thus, if area def in figure 4.5 is larger than area abc in figure 4.1, welfare declines during predation. In any event, if the predator succeeds in excluding rivals and in recouping its losses through the exercise of additional monopsony power, the predatory campaign certainly reduces total welfare. In theory, total welfare can decline as a result of predation even if the predator excludes competitors at a price no higher than the competitive price: relative to the pre-predation period, the gain in welfare during the predatory period may be less than the loss of welfare during the recoupment period. The predator’s ability to recoup losses depends upon whether monopsony pricing during the recoupment period attracts new demand, be it in the form of new entry into the input market, reentry of former competitors driven out of the market, or resurrection of capacity permanently curtailed during the predatory period. New demand will push input price up, reducing the predator’s monopsony profits. If new demand is forestalled long enough, the predator can profit on balance. The condition necessary for successful recoupment is captured imperfectly in the concept of entry barriers. A challenge confronting lawmakers is to establish liability standards for predatory buying as well as predatory pricing that take into account the institutional competence

Predatory Buying   81

of tribunals and the expected costs of error (Blair and Lopatka 2008; Hylton 2008; Easterbrook 1981). Prices during what might be a predatory period of a predatory pricing campaign will always be low relative to some standard, and low prices are generally desirable. The threat of liability may chill pricing behavior that lawmakers want to encourage. Similarly, input prices during a putative predatory bidding campaign will be high relative to some standard, and though high prices will not necessarily benefit consumers, lawmakers may wish to encourage them in many circumstances. The legal policy objective is to establish standards that minimize the costs of error—the costs of both false positives and false negatives—given the inherent limitations in the decision-making ability of tribunals. The US Supreme Court has recognized that predatory pricing and predatory bidding are analytically symmetrical (Weyerhaeuser 2007, p. 321). The Court had held in Brooke Group Ltd. v. Brown & Williamson Tobacco Corp. (1993) that unlawful predatory pricing requires that a firm (1) charge prices below an appropriate measure of its costs and (2) have a dangerous probability of recouping its investment in below-cost prices. Analogously, according to the Court, the plaintiff in a predatory bidding case must prove that (1)  “the alleged predatory bidding led to below-cost pricing of the predator’s outputs” and (2) “the defendant has a dangerous probability of recouping the losses incurred in bidding up input prices through the exercise of monopsony power” (Weyerhaeuser 2007, p. 325).7 The logic of the Court’s predatory-pricing analysis is that prices normally are not “predatory” unless they can exclude an equally or more efficient competitor, and a price equal to or above the alleged predator’s costs can only exclude a less efficient competitor. A perfect competitor prices at marginal cost, and therefore the relevant measure of cost for identifying predation is in theory marginal cost. But the first prong of the Court’s test for predatory bidding is not directed to the price a perfect competitor would pay for an input, but rather the price the alleged predator would charge for its output. A firm may be able to exclude input competitors by raising price to no more than the value of its marginal product, which is the price it would pay in a competitive input market. By analogy to predatory pricing, a competitive input price cannot be predatory. And yet a monopsonist that converts all input into output and pays a competitive input price will charge an output price below marginal cost. The Court might have held that a monopsonist’s price for an input is predatory if it is above the value of the monopsonist’s marginal product (Blair and Harrison 2010, p. 73). If the first prong of the test is based on output price rather than input price, however, the relevant measure of cost is not marginal cost, but average cost. A monopsonist may increase input purchases, even in an amount that results in an output price below average cost, for competitively legitimate or at least nonpredatory reasons. For instance, demand for the monopsonist’s output may increase or the monopsonist may increase its productive efficiency. In either case, the value of the monopsonist’s marginal product will increase, and the profit-maximizing quantity of input will 7  The Court’s use of the present tense in describing the prospect of recoupment implies that a violation can be found before recoupment is complete or even begun.

82   John E. Lopatka increase. The equilibrium output price will equal average cost. But if the monopsonist misestimates an increase in demand or an increase in its efficiency, purchasing additional input the need for which does not materialize, it may pay an input price above the actual value of its marginal product and charge an output price below average cost. Similarly, a firm is likely to maintain a reserve of inputs, warehousing them to guard against unpredictable shortages of supply or surges in output demand. Warehousing is an aspect of productive efficiency. A monopsonist may conclude that its reserve has been inefficiently low and increase input purchases accordingly. Or the monopsonist may anticipate increases in input prices or the demand for its output, and it may increase its input reserves in response. Again, if its forecast proves mistaken, it may sell its output at a price below average cost. A firm that errs in predicting the future is apt to suffer a market penalty. But mistaken increases in input purchases that drive output cost above some objective benchmark do not imply a predatory intent or effect. The threat of liability can dampen competitive zeal. The leading models of predatory pricing depend upon asymmetric information (see Kreps and Wilson 1982; Milgrom and Roberts 1982; Ordover and Willig 1981). If a firm lowers its price, its competitors may not know whether the firm is taking advantage of its superior efficiency or is engaging in predatory pricing. If the firm has lower costs than its competitors, it will be able to exclude them and generally avoid legal liability. If the competitors cannot costlessly avoid losses by curtailing production while the firm is pricing low, they are better off leaving or staying out of the market. Predatory pricing could be used to acquire a reputation as a low-cost firm, even if the firm in fact is no more efficient than its competitors, and thereby deter entry. Another model assumes that firms have symmetric information, but the incumbent has lower costs than entrants (see Edlin 2002). The incumbent is able to charge high prices indefinitely because entrants know that they could be lawfully excluded by prices above the incumbent’s costs and therefore do not attempt to enter. Consumers never receive the benefit of the incumbent’s lower costs. The logic of models of strategic predatory pricing applies to predatory bidding. A buyer might respond to entry by increasing its input purchases to convince competitors that it is more efficient than they are, or it might be able to pay low prices perpetually because potential entrants realize that they could be lawfully excluded. Theories of strategic predatory pricing have not significantly affected legal standards, however, at least partly because lawmakers doubt the ability of tribunals to identify strategic price predation that does not satisfy administrable objective criteria. Analogous theories of predatory bidding are not likely to have any greater impact on the law.

4.2.2.  Completely Different Rivals and Monopoly Power A dominant buyer may also be a dominant seller, meaning that it has monopoly power in the output market while facing a competitive fringe. A dominant seller will calculate a residual demand curve equal to market demand minus the supply of the competitive fringe. The residual demand curve implies a residual marginal revenue curve. The

Predatory Buying   83

dominant seller will set quantity where its marginal cost is equal to its residual marginal revenue and charge the associated price. At that price, the competitive fringe will supply part of the market, earning no profits, and the dominant firm will supply the rest, earning monopoly profits. Total quantity is lower and price higher than they would be in a perfectly competitive market, but quantity is higher and price lower than they would be if the dominant firm were a perfect monopolist. Figure 4.6 depicts a market with a dominant seller and a competitive fringe. The dominant firm has lower marginal costs (MCdf ) than the collective marginal costs of its competitor, shown as the supply curve of the competitive fringe (Sf ). The dominant firm calculates a residual demand curve (d) and an associated residual marginal revenue curve (mr). It sells the quantity determined by the intersection of MCdf and mr, or Q*df units. The competitive fringe takes the price as given and sells the quantity dictated by Sf, or Q*f units. Total quantity is the sum of Q*df and Q*f, or Q*. The dominant firm earns profits based on the difference between price and its average costs. If the dominant firm competes against completely different firms in the input and output markets and engages in predatory buying, it will increase input purchases during the predatory period and likely convert at least some of the extra input into output. As its output increases, the welfare loss in the output market will decline. If the firm must raise input price above the competitive level to exclude input competitors, converts the input into output, and as a result sells output at a price below its marginal cost, the firm will cause net welfare losses in both markets. But it would expose itself to predatory pricing liability.

Price and Cost

Sf

p* MCdf d

D

mr 0

Q*f Q*df FIGURE  4.6 Dominant

Q*

Seller and Competitive Fringe

Quantity

84   John E. Lopatka If predatory buying is successful in excluding fringe buyers, the dominant buyer during the recoupment period will pay a lower price for and buy less input than it did prior to the beginning of predation. Consequently, it will reduce output, and the deadweight loss in the output market will be greater than it was in the pre-predation period. Successful predatory buying by a dominant buyer with monopoly power will result in net welfare losses in both the input and output markets. Predatory buying in such a case poses a greater danger to economic welfare than it does in a pure monopsony case—one in which the monopsonist has no monopoly power. Moreover, it threatens to impose a significant loss on consumers, whereas it represents at most a trivial threat to consumers in the pure monopsony situation. If competition law is interpreted to protect only consumers or to value consumer welfare more highly than the welfare of input suppliers, then predatory buying by a monopolist is a greater concern than predatory buying by a pure monopsonist. Nevertheless, the legal standards should be the same in the two contexts, assuming the imposition of a welfare loss in the input market alone is objectionable. In both settings, the mechanism of causing a welfare loss is the same. It relates solely to the use of price to exclude input competitors.

4.2.3.  Same Competitors in Input and Output Markets A dominant buyer might compete with at least some of the same firms in both the input and output markets. A common set of competitors is a necessary though not sufficient condition for the use of overbuying to earn profits in the output market by raising output price. The idea is straightforward: By raising input price through overbuying, the predator increases the marginal costs of its input competitors, inducing them to reduce their output or raise output price (Salop 2005; Salop and Scheffman 1983; Salop and Scheffman 1987). The predator’s own marginal costs are likely to increase, but they may increase in a lesser amount. The predator functioning as a dominant firm in the output market faces a competitive fringe with higher costs and calculates residual demand and residual marginal revenue curves that lie above the comparable curves prior to predation.8 Market price increases and total quantity decreases, and the predator’s share of the market increases. Increasing the costs of input competitors can have no effect on output price if those competitors do not compete with the predator in an output market. The exclusionary strategy is profitable only if the firm’s costs of predation are less than the profits it earns from a larger market share and higher output prices. Figure 4.7 depicts the effects in the output market of successful predatory overbuying in the input market by a dominant buyer with monopoly power. Overbuying increases the marginal costs of competitors, resulting in a rotation of the fringe supply curve from 8 

A predatory buying strategy may be successful even if the predator does not have the power in the output market to raise price by restricting its output. All that is necessary is that the demand curve for the output is negatively sloped and that the competitors’ supply curve is positively sloped (Krattenmaker, Lande, and Salop 1987).

Predatory Buying   85

Sf ′

Price and Cost

Sf

d′ p*′

MCdf ′

p*

MCdf

d

mr 0

Q*f Q*f ′

FIGURE  4.7 Dominant

mr ′ Q*

Q*df Q*df ′

D Quantity

Q*′

Buyer and Seller Facing Common Competitors

Sf to Sf′. The predator’s marginal cost curve rotates from MCdf to MCdf′. The predator calculates a new residual demand curve (d′) and residual marginal revenue curve (mr′), and it produces the quantity dictated by the intersection of mr′ and MCdf′, or Q*df′. Price increases from p* to p*′, the quantity sold by the competitive firm drops to Q*f′, and market quantity decreases to Q*′. The rationality of predatory buying as a method of increasing profits in the output market by reducing the elasticity of supply depends upon two conditions. First, the higher input cost imposed on competitors must have an appreciable effect on the output market, which generally implies that the predator’s input rivals account for a significant share of output. If a large share of the output market is supplied by firms that do not use the input or acquire the input in markets unaffected by the putative predator’s overbuying, the overbuying is not likely to affect output price. Similarly, if entry into the output market is easy without use of the affected input, an overbuying predatory strategy is likely to fail. Second, the increase in the predator’s profits must exceed the predator’s costs of predation. For example, if the predator uses the relevant input less extensively than does its rivals, a strategy of overbuying will increase the price the predator as well as its competitors pays. But because the input represents a smaller proportion of total cost for the predator than for its competitors, the resulting increase in output price more than offsets the predator’s increased costs. A classic example of this situation is a predator that

86   John E. Lopatka uses labor less intensively than does its competitors and drives up wages by hiring excess workers. Both the ability to raise input price by increasing the amount of input purchased and the profitability of a cost-raising strategy are affected by the predator’s relative size. A firm is not likely to be able to raise input price appreciably through its purchase decisions unless it purchases a substantial share of the available input before increasing the amount it purchases. And the smaller the firm is before it launches the strategy, the greater the proportionate increase in its own costs will be to impose significant cost increases on rivals, thereby necessitating higher profits in the output market to achieve a net positive effect. Nevertheless, a firm need not have the level of monopoly power necessary to raise price by restricting its own output. If the conditions necessary for rational predation by overbuying are not fulfilled, the implication is that the firm’s increase in input purchases is not part of an anticompetitive strategy to increase output price. That conclusion is consistent with the observation above that a firm may increase the amount of input purchased for legitimate reasons. Although the use of predatory buying merely to raise input price on the one hand and ultimately to raise output price on the other are both anticompetitive strategies, they have certain different implications that are economically and legally important. Critically, when predatory buying is used as a method of raising output price, input suppliers benefit, consumers suffer, and the predator profits concurrently. The strategy does not involve a two-period sequence. By contrast, when predatory buying is used merely to increase monopsony profits, the predator does not begin to gain until exclusion is successful. As a result, the profits earned when overbuying causes output price to rise need not be discounted to present value. Further, because the future is inevitably unpredictable to some extent, the form of predatory buying that depends upon an increase in output price is to that extent a less risky anticompetitive strategy. In addition, the law need not be as reticent to condemn predatory buying if the legal challenge takes place when anticompetitive effects are present than when they are predicted. If the theory of the claim is that the firm is using predatory buying only to increase monopsony profits and the claim is made during the predatory period, a tribunal will have to decide at least whether the firm is planning to lower the price it pays for the input below pre-predation levels during a recoupment period, and depending upon the standards adopted it may have to decide whether recoupment is objectively probable. A mistake in finding an illicit purpose or likely effect will punish competitively desirable conduct. Conversely, if the theory is that simultaneously the firm is overbuying input and output price has increased, a tribunal will not be required to make predictions, though it will be required to determine that the overbuying was illicit and led to increased output prices. The focus of legal analysis is the output market. The tribunal will have to determine that the alleged predator’s increase in input purchases caused output prices to increase, and it will also have to determine that the input purchases were not made for competitively legitimate reasons, even if based on assumptions or predictions that later proved erroneous.

Predatory Buying   87

4.3. Conclusion Predatory buying can be a rational anticompetitive strategy. In one setting, a dominant buyer can increase its purchases and input price to exclude competing purchasers, then reduce its purchases and input price to earn monopsony profits while imposing little or no adverse effect in the output market. In another, a buyer can engage in the same conduct while exercising monopoly power as well, reducing output sufficiently to impose a significant negative impact on output consumers. In a third, a buyer can bid up the price of an input to raise the costs of input competitors who also compete with it in the output market, thereby forcing output price up, earning profits immediately in the output market, and imposing a loss on consumers. In each of these settings, predatory buying can increase the predator’s profits and is therefore rational. In each, the strategy is anticompetitive, if competition is understood to be a process that maximizes total welfare. In the first setting, however, the welfare loss is sustained almost entirely in the input market, and if competition policy is directed solely at the welfare of consumers, the strategy is not anticompetitive; in the two other settings, consumers are injured appreciably. That predatory buying can be a rational anticompetitive strategy in theory does not mean that it is prevalent. A number of obstacles can thwart a predatory campaign, and successful predatory buying may be rare. Moreover, the heart of predatory buying is an increase in the amount of an input purchased by a large buyer. A buyer may increase purchases for a number of competitively legitimate reasons. Legal standards that are ill-conceived in theory, that are imprecise and unpredictable, or that cannot be effectively administered can stifle competitive fervor. They can do more harm than good. The optimal legal approach to predatory buying is an open question.

References Bauer, Joseph P., and William H. Page. 2002. Federal Antitrust Law. Vol. 2. Cincinnati: Anderson. Blair, Roger D., and Jeffrey L.  Harrison. 2010. Monopsony in Law and Economics. New York: Cambridge University Press. Blair, Roger D., and Christine P.  Durrance. 2008. The economics of monopsony. In Wayne D.  Collins et  al., Issues in Competition Law and Policy, vol. 1. Chicago:  American Bar Association. Blair, Roger D., and John E. Lopatka. 2008. Predatory buying and the antitrust laws. Utah Law Review 2: 415–469. Brooke Group Ltd. v. Brown & Williamson Tobacco Corp., 509 U.S. 209 (1993). Easterbrook, Frank H. 1981. Predatory strategies and counterstrategies. University of Chicago Law Review 48: 263–337. Edlin, Aaron S. 2002. Stopping above-cost predatory pricing. Yale Law Journal 111: 941–991.

88   John E. Lopatka Grimes, Warren. 2005. Buyer power and retail gatekeeper power: Protecting competition and the atomistic seller. Antitrust Law Journal 72: 563–588. Hylton, Keith N. 2008. Weyerhaeuser, predatory bidding, and error costs. Antitrust Bulletin 53 (Spring): 51–73. Jones, William K. 1989. Concerted refusals to deal and the producer interest in antitrust. Ohio State Law Journal 50: 73–92. Kirkwood, John B. 2005. Buyer power and exclusionary conduct: Should Brooke Group set the standards for buyer-induced price discrimination and predatory bidding? Antitrust Law Journal 72: 625–668. Krattenmaker, Thomas G., Robert H. Lande, and Steven C. Salop. 1987. Monopoly power and market power in antitrust law. Georgetown Law Journal 76: 241–269. Kreps, David M., and Robert Wilson. 1982. Reputation and imperfect information. Journal of Economic Theory 27 (Aug.): 253–279. Lerner, Abba P. 1934. The concept of monopoly and the measurement of monopoly power. Review of Economic Studies 1 (June): 157–175. Milgrom, Paul, and John Roberts. 1982. Predation, reputation and entry deterrence. Journal of Economic Theory 27 (Aug.): 280–312. Ordover, Janusz A., and Robert D. Willig. 1981. An economic definition of predation: Pricing and product innovation. Yale Law Journal 91: 8–53. Posner, Richard A. 2001. Antitrust Law. Chicago: University of Chicago Press. Rosch, J. Thomas. 2007. Monopsony and the meaning of “consumer welfare”: A closer look at Weyerhaeuser. Columbia Business Law Review 2007: 353–370. Salop, Steven C. 2005. Anticompetitive overbuying by power buyers. Antitrust Law Journal 72: 669–715. Salop, Steven C., and David T. Scheffman. 1983. Raising rivals’ costs. American Economic Review Papers and Proceedings 73 (May): 267–271. Salop, Steven C., and David T. Scheffman. 1987. Cost-raising strategies. Journal of Industrial Economics 36 (Sep.): 19–34. Werden, Gregory J. 2007. Monopsony and the Sherman Act: Consumer welfare in a new light. Antitrust Law Journal 74: 707–738. Weyerhaeuser Co. v. Ross-Simmons Hardwood Lumber Co., 549 U.S. 312 (2007). Zerbe, Richard O., Jr. 2005. Monopsony and the Ross-Simmons case: A comment on Salop and Kirkwood. Antitrust Law Journal 72: 717–726.

CHAPTER 5

C OM P E T I T I V E D I S C OU N T S A N D A N T I T RU S T P OL IC Y KEVIN M. MURPHY, EDWARD A. SNYDER, AND ROBERT H. TOPEL

5.1. Introduction This chapter deals with a broad class of quasi-exclusive, vertical agreements in which a seller conditions price discounts on the specified quantity or share of a product line that the buyer commits to purchase from the seller. We refer to such agreements as “quantity commitment discounts” (QCDs), though they are often referred to as “loyalty discounts,” as they appear to exchange price concessions for a buyer’s loyalty to a particular brand. Both economic theory and the law recognize that in some cases pricing and business practices of sellers may harm or weaken rivals and might also reduce social welfare. The classic example is predatory pricing, in which a seller temporarily prices below incremental cost with the explicit goal of driving rivals from the market.1 Common and typically procompetitive business practices designed to increase sales—including forms of quantity-related discounts, nonlinear pricing, and various vertical restrictions on distributors—similarly might in certain circumstances harm competition by harming competitors (see, e.g., Schwartz and Vincent 2008; Wilson 1993). Of particular concern to the current US Department of Justice (DOJ) are “contracts that reference rivals” (CRR) in which the terms of one seller’s procurement contracts with downstream buyers implicitly or explicitly condition on the terms that apply to 1 

For a more detailed discussion, see Kenneth Elzinga and David Mills, “Predatory Pricing,” in this Handbook.

90   Kevin M. Murphy, Edward A. Snyder, and Robert H. Topel rivals.2 Exclusive dealing arrangements, where a buyer agrees to purchase all of a particular product from a single seller, can be viewed as a limiting case of CRRs. More generally, a seller may condition discounts from list prices on a buyer’s agreement to promote the seller’s products more prominently than certain competitors’ or to not promote them at all.3 While such practices are known to have procompetitive benefits, if widely practiced by a dominant seller the cumulative effect of such practices might be to weaken competition by impairing rivals’ ability to compete in the market. The ambiguity and difficulty inherent in attempts to balance pro- and anticompetitive effects of common pricing practices is, in our view, the great challenge of antitrust enforcement. The bedrock proposition—that harms to rivals does not imply harm to competition—does not extend to governing this core component of the economic activity. It is recognized that lax enforcement can harm competition, but so can overly aggressive enforcement that protects competitors at the expense of vigorous competition. Further, absent clear standards defining the bounds of illegal conduct the mere threat of antitrust liability may dampen rivalry among firms, with resulting harm to the competitive process and, ultimately, consumers.4 An increasing reliance on economic analysis in antitrust matters has resulted in a movement towards fewer per se illegality rules; however, per se illegality rules have not been replaced with broad safe harbors (Kobayahi and Muris 2012). As our discussion in section 5.2 indicates, law and policy have not evolved to yield clear standards for judging QCDs and indeed have taken divergent paths. In section 5.3 we cover the basic economics of QCDs, showing that they are a natural outcome of sales-promoting competition by differentiated sellers and are typically mutually beneficial for participating buyers and sellers. In section 5.4 we evaluate the circumstances when QCDs may cause harm to competition. We demonstrate that QCD agreements that would arise absent an ability or intent to exclude rivals might nevertheless do so, and might also cause ancillary harm to competition. In section 5.5 we then assess the various means of testing for potential harm. We discuss interpretations of the so-called attribution test and its flaws as well as other indicators of potential harm. 2 

Scott-Morton (2012) concludes that CRRs, such as market share discounts, “have the potential to harm consumers and competition, particularly—but not always—when they involve firms with market power. CRRs have thus been, and will continue to be, the subject of antitrust scrutiny, both at the government [sic] and in private litigation.” See Scott-Morton (2012) p. 3. Salop (1986) p. 265 discusses conditions under which pricing practices such as most-favored-nation (MFN) and meeting competition clauses (MCC) may have anticompetitive implications. 3  For example, in J.B.D.L. Corp. et al. v. Wyeth-Ayerst Laboratories, Inc., et al. (6th Cir. 2007), Wyeth’s contracts with pharmacy benefit managers granted discounts on its Premarin conjugated estrogen product in exchange for exclusive placement of Premarin in pharmacy benefit managers’ core formulary. 4  The degree of difficulty in evaluating pricing practices is underscored by the fact that private plaintiffs in the United States can bring claims of anticompetitive exclusion against their rivals under the Section 4 of the Clayton Act, which provides for treble damages and one-way fee-shifting in favor of plaintiffs (Clayton Antitrust Act of 1914, Pub. L. 63-212, 38 Stat. 730, codified at 15 U.S.C. §§12-27, 29 U.S.C. §§52-53). If the threshold to survive a motion to dismiss is that economic analysis is needed to assess the effects of a firm’s pricing practices on its rivals, then firms face the prospect to having to settle such claims or risk treble damages and the legal costs of both parties. For relevant discussions, see Easterbrook (1984); Kauper and Snyder (1991).

Competitive Discounts and Antitrust Policy   91

Our concluding remarks in section 5.6 follow most directly from two main points. First, QCDs virtually always have a clear procompetitive rationale. Second, while economic theory shows that under certain conditions the intent and effect of commitment discounts could be to harm competition, these same theories provide little guidance in identifying situations where harm actually occurs. Further, few if any past cases provide convincing evidence of competitive harm, and no evidence of outright exclusion to our knowledge. In our view, the ubiquity of procompetitive or competitively neutral reasons for QCDs, combined with the lack of reliable tests or filters that would identify anticompetitive conduct, support our overall conclusion that QCDs should be viewed as presumptively legal. Our views contrast, therefore, with current US policy whereby a broad range of single-firm pricing practices are typically judged under a rule of reason analysis.

5.2.  Legal and Policy Context The number of actual judgments evaluating claims of anticompetitive exclusion have not converged on a standard. Noteworthy is the Third Circuit’s 2004 decision in Lepage’s, upholding a jury verdict that 3M’s bundled quantity discounts violated Section 2 based on the standard that 3M’s actions made it “very difficult or impossible for competitors to engage in fair competition.”5 The Lepage’s decision has proven controversial because the court did not provide an objective standard of competitive harm and neglected to show that plaintiff LePage’s was unable to profitably compete in the sale of its private-label transparent tape against 3M’s bundled discounts. When arguing that the US Supreme Court should not grant certiorari in LePage’s, the US solicitor general described the case law as underdeveloped and pointed out that the lack of systematic assessment of alternative standards by lower courts failed to establish bases for determining how standards would be applied.6 5 See LePage’s v. 3M, 324 F3d 141 (3d Cir 2003) (en banc), cert. denied 124 S Ct 2932 (2004). LePage’s claimed that 3M set target quantities for individual LePage’s distributors such that it was impractical for the distributors to meet the targets and retain LePage’s as a supplier. See Brief for Respondents in Opposition at 5, 3M v. LePage’s, 2003 WL 22428377, at *1 (2003) (No. 02-1865). 6  The US government stated:

The court of appeals focused exclusively on petitioner’s proposed below-cost sales standard, Pet. App. 7a–8a, and the meager case law addressing bundled rebates offers little assistance in determining how alternative standards might work in practice. Because the courts below did not attempt to apply alternative standards to the facts, their decisions offer little to illuminate such potentially significant questions as whether an equally efficient supplier of private label tape could profitably have matched 3M’s discounts and rebates; whether lowered prices resulting from the bundled discounts would have increased quantities of tape purchased by an amount sufficient to make the lowering of prices profitable, even if LePage’s had matched the discounts; and whether 3M’s “discounts” and “rebates” actually resulted in reduced prices for 3M’s customers, as 3M contends, or whether the net result was a price increase structured to discourage trade with LePage’s, as LePage’s apparently claims.

92   Kevin M. Murphy, Edward A. Snyder, and Robert H. Topel The courts have not been consistent in selecting and applying standards when evaluating QCDs. In Concord Boat Corp. v.  Brunswick Corp., Brunswick—a producer of stern-drive boat engines—offered discounts to boat manufacturers who would commit to purchase at least 60% of their engines from Brunswick, which naturally reduced purchases from rival engine producers among participating buyers.7 In Concord Boat the court applied a cost-based standard. The court in SmithKline Corp. v. Eli Lilly & Co. held that Lilly’s loyalty discounts on a bundle of antibiotics would have excluded an “equally efficient” producer.8 Meanwhile, in Masimo Corp. v. Tyco Health Care Group the court sustained a jury’s verdict that Tyco’s share-based discounts on a single product—pulse oximetry sensors purchased by hospitals—violated Section 2 by maintaining Tyco’s monopoly power.9 In Eli Lilly and in Tyco, as in other cases where liability was not found, the courts sought to establish whether the defendant’s contracts were exclusionary in the sense that a hypothetical equally efficient rival could not profitably compete against them. Intel’s agreements with computer manufacturers such as Dell or Hewlett Packard were alleged to condition discounts for Intel processors on the share of total processor purchases that a buyer would commit to purchase from Intel—an implicit reference to Intel’s sole major rival, AMD.10 Intel’s practices generated challenges by the EU, Korea, and US Federal Trade Commission,11 as well as a flow of private antitrust suits by allegedly excluded rivals, but no well-defined standard.12 Finally, and most recently, is the Supreme Court’s refusal in April 2013 to review Eaton v. Meritor. Despite above-cost pricing of its heavy-duty transmissions, Eaton was found guilty of illegal monopolization by the Third Circuit because of other terms of the long-term contracts Eaton formed with heavy-duty truck manufacturers.13

Rf. Brief for the United States as Amicus Curiae, 3M Company v. LePage’s Incorporated, No 02-1865, p. 18 (footnote omitted). 7 See Concord Boat Corp. v. Brunswick Corp., 207 F.3d 1039 (8th Cir. 2000). The discounts were 1% for

purchases accounting for 60% of total purchases, 2% for purchases accounting for 70% of total purchases, and 3% for purchases accounting for 80% of total purchases. 8  SmithKline Corp. v. Eli Lilly & Co., 427 F. Supp. 1089 (E.D. Pa. 1976), affirmed 575 F.2d 1056 (3d Cir. 1978). 9  Masimo Corp. v. Tyco Health Care Group, L.P., No. CV 02-4770 MRP, 2006 WL 1236666, at *13 (C.D. Cal. Mar. 22, 2006). 10  For discussion and analysis of the antitrust claims relating to Intel’s use of loyalty discounts, see Wright (2011). 11  See, e.g., “Antitrust: Commission imposes fine of €1.06 bn on Intel for abuse of dominant position; orders Intel to cease illegal practices,” European Union press release (May 13, 2009), http://europa.eu/​ rapid/press-release_IP-09-745_en.htm?locale=en; Statement of Chairman Leibowitz and Commissioner Rosch In the Matter of Intel Corporation, Docket No., 9341, Federal Trade Commission (December 16, 2009), http://www.ftc.gov/os/adjpro/d9341/091216intelchairstatement.pdf. 12  Other antitrust cases alleging anticompetitive exclusion include Ortho Diagnostics Sys. Inc., v. Abbott Lab, Inc., 920 F. Supp. 455 (S.D.N.Y. 1996); Virgin Atlantic Airways Ltd. v. British Airways PLC, 69 F. Supp. 2d 571 (S.D.N.Y. 1999), aff ’d 257 F.3d 256 (2d Cir. 2001); Applied Medical Resources Corp. v. Ethicon Endo-Surgery (Johnson & Johnson), 03-CV-1329 (C.D. Cal. 2006). 13 See Eaton Corporation v. ZF Meritor LLC and Meritor Transmission Corporation, cert. denied, 133 S. Ct. 2025 (2013).

Competitive Discounts and Antitrust Policy   93

Given the lack of clear standards for evaluating QCDs and lack of guidance to businesses, the US Department of Justice (DOJ) in 2008 issued a detailed report clarifying its positions on liability for single-firm conduct (US Department of Justice 2008). The DOJ cited a number of potential benefits related to QCDs, including economies of scale and scope, promotional incentives for retailers, inducing customers to try new products, customer convenience, and price discrimination based on differences in customer demand elasticities.14 The DOJ report recommended a fairly high liability threshold for conduct with demonstrated procompetitive benefits and potential anticompetitive harms: for conduct to be actionable, harms would have to be “disproportionate” to benefits. The 2008 DOJ report also attempted to provide more detailed guidance based on tests. For example, though above-cost net prices may in theory exclude, the DOJ report as well as the 2007 Antitrust Modernization Commission (AMC) recommended that discounting practices that pass the so-called attribution test should be viewed as per se legal with no further inquiry into possible effects. Thus, in the case of bundled QCDs—like those featured in LePage’s—where a subset of a discounting seller’s products are contestable by rivals, all the discounts on noncontestable elements of the bundle are attributed to (subtracted from) the price of contestable units. A “net price” above incremental cost would be viewed as lawful, which is inconsistent with the recent Eaton decision. By this rule, a finding that the net price is below incremental cost would suggest that discounts might be unprofitable to the seller absent some other motive, which might include exclusion. As we discuss in section 5.5 below, even this safe harbor is fairly narrow, however, as the attribution test is known to produce “false positives,” and we provide several additional reasons why this is so. One may further object that even in combination with other (noisy) signals of competitive effects, reliance on tests risks exposing innocent contracting practices to antitrust scrutiny and possible litigation, and may blunt sellers’ incentives to engage in more aggressive and procompetitive discounting practices. The 2008 DOJ report’s recommendations, however, were formally withdrawn in 2009. Assistant Attorney General for Antitrust Christine Varney commented as follows: [T]‌ he disproportionality test reflected an excessive concern with the risks of over-deterrence and a resulting preference for an overly lenient approach to enforcement. The failing of this approach is that it effectively straightjacketed antitrust enforcers and courts from redressing monopolistic abuses, thereby allowing all but the most bold and predatory conduct to go unpunished and undeterred. (Varney 2009, pp. 8–9).

The withdrawal of the DOJ 2008 policy guidelines, as indicated, underscores the absence of a standard for determining whether any particular discounting practice merits 14 

US Department of Justice (2008). See also similar discussion in Antitrust Modernization Commission 2007. Some observers argue that the widespread use of QCDs in situations where anticompetitive intent “makes no economic sense” indicates that some such benefits must exist (e.g., efficiencies or lower costs). See Jacobson and Weick (2012).

94   Kevin M. Murphy, Edward A. Snyder, and Robert H. Topel antitrust scrutiny or liability. Current DOJ policies as reflected in its United Regional Health Care System move yet further from a common standard, by requiring sellers to tailor their discounts to the specific capabilities of particular rivals.15 Other efforts following LePage’s have been made to define standards or tests that would distinguish instances in which QCDs might be anticompetitive.16 Noting that the possible anticompetitive impact from QCDs almost always relies on the existence of rivals’ scale economies, some suggest that initial inquiries focus on whether a seller’s contracts deny the benefits of scale to affected rivals (see, e.g., Carlton and Waldman 2008). Accordingly, demonstration of rivals’ scale economies might be considered a necessary condition for a successful complaint of exclusion. The challenge herein is that various economies are difficult to quantify and it is even more difficult to ascertain whether a particular discounting practice prevents their realization. Moreover, reliance on evidence of rivals’ scale economies requires a discounting seller to calculate the impact of its contracts on rivals’ costs and profits, and to refrain from discounts that might harm them. Here we agree with Areeda and Hovenkamp (2008) that such attempts to find a middle approach would make a contracting seller “trustee for another firm’s economies of scale,” holding “a price umbrella over its rivals” even if rivals’ technologies were somehow known, with the clear danger of dampening competition (Areeda and Hovenkamp 2007, par. 749b at 249).

5.3.  Basic Economics of Quantity Commitment Discounts In this section we explain how quantity commitment contracts and associated discounts from “list” prices are natural outcomes of the competitive process, as sellers seek to increase sales.17 These increases in sales naturally occur at the expense of rivals. Key to what follows is that the typical seller does not face perfectly elastic demand for its product. As a result, almost all sellers in modern wholesale and retail markets have some control over the prices they charge—they do not simply “take” prices as given but instead have pricing strategies in which chosen prices exceed incremental costs. Simple linear pricing (“here’s my price, buy what you want”) leaves unrealized gains from trade and, therefore, establishes incentives for buyers and sellers to devise ways to unlock them.

15  Competitive Impact Statement filed by the US Department of Justice, United States of America and State of Texas v United Regional Health Care System, United States District Court for the Northern District of Texas, February 25, 2011, http://www.justice.gov/atr/cases/f267600/267653.pdf. 16  These efforts are discussed in US Department of Justice (2008). 17  The discussion that follows is based on Murphy and Topel (2011a, 2011b).

Competitive Discounts and Antitrust Policy   95

5.3.1.  Mutual Gains from Quantity Commitment Discounts Assume that seller S produces a differentiated product within a category of close substitutes also produced by rival(s) R. For the situations we have in mind, a buyer (B) will typically be a downstream business that uses the products of S and R as inputs or resells them. For example, Intel (S) is a producer of processors purchased by OEMs (B) as components for personal computers, but rival AMD makes processors that are close substitutes for at least some of the Intel line. Most large OEMs purchase processors from both Intel and AMD. Similarly, Johnson & Johnson (J&J) produces a line of endoscopic surgical tools purchased by hospitals, but rival Tyco also produces a full line of tools, while smaller, more specialized, manufacturers produce substitutes for various elements of the J&J and Tyco lines. Most hospitals purchase from multiple vendors. And grocery stores typically stock beverage lines produced by both Coke and Pepsi, who compete for scarce shelf space and other promotional advantages. In each of these examples, within-buyer competition between S and R means that increased purchases from S reduce purchases from R, and conversely. Stepping back, we should recognize that even if S had no rivals, it would want to use QCDs to realize additional gains from trade with buyers. When S does have rivals, the sales gained from QCDs may be a combination of additional purchases by buyers and buyers shifting purchases from rivals to S. To feature the issue of potential exclusion, we focus on the case where the seller S has a rival R and assume that incremental purchases from S result in a one-for-one substitution in purchases from R.18 This assumption is appropriate in some settings, such as when hospitals procure supplies or HMO pharmacies procure drugs targeted at particular ailments. The fact that these buyers purchase substitute products from multiple vendors implies a within-buyer demand for variety. For example, within a hospital some surgeons may have a strong preference for a particular manufacturer’s surgical tools while other surgeons are more willing to substitute, and some HMO members with a particular ailment may be better served by one type of drug within a class of therapeutic equivalents. In other cases the assumption of exact one-for-substitution is less compelling, such as when a retailer resells multiple brands of substitute goods to final consumers. Figure 5.1 illustrates the incentive to discount in this environment. We assume for now that buyers are identical and that demand curve Q(P;PR) represents the quantities that representative buyer B wishes to purchase from S at various hypothetical prices P, holding constant the prices charged by the rival seller, PR.19 Equivalently, at any quantity

18  We only require that S and R provide substitutes, though the one-for-one substitution is a good representation in many important cases. Whether one or the other brand of surgical tool is used does not materially affect the number of surgeries. Displacement is likely to be less than one-for-one in the grocery store example of Coke and Pepsi. 19 If R produces a differentiated product, we interpret P as its optimal Bertrand price given P . With R L many rivals who are undifferentiated from each other we interpret PR as equal to rivals’ marginal cost.

96   Kevin M. Murphy, Edward A. Snyder, and Robert H. Topel Price

PL

PD

D

IB IS

K Q(P,PR )=v(Q,PR ) Quantity QL

QD

QE

FIGURE 5.1  The non-contract price and quantity are PL > K and QL. With a quantity commitment contract buyer and seller may achieve mutual gains at combinations like D that lie above IS and below IB. The efficient quantity is QE where v(QE, PR) = K.

Q the same curve represents B’s willingness to pay for an additional unit, v(Q;PR). With these demand conditions a seller that charges a single “list” price PL for all units will set price above marginal cost (PL > K) and sell quantity QL = Q(PL,PR).20 The seller earns a profit of ΠL = [PL − K]QL. With price above marginal cost it is obvious that seller S would gain if buyer B were to purchase more than QL at the same price. But B is unwilling to do so: the meaning of the demand curve is that B wishes to purchase exactly QL units at price PL—each additional unit beyond QL is worth less than PL to the buyer. Even so, there is a set of prices P < PL and quantities Q > QL where both B and S would be better off. To find these mutually advantageous combinations we draw two “indifference curves,” IS (for the seller) and IB (for the buyer). Curve IS represents the set of all price-quantity pairs that yield to S the same profit as combination (PL,QL), equal to ΠL = [PL − K]QL.21 By construction, curve IS touches (is tangent to) B’s demand curve at (PL,QL) but otherwise lies everywhere above the demand curve—S would accept lower prices, but only in exchange for 20  The familiar solution is to set price so that marginal revenue is equal to marginal cost. We do not show the marginal revenue curve in fi ­ gure 5.1 in order to reduce clutter. 21  The formula for the combinations on indifference curve I is P = K+Π /Q. S L

Competitive Discounts and Antitrust Policy   97

greater quantity increments than are available along the demand curve.22 It follows that all price-quantity combinations above the curve IS yield greater profits. Similarly, indifference curve IB is the set of combinations that yield to B the same consumer surplus as (PL, QL). By construction IB is tangent to the horizontal line PL at the point where the demand curve crosses PL, but otherwise lies everywhere below PL—if given the choice of any quantity at price PL the buyer would choose QL, but would be willing to purchase more if compensated by a sufficiently lower price, and so on. The buyer prefers all price-quantity pairs below IB to combinations on it as they yield greater consumer surplus. Since B’s indifference curve is tangent to the horizontal line at PL and S’s indifference curve is tangent to B’s downward-sloping demand curve, there will always be mutually beneficial gains to trade available with the property that P < PL and Q > QL. The area below IB and above IS is the set of price-quantity pairs that are mutually preferred to the stand-alone price and quantity. Ignoring income effects, the quantity that maximizes the joint surplus of B and S for a given rival’s price is QE where the buyer’s marginal value is equal to the seller’s marginal cost, v(QE, PR) = K, so the “contract curve” of price-quantity pairs that maximize joint gains is vertical at QE. But full efficiency need not be achieved for discounting to be mutually beneficial. All pairs like D = (PD, QD) in the shaded region involve the buyer paying a discounted unit price PD QL)—and correspondingly fewer from R. Note that the buyer’s ability to commit is necessary to achieve this mutually beneficial outcome: At the discounted price PD an uncommitted buyer would choose to purchase less than QD along the demand curve, at which S’s profit would be less than ΠL. Therefore, realization of mutual gains requires a contract, explicit or implicit, between buyer and seller. The contract commits the buyer to purchase a larger quantity in exchange for the seller’s commitment to an appropriately large discount from the “list” (uncommitted) price PL.23 As we show later, there are many ways to achieve this commitment including nonlinear pricing, bundling, and loyalty discounts, all of which fall under our definition of QCDs. Importantly, the quantity commitment and associated price discount illustrated in figure 5.1 need not be initiated or designed by the seller. As indicated by figure 5.1, both parties stand to gain, and it is just as reasonable that the moving party is the buyer, who offers to purchase more in exchange for a negotiated discount. In practice, with heterogeneous buyers the size and location of the “football” area of mutual gain is uncertain and buyer specific—but both parties know it is there and that they would prefer to be in it. There is a deal to be made. In many cases these gains are the foundation 22 

In drawing figure 5.1 we have assumed that all buyers are the same, with identical demand curves. Then IS must be tangent to D. If buyers are heterogeneous then IS need not be tangent to any particular buyer’s demand curve, but this does not affect the following analysis. 23  Given this need for a buyer’s commitment to purchase more than it would otherwise (and hence less from S’s rivals), it is not surprising that some buyers might complain ex post that they are being “forced” to buy more from S than they would like. In a sense, this forcing is true. Without the contract, however, they would not get the discount.

98   Kevin M. Murphy, Edward A. Snyder, and Robert H. Topel for buyer-seller negotiations, be they between Intel and Dell for a favorable deal on processors, or between a grocery chain and Coke over the placement of beverages on shelves. In other cases, such as in Concord Boat, a seller may announce a common pricing schedule that offers explicit discounts in exchange for specified quantity or share commitments. This analysis has important implications for antitrust policies that target contracted discounts and quantity commitments. It says that in any situation where the uncommitted price would exceed marginal cost—which for practical purposes means always— there are mutual gains for a buyer and a seller from an agreement that offers a discount in exchange for a buyer’s commitment to purchase more. Put differently, absent transactions costs or barriers to contracting, simple linear pricing is not an equilibrium outcome of the competitive process. As a business practice subject to antitrust scrutiny, therefore, QCDs always have a procompetitive justification even if, in some circumstances, they might also have an exclusionary impact—whether intentional or ancillary—as explained below.

5.3.2.  Competition and the Division of Gains between Buyers and Sellers All combinations like D in figure 5.1 increase the joint surplus of S and B, while reducing purchases from rival(s) R. These gains exist because the stand-alone price exceeds marginal cost. The division of these gains depends on the relative bargaining powers of the two parties and the nature of competition from rivals, but without further structure it is not possible to say more than that linear pricing is not an equilibrium outcome when contracts are feasible, and that both parties stand to gain from the deal. If the parties realize the entire potential surplus, then the outcome will be on the vertical contract curve at the jointly efficient quantity QE. This is not the end of the story, however, because figure 5.1 is drawn under the assumption that S’s chosen list price and the competing offers by rivals are unaffected by commitment contracts. Assume for the moment that rivals are perfectly competitive with constant cost, so they do not offer competing contracts and PR is fixed. There are two reasons that PL will not be the same as the price that S would charge in the absence of contracts. First, so long as some buyers purchase at the list price and others through contracts, the opportunity to purchase through a commitment contract leads to sorting of buyers. For example, if there are fixed costs of negotiating and enforcing contracts then large buyers will be more likely to purchase through contracts. If small, noncontract buyers are more elastic demanders—they are more willing to substitute R for S—then this would tend to reduce the list price; and conversely if noncontract buyers are predominantly inelastic demanders. Second, with contract and noncontract customers seller S is likely to set a list price higher than the price that would maximize profits from noncontract buyers alone. To see this, let P be the price that would maximize S’s profits from noncontract buyers.

Competitive Discounts and Antitrust Policy   99

Now consider a small increase in the list price above this level. This price increase has only a second-order (approximately zero) impact on the profits earned from noncontract buyers because (by assumption) P was set to maximize profits in that segment. But for contract buyers the price increase makes the buyer’s alternative of purchasing at the list price less attractive—indifference curve IB in figure 5.1 shifts upward as the list price is increased, allowing S to earn greater profits from contract buyers.24 _ It follows that the seller would want to set PL > P . The strength of this incentive to raise the list price clearly depends on the relative numbers of contract and noncontract buyers because raising the list price by more than a “small” amount reduces profits among noncontract buyers. If the number of contract buyers is small relative to noncontract buyers, then the incentive to raise the price will also be small, and conversely. Both contract and noncontract customers of S are worse off than if S did not increase its list price, and some may be worse off than if commitment contracts were impossible. But it is difficult to argue that this feature of commitment contracts is anticompetitive—it is instead an example of first-degree price discrimination that allows S to extract more of the gains from trade. Nor are rivals harmed by the higher list and contract price— they sell more than they would otherwise, though possibly less than they would in the absence of commitment contracts.25 In most real-world cases S faces a differentiated rival, R, whose price PR also exceeds marginal cost. Because of this wedge, R would also gain by offering a discount to increase sales at S’s expense. This means that B’s alternatives are improved by “competition for the contract” among differentiated rivals, S and R. Assuming that S wins this competition, its contracts with buyers must nevertheless offer as much consumer surplus as R’s best offers, which must yield at least as much profit to R as simply selling fewer units at its own list price PR. This competition improves B’s alternative as well as its bargaining position with S, while the reduction in demand for R will generally reduce R’s list price. Both of these effects of competition benefit buyers.26 The effect on R’s list price can be very important. For example, in their analysis of exclusive dealing Klein and Murphy (2008) showed that if S and R initially compete with linear prices each would often have the unilateral incentive to offer the buyer, on an all-or-nothing basis, an exclusive contract at a price below the equilibrium linear price, since that allows them to capture more sales at a lower price. However, once the two 24 

When rival R also offers a commitment contract discount, a buyer has two alternatives to S’s contract: (1) enter a commitment contract with R and pay list for S’s good, or (2) contract with neither and pay list for both. In either case the alternative is to pay list for S’s good, so raising the list price makes the rival’s offer less attractive. 25  An extreme example: suppose buyers are identical, all contract with S, and all bargaining power resides with S. Then S will increase PL to the level that drives quantity demanded in figure 5.1 to zero. The contract leaves buyers indifferent between purchasing the efficient quantity QE via contract and doing without S’s good entirely—S’s contract and list price replicate perfect first-degree price discrimination and capture all the gains from trade. Consumers of S (and rivals) are harmed relative to a world without contracts, but (as in any case of perfect price discrimination) efficiency is enhanced. 26  Klein and Murphy (2008, 2011) and Zenger (2010) analyze “competition for the contract” in the case of exclusive dealing and illustrate how such competition benefits consumers.

100   Kevin M. Murphy, Edward A. Snyder, and Robert H. Topel sellers compete for the exclusive contract, the resulting prices will often be lower and consumer welfare higher than they would be under linear pricing. Under these conditions B’s outside opportunities are enhanced by the use of QCDs and consumers benefit more than the static analysis in figure 5.1 would imply.

5.3.3.  Loyalty Discounts and Other Forms of Quantity Commitment The contract described above has a very particular form: buyer B commits to purchase a specified quantity and seller S commits to supply those units at a discounted price PD < PL, where PL is the noncontract or “list” price. Within this class of agreements, we define a loyalty discount as a contract in which B receives a discount from the list price in exchange for B’s commitment to devote a given share (s) of its purchases in category C to the products of S. The resulting nonlinear pricing schedule may have one or several steps that result in larger discounts for greater shares. Such commitments may be individually negotiated between the seller and a particular buyer, so that different buyers may have different deals, or they may be outcomes along a single pricing schedule announced by S in which buyers may choose different price-share combinations. Or both may occur.27 At the extreme, a share commitment of 100% by buyer B is a negotiated exclusive supply contract. Note, however, that a 100% share commitment is less restrictive than oft-analyzed exclusive dealing contracts, which would specify that S will be the exclusive supplier to B under all circumstances for the duration of the contract. Here, even a 100% commitment may be abandoned by B at any time, albeit by paying a higher price. Why would a buyer and seller prefer to specify a share of purchases instead of a particular quantity? There are at least two major reasons. The first is heterogeneity among buyers. In our analysis above, we assumed that a buyer’s increased purchases from S resulted in a one-for-one reduction in purchases from R, which is a reasonable characterization in many procurement settings. Then any contract specifying a quantity that B will purchase is a share contract. Specifying contracts in terms of shares is especially useful when a seller supplies hundreds or even thousands of buyers of different sizes, and separately negotiated contracts are not cost effective. For example, J&J supplies endoscopic surgical tools to thousands of hospitals and clinics in the United States. In figure 5.1 the existence of a mutual gain from contractual discounts does not depend on the size of the buyer; all can gain from such arrangements. A simple and cost-effective way of achieving these gains is for J&J to offer discounts from the list price depending on

27 

For example, J&J’s pricing schedule for its endoscopic surgical tools offers greater discounts in exchange for greater share commitments from buyer hospitals. Any small or large hospital can avail the schedule, but many chains and group purchasing organizations (GPOs) negotiate separate deals, typically with still larger discounts and share commitments.

Competitive Discounts and Antitrust Policy   101

a hospital’s committed share of J&J tools. This contract is independent of scale, allowing small and large hospitals to benefit from the same share-based schedule of discounts.28 The second reason is demand uncertainty. When demand for the final product sold by the buyer is highly uncertain, it is inefficient to write an enforceable contract that specifies how much should be purchased in each state of the world. For example, suppose Intel’s contract with Lenovo specified a particular quantity of processors that Lenovo must purchase to qualify for a discount. If demand for Lenovo’s personal computers turns out to be unexpectedly low, then the quantity threshold to qualify for a discount should be reduced, which means that such contracts must be state contingent and verifiable. But such a state-contingent contract with an individual buyer would be nearly identical to a share contract, so long as total purchases within a product category (here, processors) are a good summary measure of the state of demand for a particular brand. Then a share contract achieves the goal of exploiting gains from trade, but with lower negotiation and monitoring costs (see, e.g., Dana and Spier 2001). In other circumstances a commitment contract specifies neither a particular quantity nor a share threshold, but rather favorable promotion or placement of a seller’s goods or other advantages that will increase sales. The challenged conduct in Brand Name Drugs29 was that drug manufacturers entered into agreements with large mail-order pharmacies such as Medco and Caremark that granted particular drugs preferred promotional status within a class of “therapeutic equivalents.” In exchange, the pharmacy received a discount from the list price of the favored drug.30 For example, Medco might have a contract granting favorable treatment to Glaxo’s antiulcer drug Zantac: if Medco received a prescription for other drugs within this class—the main rival drug at the time was Tagamet—Medco pharmacists would ask the prescribing physician to switch the patient to Zantac. Unusually for lawsuits challenging such practices, plaintiffs in Brand Name Drugs were not rival manufacturers (who had their own deals with other buyers, and were defendants) but rather brick-and-mortar drug stores that paid higher wholesale prices because they were unable to duplicate such promotional advantages.31 Similarly, food and beverage manufacturers negotiate with grocery stores for favorable placement

28  In the case of hospital procurement, hospitals and clinics typically purchase through “group purchasing organizations” (GPOs) that contract with manufacturers on behalf of member hospitals. The GPOs typically negotiate share-based commitment contracts and discounts with manufacturers, which are natural outcomes that can be applied to large and small members. 29  In re: Brand Name Prescription Drug Antitrust Litigation, Case No. 94 C 897 MDL 997 (U.S. District Court, Northern District of Illinois, Eastern Division 1996). 30  As mentioned earlier, similarly in J.B.D.L. Corp. v. Wyeth-Ayerst Laboratories, 485 F. 3d 880 (6th Cir. 2007) defendant Wyeth offered discounts (rebates) to PBMs in exchange for placing Wyeth’s conjugated estrogen product Premarin in the PBM’s “Core Formulary.” Wyeth’s discounts and contract provisions were found legal. 31  Mail order pharmacies supplied mainly drugs for chronic ailments, used over long periods. So there was time to contact the prescribing physician between receipt of a new prescription and when it must be filled. Walk-in pharmacies had less ability and incentive to do this, because a smaller share of their sales was in switchable categories and because patients were typically waiting.

102   Kevin M. Murphy, Edward A. Snyder, and Robert H. Topel or amounts of shelf space, in exchange for discounts on wholesale prices (see Klein and Murphy 2008, 2011).

5.4.  Potential Harm to Competition Much of the economics literature on nonlinear pricing and related vertical restrictions is concerned with the possibility of anticompetitive effects, mainly as a result of “exclusion” or “foreclosure” of rivals. These analyses implicate a variety of factors such as economies of scale, capacity constraints, and cost conditions.32 They also account for various characteristics of the practices in question: (i) whether the seller uses customer-specific quantity or share thresholds, (ii) whether discounts are on a single product, a product line, or are based on “bundles” of multiple products, and (iii) whether discounts are applied to all purchases (“first unit” discounts) or only to marginal purchases above the relevant threshold. It is important that any sense of anticompetitive exclusion of rival sellers due to quantity-commitment contracts must refer to factors that prevent rivals from effectively competing in the market as opposed to competing for particular buyers (Elhauge and Wickelgren 2012). Contracts that prevent or restrict a rival’s ability to sell to some buyers for a period of time, but do not impair the rival’s ability to compete—that is, do not drive the rival from the market or raise its marginal costs—do not impinge the rival’s ability to discipline market prices, including the prices paid by buyers who do not purchase from the rival and the terms that S must offer to buyers in order to induce them to sign a QCD. The main issues are easily illustrated in the context of the model presented above. Because the products of a contracting seller and its rivals are demand substitutes within buyers, it is clear that one seller’s commitment contracts must reduce the demand for rivals’ goods compared to a world without such contracts. Some rivals may be excluded and the terms that are offered must be competitive with those offered by the rival. For example, let rivals be perfectly competitive with rising supply price. Then the reduction in residual demand caused by S’s contracts will reduce the price rivals receive and the quantity they sell. Rivals’ profits are then lower, and high-cost producers may exit the market. The exit of some high-cost rivals is not, however, anticompetitive. Indeed, the resulting fall in the price charged by rivals will tighten the competitive constraint on S even though S’s market share is increased by the QCD. Nor will economic efficiency be reduced— absent commitment contracts by S, those sellers would be viable only because S priced above marginal cost, which distorts buyers’ choices away from S. Commitment contracts 32 

In an environment with simultaneous contracting, where buyers face costs of switching sellers and at least one seller is financially constrained, a ban on below-cost pricing is sufficient to prevent exclusion. See Ordover and Shaffer 2007. When negotiations between a buyer and sellers are sequential, below-cost pricing has no effect on consumer welfare. See Marx and Shaffer (2008).

Competitive Discounts and Antitrust Policy   103

reduce this distortion. The same is true if S faces differentiated rivals, some of which are more efficient than others. Some sellers may be driven from the market by S’s commitment contracts because demand for their products is reduced. Others are forced to reduce their prices (which benefits buyers) and earn lower profits. But this harm to competitors is not harm to competition. The same effects on rivals would be induced by any action by S that encourages buyers to purchase more of its product, such as an across-the-board price cut. Things can be different in the presence of fixed costs or economies of scale causing declining marginal costs, and some form of scale economy is a common element of models that generate welfare-reducing exclusion.33 To (again) keep things simple suppose that S competes with a single differentiated rival, R. Let R’s technology have constant marginal cost for output above a minimum efficient scale (MES) but infinite marginal cost for lower rates of output; an economy of scale. Absent commitment contracts, S and R would set Bertrand prices above marginal cost and divide the market, so long as in the resulting equilibrium R operates above MES. But as above this leaves unexploited gains from trade. So assume that buyers and S pursue these gains, resulting in QCD contracts with some buyers and a reduction in the residual demand for R. R will be driven from the market if its best response to S’s contracts pushes its sales below MES. Consumers and social welfare will be harmed, while S captures the market and earns “monopoly” profits. There is harm to competition. In this example, the welfare-reducing impact of S’s contracts is ancillary to their true purpose, which is to increase sales and unlock gains from trade. S’s conduct would have been the same if exclusion of R were impossible—that is, in the complete absence of scale economies. But change the example just slightly; assume that absent intent to exclude R, S’s contracts would leave just enough residual demand for R to operate above MES. If S is aware of R’s tenuous participation, it can exclude R by negotiating slightly higher quantity commitments and/or slightly more contracts with buyers.34 Then S’s conduct is (slightly) different than if exclusion were impossible. Its monopoly is gained as a result of intentional exclusion, not merely a competitive effort to increase sales and unlock gains from trade. These examples illustrate the central quandary of antitrust policy in dealing with quantity commitment discounts, including their share-based variant. We have demonstrated that such contracts are part of the competitive process and would be commonly used by buyers and sellers even if exclusion of rival sellers were impossible. In certain circumstances specifically involving rivals’ scale economies, these contracts could, in theory, harm competition by excluding rivals or raising their costs. Even if harm to competition could be demonstrated by some test, however, if this harm is ancillary to contract competition it is difficult to argue that a contracting seller should face

33  In the presence of scale economies in production, Segal and Whinston (2000b) show that exclusive contracts can prevent entry if a sufficient number of buyers agree to exclusive contracts. However, if buyers can breach the exclusive contracts and pay expectation damages, the contracts cannot prevent entry (Simpson and Wickelgren (2007)). 34  Similarly, if R would operate just above MES in the absence of contracts, then S might offer contracts even if they would not be profitable in the absence of an ability to exclude R.

104   Kevin M. Murphy, Edward A. Snyder, and Robert H. Topel antitrust liability for the outcome, or even that the seller’s contracting practices should be enjoined ex post. And how does one prove that anticompetitive effects are not ancillary to normal competition? Further, what exactly would be illegal—commitment contracts themselves, so that certain sellers in certain circumstances may not offer them, or the act of having too many or too aggressive contracts? It is clear that in some circumstances the intent of QCD contracts could be to harm competition. But given that QCDs are by definition intended to increase S’s sales and therefore reduce R’s sales, it is difficult if not impossible to reliably identify intent from S’s contracting practices. If QCD contracts are to be judged by a rule of reason, then courts and antitrust authorities must be armed with analytical tools that can distinguish exclusionary intent from the intent to simply win sales from rivals, and businesses must be able to reliably predict when their conduct will run afoul of the law or be exposed to costly antitrust scrutiny.35 One possibility is to infer intent from the data and conduct: similar to pricing below cost in the analysis of predation, certain conduct might be profitable only if its purpose is to exclude rivals and harm competition. Our view is that such fine distinctions are nearly impossible in practice, as explained below.

5.4.1.  Implicit or Explicit Bundling Consider a simple QCD contract in which a particular buyer commits to maintain its category purchases of S’s product above some designated quantity, QD, in exchange for a discounted price PD.36 We aren’t concerned with situations where the discounted price itself is below marginal cost, so we assume PD > K, as implied by figure 5.1. Since we are analyzing competition for the business of a single buyer, we can ignore scale economies for the moment. It is then clear that an equally efficient (identical cost and value) rival cannot be excluded from competing for the entirety of QD at a price above its marginal cost. It follows that in order for a contract to prevent a rival from competing for the business of a particular buyer there must be some fraction of S’s sales to the buyer that are (effectively) not “at risk” or are noncontestable—that is, that the rival cannot capture.37 This can occur for two closely related reasons. First, in the case of “single-product” commitment contracts, the fact that the buyer purchases a mixture of products from S and from rivals implies within-buyer heterogeneity of brand preferences—a demand for variety represented by downward-sloping 35 

Further, the economic literature on game-theoretic models of firm behavior has provided little in the way of guidance for antitrust law. See Kobayashi and Muris (2012). 36  To translate this analysis into a share contract, simply normalize the buyer’s total purchases to unity, so QD is the share of total purchases that come from S. 37  Some have argued that the necessity of having some units that are not “at risk” means that single-product discounting practices should be per se legal, because all units are evidently at risk and anticompetitive impact is impossible. Our analysis in the following paragraphs shows that this argument is not quite correct, but per se legality is probably a good rule because anticompetitive impact would be nearly impossible to establish in the case of single-product QCDs.

Competitive Discounts and Antitrust Policy   105

demand in figure 5.1. As explained above, for all commitment contracts that we know of, the “buyer” is better described as a purchasing agent or middleman for a group of final purchasers or users. For example, a hospital that purchases J&J surgical tools acts as a purchasing agent for staff surgeons who use them. Then a subset of surgeons in a hospital may have a strong preference for J&J’s brand of tools and be unwilling to switch, which affects the hospital’s purchasing mix. Similarly, Intel’s processors may be particularly well suited to a subset of an OEM’s product line, so those units are more difficult for a rival to displace than other marginal units. Then one might argue that a committed buyer purchases an implicit bundle of “contestable” and “noncontestable” units, even if all units purchased from S are physically identical. The second circumstance is actual heterogeneity of the units used to calculate the threshold quantity or share. For example, J&J’s share contracts for “endoscopic tools” encompass a variety of instruments used in that type of surgery, and some rivals only produce substitutes for a subset of the line.38 Then J&J’s contracts are closer to an explicit bundle of heterogeneous products that are physically and economically distinct on the supply side, only some of which may be “contestable” by a particular rival. Taking things one step further, J&J’s customers can avail further discounts on endoscopic tools by achieving an additional threshold on purchases of J&J’s popular line of sutures—an entirely different product line that is also discounted—which means that J&J offers an explicit bundle of commitment contracts.39 Similarly in Lepage’s, 3M’s discounts were conditioned on quantity thresholds for a range of 3M products, not simply its Scotch brand of transparent tape, a private-label version of which was sold by plaintiff Lepage’s.

5.5.  Potential Tests for Harm to Competition 5.5.1.  The Attribution Test for Exclusion Some form of bundling—either explicit or implicit—is necessary for S’s commitment contracts to exclude a hypothetical equally efficient rival from competing for the business of a particular buyer, B. To see this, assume that QN of B’s committed purchases 38 

Bundles are very common in the procurement of medical supplies, where GPOs negotiate commitment contracts on behalf of member hospitals that include multiple product categories. 39  Why would a seller offer a bundle of commitment contracts in which discounts on product X are partially conditioned on purchases of product Y? One answer is a form of selection or price discrimination in which willingness to substitute in many product categories is a buyer-specific trait. Then those most willing to substitute receive bigger discounts in exchange for larger quantity commitments in the bundled categories. Further, if buyers have heterogeneous tastes for elements of the bundles offered by competing sellers, then bundles are better substitutes than are individual products. Then competition in bundles may be more aggressive, benefiting buyers.

106   Kevin M. Murphy, Edward A. Snyder, and Robert H. Topel from S are “noncontestable” in that they cannot be displaced by rival R—either because some users cannot reasonably be induced to switch or because R does not produce the products in QN and cannot form an implicit or explicit joint bid with other sellers that do. On these units a buyer purchasing on the discount contract pays a discount from the list price, PDN = PLN − d N . We assume that the remaining QC of B’s committed purchases are “contestable” by R and we denote S’s discounted price for these units byPDC .40 We then ask: under what conditions would S’s contract exclude an equally efficient rival from profitably competing for the contestable units, QC? We assume that “equally efficient rival” means that B values the contestable units of S and R comparably, and that the sellers have identical marginal cost K of supplying these units. Interpreting the bundled contract literally for the moment, assume that if B purchases the contestable units from R it loses its discounts and must pay S the list price PLN for the noncontestable good. With these assumptions, R’s lowest feasible offer to supply the contestable units is a price equal to marginal cost, which would reduce B’s costs of purchasing the contestable units by [PDC − K ]Q C . But by purchasing from R the buyer sacrifices discounts on the noncontestable goods, so it isn’t enough to simply undercut S’s price. R’s offer must also offset the buyer’s loss of surplus on those purchases, denoted LN. So R can profitably compete if [PDC − K ]Q C − LN ≥ 0 , which we write on a per-unit basis as



PDC −

LN ≥K. QC

(5.1)

The left side of inequality (5.1) is the highest price that B will find attractive, which must exceed R’s marginal cost in order to be feasible. This bound is below S’s contract price because B must be compensated for sacrificing discounts on noncontestable purchases. Equation (5.1) is not yet in usable form because we haven’t specified LN, the lost surplus on noncontestable units if the buyer purchases contestable units from R. Figure 5.2 shows this loss of surplus. Absent a contract and discount dN we assume that the buyer chooses quantity Q N on the demand curve, but the discount may require a quantity commitment that is off the demand curve, Q N > Q N . The buyer’s loss of surplus is the difference in shaded areas:



LN = L1 + L2 − L3 = d N Q N − (L3 + L4 ) 1 = d N Q N − [PLN − v N ][Q N − Q N ], 2

40  This setup is consistent with either explicit or implicit bundling if QN represents units of a separate product from QC. If the latter, QN represents units of a single product that are noncontestable by R due to buyer preferences for variety.

Competitive Discounts and Antitrust Policy   107

where we have assumed that demand is approximately linear over the relevant range. If the contract sets QN efficiently then vN = KN and condition (5.1) becomes PDC −



QN QC

 N 1 N N d − 2 m η  ≥ K ,  

(5.2)

where mN = PLN − K N is S’s list price markup on noncontestable units and ηN = [Q N − Q N ]/ Q N is the percentage reduction in B’s purchases when the discount is removed. Equation (5.2) is the condition that determines whether an “equally efficient” rival can profitably compete for the contestable units, given our maintained assumptions that a noncompliant buyer loses the full amount of the discount dN specified in the contract and that QC and QN are known. As indicated above, if practiced with many buyers and if R requires sufficient scale of operations in order to effectively compete, the cumulative effect of such agreements could be to exclude R from effectively competing in the market, not just from selling to an individual buyer or buyers. But even with these (extreme) assumptions, condition (5.2) is difficult to apply as a test of exclusion because it is generally unknown how much less a buyer would purchase Price

P NL L1

L2

P NL − d N

L4

L3

vN QN (P,PR )=v(Q,PR )

QN N

N

QN

Quantity

The list price is PL with associated quantity Q . The QCD contract offers discount d N with committed quantity Q N. At this quantity the buyer’s marginal value is v N. A buyer who switches to rival R sacrifices discounts and loses surplus 1 L1 + L2 − L3 = L1 + L2 + L4 − (L3 + L4 ) = d N Q N − (L3 + L4 ) = d N D N − [PLM − νN ][Q N − Q N ]. 2 FIGURE 5.2  

108   Kevin M. Murphy, Edward A. Snyder, and Robert H. Topel in the counterfactual where discounts are removed, ηN . One possibility is simply to ignore the buyer’s ability to mitigate the lost discounts by assuming ηN = 0; the buyer will purchase the same quantity of noncontestable units from S regardless of price. This yields the so-called attribution test:



PDC − d N

QN ≥ K. QC

(5.3)

The left-hand side of (5.3) is the upper bound on offers from R that would be acceptable to a buyer that does not mitigate forgone discounts. It is typically interpreted as a “net” price for the contestable units, having “attributed” discounts granted on the noncontestable units to the contestable ones.41 Failure to satisfy (5.3) is interpreted as evidence that an equally efficient seller of the contestable units cannot profitably compete against S’s contract. Notwithstanding other flaws that we discuss below, comparison of (5.3) to (5.2) indicates that the attribution test is too likely to indicate exclusion. By assuming away mitigation, it overstates the compensation that a rival must provide to offset lost discounts. A second interpretation of the attribution test is that it determines whether a contracting seller is pricing the contestable units “below cost” to exclude equally efficient rivals. This is a form of the common “profit sacrifice” test. This may at first appear to be the same question addressed by condition (5.2), but it is not. On this interpretation, a contracting seller is “pricing below cost” if discounts would be unprofitable but for their ability to exclude, similar to predatory pricing. Discounts will be profitable in the sense that incremental revenues exceed incremental cost if the discounted price on contestable units covers incremental cost and any loss of profit, ∆Π N , from selling the noncontestable units at a discount. This yields a condition similar to (5.1):



PDC −

∆Π N ≥ K. QC

(5.4)

In condition (5.1) the discounted price is offset by the reduction in buyer’s surplus when discounts are removed on noncontestable purchases, LN. This loss is always positive because the buyer is harmed by a higher price. In contrast, the offset in (5.4) is the sacrifice in profit by S from selling noncontestable units at a discounted price. This “sacrifice”

41  Equation (3) can be evaluated under two cases. First, if all units are contestable, then the ratio, QN/ QC, in the left-hand side approaches zero and the equation reduces to a test of whether price is above incremental cost. Second, if all the discounts should be applied to the contestable units, then QN/ QC approaches one and the equation reduces to a test of whether net price, after full attribution, is above incremental cost.

Competitive Discounts and Antitrust Policy   109

need not even be positive because the discount increases sales of the noncontestable good. Indeed, if the list price for noncontract purchases is near a profit maximum this effect will be about zero, and if discounts on the noncontestable good are profitable, as in figure 5.1, then the offset is negative. For both of these cases, (5.4) is automatically satisfied. More generally LN > ∆Π N and some algebra establishes that discounts are profitable if PDC −



QN N d − mN ηN  ≥ K . QC 

(5.5)

Condition (5.5) is in the same form as the “exclusion” condition (5.2). But inspection of (5.2) indicates that the buyer’s mitigation of lost discounts gets smaller weight (1/2) in condition (5.2) than in condition (5.5) for the reasons just stated. Since mN > 0 and ηN > 0, equation (5.5) implies that discounts that pass the attribution test are always profitable, but profitable discounts may fail the attribution test. Indeed, profitable discounts are likely to fail, especially when list prices are close to the stand-alone profit-maximizing price where d N − mN ηN ≈ 0. Unlike the usual analyses of predation—which involve temporary sacrifice of profit by pricing below cost—this analysis means that QCD contracts that are profitable in the absence of any possibility of “exclusion” may nevertheless exclude an equally efficient rival from competing for a buyer’s business. Discounts that “exclude” and discounts that sacrifice profits are different things.42 As above, if we assume that buyers do not mitigate forgone discounts by purchasing less or switching to alternatives suppliers ηN = 0 , then (5.5) also reduces to the attribution test (5.3). Only then are discounts that “exclude” an equally efficient rival equivalent to a seller sacrificing profits by setting a “net price” for contestable units below marginal cost. Note that in either (5.2) or (5.5) the conditions are satisfied if all units are contestable Q N = 0 . Hence our point that some form of implicit or explicit bundling is essential if an equally efficient rival is to be “excluded” from competing for a buyer’s business.43 We return to this point below in the context of “single-product” QCDs and the possibility of inferring harm to competition.

(

(

)

)

42  A related interpretation of (5) and (2) is that exclusion can be “cheap” for a seller with substantive margins on the noncontestable good, because discounts from the profit-maximizing price are attractive to buyers and are compensated by greater sales. Further, a seller satisfying (5) may nevertheless be sacrificing profits if smaller discounts and/or quantity commitments would be profit maximizing, but somewhat larger discounts and commitments are chosen in order to exclude. 43  We have assumed that all contestable units from S and rivals are perfect substitutes. If they aren’t then B’s demand for S’s version of the contestable units is downward sloping—a taste for variety—and some units will be more costly for R to capture, as if more units are noncontestable. We think this strains the definition of “equally efficient” because it implies that R cannot reproduce some attributes of S’s product that buyers value.

110   Kevin M. Murphy, Edward A. Snyder, and Robert H. Topel

5.5.2.  The Attribution Test and Liability Thresholds The attribution test is the first leg of the 2007 AMC’s three-part test for identifying discounting practices that may violate Section 2 of the Sherman Act: Courts should adopt a three-part test to determine whether bundled discounts or rebates violate Section 2 of the Sherman Act. To prove a violation of Section 2, a plaintiff should be required to show each one of the following elements (as well as other elements of a Section 2 claim): (1) after allocating all discounts and rebates attributable to the entire bundle of products to the competitive product, the defendant sold the competitive product below its incremental cost for the competitive product; (2) the defendant is likely to recoup these short-term losses; and (3) the bundled discount or rebate program has had or is likely to have an adverse effect on competition. (See, e.g., Antitrust Modernization Commission 2007; emphasis added)

Rather than an “exclusion test” in the sense of (5.1) and (5.2), these requirements are basically a rule-of-reason test for predatory pricing. Here, a price “below its incremental cost” is determined by the logic of (5.5) with the additional assumption that ηN = 0, resulting in the attribution test (5.3) rather than a simple comparison of unit price and cost. The “each one” requirement means that contracts passing any one of the three tests would enjoy safe harbor from antitrust liability. Thus the AMC requirements at least appear to provide clear rules and, hopefully, a filter that might reliably identify anticompetitive conduct without itself dampening competition. The AMC requirements are similar to the position of the DOJ in its 2008 Section 2 guidelines, which would also grant safe harbor to discounts passing the attribution (“discount-allocation”) test: The Department believes that, when actual or probable harm to competition is shown, bundled discounting by a monopolist that falls outside the discount-allocation safe harbor should be illegal only when (1) it has no procompetitive benefits, or (2) if there are procompetitive benefits, the discount produces harms substantially disproportionate to those benefits. (US Department of Justice 2008)

We regard condition (1) above as moot because, as we explained in subsection 5.3.1, virtually all QCD practices have a procompetitive benefit. The DOJ threshold for illegality of discounts failing the attribution test in condition (2) is higher than that of the AMC, requiring not simply that harm is likely (AMC: “has had or is likely to have an adverse effect on competition”) but that the harm be both demonstrable (“produces”) and disproportionate to benefits.

Competitive Discounts and Antitrust Policy   111

5.5.3.  Flaws with the Attribution Test, and Its Potential Uses The attribution test is superficially attractive because it appears to determine whether a seller’s discounts are unprofitable, which might indicate predation. Further, comparison of (5.5) and (5.2) indicates that unprofitable discounts would exclude an equally efficient rival under the assumptions commonly used, that is, no mitigation. But the test suffers from a number of major flaws, each of which tends to bias the test toward failure, and so toward the exposure of procompetitive or competitively neutral discounting practices to antitrust scrutiny and litigation. These flaws derive from the inability of contract and sales data available in litigation to reliably estimate (i) the relative quantities that are non-contestable by rivals (QN/QC) and (ii) the “but-for” terms of trade between buyer and seller that would exist in the absence of the observed contract, particularly the magnitude of discounts on noncontestable units that a buyer would lose by switching its business to a rival (dN). As a threshold matter, it is worth emphasizing that the attribution test is meant to provide evidence that an equally efficient rival would be excluded from competing for a buyer’s business. The catalyst for this inquiry is typically a complaint by a particular rival that alleges exclusion, so it is sensible to ask whether there is direct evidence that the defendant’s contracts actually exclude the plaintiff—that is, that customers would purchase substantially more of the plaintiff ’s product but for the defendant’s contracts. For example, among buyers that do not purchase under the defendant’s contracts, is the plaintiff ’s share materially higher? If not, then it is difficult to argue that the plaintiff is equally efficient in the sense of producing goods that can replace the defendant’s sales, or that defendant’s contracts are the cause of the plaintiff ’s poor performance. We have already noted the test’s implicit assumption that buyers do not mitigate N the impact of loss of discounts (η = 0), or equivalently that the cost of the discounts to S are not reduced by the enhanced sales of noncontestable units generated by the discounts. Whether the test is interpreted as an indicator of exclusion as in (5.2) or below-cost pricing as in (5.5), this omission means that the test will generate false positives, suggesting anticompetitive impact or conduct when there is none. Put differently, if a seller’s contracts pass the attribution test then discounting practices are almost certainly “above cost” in the sense of (5.5). And while above-cost contracts may nevertheless exclude, or even be intended to exclude, there is clear danger that litigating profitable discounting practices would have a chilling effect on competition. The AMC’s safe harbor for such contracts is therefore warranted, though in our view, too many contracts would be left exposed. We are also concerned that existence of such a “guideline” will encourage sellers to satisfy it, avoiding more aggressive discounts and commitments that would otherwise enhance competition and benefit consumers. Even with the AMC safe harbor, a useful refinement would be to require evidence that

112   Kevin M. Murphy, Edward A. Snyder, and Robert H. Topel a seller actually is sacrificing profit on the noncontestable good, that is, that the bracketed term in (5.5) is positive. And of course, if a reliable estimate of ηN is available, it should be used. For example, if there are substantial sales at list price one might presume that the list price is profit maximizing. Then one can infer that the bracketed term is nonpositive, at least for small discounts. Our discussion to this point has assumed that the relative quantities of “noncontestable” and “contestable” sales (QN/QC) are known or accurately estimated, but this is rarely the case. The issue of reliably specifying QN/QC is especially problematic in the case of single-product QCDs. As explained above, the existence of such contracts generally implies a buyer-specific taste for variety (downward-sloping demand for a brand) within a product category. Then some units are more difficult to contest than others and the contract specifies an implicit bundle. Yet the very notion of “noncontestable” units in the single-product case strains the definition of an equally efficient rival. “Equally efficient” cannot simply mean “equal marginal cost” regardless of how users view the rival’s product—to qualify as equally efficient in an economic sense, the rival must also offer equal value to consumers. If so, then all units should be viewed as contestable in the single product case, and we have already seen that if Q N = 0, the attribution test is automatically satisfied so long as total sales are profitable. And even if we accept for the sake of argument that some units might not be easily contested, there is no scientifically reliable method for determining the relative number of noncontestable units—in Concord Boat, how many engines “must” a buyer have purchased from Brunswick?44 We conclude that Q N = 0 is the practically correct assumption for single-product QCDs, in which case (5.3) simply asks if price is above cost and the usual analysis of predation can follow. We then agree with Areeda and Hovenkamp (2007) that single-product QCDs should enjoy per se legal status so long as price exceeds a reasonable measure of incremental cost (Areeda and Hovenkamp 2012, chap. 17). A particularly vivid example of misapplying the “equally efficient” concept in defining contestable shares is provided by the DOJ’s recent application of the attribution test in United Regional Health Care (2011). United Regional is the largest hospital in Wichita Falls, Texas, and it has entered into QCD contracts with a number of private insurers. The DOJ calculated that noncontract buyers Blue Cross, Blue Shield and Medicare purchased only 10% of units within contestable categories from rival hospital Kell West, because “many patients are likely to choose care at United Regional even for services that competing providers offer.”45 So the DOJ assumed that 90% of products offered 44  The record in Concord Boat indicated that two previously compliant buyers switched all their purchases to rival sellers, indicating that all units were contestable, at least for those buyers. 45  Competitive Impact Statement filed by the U.S. Department of Justice, United States of America and State of Texas v United Regional Health Care System, United States District Court for the Northern District of Texas, February 25, 2011, p. 16. It is noteworthy that “foreclosure” in the government’s view applies to a rival (Kell West) that has participated in the alleged market since 1999. The DOJ’s brief goes so far as to calculate how much more profitable Kell West would be if it were to capture some of the business that currently goes to United.

Competitive Discounts and Antitrust Policy   113

by both United Regional and Kell West were, in fact, noncontestable because consumers would not purchase them from Kell West regardless of discounts. Then discounts received on all purchases were subtracted from the 10% that Kell West could allegedly contest, and it should come as no surprise that United’s discounts were found to “fail” the attribution test. The clear implication of the DOJ’s analysis is that a seller must tailor its discounts to accommodate the capabilities of its rivals, including lack of acceptance of rivals’ products by consumers. There could hardly be a better example of protecting competitors instead of competition. Things are only slightly more promising in the case of explicit bundles of physically distinct products. In many cases, a particular plaintiff who complains of exclusion may produce only a small portion of the product lines that led to the seller’s discount policy, so the “QC” for that particular rival actually is small. In suits challenging J&J’s bundled loyalty discounts on endoscopic tools and sutures, small manufacturers that did not produce sutures and could replace only a small fraction of the J&J endoscopic line (mainly tools called trocars) alleged that they could not profitably overcome the discounts that buyers might lose on all other purchases from J&J. Applied to what a single rival could displace, condition (5.3) would likely fail. But the QC relevant for antitrust scrutiny is the number of units open to competition from all rivals, not simply a one-off test of whether a buyer would be willing to switch a portion of its purchases to one rival holding constant all other purchases. Narrowly specialized plaintiffs should not enjoy preferred status in making antitrust claims, just as plaintiffs that do not meet consumer acceptance should not. In J&J the entire lines of endoscopic tools and sutures were subject to competition from another full-line supplier (Tyco), as well as from combinations of several specialized sellers.46 Then competition in the presence of QCD contracts is “bundle-to-bundle”—including the possibility that buyers or their agents may create virtual bundles from combinations of sellers, including the plaintiff. Then it is proper to think of all or almost all units as contestable, so Q N = 0 even if a particular rival can only replace a small portion of a buyer’s purchases from S. As in the single-product case the question becomes whether the “price” of the entire bundle exceeds its cost, so the usual predation analysis can be applied to the bundle as a whole. These arguments narrow the cases where Q N > 0 down to those where (a) the QCD contract consists of an explicit bundle of physically distinct products and (b) no rival or rivals produce reasonable substitutes for some elements of the bundle. Inspection of (5.3) indicates that failure to pass the test is more likely when relative number of units that might be contested by a plaintiff is small (QN/QC is large). Hence allegedly excluded plaintiffs emphasize that a committed buyer wishing to purchase a small amount “QC” of their products must forgo discounts on all its purchases from the discounting seller. But applying the test in this way makes no sense, because any discrete discount—say a 46  Tyco and J&J engaged in bundle-to-bundle competition, and each offered loyalty discounts. In recognition of this, J&J ultimately “carved out” smaller suppliers from share calculations, which were then based only on purchases from “full-line” suppliers of surgical inputs—J&J and Tyco.

114   Kevin M. Murphy, Edward A. Snyder, and Robert H. Topel 10% discount if S’s share exceeds 80%—will fail the test when the number of contested units is sufficiently small, as demonstrated by the DOJ’s analysis in United Regional Health Care. In other words, the QC in equation (5.3) must account for the best offer the plaintiff might make, which for an “equally efficient” rival is to supply all of the contestable units. In addition, if S’s contract specifies a share threshold of (say) 80% to qualify for discounts, that leaves 20% “headroom” for which any rival may compete without triggering any loss of discounts. To the extent that those units are also contestable by the plaintiff but are now supplied by other rivals, the contracting sellers’ discounts are not binding on the plaintiff ’s ability to sell more. And even if those units are not directly contestable by the plaintiff—say because they are products the plaintiff doesn’t make— the buyer could still purchase the full amount of headroom from the plaintiff if the units now supplied by others are substitutes for products supplied by S. The general principle should be to judge what sales the rival can profitably compete for using its most effective strategy (e.g., competing for all the sales in the headroom to avoid a loss of discounts or competing for all contestable units in order to spread any loss of discounts over as many units as possible). When the rival can profitably compete for all contestable sales to a given buyer, even if the rival could not compete for smaller quantities, we should conclude that the rival is not foreclosed. This principle also applies to the time dimension of the shift in sales. The question is whether the rival has the ability to compete for the sales, not whether it would be profitable to win those sales on a temporary basis. For example, if the rival must win 100 units annually in order to make matching the loss of discounts profitable, a conclusion that the buyer could switch 80 units in year one, 150 units in year two and 200 units in year three would mean the seller would not be excluded at any reasonable discount rate, even though the shift in sales would not be profitable on the basis of year one alone. A second issue is measuring the loss of discounts that the buyer would suffer, that is, how much more a buyer would have to pay if it wished to purchase less than its committed quantity or share from S. In a naive application of (5.3), the discount dN is the difference between the “list price” of noncontestable units, PLN , and the agreed-upon commitment price written in a contract—the assumption being that if a buyer wished to purchase less than the committed amount it would actually pay the list price. But contract prices and quantities are often the result of direct negotiations between seller and buyer, and contracts often have short durations. Thus Intel’s negotiations with individual OEMs resulted in buyer-specific agreements and associated percentage discounts from Intel’s menu of list prices for processors. If innovations or aggressive pricing by AMD caused an OEM to want more of AMD’s processors, its next round of negotiations with Intel could specify a smaller Intel volume or share and, perhaps, a smaller discount from the list price benchmark. But the OEM would generally not pay the list price, which is to say that the “dN” contained in a particular contract is just a convenient way of specifying the price that will be paid for the indicated quantity or share relative to a benchmark, and it overstates what would be lost if the OEM wished to purchase less from Intel. Notwithstanding its other flaws, then, proper application of (5.3) (or other variants) requires the trier of fact to estimate the outcome of individual negotiations,

Competitive Discounts and Antitrust Policy   115

and how much “dN” would actually change if a buyer wished to purchase less from the defendant.47 Some information on this might be available from econometric study of changes in contracts over time, or by comparing contracts negotiated by different buyers, which will keep the economics experts busy. But these sources are unlikely to replicate the needed conceptual experiment, which is to ask: “what if a particular buyer wants to purchase less than the amount specified in its current contract?” While the case of bundling discrete products makes measuring the relative volume of noncontestable and contestable units, QN/QC, somewhat easier, it exacerbates the problem of measuring the loss of profits from discounting and/or the buyer’s ability to mitigate. In the case of single-product QCD, the noncontestable units are a subset of all units purchased, so it might be reasonable to assume that the demand for them is inelastic at the list price because the seller is unable to price them separately. However, when the uncontestable units are sold as a separate product on a stand-alone basis, the list price is likely to be closer to the profit-maximizing price for the seller. As we discussed above, under these conditions, the profit sacrifice from a discount is likely to be small or may even be negative. Similarly, buyers will be able to mitigate their loss by purchasing less of the uncontestable product according to equation (5.2). The attribution test presumes that failure to satisfy (5.3) is evidence that the seller is forgoing current profits—it is pricing below cost. Putting aside obviously benign examples of below-cost pricing such as promoting new products, driving future scale economies, or balancing incentives in two-sided markets, the fact is that a bundled pricing scheme that fails the attribution test is consistent with both profit maximization and enhanced efficiency. For example, bundled contracts are often forms of self-sorting price discrimination in which the list price is paid by relatively inelastic demanders who are less willing to substitute. In such circumstances providing enhanced discounts to more price-sensitive customers can increase sales and social surplus. The relevant question to ask when assessing whether contracts reflect “below cost” pricing is whether the seller has sacrificed profit relative to a but-for world without QCDs, in which case the seller’s price would not, in general, be the list price charged to noncontract buyers. In most reasonable cases it will be lower, which means the attribution test would be heavily biased against a defendant seller.48 47  The point remains relevant in cases where a seller offers a pricing schedule that may be availed by any buyer—which saves on transactions costs. Individual buyers may negotiate separate off-schedule deals and “carve-outs” for particular purchases. For example, in procurement of medical supplies for hospitals standard pricing schedules of manufacturers are negotiated through GPOs. In J&J, individual hospitals negotiated quantity carve-outs for particular surgeons who insisted on using a particular rival’s product, and large hospitals and chains often negotiated separate contracts with larger discounts or smaller commitments. Then the generally available pricing schedule is a sort of starting point for buyer-seller negotiations rather than a constraint. 48  The failure of the attribution test in situations where exclusion is not possible and bundled discounts are used to price discriminate was noted by AMC commissioners Carlton and Garza, who stated their concern that the test would subject innocent pricing schemes to undue scrutiny. In these situations, incremental revenue is not properly calculated in the Commission’s recommendation, which is the point of our examples. See Antitrust Modernization Commission (2007) p. 99.

116   Kevin M. Murphy, Edward A. Snyder, and Robert H. Topel These failures of the attribution test are not isolated or contrived examples. If used as anything but a safe harbor, it is certain that the test would implicate common and procompetitive pricing and procurement practices and, at least, subject those practices to further antitrust scrutiny. This type of bias no doubt motivated the AMC’s (2007) and DOJ’s (2008) recommendation that contracts passing the attribution test be given unambiguous safe harbor. But even this narrow exemption is evidently too lenient and specific for the current DOJ, which eschewed such safe harbors in its withdrawal of the 2008 DOJ report and has provided no other guidance beyond the notion that “contracts that reference rivals” are of particular concern. As far as we can tell, these concerns are based on theoretical possibilities rather than compelling evidence of anticompetitive effects. The DOJ’s failure to endorse the attribution test as a safe harbor is particularly troubling in light of the fact that it is essentially a predation test for below-cost pricing that is heavily biased in favor of a positive finding. If pricing that is demonstrably above cost is not immune, the implication is that QCDs will face scrutiny and litigation in situations where even allegations of predation would not be deemed credible. Given the clear consumer benefits that flow from firms competing aggressively for sales, we believe it is a mistake to condemn discounting practices that are profitable and hence that firms would employ absent any intent or ability to exclude rivals. In this light, we regard the DOJ’s withdrawal of safe harbor status for above-cost single-product and bundled discount schemes as a serious policy error.

5.5.4.  Other Indicators of Potential Competitive Harm Beyond the attribution test, the AMC suggested two other necessary conditions for antitrust liability. Condition (2) is that “the defendant is likely to recoup these short-term losses” and condition (3) is that “the bundled discount or rebate program has had or is likely to have an adverse effect on competition” (see Antitrust Modernization Commission 2007). Our analysis above demonstrated that all of these conditions can be satisfied by nonpredatory and profitable QCDs that would be utilized in the absence of a possibility to exclude. In considering the efficacy of this and similar policies, the consequent efficiency loss that would be caused by condemning procompetitive discounts might be acceptable if it could be demonstrated that real-world discounting practices of the type being challenged have caused material harm to competition. But convincing examples of such harm are conspicuous by their absence. If we judge potential antitrust policies themselves by a rule-of-reason balancing of harms and benefits, it is difficult to make a case that the AMC filter or any other that we know of is useful for inferring liability. At the same time, the theoretical possibility of harm makes per se legality of all QCDs an unattractive policy—paraphrasing Potter Stewart’s inability to state a legal threshold for pornography, “we’ll know it when we see it.” In the end, the DOJ’s (2008) disproportionality standard may be the best we can do.

Competitive Discounts and Antitrust Policy   117

5.6.  Concluding Remarks Antitrust policy should put a heavy evidentiary burden on plaintiffs to demonstrate clear anticompetitive effects that clearly outweigh procompetitive benefits of challenged contracts. In this sense, our views are close to the DOJ’s 2008 disproportionality standard, and are perhaps more stringent. For this class of conduct at least, the DOJ’s 2009 decision to abandon the disproportionality standard was a mistake that has muddied the waters for effective antitrust policy, and likely itself reduced competition.

Acknowledgments We gratefully acknowledge comments, insights, and assistance offered by Shannon Seitz, Pierre Cremieux, Marc Van Audenrode, and the editors.

References Antitrust: Commission Imposes Fine of €1.06 bn on Intel for Abuse of Dominant Position; Orders Intel to Cease Illegal Practices, European Union press release, May 13, 2009. Retrieved from http://europa.eu/rapid/press-release_IP-09-745_en.htm?locale=en. Antitrust Modernization Commission. 2007. Report and Recommendations. April. Retrieved from http://govinfo.library.unt.edu/amc/report_recommendation/amc_final_report.pdf. Areeda, Phillip E., and Herbert Hovenkamp. 2007. Antitrust Law:  An Analysis of Antitrust Principles and Their Application. 2007 Supplement. New York, New York: Aspen Publishers. Areeda, Phillip E., and Herbert Hovenkamp. 2012. Antitrust Law:  An Analysis of Antitrust Principles and Their Application. 2012 Supplement. New York: Aspen Publishers. Carlton, Dennis, Patrick Greenlee, and Michael Waldman. 2008. Assessing the Anticompetitive Effects of Multiproduct Pricing. Antitrust Bulletin 53 (Fall), 587–622. Dana, James, and Kathryn Spier. 2001. Revenue Sharing and Vertical Control in the Video Rental Industry. Journal of Industrial Economics 49(3), 223–45. Easterbrook, Frank. 1984. The Limits of Antitrust. Texas Law Review 63, 1–40. Elhauge, Einer, and Abraham L. Wickelgren. 2012. Anti-competitive Market Division through Loyalty Discounts without Buyer Commitment. John M. Olin Center for Law, Economics, and Business Discussion Paper no. 723. Retrieved from http://ssrn.com/abstract=2125428. Hovenkamp, Herbert. 2011. Quasi Exclusive Dealing. Retrieved from http://papers.ssrn.com/​ sol3/papers.cfm?abstract_id=1793126. Jacobson, Jonathan M., and Daniel P.  Weick. 2012. Contracts That Reference Rivals as an Antitrust Category. Program on “Debating the Competitive Benefits and Costs of MFNs,” ABA Section of Antitrust Law, Spring Meeting.

118   Kevin M. Murphy, Edward A. Snyder, and Robert H. Topel Kauper, Thomas, and Edward Snyder. 1991. Misuse of the Antitrust Laws:  The Competitor Plaintiff. Michigan Law Review 90, 551–603. Klein, Benjamin, and Kevin M. Murphy. 2008. Exclusive Dealing Intensifies Competition for Distribution. Antitrust Law Journal 75(2), 433–66. Klein, Benjamin, and Kevin M. Murphy. 2011. How Exclusivity Is Used to Intensify Competition for Distribution—Reply to Zenger. Antitrust Law Journal 77(2), 691–99. Kobayashi, Bruce H., and Tim Muris. 2012. Chicago, Post-Chicago, and Beyond: Time to Let Go of the 20th Century. Antitrust Law Journal 78(1), 147–59. Marx, Leslie M., and Greg Shaffer. 2008. Rent Shifting, Exclusion and Market-Share Contracts. Unpublished manuscript. Ordover, James, and Greg Shaffer. 2007. Exclusionary Discounts. CCP Working Paper No. 07-13. Salop, Steven C. 1986. Practices That (Credibly) Facilitate Oligopoly Co-ordination. In New Developments in the Analysis of Market Structure, 265–94. Joseph E. Stiglitz and G. Frank Mathewson, eds. Cambridge, MA: MIT Press. Schwartz, Marius, and Daniel Vincent. 2008. Quantity “Forcing” and Exclusion:  Bundled Discounts and Nonlinear Pricing. Issues in Competition Law and Policy, American Bar Association Antitrust Section. Scott-Morton, Fiona. 2012. Contracts That Reference Rivals. Speech presented at Georgetown University Law Center. Retrieved from http://www.justice.gov/atr/public/speeches/281965.​ pdf. Segal, Ilya R., and Michael D.  Whinston. 2000a. Exclusive Contracts and Protection of Investments. Rand Journal of Economics 31, 603–33. Segal, Ilya R., and Michael D.  Whinston. 2000b. Naked Exclusion:  Comment. American Economic Review 90(1), 296–309. Simpson, John, and Abraham L.  Wickelgren. 2007. Naked Exclusion, Efficient Breach, and Downstream Competition. American Economic Review 97(4), 1305–20. US Department of Justice. 2008. Competition and Monopoly: Single-Firm Conduct under Section 2 of the Sherman Act. Retrieved from www.usdoj.gov/atr/public/reports/236681.htm. Varney, Christine A. 2009. Assistant Attorney General, Antitrust Division, US Department of Justice. Vigorous Antitrust Enforcement in This Challenging Era. Remarks as Prepared for the United States Chamber of Commerce, May 12. Retrieved from http://www.justice.gov/atr/​ public/speeches/245777.pdf. Wilson, Robert B. 1993. Nonlinear Pricing. New York: Oxford University Press. Wright, Joshua D. 2011. Does Antitrust Enforcement in High Tech Markets Benefit Consumers? Stock Price Evidence from FTC v. Intel. Review of Industrial Organization 38(4), 387–404. Zenger, Hans. 2010. When Does Exclusive Dealing Intensify Competition for Distribution? Comment on Klein and Murphy. Antitrust Law Journal 77, 205–11.

Cases and Statutes 3M v. LePage’s, 2003 WL 22428377, at *1 (2003) (No. 02-1865). Applied Medical Resources Corp. v.  Ethicon Endo-Surgery (Johnson & Johnson), 03-CV-1329 (C.D. Cal. 2006). Clayton Antitrust Act of 1914 §§ 3-4, 15 U.S.C. 14. Concord Boat Corp. v. Brunswick Corp., 207 F.3d 1039 (8th Cir.), cert. denied, 531 U.S. 749 (2000).

Competitive Discounts and Antitrust Policy   119

Eaton Corporation v. ZF Meritor LLC and Meritor Transmission Corporation, cert. denied, 133 S. Ct. 2025 (2013). In re: Brand Name Prescription Drug Antitrust Litigation, Case No. 94 C 897 MDL 997 (U.S. District Court, Northern District of Illinois, Eastern Division 1996). J.B.D.L. Corp. et al. v. Wyeth-Ayerst Laboratories, Inc., et al. (6th Cir. 2007). LePage’s v. 3M, 324 F.3d 141 (3rd Cir. 2003) (en banc), cert. denied, 124 S Ct 2932 (2004). Masimo Corp. v. Tyco Health Care Group, L.P., No. CV 02-4770 MRP, 2006 WL 1236666, at *13 (C.D. Cal. Mar. 22, 2006). Ortho Diagnostics Sys. Inc., v. Abbott Lab, Inc., 920 F. Supp. 455 (S.D.N.Y. 1996). Sherman Act, 15 U.S.C. 1 et seq., § 2, 315 U.S.C. 2. SmithKline Corp. v. Eli Lilly & Co., 427 F. Supp. 1089 (E.D. Pa. 1976), affirmed 575 F.2d 1056 (3d Cir. 1978). Statement of Chairman Leibowitz and Commissioner Rosch In the Matter of Intel Corporation, Docket No., 9341, Federal Trade Commission, December 16, 2009. Retrieved from http://​ www.ftc.gov/os/adjpro/d9341/091216intelchairstatement.pdf. United States of America and State of Texas v United Regional Health Care System, United States District Court for the Northern District of Texas, February 25, 2011. Retrieved from http://​ www.justice.gov/atr/cases/f267600/267653.pdf. Virgin Atlantic Airways Ltd. v. British Airways PLC, 69 F. Supp. 2d 571 (S.D.N.Y. 1999), aff ’d 257 F.3d 256 (2d Cir. 2001).

CHAPTER 6

SQUEEZE CLAIMS Refusals to Deal, Essentials Facilities, and Price Squeezes BARAK ORBACH AND RAPHAEL AVRAHAM

6.1. Introduction In the marketplace, firms often find themselves helpless and squeezed. Actions taken by other firms weaken them and sometimes even threaten to drive them out of business. In the past, such realities heavily influenced antitrust courts. Many old antitrust cases expressed concerns regarding “[the interests of] small dealers and worthy men,”1 “the protection of viable, small, locally, owned businesses,”2 and the “helplessness of the individual.”3 To date, many judges believe that it is “axiomatic . . . that antitrust law should prevent unfairness in commercial dealings” (Areeda and Hovenkamp, vol. IIIB, 100). Fairness perceptions may be powerful. They tend to influence the public, lawmakers, and courts (Kahneman et al., 1986; Klein, 1980; Macaulay, 1963). Fairness perceptions, however, are not equivalent to anticompetitive effects that are the focus of antitrust law. How does antitrust law treat squeezes and squeeze claims that interact with fairness perceptions? This chapter examines this question, focusing on three traditional squeeze claims—the refusal to deal, essential facility, and price squeeze. Several old antitrust doctrines, not all of which are valid today or consistent with modern antitrust, supposedly offer remedies to these claims. By contrast, the US Supreme Court is largely

1 

U.S. v. Trans-Missouri Freight Ass’n, 166 U.S. 290, 323 (1897). Brown Shoe Co. v. United States, 370 U.S. 294, 344 (1962). 3  United States v. Aluminum Co. of America, 148 F.2d 416, 428 (2d Cir. 1945) (“Alcoa”). 2 

Squeeze Claims   121

reluctant to recognize squeezing as an antitrust violation. The chapter intends to clarify the law and economics of antitrust squeeze claims. A “squeeze” is an increase in costs that cannot be passed on to consumers. A firm or a group of firms that dominates and that supplies intermediate inputs—credit, goods, or services used for the production of other goods and services—has the power to impose restrictions on their supply, which in turn could squeeze other firms. Squeeze claims take the position that, by imposing restrictions on the supply of intermediate inputs, the defendant violated antitrust law. The three traditional squeeze claims embrace this position. A refusal to deal is a unilateral or collective unwillingness to trade with a market participant. By eliminating access to needed intermediate inputs, a refusal to deal may raise a firm’s costs and adversely affects its competitiveness. A refusal to deal may be explicit or implicit, in the form of unfavorable contractual terms that a profitable firm cannot accept. An essential facility is an asset that has no practical alternatives and without access to which a firm is blocked from competing effectively in the market. An essential facility claim states that a firm is precluded from an essential facility or that its access to such facility is crippled in another way. A price squeeze occurs when a vertically integrated firm with market power in one market gains an advantage over its rivals in an adjacent market through a pricing scheme that erodes their profit margin. The legal distinction among the traditional squeeze claims is historical and not particularly helpful. It is highly artificial as it identifies two simple forms of squeezing—through refusals to deal and price squeezes. In practice, squeezing may be more nuanced and conducted also through reduced output, degraded quality of service, and other ways. The distinction follows the classification of old antitrust cases but is inconsistent with the legal understanding that “for antitrust purposes, there is no reason to distinguish between price and nonprice components of a transaction.”4 Indeed, courts often analyze one type of claim by using cases that refer to other claims. Nonetheless, because antitrust law builds on precedents, the distinction is still alive. The practical meaning of determining that squeezing is illegal is that a firm may be subject to an antitrust duty to deal, since it is not free to squeeze. Thus, it should be stated clearly that the American legal system is hostile toward duties to act and, specifically, duty to deal.5 In antitrust, the general principle is that a business has the right to refuse to deal with other businesses. Articulated a century ago, in Colgate,6 in Linkline

4 

Pacific Bell Telephone Co. v. Linkline Communications, Inc., 555 U.S. 438, 450 (2009). See, e.g., National Federation of Independent Business v. Sebelius, 132 S.Ct. 2566 (2012). 6  United States v. Colgate & Co., 250 U.S. 300, 307 (1919): 5 

In the absence of any purpose to create or maintain a monopoly, the [Sherman Act] does not restrict the long recognized right of [a]‌trader or manufacturer engaged . . . freely to exercise his own independent discretion as to parties with whom he will deal; and, of course, he may announce in advance the circumstances under which he will refuse to sell.

122   Barak Orbach and Raphael Avraham the Supreme Court restated the principle:  “As a general rule, businesses are free to choose the parties with whom they will deal, as well as the prices, terms, and conditions of that dealing.”7 In its present form, this principle is misleading. Antitrust law and many other laws impose many restrictions on such freedom. The general rule is (or used to be) that the freedom to harm is restricted by common-law maxims and modern statutes (Orbach, 2012a; Orbach, 2012b).

6.2.  Unilateral versus Concerted Squeezing Antitrust law draws a sharp distinction between unilateral conduct and concerted action. This distinction is fundamental. Certain actions, such as price setting and marketing decisions, are perfectly legal when they are made by a single firm but could be illegal when made collectively by competitors. Correspondingly the analysis of unilateral squeezing is very different from that of concerted squeezing. First, under the present jurisprudence of the US Supreme Court, “a dominant firm may incur antitrust liability for purely unilateral conduct” only in “rare instances,”8 while collusion is “the supreme evil of antitrust.”9 The analysis of unilateral and concerted squeezing loosely follows this philosophy. Second, the US Supreme Court is generally “very cautions in recognizing” antitrust duties to deal imposed on single firms.10 In Trinko, the Court declared that it had recognized only one setting in which such duty might apply: where a single firm with market power unilaterally decides “to cease participation in a cooperative venture . . . [presumably exhibits] a willingness to forsake short-term profits to achieve an anticompetitive end.”11 This setting is the Court’s interpretation of the so-called Aspen exception to the general rule that a firm is free to refuse to deal with others.12 By contrast, as discussed below, concerted squeezing is evaluated under standards that are much less favorable to defendants. While the distinction between unilateral conduct and concerted action is very basic, in the context of antitrust squeeze claims it is sometimes neglected or confused. To illustrate consider Klor’s.13 In Klor’s, the Supreme Court declared that concerted refusals to deal are illegal per se. The known facts of Klor’s, however, describe a set of 7 

Linkline, 555 U.S. at 448. Linkline, 555 U.S. at 448. 9  Trinko, 540 U.S. at 411. 10  Id. 11  Id. at 409. 12  See Aspen Skiing Co. v. Aspen Highlands Skiing Corp., 472 U.S. 585 (1985). 13  Klor’s, Inc. v. Broadway-Hale Stores, Inc., 359 U.S. 207 (1959). 8 

Squeeze Claims   123

unilateral actions without collusion. Broadway-Hale (“BH”), an appliances retailer, operated one of its stores on Mission Street in San Francisco. Many other appliances stores operated in the area, including Klor’s (“K”) that operated next door to BH’s store (see fi ­ gure 6.1). BH was an established retailer in California. It informed eighteen leading manufacturers and distributors that it would continue dealing with them only if they stopped dealing with its next-door rival, K. They complied with its demands. Losing a large number of leading brands, K claimed that BH used its “monopolistic buying power” to advance a “conspiracy” with the manufacturers and the distributors not to sell to K or to sell only “at discriminatory prices.” It further argued that this “concerted refusal to deal . . . [had] seriously handicapped its ability to compete and . . . caused it a great loss of profits, goodwill, reputation and prestige.”14 The district court concluded that the controversy was a “purely private quarrel.” The Ninth Circuit affirmed. The Supreme Court reversed, concluding that BH’s vertical agreements established a “group boycott.” It explained that the “combination” took from K “its freedom to buy appliances in an open competitive market and drives it out of business as a dealer in the defendants’ products.”15 Further, the Court stressed that “[i]‌t deprives the manufacturers and distributors of their freedom to sell to [K] at the same prices and conditions made available to [BH] and in some instances forbids them from selling to it on any terms whatsoever.”16 Put simply, the Klor’s Court found a conspiracy in a set of vertical agreements. Consider now the Adderall XR Antitrust Litigation.17 The plaintiffs, wholesale dealers in pharmaceutical products, argued that reversed payment settlement agreements may give rise to an antitrust duty to deal. Namely, a horizontal agreement could establish a liability toward downstream firms. The defendant, Shire, held a patent for a widely prescribed drug used in the treatment of attention deficit hyperactivity disorder. It entered into reversed payment settlement agreements with two manufacturers of generic equivalents. Under the agreements, the generic manufacturers would stay out of the market for three years, in exchange for licenses to make and sell a generic version of the drug after those three years. Since the generic manufacturers failed to obtain the required regulatory approvals for their generic equivalents, after three years they started selling an unbranded version of Shire’s drug. Shire, however, breached the settlement agreements and only partially filled the orders of the generic manufacturers. The plaintiffs claimed that by supplying to the generic manufacturers less than their requirements of the unbranded drug, Shire artificially maintained and raised the price of the branded drug. The antitrust duty to deal existed, the plaintiffs maintained, because Shire entered into “a cooperative venture” and its breach was a unilateral decision intending to achieve an anticompetitive end. The district court rejected the argument, stressing that “not every sharp-elbowed business practice—though potentially wrongful as a breach of 14 

Id. at 209. Id. at 213. 16  Id. 17  In re Adderall XR Antitrust Litig., 13-1232, 2014 WL 2565832 (2d Cir. June 9, 2014) (hereinafter: “Adderall XR”). 15 

FIGURE  6.1  Retailers

in the vicinity of Klor’s that sold the branded appliances carried by Broadway-Hale.

Squeeze Claims   125

contract or even fraud—necessarily amounts to an antitrust violation.”18 The Second Circuit affirmed, holding that the plaintiffs failed to show that the facts resembled those of the Aspen exception; first and foremost, Shire did not cease dealing with the generic drug manufacturers.19 In other words, the plaintiffs in the Adderall XR Antitrust Litigation failed to convince the courts to consider the economic meaning of squeezing.

6.3.  Unilateral Squeezing Unilateral squeeze claims may arise when a firm dominating the supply of intermediate inputs imposes restrictions on their supply, supposedly to gain economic benefits. To prove antitrust violation, however, a plaintiff must show harm to competition that antitrust law intends to prevent. We stress two biases that influence the legal analysis of unilateral squeezing claims. First, the unfairness appearance of squeezing prompts redundant litigation. Second, the Supreme Court is largely reluctant to recognize that squeezing could violate antitrust law.

6.3.1.  The Vertically Integrated Firm The traditional squeeze claims—unilateral refusal to deal, the essential facility doctrine, and price squeeze claims—refer to squeezing by a vertically integrated firm. The literature, therefore, has largely focused on situations where the defendant is a vertically integrated firm using restrictions on trade in intermediate inputs in the upstream market it dominates to establish and extend its power in the downstream market. Studies of foreclosure (Rey and Tirole, 2007) and price squeezes (Hovenkamp and Hovenkamp, 2009)  epitomize this framework. When a firm that dominates the supply of certain intermediate inputs also operates in an adjacent market in which those intermediate inputs are essential, the firm could gain certain economic advantages in the adjacent market through squeezing (Rey and Tirole, 2007). Undoubtedly, under certain circumstances, when a vertically integrated firm holds market power in one market, it may have the power to squeeze its rivals in the adjacent markets. For example, when proprietary rights give a firm market power in the market for replacement parts for certain machines and that firm also operates in the market for maintenance services for those machines, it could squeeze independent service providers.20 This may also be the outcome when a firm operates with no or with very limited competition a facility, such as a hospital, and also operates in an adjacent market in which 18 

Louisiana Wholesale Drug Co., Inc. v. Shire LLC, 929 F. Supp. 2d 256, 262 (S.D.N.Y. 2013). Adderall XR, at *6. 20  See, e.g., Eastman Kodak Co. v. Image Technical Services, Inc., 504 U.S. 451 (1992). 19 

126   Barak Orbach and Raphael Avraham access to that facility is critical.21 In the same spirit, when a firm dominates the supply of certain products and decides to enter into retail sales, it may have a negative impact on retailers that compete against it.22 Likewise, when Microsoft decided to enter segments related to its operating system, it effectively squeezed its rivals.23 The squeezing in such circumstances and in many others is straightforward: The control of the vertically integrated firm over the supply of intermediate inputs gives it power its rivals do not have. While perhaps unfair, in itself such squeezing is not illegal. A plaintiff must show that the vertically integrated firm excluded competition, as opposed to promoted efficiencies through vertical integration. In this spirit, Philip Areeda famously criticized the so-called essential facility doctrine, arguing that it undermines efficiencies (Areeda, 1989). In Trinko, the Supreme Court declared that it had never recognized such the doctrine.24 In Alcoa, Judge Learned Hand analyzed price squeezes. He held that it was unlawful for a firm with monopoly power “to set the price of [the finished product] so low and hold the price of the [intermediate input] so high . . . [that it] must be regarded as higher than a ‘fair price.’ ”25 This is a very rudimentary test for squeezing financed by exclusion of competition. The test supposedly provides that price squeezing is illegal when a vertically integrated firm sets the retail price and the wholesale price at unsustainable levels to exclude competition. In Linkline, the US Supreme Court rejected the Alcoa test ruling that squeezing by a vertically integrated firm does not violate antitrust laws so long as (1) the Aspen exception does not apply and (1) the Brooke Group test for predatory pricing is not met.26 As discussed above, the Aspen exception refers to circumstances where a single firm with market power unilaterally decides “to cease participation in a cooperative venture . . . [presumably exhibiting] a willingness to forsake short-term profits to achieve an anticompetitive end.”27 The Brooke Group for predatory pricing requires proving below-cost retail pricing and a “dangerous probability” that the defendant will recoup its investment in below-cost pricing.28 Linkline’s significance deserves emphasis on three points. First, the Linkline Court analyzed price squeeze claims but declared that “for antitrust purposes, there is no reason to distinguish between price and nonprice components of a transaction.”29 Specifically, the Court rejected the distinction between “insufficient assistance claims” and price-squeeze claims.”30 Linkline, therefore, applies to squeeze claims in general. Second, the Linkline Court did not bother to consider possible anticompetitive effects

21  See, e.g., McKenzie v. Mercy Hosp. of Independence, Kansas, 854 F.2d 365 (10th Cir. 1988) (hospital as a facility needed by physicians). 22  See, e.g., Eastman Kodak Co. of New York v. Southern Photo Materials Co., 273 U.S. 359 (1927). 23  United States v. Microsoft Corp., 253 F.3d 34 (D.C. Cir. 2001). 24  Verizon Communications Inc. v. Law Offices of Curtis V. Trinko, LLP, 540 U.S. 398, 411 (2004). 25  Alcoa, 148 F.2d at 437. 26  Linkline, 555 U.S. 438. 27  Trinko, 540 U.S. at 409. 28  Brooke Group Ltd. v. Brown & Williamson Tobacco Corp., 509 U.S. 209 (1993). 29  Id. at 450. 30  Id.

Squeeze Claims   127

of squeezing. Rather, the Court focused on the question whether the defendant had an antitrust duty to deal under Aspen, or engaged in predatory pricing in the meaning of Brooke Group. Finally, Linkline was the only antitrust decision that Chief Justice John Roberts wrote during his first decade at the Court. Other than Leegin,31 it was the only five-to-four antitrust decision the Supreme Court handed down during that decade. In sum, a vertically integrated firm with market power in one market may profitably engage in squeezing in adjacent markets, including for the purpose of monopolizing those markets. Under present law, with very narrow exceptions, squeezing by a vertically integrated firm is generally legal.

6.3.2.  The Nonintegrated Firm We now turn to unilateral squeezing by nonintegrated firms. Nonintegrated firms may and often do creatively devise unilateral squeezing strategies. For example, a powerful firm can set a network of exclusive contracts with upstream or downstream firms to exclude its rivals from the market by leaving them with insufficient access to intermediate inputs. Krattenmaker and Salop (1986) called this squeezing strategy “exclusionary rights contracts.” To illustrate the use of this strategy, consider Lorain Journal. A monopoly newspaper refused to sell newspaper advertising to advertisers that also used a radio station that attempted to enter the local market.32 The United States Tobacco Company (“USTC”), which dominated the market for moist snuff throughout the 20th century, deployed a similar squeezing strategy.33 The company used slotting allowances to secure its space on racks and points of sales and exclude competitors.34 These vertical arrangements, which are common and offer efficiencies, can also be used for exclusion (FTC, 2003). The facts of Klor’s, discussed in section 6.2, as described by the plaintiff, supposedly depict a similar squeezing strategy. Leveraging “a monopolistic buying power,” a powerful firm allegedly secured a network of agreements with appliances manufacturers and suppliers to exclude competition. However, the agreements Broadway-Hale secured were limited in their scope. Klor’s approach to vertical agreements belongs to the antitrust tradition of inhospitality (Turner, 1966). We compare Klor’s to Lorain Journal and USTC to emphasize that the analysis of vertical arrangements is not and should not be beyond the scope of antitrust. As contractual instruments, vertical arrangements may be used also for anticompetitive purposes. In reality, powerful nonintegrated firms devise ways limit the access of their competitors to the market and inputs in many other ways (Baker, 2013). The view that “a 31 

Leegin Creative Leather Products, Inc. v. PSKS, Inc., 551 U.S. 877 (2007). Lorain Journal Co. v. United States, 342 U.S. 143 (1951). 33  Conwood Co., L.P. v. U.S. Tobacco Co., 290 F.3d 768 (6th Cir. 2002). 34  “Slotting allowances” are payments a supplier makes to a retailer as a condition for the initial placement of the supplier’s products on the retailer’s store shelves. (FTC, 2003). 32 

128   Barak Orbach and Raphael Avraham dominant firm may incur antitrust liability for purely unilateral conduct” only in “rare circumstances”35 governs US antitrust law but does not rest on economic analysis.

6.4.  Concerted Squeezing Concerted squeezing is collusion with a feature that enables a squeeze of some sort. The ordinary collusion intends to squeeze surplus from consumers but does not exclude competition through restrictions on the supply of intermediate inputs. Concerted squeezing schemes include an enforcement mechanism that facilitates the squeeze. The enforcement mechanism may include vertical relationships with upstream or downstream firms,36 a control over trade association or industry standards,37 ownership of technological standards,38 or something else. To illustrate concerted squeezing, consider the seminal Terminal Railroad case that established the essential facility doctrine.39 In 1889, railroad tycoon Jay Gould organized the Terminal Railroad Association of St. Louis (“TRRA”) to consolidate control over railway transportation entering and leaving St. Louis and, specifically, transportation across the Mississippi River between St. Louis and Illinois. At the time, St. Louis was the fourth largest city in the United States, and in it was concentrated a significant volume of trade. TRRA was a consortium consisting of fourteen of the twenty-four railroad companies that converged at St. Louis. Its members controlled “about one third of the railroad mileage of the United States.”40 TRRA established a concerted squeezing scheme facilitated through vertical integration. It was able to squeeze other railroads through control over the transportation means across the Mississippi River: the Eads Bridge, the Merchants’ Bridge, and the Wiggins Ferry Company. Another example of an effective enforcement mechanism is the ringmaster of hub-and-spoke cartels. The ringmaster typically has vertical relationships with the cartel members, setting and enforcing a cartel for them, while squeezing its competitors and possibly also competitors of other cartel members.41 In such hub-and-spoke cartels, the ringmaster organizes a cartel among inputs sellers and, in exchange for its 35 

Linkline, 555 U.S. at 448. See, e.g., JTC Petroleum Co. v. Piasa Motor Fuels, Inc., 190 F.3d 775 (7th Cir. 1999). 37  See, e.g., Associated Press v. United States, 326 U.S. 1 (1945); American Society of Mechanical Engineers, Inc. v. Hydrolevel Corp., 456 U.S. 556 (1982); Abraham & Veneklasen Joint Venture v. American Quarter Horse Association, 2013 WL 2297104 (N.D. Tex. May 24, 2013); Areeda, 1989 (referring to “industry consortia”). 38  See, e.g., Clamp-All Corp. v. Cast Iron Soil Pipe Institute, 851 F.2d 478 (1st Cir. 1988); Fashion Originators’ Guild of America v. FTC, 312 U.S. 457 (1941). 39  See, e.g., United States v. Terminal Railroad Association of St. Louis, 224 U.S. 383 (1912). 40  Id. at 400. 41  Interstate Circuit v. United States, 306 U.S. 208 (1939); Toys “R” Us, Inc. v. FTC, 221 F.3d 928 (7th Cir. 2000); United States v. Apple Inc., 952 F. Supp. 2d 638 (S.D.N.Y. 2013); Klein, 2012. 36 

Squeeze Claims   129

organizational services (or for its ability to enforce discipline), the sellers do not deal with its competitors or deal with them in a squeezing manner. For example, in Interstate Circuit, a first-run theater chain established an agreement among the major distributors regarding the organization of movie exhibition in the region and the squeezing of its rivals.42 Similarly, Toys “R” Us, a powerful toy retailer, formed a network of vertical agreements with individual toy manufacturers, in which each manufacturer committed to restrict distribution of its products to low-priced club stores on the condition that other manufacturers do the same.43 The analysis of enforcement mechanisms that facilitate concerted squeezing may be rather complex. For example, when the mechanism is a vertical arrangement, it may be presented as a contractual instrument adopted to enhance efficiency. This was Apple’s argument in defense of the e-book hub-and-spoke cartel.44 Similarly, when the mechanism is control over industry membership, the mechanism is typically presented as a means necessary to protect quality.45 The legal rule for concerted squeezing, however, is very crude. Under Klor’s, group boycott is illegal per se. As a result, courts have been rationalizing deviations from the per se rule to the rule of reason.46 The complexity of the enforcement mechanisms suggests that per se rules may be inadequate for concerted squeezing. This point is consistent with the broad criticism of the law of group boycott.

6.5. Conclusion Squeezing is restrictions on the supply of intermediate inputs that a firm with market power or a cartel can impose on other firms, thereby raising their costs. It is a rather common strategy. Antitrust squeeze claims are as old as antitrust. Squeezed firms sometimes try to find remedy in antitrust law. The traditional antitrust analysis of squeeze claims is quite dismissive, among other reasons, because they try to revive noneconomic doctrines. In this chapter, we argue that the analysis of both unilateral and concerted squeezing is too crude and could be improved. Specifically, we argue that the traditional squeeze claims—refusal to deal, essential facility, and price squeeze—correspond to an outdated case law. They describe a limited set of restrictions on the supply of intermediate inputs and do not identify 42 

Interstate Circuit, 306 U.S. 208. Toys “R” Us, 221 F.3d 928. 44  Apple, 952 F. Supp. 2d 638. 45  See, e.g., FTC. v. Indiana Federation of Dentists, 476 U.S. 447 (1986); FTC v. Superior Court Trial Lawyers Association, 493 U.S. 411 (1990) (holding that a strike illegal per se); see, e.g., Northwest Wholesale Stationers, Inc. v. Pacific Stationery & Printing Co., 472 U.S. 284 (1985); Gregory v. Fort Bridger Rendezvous Ass'n, 448 F.3d 1195 (10th Cir. 2006); Abraham & Veneklasen, 2013 WL 2297104. 46  See, e.g., FTC. v. Indiana Federation of Dentists, 476 U.S. 447 (1986). 43 

130   Barak Orbach and Raphael Avraham anticompetitive effects or practical remedies. It may well be that antitrust should not address squeeze claims, but probably not for the reasons presently used.

References Areeda, Phillip. 1989. Essential Facilities: An Epithet in Need of Limiting Principles. Antitrust Law Journal 58: 841–52. Areeda, Phillip E., and Herbert Hovenkamp. Antitrust Law. New York: Wolters Kluwer. Baker, Jonathan B. 2013. Exclusion as a Core Competition Concern. Antitrust Law Journal 78: 527–89. FTC Stuff Study. November 2003. The Use of Slotting Allowances in the Retail Grocery Industry: Selected Case Studies in Five Product Categories. Hovenkamp, Herbert, and Erik N. Hovenkamp. 2009. The Viability of Antitrust Price Squeeze Claims. Arizona Law Review 51: 273–303. Kahneman, Daniel, Jack L. Knetch, and Richard Thaler. 1986. Fairness as a Constraint on Profit Seeking: Entitlements in the Market. American Economic Review 76: 728–41. Klein, Benjamin. 1980. Transaction Cost Determinants of “Unfair” Contractual Arrangements. American Economic Review 70: 356–62. Klein, Benjamin. 2012. The “Hub-and-Spoke” Conspiracy That Created the Standard Oil Monopoly. Southern California Law Review 85: 459–98. Krattenmaker Thomas G., and Steven C. Salop. 1986. Anticompetitive Exclusion: Raising Rivals’ Costs to Achieve Power over Price. Yale Law Journal 96: 209–93. Macaulay, Stewart. 1963. Non-contractual Relations in Business: A Preliminary Study. American Sociological Review 28: 55–67. Orbach, Barak, 2012a. Regulation: Why and How the State Regulates. St. Paul: Foundation Press. Orbach, Barak. 2012b. What Is Regulation? Yale Journal on Regulation Online 30: 1–12. Rey, Patrick, and Jean Tirole. 2007. A Primer on Foreclosure. Handbook of Industrial Organizations 3: 2147–220. Turner, Donald F. 1966. Some Reflections on Antitrust. New York State Bar Association Law Symposium 1966: 1–9.

Cases Cited Abraham & Veneklasen Joint Venture v.  American Quarter Horse Association, 2013 WL 2297104 (N.D. Tex. May 24, 2013). America’s Best Cinema Corp. v. Fort Wayne Newspapers, Inc., 347 F. Supp. 328, 333 (N.D. Ind. 1972). American Society of Mechanical Engineers, Inc. v. Hydrolevel Corp., 456 U.S. 556. (1982). Aspen Skiing Co. v. Aspen Highlands Skiing Corp., 472 U.S. 585 (1985). Associated Press v. United States, 326 U.S. 1 (1945). Brown Shoe Co. v. United States, 370 U.S. 294, 320 (1962). Brooke Group Ltd. v. Brown & Williamson Tobacco Corp., 509 U.S. 209 (1993). Clamp-All Corp. v. Cast Iron Soil Pipe Institute, 851 F.2d 478 (1st Cir. 1988).

Squeeze Claims   131

Conwood Co., L.P. v. U.S. Tobacco Co., 290 F.3d 768 (6th Cir. 2002). Eastman Kodak Co. v. Image Technical Services, Inc., 504 U.S. 451 (1992). Eastman Kodak Co. of New York v. Southern Photo Materials Co., 273 U.S. 359 (1927). Eatoni Ergonomics, Inc. v. Research in Motion Corp., 486 F. App’x 186 (2d Cir. 2012). Fashion Originators’ Guild of America v. FTC, 312 U.S. 457 (1941). F.T.C. v. Indiana Federation of Dentists, 476 U.S. 447 (1986). F.T.C. v. Superior Court Trial Lawyers Association, 493 U.S. 411 (1990). Gregory v. Fort Bridger Rendezvous Association, 448 F.3d 1195 (10th Cir. 2006). Illinois ex rel. Hartigan v. Panhandle E. Pipe Line Co., 730 F. Supp. 826 (C.D. Ill. 1990). Image Technical Services, Inc. v. Eastman Kodak Co., 125 F.3d 1195 (9th Cir. 1997). In re Independent Service Organizations Antitrust Litigation, 203 F.3d 1322 (Fed. Cir. 2000). In re Adderall XR Antitrust Litig., 13-1232, 2014 WL 2565832 (2d Cir. June 9, 2014). Intergraph Corp. v. Intel Corp., 195 F.3d 1346 (Fed. Cir. 1999). Interstate Circuit v. United States, 306 U.S. 208 (1939). K & S Associates, Inc. v. American Association of Physicists in Medicine, 2013 WL 2177938 (M.D. Tenn. May 20, 2013). Klor’s, Inc. v. Broadway-Hale Stores, Inc., 359 U.S. 207 (1959). Laurel Sand & Gravel, Inc. v. CSX Transp., Inc., 924 F.2d 539 (4th Cir. 1991). Leegin Creative Leather Products, Inc. v. PSKS, Inc., 551 U.S. 877 (2007). Lorain Journal Co. v. United States, 342 U.S. 143 (1951). Loren Data Corp. v. GXS, Inc., 501 F. App’x 275 (4th Cir. 2012). Louisiana Wholesale Drug Co., Inc. v. Shire LLC, 929 F. Supp. 2d 256, 262 (S.D.N.Y. 2013). MCI Communications Corp. v. AT&T Co., 708 F.2d 1081 (7th Cir. 1983). N.A.A.C.P. v. Claiborne Hardware Co., 458 U.S. 886 (1982). National Federation of Independent Business v. Sebelius, 132 S.Ct. 2566 (2012). Neeld v. National Hockey League, 594 F.2d 1297 (9th Cir. 1979). Northwest Wholesale Stationers, Inc. v. Pacific Stationery & Printing Co., 472 U.S. 284 (1985). nSight, Inc. v. PeopleSoft, Inc., 296 F. App’x 555 (9th Cir. 2008). NYNEX Corp. v. Discon, Inc., 525 U.S. 128 (1998). Otter Tail Power Co. v. United States, 410 U.S. 366 (1973). Pacific Bell Telephone Co. v. Linkline Communications, Inc., 555 U.S. 438 (2009). United States v. Aluminum Co. of Am., 148 F.2d 416 (2d Cir. 1945). United States v. Apple Inc., 952 F. Supp. 2d 638 (S.D.N.Y. 2013). United States v. Colgate & Co., 250 U.S. 300, 307 (1919). United States v. Terminal Railroad Association of St. Louis, 224 U.S. 383 (1912). United States v. Microsoft Corp., 253 F.3d 34 (D.C. Cir. 2001). Verizon Communications Inc. v. Law Offices of Curtis V. Trinko, LLP, 540 U.S. 398 (2004).

CHAPTER 7

I N N OVAT IO N A N D A N T I T RU ST P OL IC Y THOMAS F. COT TER

Most economists would agree that, in the long run, the welfare gains from innovation likely dwarf the relatively short-term welfare gains—the elimination or reduction of the deadweight loss attributable to market power—that economic analysis predicts will result from more competitive market structures.1 Consistent with this premise, nations offer inventors and creators intellectual property rights (IPRs), on the theory that the possibility of obtaining a (short-run) monopoly provides a necessary inducement for individuals to invest in the creation of inventions, works of authorship, and other innovative subject matter that promise to generate greater long-run social benefits. On this reasoning, an antitrust policy that focuses exclusively, or too aggressively, on promoting static (short-run) competition would appear, at best, myopic and at worst, perverse. Perhaps, then, antitrust would better serve the public interest if it focused as much or more attention on fostering long-run, dynamic competition—in penalizing what Bohannan and Hovenkamp refer to as restraints on innovation—as it traditionally has on promoting static competition. The problem, of course, lies in the details. As Bohannan and Hovenkamp also note, the standard tools of economic analysis are much better designed for predicting and measuring the effects of, say, monopoly or cartel behavior on short-run pricing and output decisions than they are at predicting the effects of such behavior on long-run innovation.2 To illustrate, under a simple decision-theoretical perspective an optimal antitrust standard is one that minimizes the sum of error (false positive and false negative) costs and enforcement costs—that is, of p1c1 + (1 − p1c2) + c3, where p1 is the probability of 1 

For discussions, see, e.g., Bohannan and Hovenkamp (2012, 8–11, 14, 241–42); Warsh (2006). Bohannan and Hovenkamp (2012, 6–7, 11, 242–56). A word on terminology may be appropriate here. Patent laws confer protection in inventions, that is, in the intangible inventive principle the inventor claims to have discovered. That principle itself may be incorporated in a wide variety of tangible embodiments that are not, strictly speaking, identical to the intangible patented invention (inventive principle). Relatedly, many scholars also differentiate invention from innovation, where the latter term refers to the embodiment of an invention in a commercial product or service. As will be noted in the text above, many patented inventions are never deployed in commercial products and thus do not result in innovation as such. 2 

Innovation and Antitrust Policy   133

wrongly condemning conduct that is, on balance, welfare-enhancing; 1  − p1 is the probability of wrongly exonerating conduct that is, on balance, welfare-reducing; c1 is the net social welfare loss from wrongly condemning conduct that is, on balance, welfare-enhancing; c2 is the net social welfare loss from wrongly exonerating conduct that is, on balance, welfare-enhancing; and c3 is the cost of adjudication. If we ignore the potential effect of the conduct at issue on innovation, c1 would be the forgone net efficiency gains—the excess of short-term efficiency gains over deadweight loss, if any—if the conduct is wrongly condemned. c2 would be the net efficiency loss—the excess deadweight loss over short-term efficiency gains, if any—if the conduct is wrongly exonerated. Economic analysis can assist in attaining a (rough) idea of whether the sum of error and enforcement costs so defined is likely to be higher or lower if the conduct at issue is evaluated under a per se rule, the rule of reason, or some other intermediate standard. By contrast, an antitrust policy that factors into the mix a practice’s effects on innovation would define c1 as the excess of efficiency gains (from both short-term efficiency gains and long-term efficiencies resulting from innovation) over efficiency losses (from both deadweight loss and any longer-term, countervailing harms to innovation). The farther out one goes in predicting those long-term consequences resulting from innovation-related efficiency gains, the less one can envision their magnitude or direction. Estimating the value of p becomes increasingly difficult as well, since the future course of innovation is, almost by definition, unpredictable. One can attempt to model these effects of dynamic innovation by incorporating into c1 and c2 the costs and benefits of IPRs and other institutions that affect the rate of innovation, but the analysis is often too abstract to produce concrete policy recommendations—one problem being that any system designed to encourage future innovation necessarily relies upon past innovation as an input.3 Innovation’s dual status as both an output and an input of, say, a patent or copyright system, in other words, means that predictions as to the net effect of a given policy on innovation are often fraught with substantial uncertainty. Seen in this light, it is not surprising that antitrust law generally focuses on the more tractable short-term costs and benefits, notwithstanding their (relative) lack of importance to social welfare. Be that as it may, whether or not policymakers consciously consider the effects of antitrust on innovation, antitrust surely will have some impact on innovation, for good or for ill; the question then is whether antitrust should at least strive to take those effects into consideration, or whether those effects are simply too unpredictable to merit any such conscious effort. I will argue below that, at least in some discrete circumstances, policymakers should (and do) take likely impacts on innovation into account in formulating antitrust standards, and that some modest expansions of such efforts would be welcome. At the same time, I will argue, it would probably be a mistake to expect antitrust to do too much. Antitrust’s role in promoting innovation inevitably will be modest in comparison with the role of intellectual property (IP) law; and IP law itself is hardly the only 3  The observation that information is both an input and an output can be traced at least as far back as Arrow (1962, 618). For discussions of the difficulties of fine-tuning a patent system to optimize innovation over time, see Green and Scotchmer (1995); Scotchmer (1991).

134   Thomas F. Cotter game in town, but rather only one component of an innovation policy that may include public or private patronage, self-help and first-mover advantages, and possibly other tools. Antitrust therefore may be part of the solution to the problem of ensuring effective dynamic competition, but it should play a dominant role only in limited circumstances. Section 7.1 provides a brief overview of the role of IPRs in promoting innovation, according to standard economic theory. Section 7.2 discusses the relationship between antitrust and IP law in promoting social welfare. Section 7.3 then highlights three ways in which antitrust can play a useful role in innovation policy. Section 7.4 concludes.

7.1.  Of Intellectual Property Rights and Innovation IPRs are probably the most prominent tool modern governments use to foster innovation, but the debate over whether IPRs are a necessary or even desirable means to this presumed end is of long standing. With regard to patents, for example—and most of the discussion that follows will focus on patents for new inventions, with less attention being paid to trade secrets, copyright in works of authorship, and other forms of IP that generally play a smaller role in innovation as such—the conventional argument is that patents solve a free-rider problem that otherwise would threaten to undermine the incentive to invent: that is, that given a sufficiently high cost of inventing and a sufficiently low cost of copying another’s invention, the dominant strategy of rational actors often would be to copy rather than to invent. Indeed, the copier often would have a competitive advantage against the inventor, since the copier does not need to make up the sunk costs of research and development (R & D) in order to meet his long-run costs. If too many people follow this strategy of copying as opposed to inventing, however, innovation stands to suffer. Of course, the conditions that give rise to the strategy are not universally present. Sometimes the cost of inventing is low, or the cost of copying—say, by reverse engineering the formula for a popular soft drink—may be surprisingly high. Or the expected gains from inventing may exceed the expected gains from copying, notwithstanding the copier’s potential cost advantage. Some people may invent or engage in other creative work for the sheer joy of it—the psychic gain, if you will—though psychic pleasures alone are unlikely to compensate for the sunk costs of, say, developing a new antiretroviral drug. In other instances, merely being the first on the market with a new product may provide the inventor with a sufficient opportunity to recoup its R & D costs, before copiers can enter and compete.4 Alternatively, various self-help measures, including contractual limitations on use or resale, or any of a variety of measures to keep the invention secret, may suffice to preserve the incentive to invent even though third 4  For arguments that first-mover advantages alone often would suffice, see Boldrin and Levine (2008, 136–40).

Innovation and Antitrust Policy   135

parties who are not bound by any duty of confidentiality might succeed in copying the invention. Many analysts nevertheless conclude that these alternatives would be insufficient to stimulate the optimal amount of invention—and that some of these measures, in particular secrecy, suffer from their own defects as tools of social policy. An invention that remains secret can be difficult to license efficiently, as Arrow’s Information Paradox (Arrow 1962; 614–16) teaches; and in any event secrecy interferes with the public’s ability to learn from, and improve upon, the technology at issue. Conventional analysis therefore assumes that some form of government intervention may be desirable in order to solve the aforementioned free-rider problem. One form that such intervention may take is the granting of exclusive rights in the form of a patent. From one perspective, this form of intervention may seem odd because the subject matter of a patent is intangible—an inventive principle (which, however, may be embodied in a variety of tangible things). Inventions therefore reflect two characteristics often associated with public goods, namely nonrivalrousness (many people can simultaneously consume the same inventive principle without depleting it) and nonexcludability (preventing someone from using an inventive principle once it has been disclosed to him can be extremely difficult). Public goods generally can give rise to free-rider problems, insofar as the nonrivalrous and nonexcludable nature of such goods make it difficult to prevent anyone—including those who do not contribute to the cost of creating or maintaining them—from enjoying their benefits. Though markets or other informal institutions sometimes may succeed in overcoming free-rider problems that threaten to undermine the provision of public goods, quite often governments step in to provide such goods themselves and finance their provision from taxes assessed from the general public; in theory, at least, such public financing overcomes the free-rider problem. To some extent, the provision of the public good of innovation is financed in much the same way, through the use or private or public institutions that fund research, either directly (for example, through grants) or indirectly (for example, through the provision of tax deductions for R & D); or that offer ex ante incentives (prizes) or ex post rewards for inventive success. One advantage of a system of IPRs, however, is that unlike a system of grants, prizes, and rewards, the system is decentralized. Rather than having government or private foundations decide what needs to be invented and how much value (either ex post or ex ante) to place upon such inventions, the government instead offers inventors exclusive rights for any and all inventions meeting the statutory requirements (novelty, nonobviousness, and so on). In theory, this decentralized approach capitalizes upon the informational advantage collectively possessed by market actors in deciding what to invent and, if they are successful in obtaining a patent, how much to ask for the use of it. The market, in other words, ultimately decides which inventions are valuable and which are not.5 5  Such an approach might not be very helpful, however, when it comes to creating incentives to develop inventions that may confer great social utility but that possess relatively little commercial value, such as essential medicines for the treatment of diseases endemic to the developing world. To fund this type of research, alternatives such as prizes may be optimal. For an overview and critical discussion of the relevant literature, see Fisher and Syed (forthcoming, ch.7).

136   Thomas F. Cotter According inventors exclusive rights in their inventions therefore may be desirable if it generates social benefits that would be more costly to generate, or that might not be generated at all, in the absence of such a system. Whether any given patent system actually is desirable, however, depends as well on its corresponding social costs. Aside from administrative costs, the most prominent costs attributable to patents are potential monopoly and (what I will refer to as) “access” costs. As for the first of these, in colloquial usage patents are often referred to as “monopolies,” but the term is a misnomer. Empirical studies generally show that less than half of all patents are ever commercialized at all, and of these only a subset enable the owner to derive substantial economic rents.6 That subset, however, can be an important one, including within it (for example) patents on new drugs for which there are no good market substitutes.7 Of course, consumers may be better off paying a monopoly price for an invention that greatly improves their lives but that would not have been invented so soon (or at all) absent the patent system, than they would be paying a competitive price for a poor substitute (Dam 1994, 251). Nevertheless, an ideal patent system would work to reduce the social costs of monopoly, for example by limiting the scope and duration of patent rights to the minimum necessary to enable the owner to recoup its R & D costs. As for access costs, the principal problem is that patents can raise the cost incurred by future inventors whose work necessarily builds upon (and perhaps improves) that which has gone before. To be sure, in a world of zero information and transaction costs, patent owners would never demand a royalty so high as to derail a potentially valuable new application of their technology. In the real world, however, information and transaction costs can sometimes be quite high. The payoff to be earned from developing an improvement is itself unpredictable; it is often difficult to determine whether a given patent actually would read on (cover) a potential improvement; and sometimes just the sheer number of patents potentially reading on a given technology can be enormous, thus giving rise to concerns that a “patent thicket” or “anticommons” of potentially blocking patent rights would be so burdensome to negotiate that some improvements will simply be abandoned instead.8 Seen in this light, the ideal patent system would be one that maximized the surplus of social benefits derived from innovation over social costs—the administrative, access, and occasional monopoly costs that are (to some degree) an inevitable byproduct of any patent system. The problem is that no one knows where this “sweet spot” lies; resolving 6 

For citations to the relevant literature, see Sichelman (2010, 362–63 and n. 121). One might imagine that only those patents that do enable the owner to accrue economic rents would be the subject of litigation. Because a patented invention for which there are adequate market substitutes can be easily designed around, an injunction would provide little leverage for extracting a lucrative licensing agreement, and damages should be minimal. Due to imperfections in the way in which patents provide public notice of their covered subject matter, however, the cost of designing around ex post may be much higher than the cost of designing around ex ante, in which case the owner of a patent that contributes little to the state of the art nevertheless sometimes may be able to negotiate a favorable settlement that reflects some portion of the infringer’s sunk costs. See Cotter (2009, 1181–82). 8  See Heller and Eisenberg (1998). Cf. Walsh et al. (2005, 2003) (reporting “little empirical basis for claims that restricted access to IP is currently impeding biomedical research,” though noting that “there is evidence that access to material research inputs is restricted more often”). 7 

Innovation and Antitrust Policy   137

the problem appears thus far to elude empirical analysis. Nevertheless, many standard patent law doctrines—including the limited term of patent rights, the exclusion from patentability of abstract ideas and laws of nature, and limiting patent protection only to novel, useful, and nonobvious subject matter that is adequately disclosed by the inventor—can be viewed as efforts, however imperfect, to attain this optimal point. In principle, moreover, policymakers could look to empirical evidence, where available, for guidance in refining patent doctrine so as to bring it more closely in alignment with optimum. As discussed below, for example, empirical evidence tends to show that patent protection is much more important in certain industries (e.g., pharmaceuticals) than in others (e.g., software). In theory, then, policymakers could adopt a system of industry-specific patent rights; and there is reason to believe that, in practice, courts do engage in a bit of “tailoring,” whether consciously or not, in determining, for example, the amount of disclosure needed to enable a person of ordinary skill in the art to practice the claimed invention.9 Explicit tailoring, however, in the sense of (say) affording pharmaceutical patents a longer term than other patents, appears to be ruled out of bounds under the TRIPs Agreement, an international treaty that forbids WTO member states from discriminating with respect to patent rights on the basis of field of technology.10 Such explicit tailoring also likely would induce rent-seeking and regulatory capture. As a result, however, in the absence of such tailoring patent law may confer stronger rights (e.g., a longer term) than is optimal for some fields of technology and weaker-than-optimal rights for others. The uncertainty that surrounds the optimal design of patent rights is further complicated by a number of less obvious effects that patents may have on innovation. Edmund Kitch’s famous “prospect” theory, for example, argues that patent rights enable inventors to coordinate follow-up investments in improvements upon an early technology. This coordination can be socially efficient to the extent it reduces the risk of duplicative efforts on the part of multiple would-be improvers (Kitch 1977, 267–71).11 Kitch’s theory, if correct, suggests that the patent system should accord “pioneering” patents broad scope, and is therefore at least broadly consistent with Joseph Schumpeter’s thesis that monopoly acts as a spur to innovation. On this view, an aggressive antimonopoly 9 

See, e.g., Burk and Lemley (2009, 59–65, 115–16). Burk and Lemley argue, however, that courts have sometimes gotten the tailoring wrong by, for example, applying doctrine so as to encourage large numbers of very narrow patents on DNA sequences. Some biotech patent cases decided since 2009 may ameliorate these consequences to some extent. 10  TRIPs art. 27(1). Moreover, TRIPs requires a minimum patent term expiring no earlier than 20 years from the date of filing, see id. art. 33, so even if discrimination as to field of technology were permitted, patents could not be granted for a shorter term. Nations may, however, extend the patent term to compensate for regulatory delay in approving a product for marketing or for delays in patent prosecution. The United States does so, and to some extent this often does result in pharmaceutical patents enjoying a longer term (though not necessarily a longer effective term). Patents also may expire earlier than 20 years for failure to pay national patent maintenance fees. 11  Some scholars have attempted to modify or extend the theory—for example, to explain why laws of nature are not patentable, see Grady and Alexander (1992), or to show how patents may assist in coordinating follow-up investments in innovation, meaning here not invention as such but rather the commercialization of invention, see Kieff (2001).

138   Thomas F. Cotter policy is misguided, because the gains from innovation, being permanent, outweigh the relatively transitory losses from monopoly; moreover, no monopolist can maintain its position forever, but rather will eventually be overtaken by the purveyor of a new generation of technology in a process of creative destruction (Schumpeter [1942] 2003, 84–85, 100–103). The prospect theory remains controversial, however, with some analysts finding empirical support for the theory lacking (Beck 1983; but see Duffy 2004, 488 n. 141), and others arguing that competition in the market for follow-up improvements is more likely to improve social welfare than is according broad control to the owner of the pioneering patent (Merges and Nelson 1990, 871–78). Kitch skeptics find support in the theoretical work of Kenneth Arrow, who (contra Schumpeter) contended that competition, not monopoly, is the greater spur to innovation (Arrow 1962, 620–21). Other theorists have focused on yet other ways in which patents can impact innovation. Clarisa Long (2002), for example, argues that patents provide a signal of a start-up company’s potential value, and thus assist venture capitalists in deciding which start-ups to fund. Some empirical evidence supports her thesis that venture capitalists place substantial weight on start-ups’ patent portfolios,12 though questions remain whether or to what extent the signal provided by a large portfolio is an accurate one. Other theoretical literature has highlighted the possible impact of patents and other IPRs on the size and structure of firms. For example, the type and scope of IPRs covering a given technology can impact firms’ and employees’ willingness to share their ideas with one another and with authorized third parties. IPRs thus can at least indirectly affect the ability of firms to organize at efficient scale (Burk and McDonnell 2007; Barnett 2011). Empirical studies on the effects of patents on invention and innovation have tended to show at most relatively modestly positive incentive effects in most industries other than pharmaceuticals and chemicals.13 Relatedly, Bessen and Meurer (2008, 141) argue that, since the late 1990s, the private costs of patenting have outweighed the private benefits for US public firms other than those in the chemical and pharmaceutical industries (see Bessen and Meurer 2008). In addition, a study of US patent litigation in the IT sector by Lemley and Cotropia (2008) shows that only a small percentage of the cases involve deliberate copying. In most, the alleged infringer independently invented (but did not itself attempt to patent) the subject matter at issue, suggesting perhaps that a good deal of innovation in this field occurs independently of the patent incentive. On the other hand, empirical work thus far has failed to show systematic 12  See Graham et al. (2009, 1270 n. 44, 1280) (“Firms that seek venture-funding appear to be patenting more actively prior to the funding event (and for the purpose of securing funding), and venture-capital investors appear much less willing to fund companies that hold no patents”). 13  See Cohen, Nelson, and Walsh (2000) (concluding, based on a “Carnegie Mellon Survey (CMS) on Industrial R&D in the U.S. manufacturing sector administered in 1994,” that patents “are still not one of the major mechanisms in most industries when the views of all firms are considered”). The authors caution, however, that “[w]‌hile patents are still not the dominant mechanism in most industries for protecting product innovations, it now appears that they can be counted among the major mechanisms of appropriation in a more sizeable minority of industries.” See also Boldrin and Levine (2008, 192–98) (discussing other empirical literature on the incentive effects of patents).

Innovation and Antitrust Policy   139

negative effects resulting from patent thickets or anticommonses. In fields such as semiconductors and biotechnology, firms and researchers have tended to reach accommodations, either of a formal nature (patent pools or other cross-licensing agreements) or of an informal one (norms counseling against filing suit for experimental purposes).14 Such studies do not prove, however, that thickets and anticommonses can never give rise to problems in the real world. Formal legal recognition of an experimental use defense, available in many countries other than the United States, might be helpful in fending off such problems in the future. Antitrust as well can (and generally does) play a positive role by recognizing that the procompetitive benefits of agreements to cross-license patents often, though not always, dominate their potential anticompetitive consequences.

7.2.  Antitrust and IP: A Complex Relationship A view commonly expressed by outsiders to the fields is that antitrust and IP are antagonistic bodies of law: antitrust, after all, condemns monopolies, while IP law offers inventors and authors the possibility of obtaining monopoly power as a goad to creating and disclosing their creations. If this view were correct, courts would have to decide whether in cases of apparent conflict antitrust or IP law would take precedence; and early in the history of US antitrust law, courts often did defer to IP owners’ right to engage in conduct that might otherwise appear contrary to antitrust norms.15 Before too long, however, the tide turned and courts began applying increasingly aggressive antitrust scrutiny to transactions involving IPRs. The high-water mark of this approach was probably in the early 1970s, when a Department of Justice (DOJ) official in a public

14 

See, e.g., Hall and Ziedonis (2001, 109) (noting cross-licensing in the semiconductor industry). For example, in Bement v. National Harrow Co., 186 U.S. 70, 92–94 (1902), the US Supreme Court held that the Patent Act permitted—and the Sherman Act did not forbid—the patentee to license the right to make and sell a patented invention, subject to an obligation on the part of the licensee/ manufacturer to resell the product at a specified price. The Court reaffirmed this rule in its 1926 decision, United States v. General Electric Co., 272 U.S. 476, 488 (1926), although it has since hinted the rule may no longer be valid. See Simpson v. Union Oil Co. of Cal., 377 U.S. 13, 24 (1964) (noting that, although “price fixing in the marketing of patent articles had been condoned” previously, resale price fixing through a “coercive type of ‘consignment’ agreement is illegal under the antitrust laws”). Similarly, in Henry v. A.B. Dick Co., 224 U.S. 1, 48–49 (1912), the Court held that a person who bought a patented machine subject to the condition that the purchaser would use only (nonpatented) ink made by the patentee engaged in an infringing use when she used another company’s ink; and the supplier of the ink, who was aware of the license restriction, was liable for contributory patent infringement. Five years later, however, the Court held that § 3 of the newly enacted Clayton Act overruled A.B. Dick. See Motion Picture Patents Co. v. Universal Film Mfg. Co., 243 U.S. 502, 517–18 (1917). 15 

140   Thomas F. Cotter speech identified the “Nine No-Nos,” nine practices involving IPRs that the DOJ viewed as being of dubious legality.16 By the late 1970s, however, as the Chicago school began to exert increasing influence among antitrust policymakers in the United States, a more nuanced view of the relationship between antitrust and IP began to emerge, in recognition of the fact that many IPRs do not confer meaningful monopoly power after all, and that many IP-related transactions (such as patent pools) frequently offer substantial procompetitive benefits. Notably as well, even when IPRs do confer monopoly power, the mere possession of such power does not state a claim for violation of § 2 of the Sherman Act, which instead requires proof of the willful acquisition or maintenance of such power.17 Thus in 1995, the US antitrust enforcement authorities published guidelines relating to the licensing of IP that expressly recognized that “[i]‌ntellectual property is . . . neither particularly free from scrutiny under the antitrust laws, nor particularly suspect under them.”18 In a similar vein, in its 2006 decision in Independent Ink, Inc. v. Illinois Tool Works Inc., the US Supreme Court overruled previous case law holding that possession of a patent conferred a presumption of market power for purposes of evaluating the legality of a tying arrangement. Henceforth, antitrust plaintiffs must prove, among other things, that a defendant patent owner has market power in the market for the tying patented product.19 On one view, then, antitrust and IP are best viewed as complementary bodies of law, each seeking to improve social welfare albeit through different means (Cotter 2008, 745–49). Any conclusion that antitrust and IP can never be in tension, however, would be premature. At least under some circumstances, antitrust may have to determine how much weight to accord a patent owner’s defense that a particular restraint or conduct is reasonably necessary to preserve the value of the patent (or other IP) incentive system; or that the restraint or conduct serves the ancillary purpose of facilitating the efficient use of a technological standard, notwithstanding (some) potential for also reducing incentives or inhibiting the development of a rival standard. Sometimes the tension is starkly presented, as in the case of grantback arrangements, where the patentee licenses its invention on condition that the licensee grant back a license to use improvements developed by the licensee. Grantback provisions may be procompetitive, to the extent they induce the licensing of the earlier (“dominant”) invention, but they also could impede dynamic competition to the extent they stifle would-be improvers’ incentives

16  See Tom and Newberg (1997, 178–79) (discussing the “Nine No-No’s” as described in a 1970 speech by Deputy Assistant Attorney General Bruce Wilson). The nine practices included “tying of unpatented supplies; mandatory grantbacks; post-sale restrictions on resale by purchasers of patented products; tie-outs; licensee veto power over the licenser’s grant of further licenses; mandatory package licensing; compulsory payment of royalties in amounts not reasonably related to sales of the patented product; restrictions on sales of unpatented products made by a patented process; and specifying prices a licensee could charge upon resale of licensed products.” 17  See United States v. Grinnell Corp., 384 U.S. 563, 570–71 (1966). 18 U.S. Department of Justice and Federal Trade Commission (1995; § 2.1). 19  547 U.S. 28, 38–42 (2006).

Innovation and Antitrust Policy   141

to create or their own ability to license third parties.20 In such cases, antitrust courts may be thrust into the uncomfortable role of determining whether the restraint on balance tends to promote or restrain innovation. Another, perhaps more controversial example, is the determination of the European Court of First Instance in Microsoft v. Commission that Microsoft had abused its dominant position by refusing to license interoperability information to potential competitors in the market for work group server operating systems. In particular, the court held that Microsoft’s “refusal to supply the relevant information limited technical development to the prejudice of consumers,” and that Microsoft had not demonstrated any impairment of its ability to innovate.21 Other times a court’s consideration of the effects of a restraint or conduct on dynamic innovation may be less explicit. In all such instances, however, the question arises whether courts are according adequate weight to the incentives of both pioneers/ licensors and improvers/licensees; unfortunately, there is often no clear answer to this question, as discussed below.

7.3.  Antitrust’s Role in Promoting Innovation In theory, antitrust could play a much more aggressive role in promoting innovation than it currently does under either US or EU law. One could imagine, for example, a regime in which it would be an antitrust offense to use a patent or other IPR to extract rents that are grossly disproportionate to the IP owner’s costs of developing the IP at issue. Such a regime would work to constrain the deadweight loss attributable to some IPRs; it also might reduce the risk of patent holdup, and by narrowing the scope of potential bargaining outcomes it might reduce the cost of licensing negotiations. Access costs faced by developers of follow-on innovation would be expected to decrease for all of these reasons, while at the same time the incentives for pioneering inventors would be adequate as long as the pioneers were assured of recovering at least some rent in excess of their R & D costs. Such an approach would be a clear break from standard antitrust policy, which generally does not focus on whether a lawful monopolist is charging too high a price. As noted above, for example, Section 2 of the Sherman Act focuses on whether the monopolist is engaging in the willful acquisition or maintenance of monopoly power

20  See U.S. Department of Justice and Federal Trade Commission (1995; § 5.6) (stating that, in accordance with Transparent-Wrap Machine Corp. v. Stokes & Smith Co., 329 U.S. 637, 645–48 (1947), the agencies will evaluate grantbacks under the rule of reason, considering both the procompetitive benefits of encouraging initial invention and disclosure, on the one hand, and the potential anticompetitive harm of discouraging licensee-generated innovation, on the other). 21  Case T-201/04, 2007 E.C.R. II-3601, ¶¶ 331–32.

142   Thomas F. Cotter by excluding would-be rivals for reasons that lack a plausible procompetitive justification.22 US antitrust law, in other words, generally focuses on the risk of unjustifiable monopoly extension rather than on an otherwise lawful monopolist’s mere extraction of monopoly rents (Carlton and Heyer 2008, 287–88). In cases involving allegations of abuse of dominant position within the EU, the legal standard is a little closer to the theoretical one sketched out above, in that courts sometimes consider whether the antitrust defendant’s conduct is justified by the social interest in providing appropriate incentives to the developers of pioneering innovations. But even in this context, in cases such as Microsoft v. Commission, the principal focus is on whether the IP owner’s conduct (e.g., a refusal to deal) directly threatens the development of new products or services, and not on whether the owner is simply charging a disproportionately high price.23 As many analysts have observed, however, an antitrust regime that required courts to set prices in the manner of a public utility commission would be misguided for several reasons, the most obvious of which is judicial competence.24 Courts are not well positioned to receive evidence and make expert determinations of what an appropriate (incentive-preserving) rate of return on an investment in R & D would be, and thus a regime of this nature would be vulnerable not only to substantial administrative costs but to error costs as well. Errors resulting from courts’ inability accurately to correlate rewards with incentives in turn threaten to impede dynamic efficiency by undermining the incentives offered to pioneering innovators. Trying to determine how much of a reward is appropriate to preserve incentives therefore is probably not a productive avenue for a significant expansion of antitrust liability. That said, antitrust’s appropriate role in promoting innovation is not negligible, though it is more modest than the proposal sketched above. In what follows, I will suggest three ways in which antitrust works (or can work) to promote innovation: first, by playing its traditional role in policing anticompetitive conduct such as horizontal price fixing and market  allocation, notwithstanding the fact that the product the price of which is being fixed happens to be IP; second, by playing (if anything) a lesser role in certain instances in which analysis suggests that greater coordination is actually necessary to promote the efficient use or development of technological standards; and third, by playing a (marginally) more aggressive role in cases in which the risk of harm to dynamic efficiency in the form of nascent technology, even if small, outweighs the corresponding risk of harm to static and dynamic efficiency from limiting the exploitation of IPRs to some degree. I will briefly explain each in turn. 22 

See United States v. Grinnell Corp., 384 U.S. 563, 570–71 (1966). See Case T-201/04, Microsoft v. Commission, 2007 E.C.R. II-3601, ¶ 332 (stating that a refusal to license IP may constitute an abuse of discretion under exceptional circumstances, namely when (1) “the refusal relates to a product or service indispensable to the exercise of a particular activity on a neighbouring market”; (2) “the refusal is of such a kind as to exclude any effective competition on that neighbouring market”; and (3) “the refusal prevents the appearance of a new product for which there is potential consumer demand”). 24  See, e.g., Verizon Communications Inc. v. Law Offices of Curtis V. Trinko, LLP, 540 U.S. 398, 408 (2004) (“Enforced sharing also requires antitrust courts to act as central planners, identifying the proper price, quantity, and other terms of dealing—a role for which they are ill suited”); Hovenkamp et al. (2010, 13-10). 23 

Innovation and Antitrust Policy   143

First, antitrust promotes innovation by playing its traditional role of penalizing practices such as horizontal price fixing and other anticompetitive practices that offer no plausible procompetitive justification, even when such practices happen to involve IPRs. Where two firms agree to pool or package potentially competing patents, for example, in such a way as to control the price of the end products that embody those patents, antitrust enforcers rightly are not reluctant to perceive an agreement to fix prices that derives no justification from the fact that the agreement also involves patent rights.25 Relatedly, antitrust is replete with examples of purported settlements or pooling arrangements that in reality have been nothing more than disguised efforts at industry-wide collusion.26 Condemning such arrangements clearly promotes antitrust’s traditional static efficiency goals, but it also can be viewed as promoting innovation as well to the extent that competition spurs market participants to differentiate themselves by offering newer, more innovative products. Indeed, one might argue that, in practice, the general rule of not exempting IPRs from antitrust scrutiny is not always applied with sufficient consistency. Section 211 of the US Patent Act, for example, expressly states that nothing in the statute “shall be deemed to convey to any person immunity from civil or criminal liability, or to create any defenses to actions, under any antitrust law,” yet at the same time another provision of the same statute appears to permit patent owners to offer licenses under which the patent owner itself agrees to market its product exclusively in one geographic area, while the licensee may market it exclusively in another.27 Outside the IP realm such an agreement normally would be per se illegal. To be sure, such an arrangement might be reasonably ancillary to the efficient licensing of a patent in a geographic area the patentee could not serve directly; and in the absence of such an enforceable arrangement, the patentee might demand a higher licensing fee to compensate for the risk of competition from the licensee within the patentee’s own territory. Nevertheless, the blanket exemption for such conduct has no counterpart in the copyright statute and may be overly broad. Similarly, the minimal scrutiny that many US courts have applied to potentially anticompetitive settlements of pharmaceutical patent litigation arising under the Hatch-Waxman Act arguably does not take seriously enough the risk of collusion among potential competitors, particularly in cases in which the owner of a patent on a brand name drug pays its would-be generic competitor more than the generic producer could have expected to earn from sales of the generic drug.28 To the extent courts have deviated from conventional antitrust principles in this latter class of cases, however, the resulting harm would appear to be fall more heavily on static efficiency (consumers paying more than is necessary to obtain pharmaceutical products) than on dynamic efficiency, which is not one of the expected benefits of generic competition. Yet another example, this one with potentially more serious consequences 25 

See, e.g., In the Matter of Summit Technology, 127 F.T.C. 208 (1999). See, e.g., United States v. New Wrinkle, Inc., 342 U.S. 371 (1952); United States v. United States Gypsum Co., 333 U.S. 364 (1948). 27  See 35 U.S.C. § 261 (stating that the patentee “may . . . grant and convey an exclusive right under his . . . patents . . . to the whole or any specified part of the United States”). 28  For an overview and citations to relevant literature, see Cotter (2008, 797–801). 26 

144   Thomas F. Cotter for dynamic efficiency, is the US courts’ arguably overcautious approach to Walker Process claims, that is, antitrust claims founded on allegations that a patentee acquired monopoly power by practicing fraud on the patent office.29 Second, however, there may be circumstances in which antitrust can promote innovation by showing greater leniency than one might otherwise expect toward joint conduct intended to make technology more widely available.30 While this might sometimes take the form of greater tolerance for, say, joint ventures organized to promote a new technology, in other instances it might involve greater tolerance for joint conduct on the part of IP users. Both examples may arise in connection with standard-setting organizations (SSOs). Because a certain degree of collaboration is necessary whenever competitors or potential competitors negotiate the adoption of technological standards, antitrust has long recognized both the potential benefits and costs of such collaboration—though it properly takes a skeptical approach to features of any such agreement that (for example) appear to serve no purpose other than placing potential rivals at a disadvantage.31 More interesting are features found in, or proposed for use in, some contemporary standard-setting arrangements under which members agree to various obligations that can be understood as serving to reduce the risk of patent holdup—for example, requirements that members disclose issued or pending patents that might read on a proposed standard; or that they license any such patents on fair, reasonable, and nondiscriminary (FRAND) terms; or (more controversial yet) that such licenses may not exceed a certain rate. The more extensive the obligation is, the greater the risk that it is nothing more than glorified price fixing; at the same time, some such obligations may be reasonably necessary to ensure that firms are willing to commit to the standard without undue risk of falling victim to opportunistic conduct on the part of other members. In recent years, many commentators (including the US Antitrust Modernization Commission) have called for greater recognition of the potential procompetitive benefits of such agreements.32 My own work has expressed agreement that in some such instances the best antitrust can do to promote innovation is to get out of the way—by which I mean not that antitrust should grant a blanket exemption from liability for any conduct that SSOs may wish to adopt, but only that a antitrust regime that is sensitive to the social benefits of interoperability, on the one hand, and to the (arguably marginal) benefits of strong patent rights in sectors such as information technology, properly may conclude that SSO conduct that plausibly and proportionately serves to limit holdup risks is on balance procompetitive. The same presumption, however, may not apply in less technologically dense industries, or in 29 

For critical discussions, see Bohannan and Hovenkamp (2012, 85–90); Leslie (2008). Reflective of the Schumpeterian perspective, one school of thought suggests that antitrust should take more of a deferential approach to certain types of conduct in “New Economy” industries, by focusing more on promoting competition for the market than on promoting competition in the market. See, e.g., Evans and Schmalensee (2002). 31  See, e.g., Allied Tube & Conduit Corp. v. Indian Head, Inc., 486 U.S. 492, 496–98, 509–10 (1988); Am. Soc’y of Mech. Eng’rs, Inc. v. Hydrolevel Corp., 456 U.S. 556, 577–78 (1982); Radiant Burners, Inc. v. Peoples Gas Light & Coke Co., 364 U.S. 656, 659–60 (1961) (per curiam). 32  For discussion and citations, see Cotter (2009, 1200–1206); Cotter (2008, 758–69, 786–94). 30 

Innovation and Antitrust Policy   145

industries such as pharmaceuticals in which patent owner incentives arguably are substantial; and courts must remain vigilant for evidence of agreements that adopt requirements that go beyond what is reasonably necessary to avoid holdup, for example requirements relating to the price at which technology-embodying products will be sold. Third, it may be the case that antitrust should be more aggressive—less concerned about false positive risks—in circumstances in which the conduct at issue poses even an objectively small risk to future innovation, if that risk (should it come to pass) threatens substantial social harm, and the potential procompetitive benefits of tolerating the conduct at issue are only modest. Imagine, for example, a case in which Firm A and Firm B each own a portfolio of patents relating to a technological standard. The two firms agree to establish a joint venture that will be authorized to offer other firms a nonexclusive package license of patents essential to the marketing of products conforming to the standard. Both firms also agree, however, not to license any of the patents other than through the joint venture; in addition, Firm B, which owns a patent (Patent X) claiming a technological solution different from the one that is enabled by the patents vested in the joint venture, agrees not to license Patent X to anyone. In theory, those portions of the agreement preventing the licensing of the essential patents other than through the joint venture and forbidding the licensing of Patent X at all could be procompetitive to the extent they are reasonably ancillary to the success of the joint venture itself: in a sense, A and B are both agreeing not to undermine the joint venture by competing against it. And to the extent the venture enables others to participate in the use of the technological standard at issue, its existence is procompetitive not only in the short-run sense but also, potentially, in the longer run as well, insofar as firms that license from the venture may be better poised to develop newer and better products by accessing the standard. At the same time, the risks to innovation should be apparent too. Firms that might seek to innovate by licensing only a subset of the patents cannot do so (though whether this puts them at a disadvantage depends on whether the price for the package exceeds the hypothetical price for a single license). More troubling, perhaps, is the fact that no one can use Patent X to develop a competing technology that may (or may not) turn out to have farther-reaching applications than the standard enabled by the package license. Perhaps, in such a case, antitrust should take a hard line against such a hypothetical agreement even if the alternative technology is merely “nascent” or even altogether unproven in its capabilities. Even though antitrust normally might require a stronger of potential anticompetitive consequences before condemning such an arrangement,33 perhaps in such a case even a small risk of substantial harm to dynamic innovation should not be tolerated, particularly if there are less restrictive alternatives that would provide an adequate, albeit not foolproof, assurance that Firm B will not undermine the venture. Perhaps as well, antitrust and patent law should be open to novel theories of private party standing, application of FTC § 5

33  See Princo Corp. v. Int’l Trade Comm’n, 616 F.3d 1318, 1338–40 (Fed. Cir. 2010) (en banc), cert. denied, 131 S. Ct. 2480 (2011) (noting that the “suppression of nascent threats can be construed as

146   Thomas F. Cotter actions, or expansions of the patent misuse doctrine to permit a challenge to the hypothetical agreement to come forward. Readers may recognize in the above hypothetical an adaptation of the facts alleged by certain patent infringement defendants in Princo v. International Trade Commission, a case in which the US Court of Appeals for the Federal Circuit decided that an allegation of technology suppression on the part of the joint venturers was insufficient to render the infringed patent unenforceable under the US doctrine of patent misuse.34 In other work, I have expressed agreement with the Princo court’s narrow interpretation of the patent misuse defense, while reserving judgment on whether a stronger antitrust response to conduct like that alleged by the infringement defendants would be desirable. For present purposes, I will note only that such a response may be warranted, assuming there were evidence of such an agreement among the two patent owners, even if the probability that the alternative technology would have been successful is somewhat speculative. If, as suggested above, the benefits from innovation are in general much higher than the benefits resulting from improvements in short-run efficiency, the better view might be to require the poolants to keep the alternative available for licensing by others. More generally, one might imagine other circumstances in which a restraint poses a small but nontrivial risk to future innovation while offering only relatively modest short-term efficiency benefits;35 perhaps some mergers that implicate so-called innovation markets would fall into this category.36 A more difficult situation arises when the purported justification itself relates to longer-term, dynamic efficiency concerns. The example above involving an agreement among SSO members to cap the price to be paid for a given patent would be one in which there may be plausible dynamic efficiency benefits on both sides. The obvious problem, once we accept the principle that any conduct that threatens some harm to innovation or creativity (no matter how speculative) properly could give rise to antitrust liability, is knowing where to stop. To illustrate, under EU law, a license agreement that forbids the licensee from challenging the validity of a licensed patent is excluded from the block antitrust exemptions given to certain other features anticompetitive behavior under certain circumstances,” but concluding that Princo bore the burden of proving, and had failed to prove, more than just a “speculative possibility” of anticompetitive harm). 34  The Federal Circuit defines misuse as the “impermissibly broaden[ing of] the ‘physical or temporal scope’ of a patent with anticompetitive effect.” Princo, 616 F.3d at 1328; Windsurfing Int’l, Inc. v. AMF, Inc., 782 F.2d 995, 1001 (Fed. Cir. 1986) (citing Blonder-Tongue Labs., Inc. v. Univ. of Ill. Found., 402 U.S. 313, 343 (1971)). 35  I develop this line of reasoning mathematically in Cotter (2006, 520–37). 36  See US Department of Justice and Federal Trade Commission (1995; § 3.2.3); Gilbert and Sunshine (1995); see also Cotter (2006, 525–26 and n. 176) (discussing relevant literature). For analysis of the role innovation should play in merger analysis, see Carrier (2009, 92–93, 295–322); Katz and Shelanski (2007). More controversially, Tim Wu (2010, 96–97) has suggested that the Supreme Court’s quondam practice of subjecting conduct such as, say, block booking of motion pictures, to serious antitrust scrutiny—an approach that seems quite dated under most contemporary understandings of antitrust economics—could play a positive role in promoting greater innovation in the production of creative expression.

Innovation and Antitrust Policy   147

of technology and R & D licenses, and thus is subject to challenge under article 81 of the EC Treaty.37 In theory, this rule could promote both short- and long-run efficiency by more effectively screening out invalid patents (the enforcement of which, by definition, is presumed to generate an inadequate or negative surplus of social benefits over costs). But the rule also forces the patentee to bear the risk of a licensee-initiated challenge, for which the patentee presumably will demand a higher licensing fee; at the margin, then, it may decrease the amount of licensing, which seems undesirable. Moreover, given the ability of other parties to challenge patent validity (in the major European nations in particular, anyone may do so at any time) and given the fact that most patents do not actually confer substantial market power, the prohibition on no-challenge clauses seems unduly broad, indeed potentially counterproductive to an effective innovation policy. My own view is that courts should apply the more expansive liability standard proposed above only in cases in which one can be highly confident that the potential shortand long-term benefits of permitting the restraint at issue are minimal. Such cases should be rare but not an empty set. In particular, it may be possible in some instances to conclude that the countervailing efficiencies are small, given factors such as the approximate cost of the R & D needed to generate the IP at issue, the strength of first-mover advantages or other empirical evidence of the relative importance of patent protection in a given industry, the presence of network effects that have already generated a supracompetitive return, and the like. Taking account of these factors does not necessarily run afoul of my admonition that courts should not try to determine appropriate rates of return in the manner of a public utility regulator, since courts would not be setting the amount of remuneration owed to the patent owner. Even so, in recognition of their institutional limitations, courts should exercise caution in determining that forbidding a given restraint will not have a negative impact on incentives. In brief, the approach sketched out above is simply a call for antitrust enforcers to show appropriate sensitivity both to what can be stated with confidence about the likely impact of market structure and conduct in specific innovation environments, and to the limits of their ability to predict and manage that environment far out into the future. As such, the approach broadly resembles an approach described in a recent paper in which Carl Shapiro (2012) argues that antitrust enforcers should not overexaggerate the difficulties of predicting the consequences of present conduct for future 37 

Article 81 of the EC Treaty—now codified as article 101 of the Treaty on the Functioning of the European Union—generally prohibits, and renders automatically void, agreements “which may affect trade between Member States and which have as their object or effect the prevention, restriction or distortion of competition within the common market, and in particular those which . . . limit or control production, markets, technical development, or investment.” The Technology Transfer and Research and Development Block Exemption Regulations exempt certain restraints from the scope of Article 81, but both regulations expressly do not exempt no-challenge clauses. See Commission Regulation (EC) No. 772/2004 of 27 April 2004 on the application of Article 81(3) of the Treaty to categories of technology transfer agreements, art. 5(1)(c); Commission Regulation (EC) No. 2659/2000 of 29 November 2000 on the application of Article 81(3) of the treaty to categories of research and development agreements, art. 5(1)(b). Ulmer-Eilfort and Boulware (2007, 769) assert that “today there are only very narrow borders for such clauses being effective.”

148   Thomas F. Cotter innovation. Instead, Shapiro proposes, policymakers should work to integrate the Arrovian and Schumpeterian perspectives into a competition policy based on principles of “Contestability” (“The prospect of gaining or protecting profitable sales by providing greater value to customers spurs innovation”); “Appropriability” (“Increased appropriability spurs innovation”); and “Synergies” (“Combining complementary assets enhances innovation capabilities and thus spurs innovation”). In a similar vein, one might argue that antitrust should make greater use of the empirical literature on the impact of market structure on innovation—as described by Bohannan and Hovenkamp (2012, 9), “there is probably more empirical literature on the relation between competition and innovation than on any topic in the field of industrial organization economics,” including influential work by Aghion and colleagues (2005)—though the nature of these studies may be such that they do not properly lend themselves to simple application.38 And by now the pioneering insights of economists in the 1980s on network externalities (see, e.g., Farrell and Saloner 1985; Katz and Shapiro 1985) are already part of the standard antitrust toolkit (as reflected most famously in the monumental Microsoft antitrust litigation), thus lending credence to the views of the Antitrust Modernization Commission (2007, 31) that “antitrust analysis has sufficient grounding in sound economic analysis, openness to new economic learning, and flexibility to enable the courts and the antitrust agencies properly to assess competitive issues in new economy industries.” Nevertheless, to the extent patent and other IP laws are perceived as conferring excessive protection or otherwise as undermining, rather than advancing, their stated purpose of promoting the progress of science and the useful arts, for the most part reform must come from the IP side, not the antitrust side. There is little if anything that antitrust, consistent with rule-of-law values, can or should do if legislatures set excessive terms of protection, confer patents on trivial inventions, or permit inventors to capture the value of others’ improvements by filing series of continuation applications.39 Antitrust’s role in promoting innovation is important but nevertheless constrained by the limited reach of the statute and by courts’ competence to second-guess legislative judgments about the appropriate scope of IPRs.

7.4. Conclusion I have argued above that antitrust has a useful, albeit limited, role to play in fostering innovation. Much of the time antitrust best plays this role simply by doing what it does best, namely focusing on practices that threaten harms to static efficiency, even when the conduct at issue happens to involve IP. At other times, for example in cases involving 38 

For overviews of the major studies and discussion of some possible limitations, see Gilbert (2006, 187–204); Greenhalgh and Rogers (2010, 140–42); Shapiro (2012, 370–82). 39  On this last point in particular, see Lemley and Moore (2004).

Innovation and Antitrust Policy   149

joint conduct on the part of SSO members, courts arguably should play a more deferential role, to the extent joint conduct is reasonably necessary for the efficient development or deployment of new technology. In yet other instances, however, US courts in particular arguably have accorded too much deference to IPRs, for example in connection with Walker Process and Hatch-Waxman litigation; and in theory a more aggressive antitrust policy could sometimes improve social welfare, to the extent such a policy can (for example) enable greater interoperability with or new uses of pioneering technology without simultaneously killing off the incentive to create such technology in the first place. Most of the heavy lifting, though, must come from other bodies of law, most notably the IP laws themselves. Antitrust tribunals are neither well positioned nor legally competent to make many of the fundamental decisions concerning the optimal scope and duration of IPRs; and even within antitrust’s own domain, uncertainties over the effects of market structure and other variables on long-term innovation counsel in favor of a relatively modest approach. Put another way, in the competition over competition policy, policymakers should ensure that antitrust itself does not come to enjoy too large a market share.

References Aghion, Philippe, Nick Bloom, Richard Blundell, Rachel Griffith, and Peter Howitt. 2005. Competition and Innovation: An Inverted U-Relationship. Quarterly Journal of Economics 120(2):701–28. Antitrust Modernization Commission. 2007. Report and Recommendations. Washington, DC: Arrow, Kenneth J. 1962. Economic Welfare and the Allocation of Resources for Invention. In Richard R. Nelson, The Rate and Direction of Economic Activity: Economic and Social Factors. Princeton NJ: Princeton University Press. Barnett, Jonathan M. 2011. Intellectual Property as a Law of Organization. Southern California Law Review 84:785–857. Beck, Roger L. 1983. The Prospect Theory of the Patent System and Unproductive Competition. Research in Law and Economics 5:193–209. Bessen, James, and Michael J.  Meurer. 2008. Patent Failure:  How Judges, Bureaucrats, and Lawyers Put Innovation at Risk. Princeton, NJ: Princeton University Press. Bohannan, Christina, and Herbert Hovenkamp. 2012. Creation without Restraint: Promoting Liberty and Rivalry in Innovation. New York: Oxford University Press. Boldrin, Michele, and David K.  Levine. 2008. Against Intellectual Monopoly. New York: Cambridge University Press. Burk, Dan L., and Mark A. Lemley. 2009. The Patent Crisis and How the Courts Can Solve It. Chicago: University of Chicago Press. Burk, Dan L., and Brett H. McDonnell. 2007. The Goldilocks Hypothesis: Balancing Intellectual Property Rights at the Boundary of the Firm. University of Illinois Law Review 2007:575–636. Carlton, Dennis W., and Ken Heyer. 2008. Extraction vs. Extension: The Basis for Formulating Antitrust Policy towards Single-Firm Conduct. Competition Policy International 4 (Autumn):285–305. Carrier, Michael. 2009. Innovation for the 21st Century: Harnessing the Power of Intellectual Property and Antitrust Law. New York: Oxford University Press.

150   Thomas F. Cotter Cohen, Wesley M., Richard R. Nelson, and John P. Walsh. 2000. Protecting Their Intellectual Assets: Appropriability Conditions and Why U.S. Manufacturing Firms Patent (or Not). NBER Working Paper No. 7552 (February). Available at http:// www.nber.org/papers/w7552. Cotropia, Christopher A., and Mark A. Lemley. 2008. Copying in Patent Law. North Carolina Law Review 87:1421–66. Cotter, Thomas F. 2006. The Procompetitive Interest in Intellectual Property Law. William and Mary Law Review 48:483–557. Cotter, Thomas F. 2008. Reflections on the Antitrust Modernization Commission’s Report and Recommendations Relating to the Antitrust/IP Interface. Antitrust Bulletin 53:745–801. Cotter, Thomas F. 2009. Patent Holdup, Patent Remedies, and Antitrust Responses. Journal of Corporation Law 34:1151–207. Dam, Kenneth W. 1994. The Economic Underpinnings of Patent Law. Journal of Legal Studies 23:247–71. Duffy, John F. 2004. Rethinking the Prospect Theory of Patents. University of Chicago Law Review 71:439–510. Evans, David S., and Richard Schmalensee. 2002. Some Economic Aspects of Antitrust Analysis in Dynamically Competitive Industries. In Adam B.  Jaffe, Josh Lerner, and Scott Stern, Innovation Policy and the Economy, 2:1–49. Cambridge, MA: National Bureau of Economic Research. Farrell, Joseph, and Garth Saloner. 1985. Standardization, Compatibility and Innovation. Rand Journal of Economics 16:70–83. Fisher III, William W., and Talha Syed. Forthcoming. Infection:  The Health Crisis in the Developing World and What We Should Do About It. Palo Alto, CA: Stanford University Press. Gilbert, Richard. 2006. Looking for Mr. Schumpeter: Where Are We in the Competition-Innovation Debate. In Adam B. Jaffe, Josh Lerner, and Scott Stern, Innovation Policy and the Economy 6:159–215. Cambridge, MA: National Bureau of Economic Research. Gilbert, Richard J., and Steven C. Sunshine. 1995. Incorporating Dynamic Efficiency Concerns in Merger Analysis: The Use of Innovation Markets. Antitrust Law Journal 63:569–601. Grady, Mark F., and Jay I.  Alexander. 1992. Patent Law and Rent Dissipation. Virginia Law Review 78:305–50. Graham, Stuart J.  H., Robert P.  Merges, Pam Samuelson, and Ted Sichelman. 2009. High Technology Entrepreneurs and the Patent System: Results of the 2008 Berkeley Patent Survey. Berkeley Technology Law Journal 24:1255–327. Green, Jerry R., and Suzanne Scotchmer. 1995. On the Division of Profit in Sequential Innovation. Rand Journal of Economics 26:20–33. Greenhalgh, Christine, and Mark Rogers. 2010. Innovation, Intellectual Property, and Economic Growth. Princeton, NJ: Princeton University Press. Hall, Bronwyn H., and Rosemarie Ham Ziedonis. 2001. The Patent Paradox Revisited:  An Empirical Study of Patenting in the U.S. Semiconductor Industry, 1979–1995. Rand Journal of Economics 32:101–28. Heller, Michael A., and Rebecca S.  Eisenberg. 1998. Can Patents Deter Innovation? The Anticommons in Biomedical Research. Science 280:698–701. Hovenkamp, Herbert, Mark D. Janis, Mark A. Lemley, and Christopher Leslie. 2010. IP and Antitruast: An Analysis of Antitrust Principles Applied to Intellectual Property Law. 2nd ed. Austin, TX: Aspen Publishers.

Innovation and Antitrust Policy   151

Katz, Michael L., and Carl Shapiro. 1985. Network Externalities, Competition, and Compatibility. American Economic Review 75:424–40. Katz, Michael L., and Howard A.  Shelanski. 2007. Mergers and Innovation. Antitrust Law Journal 74:1–84. Kieff, F.  Scott. 2001. Property Rights and Property Rules for Commercializing Inventions. Minnesota Law Review 85:697–754. Kitch, Edmund W. 1977. The Nature and Function of the Patent System. Journal of Law and Economics 20:265–90. Lemley, Mark A., and Kimberly A. Moore. 2004. Ending Abuse of Patent Continuations. Boston University Law Review 84:63–118. Leslie, Christopher R. 2008. Patents of Damocles. Indiana Law Journal 83:133–79. Long, Clarisa. 2002. Patent Signals. University of Chicago Law Review 69:625–79. Merges, Robert P., and Richard R. Nelson. 1990. On the Complex Economics of Patent Scope. Columbia Law Review 90:839–916. Schumpeter, Joseph A. [1942] 2003. Capitalism, Socialism and Democracy. Taylor and Francis e-Library. Scotchmer, Suzanne. 1991. Standing on the Shoulders of Giants: Cumulative Research and the Patent Law. Journal of Economic Perspectives 5:29–41. Shapiro, Carl. 2012. Competition and Innovation: Did Arrow Hit the Bull’s Eye? In Josh Lerner and Scott Stern, The Rate and Direction of Inventive Activity Revisited. Chicago: University of Chicago Press. Sichelman, Ted. 2010. Commercializing Patents. Stanford Law Review 62:361–411. Tom, Willard K., and Joshua A.  Newberg. 1997. Antitrust and Intellectual Property:  From Separate Spheres to Unified Field. Antitrust Law Journal 66:167–229. Ulmer-Eilfort, Constanze, and Margaret A. Boulware. 2007. Challenging Licensed Intellectual Property Rights under U.S.  and European Laws:  The U.S. Supreme Court Abandons the Principle of Licensee Estoppel. IIC 38:759–69. Walsh, John P., Charlene Cho, and Wesley M. Cohen. 2005. View from the Bench: Patents and Material Transfers. Science 309:2002–3. Warsh, David. 2006. Knowledge and the Wealth of Nations: A Story of Economic Discovery. New York: Norton. Wu, Tim. 2010. The Master Switch: The Rise and Fall of Information Empires. New York: Alfred A. Knopf.

Cases Allied Tube & Conduit Corp. v. Indian Head, Inc., 486 U.S. 492 (1988). Am. Society of Mechanical Engineers, Inc. v. Hydrolevel Corp., 456 U.S. 556 (1982). Bement v. National Harrow Co., 186 U.S. 70 (1902). Blonder-Tongue Labs., Inc. v. University of Illinois Foundation, 402 U.S. 313 (1971). Henry v. A.B. Dick Co., 224 U.S. 1 (1912). Independent Ink, Inc. v. Illinois Tool Works Inc., 547 U.S. 28 (2006). Microsoft v. Commission, 2007 E.C.R. II-3601, Case T-201/04. Motion Picture Patents Co. v. Universal Film Mfg. Co., 243 U.S. 502 (1917).

152   Thomas F. Cotter Princo Corp. v. Int’l Trade Comm’n, 616 F.3d 1318 (Fed. Cir. 2010) (en banc), cert. denied, 131 S. Ct. 2480 (2011). Radiant Burners, Inc. v. Peoples Gas Light & Coke Co., 364 U.S. 656 (1961) (per curiam). Simpson v. Union Oil Co. of Cal., 377 U.S. 13 (1964). In the Matter of Summit Technology, 127 F.T.C. 208 (1999). Transparent-Wrap Machine Corp. v. Stokes & Smith Co., 329 U.S. 637 (1947). United States v. General Electric Co., 272 U.S. 476 (1926). United States v. Grinnell Corp., 384 U.S. 563 (1966). United States v. New Wrinkle,Inc., 342 U.S. 371 (1952). United States v. United States Gypsum Co., 333 U.S. 364 (1948). Verizon Communications Inc. v. Law Offices of Curtis V. Trinko, LLP, 540 U.S. 398 (2004). Windsurfing Int’l, Inc. v. AMF, Inc., 782 F.2d 995 (Fed. Cir. 1986).

Statutes, Regulations, Guidelines, and Treaties Agreement on Trade Related Aspects of Intellectual Property Rights (TRIPs), April 15, 1994, Marrakesh Agreement Establishing the World Trade Organization, Annex 1C, arts. 27, 33, Legal Instruments—Results o the Uruguay Round vol. 31, 33 I.L.M. 1197 (1994). Commission Regulation (EC) No. 772/2004 of 27 April 2004 on the application of Article 81(3) of the Treaty to categories of technology transfer agreements. Commission Regulation (EC) No. 2659/2000 of 29 November 2000 on the application of Article 81(3) of the Treaty to categories of research and development agreements. Clayton Act § 3, 15 U.S.C. § 14. Consolidated Version of the Treaty on the Functioning of the European Union art. 101, May 9, 2008, 2008 O.J. (C115) 47. Sherman Act § 2, 15 U.S.C. § 2. US Patent Act (35 U.S.C.) §§ 211, 261. US Department of Justice & Federal Trade Commission (1995), Antitrust Guidelines for the Licensing of Intellectual Property.

CHAPTER 8

C ON T I N E N TA L D R I F T I N T H E T R E AT M E N T O F D OM I NA N T F I R M S Article 102 TFEU in Contrast to Section 2 Sherman Act PIERRE LAROUCHE AND MAARTEN PIETER SCHINKEL

The policing of significant marker power remains one of the largest areas of debate in comparative competition law. While the two leading systems, US antitrust law and EU competition law, both originate in the recognition of a need to control monopoly market power, they were set on different tectonic plates that drifted on their own individual paths. The interpretation and application of both Section 2 of the Sherman Act and Article 102 of the TFEU have been reviewed over the years. Recently, a number of leading cases have been issued on both sides of the Atlantic. While in the past decade the two systems have converged on a focus on exclusionary conduct, a number of fundamental underlying differences remain. On the fault lines, there is potential for tension, as well as motion. The aim of this chapter is to take stock of the debate, at a fundamental level. Our contribution is written from a European perspective, that is, we take US antitrust law as granted and we seek to explain particular aspects of EU competition law against it. In that, we are fully aware that US antitrust law is more diverse than presented here. Much as our representation of EU competition law is based on the most authoritative statements by the European courts and the European Commission, the representation of US law here given relies on the most authoritative primary sources. Without going in the detail of the laws, we identify a number of distinctive features of Article 102, both in wording and interpretation, in section 8.1. We subsequently discuss three lines of argument to explain these differences. The first one, considered in section 8.2, is traditionally put forward:  the interpretation of Article 102 reflects the influence of the ordoliberal school of thought on EU competition law. We will provide some new insights from this point of view. The second, more contemporary line of argument derives from the observation that competition law enforcement is fallible and is primarily institutional, as explained in section 8.3. The third argument, set out in section 8.4, introduces

154   Pierre Larouche and Maarten Pieter Schinkel dynamic elements into the analysis, whereby the interpretation of Article 102 would reflect a specifically European view of innovation policy. In section 8.5, we return to the underutilized EU category of exploitative abuses and argue that economic techniques developed in the context of damages litigation could open it up for future enforcement in a way that would be in line with ordoliberal principles. We conclude in section 8.6.

8.1.  Distinctive Features in the Wording and Interpretation of Article 102 TFEU 8.1.1.  Limited Differences in Wording The core parts of the two respective legal texts compare as follows: Section 2 Sherman Act, 15 U.S.C. § 2 Every person who shall monopolize, or attempt to monopolize, or combine or conspire with any other person or persons, to monopolize any part of the trade or commerce among the several States, or with foreign nations, shall be deemed guilty of a felony. Article 102 TFEU Any abuse by one or more undertakings of a dominant position within the internal market or in a substantial part of it shall be prohibited as incompatible with the internal market in so far as it may affect trade between Member States.

While Section 2 of the Sherman Act was enacted in 1890 as part of criminal law, with monopolization being classified as a felony, Article 102 TFEU was inspired by European administrative laws and inserted in the international Treaty of Rome in 1957.1

8.1.1.1.  “Monopolization” versus “Abuse”: Convergence on Exclusionary Behavior At first glance, “monopolization” and “abuse” seem different concepts. With the term “monopolization,” Section 2 would focus on how a monopoly position is obtained or maintained, and not so much on the actions of the monopolist once that position has been achieved, whereas Article 102 would not pay attention so much to how a dominant 1 

As a consequence, violations carry different types of sanctions, with prison sentences being unavailable at the European level—as opposed to the level of individual member states, some of which have adopted criminal sanctions for individuals in recent years, albeit first and foremost for cartel infringements, not abuses of dominance. See Cseres, Schinkel, and Vogelaar (2006).

Continental Drift in the Treatment of Dominant Firms   155

position has arisen, but would instead police particular abusive actions of the dominant firm (Elhauge and Geradin, 2011). While this distinction appears to be supported by the respective wording of these provisions, it is nowhere near as sharp as it is made out to be, once case law is taken into account. Over time, as case law and literature accumulated, these two provisions have arguably evolved away from what some presented as their original intent, and converged in their coverage. Section 2 was used in early cases to attack market power directly,2 yet it was reframed subsequently. Similarly, in the early days of Article 102, part of the commentary argued that the provision was aimed at policing the abusive exploitation of market power only (Joliet, 1970). However, it suffices to look at two recent administrative statements on these respective provisions—DOJ (2008) and European Commission (2009)—to see that as they are construed now, both provisions are primarily used to deal with exclusionary behavior on the part of firms holding significant market power. Their application requires a showing of both significant market power (monopoly power or dominance, respectively) and of exclusionary conduct (or abuse, in EU competition law). Conceptually, under EU law, an undertaking is dominant if it can behave “to an appreciable extent independently of its competitors, customers and ultimately of its consumer.”3 The prevailing measure for market power in both systems is market share in the relevant market. Under EU law, a (combined) market share of 38% can be sufficient to raise the rebuttable presumption of a dominant position (a 40% threshold is put forward in European Commission, 2009), whereas in the United States interventions based on Section 2 will usually require a market share of over 50% (DOJ, 2008).

8.1.1.2.  An EU Outlier: Exploitative Abuses While the main focus of both Section 2 and Article 102 has therefore shifted over time to lie on exclusionary abuses of market power, Article 102 TFEU continues to harbor an outlier, namely the notion of “exploitative abuses.”4 This is the use of dominance to extract rents from customers, by way of excessive pricing, for instance. While Section 2 of the Sherman Act could theoretically extend to excessive pricing, the hurdles are very high, as the Supreme Court most emphatically reaffirmed in Trinko, and accordingly no excessive pricing cases have been brought in ages.5 As the Supreme Court conceives it, exploitative behavior is quite the opposite from monopolization, on the logic that high prices would rather make customers look for alternatives and attract competitors to the market in question. In Europe, the prohibition of exploitative abuses under Article 102 has hardly been enforced, with only a handful of cases in which it was the main finding of an infringement. It is linked with the specificities of Article 102, as stated above, in ways that will be explained later in this chapter.6 2 

Most famously in Standard Oil v. US 221 U.S. 1 (1911). ECJ, Case C-27/76, United Brands [1978] ECR 207. 4  As evidenced for instance in Article 102(2)a) TFEU: “directly or indirectly imposing unfair purchase or selling prices or other unfair trading conditions.” 5  Verizon Communications v. Trinko, 540 U.S. 398 (2004). 6  Infra, section 8.5. 3 

156   Pierre Larouche and Maarten Pieter Schinkel EU law – Article 102 TFEU

Exploitative behavior

FIGURE  8.1  Coverage

US law - § 2 Sherman Act

Exclusionary behavior - Dominance/monopoly power - Abuse/exclusionary conduct

Attempted monopolization

of Article 102 TFEU and Section 2 Sherman Act

8.1.1.3.  A US Outlier: Attempted Monopolization For Section 2 of the Sherman Act, the outlier is “attempted monopolization,” which is construed along similar lines as monopolization, yet does not require the violation to have been successful in realizing its objective (DOJ, 2008). Establishing attempted monopolization requires proof that the defendant intended to destroy competition or build monopoly, which resulted in a dangerous probability of achieving monopoly power (Areeda and Hovenkamp, 2013). However, courts have been reluctant to find intent, and only few cases were tried, none in recent years (DOJ, 2008). Together, these similarities and differences can be illustrated in the Venn diagram in ­figure 8.1. Note that the outlier categories are smaller even, in terms of cases tried, than they appear. Despite convergence to a large overlap between the two provisions in terms of substantive coverage on exclusionary behavior, their interpretation and application shows a number of important divergences.

8.1.2.  Essential Differences in Interpretation Both Section 2 and Article 102 are brief and generally formulated, leaving ample room for interpretation. Famously, Frank Easterbrook (1986b, 1702) wrote: Back in 1890 Senator Sherman and colleagues protested the Sugar Trust and other malefactors and told the judiciary to do something about it. They weren’t sure just what. Their statute does not contain a program; it is instead a blank check.

Only in case law did their interpretation emerge. Likewise have Articles 81 and 82 of the Treaty of Rome, apart from two renumberings, had to go through a long process of case law developments to harden in their interpretation. As one of major figures in the development of EU competition law, Ernst-Joachim Mestmäcker (2011, 35), stated: Politicians and legislators love the ambivalence of final purposes, leaving their determination to the ensuing political or legal process. The EC’s ‘founding fathers’ were particularly resourceful in this respect. Competition rules prove the point. The

Continental Drift in the Treatment of Dominant Firms   157

drafters of the Treaty charged the European Court and the EC Commission with the task of injecting into deliberately general provisions the light of precise purpose and meaning.

In this subsection, we discuss four distinctive features of EU competition law, arising from the Commission decision practice and the case law of the European courts.

8.1.2.1.  The Special Responsibility of the Dominant Firm Since the early years of the Sherman Act, US courts have applied Section 2 with a view to avoiding, to the greatest extent possible, restrictions on the commercial freedom of firms, even if they hold monopoly power. In 1919 in Colgate the Supreme Court held that,“[i]‌n the absence of any purpose to create or maintain a monopoly, the act does not restrict the long recognized right of trader or manufacturer engaged in an entirely private business, freely to exercise his own independent discretion as to parties with whom he will deal.”7 More recently, in Trinko and Linkline, the Supreme Court reaffirmed this stance forcefully.8 In Linkline, after having cited Colgate, the Court presented the scope of Section 2 in the following terms: “But there are rare instances in which a dominant firm may incur antitrust liability for purely unilateral conduct.”9 The theory underpinning the US approach is well set out by Easterbrook (1986a, 2003). In a number of circumstances, it is beneficial for the economy and for consumers if firms holding monopoly power compete hard with other firms. In other circumstances, the conduct of those firms can be exclusionary. Since it is hard to distinguish between competitive and exclusionary conduct, it is better to allow firms to compete hard and put the burden upon authorities, in exceptional circumstances, to show that the conduct was in fact exclusionary and in breach of Section 2. In contrast, the European courts and the European Commission, in their interpretation of Article 102 TFEU, do not begin with the assumption that firms holding a dominant position can compete as any other firm would. Rather, the case law and decision practice under Article 102 TFEU constantly and consistently refers to the “special responsibility of the dominant firm not to allow its conduct to impair genuine undistorted competition on the common market.”10 The special responsibility of the dominant firm was conspicuously absent from the European Commission (2005) discussion paper circulated by Directorate-General for Competition (DG COMP), but it has been given pride of place again in the Guidance Paper (European Commission, 2009). 7 

US v. Colgate, 250 U.S. 300 at 307 (1919). Trinko, supra, note 5 at 407, as quoted infra. See the quote made infra, section 8.4. 9  Pacific Bell Telephone v. Linkline Communications, 129 S.Ct. 1109 at 1118 (2009). 10  This statement was made in one of the early Article 102 TFEU cases: ECJ, Case 322/81, Michelin v. Commission [1983] ECR 3461 at para. 57 and repeated ever since. 8 

158   Pierre Larouche and Maarten Pieter Schinkel

8.1.2.2.  Competition on the Merits Under Section 2 of the Sherman Act, there is a notion that some means of competition are allowable, and others will breach the act. As the Supreme Court put it in US v. Grinnell:11 “The offense of monopoly under § 2 of the Sherman Act has two elements: (1) the possession of monopoly power in the relevant market and (2) the willful acquisition or maintenance of that power as distinguished from growth or development as a consequence of a superior product, business acumen, or historic accident.” Under EU law, that distinction is drawn more sharply. Dominant firms are allowed to enter into “competition on the merits,” but they must abstain from other forms of competition, which are presumably “unmeritorious.” Recently, in Post Danmark, the ECJ summed up the case law as follows:12 It is in no way the purpose of Article [102 TFEU] to prevent an undertaking from acquiring, on its own merits, the dominant position on a market. . . . Competition on the merits may, by definition, lead to the departure from the market or the marginalisation of competitors that are less efficient and so less attractive to consumers from the point of view of, among other things, price, choice, quality or innovation. . . . Article [102 TFEU] applies, in particular, to the conduct of a dominant undertaking that, through recourse to methods different from those governing normal competition on the basis of the performance of commercial operators, has the effect, to the detriment of consumers, of hindering the maintenance of the degree of competition existing in the market or the growth of that competition.

Hence, while both systems thus recognize a category of meritorious competition to which the prohibition should not apply, there is a contrast between the US and EU positions. Leading US authorities are reserved with respect to active enforcement, out of recognition of the practical difficulty in telling anti- from procompetitive conduct. EU law entertains a firm belief in the distinction between competition on the merits/performance competition and other forms of competition, and this distinction is enforced unreservedly.

8.1.2.3.  Protection of the Competitive Process In mainstream economics, the majority opinion is that antitrust and competition law should focus on aggregate welfare, with some authors finding that seeking to maximize consumer welfare is the best way to maximize total welfare in practice (Farrell and Katz, 2006; Martin, 2007). The case law under Article 102 TFEU follows a different path. When dominant firms inflict injury on the competitive process as such, this is already sufficient to trigger the application of Article 102 TFEU, even in the absence of concrete evidence of consumer harm. As the Court of Justice repeatedly stated, “Article 102 TFEU 11  12 

384 U.S. 563 at 570–71 (1966). ECJ, Case C-209/10, Post Danmark, nyr, at para. 21–24.

Continental Drift in the Treatment of Dominant Firms   159

must be interpreted as referring not only to practices which may cause damage to consumers directly, but also to those which are detrimental to them through their impact on competition.”13 In its Guidance Paper, the Commission sought to put more emphasis on the need to show consumer harm (European Commission, 2009) but on that aspect the Guidance Paper has not yet been endorsed by the EU courts. In the eyes of many critics, especially from the United States, EU competition law is then protecting competitors, not competition. In US law, as the Supreme Court put it, “It is axiomatic that the antitrust laws were passed for ‘the protection of competition, not competitors.’ ”14 Instead, as Easterbrook (1986b, 1703) concluded: “However you slice the legislative history, the dominant theme is the protection of consumers from overcharges.” Indeed, the US now has—by all accounts—a strong focus on consumer welfare as the ultimate objective, whereas the EU still protects the competitive process, irrespective of consumer welfare considerations. In perceptive analyses Eleanor Fox (2003) and Gal (2004) suggest that, beyond the rhetoric, the difference between US and EU law relates to the gray zone of conduct that is not clearly output-limiting but nevertheless would injure the competitive process. US antitrust law does not thread into that gray zone, for fear of error, which would lead to prohibiting procompetitive conduct. Under EU competition law, in contrast, authorities do not hesitate to prosecute conduct falling into that gray zone, despite the error risk.

8.1.2.4. Conclusion In this section, we have inventoried four distinctive features of EU competition law, as compared to US law. A first feature appears from the text of Article 102 TFEU itself, namely the possibility of prosecuting exploitative abuses of dominant position. The three other features arise from the interpretation of Article 102. Dominant firms are put under a special responsibility. A strong distinction is drawn between competition on the merits and impermissible forms of competition in which dominant firms might engage. The protection of the competitive process is emphasized, even in the absence of concrete evidence of consumer harm. Our aim is not to justify or rebut these stated differences, but rather to provide possible explanations for them, so as to shed some light on the reasons why EU and US law would differ when it comes to the treatment of dominant firms. We start in section 8.2 by revisiting the traditional explanation for these features, that is, the influence of ordoliberal thought on EU competition law. We find that ordoliberalism provides an explanation for the latter three features, albeit not as strong as is often assumed. In section 8.3, we go on to examine the institutional context of EU law enforcement, finding there

13  Ibid. at para. 20, referring to ECJ, Case C-52/09 TeliaSonera Sverige [2011] ECR I-527 at para. 24 and to earlier cases. 14  Brooke Group v. Browne-Williamson Tobacco, 509 U.S. 209, 224 (1993), referring to a statement first made in Brown Shoe Co. v. United States, 370 U.S.294, 320 (1962). In his recent remarks on the FTC decision not to further pursue Google for alleged monopolization through search bias, Jon Leibowitz was prompt in referring to this quote.

160   Pierre Larouche and Maarten Pieter Schinkel a solid explanation for the special responsibility of dominant firms. In section 8.4, we try to introduce insights from the literature on innovation into the analysis, leading to a stronger justification for the protection of the competitive process and the emphasis on competition on the merits. This leads us to come back to exploitative abuses in section 8.5, where we argue that this is a weak spot in ordoliberalism, which can be remedied with the help of contemporary advances in economics.

8.2.  The Influence of Ordoliberalism on Article 102 TFEU A number of scholars have linked distinctive features of EU competition law to the influence of ordoliberalism (the “Freiburg School”), sometimes in a very critical way (Hawk, 1995; Ahlborn and Grave, 2006; Ahlborn and Padilla, 2008; Ahlborn and Evans, 2009). Accordingly, a first line of explanation for the differences identified above could be found in ordoliberal thinking. It is beyond the scope of this contribution to review ordoliberalism in detail. Nevertheless, since the debate on the legacy of ordoliberalism in EU competition policy (Gerber, 1998; Möschel, 2001) has been revived with the Guidance Paper (European Commission, 2009; Mestmäcker, 2011; Schweitzer, 2008; Marsden and Lovdahl Gormsen, 2010; Lovdahl Gormsen, 2007) and since the original sources are not always accessible,15 it seems appropriate to state our views here, based on our reading of the original sources.

8.2.1.  Ordoliberalism as a General Legal and Economic Theory As a starting point, it must be noted that ordoliberalism is much broader than competition policy. It is a school of economic thought that covers most areas of economic and social policy. Born in the 1930s and becoming influential after the war, ordoliberalism was profoundly influenced by the repeated failures of German economic policy in the first half of the 20th century (Eucken, 2004). Ordoliberals16 were convinced that the market economy and central planning models were so antithetical that they could not be successfully mixed (Böhm, 15  There is no shortage of ordo-liberal scholarship, but most of it is in German. There is disagreement in the German-language literature about whether English-language accounts of ordo-liberalism are accurate: see for instance the remarks of Mestmäcker (2011). 16  For the purposes of this short discussion, the authors generally associated with ordo-liberalism (including W. Eucken and F. Böhm) will be regrouped under the name “ordo-liberals.” As in any school of thought, there are differences among the various authors, but we chose not to mention them unless they are relevant for the purposes of the discussion. We chose to focus on the founding authors, such as Eucken and Böhm, since they represent ordo-liberal thought in its “purest” form. As will be seen infra, once competition laws in Germany and Europe came into force, ordo-liberal thought evolved in the light of legal practice.

Continental Drift in the Treatment of Dominant Firms   161

1980a). Central planning being unsustainable, ordoliberals preferred a market economy. Yet the experience with laissez-faire policies of the turn of the 20th century showed that a market economy left on its own would ultimately produce a concentration of economic power in private hands, with the ensuing nefarious consequences (Eucken, 2001a). In order to solve that conundrum, ordoliberals repudiated the methods of the German historicist school prevailing until then in German economics (Eucken, 2001b). In his macroeconomic textbook of the 1930s, Eucken (1947) advocated a more scientific and theoretical approach to economics. In that sense, ordoliberals were in step with the times, and indeed ordoliberal writings are not insular: they frequently refer to and engage with contemporary Anglo-American authors. Still, the main theorists of ordoliberalism wrote from the 1930s to 1950s, at a time where economics was not widely formalized. Their work follows a deductive, logical analytic method, which is not unlike that of legal scholars. It is not surprising then that lawyers and economists worked very closely within the Freiburg School. Indeed, at its core, ordoliberalism blends law and economics in a liberal project. A liberal polity is characterized by the freedom and autonomy17 its members enjoy, in a constitutional democracy under the rule of law. Such political freedom must be extended to the economic sphere as well; that is, economic actors, producers, and suppliers must enjoy the freedom to make economic choices (Eucken, 2004). This dovetails with an economic policy centered on free markets and competition. Compared to this striking and distinctive link between law and economics, ordoliberal welfare analysis, while not absent, is fairly conventional: competition in a free market economy is best able to effect the coordination of individual efforts (allocative efficiency), and prices play an essential role in conveying information as between economic actors (Eucken, 2004).

8.2.2.  The Threat from Economic Power, Public or Private In the eyes of ordoliberals, economic power forms the main threat to this free economic order based on markets and competition.18 Economic power leads to coercion, which deprives economic actors of their freedom. The ideal to be pursued by economic policy is therefore a free market without economic power, that is, a market where the only possible form of competition is “competition on the merits” (Leistungswettbewerb) (Eucken, 2004). Eucken equates the absence of economic power with “perfect competition” (vollständige Konkurrenz), a very consequential link, as we will see below.19 17 

In the sense of self-determination (Selbstbestimmung). In line with the term used in ordo-liberal writings (wirtschaftliche Macht), we use “economic power.” This term is used by and large with the same meaning as “market power” in contemporary competition law and economics. 19  It has been argued that vollständige Konkurrenz should not necessarily be translated as “perfect competition” (see for instance Möschel, 2001). Yet it is clear from Eucken’s work (1947; 2004) that he was aware of the classical theory of market structure. He renders “perfect competition” (i.e. competition on both supply and demand) as “vollständige Konkurrenz.” 18 

162   Pierre Larouche and Maarten Pieter Schinkel Economic power can be wielded by the state, in centrally planned economies. As stated above, ordoliberals reject this model, and accordingly ordoliberal thought—as one form of liberal thought—assigns a limited role to the State (Eucken, 2004). However, ordoliberals are equally concerned about coercive economic power exerted by private actors (Eucken, 2001a). As Böhm (1980b) points out, firms have an incentive to seek market power, and once they hold it, to use it to distort competition for their benefit. For ordoliberals, free competitive markets are a fragile good, threatened as they are by economic power, be it in public or private hands. As Eucken (2004, 175) puts it, “The problem of economic power has always existed, yet since the Industrial Revolution it has taken on a new dimension. The Industrial Revolution ushered in the era of proliferating economic power . . . : the power of individual firms, of trusts, of cartels, central planning offices, or unions.” For ordoliberals, therefore, protecting the free market economy from economic power held in private hands is crucial. They are very dismissive of the historical experience of the early 20th century, when the state sought to intervene via regulatory and other mechanisms in order to control private economic power:  these attempts only compounded private power with state power, leading towards central planning, with disastrous consequences (Eucken, 2001a). At the same time, the use of state power is the only avenue to confront private economic power. Competition law becomes the vehicle whereby the contradiction is solved, and state power can be wielded against private economic power without endangering the free market economy.

8.2.3.  Competition Law as a Solution to the Problem of Economic Power Competition law is therefore a key element of ordoliberal thought and yet, given the all-encompassing nature of the basic ordoliberal works, it remains summarily dealt with. The key propositions advanced by Eucken (2004) and Böhm (1980b) are the following.



1. Cartels—to the extent other elements of economic policy have not already made them unworkable—must be made illegal as such and harshly prosecuted. The fight against cartels cannot suffice in and of itself to protect the free market economy, since economic power can also arise from a single firm. 2. Competition law must attack economic power at the root and dissolve it; a mere policy of policing abuses of economic power will be ineffective. 3. In some limited cases, monopolies are unavoidable.20 In such cases, monopolies must be placed under the surveillance of the state. The aim of such supervision is to bring monopolies to behave “as if ” they were in a perfectly competitive environment.

20  Eucken (2004) refers to “a small number of cases.” He does not use the term “natural monopolies,” but he might have had them in mind.

Continental Drift in the Treatment of Dominant Firms   163



4. In order to avoid a repeat of historical experience, state control over unavoidable monopolies should be exerted via an independent monopoly supervisory office, and should be carried out strictly in accordance with the law. The applicable law should be crisp and clear, so as to leave no discretion to authorities and avoid capture by private interests (as had occurred in the 1920s and 1930s).

8.2.4.  Ordoliberalism and Article 102 TFEU Seen against that background, Article 102 TFEU cannot be considered a straightforward implementation of ordoliberal thought (Akman, 2012). It does not prohibit economic power as such, nor does it empower the Commission to dismantle firms holding economic power; rather it introduces a system of control on abuses of dominant position, the very system that was found ineffective by Eucken and Böhm. Indeed, it seems that Article 102 TFEU was the result of political bargaining between competing visions (both among and even within member states), including not only ordoliberalism, but also the traditional approach to competition law in Europe until then (however limited the experience was), which emphasized the policing of abuses rather than the direct intervention against cartels and monopoly power (Akman, 2012; Gerber, 1998). So ordoliberal writers had to adjust to a legal reality where private economic power is policed rather than outright prohibited. In order to see how ordoliberals can accommodate that significant departure from their thought, it is interesting to look at the inaugural lecture of Ernst-Joachim Mestmäcker (1959), a disciple of Franz Böhm and a leading figure in the development of EU competition law, both in theory and in practice. This lecture bears on the then new German Competition Act, and its prohibition on abuses of dominant position.21 Mestmäcker expresses its sympathy for US antitrust law (as it then was interpreted), which “gives priority to the dissolution of single monopolistic firms over any attempt to impose upon them obligations to behave properly.”22 In comparison, German law (and EU law) refuses to try to prevent the creation of dominant firms, and instead opts for a prohibition on abuse. As Mestmäcker (1959) notes, compared to the relative simplicity of the ordoliberal prescription against market power, the prohibition on abuse of dominance requires first of all a sophisticated definition of dominance, in order to single out the firms subject to the prohibition, and then a definition of abuse. As regards the latter, Mestmäcker (1959) notes the difficulty of distinguishing abuse from permissible conduct. With the help of ordoliberal thought, that distinction can be articulated along the lines of the distinction between Leistungswettbewerb (performance 21  For the purposes of this chapter, it can be assumed that the arguments made concerning § 22 of the German Competition Act (Gesetz gegen Wettbewerbsbeschränkungen) would be equally applicable to Article 102 TFEU. 22  Of course, the interpretation of Section 2 of the Sherman Act has evolved since then; nonetheless, it is striking to see how the interpretation of Section 2 in force in the 1950s is perceived to be closer to ordo-liberal thought (!) than the new German law.

164   Pierre Larouche and Maarten Pieter Schinkel competition) and Behinderungswettbewerb (hindrance competition).23 From that point on, it is only a small step to hold that, since the dominant firm will abuse its position if it engages in hindrance competition, it is under a duty to refrain from such hindrance competition, hence the special responsibility of the dominant firm.24 Furthermore, it is consistent with ordoliberal thought to hold that the mere presence of market power weakens fragile markets, thereby heightening that special responsibility.25

8.2.5. Conclusion Accordingly, even if Article 102 TFEU does not embody the model of direct intervention against market power advocated by ordoliberals, the distinctive features identified above can nonetheless be traced back to ordoliberal thought. Such is certainly the case for the focus on the protection of the competitive process, the insistence on competition on the merits, and the special responsibility of the dominant firm. If anything, the main legacy of ordoliberalism might play out at a deeper level than even the fundamental features of Article 102 TFEU analysis that we have been examining in the previous sections. Ordoliberalism underlines the fragility of markets, which are vulnerable to both public and private economic power. Underneath most of EU competition law analysis, certainly as far as Article 102 TFEU is concerned, one finds a similar concern for the resilience of competitive markets, or to put it otherwise, a skepticism towards the robustness of markets, a lack of confidence in their ability to self-heal. This basic tenet is rarely expressed,26 but it is almost always present.

8.3.  Decision Theory and the Special Responsibility of the Firm As was seen above, the special responsibility of the dominant firm can be derived from ordoliberal thought, adapted to fit a model of prohibition on abuse of dominance. In this section, we suggest that this special responsibility can be explained equally well, if not 23  Indeed, even though Eucken (2004) advocates the removal of market power pure and simple, he nevertheless discusses at length the various forms that Behinderungswettbewerb can take. Ordo-liberal authors were aware that the distinction between Leistungswettbewerb and Behinderungswettbewerb was liable to be unduly influenced by considerations coming from the unfair trade practices law. 24  As introduced by the ECJ in Michelin, supra, note 10 at para. 57. 25  As the ECJ has held since its judgment of 13 February 1979, Case 85/76, Hoffmann–La Roche v. Commission [1979] ECR 461 at para. 91. 26  Except perhaps in the recurring phrase where the Court states that dominant firms, by their very presence, already weaken the degree of competition in the market (see for instance TeliaSonera, supra, note 13 at para. 27). Interestingly enough, that phrase was not included in the reasoning of the Court in the recent Grand Chamber judgment in Post Danmark, supra, note 12.

Continental Drift in the Treatment of Dominant Firms   165

better, by looking at the institutional context of enforcement in the light of economic analysis, in particular decision theory.

8.3.1.  Decision Theory in the Context of EU Competition Law One of the legacies of the Chicago school has been to throw the spotlight on the institutional dimension of competition policy. The work of Easterbrook (1984), in particular, started from the simple observation that authorities do not and cannot possess all the information required to fully understand and analyze the practices at stake in the cases before them. Acquiring that information, if at all possible, might involve an excessive expense of time and resources, in relation to the expected gain from the application of competition law. Easterbrook sought to solve that conundrum by introducing error-cost analysis: rules should be designed in such a way as to minimize the error costs and the costs of the enforcement system. Furthermore, to the extent that trade-offs are made, it is preferable to err on the side of allowing anticompetitive practices (Type II error) than of prohibiting innocuous ones (Type I error). In the words of Easterbrook (1984), “Other things equal, we should prefer the error of tolerating questionable conduct, which imposes losses over a part of the range of output, to the error of condemning beneficial conduct, which imposes losses over the whole range of output.” Easterbrook’s analysis rests on the assumption that market forces are sufficiently strong to overcome monopoly power, so that Type II mistakes are self-correcting in the long run.27 It is not unreasonable, however, to hold that monopoly power will not always be dissipated. As was seen above, EU competition policy is based on the assumption that markets are not always self-correcting, and hence that competition policy can be needed to address the consequences of monopoly power. In addition, the institutional context remains different in the EU: for the time being, private enforcement is carried out on a much smaller scale than in the United States. Furthermore, most of the decisions are taken by competition authorities acting as investigator, prosecutor, and adjudicator. Hence, rightly or wrongly, the risk of error is perceived as more manageable in the EU. If, for the sake of argument, monopoly power is considered not to be self-correcting, then the cost of Type II errors is likely to be substantial, and the trade-off between Type I and Type II errors is not so simple to assess (Schinkel and Tuinstra, 2006). The competition

27 

The idea being that monopoly profits attract entry (as endorsed by the Supreme Court in Trinko, supra, note 5). Alternatively, if market forces do not suffice to dissolve monopoly power, intervention can always take place at a later point in time, when more information is available on the consequences of the impugned conduct. Easterbrook does not give a temporal dimension to his analysis, and in any event he is generally adverse to intervention. A dynamized Easterbrook-like error-type analysis underpins the reasoning of the General Court on the relationship between MCR review and Article 102 TFEU in Case T-80/02, Tetra Laval [2002] ECR II-4519 (upheld by the ECJ in Case C-13/03P, Tetra Laval [2005] ECR I-1113, but with the ECJ disagreeing with the General Court on that very point) and Case T-201/01, General Electric [2005] ECR II-5575.

166   Pierre Larouche and Maarten Pieter Schinkel authority will want to take a more interventionist stance, while still minimizing error costs, given information deficiencies on its part. In that context, a possible solution would be to try to shift the burden of enforcement over to firms. Since firms possess all of the relevant information about their own position and much of the information about the markets on which they evolve, a dominant firm can certainly assess the compatibility of its conduct with Article 102 TFEU with more accurate information than a competition authority. Self-enforcement of Article 102 TFEU is fraught with risks, however. First, the dominant firm may have no incentive to self-enforce and refuse to do so. Second, the dominant firm might itself commit Type I or Type II errors, for it may itself suffer from information deficiency. Third, the dominant firm might entertain an inaccurate view of the ambit of Article 102 TFEU, either interpreting it too narrowly or too broadly, and thereby not reaching the appropriate conclusion as to the legality of its conduct (assessment error). We will explore these three risks and how they can be remedied in the following three subsections.

8.3.2.  Special Responsibility as a Device to Incentivize Self-Enforcement The special responsibility of the dominant firm, as set out above, could help in addressing the first risk. It would then imply first and foremost that dominant firms are expected to question their own conduct and contribute in enforcing Article 102 TFEU by refraining from conduct that would infringe that provision. Furthermore, that special responsibility justifies, to some extent, the high fines that are then imposed when the law is actually enforced. When seen in that way, the special responsibility of the dominant firm is fully consistent with current institutional choices in EU competition policy. With Regulation 1/2003, the enforcement of Article 101 TFEU has been fundamentally modified to bring it from a system of administrative control, via individual and block exemptions under Article 101(3), towards a self-enforcement system, where Article 101(3) TFEU is directly applicable, without any need for a prior decision. Firms are now largely in charge of assessing their behavior as against Article 101 TFEU. They must conclude for themselves whether any agreement28 they enter into would breach Article 101 TFEU, and act accordingly. To be sure, competition authorities are monitoring compliance with Article 101 TFEU29 and enforcing it; yet despite the decentralization brought about by Regulation 1/2003, it is apparent that public enforcement resources are not sufficient to ensure compliance, certainly not if Article 101 TFEU, much like Article 102, is to have more than a symbolic value.30 Unless and until private

28 

Or any decision of an association or any conduct that could be construed as a concerted practice. Private plaintiffs also play a role, via claims before national courts (in addition to complaints to competition authorities). 30  If the consensus opinion in the EU were that Article 101 and 102 TFEU were by and large marginal provisions for a largely self-healing competitive market, enforcement issues would be less urgent. However, that is not the consensus opinion, which is probably correct in the context of the EU. 29 

Continental Drift in the Treatment of Dominant Firms   167

enforcement can compensate for the lack of resources of public enforcers,31 compliance with Article 101 TFEU therefore depends in large part on firms self-policing their conduct. Under Article 102 TFEU, the ever-recurring reference to special responsibility would therefore signal to dominant firms that they are expected to contribute to enforcing the law, along the same lines as under Article 101 TFEU. They would play a part in the model of administrative control—as opposed to court-centric enforcement, using public or private plaintiffs—that is now prevalent in the EU for the implementation and application of competition policy. In this respect, the special responsibility of dominant firms under Article 102 TFEU might find the most meaning when it is explained in institutional terms. Indeed, on substance, the special responsibility does not usually add much to the analysis. Typically, whether the conduct of the dominant firm is abusive or not does not depend on the special responsibility of the firm. Even though the Commission and the Courts almost never fail to mention it, the special responsibility could usually be removed from the reasoning without loss of cogency. When seen as a signal to dominant firms that they are expected to contribute actively to enforcement by self-assessing their conduct, however, it is sensible to mention that special responsibility time and again, to emphasize that dominant firms are expected to internalize the substantive analysis set out by the competition authority. In fact, viewing the special responsibility of the dominant firm in institutional terms helps to explain two recent cases where such responsibility was more central to the reasoning, namely Deutsche Telekom (price squeeze) and AstraZeneca.32 In both cases, the Commission faulted the dominant firm for its conduct in regulatory processes, invoking its special responsibility under Article 102 TFEU. In Deutsche Telekom, the German telecommunications regulatory authority had approved a set of wholesale and retail tariffs for DT, which put its competitors in a price squeeze. The Commission found that DT could not rely on the regulatory outcome, and that its special responsibility under competition law dictated that it go back to the authority to petition for new tariffs in order to remove the price squeeze. In AstraZeneca, AZ systematically took the most favorable interpretation possible33 of medicines regulation, in order to extend its patents to the maximum. The Commission found, here as well, that AZ’s special responsibility should have led it to take a more mainstream interpretation, to avoid excluding potential competitors. While on substance both cases seem to take the special responsibility of the dominant firm very far, from an institutional perspective the reference to the special responsibility underscores that these two dominant firms were bound to continue to self-assess their conduct, even when regulation came in the picture. 31 

Whether it is desirable that private enforcement takes a larger role will not be discussed here. ECJ, Case C-280/08 P, Deutsche Telekom [2010] ECR I-9555 (confirming GC, Case T-271/03, Deutsche Telekom [2008] ECR II-477) and ECJ, Case C-457/10 P, AstraZeneca (6 December 2012), nyr (confirming GC, Case T-321/05, AstraZeneca [2010] ECR II-2805. 33  Bordering on bad faith: see AstraZeneca (ECJ), ibid. at para. 74–100. 32 

168   Pierre Larouche and Maarten Pieter Schinkel

8.3.3.  Other Risks Relating to Self-Enforcement: Information Deficiencies in the Firm Nevertheless, as set out above, two other risks associated with self-enforcement by firms remain: errors due to information deficiencies by the firm itself and assessment errors when the firm does not correctly understand the applicable law. The risk of information deficiency concerns chiefly information on other firms. Indeed while the dominant firm knows more about itself, and about its own conduct, than the competition authority, it may not possess the same level of information as regards rival firms, customers, suppliers, and so on. It certainly holds enough information about others to be able to make business decisions, but that might not suffice for the purposes of competition law assessment, for which authorities have a number of privileges.34 Under Article 102 TFEU, some features contribute to reducing the risk of information deficiency. For instance, on pricing matters, the European Commission (2009) has put forward the As-Efficient-Competitor (AEC) test, whereby the impact of the prices entertained by a dominant firm is assessed by reference to a competitor with the same cost of production as the dominant firm. The AEC has been endorsed by the ECJ, among other reasons because it noted that in practice it allows the dominant firm to assess its pricing against its own information.35 Beyond that, in nonpricing cases, the effects test currently advocated by the European Commission (2009) could also help in alleviating this risk. In earlier years, a mere potential effect on competition was enough for dominant firm conduct to qualify as abuse. In practice, this meant that since an effect on competition could always be conceived, the form of conduct became determinative. The Commission is now committed to carrying out a real effects analysis, which looks at likely effects. In other words, the Commission endeavors to put together a cogent theory of harm, whereby it is shown that the dominant firm has incentives to enter into a course of conduct, which will likely cause anticompetitive harm, in the view of the likely reactions of other firms. Yet the Commission has also rejected the idea of incorporating an actual effects analysis into Article 102 TFEU, whereby it would need to show that competition has actually been harmed, that is, that competitors have actually been foreclosed or excluded, to the detriment of consumers. Unfortunately, the ECJ has not yet endorsed the introduction of a likely effects analysis (as opposed to mere potential effects) into Article 102 TFEU.36 On the assumption that the view of the Commission will eventually prevail, it could 34 

Certainly if the assessment is to be made up to the legal evidentiary standards of competition law. Deutsche Telekom, supra, note 32 at para. 202, and TeliaSonera, supra, note 13 at para. 44. Unfortunately, the AEC test can also be misinterpreted as an invitation for the dominant firm, in defense, to turn its guns against the plaintiff and shift the focus of the competition law inquiry to the efficiency of the plaintiff, as occurred in Intel [2009] OJ C 227/13. This is not what is meant by the AEC test. 36  See, for instance, ECJ, Case C-202/07 P, France Telecom (Wanadoo) [2009] ECR I-2369 or Deutsche Telekom, ibid. 35 

Continental Drift in the Treatment of Dominant Firms   169

paradoxically help to reduce the information deficiency risk in ex ante self-assessments by dominant firms.37 Indeed, ex ante it is easier for the dominant firm to measure the likely effects of its conduct, based on the information it has in hand, than to try to ascertain what the actual effect thereof will be. In the end, the information deficiency risk remains limited to information about third parties, and it is somewhat alleviated by the AEC test and a likely effects test.

8.3.4.  Other Risks Relating to Self-Enforcement: Assessment Error by the Firm The crux of the matter with self-enforcement, without doubt, is the risk of assessment error by the dominant firm, whereby the firm would not correctly understand what competition law entails. Firms could either fail to refrain from conduct that breaches Article 102 TFEU, in which case presumably competition authorities can intervene. What is more worrisome, however, is the opposite scenario, where a firm refrains from permissible conduct—in all likelihood procompetitive—because it incorrectly assesses that such conduct would breach Article 102 TFEU.38 This fear of dulling the competitive edge of dominant firms underpins much of the US approach to Section 2 of the Sherman Act, certainly as far as writers associated with the Chicago school are concerned. Under Section 2 of the Sherman Act, it is generally admitted that the dividing line between pro- and anticompetitive conduct is thin and sometimes blurry.39 One need only think of price wars versus predatory pricing. Taken ad absurdum, the only way to reduce the risk of assessment error to zero is to refrain from enforcing competition law, that is, adopt an interpretation whereby every course of conduct is deemed permissible. This would not be desirable, certainly not in the EU. Accordingly, assessment errors are unavoidable, and with them the ensuing loss of some competitive benefits from the conduct of dominant firms. It is more a matter of reducing them. In the EU, since reliance on self-assessment by dominant firms tends to increase the risk of assessment errors, one would expect a counterbalancing effort to keep competition law as understandable as possible, so as to limit the scope for divergent interpretations. Unfortunately, current trends in the interpretation of EU competition law, and of Article 102 TFEU in particular, seem to point in the opposite direction. First of all, 37  Even though on substance dominant firms are probably better served by an actual effects test. Such a test is, however, hard to reconcile with the enforcement of Article 102 TFEU in practice: in most cases, it is not desirable to wait until the actual effects of dominant firm conduct have been felt before deciding on whether there is a breach of Article 102 TFEU (and certainly where injunctive relief is at stake). 38  This is not unlike the standard error-risk analysis set out earlier, but this time the source of error is not an information asymmetry as between the authority and the dominant firm or between the dominant firm and other firms, but rather a divergence regarding the interpretation of the law, between what would be entertained by the competition authority, taking into account review by courts, and what the dominant firm uses for its self-assessment. 39  Supra, heading 1.2.2.

170   Pierre Larouche and Maarten Pieter Schinkel more economic analysis was introduced in the application of the law. Undoubtedly, this improves the substantive quality of outcomes. As a result, Article 102 TFEU is now better articulated, at a general, theoretical level, in European Commission (2009)40 with references, for example, to the need for a theory of harm and the emphasis on consumer welfare. In this respect, it might be easier to understand. At the same time, the old formalistic categories have been upended, so that predictability might have been adversely affected overall. Here as well, there is a trade-off between flexibility and certainty.41 By most accounts, recent reforms had a positive overall effect, but the loss of predictability remains. Such loss is compounded, however, by a flight into soft-law instruments (guidelines, notices) and informal conclusions (by way of settlements or commitments), instead of case work. Whereas cases offer competition authorities the opportunity to battle-test their interpretation of the law, soft-law instruments are more in the nature of general policy statements, all the more when they are not based on experience with cases (Schinkel, 2011). They might not provide as much guidance to firms as intended by the issuing authority. It is crucial, if assessment risks are to be kept under control, that competition authorities continue to process cases, so as to test and sharpen their interpretation of the law, to the benefit of dominant firms, when they self-assess their conduct.

8.3.5. Conclusion In the end, this second line of analysis of the continental drift in the treatment of dominant firms, which emphasizes the institutional dimension of competition law, provides a good explanation for the special responsibility of the dominant firm. On the assumption that markets are not self-correcting and therefore that some measure of enforcement of Article 102 TFEU is needed, one can justify, in the EU context, an enforcement approach that emphasizes self-assessment by dominant firms. This matches the new approach to Article 101 TFEU enforcement. Against that background, the special responsibility of the dominant firm signals to dominant firms that they are expected to self-assess their conduct. Nevertheless, self-enforcement also creates error risks on the part of the dominant firm, having to do with information deficiencies or divergences in the interpretation of the law. While the former risk is alleviated by some features of Article 102 TFEU, the latter could be better contained.

40  However, with the lone exception of Post Danmark, supra, note 12, the ECJ has not endorsed the approach set out by the Commission in the Guidance Paper. 41  One should not be misled by comments from practice, which tend to request more flexibility when the law is formalistic and predictable, and then go on to bemoan the lack of legal certainty when the law becomes more responsive to individual circumstances. Here as well, taken ad absurdum, the only way to achieve both flexibility and certainty is to refrain from enforcing the law altogether.

Continental Drift in the Treatment of Dominant Firms   171

The institutional line of analysis does, however, not provide a good explanation for other features of Article 102 TFEU, such as the emphasis on competition on the merits or the protection of the competitive process. Here we need to look further.

8.4.  Innovation, Dynamic Efficiencies, the Competitive Process, and Competition on the Merits The relationship between competition and innovation has been a vexed issue in economic literature. At a theoretical level, Schumpeter (1947) posited a negative relationship between competition and innovation:  as competition increases, the potential rents accruing to the innovator decrease, and so does the incentive to innovate. At the opposite end of the debate, one finds Arrow (1962), who maintained that competition rather pushes firms to innovate, because they stand to gain more from innovation if they are in a competitive environment than if they already enjoy supracompetitive profits in a monopoly environment. Some of the most interesting research lately has come from Aghion and colleagues (2005). On the basis of theoretical and empirical work, they conclude that competition and innovation are not bound in a linear relationship, but rather an inverted-U relationship. By and large, the work of Aghion and colleagues is still being received, and not all authors agree with it (Goettler and Gordon, 2011). In any event, it seems clear that neither Schumpeter nor Arrow is entirely right, and that competition policy must be carefully calibrated (Segal and Whinston, 2007). In US law, by and large, Chicago scholars have been influenced by the Schumpeterian vision of innovation as a gale of creative destruction. In essence, their main concern is that the application of antitrust law must not undermine the ex post rewards to innovators, for fear of undermining the ex ante incentives to innovate (Sidak and Teece, 2009; Evans and Hylton, 2008; Manne and Wright, 2010; Ginsburg and Wright, 2012). When combined with the Chicago school’s typical concern for productive efficiency, this could imply, for instance in a high-tech sector where fixed costs (including R & D) are high and marginal costs low, that competition on the market could be replaced by competition for the market. The US Supreme Court seems to espouse a view of dynamic efficiency that comes very close to that of Chicago school authors. In this view, innovation is seen as a linear, stand-alone process (production function): resources are invested, and innovation results as a function thereof, not unlike investment in infrastructure, for instance (Menell and Scotchmer, 2007; Scotchmer, 2004). Such investment must offer sufficient ex ante perspectives of reward if it is to be undertaken to start with. Since investment is a crucial element of dynamic efficiency, the law must be careful not to undermine

172   Pierre Larouche and Maarten Pieter Schinkel these ex ante incentives. In Trinko, for example, Justice Scalia famously wrote for the Court:42 The mere possession of monopoly power, and the concomitant charging of monopoly prices, is not only not unlawful; it is an important element of the free-market system. The opportunity to charge monopoly prices—at least for a short period—is what attracts “business acumen” in the first place; it induces risk taking that produces innovation and economic growth. To safeguard the incentive to innovate, the possession of monopoly power will not be found unlawful unless it is accompanied by an element of anticompetitive conduct.

There is no comparable statement in the EU. European competition authorities are still building up case law and decision practice in matters where innovation plays a central role, such as the recent Microsoft43 and Intel44 cases. The outcome of these cases already indicates that EU competition law is unlikely to follow the same approach to innovation as US antitrust law. In any event, it could be argued that the US Supreme Court, and the authors who inspired its vision, have a too limitative view of innovation. Innovation cannot just be reduced to an investment.45 Instead, there are two separate elements in innovation, namely the idea or invention and its implementation, that is, its successful introduction to the market. An idea alone does not make an innovation; it must be brought to the market. While investment might help on both counts, it is not the only determinant. This broader view of innovation sits well with the central elements of Article 102 TFEU protection of the competitive process, and competition on the merits.

8.4.1.  Ideas Are Scarce and Unpredictable: Protecting the Competitive Process First of all, there is no given relationship between ideas and investment: ideas are often scarce, and will not necessarily be generated by investment (Scotchmer, 2004). Ideas are unpredictable; they can, and often do, arise outside of well-planned efforts to produce them (Govindarajan and Trimble, 2010). Ex post, of course, a linear, stand-alone narrative can always be put forward, but it would be mistaken to draw the conclusion that this is what must be incentivized ex ante. Otherwise, competition law might be picking winners, at least indirectly, by being biased in favor of a certain innovation narrative,

42 

Trinko, supra, note 5. GC, Case T-201/04, Microsoft [2007] ECR II-3601. 44  Supra, note 35. 45  To some extent, the issues discussed here apply to other forms of investment as well, such as investment in new infrastructure. That point lies outside the scope of this chapter, however. See Larouche (2008). 43 

Continental Drift in the Treatment of Dominant Firms   173

namely that of the large investor. Ex ante, it is more realistic to envision that a number of firms will be vying for the next innovative step. Since it cannot be predicted which idea will successfully be brought to market (and increase welfare), all firms need to be incentivized to innovate.46 While firms are certainly incentivized by the prospect of innovation rents if they succeed, they also need to perceive that, among all firms competing to innovate, they stand a chance of succeeding. When Article 102 TFEU is seen against that background, it is sensible to use that provision to try to protect the competitive process as a value in and of itself. If innovation by its nature cannot be predicted by the authorities, and cannot even reliably be produced by the most skilled and focused firms, the best that competition policy can realistically achieve is to maximize the innovation rate by ensuring that potentially innovative firms deploy their efforts. In contrast with the ordoliberal approach, which seeks to protect the competitive process against economic power, however, when the competitive process is protected for the sake of innovation and dynamic efficiency, it would seem logical to insist that the participants in the competitive process strive for innovation. While competition authorities are hardly in a position to rule on the validity or promise of R & D efforts, there is nonetheless an outer bound: firms that are merely seeking to copy or clone the products, processes, or business methods of existing firms (“copycat firms”) do not and cannot innovate. As such, claims by copycat firms are not deserving, from a dynamic perspective. This view may explain why, in the law on refusal to deal under Article 102 TFEU, for instance, the plaintiff must show that it is seeking to bring a new product on the market, as is exemplified by the Magill,47 IMS,48 and Microsoft49 cases. In contrast to this line of case law, however, in the Deutsche Telekom price squeeze case, the ECJ has enshrined a principle of “equality of opportunity” between firms under Article 102 TFEU, which could be read as a requirement that competitors be able to replicate the business strategy and methods of the dominant firm.50 The ECJ read that principle beyond the narrow 46 

Of course, it is always possible that too many resources would be dedicated to R & D in pursuit of innovation, but we leave that possibility aside for now. 47  ECJ, Case C-241/91 P, RTE [1995] ECR I-743. 48  ECJ, Case C-418/01, IMS [2004] ECR I-5039. 49  Supra, note 43. Microsoft has been widely criticized for setting too low a threshold for novelty, when the Commission and the General Court found that a technical improvement on some features of Windows Server OS qualified as a new product within the meaning of IMS. It can be argued, however, that this holding is sensible as long as it is agreed that incremental innovation—prevalent in the software industry—is worthy of protection as well: see Larouche (2009). 50  Supra, note 32 at para. 230 and ff. In that case, DT argued that the margin squeeze test was invalid, since the Commission compared the wholesale unbundled local loop price with the retail monthly subscription price, leaving out call charges. Since DT had not yet rebalanced its tariffs, it still entertained a low price for monthly subscriptions and high call charges. It was open to competitors, of course, to charge a higher subscription price and lower call charges. Leaving out call charges made the margin squeeze clear. In essence, the ECJ answered that argument with the principle of “equality of opportunity,” which must be read as implying that Article 102 TFEU will also protect the ability of the competitors of the dominant firm to compete with the exact same business model.

174   Pierre Larouche and Maarten Pieter Schinkel confines of previous case law,51 and it is to be hoped that such a principle will not develop further to allow copycats to invoke Article 102 TFEU in their favor.

8.4.2.  Innovation Implies the Ability to Bring the Idea to the Market: Competition on the Merits Second, bringing the idea to the market is also not strictly linked to investment. There are other constraints, besides adverse investment incentives and the ensuing lack of investment, that can also explain why an idea does not unfold into an innovation. They include the inability to gain access to the market (on the supply side), which is discussed below, or rejection by prospective customers (on the demand side). For competition policy, in order to provide appropriate incentives for competing firms to innovate, it is crucial that these firms have the ability to put forward new products and services to potential customers and have customers decide whether these new products and services meet their preferences. Here as well, this implies that, under Article 102 TFEU, competition policy focuses on keeping the competitive process open, including the ability to seek customers. “Competition on the merits” takes a new dimension when seen from a dynamic perspective. Ordoliberals drew a distinction between competition on the merits and other, presumably less meritorious, means of competing.52 From a dynamic perspective, competition on the merits can be seen as the ability to present new products and services to customers and have customers decide whether these products and services, as opposed to what is already on the market or to alternative novelties, find their favor. This is the essence of dynamic competition; if new products and services cannot find their way to the market, there is no innovation. In this context, the merits should include all matters of improvement that are associated with innovation, regarding not only products and processes, but also marketing or business methods. Beyond that, to the extent that network effects play a role in customer decisions, is a firm still competing on the merits when it successfully harnesses network effects to make its products or services gain preeminence on the market? It takes considerable skill to do so, even if skill alone is usually not sufficient. Once network effects have propelled an “inferior” product to supremacy—as has already happened many times— it seems beyond the power of competition authorities to reverse the course of events and restore a “level playing field” where products or services would “compete on their merits.” In the context of Article 102 TFEU, fostering competition on the merits would imply that the law is applied so as to protect the competitive process and keep markets open for 51 

All the cases invoked by the ECJ, ibid., at para. 230 in support of this principle of equality of opportunity arose where the regulatory framework was at stake, and where the dominant operator was put at an advantage in the application of national regulation. 52 See supra, heading 1.2.2.

Continental Drift in the Treatment of Dominant Firms   175

potential innovators. This requires careful balancing. For that purpose, it might be useful to think in terms of “innovation paths,” that is, potential directions in which markets can go as a result of innovation, whereby there can be a number of potential paths, and it is impossible at the time of decision to know which one will eventually materialize. As a starting point, the task of competition policy and competition authorities is not to close innovation paths, that is, not to pick winners. Conversely, one could also argue that their task does not extend to opening new innovation paths; this would be more of a matter for R & D funding or industrial policy. Competition policy is there to keep existing paths open and ensure that the choice of path is made by those whose welfare is central to the whole enterprise, namely customers. Arguably, this reasoning underpins the rejection of Microsoft’s efficiency defense in the media player part of the case, where the General Court held that Microsoft is in fact claiming that the integration of Windows Media Player in Windows and the marketing of Windows in that form alone lead to the de facto standardisation of the Windows Media Player platform, which has beneficial effects on the market. Although, generally, standardisation may effectively present certain advantages, it cannot be allowed to be imposed unilaterally by an undertaking in a dominant position by means of tying.53

At the same time, if the authority prevents a dominant firm from closing down innovation paths and taking choices away from customers, it is not clear what kind of remedy would be appropriate. In Microsoft, Microsoft was ordered to market a version of Windows without the media player included, next to the standard-issue Windows with an integrated media player, for the same price. That remedy, while creative, proved ineffective. In the subsequent case involving the integration of Internet Explorer into Windows,54 the Commission required the introduction of a “browser choice” window into Windows. This might be more effective in ensuring meaningful customer choice, but at the same time it does fixate the browser/operating system articulation, at the expense of more integrative solutions.55 More broadly, in cases such as Microsoft (interoperability information) or Intel, the Commission analysis and remedies aim at bolstering competition and keeping innovation paths open within the current value network of the market in question, to use the terminology of Christensen (1997). The Commission fosters sustaining innovation. Yet in practice the issues identified by the Commission were solved not so much through its competition remedies, but rather by disrupting innovation, which shattered the value network. For instance, the dominance of Microsoft in PC operating systems was rendered less consequential by the rise of browser-centered computing and cloud computing, as well as the shift away from PCs towards smartphones and tablets. That latter shift 53 

Supra, note 43 at para. 1152. Microsoft (Tying) [2010] OJ C 36/7. 55  In other words, it is not totally neutral as far as innovation paths are concerned. 54 

176   Pierre Larouche and Maarten Pieter Schinkel also dampened the effects of Intel’s dominance for x86 CPUs, since smartphones and tablets do not usually run on x86 CPUs. Similarly, there are signs that Google’s market power on the search engine market is being disciplined not so much through innovation in search engines, but rather though innovative methods to circumvent the search engine when linking users to content, for instance through apps on mobile devices or social networks such as Facebook or Twitter.

8.4.3.  Innovation and Interaction between Firms Third, and consequently, innovation cannot always be perceived as a stand-alone, linear process. In many situations, innovation depends on the environment: sometimes it builds on previous innovations (cumulative innovation: Scotchmer, 2004); at other times it arises out of complex interactions between a firm and its customers (user-driven innovation:  von Hippel, 2005)  or among firms and other entities (open innovation: Chesbrough, 2003). These are the hard cases, where interaction between firms is part of the innovation process. Unavoidably, firms will disagree with one another in the course of such interactions, and one firm might try to gain an advantage over the others by affecting these interactions. In the context of Article 102 TFEU, this led to decisions such as Magill, IMS and Microsoft, for instance, where a competitor requires access to some facilities, information, or intellectual property of the dominant firm, alleging that this is necessary to enable it to innovate. The dominant firm claims that it is equally intent upon innovating, and that it should not be required to share with, or supply, its competitor. Assuming an inverted-U relationship as in Aghion and colleagues (2005), settling this claim could hinge upon whether the sector under inquiry finds itself in the ascending or descending part of inverted-U curve. In the former case, presumably increasing competition by imposing access will have a positive effect on innovation (and increase welfare), while in the latter case, it would not. Of course, finding out where a sector is on the inverted-U curve is no easy feat. In addition, if access to facilities or information is necessary to be able to compete at all, the claimant has in fact no or almost no prospect of any innovation rent, irrespective of the level of competition on the market. Under these circumstances, the assumptions underpinning the model of Aghion and colleagues are not met. In practice, when faced with this issue in Microsoft, the Commission let it slip in its decision that “on balance, the possible negative impact of an order to supply on Microsoft’s incentives to innovate is outweighed by its positive impact on the level of innovation of the whole industry (including Microsoft).”56 The General Court disavowed this statement,57 and indeed this balancing test is too rough to be accurate. In the 56 

Supra, note 43 at para. 783. The General Court noted, however, that the Commission had provided ample other reasons why Microsoft’s argument on innovation incentives should be rejected: ibid., para. 705–10. 57 

Continental Drift in the Treatment of Dominant Firms   177

end, by ordering Microsoft to provide access, the Commission in fact reduced the cost of making a technological jump (even if the jump is also smaller, i.e. more incremental than breakthrough innovation) for Microsoft’s competitors. Yet there are not so many competitors. If the effect of the decision is to turn the market for server operating systems into a neck-and-neck market, with a few competitors being at the same technological level, then in theory the market could become “sleepy” unless still more competition is stimulated; it is not clear how a competition authority could achieve that. In particular, from the moment that a more interventionist stance is chosen when it comes to innovation and competition policy,58 competition authorities will find it very difficult not to affect the course of innovation, by preempting certain paths or even picking winners. In this respect, the first signals emanating from EU competition policy are not very encouraging. For instance, in cases such as Microsoft or Intel, the Commission could not manage to set out cogent reasoning without making a negative judgment on the quality of the research efforts and the products of the dominant players, Microsoft and Intel. Similarly, in Microsoft, the Commission decided that the level of royalties for the disclosed information should not reflect its strategic value, that is, the use to which it can be put (obtaining or maintaining a presence on the server OS market).59 That principle was further developed in the course of implementing the Commission decision, as explained in a subsequent decision where the Commission fixed the periodic penalty upon Microsoft for having failed to implement the disclosure remedy on time.60 In the end, the rate of royalties was set to reflect the “intrinsic” value of the intellectual property, which was a function of whether the property was innovative. Innovativeness was assessed by reference to criteria taken from patent law, that is, novelty and nonobviousness.61 The Commission therefore put itself in the position of a patent office, ruling on the value of a firm’s efforts, and introduced the notion that intellectual property has an “intrinsic” value that a competition authority can figure out.62

8.4.4. Conclusion In short, once innovation and dynamic effects are taken seriously and integrated into competition analysis, central elements of Article 102 TFEU—the protection of the competitive process as such, competition on the merits—are infused with new meaning and

58  A competition policy that is geared simply to ensure that innovators can look forward to innovation rents, as in the United States, will typically not result in much intervention. 59  Decision of 24 March 2004 [2007] OJ L 32/23, at para. 1008 (ii). This is to avoid the possibility that the royalties received as a result of the compulsory disclosure would put Microsoft in the same position as if it had agreed voluntarily on disclosure. 60  Microsoft (Fixation of Penalty Payments) [2009] OJ C 166/20, confirmed by the General Court, Case T-167/08, Microsoft (27 June 2012), nyr. 61  Ibid. at para. 138–51. 62  It will be interesting to see if the same approach will be followed to solve the cases concerning the pricing of Standard-Essential Patents (SEPs) now piling up before the Commission.

178   Pierre Larouche and Maarten Pieter Schinkel significance. Nevertheless, as shown above, many fundamental issues are still open, and we are still far from a solid approach to innovation under Article 102 TFEU.

8.5.  Excessive Pricing Revisited In the previous sections, we explored alternative explanations for the three distinctive features identified as regards the implementation of Article 102 TFEU: the special responsibility of the dominant firm, the protection of the competitive process, and the focus on competition on the merits. The European “outlier” prohibition of exploitative abuses, in particular in the form of excessive pricing, still remains to be explained. As mentioned in section 8.1, this category has been little enforced. In this section, we revisit it and suggest how it could be brought to life by recognizing the contradictions limiting the usefulness of ordoliberal thought for this category, overcoming practical objections, and applying more recent economic methods from the calculation of antitrust damages.

8.5.1.  Ordoliberal Contradictions and Excessive Pricing The treatment of exploitative abuses highlights a contradiction in ordoliberal thought. As was seen under section 8.2, ordoliberals advocated outright intervention against economic power as such.63 Nevertheless, Eucken and others believed that some “unavoidable monopolies” might remain. The starting point of the ordoliberals was that this—presumably very limited—class of monopolists is to be regulated by a monopoly surveillance authority, with a view to making them behave “as if ” they were evolving in a competitive setting (Eucken, 2004; Böhm, 1980b). This also implies that price-related conduct is controlled so that regulated prices approximate those that would materialize under perfect competition. Therein lies a contradiction that Eucken struggled with: while the freedom to set prices is a central tenet, “as if ” regulation substitutes regulated prices for freely set prices. Regulated prices can hardly fulfill the signaling function of market prices. Eucken’s way of minimizing the contradiction was to restrict the “as if ” regulation to that narrow set of unavoidable monopolies, all other manifestations of economic power having been uprooted. Other ordoliberal writers—in particular Leonhard Miksch (1948–49), who invented the term “as if ”—were less hesitant about regulatory intervention. As a starting point, Miksch considers that, while perfect competition might be the theoretical ideal, it is rarely achieved in practice. The task of the state is then to intervene in order to bring about the state of affairs that would have resulted if perfect competition had prevailed: it is a matter of “finding a price that, first, does not depend on the will of the [dominant 63 

Supra, heading 2.4.

Continental Drift in the Treatment of Dominant Firms   179

firm]; second, leads to rational production because of its relationship to production costs, and, third, evolves in parallel to free market prices” (Miksch, 1948–49, 333). Once a system of abuse control was enshrined in the EEC Treaty in 1957, it was clear that a much larger number of dominant firms would be present than the original ordoliberal ideas would have countenanced.64 Accordingly, the more interventionist approach advocated by Miksch took hold, without necessarily having been proven superior. Indeed, in essence the range of the “as if ” principle was extended to all dominant firms, irrespective of whether their position is “unavoidable” or not. Article 102 TFEU was applied to dominant firms on the assumption that the behavior of these firms should be analyzed, and remedies imposed on them, “as if ” they would evolve in a competitive environment. In the realm of pricing abuses, price-cost tests, designed to assess the pricing of the dominant firm as against cost-based “as if ” prices, became a fixation, as can be seen to this day in the application of Article 102 TFEU to predatory pricing, rebates, or price squeezes.65

8.5.2.  Practical Objections against Enforcing a Prohibition on Excessive Pricing However popular Miksch’s “as if ” approach has proven as regards exclusionary abuses, the situation is different with exploitative abuses. In principle, under this approach, competition authorities should seek to bring the dominant firm to a price level corresponding to what it would entertain if it were active on a competitive market. But in practice, this has proven very difficult. According to the ECJ, an excessive price within the meaning of Article 102 TFEU bears “no reasonable relation to the economic value of the product.”66 That value is established by comparing the price with the production costs (cost-plus test), and then by determining whether the price is unfair by itself or in comparison with competing products (comparator test).67 The cost-plus test is directly inherited from ordoliberal thought; neither test is defined with any greater precision in the ECJ case law. The Commission has a mixed track record as regards excessive pricing under Article 102 TFEU, having lost two early cases before the ECJ.68 Subsequently, the Commission brought few excessive pricing cases. Following a review of the decision 64 

Supra, heading 2.4. Even if Post Danmark, supra, note 12 might mark some progress in this respect, towards a general test applicable to pricing practices, it still remains a price-cost test. The ECJ refused to move the analysis of predatory pricing away from costs, and towards a focus on recoupment, in Wanadoo, supra, note 36. Similarly, in Case C-549/10 P, Tomra (19 April 2012), nyr, a judgment issued after Post Danmark, the ECJ reiterates its classic position on rebates, which are to be assessed by looking at cost savings against an ideal cost-based list price. Furthermore, in Microsoft, supra, note 60 the General Court approves an extension of this approach to the pricing of IP royalties, as advocated by the Commission. 66  United Brands, supra, note 3 at para. 250. 67  Ibid. at 251–52. 68  ECJ, Case 26/75, General Motors [1976] ECR 1367 and United Brands, ibid. 65 

180   Pierre Larouche and Maarten Pieter Schinkel practice of the Commission and of the British, Dutch, and German competition authorities, Geradin (2007) concluded that excessive pricing cases are few and often unsuccessful. The Commission and national authorities tend not to prioritize these cases, for a number of reasons: a belief—in line with US law—that excessive prices are self-correcting through market entry, methodological difficulties in ascertaining whether prices are excessive, and a reluctance to enter into what is perceived as a price regulation exercise.69 In recent times, a number of authors have sought to improve upon the United Brands test (Motta and de Streel, 2006; Evans and Padilla, 2005; Röller, 2008; Paulis, 2008; Geradin, 2007; Akman and Garrod, 2011). While their proposals differ on some points, they all insist upon the need to restrict enforcement to markets evidencing high barriers to entry; in that sense, these proposals attempt to limit the application of Article 102 TFEU rather than modify or replace the substantive test for excessive pricing.

8.5.3.  Applying Modern Economics to Determine “but for” Prices in Excessive Pricing Cases The price-cost test approach that stems from the ordoliberal fixation on natural monopolies versus perfect competition is, however, only one of several methods developed in modern economics to determine so-called “but for” prices, in particular in cartel damage litigation. Comparator-based methods are much more prominent in this area, including time-series analysis of the same market before and/or after the infringement and comparisons to a different but similar geographic market, and to different but similar products (van Dijk and Verboven, 2008; Davis and Garces, 2009, Korenblit, 2012). In these methods, the key assumption is that the period or market to which a comparison is made can serve as a reasonable model for what competitive conditions in the market under investigation could have been, absent the anticompetitive behavior. The before-and-after method involves sophisticated econometrics in estimating, from actually observed prices and relevant additional explanatory variables, the “but for” price development that would have occurred between endogenously determined structural breaks around the beginning and the end of the infringement period.70 Yardstick methods likewise can include a number of variables to control for any differences that might be present between the markets compared, besides the antitrust violation. An alternative approach to damages assessment uses structural industrial organization models for simulation analysis with calibrated parameters, which can endogenize the determination of costs. 69  A recent example of a regulation-like intervention is Standard & Poor’s. See European Commission, Antitrust: Commission Makes Standard & Poor’s Commitments Abolish Fees for Use of US International Securities Identification Numbers Binding, IP/11/1354, 15/11/2011. 70  A prominent example is B. D. Bernheim, Expert Report in RE: Vitamins Litigation, M.D.L. No. 1285, United States District Court for the District of Columbia, May 24, 2002.

Continental Drift in the Treatment of Dominant Firms   181

While primarily developed in applications to cartel cases, these economic methods can, in principle, be modified rather straightforwardly for use in assessing the effects of abuses of dominance as well. The European Commission (2011) has so far always maintained that damages claims were also available against exploitative abuses under Article 102 TFEU. Recently, the OECD (2011) organized a roundtable about the topic, on the observation that competition authorities are often hesitant to enforce, while direct exploitative abuses can cause extensive harm to consumers. In the lengthy background paper by the Secretariat, only slightly more than one page, Section 6.3, discusses the use of price comparisons to determine to what extent prices may be excessive. The main author of the OECD paper, Maier-Rigaud (Maier-Rigaud and Schwalbe, 2013), also briefly mention application to exploitative abuse cases in a recent survey paper on the quantification of antitrust damages. These more contemporary methods, while developed primarily to deal with private damage claims for breaches of Article 101 TFEU, can very well be modified and applied to assess whether, and by how much, dominant firm prices are excessive. As the European Commission (2011) itself notes, damage claims for excessive pricing in breach of Article 102 TFEU do not differ significantly from damage claims for overcharging in connection with a cartel, in breach of Article 101 TFEU. Nothing seems to stand in the way of the Commission itself applying these economic methods in Article 102 cases. Doing so may well revive public enforcement in this stale European outlier category. Indeed, as appears from the previous section, much of the unease with excessive pricing under Article 102 TFEU stems from the practical difficulty of applying what remains a fairly vague price-cost test, and from the more theoretical unease at the perspective of having to force the dominant firm to bring its prices down to the “right” cost-based price, if and once it is identified. The key to staying within ordoliberal principles is to understand that the “but for” comparator that authorities are ascertaining in excessive pricing cases is not perfect competition, but any form of imperfect competition that would have prevailed without the abuse of dominance. In this sense, excessive prices are conceptually the same as cartel overcharges, with one important complicating factor. Whereas in cartel cases there typically are periods or geographically different but otherwise comparable markets, where there was no collusion, so that before-and-after or yardstick-methods can be applied, this may not always be the case in situations of dominance that are caused by structural characteristics, such as network effects (OECD, 2012). Yet it is exactly here that the difference appears between natural monopolies and “garden variety” dominance cases:  the latter, being the results of exclusionary practices, would potentially leave comparable competitive conditions to use as a basis for a “but for” estimation. In contrast, “natural monopoly” markets typically have structural market characteristics, in particular large fixed costs in relation to the size of the market, such that optimal prices, even in the absence of information problems, cannot be found with the simple rule of thumb from perfect competition, that is, price equals marginal costs (Viscusi, Vernon, and Harrington, 2005). Yet despite these complexities in determining optimal price regulation of natural monopolies, this is a large and very active area of interference with actual markets, mostly via sector-specific regulation.

182   Pierre Larouche and Maarten Pieter Schinkel Moving away from a reliance on the hypothetical “right” price under perfect competition would overcome a key weakness in ordoliberal thought. Indeed the starting normative aim for Eucken, Böhm, and others is the absence of economic power on markets; that is a negatively formulated aim, which can be achieved by removing economic power, without encroaching on price mechanisms. The “as if ” approach goes beyond that negative aim; it rests on a positive norm, namely perfect competition. In the writings of Eucken (2004) and Böhm (1980b), while the negative aim (absence of market power) is solidly argued, the link with the positive aim (perfect competition) is less clearly made. Not all authors agree with Eucken, indeed: later authors such as Möschel (2001) downplay the link between ordoliberalism and perfect competition. Furthermore, since these methods would be applied to historical data, with the aim being to establish to what extent prices have been excessive in comparison to a relevant comparator, modeling a public enforcement practice in excessive pricing on them would stay a distance away from regulation. After all, damages claims are aimed at obtaining compensation for past overcharges, rather than regulating future conduct. In a finding of an infringement of Article 102 by an exploitative abuse of dominance, assessment of the “but for” price could, but would not need to, be the base of the fine. It suffices as a foundation for the finding of an infringement. Private damages cases for the breach of 102 may follow and, based on the same principal methods, add to the liability of the infringing undertaking. In either case, the only effect going forward would be through the internalization of the possibility of future conclusions on excessive prices when a case might be brought, which will discipline undertakings by way of a deterrence effect, within the realm of the undertaking’s special responsibility. The methods outlined above—save for cost-based methods—offer an alternative to the price-cost test. With these methods, the aim of the inquiry is not to identify the positively “right” price, but rather to tease out, from available data, the effect of dominance on prices. This corresponds to the negative normative aim of ordoliberalism, that is, the absence of economic power. The effects of dominance are isolated, and to the extent that they constitute excessive pricing, a damage claim will be entertained.

8.6. Conclusions Despite visible divergences in their wording, Section 2 of the Sherman Act and Article 102 TFEU, like two tectonic plates, drifted away from the interpretation commonly attributed to them at the outset, in order to converge in substance on the policing of exclusionary conduct by dominant firms. Yet significant differences remain in their application in individual cases. These differences must be attributable to basic features of the analysis carried out under each provision. We have inventoried three such distinctive features of EU competition law, namely: the special responsibility of the dominant firm; the strong distinction between competition on the merits and impermissible

Continental Drift in the Treatment of Dominant Firms   183

forms of competition; and the protection of the competitive process even in the absence of proven consumer harm. A fourth feature appears from the text of Article 102 TFEU itself, namely the possibility of prosecuting exploitative abuses of dominant position. In this chapter, we sought to provide explanations for these distinctive features. A first line of explanation, traditionally put forward, is the influence of ordoliberal thought (the Freiburg school) on Article 102 TFEU. The focus on the protection of the competitive process, the insistence on competition on the merits, and the special responsibility of the dominant firm can all be traced back to ordoliberalism. Yet Article 102 TFEU does not correspond to the model of direct intervention against market power advocated by ordoliberals. It merely polices abuses, as opposed to market power as such. Considering that the main works of ordoliberalism date from the mid-20th century, it should come as no surprise that the approach suffers serious shortcomings. Beyond the sometimes exaggerated criticism leveled in the literature, it can be argued that, for instance, ordoliberals were too quick to jump from a negative normative ideal of absence of market power, to a positive ideal of perfect competition. This implied that there was a “right” price that dominant firms should charge, and accordingly the application of Article 102 TFEU became rife with price-cost tests. In the case of exploitative abuses, and especially of excessive pricing, this led to a marked reluctance on the part of competition authorities to entertain cases. Yet when re-examined in the context of private enforcement, with the help of newer economic tools to ascertain price overcharges, it no longer seems so odd that dominant firms would have to pay damages for excessive prices. In light of these gaps in ordoliberal thought, it is worth looking at alternative lines of explanation. A second line of explanation emphasizes the institutional dimension of competition law, and it provides a good explanation for the special responsibility of the dominant firm. The special responsibility of the dominant firm signals to dominant firms that they are expected to self-assess their conduct. This matches the new approach to Article 101 TFEU enforcement. Nevertheless, self-enforcement also creates error risks on the part of the dominant firm, having to do with information deficiencies or divergences in the interpretation of the law. While the former risk is alleviated by some recent developments in the interpretation of Article 102 TFEU, the latter could be better contained. In order to explain the emphasis on competition on the merits or the protection of the competitive process, however, a third line of explanation is needed. It requires that innovation considerations be brought into Article 102 TFEU analysis. Innovation must be seen as more than a simple production function, depending on investment, but rather as a combination of novel ideas and successful implementation. Against that background, Article 102 TFEU can be seen as an instrument to ensure that markets remain open for all potential innovators, hence the need for enforcement to focus on keeping the competitive process open. Similarly, competition on the merits takes on a new meaning, since it is in the essence of dynamic competition that firms seek to compete on the strength of innovative products that they bring to the market. Nevertheless, as the chapter shows, many fundamental issues are still open, and we are still far from a solid approach to innovation under Article 102 TFEU.

184   Pierre Larouche and Maarten Pieter Schinkel The following table sums up our analysis:

Special responsibility of the dominant firm Competition on the merits Protection of the competitive process

Ordoliberalism

Decision theory

X X X

X

Innovation X X

Accordingly, the features identified above can be explained otherwise than strictly through ordoliberal thought. Still, this does not imply that ordoliberalism is superfluous to the understanding of Article 102 TFEU. At an even deeper level, the interpretation and enforcement of Article 102 TFEU—and to some of extent EU competition law as a whole—assumes that markets are fragile, that they are vulnerable to both public and private economic power. This is the starting point of ordo-liberal thought, and it stands in contrast with the more confident attitude displayed throughout US antitrust law nowadays. EU competition authorities and courts are skeptical of the robustness of markets and their ability to self-heal. At this juncture, there is no conclusive evidence to support or disconfirm either the US or the EU basic assumption, and maybe there never will be.

Acknowledgments The authors wish to thank (in alphabetical order) Pinar Akman, Roger Blair, Victoria Daskalova, David Gerber, Imelda Maher, Matteo Negrinotti, Heike Schweitzer and Daniel Sokol for their useful comments at various stages in the preparation of this chapter, as well as the participants in seminars held in Tilburg (TILEC) and the University of East Anglia (CCP), and in the Loyola Antitrust Colloquium, for their comments.

References Aghion, P., N. Bloom, R. Blundell, R. Griffith, and P. Howitt. 2005. Competition and Innovation: An Inverted-U Relationship. Quarterly Journal of Economics 120: 701. Ahlborn, C., and D. S. Evans. 2009. The Microsoft Judgment and its Implications for Competition Policy towards Dominant Firms in Europe. Antitrust Law Journal 75: 887. Ahlborn, C., and C. Grave. 2006. Walter Eucken and Ordoliberalism: An Introduction from a Consumer Welfare Perspective. Competition Policy International 2: 197. Ahlborn, C., and A. J. Padilla. 2008. From Fairness to Welfare: Implications for the Assessment of Unilateral Conduct under EC Competition Law. In C. D. Ehlermann and M. Marquis, European Competition Law Annual 2007: A Reformed Approach to Article 82 EC. Oxford: Hart. Akman, P. 2012. The Concept of Abuse in EU Competition Law. Oxford: Hart.

Continental Drift in the Treatment of Dominant Firms   185

Akman, P., and L. Garrod. 2011. When Are Excessive Prices Unfair? Journal of Competition Law and Economics 7: 403. Areeda, P., and H. Hovenkamp. 2013. Antitrust Law: An Analysis of Antitrust Principles and Their Application, 4th ed. New York: Aspen. Arrow, K. J. 1962. Economic Welfare and the Allocation of Resources for Invention. In R. Nelson, The Rate and Direction of Inventive Activities:  Economic and Social Factors. Princeton, NJ: Princeton University Press. Arts, G., W. Dicke, and L. Hancher. 2008. New Perspectives on Investment in Infrastructures. WRR Studies. Amsterdam: Amsterdam University Press. Ashton, D., and D. Henry. 2013. Competition Damages Actions in the EU: Law and Practice. Cheltenham: Edward Elgar. Böhm, F. 1980a. Die Idee des Ordo im Denken Walter Euckens. In E. J. Mestmäcker, Freiheit und Ordnung in der Marktwirtschaft. Baden-Baden: Nomos. Böhm, F. 1980b. Wirtschaftsordnung und Staatsverfassung. In E. J. Mestmäcker, Freiheit und Ordnung in der Marktwirtschaft. Baden-Baden: Nomos. Chesbrough, H. W. 2003. Open Innovation: The New Imperative for Creating and Profiting from Technology. Boston: Harvard Business School Press. Christensen, C. M. 1997. The Innovator’s Dilemma. Boston: Harvard Business School Press. Collins, W. D. 2008. Issues in Competition Law and Policy. Chicago: ABA Section of Antitrust Law. Cseres, K., M. P. Schinkel, and F. O. W. Vogelaar. 2006. Criminalization of Competition Law Enforcement: Economic and Legal Implications for the EU Member States. London: Edward Elgar. Davis, P., and E. Garces. 2009. Quantitative Techniques for Competition and Antitrust Analysis. Princeton, NJ: Princeton University Press. Easterbrook, F. H. 1984. The Limits of Antitrust. Texas Law Review 63: 1. Easterbrook, F. H. 1986a. On Identifying Exclusionary Conduct. Notre Dame Law Review 61: 972. Easterbrook, F. H. 1986b. Workable Antitrust Policy. Michigan Law Review 84: 1696. Easterbrook, F. H. 2003. When Is It Worthwhile to Use Courts to Search for Exclusionary Conduct? Columbia Business Law Review: 345. Ehlermann, C.-D., and I Anastasiu. 2006. European Competition Law Annual 2003: What Is an Abuse of a Dominant Position? Oxford: Hart. Ehlermann, C. D., and M. Marquis. 2008. European Competition Law Annual 2007: A Reformed Approach to Article 82 EC. Oxford: Hart. Elhauge, E., and D. Geradin. 2011. Global Antitrust Law and Economics. 2nd ed. New York: Foundation Press. Eucken, W. 1947. Die Grundlagen der Nationalökonomie. 5th ed. Godesberg: Küpper. Eucken, W. 2001a. Das Problem der wirtschaftlichen Macht. In W. Eucken, Wirtschaftsmacht und Wirtschaftsordnung. London Lectures on Economic Policy. Münster: LIT. Eucken, W. 2001b. Die Entwicklung des ökonomischen Denkens. In W. Eucken, Wirtschaftsmacht und Wirtschaftsordnung. London Lectures on Economic Policy. Münster: LIT. Eucken, W. 2001c. Wirtschaftsmacht und Wirtschaftsordnung. London Lectures on Economic Policy. Münster: LIT. Eucken, W. 2004. Grundsätze der Wirtschaftspolitik. 7th ed. Tübingen: Mohr Siebeck. European Commission. 2005. DG Competition Discussion Paper on the Application of Article 82 of the Treaty to Exclusionary Abuses. European Commission. 2009. Guidance on the Commission’s Enforcement Priorities in Applying Article 82 of the EC Treaty to Abusive Exclusionary Conduct by Dominant Undertakings. OJ C 45/7.

186   Pierre Larouche and Maarten Pieter Schinkel European Commission. 2011. Draft Guidance Paper on Quantifying Harm in Actions for Damages Based on Breaches of Article 101 or 102 TFEU. Evans, D. S., and K. Hylton. 2008. The Lawful Acquisition and Exercise of Monopoly Power and Its Implications for the Objectives of Antitrust. Competition Policy International 4: 203. Evans, D. S., and J. Padilla. 2005. Excessive Prices: Using Economics to Define Administrable Legal Rules. Journal of Competition Law and Economics 1: 97. Farrell, J., and M. L. Katz. 2006. The Economics of Welfare Standards in Antitrust. Competition Policy International 2. Fox, E. M. 2003. We Protect Competition, You Protect Competitors. World Competition 26: 149. Gal, M. S. 2004. Monopoly Pricing as an Antitrust Offense in the U.S. and the E.C.: Two Systems of Belief about Monopoly? Antitrust Bulletin 49: 343. Geradin, D. 2007. The Necessary Limits to the Control of “Excessive” Prices by Competition Authorities: A View from Europe. Tilburg University Legal Studies Working Paper. Available on SSRN. Gerber, D. 1998. Law and Competition in Twentieth Century Europe. Oxford: Oxford University Press. Ginsburg, D. H., and J. D. Wright. 2012. Dynamic Analysis and the Limits of Antitrust Institutions. Antitrust Law Journal 78: 1. Goettler, R. L., and B. R. Gordon. 2011. Does AMD Spur Intel to Innovate More? Journal of Political Economy 119: 1141. Govindarajan, V., and C. Trimble. 2010. The Other Side of Innovation. Boston: Harvard Business School Press. Hawk, B. 1995. System Failure: Vertical Restraints and EC Competition Law. Common Market Law Review 32: 973. Joliet, R. 1970. Monopolization and Abuse of Dominant Position. The Hague: Martinus Nijhoff. Korenblit, C.  M. 2012. Quantifying Antitrust Damages-Convergence of Methods Recognized by U.S. Courts and the European Commission. Competition Policy International Antitrust Chronicle 3(1), Spring. Larouche, P. 2008. Europe and Investment in Infrastructure, with Emphasis on Electronic Communications. In G. Arts, W. Dicke, and L. Hancher, New Perspectives on Investment in Infrastructures. WRR Studies. Amsterdam: Amsterdam University Press. Larouche, P. 2009. The European Microsoft Case at the Crossroads of Competition Policy and Innovation. Antitrust Law Journal 75: 933. Lovdahl Gormsen, L. 2007. The Conflict between Economic Freedom and Consumer Welfare in the Modernization of Article 82 EC. European Competition Journal 3: 329. Maier-Rigaud, F. 2012. Excessive Prices. In OECD, OECD Best Practice Roundtables in Competition Policy. Maier-Rigaud, F., and U. Schwalbe. 2013. Quantification of Antitrust Damages. In D. Ashton and D. Henry, Competition Damages Actions in the EU: Law and Practice. Cheltenham: Edward Elgar. Manne, G., and J. D. Wright. 2010. Innovation and the Limits of Antitrust. Journal of Comparative Law and Economics 6: 153. Marsden, P., and L. Lovdahl Gormsen. 2010. Guidance on Abuse in Europe:  The Continued Concern For rivalry and a Competitive Structure. Antitrust Bulletin 55: 875. Martin, S. 2007. The Goals of Antitrust and Competition Policy. Purdue CIBER Working Papers. Menell, P. S., and S. Scotchmer. 2007. Intellectual Property Law. In M. Polinsky and S. Shavell, Handbook of Law and Economics. Amsterdam: North Holland.

Continental Drift in the Treatment of Dominant Firms   187

Mestmäcker, E.  J. 1959. Das marktbeherrschende Unternehmen im Recht der Wettbewerbsbeschränkungen. Tübingen: Mohr-Siebeck. Mestmäcker, E. J. 1980. Freiheit und Ordnung in der Marktwirtschaft. Baden-Baden: Nomos. Mestmäcker, E.  J. 2011. The Development of German and European Competition Law with Special Reference to the EU Commission’s Article 82 Guidance of 2008. In L. F.  Pace, European Competition Law:  The Impact of the Commission’s Guidance on Article 102. Cheltenham: Edward Elgar. Miksch, L. 1948–49. Die Wirtschaftspolitik des Als-Ob. Zeitschrift für die gesamte Staatswissenschaft 105: 311. Möschel, W. 2001. The Proper Scope of Government Viewed from an Ordoliberal Perspective: The Example of Competition Policy. Journal of Institutional and Theoretical Economics 157: 3. Motta, M., and A. de Streel. 2006. Exploitative and Exclusionary Prices in EU Law. In C.-D. Ehlermann and I Anastasiu, European Competition Law Annual 2003: What Is an Abuse of a Dominant Position? Oxford: Hart. Nelson, R. 1962. The Rate and Direction of Inventive Activities: Economic and Social Factors. Princeton, NJ: Princeton University Press. OECD. 2011. Policy Roundtable, Excessive Prices. Paris. OECD. 2012. Excessive Prices. OECD Best Practice Roundtables in Competition Policy. Pace, L. F. 2011. European Competition Law: The Impact of the Commission’s Guidance on Article 102. Cheltenham: Edward Elgar. Paulis, E. 2008. Article 82 EC and Exploitative Conduct. In C. D. Ehlermann and M. Marquis, European Competition Law Annual 2007: A Reformed Approach to Article 82 EC. Oxford: Hart. Polinsky, M., and S. Shavell. 2007. Handbook of Law and Economics. Amsterdam: North Holland. Röller, L.  H. 2008. Exploitative Abuse. In C. D.  Ehlermann and M. Marquis, European Competition Law Annual 2007: A Reformed Approach to Article 82 EC. Oxford: Hart. Schinkel, M. P. 2011. Bargaining in the Shadow of the European Settlement Procedure for Cartels. Antitrust Bulletin 56: 461. Schinkel, M. P., and J. Tuinstra. 2006. Imperfect Competition Law Enforcement. International Journal of Industrial Organization 24: 1267. Schumpeter, J. 1947. Capitalism, Socialism and Democracy. New York: Harper & Row. Schweitzer, H. 2008. The History, Interpretation and Underlying Principles of Section 2 Sherman Act and Article 82 EC. In C. D. Ehlermann and M. Marquis, European Competition Law Annual 2007: A Reformed Approach to Article 82 EC. Oxford: Hart. Scotchmer, S. 2004. Innovation and Incentives. Cambridge, MA: MIT Press. Segal, I., and M. Whinston. 2007. Antitrust in Innovative Industries. American Economic Review 97: 1703. Sidak, J. G., and D. Teece. 2009. Dynamic Competition in Antitrust Law. Journal of Comparative Law and Economics 5: 581. US Department of Justice. 2008. Competition and Monopoly:  Single-Firm Conduct under Section 2 of the Sherman Act. Van Dijk, T., and F. Verboven. 2008. Quantification of Damages. In W. D. Collins, Issues in Competition Law and Policy. Chicago: ABA Section of Antitrust Law. Viscusi, W. K., J. M. Vernon, and J. E. Harrington. 2005. Economics of Regulation and Antitrust. 4th ed. Cambridge, MA: MIT Press. Von Hippel, E. 2005. Democratizing Innovation. Cambridge, MA: MIT Press.

CHAPTER 9

TREATMENTS OF MONOPOLIZATION IN JAPAN AND CHINA PING LIN AND HIROSHI OHASHI

9.1. Introduction Monopolization (or abuse of a market dominant position) is prohibited under both Japan’s Action of Prohibition of Privation Monopolization and Maintenance of Fair Trade, enacted in 1947, and China’s Antimonopoly Law, passed sixty years later in 2007 (hereafter AMA and AML respectively). This chapter describes how monopolization is treated under the competition laws in the two largest economies in the world behind the United States. In addition to descriptions of the relevant legal provisions, it also focuses on the economic principles and reasoning that have been adopted in Japan and China, as reflected in their laws and supplementing guidelines and regulations, and court rulings. Lastly the chapter attempts to provide a comparative perspective to the legal and enforcement structures of these two countries.

9.2. Japan 9.2.1.  Overview of Japan’s Antimonopoly Act The principal antimonopoly act in Japan was drafted under pressure from the occupation forces and issued on 14 April 1947.1 Upon its enactment three months later on 20 1  Japan’s Anti-monopoly Act is an abbreviated name of “the Act on Prohibition of Private Monopolization and Maintenance of Fair Trade.” An overview of the AMA is available, for example, in Inoue (2007) and Wakui (2008).

Treatments of Monopolization in Japan and China   189

July, the AMA established the Japan Fair Trade Commission (hereafter JFTC), as the primary antimonopoly regulatory body. The Commission’s duty has been to achieve the purposes set forth in Article 1 of the act, which aims to promote free and fair competition in markets, to stimulate the creative initiative of entrepreneurs, to encourage business activities, to heighten the level of employment and actual national income, and thereby to promote the democratic and wholesome development of the national economy as well as to assure the interests of general consumers.2

While Article 1 contains a multitude of purposes, most legal scholars and practitioners consider the JFTC’s chief duty to be “to promote free and fair competition in markets.” Looking back over the sixty-five years that have passed since the establishment of Japan’s AMA and JFTC, we may observe a number of developments in both the legal status and the enforcement structure associated with the general prohibition of monopolization in Japan. In particular, the enforcement of monopolization was notably lax until the late 1990s, since which time it has been heightened considerably. To understand the past lack of enforcement and the recent changes in the JFTC’s attitudes towards regulations on monopolization, it is helpful to adopt a historical perspective when assessing the AMA. Therefore, before proceeding directly to the issues of monopolization, this section begins with an overview of the AMA, followed by accounts of the historical background to Japan’s antimonopoly act (section 9.2.2) and its enforcement structure (section 9.2.3). Section 9.2.4 details specific regulations concerning monopolization in Japan, with a particular focus on four types of exclusionary monopolization in section 9.2.5. Section 9.2.6 provides a future perspective on Japan’s AMA and private monopolization. Drafted under the strong leadership of the American occupation forces, many provisions of the AMA resemble the US federal antitrust laws, namely, the Sherman Act,3 the Clayton Act,4 and the Federal Trade Commission Act.5 The AMA includes four major categories of regulations: the prohibition of private monopolization (the first clause of Article 3), the prohibition of unfair trade practices (Article 19), the prohibition of unreasonable restraint of trade (the final clause of Article 3), and regulations on mergers and acquisitions (Chapter 4). We briefly explain the first two categories below, as they are relevant to the topic of this chapter. Private monopolization is the legal name for monopolization in Japan. The word “private” originally referred to both for-profit and non-profit businesses, as distinct from public businesses and those with governmental functions. However, such a distinction is no longer clearly maintained, and the regulations on private monopolization can be 2  An unauthorized English translation of the act is available at www.jftc.go.jp/e-page/legislation/ama/​ amended_ama.pdf. 3  Sherman Act 15 USC §§ 1–7 (2000). 4  Clayton Act 15 USC §§ 12–27 (2000). 5  Federal Trade Commission Act 15 USC §§ 41–58 (2000).

190   Ping Lin and Hiroshi Ohashi applied to state-owned enterprises.6 The regulations prohibit any exclusion or control of the business activities of other firms resulting in a substantial restraint of trade in a relevant market, as we will define later in section 9.2.4. Unfair trade practices refer to certain business activities defined in Article 2.9 of the act and are designated by the JFTC. They are regarded as a concept unique to the AMA. Conduct subject to the regulations of unfair trade practices includes activities such as abuse of superior bargaining position, below-cost pricing, refusal to sell, and nine other such violations. Indeed, the regulations on unfair trade practices were originally viewed as a supplemental role in antitrust enforcement, and cover situations not usually considered to be the domain of competition laws, such as the protection of consumers and small- and medium-size enterprises (SMEs). It is widely believed that it is easier to establish a violation of unfair trade practices, which requires “a tendency to impede fair competition,” than it is to establish a violation of the provisions prohibiting private monopolization, which requires “a substantial restraint of trade” (Takigawa, 2009; Negishi and Funada, 2010). The provisions in the act against unfair trade practices have been criticized by business and academic groups for their unclear standards delineating illegal conduct. This issue is one of the primary reasons why the regulations on private monopolization were inactive until the late 1990s, as we will discuss further in the remaining sections.

9.2.2.  Historical Background of the Japanese Act 9.2.2.1.  The Original AMA: 1947 to 1953 The original AMA7 was drafted by Judge Posey T. Kime of the Antitrust and Cartels Division of the General Headquarters of the Allied Forces, and passed by the Imperial Diet on 31 March 1947. This date was the last day before the Diet was abolished by the order of the Allied Forces. The original AMA contained the same provisions in its Article 3 (the prohibition of private monopolization and unreasonable restraint of trade) as those in the current AMA; however, it contained several unique features not found in the current act that were sometimes considered severer than US antitrust laws. These measures mandated (1) the strict control of business combinations (Articles 9–16), (2) the per se illegality of cartels (Article 4), and (3) the prohibition

6 

The cases against state-owned enterprises include the Metropolitan Slaughter House Case (Supreme Court, 18 November 1989; 43 12 MINSHU 2078), Municipal Bus for Aged Citizens case (Yamaguchi District Court Shimonoseki Branch, 16 January 2006; 52 SHINKETSUSHU 918), Yamato Transport v. Japan Post (Tokyo High Court, 28 November 2007; 2006 (Ne) No. 1078), New Year Postcard Case (Supreme Court, 18 December 2008; 45 SHINKETSUSHU 467), among others. 7  JFTC (1997) is an authoritative reference to the history of the AMA and JFTC. Kisugi (1999) summarizes the history before 1977, and Kojo (1999) summarizes the period after 1977. As we will discuss below, the name of the original AMA, The Law Relating to Prohibition of Private Monopoly and Methods of Preserving Fair Trade, differs from the current one.

Treatments of Monopolization in Japan and China   191

of unreasonable differences in terms of economic power among business entities (Article 8). There was a view in the office of the Supreme Commander of the Allied Powers (the so-called SCAP) that the concentration of economic power was primarily responsible for the sense of national militarism that had carried Japan into the Second World War. In particular, the SCAP looked critically upon the zaibatsu (at the time principal four were Mitsui, Mitsubishi, Sumitomo, and Yasuda). Through such conglomerates, often a single family alone held the ownership of hundreds of businesses. The zaibatsu were viewed as controlling not only finance, industry, and commerce, but also major functions of the government. No doubt the SCAP perceived it essential to dissolve the zaibatsu in order for Japan to govern itself as a free and democratic nation. Indeed, in the same year as the original AMA was passed, the SCAP enacted the Law of the Elimination of Excessive Concentration of Economic Power (hereafter LECP), and designated a total of 325 companies to be dissolved for standing in violation of the law. As shown in figure 9.1, the JFTC appeared to spend most of its resources in enforcing the aforementioned three activities along with LECP (under the heading of “others”), and only three violations were reported under private monopolization during the period. A few years of experience enforcing the original AMA made apparent that the concentration measure was not useful, even sometimes harmful, to the recovery of the Japanese economy. Indeed, damages incurred from the Second World War were so devastating8 that a quick economic recovery would not have been possible without the help of the zaibatsu and other large companies. To rebuild the Japanese economy and to transform the country into a fully industrialized nation capable of serving as a wall against communism, it was deemed essential to maintain some economies of scale in industries. This feeling culminated when Japan signed the San Francisco Peace Treaty to regain its independence in 1951. Two years later, the original AMA was amended and the major provisions of the act were relaxed or eliminated with regard to the three features unique to the original AMA listed above. First, the control of business combinations was eased so much that banks were allowed to hold shares—a prime example of the fact that Japanese anticompetition policies became more tolerable of the zaibatsu. Second, concerted practices were no longer deemed per se illegal, but instead were judged according to a rule of reason; “[a]‌substantial restraint of competition in any particular field of trade” was required for concerted practices to be considered illegal under the regulations pertaining to the unreasonable restraint of trade, then as today. The change in the treatment of the cartel regulations opened up many types of cartelized activities led by government authorities, such as MITI, then the Ministry of International Trade and Industry, under the names of “depression cartels” or “rationalized cartels.” As many as 49 cartel activities were legally exempted from the AMA in the 1960s.

8  To take an example from the steel industry, more than 70% of blast furnaces were rendered inoperable by wartime bombardment.

192   Ping Lin and Hiroshi Ohashi 70

Others Unfair trade practice Cartels

60

50

40

30

20

10

1947 1949 1951 1953 1955 1957 1959 1961 1963 1965 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011

0

FIGURE  9.1  The Number of Administrative Procedures (Cease-and-Desist Orders and Surcharge Payment Orders)

Third and finally, the prohibition of unreasonable differences in terms of economic power among business entities was eliminated from the act. However, at that time, public pressure grew, calling for the protection of small and medium-sized enterprises conducting business transactions alongside larger-sized corporations. Thus, the government introduced two new provisions in the AMA. One banned the abuse of bargaining position, which has since become known as the abuse of superior bargaining position under the current act.9 The other provision was the Act against Delay in Payment of Subcontract Proceeds, Etc. to Subcontractors (the so-called Subcontract Act), which entered into force in 1955. Both of the new provisions have been powerful tools to protect small and medium-sized enterprises. In particular, in cases where the 9 

Note that abuse of bargaining position was not considered an unfair “method” of competition, but rather regarded as an unfair trade “practice.” This new provision regarding the abuse of bargaining position presumably led to this new provision.

Treatments of Monopolization in Japan and China   193

payment of subcontract proceeds is delayed, the Subcontract Act automatically charges main subcontracting firms, defined as entities with capital or total contribution exceeding 300 million yen, of abusing their superior position, without considering the subcontracting firms’ situations to the contrary.

9.2.2.2.  Competition Policies versus Industrial Policies: 1953–1977 During the late 1950s and the 1960s Japan experienced rapid economic growth. This coincides with the period when transactions of goods and capital were liberalized, as Japan became a member of the GATT and IMF. Faced with increasing pressure from international competition, MITI came to hold the view that it was essential for Japanese firms to expand in size and achieve the maximum potential from economies of scale. A set of policies led by MITI that aimed to restructure industries through mergers and acquisitions, or through depression and rationalized cartels, were known as industrial policies,10 tensely squaring off against competition policies. The tension reached new heights with the merger case between the Yawata and Fuji Iron and Steel companies in the late 1960s. It is interesting to note that the two companies were originally the same company under the name of Nippon Steel, which had been previously dissolved by the LECP, the stringent competition law mentioned above. Thus, it would be a symbolic defeat for competition policies and a victory for industrial policies should the new proposed merger pass muster. As a matter of fact, however, the JFTC had only a minor influence in this merger case. The unification of Yawata and Fuji took place in 1970, and Nippon Steel was created once again, leading to the second largest steel producer in the world after U.S. Steel. Competition policies would not be enforced afterwards in Japan until the next amendment to the AMA in 1977. Indeed, violations in private monopolization led to merely four cases during the quarter-century from 1953 to 1977.

9.2.2.3.  The Amendment of 1977 Inflation rates as high as 23 percent after the Oil Crisis of 1973 shed new light on the importance of competition policies. Public outcry against the high inflation rates and possible price-gouging behavior demanded strict enforcement of the AMA from the government. To effectively enforce the law to fight against market concentration, an amendment to the AMA was passed in 1977. Among a number of changes introduced with this amendment, a primary one worth mentioning concerns the introduction of surcharges. The JFTC was empowered to order participants in pricing cartels to pay administrative fines to deprive them of the excess profits reaped by such behavior. The amount of the surcharges was calculated according to a formula, namely, as a percentage (stipulated in the act) of the sales amount for products or services related to the 10 

While there has been no clear definition of the term (and this is perhaps the reason why industrial policies gained extensive support in the first place), from an economics perspective, industrial policies are defined as those that facilitate a shift in inter- and intraresource allocations among sectors to correct for possible market failures.

194   Ping Lin and Hiroshi Ohashi anticompetitive activity involved. The types of anticompetitive activities applicable to surcharges under the act expanded to private monopolization in 2006. However, the nature of the surcharge system has essentially remained the same since its introduction. We discuss the surcharge system in greater detail in section 9.2.3.2. The introduction of surcharges, however, did not lead to an increase in the number of JFTC cases involving anticompetitive behavior. As figure 9.1 shows, the number remained quite low in the period immediately following 1977. While the JFTC for the first time lodged a criminal complaint against a wholesale gasoline cartel,11 it did not seek subsequent charges against other firms potentially conducting price-gouging behavior afterwards. Private action for compensation damages was difficult to seek given the high standards of proof demanded on the side of plaintiffs.

9.2.2.4.  The Japan-US Structural Impediments Initiative in the 1980s and Afterwards The bilateral relationship between Japan and the United States was put to the test on several occasions from 1989 through the first half of the 1990s. Against the backdrop of continued large trade imbalances, pressure for tougher policies towards Japan intensified in the United States. Individual trade issues, especially involving semiconductors, raised in connection with the Super 301 provision of the Omnibus Trade Act and how to deal with the Structural Impediments Initiative (hereafter SSI), all emerged as a source of diplomatic contention between the countries. The SSI talks were initiated by President George Bush and Prime Minister Sosuke Uno in an attempt to resolve the structural problems that stood as impediments to the adjustment of trade and the trade balance between the two countries. The United States faulted Japan for one such structural impediment:  namely, its exclusionary business practices and accompanying poor enforcement record. Japan’s stance toward competition proved the subject of intense discussions during the talks and changed as a direct result of them. For, following the SSI, the JFTC increased the rates of surcharges in response to US criticisms of the JFTC’s underdeterrence of anticompetitive behavior. We further elaborate upon the Japanese surcharge system in section 9.2.3.2. The JFTC also increased the number of staff and other human resource capacities to improve the enforcement system of the JFTC. The scope of exemptions under the AMA was also narrowed; the JFTC scaled down the exemption of resale price maintenance and wholly abolished the existing depression and rationalization cartels. As a consequence, the number of JFTC cases increased in the 1990s. It is worthwhile to note, however, that no cases were reported regarding violation of private monopolization for 24 years in the period from 197212 to 1996.13

11 

Supreme Court, 24 February, 1984; KEISYU 38.4.1287. Toyo Seikan, 18 September 1972; SHINKETSUSYU 19.87. 13  Japan Medical Food Association, 8 May 1996; SHINKETSUSYU 43.209. 12 

Treatments of Monopolization in Japan and China   195

9.2.3.  Structure of Japanese Antitrust Enforcement The United States delegates antitrust enforcement to two agencies: the Antitrust Division of the Department of Justice (hereafter DOJ) and the Federal Trade Commission (hereafter, US-FTC). While the AMA did emulate American antitrust laws, antitrust enforcement in Japan bears no resemblance to the practices in the United States. Unlike the United States, Japan has only one antitrust agency. The JFTC was apparently modeled after the US-FTC: it is an independent agency under the Prime Minister’s Office. The JFTC has five commissioners, including one chairperson, who are appointed by the prime minister with the consent of both houses of the Diet (Article 29). Commissioners have five-year terms (Article 30), and they are protected against pay cuts (Article 36) or removals without cause or without their consent (Article 31). Having only one antitrust agency responsible for enforcing the act has been advantageous in the sense that it has helped promote the development of a uniform national competition policy. On the other hand, it makes the JFTC maintain a jurisdictional monopoly in antitrust enforcement. Miwa and Ramseyer (2005) and Harris and Ohashi (2011) argue that this kotorii-chusinsyugi, or JFTC-centrism, affords the JFTC a degree of discretion in the application of the law. An illuminating example comes from the relationship between the private monopolization and the unfair trade practices regulations. Figure 9.1 indicates that, throughout the period from 1947 to 2011, the JFTC’s enforcement of private monopolization was weak. This is perhaps because the JFTC required high standards of proof to establish evidence of adverse effects on competition in the relevant market. Before 2009 only cease-and-desist orders were imposed for the regulation of exclusionary conduct, as was the case for the regulation of unfair trade practices. Since the unfair trade practice regulations do not require proof of adverse competitive effect and they essentially cover the regulations on private monopolization, the JFTC tends to make use of the former regulations, even if the application of private monopolization regulations seems appropriate. It was indeed expected that the regulation of unfair trade practices would elicit less opposition from the private parties involved because it did not come with surcharges or criminal penalties. With the aim of improving the act’s effectiveness and deterrence effects, the JFTC brought in administrative surcharges under the category of private monopolization; controlling conduct in the 2006 amendments and exclusionary conduct in 2009. The JFTC also made public the new guidelines against exclusionary conduct. This series of revisions leaves the clear impression that the JFTC will intensify enforcement activities against private monopolization, especially in the area of exclusionary conduct. A troubling issue is that the 2009 amendments to the AMA also extended coverage of surcharges to include certain types of unfair trade practices, such activities as unjust low-price sales and abuse of a superior bargaining position, as the result of strong political influence. Note that the regulations concerning unfair trade practices are often considered a precautionary measure to protect consumers and SMEs, on which Japanese politicians are very keen as a voting bloc. Since private monopolization and unfair trade

196   Ping Lin and Hiroshi Ohashi practices are now subject to surcharges, some observers believe that the unfair trade practices regulations should no longer be supplemental to the private monopolization regulations and that these two sets of regulations should be unified (Murakami, 1997).

9.2.3.1.  Procedure for Antitrust Cases The JFTC formally takes two types of administrative measures—a cease-and-desist order and a surcharge payment order—if it finds a violation of AMA. Subsequently, the Decision and Lawsuit Division under the JFTC initiates a formal administrative hearing (hereafter Shinpan hearing) upon the request of the respondent. The Shinpan hearing adopts the adversary system, in which independent referees (namely hearing examiners), who are members of the JFTC staff, make a judgment. The structure of the Shinpan hearing has long been criticized as being rather inquisitorial, because within the JFTC staff no clear firewall has been erected between Shinpan referees and Shinpan prosecutors (see, e.g., Murakami, 1997). During the formal hearing, the respondent firm may challenge the JFTC’s fact-finding and application of the law and may make a remedial proposal. If the JFTC finds that the remedy proposed by the respondent is appropriate, the JFTC drops the case,14 or orders the firm to perform the remedy. If the JFTC finds that a violation exists, the JFTC issues a formal decision at the end of the Shinpan hearing procedure, ordering the firm to perform the remedy. If the parties are dissatisfied with the decision delivered after the Shinpan hearing, they may enter into litigation at the Tokyo High Court, which has exclusive jurisdiction to revoke the JFTC judgment. In an instance where the judicial review is exercised, the court is bound by the facts found in the JFTC’s decision if such facts are “supported by substantial evidence” (Article 80). If the court finds that a reasonable person would reach the same conclusion as did the JFTC, based on the same facts the court is obliged to uphold the JFTC’s findings. Thus, under the judicial review system, the plaintiff tends to argue that the facts upon which the JFTC’s decision was made are not supported by substantial evidence. The plaintiff may also file an appeal against the judgment of the Tokyo High Court to the Supreme Court.

9.2.3.2. Sanctions Along with the formal administrative procedures (namely, a cease-and-desist order and a surcharge payment order), criminal and civil procedures in court may also be instituted, and violations of private monopolization are no exception. Civil procedure is discussed in section 9.3.3. Regarding criminal procedure, the JFTC has the sole authority to file an accusation with the prosecutor general, but only when the public investigator commences the investigation. In instances where the prosecutor general decides not to prosecute a case, the decision must be reported with reasons to the prime minister (Article 74(3)). No criminal charges have been executed against private monopolization, and no one has been imprisoned under violations of the AMA. 14 

The recent case is JASRAC (2012), as we discuss in section 4.

Treatments of Monopolization in Japan and China   197

As discussed in section 9.2.2, the surcharge system was introduced in 1977. The original purpose of the surcharge was to disgorge undue profits obtained from anticompetitive activities, and this purpose still remains alive in the current system. Surcharges are a rigid administrative fine. There are two aspects to the rigidity: first, the JFTC is wholly obliged to issue a surcharge order if it is applicable. Second, the amount of the surcharge is calculated according to a formula, namely, as a percentage (stipulated in the act) of the sales amount for products or services related to the anticompetitive activity involved.15 The surcharge rate differs according to each firm’s capital or total amount of contribution and according to line(s) of business, not according to the severity of damages or the amount of excess profit accrued by the alleged anticompetitive behavior. The surcharge payment system has significantly evolved since 1977, when it was first introduced into the regulations on cartels. Initially, surcharges were a set rate of 1.5 percent of sales of products related to the alleged anticompetitive activity, and subsequently increased to a rate of 6 percent in 1991.16 In order to improve the deterrence effects of the act, activities under the heading of private monopolization were added under the scope of surcharge payments in the 2006 amendment to the AMA. The surcharge rate increased still further, from 6 percent to 10 percent, at the same time. Because of the possibility that, at this higher rate, the surcharge could exceed the amount of undue profits accrued from the alleged anticompetitive behavior, the JFTC effectively added a new function to the role of surcharges in that they would serve as deterrence against violations of the AMA. In the 2009 amendments, the scope of the surcharge payment order was further expanded to incorporate exclusionary conduct falling under the private monopolization regulations and five designated activities under the unfair trade practices category (namely concerted refusal to trade, discriminatory consideration, retail price maintenance, abuse of superior bargaining position, and unjust low-price sales). The surcharge rates were set at 6 percent of sale values for exclusionary conduct and in the range of 1 to 3 percent for the five unfair trade practices specified (see table 9.1 for details).17 Two other changes to the surcharge system accompanied the 2009 amendments: the introduction of a leniency program to promote voluntary notification of violations from cartel members themselves; and the introduction of greater flexibility in surcharge calculations. While these changes to the available pecuniary sanctions may enhance the enforcement effectiveness of the AMA, there remains room for improvement. In particular, note that in calculating the surcharges, the JFTC takes into account only domestic sales. Thus, it is currently impossible for the agency to collect fines from companies that 15  Note that a reduction or exemption through an application of the leniency program and coordination with criminal penalties should be considered an exception. We will discuss this point shortly in this subsection. 16  The rates shown here are applied to the category of a large company mainly in the manufacturing sector; small and medium-sized companies were charged at lower rates. 17  To be precise, the rate of 3 percent of sale values was set for concerted refusal to trade, discriminatory consideration, retail price maintenance, and unjust low-price sales; and the rate of 1 percent of transaction values (sales or purchase values) was imposed for the abuse of superior position.

198   Ping Lin and Hiroshi Ohashi conduct their business abroad, that is, that participate in cartel activities but do not sell to Japan. This structure of the surcharge system appears particularly troublesome given that the JFTC must enforce the law in a globalized economy where cross-border transactions are common. For example, in the Marine Hose Case (2008),18 surcharges were calculated based on sales to consumers located in Japan, and foreign companies that were identified as taking part in the cartel activity under the cease-and-desist order were exempted from the surcharge payment. In the Cathode-Ray TV Tube Case (2009),19 surcharges were levied based on the sales ordered by Japanese TV manufacturers; however, a majority of cathode-ray TV sets under the surcharge payment order were sold outside of Japan. As multiple jurisdictions have introduced competition laws and begun to enforce cartelized activities, double or even triple punishments for the same conduct across nations have emerged as a real possibility. While it is often argued in Japan that a discretionary administrative fine violates transparency and predictability concerns from the respondents’ point of view, it may be worthwhile to reconsider introducing a discretionary surcharge system. Such a system is more in line with the original purpose of the surcharge, in that it accounts for the difference in the amount of disgorged undue profit. Of course, calculating the appropriate

Table 9.1  Surcharge Rates Sector Types Manufacturing and others Unreasonable Restraint of Trade Private Monopolization Controlling Conduct Exclusionary Conduct Unfair Trade Practices Concerted refusal to trade, discriminatory consideration, retail price maintenance, and unjust low price sales Abuse of superior bargaining position

10% (4%)a

Retail 3% (1.2%)a

Wholesale 2% (1%)a

10% 6%

3% 2%

2% 1%

3%

2%

1%

1%

Note: a: The number inside the paranthesisis the rate applied to small and medium-sized enterprises.

18 

JFTC, 20 February 2008; 54 SINKETSUSYU 512. The cease-and-desist order and surcharge payment order is also available at http://www.jftc.go.jp/en/pressreleases/uploads/2008-Feb-22.pdf. 19  JFTC, 9 October 2009; The cease-and-desist order and surcharge payment order are also available at http://www.jftc.go.jp/en/pressreleases/uploads/2009-Oct-7.pdf.

Treatments of Monopolization in Japan and China   199

amount of a discretionary surcharge requires the discipline of economics, which may be lacking in current JFTC practices.20

9.2.3.3.  Civil Procedure The lack of effective consumer involvement in enforcement was long a feature of the AMA. Even though the AMA was strongly influenced by its American counterparts, as we have previously noted, there was a marked tendency in the law for both criminal and administrative sanctions to be exercised solely by the JFTC, undermining the role of private initiatives. In response to criticisms, the civil injunction system was finally implemented in 2000. Unfortunately, this system currently demands extremely high standards of proof for the existence of damage from complainants, and therefore no civil injunctions have successfully been approved as of today. The no-fault compensation suit is a private lawsuit specifically prescribed under Article 25 of the AMA. The damage award is not trebled: instead, single damages are payable. A firm whose interest is seriously injured by unfair trade practices may seek injunctive relief from the court for the suspension of such conduct (Article 24 of the act). Both of these private resources have failed to work properly, however. That is presumably because plaintiffs, under the civil litigation procedures in Japan, do not have sufficient capacity to collect evidence to prove the existence of violations. As we will discuss below, these private actions can in principle be a complement to, rather than a substitute for, JFTC’s enforcement, and they should thus be activated effectively.

9.2.4.  The AMA on Monopolization The regulation of monopolies is defined as the regulation of “private monopolization” in the AMA. Private monopolization is an important pillar of the AMA. It is therefore interesting to note, as has already been pointed out in section 9.2.2, that the regulation of private monopolization was rarely enforced, resulting in only a handful of cases before the late 1990s (as shown in figure 9.1). This section describes the general rules that govern private monopolization under the AMA. Section 9.2.5 then takes up certain types of activities that fall under the category of exclusionary conduct.

9.2.4.1.  Private Monopolization under Japanese Law In the provision contained in Article 2(5), private monopolization is defined as such business activities, by which any entrepreneurs, individually or by combination or conspiracy with other entrepreneurs, or by any other manner, excludes or controls the business activities of other entrepreneurs, thereby causing, contrary to the public interest, a substantial restraint of competition in any particular field of trade. 20  According to Global Competition Review (2012), the share of economists employed at the JFTC is 5 percent, which is smaller than in other industrial countries.

200   Ping Lin and Hiroshi Ohashi Private monopolization is prohibited per the terms outlined under Article 3. Before considering Article 3, we first provide an explanation of each component of Article 2(5) in detail. Entrepreneurs: The entity referred to Article 2(5) is an individual entrepreneur, or a multiple number of entrepreneurs.21 According to the article, the J-FTC regulates under the private monopolization heading any conspiracy among many entrepreneurs that involves excluding or controlling competitors. In practice, however, the private monopolization label is applied to unilateral conduct; coordinated conduct such as conspiracy is usually considered under the regulation of unreasonable restraint of trade. As discussed below, since exclusion and control exhibit wrongful behavior on the part of a monopolizing party, private monopolization is usually considered as inherently “contrary to the public interest.” In this chapter, we use the terms “entrepreneur” and “firm” interchangeably. Exclusion: The term “exclusion” indicates an act that makes it difficult for other entrepreneurs to continue their business operations (thus eventually forcing their exit) or an act that deters the entry of new entrepreneurs into a relevant market. The types of conduct that fall into this group include predatory pricing, exclusive dealing, tying, refusal to supply, and discriminatory treatment. As economists know, it is difficult to draw a clear line between normal competitive business practices and anticompetitive exclusion, and the JFTC is well aware of the issue. The Guidelines for Exclusionary Private Monopolization under the Antimonopoly Act, issued for the first time in 2009 (hereafter, “the Guidelines”), stipulate “in the case that an entrepreneur supplies a low-cost and high-quality product by its own efforts such as improving efficiency [to which the JFTC refers as “competition on merit”], and if such conduct would make it difficult for competitors to continue their inefficient business activities, it does not fall under exclusionary conduct because it is a result of fair and free competition, which the Antimonopoly Act intends to promote.” To constitute exclusionary conduct, such conduct must cause the “substantial restraint of competition in any particular field of trade,” and does not have to result in the actual elimination of competitors from the market. At the same time, if the elimination of competitors from the market does not lead to substantially hindered competition, the conduct does not fall under the heading of “exclusion.” The Guidelines also make it clear that the intent to exclude is in itself not essential for the JFTC to establish exclusion; however, the intent of entrepreneurs, if identified, would help convince the JFTC of an important fact that might lead to the presumption that the alleged conduct is exclusionary.

21  Article 2(1) of the AMA defines an entrepreneur as “a person who operates a commercial, industrial, financial, or any other business.” Any other business is interpreted as “any entity that engages in business activities in which economic interest of any nature is provided and performance in return is repeatedly and continuously received.” A natural person and a legal entity can both be entrepreneurs, either incorporated or unincorporated.

Treatments of Monopolization in Japan and China   201

When the market is correctly defined, an entrepreneur (or entrepreneurs operating collectively) must have a large market share to exercise exclusionary conduct. The JFTC prioritizes its investigation on cases in which the entrepreneur owns a market share of the product in question exceeding 50 percent after the alleged exclusionary conduct takes place.22 Of course, the JFTC quickly adds the caveat in the Guidelines that how it prioritizes its investigation resources is subject to the type of activities in question, the degree of product differentiation, and other features of market structure that we will discuss in section 9.2.5.4. One of the unique features of the AMA, compared with EC and US laws, is that, even when the JFTC finds no exclusionary practice, the alleged conduct can be deemed anticompetitive under the regulation of unfair trade practices, as discussed in sections 9.2.1 and 9.2.3. Control: Another form of conduct that is deemed to be monopolization under the AMA is the act of “controlling” the business activities of other entrepreneurs, causing a substantial restraint of competition in a particular field of trade. Compared with the classes of conduct associated with exclusion as discussed above, the concept of control is not as clearly defined in the JFTC’s practice from the vantage of economics. Only three cases have been identified by the JFTC as private monopolization control violations, two of which—the Japanese Medical Foods Case (1996)23 and the Paramount Beds Case (1998)24—were found to be exclusionary conduct as well. Thus, the only private monopolization case involving a control violation occurred more than half a century ago: the Noda Shoyu Case of 1957.25 Therefore, exclusion has been and will be the primary focus of private monopolization regulations in Japan. Particular field of trade: The JFTC judges the applicability of exclusionary private monopolization according to the degree of impact the alleged conduct has on competition in “a particular field of trade.” The term “particular field of trade” is mentioned also in Article 2(6), concerning the unreasonable restraint of trade, and in Article 15 on business combinations. As stated in the Guidelines, “a particular field of trade” has two main dimensions: the scope of the products (namely goods and services) under investigation and the geographical scope. Both types of scope are determined on the basis of substitutability from the consumers’ perspective. If necessary, the JFTC would also consider suppliers’ substitutability—how easily and quickly suppliers are able to switch over their sales and production processes from one product to another. One primary method of defining a particular field of trade is to rely on the small but significant nontransitory increase in prices (hereafter SSNIP) test, introduced in the US Merger Guidelines of 1982. These guidelines have been widely used by competition authorities to define relevant markets in a variety of contexts. The SSNIP test was 22  The “share of the product” refers to the share of the primary tying product in the case of tying discussed in section 5.2 and to the share of the upstream market in the case of refusal to supply and discriminatory treatment described in section 5.3. 23  JFTC, 8 May 1996; 43 SHINKETSUSYU 32. 24  JFTC 31 March 1998; 44 SHINKETSUSYU 362. 25  Tokyo High Court, 25 December 1957; 10 12 KOUSAIMINSYU 743.

202   Ping Lin and Hiroshi Ohashi brought in 2006 under an amendment to Japan’s Guidelines for the Application of the Antimonopoly Act Concerning the Review of Business Combination.26 However, it has not been intensively used in Japan, even in the course of horizontal merger investigations. Substantial restraint of competition: In the NTT East Case (2011) discussed in the next section, the Supreme Court held that “substantial restraint of competition in a particular field of trade” should be interpreted as “establishing, maintaining, or strengthening the state in which a certain entrepreneur or a certain group of entrepreneurs can control the market at will by being, to some extent, free to influence price, quality, quantity, and other various conditions after competition itself has lessened.” Given the Supreme Court’s assessment of substantial restraint of competition, it may be necessary but not sufficient for the firm to possess a certain level of market share. The exclusion of a firm that holds merely a percentage of the total market share does not seem to generally restrain market competition. Indeed, as described in section 9.2, the JFTC prioritizes investigating cases in which the product in question accounts for more than half of the relevant market after the alleged conduct takes place. Defining “a particular field of trade” is a task necessary to calculate this market share, but the market share threshold provides a mere guideline as to how to allocate the JFTC’s investigatory resources. As stated in the Guidelines, whether or not the agency challenges the practice as a violation of private monopolization is assessed on a case-by-case basis. The JFTC does not apply a specific criterion; according to the Guidelines, it considers the following four aspects: (1) market structure; (2) potential competitive pressure; (3) countervailing buying power, and (4) efficiency. We discuss how these four aspects are interlinked in the JFTC’s assessment of exclusionary practice in the following section. It should be noted that, as with horizontal merger investigation, exclusion resulting in (near) monopoly is usually not considered procompetitive by the JFTC.

9.2.5.  Four Types of Exclusionary Conduct While it is difficult to characterize all possible forms of exclusionary conduct, the JFTC describes four typical categories of exclusion in the Guidelines. They are below-cost pricing, exclusive dealing, tying, and the refusal to supply and discriminatory treatment. We describe each of them in turn below, along with a selective set of recent JFTC decisions and court judgments as available. Note that exclusionary conduct under the regulations of private monopolization and unfair trade practices in the AMA is not limited to these four types, and that the JFTC will investigate alleged conduct in the context of individual cases.

9.2.5.1.  Below-Cost Pricing The JFTC acknowledges that price-cutting competition based on entrepreneurs’ own efforts constitutes the core of competition, and thus that intervention in this process should be kept at a minimum to promote fair and free competition. The practice of selling 26 

http://www.jftc.go.jp/en/legislation_guidelines/ama/pdf/110713.2.pdf.

Treatments of Monopolization in Japan and China   203

a product at a very low price—at a level that would not allow an entrepreneur to recover its average avoidable cost (hereafter, AAC)—would normally be considered lacking in economic rationality and thereby undermine the competition process. The AAC is defined as the average of product-specific fixed costs and variable costs that could have been avoided if the entrepreneur had not produced extra output. The JFTC, on the practical level, takes the AAC as the cost that would not be generated unless the product was supplied. Below-cost pricing, setting a price below “the cost that would not be generated unless the product was supplied” so as to make it difficult for an equally or more efficient competitor to stay in business, may fall under exclusionary conduct under the AMA. While they can be defined at the theoretical level, determining that costs are actually “the cost that would not be generated unless the product was supplied” is critical in the JFTC’s practical assessment of below-cost pricing. The Guidelines explain that whether or not the cost will vary with the quantity of the product supplied is an important consideration in the assessment. For example, variable costs in economics, or production and purchasing costs in corporate accounting terms, are presumed to be “the cost that would not be generated unless the product was supplied.” Setting a product price below the average total cost (namely the cost required for supplying the product) but not less than “the cost that would not be generated unless the product was supplied” is unlikely to be deemed significant enough to drive an equally or more efficient competitor out of the market. Thus, such price-setting behavior may not fall under the category of exclusionary conduct, unless there are exceptional circumstances in which large quantities of the products were supplied at that price over so long a period that such conduct would not be considered economically rational even from a long-term perspective. A recent case is K.K. Usen Broadband Networks, K.K. Nihon Network Vision.27 Usen Broadband Networks (hereafter Usen) is the largest cable broadcasting company in Japan, and engaged in a sales contract with Nihon Network Vision (hereafter NNV). Usen, jointly with NNV, offered a discounted monthly listening fee only to the customers of their major competitor, Can System, thereby increasing its market share from 68 percent to 72 percent, while Can System lost 6 percent of its market share in one year. The JFTC issued a cease-and-desist order in 2004 against Usen, finding that it excluded the business activities of Can System. While the JFTC did not quantitatively assess whether the discounted fee offered to the Can System customers was below the “cost” of either Usen or NNV, it appeared to have found both Usen’s and NNV’s pricing strategy anticompetitive in that they targeted specifically Can System among other rivals. It remains to be seen how the JFTC will determine below-cost pricing before one can assign more cases to this category.

9.2.5.2.  Exclusive Dealing Exclusive dealing involves a supplier conditioning its sales to customers on their agreement not to purchase from its rivals. Prohibiting customers from purchasing rivals’ 27 

13 October 2004, 51 SHINKETSUSYU 518.

204   Ping Lin and Hiroshi Ohashi products is not of itself considered to be anticompetitive, in that every unit sold by a supplier would have been sold by its rivals given the market size. However, if such prohibition or restraint of transactions with competitors results in driving rivals out of the market, such conduct is deemed to be anticompetitive and falls under exclusive dealing regulations. Exclusive dealing takes various forms other than prohibiting the customers from transacting with rivals. For example, one form occurs when an entrepreneur requires customers to purchase a minimum amount. Another form might occur if an entrepreneur should require one or more customers to obtain approval to deal with the entrepreneur’s rival. The latter form could be anticompetitive if the given entrepreneur provides benefits to customers not to make transactions with its rivals. The JFTC considers various aspects of market structure as a whole in its assessment of exclusive dealing, including the degree of market concentration, product characteristics and differentiation, economies of scale in production and distribution, market dynamics, and the difficulty of entry. For example, if the market exhibits strong network effects and/or reputation, or if competitors are subject to capacity constraints, exclusive dealing would likely be effective in driving rivals out of the market. The Guidelines also mention an anticompetitive aspect in the use of stipulated damages in the contract. Aghion and Bolton (1987) and Whinston (2006) showed that a buyer and seller could use stipulated damages clauses to extract profits from a potential entrant. In their models, the damage provision typically creates inefficiency, as it makes the buyers transact with the potential entrant less frequently than is optimal. The JFTC expresses its concern in the Guidelines that the rebates could serve as means of facilitating exclusive dealing, although there are many instances when rebates work procompetitively through stimulating demand. Especially in cases in which the rebate amount is tied to the volume of transactions conducted with the entrepreneur who provides the rebates, and this amount is progressively set in accordance with volume, the JFTC is careful to monitor possible instances of exclusive dealing. The Intel Case (2005) below provides insight into what the JFTC considers to be unlawful rebates.

9.2.5.2.1.  The Intel Case (2005) Intel Corporation of the United States has a subsidiary in Japan, which sells central processing units (CPUs) to Japanese PC manufacturers with a substantial market share. Advanced Micro Devices (hereafter AMD) in Japan began selling CPUs at a lower price than Intel Japan, and improved its market share from 17 percent to 22 percent in the period from 2000 to 2002. In response to this aggressive pricing by AMD Japan, Intel Japan offered price rebates to the five major Japanese PC manufacturers and subsidies related to the “Intel Inside Program” supporting advertisements for PC manufacturers products using Intel’s microprocessors. In order for the leading manufacturers to be eligible for the rebates and subsidies, however, Intel required the PC manufacturers to purchase “X” percent of their required microprocessors from Intel, in which Intel set a different value of X for a different manufacturer, such as Toshiba and Sony. After the introduction of the rebates and subsidies program, the total market shares of Intel’s rivals

Treatments of Monopolization in Japan and China   205

(mainly AMD and Transmeta) plummeted from 24 percent in 2002 to a mere 11 percent in 2003. The JFTC found that Intel’s exclusivity conditions prevented the business activities of Intel’s rivals, and thus violated the prohibition of private monopolization.28 In the decision, the JFTC was not very clear as to how it evaluated a substantial restraint of competition. As Amemiya (2005) clearly pointed out, the JFTC’s recommendation order did not contain discussion of how the alleged harm to these competitors would have resulted in adverse consequences to the PC manufacturers that purchased the relevant products, nor did it explain why the competitors were unable to match the rebates or subsidies offered by Intel to retain the business of these customers.

9.2.5.2.2.  The JASRAC Case (2012) The JFTC found on 27 February 2009 that JASRAC (Japanese Society for Rights of Authors, Composers and Publishers), a dominant copyright management organization, excluded other copyright management entities from the market by entering into “comprehensive contracts” with broadcasting companies for their musical works.29 Under the contracts, JASRAC charged a fixed amount of total royalty payment, regardless of how many licensed musical works were broadcast. JASRAC requested the commencement of Shinpan procedures in order to challenge the cease-and-desist order levied against it by the JFTC. During Shinpan procedures that commenced in May 2009, new evidence was submitted indicating that JASRAC’s competitors were not really excluded from broadcasting business, contrary to the claim in the cease-and-desist order. The JFTC delivered a decision requiring that the cease-and-desist order be rescinded, its reason being that there was no evidence to show that JASRAC’s royalty collection method had the effect of damaging the business activities of other copyright management organizations. The JFTC judged no exclusion be present on the basis of the evidence that no actual elimination of competitors was observed in this case. This decision is inconsistent with the JFTC’s prior findings including the decision made in the Intel case (2005) discussed above. An action for the revocation of the JFTC’s decision has been filed with the Tokyo District Court.

9.2.5.3. Tying The practice of tying, under which a firm requires customers who are purchasing product A also to purchase product B, has received a good deal of attention in economics literature. Tying itself should not be an anticompetitive practice, as it may create new value for buyers that would not be realized if the products were to be independently offered untied. However, tying the sales of a monopolized product to sales of partial substitutes at an appropriate margin can be considered as leverage of monopoly power, particularly when the substitutes could be otherwise obtained at competitive prices.

28 

JFTC, 13 April 2005, 52 SHINKETSUSYU 341. JFTC 12 June 2012. An English translation of the cease-and-desist order of 27 February 2009 is available at http://www.jftc.go.jp/en/pressreleases/uploads/2009-Feb-27.pdf. 29 

206   Ping Lin and Hiroshi Ohashi In practice, the essential matter is how the authority assesses two tied products, A and B, and how it envisages the two sold separately. Considerable attention is given to identifying the market structure for the respective products, including the nature of user demand, the degree of market concentration, and product characteristics, to name a few. Take the example of a mobile phone set that integrates a digital camera. In the Guidelines, the JFTC mentions that the features and usability of mobile phone sets are substantially altered without digital cameras. Since a mobile phone set with a digital camera can be regarded as a single product with distinct functions, the sales of a mobile phone set that integrates a digital camera should not be considered a tying practice under the current views held by the JFTC.

9.2.5.3.1.  The Microsoft Case (1998) The JFTC considered a Microsoft Japan licensing policy whereby Japanese manufacturers, including NEC and Toshiba, were offered licenses for the Excel spreadsheet application, but only on the condition that they also took licenses for Microsoft’s word-processing application, Word.30 In the process of negotiation, Microsoft rejected proposals from personal computer manufacturers that requested the separate licensing of Excel and Word. At that time in Japan, Ichitaro, produced by Just Systems, was the most popular word-processing application, followed by Microsoft Word, whereas Lotus 1-2-3 had the largest share next to Microsoft Excel in the spreadsheet market. In 1998, the JFTC found that Microsoft Japan’s conduct constituted an unfair trade practice under Item 10 of the General Designation (tie-in sales), and issued a recommendation decision to the company to take remedial measures. Microsoft Japan accepted the recommendation. Therefore, the JFTC issued a cease-and-desist order requiring Microsoft Japan to cease imposing conditions on individual Excel licenses, to accept proposals by PC manufacturers to amend existing contracts to allow licenses for one or both of the Excel and Word applications, and not tie the licenses given to PC manufacturers for any other software to licenses for Word or Excel.

9.2.5.3.2.  The Toshiba Elevator Case (1997) Toshiba Elevator, a subsidiary under the Toshiba Group that manufactures elevators, refused to supply spare parts to independent maintenance service providers. At the time, there existed six competing manufacturers in the Japanese elevator market, of which Toshiba Elevator accounted for approximately a 20 percent share. Replacement parts were made incompatible with elevators made by other manufacturers, and Toshiba Elevator took almost the entire share of the maintenance and repair market for its own elevators. The Osaka High Court held on 30 July 1997 that tying the provision of replacement parts with the use of its own Toshiba maintenance service violated Article 10 of the

30 

JFTC, 14 December 1998; 45 SHINKETSUSHU 153.

Treatments of Monopolization in Japan and China   207

general designation (tie-in sales), and ordered Toshiba Elevator to pay damages to independent maintenance service providers.31 In the course of the proceedings, Toshiba Elevator argued that the independent service providers had an insufficient capability to maintain and repair the elevators manufactured by Toshiba Elevator; however, the court found otherwise. This case is regarded as being similar in nature to the Kodak case in the United States.32

9.2.5.4.  Refusal to Supply and Discriminatory Treatment A firm has relative freedom to determine which buyers to supply and on what conditions. However, if a firm, beyond a reasonable degree, imposes quantity restrictions or other discriminatory conditions on its supply to buyers, so as to place those buyers in such difficulty that they may struggle to continue doing business, this behavior may be regarded as exclusionary conduct. According to the Guidelines, the JFTC often considers the refusal to supply and discriminatory treatment to take place in transactions between firms in upstream and downstream markets. Whether or not the product in question is considered as a product indispensable for the trading customers to carry out their business activities is judged primarily on two factors: (1) whether the product is indispensable for the buyers to carry out their business downstream, such that no alternatives are available to them, and (2) whether it is impossible for the buyers to manufacture the products in question on their own. Products that satisfy both criteria may exhibit economies of scale or scope in production, or network effects in distribution. Alternatively, the right of use may be exclusively assigned to a particular supplier under regulatory oversight. Whether or not the refusal to supply or discriminatory treatment goes beyond reasonable degree will depend on the market structure in which the particular transaction at hand takes place. An example might arise in a scenario in which supply conditions differ between incumbents and new entrants in such a way that the difference cannot reasonably be explained by market conditions. Nevertheless, the JFTC explicitly mentions that it should respect the decisions made by firms regarding their transactions, and thus “whether or not refusal to supply and discriminatory treatment by a single entrepreneur falls under exclusionary conduct should be assessed especially prudently.” When assessing a potential case of refusal to supply and discriminatory treatment, the JFTC considers both the upstream and downstream market conditions. The JFTC examines both markets for information about market shares and product concentration, product characteristics and degree of product differentiation, economies of scale in production and distribution, difficulty in entry, and market dynamics. For instance, in a scenario where the products have strong network externalities in the upstream market, the buyers in the downstream market will not easily find an alternative source of supply, which makes the products more critical for the buyers. However, if the downstream buyer owns a large market share, the likelihood of monopolization would not be 31  32 

Osaka High Court, 30 July 1993; 833 HANREITIMES 62. Eastman Kodak Co. v. Image Technical Services Inc. 504 US 451 (1992).

208   Ping Lin and Hiroshi Ohashi considered substantial. Product prices and the duration of the practice in question also serve as important pieces of evidence for the JFTC to assess the legality of the conduct. Before we discuss two recent cases associated with refusal to supply and discriminatory treatment, let us briefly mention unfair trade practices. As we discuss in section 9.3, if a firm’s conduct is deemed not to engender adverse effects on market competition, but to enhance the likelihood of impeding fair competition, the alleged conduct may fall under unfair trade practices.

9.2.5.4.1.  The NTT East Case (2010) One interesting case concerns a decision made by the Supreme Court of Japan on 17 December 2010.33 Nippon Telegraph and Telephone East Corporation (hereafter NTT East) is a single company whose shares are owned by NTT under the NTT Law. NTT East operations accounted for high shares of the market in the supply of optical fiber telecommunications services to homes (hereafter, a service we will refer to by the acronym FTTH) in eastern Japan. Because of its high market share, NTT East falls under the regulatory oversight of the Ministry of Internal Affairs and Communications (hereafter MIC), and is obliged to offer its services to other telecommunications carriers for a set, authorized connection fee. However, NTT East set its user’s fee for the FTTH service at 5,800 yen for the first month and then at 4,500 yen per month thereafter, which is lower than the authorized connection fees. Therefore, in order to be competitive with NTT East in terms of price, other telecommunications carriers had to set a user fee that would make a negative margin under the homogeneous-good assumption. Since the connection provided by the NTT East facility is an input essential for other carriers to provide the FTTH service, these other carriers have to accept the price authorized by the MIC, which was higher than the price for the NTT East service itself. Noting that it had no choice but to offer the FTTH connection service at the authorized price, NTT East argued before the Supreme Court that consumers must have benefited from the lower price they offered. Although it was not explicit in the judgment, the Supreme Court essentially examined the welfare trade-off to conclude that NTT East’s pricing behavior was exclusionary towards the business activities of other telecommunication carriers intending to enter the subscriber FTTH market.34

9.2.5.4.2.  The Pachinko Patent Pool Case (1997) Ten leading manufacturers of pachinko (Japanese pinball) machines had formed an association and pooled a portfolio of their patents essential for the manufacture of pachinko machines. They manufactured almost all of the pachinko machines sold in Japan. Through that association (named Nittokuren), the manufacturers adopted a policy of granting no licenses to any of the pooled patents to any entity that was not a member of the association. The JFTC found in 1997 that this joint refusal to license 33  Supreme Court, 17 December 2010. HANREITIMES 1339·55. An English translation is available at the website http://www.courts.go.jp/english/judgments/text/2010.12.17-2009.-Gyo-Hi-.No..348.html. 34  A similar analysis is also made in Arai (2012).

Treatments of Monopolization in Japan and China   209

substantially restrained competition in the market for pachinko machines.35 The JFTC ordered the association members to abolish their no-license policy and to cancel all decisions taken in accordance with it. This decision is now used as an illustrative example in the Intellectual Property Guidelines, Part 3(1)(i)(a).36 The JFTC judged that while patent pooling that limits the use of licenses has some procompetitive effect, the policy of granting no licenses to non-pool members would eliminate such effects and substantially hinder market competition.

9.2.6.  A Future Agenda for the AMA and Summary The 2009 amendments to AMA discussed in sections 9.2.2 and 9.2.3 have elements that could fundamentally alter the way the JFTC will enforce private monopolization in the future. In this final section, we discuss two primary challenges that the JFTC may have to tackle. The first issue is related to the expanding coverage of surcharge payment orders. Since surcharges now have been introduced for unfair trade practices, the agency may have to reconsider its enforcement practices pertaining to such violations and instances of private monopolization. As figure 9.1 shows, the JFTC has enforced the regulations on unfair trade practices far more frequently than those pertaining to private monopolization. This is presumably because it has been easier to establish violations of unfair trade practices (as explained in sections 9.1 and 9.3.1), and at least before the 2009 amendment was enacted, the regulations of unfair trade practices were expected to elicit less opposition from the parties involved because it did not come with surcharges (nor criminal penalties). The latest such example is the DeNA case (2011) on social games, by which more than one person can play in the game through mobile networks. Social games are a booming market, whose size has tripled over the year to over 130 trillion JPY, equivalent to 1.33 billion USD. DeNa Co., Ltd. (hereafter DeNA) kept a group of social game developers from providing their games through DeNA’s primary competitor, GREE, by disconnecting the website links of the games the developers provided through DeNA’s platform (called “Mobage-Town”), if the developers provided the games through GREE. The JFTC did not establish the extent to which DeNA’s conduct affected market competition. However, since the group of social game developers, who were forced by DeNA not to provide GREE, accounted for a major share of the market, the JFTC found that DeNA’s conduct interfered with GREE’s business activities, in violation of Paragraph 14 of the designation of unfair trade practices. The JFTC issued a cease-and-desist order without surcharges payment orders (as was stipulated under the violation of Paragraph 14 of the unfair trade practices), while it could have ordered payments for exclusionary private monopolization.

35 

36 

JFTC, 6 August 1997; 44 SHINKETSUSYU 238. http://www.jftc.go.jp/en/pressreleases/uploads/2007-Sep-28.pdf.

210   Ping Lin and Hiroshi Ohashi As surcharges have been implemented against some exclusionary conduct of unfair trade practices, however, the JFTC has to consider how “a tendency to impede fair competition” differs from “a substantial restraint of competition.” There is a view expressed in both legal and economic circles that unfair trade practices should now require an adverse effect at a level similar to that indicating a substantial restraint of trade (Negishi and Funada, 2011). Although some legal scholars argue that the JFTC has already set the same standard for both private monopolization and unfair trade practices, this is not the case in injunction relief under Article 24, as discussed in section 9.2.3.2, in that the current AMA limits injunction relief only to cases involving unfair trade practices. The JFTC will have to clarify how private monopolization differs from several types of unfair trade practices. Perhaps it is a good occasion for the JFTC to adopt a single standard under private monopolization, rather than the current “double” standards under the two types of regulations. A second point is in need of more economic and econometric analysis in the enforcement procedure. While economic and econometric evidence has been submitted in some merger investigations, it plays virtually no role in private monopolization cases. It is, however, worthwhile to note that the Competition Policy Research Center (CPRC) established inside the Economic Research Office of the JFTC in 2002 offers a platform for academic researchers in both economics and legal fields to jointly conduct economic and econometric research on topics not directly linked to actual cases.37 It remains to be seen whether the CPRC’s research activities lead the JFTC to employ economic analysis more prominently in its decision to bring cases and in the manner in which the cases are argued. In the meantime, according to Global Competition Review (2012), the JFTC remains a largely law-dominated organization; a mere 5 percent of its staff are economists, including four economics Ph.D.  holders, among 748 competition staff members. To bring Japan closer to the goal of the AMA cited in the introduction to this section, it appears to be essential to invite private parties and the courts to be more actively involved in the enforcement process—essentially to introduce “competition” into the enforcement procedure. Introducing competition in Japan’s antitrust enforcement would further improve transparency and accountability, increase the opportunity to introduce rigorous economic analysis by private and academic economic consultants into AMA cases, and eventually ensure the further penetration of economics to levels hopefully equivalent to those of the United States and Europe.

9.3. China China’s AML was passed on 30 August 2007 and took effect on 1 August 2008, nearly 14  years after it was first proposed. The AML was enacted to “prevent and prohibit 37  The CPRC publishes working papers on topics such as unjust low bids under public procurement (CPRC, 2012a), and margin squeezes in network industries (CPRC, 2012b), both of which are available in Japanese.

Treatments of Monopolization in Japan and China   211

monopolistic conduct, protect fair competition, improve efficiency of operation, safeguard consumer and public interests, and promote the healthy development of the socialist market economy” (Article 1). Article 17 of the AML prohibits selling at unfairly high prices and buying at unfairly low prices and the following business practices without valid reasons:  predatory pricing, refusal to deal, exclusive dealing, tying, price discrimination, and other abusive conduct as might be recognized by the relevant antimonopoly enforcement authorities. This section describes how the AML treats the abuse of a market dominant position (or monopolization).38 It focuses on the economic reasoning/principles that are adopted or reflected in the AML and its supplementary guidelines and regulations for determining abuse violations of the law.

9.3.1.  Legal Provisions and Enforcement: An Overview 9.3.1.1.  Legal Provisions Article 17 of the AML provides that an undertaking with a dominant market position shall not abuse its dominant market position to carry out the following acts: (1) selling commodities at unfairly high prices or buying commodities at unfairly low prices; (2) selling products at prices below cost without valid reasons; (3) refusing to deal with a trading party without valid reasons; (4) requiring a trading party to trade exclusively with itself or trade exclusively with a designated undertaking(s) without valid reasons; (5) tying products or imposing unreasonable trading conditions at the time of trading without valid reasons; (6) applying dissimilar prices or other transaction terms to counterparties with equal standing without valid reasons; and (7) other conduct determined as abuse of a dominant position by the relevant antimonopoly authorities under the State Council. The AML applies to all “undertakings,” defined as any natural or legal person, or any other organization, that produces or deals in goods or provides services. If a conduct eliminates or restricts competition within the territory of the People’s Republic of China (Hong Kong, Macau and Taiwan are excluded), it can constitute an infringement of the AML even if it takes place outside the territory of the People’s Republic of China. The AML establishes a two-tier enforcement structure under the State Council. The Anti-Monopoly Commission, set up by the AML, is responsible for promulgating guidelines and coordinates the work of three antimonopoly enforcement authorities. In particular, the Price Supervision and Anti-Monopoly Bureau of the National Development and Reform Commission (NDRC) is responsible for enforcement of the provisions on cartels involving pricing decisions and price-related abuse of a dominant position; the Anti-Monopoly and Unfair Competition Enforcement Bureau of the State Administration for Industry and Commerce (SAIC) for enforcement with respect to non-price-related abuse of a dominant position and monopoly agreements involving 38 

For an overview of the AML, see, e.g., Fels (2012).

212   Ping Lin and Hiroshi Ohashi non-price coordination, and the Anti-Monopoly Bureau of the Ministry of Commerce (MOFCOM) for enforcement of the merger control provisions. The Anti-Monopoly Commission reports directly to the State Council.39 In the area of abuse of a dominant position, NDRC and SAIC may carry out investigations by following a complaint, taking a case transferred from another authority, or acting on their own initiative. They may also delegate enforcement functions to relevant divisions at the provincial and municipal level where the conduct occurs within a particular administrative region. Pursuant to Article 53 of the AML, any undertaking that wishes to challenge a decision by the antimonopoly authorities may choose whether to launch an administrative appeal or seek judicial redress.40 The general provisions of Article 17 of the AML are supplemented by the Regulation on Anti-Price Monopoly and the Regulation on Prohibiting the Abuse of a Market Dominant Position, issued by NDRC and SAIC respectively in December 2010.41 The two regulations (NDRC APM Regulation and SAIC AMD Regulation respectively) provide more specific guidance on the provisions of Article 17. The presentation below is primarily drawn from the AML provisions and the two sets of supplementing regulation, as well as some (albeit very few up till now) rulings by the Chinese courts.

9.3.1.2.  Private Action The Intellectual Property Division of the Chinese courts is responsible for handling private litigations in the area of antitrust.42 According to data released by the Supreme People’s Court, courts had accepted 61 civil AML cases as of the end of 2011, most of which are abuse-of-dominance cases. However, most of these cases were dismissed by the court on the ground that the plaintiff failed to provide sufficient evidence to show that the defendants possessed dominance in the relevant market.43 39  The new competition enforcers have rather small staffs, with NDRC having around 20 officials, SAIC 10 officials, and MOFCOM around 30 staff. 40  According to information disclosed by NDRC and SAIC officials at the Asia Competition Association 2012 Annual Conference, which was held in Beijing on 21 October 2012, the SAIC had authorized Jiangsu Province and other provinces and cities to investigate 16 suspected cases so far, one of which was a suspected abuse of dominance and 15 others were suspected non-price cartel agreements. NDRC has directly investigated 15 cases involving price-related monopoly conduct (including cartel cases) since 2011. See The China Competition Bulletin, September–October 2012 (available at http://​ www.anzsog.edu.au/research/publications/the-china-competition-bulletin). There were no published cases by either of these two competition authorities during 2008 and 2009 and one case by NDRC in 2010 (Norton Rose, 2012). Thus, enforcement efforts by the two competition authorities increased substantially during 2011, though the number of investigated cases is still very small. 41  The regulations are available (in Chinese) at http://jjs.ndrc.gov.cn/zcfg/t20110104_389399.htm and http://www.saic.gov.cn/fldyfbzdjz/zcfg/zcfg/201101/t20110107_103379.html, respectively. 42  The Supreme People’s Court of China designated its IP divisions the responsibility because of the fact that Chinese courts generally have not developed an expertise in complex economic analysis. 43  Press conference of the Supreme People’s Court, 8 May 2012. See http://www.court.gov.cn/xwzx/​ xwfbh/twzb/201205/t20120508_176702.htm (in Chinese). For example, in Renren v. Baidu, a widely publicized case, the defendant submitted estimates of Baidu’s market share in the search engine service market in China as between 65 percent and 70 percent based on business periodicals. The Court considered the evidence insufficient, as the underlying data and methodology used to calculated the

Treatments of Monopolization in Japan and China   213

According to the Judicial Interpretation on the Application of Laws to Anti-Monopoly Private Actions (Judicial Interpretation) issued by the Supreme People’s Court of China on 8 May 2012, parties harmed directed or indirectly by anticompetitive conduct have the standing to sue. Both stand-alone and follow-on cases are allowed. Regarding the burden of proof, Article 9 of the Judicial Interpretation provides that the plaintiffs shall be responsible for identifying the relevant market within which the alleged abusive conduct has taken place, and for proving that the defendant possesses a dominant market position and that the alleged conduct falls within the categories prohibited by Article 17 of the AML. However, Article 9 provides that for cases involving public utilities or other undertakings with monopolistic licenses issued by the state, the court may determine that the defendant has the dominant position in the relevant market, based on specific circumstances of the market structure and competition landscape. It is noteworthy that the Juridical Interpretation (Article 13) provides that the parties may apply to the court to appoint economic or trade experts to provide explanations on technical issues. These provisions are considered responses to the fact that the plaintiffs had a rather low success rate in abuse-of-dominance cases during the first three years of AML private enforcement mainly because of the difficulties of proving that the defendant holds a dominant market position and has abused such a position.

9.3.1.3.  Legal Liability under the AML The following penalties for abuse of a dominant market position are available under the AML: (a) fines of up to 10 percent of the total turnover in the preceding year;44 (b) confiscation of illegal income; and (c) the invalidation of agreements concluded in violation of the law, and cease-and-desist orders in respect of abuse of dominant position. There are no punitive damages or criminal penalties under the AML.

9.3.1.4.  Other Relevant Laws The Anti-Unfair Competition Law of China enacted in 1993 was China’s first competition law. While mostly aiming at consumer protection and the prevention of fraudulent commercial practices (such as infringement of trademarks), part of the law is related to preventing abusive behavior of dominant firms. It particular, it prohibits exclusive dealing by public utilities or other enterprises with legal monopoly status (Article 6), predatory pricing (Article 11), and tie-in sales and bundling (Article 12). For a violation of Article 6 a fine may be imposed of between RMB 50,000 and RMB 200,000 (approximately US$79,400 to US$317,500), as well as confiscation of between 100 percent and 300 percent of the illegal revenues (Article 23). Enforced by the NDRC, the 1997 Price Law was enacted to regulate the pricing behavior of both firms and the government.45 Relevant to conduct that constitutes an abuse market shares were not provided. See Tong (2010) for a detailed description and analysis of the case by Judge Tong Shu of the IP Tribunal of the Beijing No. 1 Intermediate People’s Court. 44 

It is not clear whether the fine is based on total turnover or turnover in the relevant market. Pricing behavior of the government includes the government-guided prices and the government-set prices, as explained in Article 3 of the Price Law. 45 

214   Ping Lin and Hiroshi Ohashi of a dominant market position, Article 14 prohibits any firm from selling products at a below-cost price in an attempt to hurt competitors, spreading rumors of price hikes, attracting business through deceptive pricing strategies, and price discrimination. The Price Law allows for fines up to five times the illegal gains.

9.3.2.  Determination of Market Dominance and Abuse under the AML 9.3.2.1.  Definition of Market Dominance Article 17 of the AML defines a dominant market position as “a market position held by an undertaking having the capacity to control the price, quantity or other trading conditions in relevant market, or to hinder or affect any other undertaking to enter the relevant market.” The language of Article 17 is rather general, particularly about the meaning of “other trading conditions in the relevant market.” Article of the 2011 SAIC AMD Regulation clarifies that (1) “other trading conditions” refer to factors other than price and quantity that can have a substantial impact on market transactions, including product quality, payment terms, mode of delivery, and after-sale services. Regarding the meaning of the ability to hinder or affect market entry, the Regulation states that it includes the ability to exclude, delay, or make more difficult for other undertakings to enter or expand into the relevant market within a reasonable period of time. This includes the ability of the undertaking to raise the cost of entry for its competitors so that they are unable to compete effectively in the relevant market postentry.

9.3.2.2.  Determination of Dominant Market Position In determining whether an undertaking possesses a dominant market position, the enforcement authorizes will consider the following factors, as stipulated in Article 18 of the AML: (1) the market share of an undertaking and the competition condition in the relevant market; (2) the capacity of an undertaking to control the product market or the raw material procurement market; (3) the financial resources and technical capabilities of the undertaking; (4) the degree of dependence of other undertakings on the undertaking in transactions; (5) the level of difficulty for other undertakings to enter the relevant market; and (6) other factors relating to the determination whether the subject undertaking has a dominant market position. The 2011 SAIC AMD Regulation provides further guidance on the above factors. Specifically, the market share of an undertaking is calculated both in value and in volume, and in reviewing an undertaking’s market share SAIC will also take into account the state of development of the market, product differentiation, and potential competition, among other factors. When reviewing the financial and technical resources of an undertaking, SAIC looks at the undertaking’s assets, its financial position and ability to raise funds, its R&D capabilities, and any intellectual property rights it may have. SAIC also consider the scale of transactions between other relevant undertakings and

Treatments of Monopolization in Japan and China   215

the dominant operator concerned, the duration of such relationships, and the degree of difficulty of those undertakings switching to other trading parties. When determining the degree of difficulty of entry by potential entrants, SAIC considers such factors as market entry regulations, the essential facility conditions, distribution channels, capital and technological requirements, and the costs of entry (Article 10). While the above guidance is rather detailed and broadly in line with international standards, there have not been any published cases by SAIC up till now concerning abuse of dominance.

9.3.2.3.  Presumption of Market Dominance Article 19 of the AML provides that where an undertaking exhibits any of the following circumstances, it may be presumed to have a dominant market position: (1) the relevant market share of an undertaking accounts for one-half or more in the relevant market; (2) the joint relevant market share of two undertakings accounts for two-thirds or more; or (3) the joint relevant market share of three undertakings accounts for three-fourths or more. An undertaking with a market share of less than one-tenth shall not be presumed as having a dominant market position even if it falls within the scope of the second or third item. If an undertaking presumed to have a dominant market position can present evidence showing otherwise, it shall not be deemed as having a dominant market position. Market share is not the only presumption available to help establish dominance. According to the Judicial Interpretation issued by the People’s Supreme Court of China on 8 May 2012, a defendant may be presumed dominant when it is a public utility company, or when it is conferred with exclusive license by law or administrative rules, or when the defendant’s trade counterparts are highly reliant on its supply due to a lack of effective competition in the relevant market.

9.3.2.4.  Joint Dominance Although not stated explicitly, it appears from Article 19 of the AML that two or more independent undertakings can be found to hold joint dominant position in the relevant market, even if they are not actively coordinating their conduct in the market. The concept of joint dominance was apparently used in a recent NDRC investigation of China Telecom and China Unicom, two state-owned firms in the telecommunications industry, which were alleged to have engaged in abuse of dominance in the broadband access market. Section 9.3.3 contains more information about this case.

9.3.2.5.  Definition of Relevant Market On 24 May 2009 China’s Anti-Monopoly Commission announced its Guidelines Concerning the Definition of Relevant Markets.46 The Guidelines recognize that “any competitive behavior (including any behavior that has resulted or is likely to result in eliminating or restricting competition) occurs within a particular market scope. The relevant market definition is to define the market scope within which the business operators compete with each other. In the work of anti-monopoly law enforcement, such 46 See http://www.gov.cn/zwhd/2009-07/07/content_1355288.htm (in Chinese).

216   Ping Lin and Hiroshi Ohashi as prohibiting monopoly agreements among business operators, prohibiting the abuse of dominant market positions and controlling the concentrations of business operators that has resulted or are likely to result in eliminating and restricting competition, defining the relevant market may be involved” (Article 2). The Guidelines further make it clear that defining the relevant market “plays an important role in key issues such as recognizing competitors and potential competitors, determining the market share of the business operators and the degree of market concentration, deciding the market position of the business operators, analyzing the impact of the business operators’ behaviors on market competition, judging whether such behaviors are illegal or not and determining the legal liabilities they need to bear if their behaviors are illegal. As a result, defining the relevant market is usually the starting point of conducting an analysis on competitive behavior and an important step of antimonopoly law enforcement” (Article 2). In line with international standards, the Chinese guidelines define the relevant product and geographical markets based on demand substitution and supply substitution factors, and take into account factors giving rise to a temporal market. They also recognize the effectiveness of the hypothetical monopolist test (the SSNIP test) in ascertaining the boundaries of the market, with a price increase in the range of 5 percent-10 percent and for normally one year as thresholds for testing. One difficulty in defining the relevant (geographical) market in abuse-ofmarket-dominance cases in China was encountered in the recent case against Tencent by Qihoo, the details of which is contained in section 9.3.4.6.1. The plaintiff and the defendant disagree on what constitutes the relevant product market. In particular, the plaintiff is of the view that Tencent’s instant message software and relevant services should be defined as the relevant product market and that the market should be defined as the domestic market of China because of the unique Chinese language and cultural background of the instant message services. Tencent on the other hand argues that because of the openness and interoperability of the Internet, the users of the instant messaging are not limited to those in the Mainland China. The case is still pending at the time of this writing. More details of this case are contained in section 9.3.4.6.1. An interesting question of whether “free” services can constitute a relevant market arose in Renren v. Baidu (2009), in which the defendant argued that its Internet search services are free to the general Internet users and thus are not regulated by the Antimonopoly Law. The court disagreed and reasoned that free provision of some products or services is often closed related to its charged provision of some other products or services. The court ruled that “free” services can constitute a relevant market in the sense of the AML (see Tong 2010). This ruling is consistent with insight from the modern economics of networks on cross-subsidization of services under optimal pricing in two-sided markets (see, e.g., Laffont et al., 2003).

9.3.2.6.  Rule-of-Reason Approach Consistent with international best practice, whether a conduct constitutes an abuse of market dominance is determined by the rule of reason, rather than the per se rule

Treatments of Monopolization in Japan and China   217

under the AML. This approach was made clear by the SAIC in its press conference following its announcement of its AMD Regulation.47 In fact, the rule of reason had been also explicitly adopted by the court in the early high-profile private action cases. In Renren v. Baidu, the court explicitly stated that a dominant market position or economies of scale per se do not violate the AML. The issue for antitrust scrutiny is abuse of a dominant position, or anticompetitive behavior of dominant firms.48 In the ongoing Qihoo v. Tencent, Judge X. J. Zhang of the Guangdong Province People’s Court summarized the focuses of this case, which are to determine (1) what constitutes the relevant market; (2) whether Tencent possesses a dominant position in this market; (3) if it has a dominant position, whether or not Tencent abused its dominant position; and (4) what would be the proper penalty if Tencent is found to have violated the abuse provisions of the AML.49 In the next section, we focus on the type of abusive conduct that falls within the scope of responsibility of NDRC, namely abusive conduct related to pricing behaviors.

9.3.3.  Price-Related Abuse of Dominance Among the types of abusive conduct listed in Article 17 of the AML, three concern price decisions, namely setting unfairly high selling prices or unfairly low buying prices, and predatory pricing, without valid reasons.

9.3.3.1.  Selling at an “Unfairly High Price” or Buying at an “Unfairly Low Price” Similar to the EU competition law, China’s AML prohibits dominant firms from “selling at unfairly high prices or buying at unfairly low prices” (Article 17). What constitutes “unfair” is not defined in the AML. Article 11 of NDRC APM Regulation lists a range of factors that NDRC will consider when determining whether a price is excessively high or low. These include whether the price is significantly higher or lower than the price at which other operators buy or sell the same type of goods; whether the increase in selling price or decrease in purchase price exceeds the normal range when production costs are generally stable; and when production costs vary, whether the change in price is significantly larger than the change in costs. No definition of costs is provided. One decision of the NDRC may be indicative of what may be considered an “unfairly large increase in price.” In March 2011, in its informal investigation of the pricing policy of the industry leader in the instant noodle market, Master Kang, the NDRC found the planned price increase by Master Kang in mid-March 2011 for its canned instant noodle “overly high” and “improper.” In reaching this conclusion, NDRC specifically stated that Master Kang’s market share in canned instant noodles was 73.6 percent in 2010; it thus had a dominant market position. The NDRC referred to the following figures: The price increase would raise Master Kang’s net profit margin of this product line to 14.3 percent, 47 See http://www.saic.gov.cn/fldyfbzdjz/gzdt/201101/t20110107_103382.html (in Chinese). 48  49 

See Liu and Qiao (2012) for a more detailed description of the case. See, e.g., the report on China Webmaster: http://www.chinaz.com/news/2012/0418/246478.shtml.

218   Ping Lin and Hiroshi Ohashi a 4.2 percent jump compared with its January 2011 level and an 1.1 percent jump compared with its average level in 2010.50

9.3.3.2.  Predatory Pricing The AML specifies predatory pricing as one of the six types of conduct that are explicitly prohibited for dominant firms. In particular, Article 17 (2) AML prohibits an undertaking in a dominant market position from selling products at prices below cost without valid reasons. Article 12 of NDRC APM Regulation lists some of the legitimate reasons for pricing below cost: undertakings may sell below cost if they are having a closing-down sale, selling overstocked or perishable goods, or promoting new products. As under the EU competition law, the AML states that only firms in a dominant market position are prohibited from predatory pricing. However, the predatory pricing provisions in early competition laws of China such as the 1993 Anti-Unfair Competition Law and the 1998 Price Law do not specifically require an undertaking to be dominant for its below-cost pricing to be predatory. In Chuanglian v. Huimeng, for example, the Haidian District Court in Beijing found the defendant—an Internet portal—to have engaged in illegal below-cost pricing of advertising space, although there was no evidence that it occupied a dominant market position. The AML does not define the term “cost” in relation to predatory pricing. China’s Price Law suggests that costs include both production and operation costs, as confirmed in Article 4 of the Regulation on the Prevention of Below Cost Dumping Conduct (1999), Order No. 2 by the National Development and Plan Commission.51 It appears that the cost benchmark set out in the Regulation is total unit cost. The Chinese competition laws and regulations and any court rulings have not made any particular reference to the notions of Areeda-Turner’s average variable cost (AVC) test or Baumol’s proposal of use of “average avoidable cost” (AAC). Moreover, neither the AML nor NDRC APM Regulation has any specific requirements regarding a showing of recoupment, consumer harm, or, in fact, the rival’s exit. Whether there such requirements are needed in future court and enforcement decisions remains to be seen.

9.3.3.3.  Discriminatory Practices Article 17 of the AML prohibits discriminatory treatment without valid reasons by a dominant undertaking of trading counterparties in prices or other trading terms that constitutes abuse of its dominant position. It covers but is not limited to price discrimination. Article 5 of SAIC AMD Regulation provides the following list of (mostly) non-price-related discriminatory practices by an undertaking towards trading 50  As released at the NDRC press conference held on 6 May 2011 on its formal investigation of the Unilever price-hikes case. See 国家发展改革委有关负责人就查处联合利华(中国)有限公司散布 涨价信息扰乱市场秩序的有关问题答记者问; http://www.ndrc.gov.cn/xwfb/t20110506_410543.htm. Advised by the NDRC, Master Kang did not implement the price increase as planned. (It did, however, raise its price in August 2011 by about 10 percent.) 51 See http://jjs.ndrc.gov.cn/zcfg/t20070727_150273.htm (in Chinese).

Treatments of Monopolization in Japan and China   219

counterparties with equivalent tradition conditions without valid reasons:52 (a) imposing different transaction quantities, types of goods, or quality levels; (b) imposing different preferential conditions, such as quantity discounts; (c) imposing different payment conditions or delivery methods; or (d) imposing different after-sales service conditions, such as warranty terms and term, repair conditions and times, provision of spare parts and components, and technical guidance. No mention, however, is made of what might constitute “valid reasons” for such discriminatory practices. There have some price discrimination cases since the AML took effect. In the recent high-profit investigation of China Telecom and China Unicom, NDRC clearly adopted the theory of price squeeze, though it did not use this term.

9.3.3.3.1.  NDRC Investigation of China Telcom and China Unicom (2011) On 9 November 2011, in a News 30’s Program of China Central Television’s (CCTV), Ms. Li Qing, deputy director of the Price Supervision and Anti-Monopoly Bureau of the NDRC, publicly acknowledged that NDRC had been conducting an investigation into China Telecom and China Unicom since the first half of 2011. Implicitly referring to Article 19 of the AML, Deputy Director Li released the preliminary findings of that investigation: with a combined share of more than two-thirds of the “broadband access” market in China, China Telecom and China Unicom held a dominant position in the broadband access market. She also disclosed that, by charging higher prices for access to their broadband network to direct competitors than to noncompetitors, China Telecom and China Unicom were guilty of price discrimination.53 The investigation is still ongoing at the time of this writing. Price discrimination in final product markets has also been brought to investigation in China. In Li Fangping v. China Netcom (Group) Co., Ltd., filed on 1 August 2008, the very day that the AML took effect, Beijing Netcom, a telecommunications service provider and China Netcom’s Beijing branch, was accused of charging different rates to consumers with a Beijing residence permit (known as a “hukou”) and those without. The case was accepted by the Beijing Chaoyang District People’s Court on 24 September 2008, but later referred to Beijing No. 2 Intermediate People’s Court. The trial court rejected Li’s requests, and Li appealed to the Beijing High People’s Court. The High Court upheld the trial court decision. This case attracted wide attention as the first major judicial antimonopoly case in China. It has been instrumental in establishing principles concerning rules of evidence and market definition in antimonopoly litigation. First, Beijing Netcom argued that its discriminatory pricing was reasonable because under its pricing policy, only those 52 

The NDRC APM Regulation also contains an article prohibiting discriminatory treatments in price by a dominant undertaking of trading parties with the same transaction conditions without valid reasons (Article 16). However, the NDRC Regulation does not provide additional information; it just covers the price-related abusive discriminatory treatment as provided in Article 17 (6) of the AML. 53  The NDRC has yet to complete its investigation of the alleged price discrimination case. The CCTV program exclusive interview with Deputy Director Li can be found in http://v.qq.com/​ cover/9/9ux3zxaeki17i7s.html?vid=8Effma9lZ7V (in Chinese).

220   Ping Lin and Hiroshi Ohashi non-Beijing resident customers who did not have real estate in Beijing were charged a higher rate. Without such a restriction, it would face “business risks.” The court accepted this justification by the defendant.54 Second, the principle of allocating the burden of proof in antimonopoly litigation was for the first time specified by a court decision. The Beijing High People’s Court held that in civil actions involving the abuse of market dominance, the plaintiff was responsible for providing evidence concerning the definition of the relevant market, and proving that the defendant held a dominant position in said market, that it had abused that dominance, and that such abuse had caused actual losses to the plaintiff. Third, the court made a direct determination on the issue of what was the relevant market. In the final judgment, the Beijing High People’s Court supported the trial court’s decision that fixed telephone, Little Smart (xiaolingtong) phones, and mobile phones were closely substitutable, as were ADSL and wireless access to the Internet. Previously, such determination was generally made by a competition authority in merger assessments.55

9.3.3.3.2.  Prohibition of Price Discrimination under the Price Law China’s Price Law, which took effect in 1998 and is still in effect together with the AML, also prohibits price discrimination. In particular, Article 14 (5) of the Price Law prohibits “price discrimination against other undertakings with the same trading conditions.” In comparison, the AML prohibits undertakings from abusing their dominant market position by engaging in “discriminatory treatment of trading counterparties with the same trading conditions in terms of prices and other trading conditions, without justifiable reasons” (Article 17 (6)). There are three important differences between the treatments of price discrimination under the AML and under the Price Law. Firstly, the Price Law prohibits price discrimination in the intermediate product markets only, as is clear in the wording of Article 14 (5). On the other hand, “trading counterparties” under the AML include final consumers as well business operators. In fact, the plaintiffs in AML cases of Li Fangping v. China Netcom and Zhou v. China Mobile were both individual consumers. Second, the Price Law does not require that the undertakings engaging in price discrimination hold a dominant market position; any undertakings are banned from price discrimination. 54 

See, e.g., http://www.chinalawinsight.com/2010/09/articles/corporate/antitrust-competition/​ aeaeaeaaceaecaaaeeaaaeeaea/. 55  In a similar case, Zhou vs. China Mobile, an activist lawyer, Mr. Zhou Ze, filed an antitrust suit with a Beijing district court, which later transferred the case to Beijing No.2 Intermediate People’s Court on 4 March 2009, alleging that China Mobile abused its dominant position in China’s mobile telecommunications market by engaging in unlawful price discrimination, in violation of Article 17(6) of the AML. Specifically, Zhou alleged that China Mobile discriminatorily charged him a “monthly rental fee” that was not charged to other customers. He sought damages and an injunction to stop the discriminatory charges. On 23 October 2009 the Beijing Dongcheng District People’s Court announced a settlement of the case in which China Mobile allowed Zhou to change his plan to one without a “monthly rental fee” and to keep his same mobile phone number. As compensation for Zhou’s “input to the pricing issues,” China Mobile rewarded him RMB 1,000 (about US$146). Zhou withdrew the case.

Treatments of Monopolization in Japan and China   221

Third, the Price Law adopts the per se illegal rule toward price discrimination, whereas the AML uses the rule of reason. The last two differences were confirmed by Xu Kunlin, director general of the Price Supervision and Anti-Monopoly Bureau of the NDRC in his interview by the Antitrustsource in 2011.56 A possible explanation for the very different treatments of price discrimination under the two laws is that China puts more emphasis on fairness during the early stage of its economic development. In 1998, it was stated in the Interpretation of the Price Law issued by the State Reform and Planning Committee (the predecessor of NDRC) that price discrimination against downstream undertakings leads to unfair competition in the downstream market and is thus prohibited.

9.3.4.  Treatment of Non-Price-Related Abuse of Dominance The SAIC ADM Regulation provides guidance as to what constitutes abuse of dominance in the form of non-price business practices, in relation to Article 17 of the AML. Specifically, Article 4 of the Regulation lists different types of refusal-to-deal behavior by dominant firms. Articles 5 to 7 specify in detail the abusive practices of dominant firms in terms of exclusive dealing; tying or restrictive sales; and discriminatory trading, respectively. However, for refusal to deal and exclusive dealing, the NDRC APM Regulation may also be applicable if the abuse is disguised in price contracts of the dominant firm with its suppliers or customers.

9.3.4.1.  Refusal to Deal Article 4 of the SAIC AMD Regulation provides that an undertaking with a dominant market position shall be prohibited from refusing, without legitimate reason, to transact with the counterparties to the transaction by means of any of the following: (1) reducing the volume of an existing transaction with the counterparties to the transaction; (2) delaying in executing or discontinuing an existing transaction with the counterparties to the transaction; (3) refusing to engage in a new transaction with the counterparties to the transaction; (4) imposing restrictive conditions that make it difficult for the counterparties to the transaction to continue to trade; (5) refusing the use of any essential facility under reasonable conditions by the counterparties to the transaction in the course of production and operation activities. There is also so-called disguised refusal to deal. Article 13 of the NDRC APM Regulation prohibits undertakings from setting a very high selling price that, in effect, amounts to a refusal to deal with other trading parties without legitimate reasons. 56 

With respect to regulation approaches, Mr. Xu stated that “the Price Law emphasizes more the examination of whether the conduct itself stays in compliance with provisions of the law; the AML emphasizes the examination of the impact of such conduct on competition in the market.” See F. H. Deng, S. Harris Jr., and Y. Zhang (2011), “Interview with Xu Kunlin, Director General of the Department of Price Supervision Under the National Development and Reform Commission of People’s Republic of China,” available at http://www.americanbar.org/content/dam/aba/migrated/2011_build/antitrust_law/​ feb11_fullsource.authcheckdam.pdf.

222   Ping Lin and Hiroshi Ohashi According to the NDRC APM Regulation, an excessive high selling price by an undertaking may be legitimate if its trading party has a extremely poor credit record, or is in financial distress so that continuous dealing may cause financial risks to the undertaking, or if the party is able to obtain similar goods at a reasonable price from a third company. There have not been any formal cases against refusal to deal by either the NDRC or the SAIC. However, it is important to note that the SAIC AMD Regulation explicitly refers to the issue of essential facility.57 This provision may potentially be applied in technological licensing (or refusal to license) cases. Although there have not been formal such cases, the fact that the SAIC AMD Regulation contains a detailed paragraph particularly pertaining to the issue of essential facility may be an indication that the issue of refusal to deal is of substantial importance.

9.3.4.2.  Exclusive Dealing Article 6 of the SAIC AMD Regulation provides that, without legitimate reasons, an undertaking with a dominant market position shall not restrict a trading partner by requiring it to deal with the dominant undertaking or with other designated undertakings; or requiring it not to deal with its competitors. The SAIC Regulation does not provide additional information as to what constitutes legitimate reasons for exclusive dealing. However, Article 14 of the NDRC APM Regulation clarifies that, without justifiable reasons, a dominant undertaking shall not use price discounts or other means to require its trading partners to exclusively deal with it or with the undertaking(s) it designates. The NDRC APM Regulation provides a list of circumstances where imposing an exclusivity requirement may be justifiable by the dominant undertaking. This would be the case where the exclusivity is required to (1) protect product quality and safety; or (2) to maintain brand image or improve services; or (3) to lead to significant cost-reductions or efficiencies gains, the benefits of which must be partially passed on to consumers. From the language used in the Regulation, the listed cases where exclusive dealing can be justified seem consistent with modern industrial economics. For instance, exclusive dealings may enhance efficiency when used to improve and protect product quality, or as a response to free riding among manufacturers in the upstream market, or to facilitate R&D investments by upstream and downstream firms. Whether the above modern economic reasoning is the rationale behind the Regulation or whether or not it will be adopted in real exclusive dealing cases remains to be seen. One thing that is interesting to note is that the NDRC Regulation specifically mentions the type of exclusive dealing using price discounts, which is presumably often 57  Article 4 states that when determining whether there exists a refusal to deal in essential facilities, SAIC shall consider such factors as the feasibility of otherwise investing in and constructing or otherwise developing and building such facilities, the degree of dependence of the counterparties to the transaction on such facilities in order effectively to conduct their production and operation activities, the feasibility of undertaking to provide such facilities and the impact of providing such facilities on its own production and operation activities, etc.

Treatments of Monopolization in Japan and China   223

used by an upstream supplier to induce its downstream buyer(s) not to purchase from other supplier(s). However, the Regulation does not explicitly mention the exclusive dealing arrangements that might be initiated by a downstream buyer that agrees to pay a higher price for the good concerned in exchange for the supplier’s agreement to not supply other downstream rivals, although the wording in the Regulation (“by the use of price discounts or other means”) should capture such arrangements initiated by downstream firms. What leads to the asymmetric wording with regard to the above two types of exclusive dealing arrangements in the Regulation is not known. The downstream-initiated exclusive dealing arrangement is at the core of the recent NDRC decision against two drug companies in Shandong Province. On 14 November 2011, the NDRC announced fines of nearly RMB 7  million (US$1.1 million) against two private pharmaceutical companies in Shandong Province for various abuses of dominant market position, including an exclusive supply contract and setting excessively retail prices. According to the public announcement issued on NDRC website on 14 November, the offending companies were two distributors of the active ingredients used in the manufacture of Reserpine, an antihypertensive drug that is commonly used for the control of high blood pressure.58 NDRC investigation found that two companies authorized to manufacture the active ingredient used in Reserpine in China, promethazine hydrochloride, Donggang Hongda Pharmaceutical Co., Ltd. (“Donggang Hongda”) and Dandong Yichuang Yaoye Co., Ltd. (“Dandong Yichuang”), each entered into an exclusive distribution agreement with Shandong Weifang Shuntong Pharmaceuticals Co., Ltd. (“Shuntong”) and Weifang Huaxin Pharmaceuticals & Trading Co., Ltd. (“Huaxin”) under identical terms on 9 June 2011. Under the terms of these agreements, the two manufacturers of promethazine hydrochloride on the Chinese market were prohibited from selling the product to third parties without first obtaining the consent of Shuntong and Huaxin. Although the NDRC announcement did not contain information about price premiums contained in such exclusive dealing agreement, it was reported that the two distributors agreed to raise the price of promethazine hydrochloride from RBM 178 prior to the agreement to RMB 300.59 Having secured exclusive distribution rights over the active ingredient promethazine hydrochloride, Shuntong and Huaxin then raised the retail prices to their downstream manufacturers to increase their own retail price of Reserpine by more than 400 percent. The proceeds of the required price increases were then to be shared between the manufacturers (Dongang Hongda and Dandong Yichuang) and their exclusive suppliers (Shuntong and Huaxin). With more than 10 million patients taking the drug nationwide, Reserpine has been listed as an “essential” medicine in the National Essential Drugs List. Since the payment for drug is fully covered by the national social insurance policy, the decision noted that 58  The NDRC’s announcement can be found here: http://jjs.ndrc.gov.cn/fjgld/t20120306_465386.htm (in Chinese). 59  See, e.g., the report at http://www.fh598.com/article-4856892-1.html (in Chinese).

224   Ping Lin and Hiroshi Ohashi these increased costs were directly borne by society. With the combined total of the fines of more than RMB 7 million (approximately US$1.1 million), this case marks the first occasion on which NDRC has sought to impose significant fines for price-related abuse of dominance under the AML. NDRC announcement stated that Shuntong and Huaxin had abused their dominant position, thereby violating the AML and the Price Law. It appears that the NDRC’s decision was based on the grounds that Shuntong and Huaxin had controlled the supply of promethazine hydrochloride via its exclusive dealing agreement with the upstream suppliers and that they had then raised the retail price of the final product Reserpine to an excessively high level.

9.3.4.3.  Tie-in Sales SAIC AMD Regulation offers more explanations on what might constitute an illegal tie-ins. According to Article 4, an undertaking with a dominant market position shall be prohibited from tying the sale of goods or, in the course of a transaction, imposing other unreasonable transaction conditions: (a) in a manner counter to trading practice, consumer habits, and so on, or ignoring the functions of the goods, forcibly tying different goods together for sale or selling them as a bundle; (b) imposing unreasonable restrictions on the contract term, payment method, method of transportation and delivery of the goods, or the method of provision of the services, and so on; (c) imposing unreasonable restrictions on the sales territory or sales counterparties for the goods, after-sales services, and so on; or (d) imposing transaction conditions that have no connection with the subject matter of the transaction. (The NDRC Regulation does not provide any guidance on tying practices.) As can be seen, SAIC AMD Regulation specifies that both demand-side factors and supply-side factors (including the technical “functions” of the goods), as well as industry practice, are relevant in determining whether a tie-in practice is illegal, which is in line with international best practice.60 However, there has been very few tying cases in China, either public or private, so as to provide more guidance on important issues concerning how tie-in sales will be treated under the law. For example, how are relevant markets defined in tie-in cases? Specifically, how can one determine the two goods involved in tie-in sales are in two separate markets or just two components of an integrated product? What are the economic theories about the competition effects of tie-in sales that the enforcement agencies or the court might have accepted? Perhaps the only tie-in decision under the AML so far is Wuchang Salt (2010).61 On 12 August 2010, the local price bureau of the National Development and Reform Commission (NDRC) in Hubei (Hubei Price Bureau) launched an investigation into the conduct of the Wuchang branch (Wuchang Salt) of the Hubei Salt Industry Group (Hubei Salt) in response to local media reports that Wuchang Salt had tied the sale 60 

For example, see Report on Tying and Bundled Discounting, International Competition Network, 2009. This case occurred before NDRC and SAIC issued their 2011 regulations, which specify that SAIC would be responsible for handling non-price-related abuse of dominance. 61 

Treatments of Monopolization in Japan and China   225

of salt to the sale of a local brand detergent washing powder in the wholesale market. (Wuchang Salt had previously tied the sale of nutritious salt with ordinary salt.) Wholesalers then either supplied the salt and the tied products to retailers separately or as a bundle. Most retailers, when selling to end users, did not tie the sale of salt to the sale of washing powder. The Hubei Price Bureau held that the conduct of Wuchang Salt had violated the AML. First, Wuchang Salt holds a dominant position in the local wholesale market for salt. Second, the conduct of Wuchang Salt constitutes a tying (without justifiable reasons) prohibited by Article 17(5) of the AML. Third, Wuchang Salt, as an undertaking having a status as a statutory monopoly, infringed Article 7 of the AML, which requires that statutory monopolists should “conduct business in accordance with laws in an honest and trustworthy manner, impose strict self-discipline, and accept supervision from the public . . . [and] should not harm consumer interest by taking advantage of their position.”62 With respect to the economic motivation of the tying, the NDRC’s published decision stated that Hubei Price Bureau found that Wuchang Salt’s tying conduct was pursuant to a policy of “developing businesses other than salt.” The decision did not make reference to any economic theory about tying sales, although the pursuit of business in the salt market can be explained by the standard “leverage theory” of tying sales, given the fact that Wuchang Salt was the only salt wholesaler in the Wuhan city and that there were other independent sellers of washing power in the market.63 Regarding the effects of the typing, the NDRC concluded that “the sales volume, number of consumers affected and the social impact of the tying were limited.” Hubei Price Bureau issued a warning to Hubei Salt and Wuchang Salt pursuant to Article 45 of the AML. No fine was imposed.

9.3.4.4.  Abuse of Intellectual Property Rights Article 55 of China’s AML expressly prohibits abuse of IPRs that restrict or eliminate competition. To supplement the very brief Article 55, SAIC has been drafting IPR guidelines since 2009. While the official guidelines have yet to be released, consensus regarding some fundamental issues seems to have been reached. For example, the most recent Draft Guidelines for Anti-Monopoly Enforcement in the Field of Intellectual Property Rights (Draft IPR Guidelines) dated 8 August 2012, contain promising signs of alignment between the Chinese approach in this field and the current approach in the United States and the EU. For example, the Draft IPR Guidelines explicitly state that “IPR laws and antimonopoly laws share the same goals and functions, namely to promote innovation and competition, increase efficiency, and safeguard consumers interest and the public interest” (Article 2). Thus, the Chinese antimonopoly 62  The NDRC’s published decision is available at http://jjs.ndrc.gov.cn/gzdt/t20101115_380421.htm (in Chinese). 63  There are media commentaries that point out the fact that the tied product was a local brand of washing power in Hubei Province and suggested that the provincial government may have had an incentive to use its monopoly position in salt wholesale as a means to promote its local washing power brand under the tying strategy. See, e.g., http://www.tech-food.com/news/2010-8-12/n0412110.htm (in Chinese).

226   Ping Lin and Hiroshi Ohashi authorities will not take the narrower and outdated view that IPR laws, by creating legal monopolies, are in conflict with competition laws. The Draft IPR Guidelines also state that IPRs are treated in the same way as other assets of an undertaking and that an undertaking that possess IPRs will not be automatically deemed to enjoy a dominant position in the relevant market. When IPRs are an important factor that gives rise to a dominant position, the basic principles and analytical framework of the AML will be followed in determining whether the undertaking possesses a dominant position in the relevant antitrust market (Article 9). In accordance with and parallel to Article 17 of the AMIL, the Draft IPR Guidelines (Article 16) also lists six types of abuse of IPRs by undertakings with a dominant market position: (1) setting unfairly high prices in the IPR licensing contract; (2) refusal to license IPR without valid reasons; (3) exclusive dealing in the context of IPR licensing without valid reasons; (4) tying in IPR licensing without valid reasons; (5) discriminatory treatment in setting IPR licensing fees; and (6) other abusive conduct as might recognized by the relevant antimonopoly enforcement agency. Thus, all six forms of abuse of market dominance stipulated in Article 17 of the AML can potentially occur in relation to IPRs. A particularly important issue in IPR-related antitrust conduct is that of refusal to license. The Draft IPR Guidelines state that the antimonopoly authorities will not investigate a unilateral refusal to license that is unconditioned or nondiscriminative, but may investigate relevant cases of unfair and discriminative refusals to license, or cases where the IPRs concerned constitute the “essential facilities” for the would-be licensee to be able to compete effectively in the relevant market. The antimonopoly authority may also investigate refusals to license that occur without justification or as a means of enforcing other restrictive terms or tying arrangements. While not official yet, the Draft IPR Guidelines are quite detailed and broadly in line with international standards. It is expected that the official guidelines will not differ too much from the current draft, which has gone through four rounds of revision already.

9.3.4.5.  Treatment of SOEs Article 7 of the AML contains somewhat ambiguous language concerning the law’s application to state-owned enterprises. Specifically, on the one hand, it states that it “the state shall protect the lawful activity of undertakings in industries controlled by the Stated-owned enterprises and relied upon by the national economy and national security.” At the same time, Article 7 also provides that undertakings in the above-mentioned industries “shall not use their controlling position or exclusive operation rights to hurt consumers.” Although some commentators have interpreted Article 7 as potentially aiming at protecting SOEs, especially right after the passage of the AML, the official and the majority view is that the AML applies equally to all undertakings. Some of the cases investigated by the government in fact involve SOEs. For example, in Wuchang Salt, NDRC’s local bureau in Hubei province took action against state-owned companies that enjoyed exclusive operation rights for abuse of dominance in the form of tying and for

Treatments of Monopolization in Japan and China   227

their violation of Article 7. Equal treatment of SOEs and privately owned firms is further reflected in the more recent NDRC investigation of state-owned China Telecom and China Unicom for suspected abuse of dominance in the broadband interconnection market by way of setting discriminatory prices. In the area of merger control, the Ministry of Commerce has also imposed conditions on proposed mergers involving SOEs (e.g., in case General Electric/China Shenhua [2011]).

9.3.4.6.  Summary of China’s Treatments of Monopolization The provisions in China’s 2008 AML prohibiting the abuse of a market-dominant position follow very closely those under the EU competition law. NDRC APM Regulation and SAIC AMD Regulation, which went into effect in February 2011, provide useful guidance on how conduct of dominant firms might be treated as abusive under the very general language of the AML.64 Although the Chinese antimonopoly authorities have yet to demonstrate clearly how the newly issued NDRC and SAIC regulations will be applied when assessing abusive conduct, overall the economic principles and reasoning that can be seen from the regulations are broadly in line with international standards. As a unique feature of China’s AML, the three-pillar enforcement structure has raised concerns over whether coordination will be effective among the three enforcers, and between NDRC and SAIC in particular when it comes to abuse-of-dominance cases (see, e.g., Fels 2012 and Zhang 2011). While NDRC and SAIC seem to have a clear division of responsibilities, namely that NDRC is responsible for price-related abusive conduct and SAIC for non-price-abusive conduct, whether the two enforcers will follow the same economic principles when assessing abuse cases that have both price and nonprice elements (for instance, refusal to deal and exclusive dealing) may present a big challenge in future enforcement of the AML. Compared with almost no public enforcement in the area of abuse of market dominance (with the exception the NDRC investigation of price discrimination practices by China Telecom and China Unicom and the Wuchang Salt [2010] tying case), the limited number of private cases dealt with by the Chinese courts so far has had a significant “demonstration effect” in the sense that several important legal and economic issues have been clarified by the courts. For instance, the court in Renren v. Baidu (2009) ruled that possession of a dominant market position per se does not constitute a violation of the AML, thus establishing the rule-of-reason approach. It also made clear that the burden of showing possession of dominance and its abuse both rest on the shoulders of the plaintiff.65 The national coverage of the ongoing, high-profile Qihoo v. Tencent case also has the effect of educating the public how the court will go about assessing abuse-of-dominance cases. On the economic side, the judge in Renren vs Baidu rejected 64  Prior to this, the Chinese courts had handled several private action cases (e.g., Baidu vs. Renren and Li vs. China Netcom). In all these cases the courts have ruled in favor of the defendants, on the ground of lack of evidence to establish possession of position. It has been suggested that the courts were taking a cautious approach until they receive more guidance from enforcement agencies (Cleary Gottlieb, 2010). 65  The recent Judicial Interpretation by the People’s Supreme Court partly alleviates the proof burden on the plaintiffs, as presented in section 3.1.2.

228   Ping Lin and Hiroshi Ohashi the defendant’s claim that “free services” cannot constitute a part of an antitrust-relevant market, which is perfectly consistent with modern economic understanding of monopoly pricing in two-sided markets. These developments will undoubtedly continue to have a positive influence on future private (and public) enforcement of the AML provisions on abuse of dominance. The Chinese antimonopoly authorities have yet to demonstrate how they will apply modern economic reasoning in presenting/analyzing evidence, organizing views, and assessing competition impacts of alleged abuse-of-dominance conduct under the AML. They have shown signs, on the other hand, of taking the lead in private action cases, as in the ongoing Qihoo v. Tencent case. Because it is understood to be the first abuse-ofdominance case under the AML in which economics experts were engaged, we describe the case below with the focus on the types of evidence presented and the economic arguments adopted in court (which are quite detailed, relatively speaking) by both parties of the case.66

9.3.4.6.1.  Qihoo v. Tencent (2012) On 18 April 2012, the case of alleged abuse of market dominance by Tencent Technology Co. Ltd. and Shenzhen Tencent Computer System Co. Ltd. was heard publicly in the High People’s Court of Guangdong Province. In this case, the plaintiff, Beijing Qihoo Technology, accused Tencent of exclusive dealing and tying in the market of instant message services on the Internet. Both parties in this high-profile case invited in expert witnesses. The plaintiff engaged scholars from overseas, including a former official of the Office of Fair Trading of England, who submitted a report covering the relevant market in the context of the litigation, the extent of competition in the market, and the defendant’s dominance. At the court hearing, both parties followed the State Council’s 2009 Guidelines Concerning the Definition of Relevant Markets in defining relevant markets. While the details of their arguments are not available, the plaintiff utilized such notions as demand substitutability, market shares, and indirect evidence obtained from event study price correlation analysis and so on. The analyses of the plaintiff were conducted, and conclusions drawn, in the context of the SSNIP framework. The court concluded that the QQ instant messaging software of Tencent and the associated service market constituted the relevant product market, because according to its unique pricing and profit mode, it is not easily replaced by any other messaging service. Furthermore, due to the unique Chinese language and cultural background in this market, there is no substitution relationship between geographical areas. Hence, the market was defined as the domestic market of China. Tencent argued that the plaintiff ’s market definition was incorrect. Specifically, it argued that in addition to QQ, there are various instant messaging services in the market, such as 66 

The presentation below is drawn from a report by Chinese Court Daily (19 April, 2012) available at the web page of the High People’s Court of Guangdong Province (in Chinese): http://www.gdcourts.gov.​ cn/gdcourt/front/front!content.action?lmdm=LM20&gjid=20120419103006504217. Also see Chen and Schoneveld (2012).

Treatments of Monopolization in Japan and China   229

MSN, microblog, emails, and MiTalk. Also, instant messaging service is only one of several communication services, and traditional communication products such as SMS, mobile phones and telephones, and e-mails, microblog, SNS, and other social networking services, all are substitutes. Lastly, Tencent held that, because of the openness and interoperability of the Internet, the users of instant messaging are not limited to those in Mainland China. In attempting to establish Tencent’s dominance, Qihoo presented the following evidence. First, the market share of QQ significantly exceeds 50  percent, the threshold specified in the AML. In fact, Qihoo pointed out that according to the data from iResearch, a leading consulting company, the market share of the defendant is as high as 76.2 percent, and that the penetration rate of QQ software is 97 percent according to a report by the China Internet Network Information Center (CNNIC). Second, according to the financial reports of the defendant for 2010, the annual revenues of the defendant were as high as RMB 1.96 billion (approximately $311.11 million) and its profitability was well above that of similar companies. Third, in terms of technological capability, the defendants possess more than 80 percent of all the patents in the area of instant messaging in China. Tencent claimed that it does not have a dominant position (even in the market for instant messaging). First, Tencent argued that there is significant difference between iResearch’s definition of market share and the provisions of the AML in terms of time and sale volumes, so the weight of the evidence is light. Additionally, the CNNIC data refers to the penetration rate rather than market share. Second, the number of users is not equivalent to the market share because a user may create multiple QQ accounts. Moreover, the users are generally sensitive to price. According to a survey conducted by the famous website eNet, if the QQ software becomes fee-based, 81.71 percent of the users will be lost. This implies that Tencent possesses no pricing power in the market. Qihoo accused the defendants of acts of exclusive dealing and tie-in sales. Specifically, Qihoo pointed out that on 3 November 2010, Tencent issued a letter to QQ users, forcing them to uninstall 360 software, a privacy protector developed by Qihoo that specifically targets Tencent’s QQ software, and refusing access to QQ spaces to users of the 360 browser. The forced uninstalling accounts impeded normal competition, as 360 software competes with Internet security software products of Tencent (QQ Doctor and QQ Software Housekeeper in particular).67 Furthermore, QQ tied its Software Housekeeper with its popular QQ messaging software, and forced users to install QQ Doctor in the name of upgrading instant messaging software, in violation of the AML. Tencent did not dispute the factual aspect of the alleged conduct but claimed that it did not constitute an abuse of dominance under the AML. Tencent defended its conduct by saying that QQ took incompatibility measures against the 360 products as a self-defense, because Qihoo used infringed software, namely, the 360 privacy protector, Koukou Guard, and other software to commercially defame and maliciously harm QQ products. Meanwhile, the packing of QQ software with the QQ software Housekeeper

67 

This can be understood as conduct of exclusive dealing, although not stated explicitly.

230   Ping Lin and Hiroshi Ohashi does not constitute tie-in sale. Rather, it is a normal practice of software integration. Both products are installed free of charge and the users may uninstall them conveniently. As can been seen, both the plaintiff and the defendant utilized economic arguments in each of the three stages of court hearing, namely market definition, and determinations of a dominant position and abuse, within the framework set by the AML. The court is yet to make a decision on the outcome of Qihoo v. Tencent at the time of this writing. It will be interesting to see to what extent the court will employ economic reasoning in its assessment of the facts and arguments of the plaintiffs and defendant. Since the NDRC and SAIC Regulations (mentioned earlier) are already in effect, unlike earlier private cases where the courts had no guidance from the antimonopoly authorities on abuse cases, it will also be interesting to see how the court interprets these regulations. Whatever the final outcome of Qihoo v. Tencent, it can be expected that more and more application of modern economic reasoning will be seen in future private cases in AML enforcement. While the Chinese judges are new learners in AML enforcement, the parties in private cases certainly have strong incentives to hire competition economists, both domestically and from overseas. One may even expect that before long, private enforcement may lead public enforcement by either NDRC or SAIC in applying state-of-art economic analyses in AML cases, given the staff constraints of NDRC and SAIC, which may make it infeasible for them to hire overseas economists.

9.4. Concluding Remarks: A Comparison We in this chapter have described treatments of monopolization under Japan’s 1947 antimonopoly act (AMA) and China’s 2008 Antimonopoly Law (AML). While Japan’s AMA has its roots in US antitrust law, China’s AML is drafted following the European model, which leads to some differences in their coverage of monopolization conduct. For instance, excessively high selling prices set by a dominant firm without valid reasons constitutes a violation of China’s AML, but not under Japan’s AMA. Moreover, China’s AML recognizes the notion of joint dominance, whereas Japan’s AMA does not. Third, China’s AML contains a presumption of market dominance based on market shares. Another distinctive feature of both Japan’s and China’s competition laws is that prohibitions of certain monopolization conduct are contained in either different sections of the same law or in different laws, which have led and can lead to systematic inconsistencies. Specifically, while they are subject to regulation of unfair trade practices of Japan’s AMA (Article 2.9), such exclusive conduct as predatory pricing and refusal to supply may also fall into the range of exclusive private monopolization under Japan’s AMA (Article 2.5). However, the competition tests for unfair trade practices and exclusion as set out in the law differ: finding unfair trade practice requires showing “a tendency to impede fair competition,” whereas finding violation of monopolization requires a

Treatments of Monopolization in Japan and China   231

proof of “a substantial restraint of competition.” Furthermore, unfair trade violations did not carry administrative surcharges (or criminal sanctions) until 2009, when some of the violations became subject to surcharges. As a result, the JFTC has pursued fewer monopolization cases than unfair trade practice cases (as discussed in section 9.2.3.2). A similar situation exists in China. Predatory pricing and price discrimination (against other undertakings) are prohibited by both China’s AML and its 1998 Price Law. However, unlike the AML, the Price Law does not require possession of a market-dominant position as a necessary condition for violation. Moreover, the Price Law adopts the per se rule rather the rule of reason. In terms of sanction, the Price Law allows for fines up to five times the illegal gains, where the AML does not permit punitive damages. The enforcement agency (NDRC) can in fact choose which law to apply towards predatory pricing and price discrimination. The major reason for the presence of the above-mentioned double coverage in both Japan’s and China’s antimonopoly laws is historical. In Japan, the provisions concerning unfair trade practice were originally meant to be supplemental to the major goal of the AMA, and they have often been regarded as precautionary means of protecting consumers and small and medium-sized enterprises. China’s Price Law was enacted in 1997, ten years before its AML, with the aim being to regulate pricing behavior of both the firms and the government, as China had just started to transform on a broad scale its central-planning economy to a market-based one.68 While the JFTC may well intend to enforce the private monopolization regulations by issuance of the 2009 Guidelines for Exclusionary Private Monopolization, the introduction of surcharges for unfair trade practices has made the distinction between private monopolization and unfair trade practices rather obscure. In the years to come, it will be a sensible task for China and Japan to harmonize the treatment of certain types of monopolization/abuse of market dominance under their respective legal regimes. On enforcement, Japan and China adopt quite distinct enforcement structures in their competition laws: the JFTC is the sole enforcer of Japan’s AMA, whereas China’s AML enforcement duties are carried out by three government agencies: NDRC, SAIC, and MOFCOM. It is interesting to note that, while the jurisdictional monopoly of the JFTC has been criticized by some scholars (e.g., Miwa and Ramseyer, 2005; and Harris and Ohashi, 2011) for having given the JFTC a degree of discretion in the application of the laws, as well illustrated in its treatment of private monopolization and unfair trade practices as discussed above, China’s three-pillar enforcement system has received wide criticism for its inherent tendency to lead to inconsistent application of the law (e.g., Fels, 2012; Zhang, 2011). Japan and China may benefit from sharing their experiences and lessons from their respective competition law enforcement. Enforcement of monopolization cases in Japan was notably lax until the late 1990s. Specific guidance of how private monopolization should be treated under Japan’s AMA 68 

The introduction of an antimonopoly law started in 1994 and was delayed until its final passage in 2007, due to debates about whether, and at which stage of China’s economic development, a comprehensive competition law was necessary (see, e.g., Lin, 2005).

232   Ping Lin and Hiroshi Ohashi was first introduced in the Guidelines for Exclusionary Private Monopolization under the Antimonopoly Act in 2009. In China, the two sets of regulations issued by China’s competition authorities, NDRC and SAIC, in 2011 provide guidance on how China treats monopolization. We are of the view that the principles and economic reasoning set forth under both Japan’s and China’s competition laws are broadly in line with international standards, although the competition enforcers and courts in both countries have yet to demonstrate how their respective guidelines are fully implemented in practice. In both Japan and China, there is a need for more economic and econometric analysis in competition law enforcement. In Japan, while economic and econometric evidence has been submitted in some merger investigations, it has played virtually no role in private monopolization cases, largely due to the lack of private enforcement and to the fact that the JFTC remains a largely law-dominated organization (Global Competition Review, 2012). Likewise, the Chinese antimonopoly authorities, as new members of the international antitrust community and facing tight staffing constraints, have not yet utilized much economic reasoning and analysis in abuse cases so far. However, private enforcement in China has been rather active, and there are signs that private parties have taken the lead in applying economic analysis in monopolization cases, as demonstrated by the ongoing Qihoo v. Tencent, in which the plaintiff engaged competition experts from overseas who presented systematic economic analysis in court. Active private enforcement, as a “competitive force,” seems to have an essential role to play in competition law enforcement in both Japan and China, as private parties have a strong incentive to employ advanced economic analyses and techniques available in today’s globalized world. Interaction of private action and public enforcement will further help improve the transparency and accountability of competition law enforcement in both countries.

References Aghion, Philippe, and Patrick Bolton. 1987. Contracts as a Barrier to Entry. American Economic Review 77 (June): 388–401. Amemiya, Kei. 2005. JFTC Challenges Intel’s Rebate and Marketing Programs: Key Implications for Doing Business in Japan. Client Alert, published by Morrison and Foerster, 3 October. Arai, Koki. 2012. Recent Developments in Private Monopolization. Mimeo, Japan Fair Trade Commission. Baumol, William. 1996. Predation and the Logic of the Average Variable Cost Test. Journal of Law and Economics 39: 49–72. Chen, Henry, and Frank Schoneveld. 2012. An RMB 150 Million Litigation in China for the Abuse of a Dominant Market Position. National Law Review. Available at http://www.natlawreview.​ com. CPRC. 2012a. Unjust Low Bids under Public Procurement. CR04-12 (in Japanese). CPRC. 2012b. Margin Squeeze in Network Industries. CR02-12 (in Japanese).

Treatments of Monopolization in Japan and China   233

Fels, Allan. 2012. China’s Anti-Monopoly Law 2008:  An Overview. Review of Industrial Organization 41: 7–30. Global Competition Review. 2012. Rating Enforcement 2011. United Kingdom:  Law Business Research Ltd. Harris, H. Stephan, and Hiroshi Ohashi. 2011. Japan, ­chapter 9 in R. I. McEwin, ed., Intellectual Property, Competition Law and Economics in Asia. Oxford: Hart. Inoue, Akira. 2007. Japanese Antitrust Law Manual: Law, Cases and Interpretation of the Japanese Antimonopoly Act. Frederick, MD: Kluwer Law International. Japan Fair Trade Commission. 1997. Fifty Years’ History of Japan’s Anti-monopoly Policies, Tokyo (in Japanese). Kisugi, Shin. 1999. Historical Overview of Japan’s Competition Policies (1): Pre-war to the 1997 Amendments. Chapter 1 in A. Goto and K. Suzumura, eds., Competition Policy in Japan. Tokyo: University of Tokyo Press (in Japanese). Kojo, Makoto. 1999. Historical Overview of Japan’s Competition Policies (2): The 1997 Amendments and Thereafter. Chapter 2 in A. Goto and K. Suzumura, eds., Competition Policy in Japan. Tokyo: University of Tokyo Press (in Japanese). Laffont, J. J., S. Marcus, P. Rey, and J. Tirole. 2003. Internet Interconnection and the Off-Net-Cost Pricing Principle. Rand Journal of Economics 34: 370–90. Lin, P. 2005. Competition Policy in China: Interactions with Industrial and FDI Policies. In S. Evenett and D. Brooks, eds., Competition Policy and Development in Asia. New York: Routledge, 71–106. Liu, Zhiyong, and Yue Qiao. 2012. Abuse of Market Dominance under China’s 2007 Anti-Monopoly Law: A Preliminary Assessment. Review of Industrial Organization 41: 77–107. Miwa, Yoshiro, and Mark Ramseyer. 2005. Toward a Theory of Jurisdictional Competition: The Case of the Japanese FTC. Journal of Competition Law and Economics 1(2): 247–77. Murakami, Masahiro. 1997. Dokusen Kinshiho Kenkyu (Anti-Monopoly Law). Tokyo: Kobundo (in Japanese). Negishi, Takashi, and Masayuki Funada. 2010. Japanese Antitrust Law. 4th ed. Tokyo: Yuhikaku (in Japanese). Norton Rose Group. 2012. Antimonopoly Law in China. Available at http://www.nortonrose.​ com/files/download-antimonopoly-law-in-china-63824.pdf. State Administration for Industry and Commerce of the People’s Republic of China. 2012. Draft Guidelines for Anti-Monopoly Enforcement in the Field of Intellectual Property Rights. Takigawa, Toshiaki. 2009. Competition Law and Policy of Japan. Antitrust Bulletin 54(3). Tong Shu. 2010. Reflection on Baidu Monopoly Litigation: Comments on Renren and Baidu. China Patent v. Trademarks 1: 66–71 (in Chinese). Wakui, Masako. 2008. Antimonopoly Law:  Competition Law and Policy in Japan. Suffolk, UK: Arima Publishing. Whinston, Michael. 2006. Lectures on Antitrust Economics. Cambridge, MA: MIT Press. Zhang, H. Angela. 2011. The Enforcement of the Anti-Monopoly Law in China: An Institutional Design Perspective. Antitrust Bulletin 56: 630–63.

CHAPTER 10

M O N O P O L I Z AT IO N I N DEV E L OP I N G COUNTRIES ALBERTO HEIMLER AND KIRTIKUMAR MEHTA

10.1. Introduction More than 100 jurisdictions in the world have adopted an antitrust law. Most of them (if not all) have legal provisions regarding abuse of a dominant position. Together with the prohibition of cartels, these provisions represent a central element of these laws. However, contrary to cartels, which are always prohibited, aggressive practices by dominant companies may only sometimes be an abuse (and prohibited). Furthermore the same practices are not prohibited if they are put in place by nondominant firms. These are the reasons why the sound application of provisions against abuse of dominance involves significant economic and legal challenges. Abuses can be directed towards the customers of a dominant firm (exploitative abuses) or towards the competitors of a dominant firm (exclusionary abuses). In recent years, enforcement agencies have for the most part applied the legal provisions against abuse of dominance to prevent exclusionary as opposed to exploitative abuses,1 as exemplified by the reports of the “Unilateral Conduct” working group of the International Competition Network (ICN 2007). The major reason why exploitative abuses are not even mentioned in these reports is that proving exploitation requires defining what is an acceptable exercise of market power. In other words it entails identifying the price above which a dominant firm would abuse its

1 

In some jurisdictions, notably the United States, only exclusionary abuses are addressed under the relevant legal provisions (in this case, Section 2 of the Sherman Act).

Monopolization in Developing Countries   235

position, a very difficult task for an administrative agency or for a judge. Furthermore, a rational dominant firm would price just below this threshold, possibly reducing the incentives of competitors to enter (that otherwise would have faced a much more profitable opportunity), reducing more than enhancing competition. In fact antitrust intervention against “excessive” prices can have the perverse effect of deterring entry by new competitors. However, especially in developing countries where regulatory institutions are not fully operational, it may be necessary to use competition law provisions as a residual tool of price regulation, as Ordover and Pittman (1993) argue, often in order to overcome unequal contract terms. More importantly, however, firms’ strategies that may be disciplined by the antitrust provisions against exclusionary abuse of dominance range from the refusal to provide access to an essential infrastructure, to the most complex cases in the field of competition law, involving, for example, the identification of the information requirements on its own products characteristics a dominant firm may be obliged to provide in order not to block the entry of competitors, the circumstances according to which intellectual property rights need to be shared, the exclusionary nature of target discounts, and so on. While a clear refusal to permit access to a physical, essential facility may indeed be easy to consider abusive, a target discount scheme (even if it achieves the same result as a refusal to deal) becomes abusive only insofar as it excludes from the market efficient competitors. In other words, in the case of indirect exclusions (predation, discounts, margin squeezes, etc.) only practices that are able to block the entry of equally efficient competitors should be prohibited, See Vickers (2005), a quite complex type of analysis to be performed. A further source of complexity is that if a firm is not dominant, not only is the conduct under consideration legitimate, it may even be procompetitive or, in any case, beneficial to consumers. The enforcement of the provisions against abuse of a dominant position should therefore be based on a rigorous evaluation of the actual exclusionary effects of the alleged abusive conduct. In particular, the more open markets are, even to potential competitors, the more unstable are the observed dominant positions. In such circumstances, abuse-ofdominance cases may be without merit, or even achieve the opposite of what they are meant to. By making sure that dominant companies do not exercise too much market power, abuse-of-dominance cases may suppress the signals that some markets are profitable and as a result may keep new entry low. This is the reason why, many argue, it is better not to intervene against abuse-of-dominance cases (see, e.g., Posner 1979). Arguably such an approach may be appropriate in the United States, where dominance mostly originates from innovation and entrepreneurship and therefore most markets are open to outsiders’ entry. In developing countries, where the most common origin of dominance is protectionist regulation, abuse-of-dominance cases are often the only discipline these firms may ever face. It is, however, worth reflecting that even in such situations intervention by the authority may result in losses in consumer welfare if the remedies imposed are inappropriate or even perverse in that they reinforce the dominant position. Very often, regulation, more than antitrust, is the most effective instrument.

236   Alberto Heimler and Kirtikumar Mehta

10.2.  Legal Standard As for the legal standard, the European concept of abuse of a dominant position overlaps but is not synonymous with that of monopolization in US antitrust law. Whereas a finding of abuse of dominance requires, as a basic element of a violation, the existence of a dominant position in a market, monopolization is concerned with the process by which firms achieve or maintain monopoly power. As a result, at least in principle, US law has a wider range of applications than European law since the rules against monopolization could easily be applied to firms that are not yet dominant, but may become so thanks to the adopted abusive behavior. In practice, however, at least in recent decades, as a result of case law developments, in the United States there is a higher threshold than in Europe for the finding of the required degree of market power and a wide acceptance of firm strategies that would be considered abusive under European law (Heimler 2007; Fox 2008). However, abuse of dominance laws in developing countries often have a wider application than in the EU or US models. China’s law, for example, is also rightly concerned with abuse of administrative power, Russia’s law stresses unequal contracts conditions, while India’s law takes over the EU idea of “unfair” pricing and often considers leveraging an abuse. Most developing countries take over the UNCTAD model law, which also focuses on compulsory contract terms that may directly or indirectly limit competitors. The recognition of an abuse of dominance typically requires three steps. First, it is necessary to identify the relevant market in which a firm is allegedly dominant. Then the existence of a dominant position needs to be established. Finally it is necessary to show that the specific practices under review are harmful to competition and are indeed abusive. The content and application of these elements varies between jurisdictions, the major difference being the role played by market shares in the identification of dominance. While it is clear that market shares represent only an indication of dominance, the full analysis of harm requires proof that a given practice is indeed harmful to competition. As a result, any lack of rigor in the definition of dominance can be more than taken up by the rigorous analysis of harm. Especially in developing countries, where authorities are small and weak and where vested interests are strong and connected, a market share threshold above which dominance is presumed may strengthen the power of the authorities with respect to powerful, dominant firms. Irrespective of the legal requirements associated with proving a dominant position, both economic theory and case experience underline the difficulties that are involved in reaching sound conclusions in abuse cases. The major reason is that firms can achieve a dominant position in a market through innovation, through superior production or distribution methods, or simply through greater entrepreneurial efforts. The enforcement of competition law should not interfere with such developments. However, in developing countries many dominant firms have achieved their positions via the protection of

Monopolization in Developing Countries   237

regulatory barriers to entry, national champion policies, or legal monopolies, so that the disincentive of innovation originating from an excessive enforcement activism may be less relevant. Indeed the provisions against abuse of dominance may in some occasion be the only instrument available for establishing a level playing field vis-à-vis less protected competitors. Nonetheless, the application of legal provisions relating to abuse of dominance in developing or transition economies entails particular challenges. As Anderson and Heimler (2007a and 2007b) argue, in analyzing cases in developing economies, account should be taken of some special characteristics: (1) thinly developed capital markets; (2) high entry barriers due to inadequate distribution channels and other business infrastructure; (3) intrusive regulatory and licensing regimes; (4) more extreme asymmetries of information than normally occur in developed economies, in both product and credit markets; (5) smaller and poorly staffed institutions (see also Beato and Laffont 2003 and Anderson and Jenny 2005). However, such shortcomings do not imply that developing countries are incapable of applying abuse-of-dominance provisions. To the contrary, as Brusick and Evenett (2008) argue, there are many abuse cases in developing countries and the most common types of abuses originate from the state or from privatized monopolies, suggesting that enforcement is complementary to liberalization and to the promotion of competition. As for markets unaffected by the liberalization drive, there seems to be no problem in developing countries with the effect-based approach. However, because authorities are small and weak, the number of cases remains quite limited.

10.3.  Competition Law Enforcement in Developing Countries Although most countries of the world have adopted a competition law, the number of cases has not grown proportionally. Indeed, enforcing antitrust laws is much more difficult than enacting them. Authorities must be funded, must be staffed with high-quality personnel (both economists and lawyers) and with salaries high enough so as to keep them in the job. Also, corruption is a problem developing countries face more often than developed economies, and it should be taken into account when institutions are set up. Making sure that the competition authority is of high standing ensures that professionals apply, that they keep their job, and that the institution maintains its independence. According to the 2003 ICN report on capacity building (see ICN 2003), some authorities, for example those of “Lithuania, Romania, Uzbekistan, the Slovak Republic, Poland and Hungary—whether they are formally inserted in the executive apparatus or are independent agencies—appear to be inserted into some of the highest policymaking councils of government.” The report suggests that similar status is enjoyed by the antitrust authorities of Jamaica, India, Mexico, South Africa, and Zambia. In many of

238   Alberto Heimler and Kirtikumar Mehta these countries salaries of the staff of competition authorities are higher than normal government officials and in some they are in line with those of central bank employees, usually the most reputable institution of any country. In order to strengthen the role and reputation of antitrust authorities, an effort should be made in that direction, so that good people are hired and kept in the job. As for corruption, it is well known that it is very difficult to eliminate it because even the existence of the crime is unknown.2 As a result, governments can fight corruption only indirectly, for example by reducing the relative gain associated with it, paying high enough salaries, and making it unnecessary for employees’ incomes to be supplemented by illegal activities. This is why granting the competition authority the same status as the central bank is so important. Furthermore, especially because in developing countries vested interests are highly concentrated and well connected, the system of antitrust enforcement has to be structured so as to promote as much independence of judgment as possible. One way to reduce the influence of the pressure of vested interests is to make the opening of a case more or less mandatory, especially in the area of merger control, since the best outcome for a well-connected dominant firm is not to have the merger analyzed at all. With an ex ante notification system (rather than a system of voluntary notification) the authority is forced to make a decision in favor of any merger above the threshold. When a case is opened a decision needs to be made, which makes it more difficult to conclude in writing with a coherent argument that an anticompetitive merger is not to be prohibited. The same is true for abuse of dominance. If the law introduces a presumption of dominance above a certain market share, the authority is obliged to take seriously complaints against firms that are legally presumed to be dominant. Otherwise, the political pressure not to open a case (on the argument that the firm is not dominant) may be difficult to resist. This is the main reason why in many jurisdictions (especially in developing countries) dominance is identified primarily by an established market share threshold that may allow for possible situation-specific deviations. For example, in South Africa, a firm is considered dominant in a market if it has at least 45% of that market, or if it has the power to control prices, or to exclude competition, or to behave to an appreciable extent independently of its competitors, customer, or suppliers. The ICN report on unilateral conduct is critical of jurisdictions that adopt a structural definition of dominance, praising countries like Latvia and Ukraine that have decided to move to a behavioral definition of dominance. According to the report, “a key advantage of the behavioral definition is that it better reflects dominance/SMP assessment being a multi-facetted analysis that reaches well beyond market shares.” However the report rightly recognizes that “while the use of behavioral and structural definitions [implies] different breadth and detail of analysis, the definition used does not appear to necessarily affect the assessment practice of specific jurisdictions directly.” In other words, by 2  Heimler (2012) argues that crimes that are undetected by the victims, in the case of corruption the taxpayer or the competitors, are the most difficult to eliminate.

Monopolization in Developing Countries   239

adopting an effect-based approach when assessing the abusive behavior under scrutiny, jurisdictions can more than make up any shortcoming encountered in the definition of dominance. In any case, the law should not oblige the authority to formally open a case for every complaint it receives, but only in those instances when there is a high presumption that a violation will be found. This means that if the authority is not convinced of the validity of a complaint, it should ask for additional information from the complainant, opening a procedure only when the probability of finding a violation is high. At the same time, in order to avoid corruption and to make sure that parties are treated fairly, antitrust rules should be accompanied by strict procedural requirements: reasonable deadlines for concluding cases, requirements that decisions be preceded by statements of objections, access to files, and so on. Also, judicial review is a key component of a fair antitrust procedure. This means ensuring that judicial review is timely and that judges to whom antitrust cases are assigned build expertise in antitrust enforcement so that they are able to exercise effective control over the authority’s decision-making. Especially in jurisdictions where the number of antitrust cases that arrive to the courts is limited, it is better to make sure that cases arrive to the same section of the court so that judges are able to gain experience, whereas creating a specialized tribunal would be a waste of resources and, paradoxically, could lead to a very understaffed body (no assistance provided to judges since the number of cases is low, but when a case arrives the judge is left alone and the process of judicial review may as result be very slow). However, in order to build expertise in overseeing successful cases, it may be better for new authorities to concentrate first on less complex cases (under a technical perspective) of exclusion of competitors by compulsory contract terms or on monopolization of raw material sources. In both instances the remedies are also easier to identify, such as government removing licenses or revising regulations and dominant firms being obliged to grant access to newcomers. Finally it would be good to make sure that the law indicates some defenses that firms can use. In particular efficiency defenses can make it easier to make sure that an effect-based approach is actually adopted.

10.4.  Abuse-of-Dominance Cases in Developing Countries: Some Examples In the past, actual examples of abuse-of-dominance cases (or other types of competition law cases) from developing countries were relatively rare and not well documented. Increasingly, however, this is no longer the case. Data on competition law cases in developing countries is becoming more systematically available (see, e.g., Clarke, Evenett, and Lucenti 2005; OECD 2005; Evenett and Jenny 2006; and Evenett and Brusik 2011). Also, the International Competition Network (ICN) has provided some new and

240   Alberto Heimler and Kirtikumar Mehta relevant information on abuse-of-dominance cases in developing countries, but not as much as expected.3 The available evidence shows that in developing countries the legal provisions against abuse of dominance are seldom enforced: the major reason being that abuse-of-dominance cases are very demanding in terms of the burden of proof that the agency has to meet. In particular, legal provisions on abuse of dominance are mostly enforced against firms that benefited from special and exclusive rights and that artificially excluded competitors from the market. In other words developing countries’ agencies are quite narrow in the way they enforce the rules: market-based dominance is rare, and the law is very seldom applied with respect to exploitative abuses. Jamaica has dealt in the past ten years with many abuse cases and in all of them has conducted a thorough analysis of whether the company was actually dominant and the alleged behavior was indeed exclusionary. What is interesting in the Jamaican experience is that many alleged cases of abuse ended with the finding that the abuse did not take place. At the same time, in the few cases where the agency concluded that the firm abused its dominant position, the case is well argued and is in line with international best practices. For example in the investigation into the conduct of Mossel Jamaica Limited with respect to the pricing of its fixed to mobile (FTM) voice termination service (Case no: 6489-09 June.22.2010), the Jamaican competition authority concluded that a margin squeeze by the dominant telecommunications provider was indeed abusive and that in order not to be abusive “the price Digicel charges its business fixed-line voice (BFV) subscribers to call its mobile voice (MV) subscribers [must] be no different from the price Digicel causes LIME to charge its BFV subscribers to call Digicel’s MV subscribers; and further this price [must] be no greater than is necessary to cover an appropriately measured cost of providing the service,” a conclusion very similar to the one taken by the European Commission in its Deutsche Telekom case.4 Very similar experiences can be found in other countries of Central America. In Costa Rica the competition authority (COPROCOM) decided in 2011 that an electricity company, Empresa de Servicios Publicos de Heredia, ESPH, abused its dominant position by not allowing a competing telecommunications company to use the electricity poles for laying cables to provide TV services. According to COPROCOM the refusal was not objectively or technically justified. Similarly in El Salvador the authority prohibited practices of refusal to deal in three cases. In one case, the TRANSAE v. MIDES case, the only waste-processing plant in the region, once it started to provide collecting services of infectious waste from hospitals, refused access to the waste-processing plant to competitors. In the other two cases the dominant electricity distributor did not allow access to competing generators. The same pattern, that is, that most cases originate in regulated sectors, emerges in the cases presented to the OECD Global Forum on Competition Policy in 2005. These were 3 

While the ICN started as a forum for all agencies of the world, the active participation of non-OECD countries is quite limited. 4  European Commission, Decision 2003/707/EC, Case Comp/C-1/37.451, 37.578 Deutsche Telekom AG, 2003 O.J. (L 263).

Monopolization in Developing Countries   241

nine cases of abuses of a dominant position in developing economies in Asia, Africa, Latin America, and the Caribbean, briefly summarized in table 10.1. The sectors covered by the cases comprised telecommunications, insurance, stevedoring services, electrical energy, pharmaceuticals, travel agencies, and marine ports. The overwhelming majority of these cases relate to infrastructure markets, including telecommunications (three of the cases), port facilities, stevedoring services (also three cases), and electricity transmission. In such cases, the decisions by competition authorities are clearly meant to promote the entry of competitors in newly liberalized markets characterized by large vertically integrated incumbents. In fact, the majority of these cases involve, in one way or another, the denial of access by competitors to unique or potentially unique facilities owned or managed by a vertically integrated dominant firm. Dominant firms employ a variety of methods for denying access to competitors, including a straightforward refusal to deal (Jamaica and the Russian Federation); denial of interconnection (Peru); tying arrangements (Chinese Taipei) and “degrading” of the quality of service provided (Zambia). In any case, clearly, the issue of denial of access is a key concern for developing and transition economies.

Table 10.1  Summary Characteristics of Nine Abuse-of-Dominance Cases from Developing/Transition Economies in the Early 2000s Economy

Sector affected

Key issue(s)

China

Insurance for schools

Disposition of management fees received by schools for monopoly-supplied, compulsory insurance

Jamaica

Stevedoring services

Latvia

Telecommunications

Peru

Telecommunications

Russian Federation Senegal

Electricity

South Africa

Pharmaceuticals

Denial of access by independent stevedoring companies to wharves owned by an integrated service provider Alleged pricing of a bundle of telecom services by a historical monopolist so as to exclude competition from a single-service provider Refusal to allow a competitor’s customers to “roam” using the facilities of a dominant mobile telephone service provider Denial of access, by competing generators, to electricity transmission facilities owned by a dominant generator Alleged abuse by airlines of “economic dependence” of local travel agents Establishment of a sole distributor for certain pharmaceuticals, jointly owned by major manufacturers and importers Whether rivals must be given access to a new service provided by a historical monopolist Exclusion of competitors from access to unique port facilities managed by a long-term concessionaire

Travel agents

Taipei, Chinese Telecommunications Zambia

Marine ports

Source: OECD (2005).

242   Alberto Heimler and Kirtikumar Mehta The majority of the cases (especially those concerning refusals to deal) appear to be concerned with exclusionary as opposed to exploitative abuses. Only two jurisdictions, China and Senegal, reported to the OECD cases concerned with exploitative abuses. However, for China, the decision in the insurance market is clearly related to overcoming the possibility that the monopoly supplier will exercise significant market power in the process of fee determination. A legitimate regulatory issue was solved through antitrust enforcement. As for Senegal, the case, an abuse of “economic dependence” by major airlines on the part of small travel agents, is meant to protect the income of these travel agents and is clearly not fully in line with an efficiency-oriented approach to antitrust enforcement.

10.5.  Abuse Cases in Russia, China, India, and South Africa 10.5.1. Russia In Russia the number of abuse-of-dominance cases has been very high and relatively stable for many years (see table 10.2). Contrary to most other jurisdictions, Russia enforces the provisions against abuse of dominance mainly with respect to exploitative abuses by public utility service providers (distance heating, electricity, telecommunications, etc.) or by monopsony purchasers. In other words, the competition authority mostly serves as the regulator of behavior for natural monopolies in the infrastructure and utilities sectors. Most of the cases are meant to stop public utility services from pricing above the regulated prices with limited effect on competition, while a significant portion of Russia’s refusal-to-contract cases also have limited effects on competition, being cases of simple denial of service or a result of a dispute over terms rather than strategic denials of access to competitors. Indeed most cases are opened on the basis of complaints by final consumers, and cases concerning imposition of burdensome contract provisions, refusal to conclude a contract, and violations of pricing regulatory rules account for a large share of the caseload. This overflow of cases that are mostly aimed at consumer protection influences the work priorities of the Russian agency (Federal Antimonopoly Service, FAS), which as a result is not so active in the more severe violation of the antitrust rules, including cartels and abuse of dominance, where the number of ascertained violations is quite limited. Some cases on abuse of dominance are reported by the ICN and are, at least partly, in line with international standards. For example, in a case of refusal to deal, FAS imposed a fine of more than 10 million rubles on the Sakhalin-based supplier of aviation fuel TOK, a joint-stock company, for monopoly high pricing of aviation kerosene for the Youzhno-Sakhalinsk airport and for refusing to allow the Vladivistok Avia Open

Table 10.2  Abuse-of-Dominance Cases by Type in the Russian Federation 2008

2009

All cases Complaint Own opened based initiative All abuse-ofdominance cases Monopoly high prices Monopoly low prices Withdrawal of goods from circulation Abusive conditions of contract Groundless reduction of production Refusal to contract Unreasonably high prices of financial services Unreasonably low price financial services Discriminatory conditions Creation of entry/exit barriers Violation of legal rules on pricing Other violations

2010

All cases Complaint Own All cases Complaint Own opened based initiative opened based initiative

1,639

1,481

158

2,411

2,061

350

2,736

2,419

317

66 4 2 393 217 231 41 3 77 53 210 342

46 3 2 373 211 227 25 2 68 49 153 322

20 1 0 20 6 4 16 1 9 4 57 20

161 7 6 448 317 380 72 2 56 88 366 508

91 4 2 415 312 371 43 2 45 84 233 459

70 3 4 33 5 9 29 0 11 4 133 49

130 6 9 543 296 357 63 4 58 98 394 778

100 6 8 513 288 345 37 3 47 85 265 722

30 0 1 30 8 12 26 1 11 13 129 56

244   Alberto Heimler and Kirtikumar Mehta joint-stock company to store aviation fuel at the airport (ICN, 2010). Analogously, in 2008 FAS concluded that it was abusive for a monopolist gas supplier to make it mandatory for customers to specify the gas quantities needed for a given year and then require customers to pay for storage in case the target set for the year was not reached. The reason for the relatively few cases on exclusionary abusive practices, is that the Russian law is mainly concerned with monopoly pricing and as a result requires dominant companies to be registered so that consumers would know that they should not overprice. The register, maintained by FAS, lists all enterprises with a share of the market of more than 35% with thousands of listings.5 A significant bureaucracy is required to keep the register up to date, so that keeping the register is quite costly. On the other hand, the benefits of the register are very limited since the public is generally not informed on which enterprise is still dominant, because many entries are old, sometimes even ten years old. As for monopoly prices, the Russian law defines what an abusive high price is (a price that exceeds the cost of production and a normal profit), ignoring the difficulties associated with the proper assessment of these values. Although the number of cases against monopoly prices is a low percentage of all abuse-of-dominance cases, they are numerically quite important (130 in 2010) and represent a big and unnecessary burden on FAS resources.

10.5.2. China In China the antitrust law entered into force in August 2008. The law follows the EU approach prohibiting the abuse of a dominant position. Like the European Commission in the “Guidance on the Commission’s enforcement priorities in applying Article 82 (102) of the EC Treaty to abusive exclusionary conduct by dominant undertakings,”6 the Chinese law identifies a rebuttable presumption of a dominant position, defined in terms of market shares. Furthermore, in December 2010 the authorities in charge of nonprice (SAIC) and price (NDRC) abuse issued guidelines7 to clarify for the courts and for dominant firms the way legal provisions are to be interpreted. Since 2008 quite a number of abuse-of-dominance cases have been decided by the courts. All of these cases (at least those reported in English sources) show that Chinese courts adopt an effect-based approach with respect of the definition of dominance and the identification of an abuse. Even in the first court ruling in a dominance case (in 2009), Sursen v. Shanda and Xuanting, where Shanda and Xuanting were charged with abuse of a dominant position in the Chinese online literature market, the court made it

5  The full title of the register is the “Register of Economic Subjects (with the Exception of Financial Organisations) That Have a Share of the Market for a Specific Product Greater Than Thirty-Five Percent.” 6  European Commission (2009), Communication from the Commission, available at http://eur-lex.​ europa.eu/LexUriServ/LexUriServ.do?uri=OJ:C:2009:045:0007:0020:EN:PDF. 7  In December 2010 SAIC released the Regulation on Prohibiting Abuse of Dominant Market Positions and NDRC released the Regulation on Anti Price Monopoly.

Monopolization in Developing Countries   245

clear that the evidentiary threshold in supporting a plaintiff ’s claims for dominance was to be “objective, fair and accurate.” Anecdotal evidence was not considered sufficient, and the case was dismissed. In the Renren v. Baidu case, Renren, an operator of a medical information intermediary website, after reducing its spending on Baidu’s search advertising service, immediately witnessed a sharp reduction of visits to its website, as a result of the fact that Baidu had actually blocked its website from searches. The court agreed with the complainant that the “search engine” was a relevant antitrust market even though the search engine service was free to Internet users. However, the court accepted Baidu’s defense that it blocked Renren’s website because it contained too many links to dangerous sites that could threaten the reliability of the search service.8 Contrary to the court cases, Chinese agencies are often influenced in their decision-making by the peculiar institutional structure China chose for antitrust enforcement. NDRC is responsible for price-related infringements, while SAIC for nonprice ones. As a result, once one of the agencies opens a case, it tends to refrain from entering into the jurisdiction of the other, leading to a very formal approach in decision-making. For example, in November 2011 NDRC imposed fines of almost CNY 7 million, confiscated illegal gains, and imposed a cease-and-desist order on two pharmaceutical companies for monopoly pricing, the first time that the NDRC has imposed significant penalties for a breach of the antitrust law. The two firms, Shuntong and Huaxin, had signed exclusive distribution agreements with the only two manufacturers of an essential input for a drug that reduces blood pressure. As a result of the exclusive arrangements, the price of the drug increased from CNY 200 per kilogram to as much as CNY 1,350 per kilogram. The origin of the price increase was the exclusive arrangements, which by the way might have been a facilitating practice for a cartel between Shuntong and Huaxin. This more complex reasoning might have put into question NDRC’s jurisdiction on the case, and this was probably the reason why NDRC failed to argue along these lines.

10.5.3. India Section 4 of the 2002 Competition Act (amended in 2007) contains a prohibition of both exploitative and exclusionary abuses by dominant enterprise, singly or collectively. The content of Section 4 is very much in line with standard formulation of prohibited conducts following the European approach.9 It is useful nevertheless to note certain specific aspects of the legislation on abuse under Indian competition law. 8  The case in Chinese is summarized in http://bj1zy.chinacourt.org/public/detail.php?Id=675 and is discussed by Liu and Qiao (2012). 9  For a thorough discussion of Indian abuse-of-dominance cases see Chakravarthy (2010); Aditya Bhattacharjea (2008) provides a critical assessment of the Indian Competition Law and its development from the previous legislative framework.

246   Alberto Heimler and Kirtikumar Mehta First, while liability for abusive conduct can be attributed to an enterprise or to a group that consists of a parent and its associated subsidiaries, the law does not cover abusive conduct by “an association of enterprises,” although provisions on anticompetitive agreements provide a way to intervene in the case of abusive conduct. For unfair, discriminatory or predatory prices, Section 4 explicitly recognizes the “meeting the competition” clause. This attempt to circumscribe instances of unfair, discriminatory, or predatory prices would have been better achieved by relating to some measure of cost—for example, the simple existence of differences according to customers in the price/cost ratio for the same good or service could be labeled as discriminatory. As defined, however, an incumbent that is pricing to “meet competition” across all its customers is unlikely to be found to have abused its dominant position, while if only applying low pricing to customers of rivals to “meet competition,” it would be considered to abuse its dominant position by practicing price discrimination. A dominant incumbent in one relevant market risks being considered to be abusing its dominance if it merely “enters” another relevant market, which seems to create unnecessary entry barriers. By far the majority of the Competition Commission of India’s orders relate to Section 4 complaints. In comparison there are far fewer cases of cartels being uncovered and fined. Procedurally, if the complainant does not submit sufficient information to permit a decision by the Commission that there is prima facie indication of abuse, then the case is closed without an investigation by the director general of investigations. Where there is a prima facie indication, a full investigation takes place.10 A number of orders of the CCI outline in some detail the findings of the investigation and generally a reasoned opinion leading to the final decision. Cases of interest from the point of view of understanding the decisional practice of the CCI are those where after a fact-intensive investigation the CCI came to a decision on a finding or lack of finding of abuse, which we describe in what follows. In the manufacturing sector, an intensive investigation by the CCI (order 22/2010, Kapoor Glass Private Limited vs Schott Glass India Private Limited) showed that Schott Glass was dominant in the supply of borosilicate glass of the specific quality that is used to manufacture ampoules, vials, and cartridges for use in pharmaceutical industry. Furthermore, it reinforced its dominance through a downstream joint venture with a major India-based manufacturer of ampoules, vials, and cartridges. The abuse concerned unfair and discriminatory pricing to third-party transformers. The complainant alleged predation, loyalty discounts, and refusal to supply, but the investigation did not fully sustain a finding on all these aspects. The decision imposed a fine and also ordered Schott to cease several exclusive vertical agreements and desist from implementing its scheme of discriminatory discounts to transformers. 10 

See Sharma (2014)

Monopolization in Developing Countries   247

In the real estate sector, the CCI found a major construction landowner and property developer had abused its dominant position by charging unfair, arbitrary, and unreasonable terms in allotting apartments in a subnational geographical market near Delhi (order 18, 24, 30, 31, 32, 33, 34, and 35/2010 DLF Park Place Residents v. DLF Limited and DLF Home Development). The relevant market was defined as the provision of apartments in the region around Delhi where DLF was found to be dominant. As with a large number of other real estate or property contractual cases dealt with by CCI, in this particular case the complaints were all made by individuals purchasers who had been directly affected by unreasonable and unfair pricing and a pattern of arbitrary terms imposed on them. The assessment of excessive pricing in the decision is based in particular on arbitrary terms and retrospective changes to those terms, an approach that is not entirely a structural one (Simon Roberts 2008) nor one based on excessive departures from long-run costs of an efficient supplier (Calcagno and Walker 2010). CCI’s approach is adapted to the specificities of the situation under assessment, where the allocation mechanism for development of urban land did not follow any robust tender procedures. Under the abuse provisions of Indian competition law, thus, individuals singly or in groups can file for an intervention by the CCI rather than use consumer protection law. An important abuse-of-dominance case in the financial sector (order 13/2009 MCX Stock Exchange v. National Stock Exchange of India) concerned the alleged predatory and entry deterring practices of National Stock Exchange (NSE) against the entry of MCX into the currency derivative market. The investigation confirmed the dominant position of NSE in “all stock exchange segments” and concluded that the waiving by NSE of transaction fees, subscription fees, and deposit margin requirements for trading currency derivatives and limiting access to contemporaneous trading data software amounted to an abuse of a dominant position. In contrast to other cases where CCI has found an abuse, here the question of harm to competition or to investors was not raised; nor did the investigation recognize that in regard to stock exchanges the competition is “for the market” rather than “within the market.” The dissenting opinion by two commissioners is based on a robust market definition exercise, and they rightly raise the question of harm to competition when the entry of a new rival leads to significant reductions in charges and fees paid by users. In applying the abuse-of-dominance provisions of the Indian Competition Law, the CCI has also dismissed a number of complaints after investigation. Some of the important ones are listed below: (1) Order 01/2010 relating to the complaint of GKB High Tech concerning abusive conduct of Transition Opticals. (2) Order 1/28 relating to MRTP Case No. 1/28, M/s Royal Energy Ltd. v. M/s Indian Oil Corporation Ltd. & Ors. (3) Order 13/2011 relating to Manappuram Jewellers Pvt. Ltd., Thrissur, Kerala v. Kerala Gold & Silver Dealers Association, Thrissur, Kerala & Ors. (4) Order 50/2011 relating to Gujarat Textile Processors Association, Surat, Gujarat and Government of Gujarat v. Gujarat Gas Company Ltd., Ahmedabad. (5) Order 11/2009 relating to Jindal Steel & Power Ltd. v. Steel Authority of India Ltd.

248   Alberto Heimler and Kirtikumar Mehta In regard to order (1), the investigation on the relevant market found that the complainant party itself had a significant market position on the market for plastic photochromatic lenses. Similarly for order (3), the investigation did not uncover a dominant position on a relevant market for the alleged entity abusing market dominance. In orders (2) and (4) the issue was not the market position of the defendant but was essentially that of the state measures facilitating or underpinning conduct, as the dissenting opinion in order (2) very cogently illustrates. As regards order (5), the investigation and final order dismissing the complaint reflects a benchmark example of an “effect”-based analysis of alleged abusive conduct. Finally it is important to note that unlike in most other jurisdictions, in India there are very few cases against state-owned regulated monopolies in telecoms or other governmentally run utilities.

10.5.4.  South Africa In South Africa in recent years the Competition Commission has prosecuted three important cases of abuse of dominance, confirming the capabilities of developing countries to apply an effect-based approach.11 South African Airways was fined by the Competition Commission in 2005 for operating loyalty schemes for travel agents that induced travel agents to divert passengers away from competing airlines (Nationwide Airlines and BA/ Comair). The loyalty schemes granted to travel agents a flat basic commission on amounts up to a target, and if sales exceeded that target, then agents received a commission on the whole amount, not just on the amount above the target. In addition, amounts above the target received a higher commission rate. The Competition Tribunal concurred with the decision of the Commission. According to Federico (2013), this case represents a bestpractice example of how to apply an effect-based approach to exclusionary discounts. Two other important recent cases concern the sector of agricultural processing. Following a complaint from maize traders, the Commission brought a case against Senwes in 2006, finding that it abused its dominant position. Senwes is a silo owner that is vertically integrated into grain trading. It is dominant in the storage of grain (mostly white maize) for farmers close to its silos, which are located mainly in the Free State. It charged a favorable price to farmers, as opposed to traders, to store their maize beyond a period of 100 days. The Tribunal dismissed the price discrimination charge but agreed that Senwes was practicing a margin squeeze. The final court judgment ruled against the Tribunal since neither the original complaint nor the Commission’s decision referred to margin squeeze. A second case in agriculture concerned of Pioneer Foods in the bakery sector. As a dominant incumbent, Pioneer Foods’ subsidiary could aggressively threaten a price war on those independent bakeries that did not follow prices set by Pioneer. South Africa has been very active in enforcing the law against abuses by public utilities in order to promote competition in liberalized markets. For example, in August 2012, 11 

The description of the cases is based on media releases by the South African Competition Commission.

Monopolization in Developing Countries   249

the Competition Tribunal imposed a penalty of R 449,000,000 on Telkom SA Limited for abusing its dominance in the telecommunications market between 1999 and 2004, a period in which Telkom was a monopoly provider of telecommunications facilities. The Tribunal concluded that Telkom leveraged its upstream monopoly in the facilities market to advantage its own subsidiary in the competitive value-added network market and in particular in the market for the provision of Internet services.12 A criticism of the South African system, unfortunately shared by many jurisdictions, is the long time necessary for a decision to be taken. The Telkom case, decided in 2012, referred to a practice that had taken place eight years before, substantially reducing the guiding effect of the decision and the deterring effect of the fine.

10.6.  Developing Appropriate Remedies in Abuse-of-Dominance Cases Remedies in abuse-of-dominance cases typically comprise one or more of the following types of measures: (1) behavioral orders prohibiting the continuation of specific actions or behavior; (2) the levying of fines or financial damages on firms that have engaged in such conduct; (3) divestiture of specific assets of the firm, including vertical disintegration, for the purpose of re-establishing competition in the market; (4) other “proactive” measures such as mandating access to essential facilities and compulsory licensing. Probably the most common are the first category of measures—that is, orders prohibiting the continuation of specific actions or behavior. This is particularly the case in developing countries, where antitrust agencies very often lack effective sanctioning powers or are unable to impose structural measures. However, by themselves, cease-and-desist orders may not be sufficient to deter similar abuses by other firms or be able to restore competition in the markets affected by the anticompetitive conduct. For this reason, other measures such as fines or structural divestiture may need to be considered. Where simple orders prohibiting abusive conduct are imposed, it is important that they be sufficiently clear and specific to prevent recurrences. For example, if the core abuse relates to the use of loyalty rebates to exclude potential competitors, guidance must be provided as to the circumstances in which such rebates will and will not be considered acceptable. A mere prohibition of “continued anticompetitive conduct” will not suffice. Pecuniary sanctions have clear advantages vis-à-vis simple prohibitory orders in deterring similar conduct by other firms. However, while fines are an appropriate remedy in many cases, they are not appropriate in the circumstances where the abuse is somehow new or the firm may not have been aware that it had reached a dominant position.

12 See http://www.compcom.co.za/assets/Uploads/AttachedFiles/MyDocuments/​ Tribunal-imposes-R449M-on-Telkom.pdf.

250   Alberto Heimler and Kirtikumar Mehta On the other hand, where a clearly dominant firm violates known standards, that is, those that have been enunciated in relevant guidelines or case decisions, it may well be appropriate to seek/impose fines or other punitive sanctions. Heimler and Mehta (2012) provide some indications on the range of fines that may be necessary for achieving deterrence in abuse cases, somewhere between 3% and 6% of the relevant turnover. Still another remedy that is often considered in cases involving network industries, sometimes as an alternative to vertical divestiture, is the mandating of access to facilities that are deemed to be essential. In the EU, the “essential facilities doctrine” permits the mandating of access in cases where the following four criteria are met: (1) control of the essential facility by a monopolist; (2) indispensability of access to the facility by a downstream/upstream competitor; (3) denial of the use of the facility to a competitor supplying a new product; and (4) the absence of an objective justification for denial of access (see Oscar Bronner GmbH & Co. K.G. v. Mediapring Zeitungs und Zeitcshriftenverlag GmbH & Co. K.G., Case C-7/97, 1998 E.C. R. I-7791, [1999] 4 C.M.L.R. 112).13 A key challenge in such cases is to set an appropriate price for the services to be provided. Without minimizing the problems inherent in setting such a price, one possible approach to this issue is given by the so-called Efficient Component Pricing Rule (ECPR) (see Baumol and Sydak 1994). Under this approach, the incumbent operator’s price to final consumers is accompanied by an access charge low enough to allow equally efficient competitors to compete (and equal to the final price minus the cost avoided by not supplying in the competitive segment of the market). In practice the remedy should be narrowly tailored to the violation and should not be used to obtain “more.” In addition, behavioral remedies should not be imposed permanently, so as not to limit too much the flexibility of the dominant company vis-à-vis its competitors. And third, structural divestiture remedies should be used only in extraordinary circumstances.

10.7. Conclusions Three conclusions can be reached from this overview. First, in contrast to mature competition jurisdictions, authorities in young jurisdictions do not hesitate to apply the provisions on abuse of dominance that exist in their laws. Second, and again compared to mature jurisdictions, the investigation and decision periods are sometimes relatively short (with some exceptions); thus there exists a reasonable prospect that remedies are timely. Third, and perhaps not too surprisingly, the risks for false positives of the formalistic approach to abuse of dominance have been well understood. Thus in developing countries overwhelmingly the approach is based on economic effects and harm to competitive processes with a significant emphasis on good empirical investigations. The 13 

It should be noted that the essential facilities doctrine concerns the availability of a remedy and not a finding of liability; before access can be granted under this doctrine an underlying antitrust violation must be shown.

Monopolization in Developing Countries   251

cases to date reveal that the focus is on actual foreclosure in the relevant markets concerned rather than on restrictions of competition that may give grounds for assuming potential foreclosure, or generally on market situations where there is already discernible empirical evidence of damage to competitive process of a dominant incumbent. A notable specificity of developing countries is that sometimes consumer protection issues are addressed by enforcing the abuse-of-dominance provisions instead of the specific consumer protection legislation often administered by the sectoral regulator, the reason being that the antitrust rules are often the only ones available in such jurisdictions or the most immediately effective. As soon as the legislative and institutional pillars are in place, these countries may switch to the best-practice approach. Finally, the peculiarity of the Chinese institutional structure, where one agency is in charge of enforcing the law against nonprice and another one against price abuses, may lead agencies to stick to the original reconstruction of the substance of the case, fearing otherwise to lose jurisdiction. The same pattern does not emerge in court cases. The focus of enforcement actions in developing countries in the last decade, which was characterized by a process of restructuring and privatizations in the network industries, was initially on cases having to do with refusals to provide access to entrants by former monopolies. Over time the focus has moved to other markets in the industrial and distributive industries where emerging dominant incumbents, several being joint ventures between domestic and international partners, have been found to exercise market power through a wide range of classic exclusionary practices. Sanctions for abusive conducts would still appear to be modest; however, certain cases show that even competition authorities of developing countries are not reluctant to impose mandatory cancellations of contracts or structural remedies.

References Anderson, Robert D., and Frédèric Jenny. 2005. Competition Policy, Economic Development and the Possible Role of a Multilateral Framework on Competition Policy: Insights from the WTO Working Group on Trade and Competition Policy. In Erlinda Medalla, Competition Policy in East Asia. New York: Routledge/Curzon. Anderson, Robert D., and Alberto Heimler. 2007a. Abuse of Dominant Position: Enforcement Issues and Approaches for Developing Countries. In Vinod Dhall, Competition Law Today. New York: Oxford University Press. Anderson, Robert D., and Alberto Heimler. 2007b. What Has Competition Done for Europe: An Interdisciplinary Answer. Aussenwirtschaft 4. Bhattacharjea, Aditya. 2008. India’s New Competition Law: A Comparative Assessment. Journal of Competition Law and Economics 4(3): 609–38. Baumol, William J., and Gregory J. Sydak. 1994. The Pricing of Inputs Sold to Competitors. Yale Journal of Regulation 11: pp. 171–202 . Beato, Paulina, and Jean-Jacques Laffont. 2003. Competition Policy in Regulated Industries:  Approaches for Emerging Economies. Washington, DC:  Inter-American Development Bank.

252   Alberto Heimler and Kirtikumar Mehta Brusick, Philippe, and Simon Evenett. 2008. Should Developing Countries Worry about Abuse of Dominance? Wisconsin Law Review. Calcagno, Claudio, and Mike Walker. 2010. Excessive Pricing: Towards Clarity and Economic Coherence. Journal of Competition Law and Economics 6(4): 891–910. Chakravarthy, S. 2010. Dominance and Its Abuse. Manupatra Competition Law Reports, July– September, B153–B160. Clarke, Julian L., Simon J. Evenett, and Krista Lucenti. 2005. Anti-competitive Practices and Liberalizing Markets in Latin America and the Caribbean. World Economy 28:1029–56. Evenett, Simon J., and Frédéric Jenny. 2004. An Inventory of Allegations of Anticompetitive Practices in Sub-Saharan Africa. Working Paper, September 5. Available at http://www.even​ ett.com/working/evenettjennycapetownsummary.pdf. Federico, Giulio (2013), SAA II: Abuse of dominance in the South African skies, Journal of Competition Law & Economics, 9 (3): 709–37 Fox, Eleanor. 2008. Market Power Elements of Abuse of Dominance: Parallels and Differences in EU and US. In Claus-Dieter Ehlermann and Mel Marquis, European Competition Law Annual 2007, A Reformed Approach to Article 82 EC. Oxford: Hart. Heimler, Alberto. 2007. Abuse of Dominance: How Far Is the EU from the US? Economia delle scelte pubbliche 25(1):31–52. Heimler, Alberto. (2012). Cartels in public procurement. Journal of Competition Law and Economics 8(4): 849–62. Heimler, Alberto, and Mehta Kirtikumar. 2012. Violations of Antitrust Provisions: The Optimal Level of Fines for Achieving Deterrence. World Competition, March. International Competition Network. 2003. Capacity Building and Technical Assistance: Building Credible Competition Authorities in Developing and Transition Economies. Brussels. International Competition Network. 2007. Report on the Objectives of Unilateral Conduct Laws, Assessment of Dominance/Substantial Market power and State Created Monopolies. Available at http://www.internationalcompetitionnetwork.org/library.aspxsearch=&group =1&type=0&workshop=0. International Competition Network. 2010. Report on the Analysis of Refusal to Deal with a Rival under Unilateral Conduct Laws. Presented at the Ninth Annual Conference of the ICN, Istanbul, Turkey, April. Liu, Zhiyong, and Yue Qiao. 2012. Abuse of Market Dominance under China’s 2007 Anti-monopoly Law: A Preliminary Assessment. Review of Industrial Organization 41(1–2): 77–107. Ordover, Janusz, and Russell Pittman. 1993. Comments on Competition Policy during Transformation of a Centrally Planned Economy. In Barry Hawk, International Antitrust Law and Policy. New York: Annual Proceedings of the Fordham Corporate Law Institute. Posner, Richard. 1979. The Chicago School of Antitrust Analysis. University of Pennsylvania Law Review: 925 41 (1–2): 77–107. Roberts, Simon. 2008. Assessing Exclusive Pricing: The Case of Flat Steel in S. Africa. Journal of Competition Law and Economics 4(3): 871–91. Sharma, Kaushal Kumar (2014): “Competition Commission Cases- A Compendium of CCI cases from 2009-2014” LexisNexis (India) Thanitcul, Sakda. 2005. SMEs and Competition Law: A Case Study of Suppliers of Goods to Large Retail Stores. Paper presented to the International Expert Seminar “Mapping Policy Experience for SMEs,” organized by Thai APEC Study Center, Thammasart University, May 19–20, Phuket, Thailand. Vickers, John. 2005. Abuse of Market Power. Economic Journal 115: F244–61.

CHAPTER 11

BU SI N E S S S T R AT E G Y A N D A N T I T RU S T P OL IC Y MICHAEL J. MAZZEO AND RYAN C. MCDEVIT T

11.1. Introduction Business strategy is fundamentally about firm decision-making. Antitrust policy and enforcement, in turn, evaluate the decisions made by firms and the market outcomes that result. To the extent that firms’ decisions will be scrutinized ex post, managers must understand how antitrust concerns might constrain their actions and, thus, suggest alternative optimal decisions. Owing to this importance, most business strategy courses broach the subject of antitrust, and managers frequently confer with antitrust attorneys when making important strategic decisions. Correspondingly, it is also useful for the antitrust community to understand how firms use the concepts and frameworks of business strategy to make the decisions that they will be evaluating. Business strategy maintains a holistic orientation, drawing on traditional functional areas such as operations, finance, accounting, and marketing to inform the firm’s overarching direction. Through the effort of academics and management consultants, business strategy has grown in prominence and is pervasive at the top levels of most firms. Over the past several decades, economics has emerged as the guiding discipline for businesses making strategic decisions. This immediately suggests a potential conflict between the goals of practitioners in strategy and antitrust. Economics-based strategy will inevitably aim towards the maximization of producer surplus, while antitrust policy and enforcement puts more emphasis on protecting consumer welfare. In this chapter, we will discuss circumstances in which conflicts may arise between these goals; however, we will also highlight situations in which firms can increase their profits without shifting consumer surplus to producer surplus. These represent potential opportunities

254   Michael J. Mazzeo and Ryan C. McDevitt where firms can be successful—even to the point of enjoying market power—without being sanctioned by antitrust authorities. Using economics as a framework for understanding business strategy requires that a firm (or any organization) start with a clearly articulated objective. This objective typically centers on maximizing shareholder wealth for public firms, though for private companies and nonprofit organizations modified objectives are common. The objective provides a structured rubric managers can use to evaluate alternative strategies and make judgments about the optimal approach. We often use profit-maximization as a shorthand description for maximizing shareholder wealth, reflecting the fact that the value of a public companies’ shares is based on its current assets plus the future stream of profits expected to result from its activities. Definitions of strategy generally emphasize “big picture” issues for a business and “long run” rather than “short run” decision-making. Distinguishing between strategy and “tactics” is useful here, with the latter being more the purview of operations than strategy. As Besanko and coauthors (2009) state in their leading textbook, “strategy is revealed in terms of consistent behavior, which in turn implies that strategy, once set, is not easy to reverse.” This may hold relevance for antitrust insofar as individual behaviors such as predatory pricing may be thought of as tactics rather than strategy. Nonetheless, the overarching approach to decision-making that reflects a consistent business strategy may contain elements that put a firm at risk of scrutiny from antitrust authorities. The role of economics in providing structure for understanding business strategy is neither universal nor uncontroversial. However, this discipline has influenced both theory and practice because economics requires precision regarding the inputs to its models and the identifying assumptions needed to make conclusions based on empirical evidence. In her article “Why Economics Has Been Fruitful for Strategy,” Scott Morton (2003) notes that “economists have powerful tools: formal modeling, the assumption of maximizing behavior by agents, and the notion of equilibrium. Using these techniques produces crisp, testable conclusions.” Managers benefit from the structure of formal economic modeling because the resulting insights suggest prescriptions that depend on the specific economic environment that firms face. However, economics rarely provides the “answer” in strategy—instead, it demonstrates the trade-offs associated with alternative strategies and can identify the conditions under which they will be more or less successful. Business strategy is often analyzed from the perspective of three related audiences: practitioners who actually make the decisions (and the consultants who advise them), researchers who study management and organizational decision-making, and business school instructors who teach the concepts and frameworks of business strategy to students ranging from undergraduates and MBA candidates to participants in executive education programs. This chapter is largely organized around the last perspective, as it tends to act as a nexus by incorporating both academic research and real-world applications. Indeed, many practitioners have received training in business schools, increasing the relevance of what gets taught in the classroom. With that in mind, this chapter is organized around three fundamental concepts taught in business strategy classes. The first is “Value Creation and Capture,” which

Business Strategy and Antitrust Policy   255

establishes the connection between firms’ activities and the notion of economic surplus. In so doing, consumers are brought into consideration explicitly, as consumers’ willingness-to-pay represents a bound on the amount of economic surplus that a firm can create through its activities. A  firm’s interaction with its external environment determines how the total surplus it generates is divided between consumer surplus and producer surplus. Profit-maximizing firms will focus inevitably on the latter—otherwise termed value capture—in their decision-making. However, this construct makes it clear that a firm can generate more producer surplus either by increasing the share of total surplus captured (relative to consumers) or by increasing the total value created. Two influential business strategy frameworks help students and practitioners understand the role of a firm’s external environment in capturing value. Michael Porter’s “Five Forces” framework for industry analysis provides a comprehensive checklist of economic factors that complicate the conversion of created value into captured value. To the extent that a business strategy is designed to mitigate such factors, this may present antitrust concerns. A second important framework emphasizes “added value,” which represents the unique contribution that an individual firm can provide to generate surplus. Successful firms capture value as a consequence of scarcity in the added value framework, and this scarcity could result potentially from anticompetitive actions. The second fundamental concept of business strategy is “competitive advantage,” which focuses on a firm’s ability to create and capture value better than current or future competitors. In that sense, competitive advantage concerns intraindustry heterogeneity in performance and has less to do with generating profitability through the concentration of market power. Competitive advantage is a firm-centered concept and, as such, has been influenced by the academic literature in management, particularly the so-called resource-based view of the firm. This topic also considers the sustainability of competitive advantage—how firms can maintain a superior position over time in the face of potential imitation. Finally, business strategy covers the foundational issue of the “scope” of the firm. A firm’s decisions regarding exactly what activities it will perform (and which ones it will not) are critical components of its overall strategy. Since firms can undertake mergers and acquisitions to alter the set of activities that they perform, a natural connection exists between this topic and antitrust. In the last section of this chapter, we will discuss motivations for firms’ scope strategies using the Value Creation and Capture framework. This approach can help managers and policymakers identify which mergers will likely pose problems from an antitrust perspective.

11.2.  Value Creation and Capture By emphasizing an economics-based approach for business strategy, we evaluate the decisions that firms make in the context of optimizing behavior. Based on the goal of

256   Michael J. Mazzeo and Ryan C. McDevitt enhancing shareholder wealth, that means profit maximization. Immediately, this provides a simple, straightforward metric for evaluating firm strategy—policies that improve profitability, Π, must either generate higher prices, P, reduce average costs, C, or increase quantity sold, Q. A fundamental organizing structure for undertaking strategy is the Value Creation and Capture framework. This framework explicitly incorporates an economic treatment of consumers, whose behavior is of course critical to a firm’s decision-making process. To insert consumers into the framework, we denote the maximum amount that a consumer is willing to pay for a product or service as B. From this, we define the following: • “Value Created” is equal to B − C , and represents the total societal benefit that the firm generates with its product or service. • “Consumer Surplus” is equal to B − P , and represents the share of value created that flows to consumers. Importantly, consumers will choose among alternatives by selecting the product with the highest consumer surplus. • The remainder of the Value Created, P − C , flows to the firms as profits. We call this portion “Value Captured” or “Producer Surplus.” It is the most relevant concept for formulating strategy, as producer surplus falls directly into the profit function:

Π = ( P − C ) *Q .



Necessarily, where P falls will be a crucial determinant of firm profitability, conceptualized in the Value Creation and Capture graph in figure 11.1. The Value Creation and Capture framework suggests two generic approaches for enhancing profits. Profitability is equal to Value Created times the share of Value Created that a firm can capture. So, a firm can increase profits by creating more value— either through cost reductions or by making its product more attractive to consumers. As long as the share captured does not decrease, profits will be higher. In addition, firms can also focus on capturing a greater share of the value they create.

11.2.1.  Enhancing Value Creation The process and approach by which a firm achieves a particular combination of B and C represent a starting point for understanding its strategy. In his influential article “What Is Strategy?” Porter (1996) uses the term “operational effectiveness” to describe how efficiently firms translate C into B. Firms should strive to generate the largest B for a given level of C, or conversely, to achieve a particular level of B at the lowest cost possible. In effect, operational effectiveness is a necessary condition for profit maximization—if a

Business Strategy and Antitrust Policy   257 VALUE CREATION AND CAPTURE $/unit Benefit of product to customer; willingness-to-pay

B

Value that goes to customers as CONSUMER SURPLUS

Price customer pays for product in the market

P

Average (per unit) cost of production

C

VALUE CREATED Value that goes to firm as PROFITS or VALUE CAPTURED

units sold Q Units sold, or number of customers who purchase FIGURE  11.1 Graphical

Illustration of the Value Creation and Capture Framework

firm could be more efficient, all else equal, it could provide the same B at a lower C, or a higher B at the same C. In either case, profits could be higher. Generating profits through operational effectiveness, however, depends critically on what other firms in the market can achieve through their own efficiency. If multiple firms are equally efficient, they may engage in destructive price competition that drives down profits by shifting value created from producer surplus to consumer surplus. As a result, a firm must achieve operational effectiveness through a unique value creation proposition in order to generate robust profitability. If the firm’s strategy is somehow unique, destructive price competition is much less likely. Uniqueness can come through either doing a different set of activities than competitors, or doing the same set of activities in a different way (or both). Porter further argues that selecting a strategy that is both unique and operationally effective should be the goal of all firms. At the same time, firms must be wary that imitators will copy unique and operationally effective business strategies. In this context, a strategy will have a greater chance of resulting in continued profitability in the face of potential imitation if it includes the following: • Trade-offs. A strategy that gains part of its operational effectiveness and/or uniqueness by explicitly excluding specific elements as a part of the strategy is said to exhibit trade-offs. Trade-offs are particularly effective to the extent that potential

258   Michael J. Mazzeo and Ryan C. McDevitt imitators are already established in performing the activities that have been explicitly excluded by the profitable firm. • Complementarities. A complementary exists between two elements of a firm’s strategy ( X and Y ) if the return to doing activity X is higher if the firm also does activity Y (as compared to if it only did X ), and vice versa. To the extent that a strategy has complementary elements, a potential competitor must imitate all of these elements to become an effective imitator. Importantly, achieving this becomes exponentially more difficult as the number of complementary elements in a firm’s strategy increases. Porter refers to this concept as “strategic fit” and describes the set of complementary elements as an “activity system.” It is worth noting that value-creating strategies are much less likely to generate scrutiny from antitrust authorities. Rather, regulators are likely to be more concerned about the split of value created into consumer surplus and producer surplus. Industrial organization economics has a lot to say about the conditions under which consumer surplus will be squeezed to the benefit of firms. Earlier work by Michael Porter helped to bring these issues to the forefront of strategic thinking.

11.2.2.  Shifts to Produce Surplus: Porter’s Five Forces The competitive conditions that influence the split between consumer and producer surplus are the focus of Michael Porter’s influential “Five Forces” framework of industry analysis (Porter 1979). Using microeconomic insights, the framework provides a template that a firm can use to perform a comprehensive audit of all the factors that potentially could reduce the share of value created that firms in its industry can capture as profits. By tying industry profits to economic principles, the Five Forces moves beyond simple rubrics such as industry concentration to assess issues of direct concern to antitrust regulators, such as the competitive effects of mergers. In order to ensure a comprehensive treatment, Porter divides the possible threats to value capture into five categories: industry rivals, potential entrants, substitute industries, buyers, and suppliers. For each of these categories, Porter provides a checklist of economic conditions that tend to strengthen the threat from that group. An analyst can then use this checklist to determine the sources of competitive threats and potentially design strategies around limiting their strength (and thereby increase profitability). The Five Forces framework uses the industry as the unit of analysis; therefore, using the framework to understand a firm’s strategic position requires a clear definition of the relevant industry and market. The tools of industry definition for strategy will be similar to those in antitrust, with particular focus on products and geography. However, while the precise details of market definitions are often critical for antitrust cases, the structure of the Five Forces framework renders the stakes relatively low for strategy. To the extent that analysts define the market narrowly, there will be less industry rivalry. But,

Business Strategy and Antitrust Policy   259

the analyst will still account for all the parties that affect the industry’s profits in their consideration of substitute products. Rivalry stands at the center of the Five Forces framework, and all the other forces point towards it. Most forms of rivalry result in price competition that reduces industry profits. The presence of several industry characteristics often coincides with intense industry rivalry or indicates that one of the market’s participants might have a strong incentive to cut its prices, potentially initiating a price war. For instance, in a market characterized by high inventory costs, firms may cut prices to unload products nearing obsolescence. In a market with high fixed costs, firms may lower prices to increase quantity sold and reduce average costs. For undifferentiated products, firms often compete on price to distinguish themselves. To counteract each of these industry characteristics, firms may pursue policies that run afoul of antitrust regulations. To avoid the destructive nature of obsolescence, firms may join trade organizations that set standards, however implicit, for product redesigns and innovation cycles. To combat the incentive to move down the average cost curve, firms may collude to set quotas. To instill a measure of differentiation, firms may divide up exclusive territories. In Porter’s framework, rivalry refers only to the firms operating in the same industry or market. The other forces describe the competitive strength of less-direct competitors, such as those that sell substitute products. When one of these other forces is strong, downward pressure on prices may also result. For example, in an industry that does not have intense rivalry, we might expect more firms to join the market and subsequently increase industry rivalry (Bresnahan and Reiss 1991). Firms may not suffer from this competitive threat if other firms cannot easily enter the market—that is, when barriers to entry exist. Using the Value Creation and Capture framework, a strategist would think of a barrier to entry as any factor that increases the costs of new firms in the market relative to established firms, or that increases a consumer’s willingness-to-pay for an incumbent’s products relative to new entrants’. Features such as economies of scale, product differentiation, switching costs, and access to distribution channels all affect a potential entrant’s value creation upon joining a market and, consequently, its ability to compete with incumbents. Again, antitrust concerns may readily apply in situations where firms take actions to forestall entrants. For instance, Microsoft’s decision to bundle Internet Explorer with Windows effectively reduced the relative value created by Netscape in the market for browsers. In the event that an industry (1) does not have intense rivalry, (2) has meaningful barriers to entry, and (3) lacks compelling substitute products, two remaining forces still may dampen profits. First, if consumers (here, “buyers”) are more powerful, they may be able to negotiate lower prices and capture a larger share of the value created. Some of the factors related to buyer power directly reflect the competitive conditions outlined under rivalry. For example, with more or undifferentiated rivals, it is easier for buyers to pit one industry rival against another for price negotiations. Other factors relate to the ability or motivation buyers have to negotiate more intensely with firms in the industry. A buyer, or group of buyers, is typically better able to negotiate lower prices if it purchases a large

260   Michael J. Mazzeo and Ryan C. McDevitt volume of the seller’s output, if it has full information about the deal’s specifications and past transactions, or if it earns lower profits.1 A company can improve its strategic position by serving customers that do not possess much negotiating power, a tactic known as buyer selection. Cartels represent a classic example of attempts to reduce buyer power that violate antitrust acts. Analogous conditions will permit powerful suppliers to extract profitability from a given industry. If suppliers have power, they can force price increases (or quality reductions) onto firms in the industry, which increases their production costs. Supplier power is unique among the Five Forces insofar as the effect is on value created and not the share of value that is captured. Nonetheless, there may be regulatory concerns regarding strategies that aim to limit supplier power, as they transfer value created into one industry from another. Note that labor must be recognized as a supplier as well, and it may exert substantial power in many industries. Recently, Apple, Google, Intel, and others faced antitrust scrutiny for conspiring not to recruit each other’s employees (Helft 2010).

11.2.3.  Added Value In a number of productive ways, strategists have incorporated the principles of game theory to bolster the Value Creation and Capture framework. A particularly prominent concept from this evolution is added value, which is defined as the total economic value created less the counterfactual value that would be created absent a given participant (Brandenburger and Nalebuff 1998). Added value is greater in circumstances where more value is created and in situations where an agent’s contribution is scarce—that is, when no other participant can generate the same value. A participant’s added value then bounds the amount of value that he can capture; intuitively, a player cannot take away more than he brings to the table. This reasoning maps directly to the Five Forces, particularly buyer and supplier power, and the negotiating ability of firms to transfer surplus. Firm tactics aimed at increasing added value may straddle the line of antitrust regulations. For instance, restricting the number of licenses available for a given technology protects its scarcity value, but may draw the attention of regulators who may view it as anticompetitive. Strategy frameworks that analyze competition using economic principles provide analysts with a more comprehensive picture of its industry’s prospects for profitability. Furthermore, an effective competitive strategy does not just accept the industry assessment, but creates a defendable position against the Five Forces. This is the heart of competitive advantage outlined in the following section. Distinguishing between value creation and value capture provides a useful construct for thinking about the types of

1  For example, Dafny (2010) finds evidence that firms with positive profit shocks subsequently pay higher health insurance premiums. The interpretation is that the firm’s relative profitability will motivate them to fight harder or less hard to get a good deal on health insurance for their employees.

Business Strategy and Antitrust Policy   261

strategies on which antitrust authorities will focus. We will continue with this theme in the next section as well. Before leaving the topic of value creation and capture, it is worth noting that, from a business strategy perspective, decisions are typically based on maximizing long-term profitability. As a result, a firm may not appear to be maximizing short-run profits in certain contexts related to pricing, R & D, network effects, and so on (Oster 1999). In contrast, many antitrust analyses are more explicitly short-run in nature.2 We can see this most readily in empirical merger evaluations, such as Nevo (2000), where price effects are simulated assuming a change in ownership for previously competing differentiated products but no change in the merged firm’s product portfolio. Efforts to incorporate longer-term product decisions in merger analyses (Draganska, Mazzeo, and Seim 2009, Fan 2013) represent a way to bridge the antitrust and strategy literatures, and is an important avenue for future research.

11.3.  Competitive Advantage and Sustainability The Value Creation and Capture framework provides an analytical link between economics and firm strategy. An analysis of competitive advantage takes the next logical step in developing a structure for evaluating the success of individual firms. This framework helps us understand—and predict—why firms in the same industry that face the same underlying economic forces may nevertheless have very different profits. Perhaps more importantly, this framework allows us to diagnose the sources of a firm’s superior (or inferior) profitability to guide strategic decisions. As such, the competitive advantage framework focuses on the individual firm, not its industry. We will consider how the firm’s “resources” contribute to its capabilities, and how these capabilities explain its performance.3 To start, we need a precise definition for competitive advantage: the resources or capabilities that allow a firm to capture value better than existing or potential competitors. We think of these “resources” in fairly general terms. They are anything that directly affects the quality, costs, and other attributes of a firm’s product or service. Examples of resources could include tangible elements like location, physical plant/ equipment, or product offerings, as well as intangibles such as brand identity, people/ culture, relationships, and so on. “Capabilities” represent the activities that a firm’s resources enable it to do. Some analysts will collectively refer to a firm’s resources and capabilities as its “assets.” 2 Ghosal (2011) discusses this phenomenon and provides relevant citations.

3  This contrasts with Porter’s Five Forces framework, which considers the conditions that make an entire industry more or less profitable, on average.

262   Michael J. Mazzeo and Ryan C. McDevitt A key assumption here involves heterogeneity among firms’ resources and capabilities. All firms are different—even firms producing goods that are very close substitutes may produce them in very different ways. We need to build on the notion of heterogeneity to consider competitive advantage: a firm must possess some resource that leads to a unique capability that, in turn, results in superior performance compared to its rivals. It is important to emphasize that a useful analysis of competitive advantage must at all times be framed in terms of comparisons between a firm and its rivals. With this perspective, firms may confront challenging, and potentially uncomfortable, realities— even activities that a firm does well may not be superior, as compared to its rivals.4 Firms pursue three broad types of advantages, with the first being a cost-based advantage. Consider an industry with many firms. Even if they produce the same product, the firms may be heterogeneous in terms of the resources they use to produce their output. A firm with superior resources may produce the good at a lower cost than its rivals. If there is not enough low-cost capacity to satisfy market demand, the low-cost firm will be able to earn the difference between its own costs and its rivals’ (which will determine market prices) as economic profits. A firm with a worse set of resources may just break even. A cost-based advantage may arise from a number of sources. A firm may increase its size or scope: economies of scale, economies of scope, volume purchases, increased capacity utilization, and specialization all potentially reduce a firm’s costs. Relatedly, a firm may increase its cumulative experience, which can bring down marginal costs in a setting characterized by a learning curve. Firms may increase their organizational efficiency through vertical integration, long-term contracts, or management and control. Finally, a firm may increase its technological efficiency through automation, production processes, coordination, transportation, or communication. To the extent that cost-based advantages derive from efficiencies, antitrust authorities typically remain passive. When the advantages come from wielding buyer power or from taking actions that raise rivals’ costs, however, scrutiny may result. Firms may also pursue a benefit-based advantage, a natural companion to cost-based differentiation. Typically, firms have to make trade-offs related to the price-quality preferences of consumers because increasing quality involves increasing costs. Suppose, however, that a firm possessed the resources and capabilities to produce a higher-quality product at a lower or similar cost than rivals. Charging the same price would not maximize the firm’s profits in such a case. The firm could increase its price and not lose customers—consumer surplus, B − P , for the firm’s superior product would still exceed competitors’. Benefit-based advantages stem from various origins. For instance, a firm may improve the physical characteristics of its product by improving performance, durability, quality, features, aesthetics, or ease of use. A firm might also be able to increase 4  The related concept of core competencies puts somewhat less emphasis on comparison across organizations, but nonetheless has contributed to the development of competitive advantage (Prahalad and Hamel 1990).

Business Strategy and Antitrust Policy   263

the quality of complementary goods such as postsale service, spare parts, warranties, maintenance and repair, or characteristics associated with sale or delivery in regards to timeliness, convenience, and the quality of sales staff. Benefit-based advantages could also result from factors that influence customers’ perceptions or expectations in terms of reputation, an installed base of users, and network externalities. Finally, a firm may improve its subjective image through prestige, status, or association. A firm’s key challenge when pursuing a benefit-based advantage is to choose which product characteristics to improve. Consumers must be made better off despite the price increase, which is necessary to offset the firm’s higher costs. When such price increases are associated with firms having a large market share following a merger, antitrust authorities may become concerned. For example, practices such as bundling and tying may result in higher prices but nonetheless improve consumer welfare. Uncertainty regarding the treatment of such outcomes may yet exist, as the proper role for consumer welfare in competition policy remains the source of considerable debate (see, e.g., Crandall and Winston 2003). Finally, firms can pursue a niche-based competitive advantage in which the firm produces a good that some consumers prefer over alternatives at the same price—that is, customers have heterogeneous preferences in the sense that not everyone would purchase the same product at the same price. In a way, this firm is somewhat like a monopolist:  it occupies a unique location in “product space.” Once the firm has chosen its product-space location, the resources and capabilities that allow the firm to occupy it efficiently generate the niche-based advantage. As in Hotelling (1929), a strategist would conceptualize product heterogeneity as the distance between the firm’s location in product space and its nearest rival’s location. Because of their heterogenous preferences for different varieties, consumers do not consider competing products perfect substitutes. As a result, a firm that is differentiated in product space can maintain a price above costs without losing all of its market share to competitors. Since price exceeds cost, the firm with a niche-based competitive advantage earns an economic profit, P − C > 0. The size of the subsequent profit margin depends on the intensity of consumer preferences relative to other available substitutes in the market. Note that “positioning” in a market cannot deliver, by itself, a niche-based competitive advantage. The firm’s offering must (1) be unique relative to the competition and (2) have sufficiently high demand to cover its fixed costs of production. These conditions require the firm to possess distinct resources and capabilities. There may, in this case, be an inherent conflict between the goals of strategy and antitrust policy. As demand grows, what was once a profitable “niche” can potentially be construed as a “market,” resulting in scrutiny by regulators. The questions of market definition presented in the previous section become especially relevant in such circumstances. While the competitive advantage framework is not nearly as formal as Porter’s Five Forces, the underlying microeconomic foundations remain critical. The key conceptual touchstone for competitive advantage is consonance. Successful firms have consistency among the activities they pursue, the resources they possess, and the capabilities that

264   Michael J. Mazzeo and Ryan C. McDevitt these resources confer. Firms can achieve greater consonance by undertaking strategic initiatives that are consistent with the resources and capabilities they possess and by developing and acquiring resources that fit well (i.e., are co-specialized) with their existing resources. While consonance may be difficult to measure or quantify, some strategists believe that systems of activities that reflect this internal consistency are crucial for the success of firm strategies.5

11.4.  Sustainability of Competitive Advantage To this point, our discussion of competitive advantage has taken mostly a static view. We have looked at how a firm might generate a competitive advantage, but have not dwelled on whether or not the firm would achieve only a short-term gain. Clearly, market conditions change, and a robust competitive advantage framework must incorporate a dynamic perspective into firm strategy. As mentioned previously, successful strategies will attract imitators. From an evolutionary perspective, this increases social welfare: good strategies replicate, bad ones die out. From the perspective of a firm that currently enjoys a profitable competitive advantage, however, the threat of imitation looms large. The types of strategies that a firm employs to protect its competitive advantage over time ultimately will determine its success. These strategies can also be troubling and problematic from an antitrust perspective. The competitive advantage framework outlined above suggests that a firm can sustain its competitive advantage by protecting the resources and capabilities responsible for generating it. In that spirit, the following set of conditions represents minimum and necessary conditions for a firm to maintain its competitive advantage in the long run: limits to resource competition, limits to resource acquisition competition, and resource immobility (See, for example, Peteraf (1993) and Wernerfelt 1984). Limits to resource competition are key to preserving a competitive advantage. Subsequent to a firm gaining a superior position and earning profits, a protection must be in place to limit the competition for those profits. Possessing uniquely valuable resources drives competitive advantage; therefore, if another firm can obtain the same resources, the competitive advantage will not persist. Two important aspects further distinguish the limits to resource competition. First, isolating mechanisms prevent other firms from copying the resources responsible for a firm’s superior profits. An isolating mechanism prevents firms from imitating either the production efficiencies or the characteristics of the end product (of the

5 

One recent example can be found in Leinwand and Mainardi (2010).

Business Strategy and Antitrust Policy   265

superior firm) that make it uniquely desirable to users. Some examples of isolating mechanisms are formal, such as property rights, patents, or copyrights. These legal protections prevent others from using the resource. Indeed, resources such as patents and copyrights explicitly confer monopoly power—society is willing to grant pricing power as an incentive for innovation (presumably with a net benefit for consumer welfare). However, some argue that, especially in industries susceptible to monopolies, firms can abuse the legal protections of patents to protect themselves from competition.6 Other isolating mechanisms may be less formal but can be equally effective in protecting resources and competitive advantages. For example, if scale is crucial for achieving lower costs, an entrant may not be able to achieve the same size as incumbents. Intangibles such as culture and reputation are particularly difficult to replicate. By definition, culture and reputation need time to grow and can become stronger over time. In the strategy field, we often point to Southwest Airlines as the prototypical example, though many others exist as well (O’Reilly and Pfeffer 1995). Firms must also guard against resource substitutability: situations where competitors may possess a different resource that delivers the same advantage. The rival firm is not copying a resource per se, but it nevertheless achieves the same result. For example, if another farmer developed a very low-cost fertilizer, a firm that owned more-fertile land than its rivals would no longer have a competitive advantage (even though his property right to the fertile land remains intact). An additional condition necessary for sustaining profits is that other firms do not foresee the value that the resource will create—that is, the firm has strategic foresight. Consider the alternative: if all firms recognize a resource’s value, competition for the resource will drive up its acquisition price to a point that offsets any profits generated by it. This is why, in general, strategists are skeptical that “exclusive” arrangements from suppliers will be profitable. Obtaining exclusivity should be costly, assuming competition. As such, this is the place to look to ensure that markets are sufficiently competitive; indeed exclusive dealing is a very rich area for antitrust theory and practice (Marvel 1982). The third necessary condition for a firm to achieve a sustainable competitive advantage is resource immobility. Immobility generally requires that the superior resource cannot be profitably traded. If the resource would be more productive in the hands of another firm, then the firm that does control it is not maximizing economic profits. For instance, co-specialization occurs when a resource must be used in conjunction with other firm-specific resources in order to create the most economic value. In cases where resources are not co-specialized, a firm can benefit its shareholders by trading the asset (even if it is a profitable asset). Presumably, a more co-specialized firm would be willing to pay a premium to acquire the asset, and the net profit from the trade would contribute positively to the value of the firm.

6  This argument is laid out nicely in Feldman (2003). It is being tested in the contemporary strategic and legal battles playing out in the handheld device industry (Catan 2011).

266   Michael J. Mazzeo and Ryan C. McDevitt In addition to resource protection, established firms can extend their sustainable competitive advantage through accumulated market experience. Here, we can again turn to the Value Creation and Capture framework for conditions under which operations in the past can either lower C or increase B . Such factors would enhance the firm’s prospects for maintaining profitability by creating additional value. A firm’s learning curve describes any situation in which cumulative production experience reduces a firm’s average variable costs. Note the distinction between the learning curve and economies of scale—an experienced firm (with learning curve economies) would have lower costs at any particular scale of production. Firms with a steep learning curve may attempt to underbid rivals for business or subsidize consumption initially in order to build up their cumulative experience. Note, however, that just as with scale economies, a firm might reach a point of diminishing incremental cost savings at very high levels of cumulative experience. When learning curve economies are particularly compelling, it may be sensible for firms to engage in predatory pricing behavior. Analysts have noted that these cases are often very difficult to prove (Cabral and Riordan 1997). For some products, a consumer’s willingness to pay is partially determined by the total number of consumers who use the product. Here, there are “network externalities”: a “network”of users creates an “external” benefit to additional consumers.In this environment, firms can gain an advantage by building up sales in early periods and developing a large “installed base” of users who have purchased the product in previous periods and still use it. Studies have shown that if network externalities are strong, there may be a welfare benefit associated with a monopolist, complicating antitrust analysis (Katz and Shapiro 1985). Finally, switching costs can increase the effective price of a new product relative to an established one, conferring an advantage to a firm that has achieved more sales in earlier periods. Clearly, a strategy of building switching costs into a product or encouraging early adoption can permit firms to extend their competitive advantage over time. As Farrell and Klemperer (2007) point out, this can lead to competition “for the market,” and competition policy behaves somewhat differently. Note, in addition, that forward-looking consumers will take the effects of switching costs into account when they make their initial purchase. Knowing that a firm will have them “locked in” and vulnerable to price increases in future periods renders consumers more wary at the initial point of purchase. This may limit the potential efficacy of such strategies.

11.5.  Scope of the Firm Among the specific topics that we address using the principles of economics in strategy, none is more fundamental and relevant than issues surrounding the scope of the firm. The first substantial section of the Besanko and coauthors (2009) strategy textbook, for

Business Strategy and Antitrust Policy   267

example, covers firm boundaries—both the vertical boundaries of the firm, as well as questions related to diversification. This partly reflects the historical context—among the earliest influential strategy frameworks from consulting was the BCG “growth/share matrix” that classified a firm’s business units based on their market share and growth prospects, identifying some for investment and others for divestiture. In addition, evaluating potential mergers and acquisitions is a principal responsibility of the strategy group within many organizations. In using economics as the underlying framework for making strategic decisions, we recall Coase’s definition of a firm’s role as “organizing transactions” for the economy. Coase (1937) suggests that a firm’s decision about its scope should address the question, “Why does the entrepreneur not organize one less transaction, or one more?” Of course, the answer to this question depends on the context considered—a particular firm’s resources, products, and relevant markets will determine how it should set its boundaries in order to maximize profits. In strategy, we focus on the economic issues that affect a firm’s decision to conduct certain tasks internally or to “use the market” instead. In so doing, we embellish the Value Creation and Capture and Competitive Advantage frameworks to address how a firm should organize its activities to maximize profits and shareholder wealth. On the surface, strategic decisions regarding economies of scope can be evaluated using another simple, straightforward rule: a single firm should perform two activities, X and Y , if and only if the profits from doing both activities within a single firm exceed the profits from doing each activity across two distinct firms:

Π ( X + Y ) > Π ( X ) + Π (Y ) .



Otherwise, one firm should perform activity X, and a separate firm should perform activity Y . We can think of activities X and Y in fairly general terms—they could refer to the same activity in different geographic markets, differentiated products, rival firms in the same industry, or completely unrelated activities. When practitioners use the term “synergy,” they are essentially referring to the idea behind this comparison of profit functions—there would be a synergy between activities X and Y if profits increased when those activities were done within the same firm. In other words, the “synergy” is the explanation for why profits are greater when activities are combined.7 7  In this spirit, a firm derives no synergy if it vertically integrates simply to “obtain the activity at cost” (i.e., to avoid paying a supplier’s high markup) or if it horizontally integrates to obtain the buyer’s or supplier’s profit. The buying and selling of firms occurs in a market—to the extent that “excess profits” exist, the potential integrator would have to pay for these profits in the acquisition price. While “bargains” may be found, an acquisition premium typically drives final bids to the second-highest value among the potential acquirers. Even if it seems like the combined firm is obtaining the activity for itself “at cost,” this masks opportunity costs, in terms of what the firm would earn by selling the output at the market price.

268   Michael J. Mazzeo and Ryan C. McDevitt By linking synergies directly to the profit function, we can immediately categorize potential ways in which firms could increase profits by combining activities. The combined activities must result in at least one of the following for the firm—higher prices, lower costs, or greater demand (quantity)—and must do so without counteracting the gains with offsetting losses on other dimensions. From the firm’s perspective, identifying what these potential synergies are (and attempting to quantify them) is crucial for making effective strategic decisions. However, as long as synergies exist and are substantial, it would not matter from the firm’s perspective where the important synergies associated with a merger derive. Of course, from an antitrust perspective, the source of the synergy is crucial for the regulatory evaluation of a merger. Of particular concern are circumstances in which a firm may achieve price-based synergies from a potential acquisition that reduces competition in a market. In the United States, the Hart-Scott-Rodino Act stipulates that the parties involved in substantial mergers and acquisitions must notify the Federal Trade Commission and the Department of Justice before commencing the proposed transaction. In their notification filing, the parties provide information about the industry and their respective firms that may be used (along with additional data and detailed analysis) to determine whether the merger will reduce competition and cause harm to consumers. These agencies may bring legal action to block potential mergers in such circumstances, though these cases often settle prior to litigation following mutually agreed-upon conditions aimed at limiting the potential ex post harm to consumers. For example, a settlement was reached in 2007 between the Federal Trade Commission and two merging northeastern US supermarkets, A&P and Pathmark. Of the roughly 450 stores in the combined company, it was agreed that six in specific towns in New York State would be sold in order to ensure that consumers would not face substantial price increases after the merger.8 In a similar vein, firms may be able to raise revenues (or reduce costs) by expanding their portfolio of activities to increase their negotiating power over a common buyer (or supplier). This strategy may be achievable even if the firms are producing noncompeting products—the idea is that in a negotiation, a firm’s threat point is more damaging if the firm constitutes more of the buyer or supplier’s overall business. The merger between Gillette and Procter & Gamble is an illustrative example; by merging, the firm could potentially increase its negotiating power with powerful retailers such as Walmart based on the size of its overall portfolio of products sold there. The regulatory authorities considered a “portfolio effect” along with concentration in particular product categories where the firm offered competing products premerger. Revenue increases that come from shifting surplus to the merged firm from another firm rather than from

8  The FTC commented that “Absent the relief provided by the Commission’s consent order, consumers in these areas likely would face higher prices and lower levels of service when shopping for their weekly groceries.” Cf. http://www.ftc.gov/opa/2007/11/pathwork.shtm.

Business Strategy and Antitrust Policy   269

consumers may well generate less regulatory scrutiny, particularly (as cited in this case) where the countervailing power of buyers is also quite high.9 This detailed level of regulatory concern and analysis suggests that firms may not be able to rely solely on price synergies as justifications for mergers. As a consequence, cost-based synergies and demand-related benefits become more important for managers that are evaluating opportunities to add further activities to their firm’s scope. For demand synergies, we attempt to take account of situations in which the products or services offered by a firm would generate a greater willingness-to-pay for consumers as a consequence of the firm increasing its scope and adding an activity. The 2004 merger between health insurers Anthem and WellPoint illustrates this concept. Providing health insurance requires making contractual arrangements with local providers and receiving regulatory approval from individual states. Prior to their merger, each firm offered such services in a network of states that did not overlap. Thus, the merger would not eliminate a competitor in any geographic market. However, the merger did increase the network of states in which the combined firm offered coverage. An insurer with a broader geographic coverage network is likely more attractive to an employer with a presence (and employees that need to be covered) in multiple states. There is a potential for higher prices as a consequence of such a merger, but the price increases would likely come from having created more value rather than from a change in the share of the value created that is captured by firms. In practice, firms rarely rely on these sorts of demand-side synergies to justify merger decisions. Situations like the Anthem-WellPoint merger are rare, and even if a demand expansion seems compelling, it may be difficult to quantify. Furthermore, alternatives to integration such as joint marketing agreements can often be employed when independent firms want to take advantage of demand-side complementarities between products or services that they own. Unless there are incentive or informational issues that would decrease the efficacy of a contractual arrangement, it is often a better strategy to avoid merging activities to exploit demand-side synergies. This leaves cost-side synergies, which are typically the most straightforward for firms to quantify premerger and the least objectionable from the perspective of the regulatory agencies. Synergies on the cost side require some kind of reduction in costs that comes from the integration of multiple activities within the same firm. In the case of within-industry horizontal mergers, cost savings would be associated with traditional economies of scale; the extent of the synergy would depend on premerger capacity utilization and the scalability of the underlying activities. If a firm would have to incur additional fixed costs as a result of the combination, any potential synergy would be offset to some degree. 9  In its decision on this merger, the European Commission noted that “The risk of portfolio effects resulting from the merger is mitigated considerably by the ability and incentive of retailers to exercise countervailing buyer power. Large retailers can exert pressure on the parties as they can more credibly threaten to integrate private labels on their shelves of by sponsoring new entry through active in-store promotion.” Cf. http://ec.europa.eu/competition/mergers/cases/decisions/​ m3732_20050715_20212_en.pdf.

270   Michael J. Mazzeo and Ryan C. McDevitt In other horizontal combinations, the source of cost-side synergies may be more subtle. Conceptually, however, the notion of shared production infrastructure remains. For example, firms in different industries that sell to similar consumers may be able to share customer lists or generate other marketing efficiencies. Managerial infrastructure can also potentially be shared, especially in cases where the merged activities have a similar overall strategy. These factors are illustrated in the long-standing mutually beneficial combination within PepsiCo of a soft-drink division and snack foods division (Frito Lay). The company historically has not attempted to recognize synergies through negotiations or joint promotions; in fact, the divisions have operated completely independently from each other. Instead, the company cites managerial efficiencies at the highest levels of the organization that come from overseeing businesses with a common approach (particularly relying on national and international marketing). Indeed, PepsiCo executives typically rotate through both divisions at various points in their careers, gaining experience in their shared endeavors. While most of the discussion above has focused on horizontal combinations, the same analysis of value creation and capture applies to vertical combinations. In this sense, the distinction between horizontal and vertical integration is somewhat arbitrary; however, subtle issues that affect firms as they extend their vertical scope warrant special consideration in strategy formulation. In particular, considering vertical integration provides us the opportunity to introduce incentive and organizational issues that are critical to effective strategy. These issues are relevant even for firms that do not typically consider mergers and acquisitions an important part of their overall strategy. Indeed, for most firms, vertical combinations reflect a fundamental strategy consideration, insofar as they must decide whether to perform activities along the vertical chain themselves or “purchase” services from independent firms in the market, if they even want to be in the business at all. As such, we pose the vertical integration question in the context of a “make versus buy” problem. A key consideration, then, is the opportunity cost of not integrating. For example, if the downstream firm does not produce one of its inputs, it must purchase that input from an upstream supplier. For the firm to pursue a vertical integration strategy, the overall profits associated with producing that input must exceed the overall profits associated with purchasing it from the market instead. Again, much of the focus in analyzing the make-versus-buy problem is about value creation. Often, straightforward cost-based reasoning applies. For example, a firm should outsource activities for which it does not have enough scale when (competitive) market specialists are larger, whereas vertically integrating may allow firms to avoid the (often expensive) transaction costs associated with exchange across firms. For example, the coordination of production flows may be compromised when a firm purchases an input from the market. There may be additional inventory costs if the inputs arrive too early, or costly idle time if they arrive too late. Dealing with such problems using technology or contracting may be more difficult if transactions take place outside the firm. In addition, monitoring in a vertically integrated organization may prove challenging. We introduce the concept of agency costs in terms of the potentially misaligned incentives along the vertical chain. The “cost” to the firm represents the forgone profits

Business Strategy and Antitrust Policy   271

when incentives are not appropriately aligned. For example, franchising relationships are common in industries like lodging and food service because success (at least in part) requires intense managerial effort that is difficult to monitor and reward. If technological improvements allow greater monitoring (as in the trucking industry), it may be optimal to bring once-outsourced activities in-house (Baker and Hubbard 2003). Beyond these efficiencies, however, there is scope for vertical integration to increase a firm’s profits without creating additional value. As an example, such a situation can occur when an upstream monopolist sells to a competitive industry. By vertically integrating with one of the downstream competitors, the firm could, in effect, extend its monopoly to another industry by limiting access of its unique input to only its integrated division. Providing downstream rivals with the input, but at less attractive terms, would accomplish a similar goal by raising the rivals’ (relative) costs. In some markets, limited access to a key input could generate barriers to entry that help maintain the profitability of the downstream firm, though attempts at market foreclosure often incite regulators. As with horizontal integrations, we see that applying the business strategy objective of profit maximization can lead to vertical integrations because of efficiencies and because of greater opportunities to capture value that could generate antitrust scrutiny. To the extent that either could increase profits, there would be no need to distinguish between them conceptually. A manager must, however, confront the reality that the firm’s activities may be constrained by regulators enforcing competition laws. We believe that the Value Creation and Capture framework provides a useful first step in distinguishing benign value-creating explanations for acquisitions from ones that may prove problematic based on reliance on value capture synergies.

11.6. Conclusion At the core of business strategy lies the Value Creation and Capture framework. Owing to its central importance, we focused our discussion of strategy’s relevance to antitrust policy on its principal tenets. From the firm’s perspective, a firm may seek to maximize profits by creating more value or by capturing the largest possible share of this value. It is this latter objective that most concerns antitrust authorities. Distinguishing between the economics of value creation and value capture can help practitioners understand how antitrust policy might constrain their activities. The remaining topics of industry analysis, competitive advantage, and firm scope build on the basic strategic foundation of creating and capturing value. An additional aim of our chapter was to provide a link between antitrust policy and business strategy, reframing intuition within the particular lexicon of strategy. While these frameworks influence strategy academics and practitioners, their terminology may be unfamiliar to those outside of this community. Understanding the foundations of strategy frameworks

272   Michael J. Mazzeo and Ryan C. McDevitt and how they are used by practitioners can potentially help regulators apply and interpret antitrust policies in applied business situations. Given the limitation of this review chapter, we could not possibly address every relevant topic. Chief among the omitted areas is understanding how firms tactically engage with antitrust, which remains an open area of research in strategy and economics. Private firms may file antitrust suits against competitors, and doing so often represents a strategic choice to gain an advantage. For instance, many have speculated that Microsoft played an important role in recent antitrust investigations against Google (Gans 2010). Given the nature of competition between Microsoft and Google across several interrelated markets, the decision by Microsoft to act as a complainant represents a deliberate tactical choice. Thinking through the optimal strategy in this regard will be an active area of research for the near future.

References Baker, G., and T. Hubbard. 2003. Make versus Buy in Trucking: Asset Ownership, Job Design, and Information. American Economic Review 93(3): 551–72. Besanko, D., D. Dranove, M. Shanley, and S. Schaefer. 2009. Economics of Strategy. New York: Wiley. Brandenburger, A., and B. Nalebu. 1998. Co-opetition. Crown Business. Bresnahan, T., and P. Reiss. 1991. Entry and Competition in Concentrated Markets. Journal of Political Economy 99: 977–1009. Cabral, L., and M. Riordan. 1997. The Learning Curve, Predation, Antitrust, and Welfare. Journal of Industrial Economics 45(2): 155–69. Catan, T. 2011. When Patent, Antitrust Worlds Collide. Wall Street Journal, November 14. Coase, R. 1937. The Nature of the Firm Economica 4(16): 386–405. Crandall, R., and C. Winston. 2003. Does Antitrust Policy Improve Consumer Welfare? Assessing the Evidence. Journal of Economic Perspectives 17(4): 3–26. Dafny, L. 2010. Are Health Insurance Markets Competitive? American Economic Review 100(4): 1399–431. Draganska, M., M. Mazzeo, and K. Seim. 2009. Beyond Plain Vanilla: Modeling Joint Product Assortment and Pricing Decisions. Quantitative Marketing and Economics 7(2): 105–46. Fan, Y. 2013. Ownership Consolidation and Product Characteristics: A Study of the US Daily Newspaper Market. American Economic Review 103(5): 1598–628. Farrell, J., and P. Klemperer. 2007. Coordination and Lock-In: Competition with Switching Costs and Network Effects. In M. Armstrong and R. Porter, Handbook of Industrial Organization, vol. 3: 1967–2072. Feldman, R. 2003. The Insufficiency of Antitrust Analysis for Patent Misuse. Hastings Law Journal 55: 399. Gans, J. 2010. Microsoft’s New Antitrust Strategy. Harvard Business Review Blog Network, September 28. Ghosal, V. 2011. Regime Shift in Antitrust Laws, Economics, and Enforcement. Journal of Competition Law and Economics 7(4): 733–74.

Business Strategy and Antitrust Policy   273

Helft, M. 2010. Tech Firms Said to Be in Talks to Settle Inquiry over Recruiting. New York Times, September 17. Hotelling, H. 1929. Stability in Competition. Economic Journal 39(153): 41–57. Katz, M., and C. Shapiro. 1985. Network Externalities, Competition, and Compatibility. American Economic Review 75(3): 424–40. Leinwand, P., and C. Mainardi. 2010. The Coherence Premium. Harvard Business Review. Marvel, H. 1982. Exclusive Dealing. Journal of Law and Economics 25(1): 1–25. Nevo, A. 2000. Mergers with Differentiated Products:  The Case of the Ready-to-Eat Cereal Industry. Rand Journal of Economics 395–421. O’Reilly, C., and J. Pfeffer. 1995. Southwest Airlines: Using Human Resources for Competitive Advantage. Stanford Graduate School of Business. Oster, S. 1999. Modern Competitive Analysis. New York Oxford University Press. Peteraf, M. 1993. The Cornerstones of Competitive Advantage: A Resource-Based View, Strategic Management Journal 14(3): 179–91. Porter, M. 1979. How Competitive Forces Shape Strategy. Harvard Business Review, March– April, 1–10. Porter, M. 1996. What Is Strategy? Harvard Business Review, November–December, 1–20. Prahalad, C., and G. Hamel. 1990. The Core Competence of the Corporation. Harvard Business Review 68(3): 79–91. Scott Morton, F. 2003. Why Economics Has Been Fruitful for Strategy. International Library of Critical Writings in Economics 163: 34–39. Wernerfelt, B. 1984. A Resource-Based View of the Firm. Strategic Management Journal 5(2): 171–80.

PA R T I I

V E RT IC A L I N T E G R AT ION A N D C ON T R AC T UA L E QU I VA L E N T S

CHAPTER 12

R E S A L E P R IC E M A I N T E NA N C E OF O N L I N E R E TA I L I N G BENJAMIN KLEIN

The growth of online retailing has magnified the competitive economic forces that are likely to lead manufacturers to adopt resale price maintenance. Because manufacturers are, in effect, purchasing retailing services at a cost equal to the retail margin, online retailer price discounting that reduces the retail margin reduces an implicit manufacturer cost of doing business. However, manufacturers do not merely “throw their products on the market” and leave it entirely up to price competition among retailers to minimize retailing costs. It is an obvious business reality that manufacturers actively determine what they consider to be the most effective retail distribution for their products, including the number and type of retailers that sell their products and the point-ofsale services they expect those retailers to provide. These retail distribution decisions, which often are essential determinants of the ultimate success or failure of a manufacturer’s products in the marketplace, are more likely to include the use of resale price maintenance when there is significant online retailing. The common procompetitive rationale for manufacturer control of retailer price discounting, widely accepted in both economics and antitrust law, is the prevention of free-riding. Free-riding occurs when consumers obtain, for example, salesperson assistance or product demonstrations at full-service retailers before purchasing the product at a lower price at a discount retailer that does not provide such services. The existence of such free-riding reduces the incentive for retailers to bear the costs of supplying these valuable services that expand the demand for a manufacturer’s products. Because online retailing substantially magnifies the free-riding problem by permitting consumers to more easily buy products at a lower price online after first obtaining desired services at a full-service brick-and-mortar retailer, resale price maintenance to prevent free-riding has broader potential applicability when there is significant online retailing. However, many resale price maintenance arrangements, including arrangements designed to control the extent of online retailer price discounting, are not motivated by the desire to prevent free-riding consumers from purchasing at a discount retailer

278   Benjamin Klein after shopping at a full-service retailer. Resale price maintenance often is used by manufacturers even when there is not a consumer free-riding problem as a way to preserve effective retail distribution by ensuring a retail margin that encourages retailers to stock and promote the manufacturer’s products. This non-free-riding procompetitive motivation for controlling retailer price discounting is particularly significant when there is online retailing. Even when consumers who purchase online at a low price do not first take advantage of the services provided by brick-and-mortar retailers, uncontrolled online retailer price discounting has the potential to significantly disrupt a manufacturer’s efficient retail distribution network by reducing retailer compensation below the level required for brick-and-mortar retailers (as well as established online retailers) to effectively distribute the manufacturer’s products. In what follows, the standard prevention of free-riding procompetitive explanations for resale price maintenance are contrasted with the procompetitive use of resale price maintenance as a way to preserve effective retail distribution when there is not a free-riding problem. The non-free-riding procompetitive rationale for resale price maintenance is shown to provide insights into the anticompetitive theories of resale price maintenance and to have important implications for United States and European Union antitrust regulation of resale price maintenance, especially resale price maintenance of online retailing. As a result of the US Supreme Court’s adoption of a rule-of-reason antitrust standard for resale price maintenance in Leegin,1 manufacturers in the United States have gained substantially greater flexibility to use resale price maintenance to control online retailer price discounting whether or not consumer free-riding is occurring. In contrast, European Union antitrust law continues to consider resale price maintenance as a “hardcore restraint,” now rebuttable but with the prevention of free-riding as the primary possible offsetting procompetitive efficiency justification. Consequently, it is much more difficult for manufacturers in the European Union to use resale price maintenance to control online retailer price discounting when resale price maintenance is motivated by the preservation of efficient retail distribution and not by the prevention of free-riding.

12.1.  Resale Price Maintenance Controls Consumer Free-Riding 12.1.1.  Resale Price Maintenance Controls Free-Riding on Retail Services The widely accepted economic explanation for resale price maintenance relies on the prevention of an interretailer externality when discount retailers free-ride on the

1 

Leegin Creative Leather Prods., Inc. v. PSKS, Inc., dba Kay’s Kloset, 551 U.S. 877 (2007).

Resale Price Maintenance of Online Retailing   279

services supplied by full-service retailers (Telser 1960). Resale price maintenance eliminates the incentive of consumers to first obtain free services, such as a product demonstration, at full-service retailers before buying the product at low-service retailers because low-service retailers cannot discount price. As a result, retailers capture a greater return on their retailing efforts and are incentivized to supply increased amounts of the retailing services that are desired by consumers and the manufacturer. The potential for consumers to engage in free-riding, and hence this economic motivation for resale price maintenance, is often particularly important when there is online retailing. Many consumers may first obtain retail services, such as examining the product and obtaining salesperson assistance, at a costly brick-and-mortar retailer before buying the product at a lower price online. This phenomenon, described as “showrooming,” has become more prevalent with the growth of mobile phone scanner apps that permit consumers to take a photo of the product bar code at a brick-and-mortar store and instantly see the prices at which the identical item can be purchased at online retailers. In fact, consumers may actually purchase a product online while at the brick-and-mortar retailer. Van Baal and Dach (2005, 75, 81–82) find that 26% of online purchasers visited a brick-and-mortar store before their online purchase and less than 2% purchased online at the online retail website of the brick-and-mortar store they visited. Critics of resale price maintenance have responded to this empirical evidence of online retailers free-riding on brick-and-mortar retailers by noting that there also is a potential for “reverse” free-riding by brick-and-mortar retailers on the services supplied by online retailers. This occurs when consumers first search online to determine what they wish to purchase and then use the valuable information acquired free of charge to purchase the product at a brick-and-mortar store. A survey of marketing studies (Gundlach, Cannon, and Manning 2010) concludes that this second type of reverse free-riding may occur more frequently than the free-riding by consumers who first shop at brick-and-mortar retailers before purchasing at a lower price online. These two forms of free-riding, however, are separate “distortions” that should not merely be netted out against one another. Manufacturers wish to mitigate the malincentive effects from both types of free-riding, preventing both (1) an insufficient amount of face-to-face retailing services provided by brick-and-mortar retailers, and (2) an insufficient amount of information provided for their products on the Internet. Types (1) and (2) are separate potential problems the manufacturer wishes to avoid. A manufacturer that uses resale price maintenance to reduce the first potential problem by incentivizing and protecting brick-and-mortar retailer selling efforts from online retailer free-riding is unlikely to significantly exacerbate the second problem of less than the desired amount of product information available to consumers online. Furthermore, in contrast to insufficient brick-and-mortar retail services, manufacturers sometimes may be able to more easily directly mitigate the second problem by supplying online product information, at their own website, and by contracting for online advertising on other websites (Carlton and Chevalier 2001).

280   Benjamin Klein

12.1.2.  Free-Riding on Brand Image Retail Services While the economic foundation and empirical relevance of the prevention of free-riding is widely recognized by economists and accepted in antitrust law as a competitive rationale for resale price maintenance, a great many cases of resale price maintenance, including the use of resale price maintenance to control online retailer price discounting, do not fit the prevention of free-riding paradigm. Robert Pitofsky (1984, 29) has perceptively asked us to “think for a moment about the product areas in which resale price maintenance has appeared—boxed candy, pet foods, jeans, vitamins, hair shampoo, knit shirts, men’s underwear. What are the services we are talking about in these cases?” Economists have attempted to answer this critique, in the first instance, by expanding the type of services upon which there may be interretailer externalities from consumer free-riding. Specifically, it is posited that retailer decisions to stock a manufacturer’s products provide brand image information to consumers and that a potential interretailer externality exists with regard to the supply of these information services. There are two variants of this brand image externality analysis. The first variant involves the positive externality a high-quality retailer (that is, a retailer with an established reputation for selling high-quality products) is presumed to have on the demand for a manufacturer’s products at all other retailers when the high-quality retailer stocks a product because stocking provides product quality “certification services” (Marvel and McCafferty 1984). Under this theory, some consumers who decide to buy a manufacturer’s products at a discount store do so in part because they first see the product stocked at a full-priced retailer that has a reputation for selling high-quality merchandise. Discount retailers therefore are free-riding on the “certification services” supplied by the high-quality retailer. Because high-quality reputable retailers share part of the return on their “certification services” with other free-riding retailers, high-quality retailers are less likely to stock and therefore “certify” the manufacturer’s products. While this free-riding hypothesis makes theoretical sense, the products at issue in most resale price maintenance cases are established brands that consumers already recognize as high quality without first seeing the products stocked at a reputable retailer. For example, consider Rolex watches. Consumers do not have to see the Rolex watch sold at an authorized high-quality retailer to know that the product is high quality. While manufacturers certainly wish to have their products sold at reputable, high-quality retailers and such retailers are less likely to carry a manufacturer’s products when they are also sold at discount retailers, this is not because the discount retailers are free-riding on quality “certification services” provided for the manufacturer’s products by authorized retailers. As we shall see in section 12.2.3, high-quality retailers are less likely to stock and promote a manufacturer’s products when the products are also sold by discount retailers not primarily because of consumer free-riding on certification services but because of the more fundamental economic reason that the high-quality retailers earn significantly less in these circumstances from selling the manufacturer’s products.

Resale Price Maintenance of Online Retailing   281

The second variant of the retailer brand image services theory involves a negative interretailer externality. Rather than a high-quality retailer creating a positive “certification” externality on a discount retailer’s demand for the manufacturer’s products, a discount retailer is assumed to impose a negative externality on high-quality retailer demand through deterioration of the perceived quality or brand image of the manufacturer’s products when consumers observe the product sold at the discount retailer. To avoid this negative perceived quality externality and the associated reduction in the demand for its products, a manufacturer may wish to have only higher quality retailers sell its products. The manufacturer thereby protects the quality brand image of its products and avoids the association with lower quality products that are sold at discount stores (Jacoby and Mazursky 1984). However, while eliminating this negative externality and protecting a product’s image by preventing distribution through lower quality retailers often may be a legitimate economic motivation for manufacturers, this alone is unlikely to explain manufacturer use of resale price maintenance in such cases. The negative externality theory implies that a manufacturer of high brand name products would want to prevent distribution of its products through retailers that specialize in lower quality merchandise even when such lower quality retailers sell the manufacturer’s products at suggested prices. The negative brand image externality analysis provides an economic rationale for the establishment of a distribution network that avoids lower quality retailers; it does not by itself provide an economic rationale for resale price maintenance.

12.2.  Resale Price Maintenance Preserves Effective Retail Distribution When There Is No Free-Riding 12.2.1.  Retailers Have Discretion over Retail Services That Have Interbrand but Not Interretailer Demand Effects The economic rationale for manufacturer use of resale price maintenance in cases where there is no obvious consumer free-riding (Klein 2009) relies, first of all, on the fact that retailers often have considerable discretion in deciding which particular manufacturer’s products they will stock, prominently display, or otherwise promote. Retailer discretion exists with regard to such product promotional decisions because in many cases these retailing promotional services have primarily interbrand demand effects with little or no interretailer demand effects.

282   Benjamin Klein For example, consider a retailer’s decision to prominently display one particular manufacturer’s products rather than another manufacturer’s products. This will have interbrand demand effects, positively influencing consumer purchases of the prominently displayed manufacturer’s products at the expense of products consumers otherwise would have purchased if they were prominently displayed. As is well known in the marketing literature, a significant fraction of sales are made by consumers who purchase a product only because they become aware of it when shopping (Areni, Duhan, and Kiecker 1999; Underhill 2009). While some consumers will choose to purchase the brand of product that is prominently displayed, few if any consumers are likely to shift the store at which they commonly shop based on this retailer decision. Consumers who prefer an alternative brand that is stocked but not prominently displayed can still obtain their preferred brand without switching retailers. Given the absence of interretailer demand effects, there is no consumer pressure in the competitive retail market for retailers to provide prominent display space in connection with the sale of one or another particular manufacturer’s products. Therefore, retailers often have significant discretion regarding which products they will prominently display. Another retailer service that similarly has large interbrand demand effects and little interretailer demand effects is increased salesperson attention devoted to the sale of a particular manufacturer’s products. While such selling efforts will influence one manufacturer’s sales relative to another manufacturer’s sales at the retailer, these retail services also are not likely to significantly shift where consumers shop. Consumers who prefer a brand that is not promoted by increased sales efforts will, once again, not shift the retailer they generally purchase from because they are still able to obtain their preferred products. Consequently, retailers will not be forced by competitive consumer pressures to actively promote a particular product with salesperson efforts at the point of sale. A retailer’s decision to stock one particular brand rather than another brand also often has primarily interbrand demand effects and relatively small interretailer demand effects. While retailer failure to stock a well-known, highly demanded product will cause some consumers to decide to shop elsewhere, retailers often will have considerable discretion regarding which of many somewhat lesser known brands in a product category they decide to stock. And when the retailer is a supermarket or department store that sells hundreds of different products, this retailer stocking discretion will be significant because it is unlikely that many consumers will switch the retailer they commonly purchase from solely because of the retailer’s failure to stock a single product. Although retailing is highly competitive, there are not an infinite number of retailers offering every possible product selection combination. Individual consumers decide where to shop and develop loyalty to particular retailers based on a retailer’s overall product selection that best fits their tastes, as well as other retailer characteristics such as average price, convenience of store location, and the service and amenities provided. Depending upon the degree of consumer loyalty, all retailers therefore have the ability to shift sales between brands to some extent by their stocking and promotional decisions,

Resale Price Maintenance of Online Retailing   283

which often will have small interretailer demand effects relative to much larger interbrand demand effects. The fundamental economic distinction between retailer actions that have primarily interbrand demand effects in contrast to interretailer demand effects does not coincide with the distinction between the demand effects associated with retailer nonprice services versus retailer price discounts (Winter 1993). Consumers may in fact be highly sensitive across retailers with regard to the supply of some nonprice retailer services such as free convenient parking or a pleasant selling environment and other amenities. Because these services are not brand-specific, we know they have little or no interbrand demand effects. On the other hand, price discounts, which generally have large interretailer demand effects, in some circumstances may have primarily interbrand demand effects. For example, when a supermarket has a sale on Coke, this may substantially shift sales at the supermarket to Coke from Pepsi, as marginal consumers who are relatively indifferent between Coke and Pepsi shift their purchases to Coke in response to the price discount. Because Coke and Pepsi are only one product in the consumer’s total supermarket purchase basket, the price promotion may not dramatically shift sales across supermarkets. Bottlers further substantially reduce interretailer demand effects by running the price promotion simultaneously at all supermarkets in their geographical area.2

12.2.2.  Manufacturers Must Compensate Retailers for Providing Interbrand Retailing Services Because retailers have substantial discretion with regard to how they decide to allocate their retail selling efforts that have primarily interbrand demand effects, such as product stocking, display space, and sales staff promotion, the products that receive these retailing services will not be determined primarily by retailer competition for consumers. The marginal consumers who are significantly influenced by these point-of-sale promotional services do not pay for the services; the services are provided free of charge to lower the effective price (net of the value of the services) and thereby generate incremental sales that primarily benefit the manufacturer. Therefore, rather than retailer competition for consumers determining which products receive retailer promotional sales 2 

Coke and Pepsi bottlers also maximize the interbrand demand effects by contracting with supermarkets for temporary exclusive price discount promotions. If Coke and Pepsi prices were both reduced, it would not be worth very much to either bottler in terms of incremental sales; when only one price is reduced, consumers who are marginal with respect to their preference for the two brands will shift to the price-discounted brand. The competitive outcome is for Coke and Pepsi to each offer exclusive price discounts for limited periods because this maximizes the extent of interbrand share shifting and hence the supermarket’s return, and consequently the ultimate benefit to consumers through supermarket competition. Coca-Cola Co. v. Harmar Bottling Co., 218 S.W. 3d 671 (Tex. 2006), discussed in Klein and Murphy 2008.

284   Benjamin Klein efforts, retailers supply these services in connection with the sale of particular products in response to the compensation they receive from manufacturers.3 Manufacturers compete with one another for retailer selling efforts because the interbrand demand effects from the retailer services means that consumers are induced to switch from rival brands to the manufacturer’s products and therefore represent clear additional sales of the manufacturer’s products. This is valuable to manufacturers because manufacturer wholesale prices are generally greater, frequently significantly greater, than manufacturer marginal costs. As a result, manufacturers often earn significant profit on incremental sales induced by retailer selling efforts. The profit earned by manufacturers on incremental sales is a consequence of the fact that manufacturers commonly sell differentiated products and face negatively sloped demands for their products; it certainly does not imply the existence of any manufacturer antitrust market power (Klein 1993). The resulting gap between wholesale price and marginal cost is what permits competitive manufacturers to cover their marketing, R & D, and other fixed overhead costs. Given this equilibrium under which most manufacturers operate, manufacturers will earn additional profit if they can convince retailers to supply point-of-sale services that induce incremental sales to marginal customers who would not otherwise purchase their products. Because an effective retail distribution network where retailer promotional services that induce incremental sales are provided in connection with a manufacturer’s products is valuable, often extremely valuable, manufacturers will bid in competition with other manufacturers for such retailing services. In general, manufacturers with the largest profit margin on incremental sales and the largest expected sales gain from retailer promotional selling efforts will bid the most for the retailer services. Retailers will choose to stock and promote the products they expect will yield them the greatest return, taking account of both the short-term and long-term effects of their decisions. Manufacturers must convince retailers that if they devote their valuable retailing assets such as their shelf space and sales staff to the promotion of the manufacturer’s products, the return on their assets will be greater than the return the retailers would earn if their retailing assets were devoted to the sale of another manufacturer’s products. Manufacturer bidding for retailer promotional selling efforts takes the form of an implicit manufacturer promise to retailers of an expected return if they devote their selling efforts to the manufacturer’s products. This implicit promise is frequently made by manufacturers by adopting some form of restricted distribution arrangement that reduces the number of authorized retailers of the manufacturer’s products, sometimes with the grant to each retailer of an explicit or de facto exclusive sales area. This induces 3 

Retailer competition for consumers, however, ultimately drives the competitive process by determining the investments made by retailers in the construction of attractive, conveniently located stores, the hiring and training of a sales staff, and the development of a reputation for reasonable prices, good product selection, and knowledgeable and friendly service that creates the loyal customer base underlying the retail assets manufacturers are bidding for. Competition and free entry into retailing benefits consumers and drives the equilibrium return earned by retailers on their tangible and intangible retailing assets to a normal rate of return.

Resale Price Maintenance of Online Retailing   285

desired retailer selling efforts because it increases the sales and the profit margin retailers expect to earn on the manufacturer’s products. Alternatively, when the number of retail outlets that carry a manufacturer’s products is an especially important determinant of total consumer demand, the manufacturer may compensate retailers for their selling efforts by increasing each retailer’s expected margin through a decrease in the wholesale price and the use of resale price maintenance.4 If incremental sales induced by retailer promotional services involve merely a movement down the demand curve or a percentage shift out of demand at every price, the profit-maximizing suggested retail price will not change. The manufacturer merely lowers the wholesale price to increase the price gap between retail and wholesale prices that is protected with resale price maintenance. In many circumstances where desired retailer promotional selling efforts are significant, this gap may be very large. For example, ready-to-wear clothing normally has a suggested retail price that is 2 to 2.5 times the wholesale price (Binkley 2012) that is protected with either restricted distribution or resale price maintenance. When the desired retail promotional services, such as cooperative advertising or hiring an employee dedicated to the demonstration of a manufacturer’s products, can be precisely specified and monitored, manufacturers may compensate retailers directly for the provision of such promotional services with a per service payment (Steiner, 1991).5 Manufacturer compensation for retailer product stocking decisions therefore often may take the form of a per unit time slotting fee payment (Klein and Wright 2007). However, efficient retailer compensation mechanisms more commonly incentivize retailer promotional efforts with a payment based on sales, such as occurs with resale price maintenance and exclusive territory arrangements. The increased profit margin on sales that is created provides the retailer with an increased profit incentive to promote the manufacturer’s products.6 A promotional payment that compensates retailers on the basis of incremental sales relative to some prior sales benchmark may appear to be an even more efficient way for a manufacturer to purchase retailer promotional services. In contrast to resale price maintenance, which compensates retailers on all their sales, such a promotional payment compensates retailers only on incremental sales and thereby appears to provide 4  The US Supreme Court recognized in Business Elecs. Corp. v. Sharp Elecs. Corp. 485 U.S. 717, 728 (1988) that all vertical restraints, exclusive territories as well as resale price maintenance, operate by inducing retailers to devote their selling efforts to a manufacturer’s products by “reduc[ing] intrabrand price competition to the point where the dealer’s profit margin permits provision of the desired services.” 5  Contrary to the inventory models developed by Deneckere, Marvel, and Peck (1996) and Butz (1997), where manufacturers use resale price maintenance to induce retailers to hold increased inventories of their products, increased inventory is a retailer service that manufacturers also are more likely to use more direct financial incentives to induce, such as subsidized financing of inventories, liberal return policies, or wholesale price refunds for unsold products. 6  For this reason, a slotting fee arrangement may sometimes include a lower wholesale price together with the per unit time slotting fee, which is combined with resale price maintenance if there are large interretailer demand price effects or with the requirement that a minimum fraction of the wholesale price reduction be passed on in lower retail prices when there are small interretailer demand effects.

286   Benjamin Klein retailers with a greater direct marginal incentive to expand sales of the manufacturer’s products. However, incremental sales are often difficult to measure. Moreover, a promotional payment based on incremental sales absent resale price maintenance would lead retailers to engage in price discounting that has largely interretailer demand effects, that is, an increase in sales of the manufacturer’s products at the price discounting retailer at the expense of sales of the manufacturer’s products at other retailers, and less likely to involve increased manufacturer sales at the expense of rival products. While this may result in a short-term increase in total manufacturer sales due to a reduction in the average retailer margin and resulting movement down the manufacturer’s aggregate demand curve, the reduction in the retailer profit margin also disrupts efficient retail distribution by decreasing retailer compensation for interbrand shifting retail services and hence reduces long-run sales.

12.2.3.  Retailer Price Discounting May Damage a Manufacturer’s Distribution Network When There Is No Consumer Free-Riding The potential for retailer price discounting to disrupt a manufacturer’s efficient retail distribution network designed to induce retailer selling efforts explains why manufacturers may use resale price maintenance when there is not a consumer free-riding problem. A retailer that discounts prices reduces the sales and/or the profit margin earned by other retailers that are distributing the manufacturer’s products. Because retailers require a minimum expected return to stock, display, and otherwise promote a manufacturer’s products equal to what they could earn by devoting their retailing assets to an alternative manufacturer’s products, this reduction in compensation caused by retailer price discounting will lead other retailers to reduce the selling efforts they devote to the manufacturer’s products. It may even lead some retailers to drop distribution of the manufacturer’s products entirely. Consequently, even when there is no consumer free-riding, retailer price discounting may deteriorate a manufacturer’s retail distribution network, in terms of reducing both the number and quality of retailers that stock the manufacturer’s products and the point-of-sale promotional services the remaining retailers provide in connection with the manufacturer’s products. The effects of retailer price discounting on reducing the return earned by other retailers may sound superficially like the effects of normal price competition among retailers, which manufacturers generally are in favor of. However, although manufacturers wish, everything else equal, to have their products sold at the lowest retail margin, to maximize long-run sales and profits manufacturers also wish to have their products effectively distributed and adequately promoted by an efficient number and type of retailers. To accomplish this manufacturers must establish a retail distribution network where retailers are adequately compensated for devoting their retailing efforts to the manufacturer’s products.

Resale Price Maintenance of Online Retailing   287

For example, consider resale price maintenance of boxed candy, the first example noted above by Pitofsky 1984 as a resale price maintenance arrangement that does not appear to be explainable by the prevention of consumer free-riding.7 The retailer services in that case consist primarily of stocking and displaying the product. It is incredible to believe that without resale price maintenance a significant number of consumers would free-ride by first going to a “full-service” retailer to obtain these stocking services before purchasing the candy at a discount retailer. However, although there is not a consumer free-riding problem on stocking services, these services are very valuable to the candy manufacturer. Candy is the type of product where display creates significant incremental sales because substantial sales are made to consumers when they observe the product displayed while shopping for other products. Consequently, wide retail distribution is valuable because it creates the increased potential for unplanned or “impulse” consumer purchases. A price-discounting retailer therefore can disturb the manufacturer’s effective retail distribution by reducing the number of other retailers that distribute the manufacturer’s products even when the price-discounting retailer is not free-riding on the services supplied by other retailers. The use of resale price maintenance by a manufacturer to create and protect an increased retailer margin and thereby ensure that the retail distribution network for its products consists of a larger number of retailers than would otherwise exist under uncontrolled retail competition is similar to the “outlets hypothesis” explanation for resale price maintenance (Gould and Preston 1965). Gould and Preston assume that the demand for a manufacturer’s products is related to the number of outlets that are selling the manufacturer’s products. However, while demand for a product often may be empirically related to the number of outlets that carry the product, Gould and Preston do not provide an economic basis for why the number of outlets cannot be left entirely up to consumers who choose where to shop in an uncontrolled competitive retail environment. The economic analysis of resale price maintenance presented here indicates that manufacturers often may wish to purchase an increased quantity of retail display space in a greater number of outlets than would otherwise exist because this induces profitable incremental sales by marginal consumers. Resale price maintenance in these circumstances is an efficient way for the manufacturer to purchase this retailing service. Manufacturers similarly do not want to leave the quality of retailers that distribute their products entirely up to uncontrolled retail competition for the same analytical reason that high-quality retailers stocking their products induces profitable incremental sales. Manufacturers wish to have reputable high-quality retailers stock and promote their products not primarily because of the positive “certification externality” hypothesized by Marvel and McCafferty 1984 described above, where such retailers increase the overall demand for the manufacturer’s products at all retailers. Instead, it is because high-quality retailers have a group of loyal customers and distribution through these retailers provides the manufacturer with valuable access to

7 

Russell Stover Candies, Inc., 100 F.T.C. 16 (1982).

288   Benjamin Klein this customer base. High-quality retailers, however, will decide to stock and promote a manufacturer’s products and thereby provide the manufacturer with access to their customers only if they expect to earn a sufficient return on the sale of the manufacturer’s products. When discount retailers sell a manufacturer’s products, the reduced retail profit creates a situation where it may no longer be worthwhile for high-quality retailers to devote their valuable shelf space and retailing efforts to the manufacturer’s products. For example, consider the retail distribution of Rolex watches discussed above. If Rolex watches are sold at a substantially lower price at discount retailers, this ultimately would significantly reduce Rolex sales at full-price authorized retailers. The insufficient return earned by authorized retailers on their valuable retailing assets would then lead such retailers to reduce their display and selling efforts devoted to Rolex watches, and possibly to stop carrying Rolex watches entirely. Although total Rolex sales may increase in the short run as a result of the lower prices set by discount retailers, Rolex sales ultimately are likely to decrease when authorized retailers no longer carry or actively promote Rolex products. This long-run reduction in demand is distinct from any free-riding by consumers on the services, including certification services, supplied by authorized Rolex retailers or from the possible negative externality effect on Rolex’s image from distribution by discount retailers. It is merely due to the fact that Rolex has lost its authorized retailers and authorized retailer sales efforts because of inadequate authorized retailer compensation, resulting in less effective retail distribution of Rolex’s products. Because the retailing promotional services a manufacturer may purchase from its authorized retailer network with resale price maintenance are not services that are valued by all consumers, a manufacturer that controls retailer price discounting with resale price maintenance may appear to imply that consumers who do not value the services and would purchase at a discount retailer in the absence of resale price maintenance are worse off. While marginal consumers who respond to the retailing services purchased by a manufacturer with resale price maintenance may benefit, inframarginal consumers seem to be paying higher prices without receiving any value from the purchased retail services. Frederic Scherer (1983) and William Comanor (1985) therefore argue that the net welfare effect and potential antitrust liability of a resale price maintenance arrangement should depend on the relative number of inframarginal versus marginal consumers of the manufacturer’s products. This analysis, however, fails to recognize that, as described above, manufacturers who purchase promotional retailing services with resale price maintenance will do so by lowering wholesale prices and setting resale price maintenance at unchanged retail prices. Therefore, discount retailers that evade resale price maintenance to serve consumers who do not value the promotional services are merely taking advantage of the manufacturer’s low wholesale price used to compensate retailers for the supply of such services. Consumers who do not value the promotional services are not being forced to pay for the services. If a manufacturer did not purchase the promotional retailing services with resale price maintenance, wholesale prices would be higher.

Resale Price Maintenance of Online Retailing   289

Furthermore, even if we ignore the adjustments that would occur in the wholesale price, differential effects between consumers are not an appropriate basis for determining whether a practice is anticompetitive. Competitive retailers provide many services that are not consumed by all customers, such as free sales assistance. One customer may get the benefit of trying on two dozen pairs of shoes using an hour of salesperson time before making a purchase, while another customer purchases the same pair of shoes in five minutes without trying on any shoes. The fact that retailers provide free sales assistance is reflected in the average retail margin, and we leave it up to competition to determine what retail services are supplied to consumers on what terms. A manufacturer’s implicit purchase of promotional retail services with resale price maintenance or with some other substitute arrangement used to compensate retailers for supplying retailing services valued only by some customers is similarly a normal aspect of competition. Although the purpose of the antitrust laws is to establish rules that have the ultimate effect of maximizing consumer welfare, this does not imply that antitrust regulation should involve empirically estimating whether or not a competitive firm’s particular marketing practice produces a net overall consumer welfare gain or loss from actions that are a part of the normal competitive process.

12.2.4.  Online Retailing Makes It More Likely a Manufacturer Will Use Resale Price Maintenance to Protect Its Retail Distribution Network The growth of online retailing has increased the importance of resale price maintenance as a way for manufacturers to protect their retail distribution networks. Most manufacturers will include online retailing as part of their efficient retail distribution network because online retailing increases shopping convenience and permits consumers to obtain product information cheaply, reducing retail distribution costs. However, manufacturers generally will not want to leave the terms of online retailing and the share of retail distribution that moves online entirely uncontrolled. A fundamental concern of manufacturers is that uncontrolled online retailing is likely to result in less than the desired amount and type of brick-and-mortar retail distribution. Even when there is no free-riding by consumers who first visit a brick-and-mortar retailer before purchasing at an online retailer, online retailer price discounting is likely to disrupt a manufacturer’s efficient brick-and-mortar retail distribution network. The effects are similar to the disruption described above that is created when price discounting by an individual retailer reduces retailer compensation below the level required to induce other retailers to provide the manufacturer’s desired point-of-sale retail services. Online retailer price discounting has the potential to greatly magnify the detrimental effects of retailer price discounting on a manufacturer’s efficient retail distribution network because the interretailer demand effects of online retailer price discounting are likely to be substantially larger than brick-and-mortar retailer price discounting. The

290   Benjamin Klein result of uncontrolled online retailer price discounting therefore may be a very substantial reduction in the number and type of brick-and-mortar retailers that stock and promote the manufacturer’s products. Interretailer demand effects from online retailer price discounting are likely to be substantially larger than brick-and-mortar retailer price discounting for two reasons; one, the uncontrolled online retailer price discount is likely to be substantially larger, and two, the interretailer demand elasticity is likely to be substantially larger. With regard to the first effect, the online retailer price discount is likely to be substantially larger because online retailers can take advantage of their relatively low retailing costs in addition to the lower wholesale prices set by manufacturers to compensate retailers for the provision of retailer promotional services. Consequently, the potential decrease in the retail price by a discounting online retailer will be substantially larger than the retail price decrease by a discounting brick-and-mortar retailer. With regard to the second effect, the likely demand response from an individual online retailer price discount is likely to be substantially greater than an individual brick-and-mortar retailer price discount because of the increased ability of consumers to shop alternative retailers online. For both these reasons, the potential disruption to a manufacturer’s retailer network (brick-and-mortar retailers as well as established online retailers with loyal customers) caused by online retailer price discounting is likely to be substantially greater than when only brick-and-mortar retailers are discounting price. Even if consumers do not free-ride by shopping at a brick-and-mortar retailer or an established online retailer before purchasing at an online discount retailer, discount online retailers are taking advantage of the fact that manufacturers set relatively low wholesale prices to compensate retailers for supplying costly retailing promotional services, including access to a retailer’s loyal customers. Online retailers that discount price close to this low wholesale price are able to expand their sales and profits solely because of the way manufacturers pay for the provision of retailer promotional services. As described above, the use of resale price maintenance to prevent such discount online retailing does not mean that consumers who would otherwise purchase from the discount retailer are being charged for promotional services they do not value and are not receiving. Online retailers operating in an environment where there are primarily brick-and-mortar retailers also are facing lower wholesale prices than otherwise because their costs of retailing are lower than brick-and-mortar retailers. (If there were only online retailing, manufacturer profit-maximizing wholesale prices would be set by the manufacturer at a substantially higher level because, from the manufacturer’s point of view, retailing is a complementary input.) Given the lower wholesale prices set by the manufacturer to compensate brick-and-mortar retailers for their supply of interbrand promotional services, as well as their higher retailing costs, totally uncontrolled online retailer price discounting is likely to result in a great many brick-and-mortar retailers earning an insufficient return on the retailing assets they devote to the sale of the manufacturer’s products. This will lead many brick-and-mortar retailers to decrease their retailing efforts focused on the manufacturer’s products and in some cases to drop

Resale Price Maintenance of Online Retailing   291

distribution of the manufacturer’s products entirely. The presence of online retailing therefore makes it even more crucial for manufacturers who wish to continue to purchase brick-and-mortar point-of-sale retailer services that induce profitable incremental sales to control price discounting in order to ensure that the number and type of retailers that sell their products remains optimal. As long as online sales are a relatively small fraction of a manufacturer’s total sales, uniform resale price maintenance with a relatively low wholesale price across all retail distribution is likely to be an efficient way for a manufacturer to preserve the desired amount of brick-and-mortar distribution. However, once the efficient amount of online retailing grows, manufacturers may find it profitable to set differential price terms that include small permitted online price discounts combined with higher wholesale prices for online retailers. However, the potential for retailers to underreport the extent of their online sales to obtain more favorable wholesale prices or for brick-and-mortar retailers to transship product purchased at lower wholesale prices to online retailers makes this solution difficult to implement, and it could face legal challenges. Some manufacturers may find it economic to move to an alternative retailer compensation arrangement, such as making per unit time payments to brick-and-mortar retailers for desired point-of-sale retailer promotional services. Retailers would not have the same incentive to provide dedicated sales efforts under such an arrangement because they would be earning less (or even losing) money on the margin on each incremental sale. Retailers, for example, may not devote as many salespeople or as much preferred shelf space to the manufacturer’s products. As a result, per unit time payments would have to be combined with more direct manufacturer monitoring and enforcement of purchased retailer point-of-sale promotional efforts than exists with resale price maintenance, an increased real resource cost borne by both manufacturers and retailers. The other alternatives to resale price maintenance available to manufacturers that wish to preserve their desirable brick-and-mortar retail distribution networks also have significant economic disadvantages because of the economic pressures created by the large potential interretailer demand effects from online retailer price discounts. As noted above, manufacturers may sometimes compensate retailers for increased retailer services not with resale price maintenance but by reducing the number of authorized retailers that distribute the manufacturer’s products. Adequate retailer compensation for the services they devote to the sale of a manufacturer’s products is achieved because each individual retailer then has some potential pricing power and also expects a higher level of sales of the manufacturer’s products. However, with uncontrolled online retailer discounting it becomes extremely difficult for manufacturers to create adequate compensation of brick-and-mortar retailers in this way. While consumers generally travel only a limited distance from home or work to shop and compare prices, when there is online retailing the fact that there are relatively few authorized brick-and-mortar retailers will not guarantee that these retailers obtain a significant share of sales. In addition, online retailing makes it difficult for manufacturers to generate a sufficient retailer return necessary for efficient retail distribution with the use of a “minimum advertised price” (MAP) policy. In contrast to resale price maintenance, which

292   Benjamin Klein controls all prices, a MAP policy only limits retailer public advertisements of prices to manufacturer-suggested minimum levels but permits retailers to lower prices to whatever level they wish within their stores. This often makes economic sense because within-store price discounts may have primarily interbrand demand effects that manufacturers wish to encourage, while publicly advertised discount prices are likely to have large interretailer demand effects that may produce deleterious effects on overall retailer compensation and therefore the manufacturer’s efficient retailer network. The presence of online retailing, however, largely destroys these differential demand effects upon which a MAP policy is based. Because of the ease with which consumers shop across online stores, within-store online retailer price discounting has much greater interretailer demand effects than within-store brick-and-mortar retailer price discounts. Even when a manufacturer utilizing a MAP policy requires online retailers to show consumers their discounted price only at checkout, comparing actual sales prices across online retailers by consumers before final purchase is fairly easy. The online retailer may merely include a statement such as “Add to cart to see price.” Within-store online retailer price discounts therefore have a much greater potential to disrupt a manufacturer’s retail network than within-store brick-and-mortar retailer price discounts. As a result, rather than a MAP policy, manufacturers that wish to preserve a sufficient number of brick-and-mortar retailers as part of their efficient retail distribution network will often find it necessary to control online retailer pricing more directly with resale price maintenance.

12.3.  Anticompetitive Motivations for Resale Price Maintenance 12.3.1.  Manufacturer-Motivated Anticompetitive Resale Price Maintenance In addition to the procompetitive motivations discussed above, resale price maintenance also may be motivated by anticompetitive considerations. The potential for anticompetitive effects is generally considered significantly lower when a resale price maintenance arrangement is initiated by a manufacturer rather than by a retailer. Because manufacturers are in effect purchasing retailing services at an implicit price equal to the retail margin defined by the gap between the retail price paid by consumers and the wholesale price manufacturers receive, a manufacturer would not appear to have an economic interest in instituting resale price maintenance that increased retail margins unless doing so compensated retailers for supplying services that shifted out the demand for their products. Even if the manufacturer instituting resale price maintenance were

Resale Price Maintenance of Online Retailing   293

a dominant firm with significant market power, it would not want to have its retailers earn a greater profit margin or prevent innovations in retailing that lowered retail margins unless the greater retail margin was serving some demand-increasing competitive purpose. There are two primary exceptions to this general presumption that manufacturerinitiated resale price maintenance must be procompetitively motivated. One exception involves the joint adoption of resale price maintenance by a group of manufacturers to stabilize a collusive manufacturer cartel. Because resale price maintenance prevents retailers from lowering prices, it substantially reduces the sales gain a manufacturer can achieve by undercutting the collusively set wholesale price. Resale price maintenance therefore decreases the incentive of manufacturers to cheat on the cartel. To establish the likely existence of resale price maintenance as a cartel-facilitating practice one must have evidence of a collusive horizontal conspiracy among manufacturers to fix wholesale prices. The mere fact that a large fraction of firms in an industry use resale price maintenance is far from sufficient evidence of such a horizontal conspiracy. The industry must be conductive to a conspiracy and evidence must exist that individual manufacturers are acting contrary to their independent economic self-interests absent such a conspiracy. Without evidence of a horizontal agreement among manufacturers to fix wholesale prices and to jointly adopt resale price maintenance, the widespread use of resale price maintenance in an industry, by itself, is more likely to suggest a common underlying efficiency reason for manufacturers in the industry to have unilaterally adopted resale price maintenance. The second case where it has been claimed that a manufacturer may anticompetitively adopt resale price maintenance involves a dominant manufacturer’s use of resale price maintenance to maintain its market power by placing smaller manufacturing rivals and potential new manufacturing entrants at a significant disadvantage. Although a dominant manufacturer does not want to overpay for retailing services, this anticompetitive theory posits that a dominant manufacturer may use resale price maintenance to reduce the incentive of retailers to carry rival brands by providing retailers with a protected, higher than normal margin on sales of the dominant manufacturer’s brand. Dominant manufacturer use of resale price maintenance, it is therefore argued, may have the effect of driving rival firms out of the market and preventing new firms from entering because of the inability of rivals to obtain retail distribution. However, absent the dominant manufacturer’s threat to terminate retailers that carry rival products, that is, absent a de facto exclusivity agreement between the dominant manufacturer and its retailers (a claim that should be evaluated for anticompetitive effects separately from resale price maintenance), there is no reason why retailers would not also carry competing products. Absent exclusivity, all that is necessary for retailers to decide to carry rival brands is that retailers expect to earn a sufficient return on the sale of the rival brands. This anticompetitive theory of resale price maintenance amounts to a claim of predatory overpayment for retail distribution. When a retailer’s return on the dominant manufacturer’s established brand is increased with resale price maintenance,

294   Benjamin Klein competing manufacturers of rival brands must similarly increase what they offer retailers to obtain distribution, possibly also compensating retailers with resale price maintenance. Rather than exclusionary predatory competition, however, it is much more likely that a dominant manufacturer’s use of resale price maintenance in these circumstances is just an element of the normal competitive process by which manufacturers compete for retail distribution. Small new manufacturers, as well as large established manufacturers, may use resale price maintenance within this competitive framework as one of the effective ways to compensate retailers for stocking and promoting their products.

12.3.2.  Retailer-Motivated Anticompetitive Resale Price Maintenance In contrast to manufacturer-initiated resale price maintenance, retailer-initiated resale price maintenance may superficially appear to be more troubling. If a demand-increasing purpose were served by resale price maintenance, it would seem that the manufacturer would initiate it. If instead it is a retailer that demands manufacturer institution of resale price maintenance, this may incorrectly suggest that resale price maintenance is being used anticompetitively to protect the retailer from retail price competition and not to ensure the supply of desired retailing services. A key question that must be answered, in terms of both the economics and the legal standard, is whether coercive demands are being made by the retailer on the manufacturer to adopt resale price maintenance. Coercion, rather than mere initiation of a request for resale price maintenance, indicates both the existence of retailer market power and the exercise of such power. By acquiescing to coercive retailer demands, the manufacturer is adopting resale price maintenance contrary to its belief that an alternative arrangement where sales by lower priced retailers are uncontrolled by resale price maintenance would be in its economic interests. Retailer market power sufficient to accomplish such coercion generally requires a dominant retailer market share. Alternatively, a group of smaller retailers may jointly coerce a manufacturer to institute resale price maintenance. However, this requires a collusive agreement among numerous small retailers and an effective cartel enforcement mechanism, an extremely rare condition. In fact, a very small number of resale price maintenance arrangements are instituted as a result of organized group retailer pressure of manufacturers consistent with the existence of an anticompetitive retailer cartel (Overstreet 1983). Multiple retailers may make simultaneous demands on a manufacturer to stop retailer price discounting and to adopt resale price maintenance, but this should not be interpreted as analogous to a coercive retailer cartel. Extensive retail price discounting may mean that most retailers are not earning a competitive return on their retailing assets, and therefore it would be in each individual retailer’s economic interests to independently demand that the manufacturer stop such discounting.

Resale Price Maintenance of Online Retailing   295

A dominant retailer may initiate the discussion of resale price maintenance with a manufacturer for a similar reason, that it is earning less than normal profit in selling the manufacturer’s products. It certainly does not imply that the dominant retailer is coercing the manufacturer to act contrary to its interests. The retailer may initiate the discussion because it is more immediately aware of retail market conditions and its own reduced profitability. Retailer communication of market realities to the manufacturer, such as the extent and magnitude of price discounting occurring in the marketplace, the retailer’s estimate of its resulting lost sales, and the communication of its likely response to these market realities do not mean that the manufacturer is being coerced by the retailer to adopt resale price maintenance. If the manufacturer had this information about current and likely future market conditions, the manufacturer independently may have decided that the adoption of resale price maintenance was in its economic interests. We would expect a retailer to communicate to a manufacturer its requirements regarding minimally acceptable retail margins and its desire for the manufacturer to adopt resale price maintenance as a normal part of competitive business activity. As we have seen, resale price maintenance often is an element of an exchange between a manufacturer who is buying retail promotional services and its retailers who are selling retail promotional services, both of whom are benefiting from the transaction. In these circumstances, who initially suggests resale price maintenance may not have much economic significance. It certainly would be incorrect to infer that the transactor that initiates discussion of resale price maintenance is the only one benefiting from the arrangement and is coercing the other party to adopt the arrangement against its economic interests. For example, consider a case where a manufacturer asks a retailer to stock or intensively promote a new product and the retailer in response asks the manufacturer how it intends to organize distribution of the product—a clearly relevant business consideration for the retailer. The retailer may ask the manufacturer who else is (or will be) carrying the manufacturer’s product in the retailer’s area and what the manufacturer’s other terms of retail distribution will consist of. No economic significance should be attached in these business communications that are part of the normal competitive process to whether (a) the manufacturer says it plans to institute resale price maintenance and expects the retailer in return to intensively promote the product, or (b) the retailer offers to carry and more intensively promote the product and says in return it expects the manufacturer to maintain reasonable retail margins. The manufacturer may have already decided to adopt some minimum pricing arrangement or may realize during its conversations with retailers that such an arrangement would be the best way to maximize its sales and effectively compete with other manufacturers. Continuous communications between manufacturers and their retailers about market conditions are expected as part of the normal, noncoercive competitive process. As noted by the US Supreme Court in Monsanto, “the fact that a manufacturer and its distributors are in constant communication about prices and marketing strategy does

296   Benjamin Klein not alone show that the distributors are not making independent pricing decisions.”8 The Supreme Court concluded that “something more than evidence of complaints is needed” to infer the existence of an agreement (Monsanto at 764). Furthermore, even if a retailer warns or threatens a manufacturer that it will drop distribution if resale price maintenance is not instituted, such communications do not imply the existence of a coercive agreement. We cannot infer that the retailer has forced the manufacturer to agree to do something that would be contrary to its interests absent retailer coercion. It may merely involve clarification of the expected quid pro quo in a voluntary transaction between a retailer that is providing access to valuable distribution through its outlets in return for reasonable manufacturer compensation for these retailing services. Economic evidence of retailer coercion of a manufacturer to act against its economic interests in establishing resale price maintenance requires, first of all, that the retailer has the ability to coerce a manufacturer or manufacturers to act contrary to their economic interests. That is, there must be evidence that the retailer possesses effective market power. This requires, as an initial screen, that the retailer possesses a sufficiently large market share in the relevant retailing market so that the retailer is essential for effective distribution by the manufacturer of its products. It further requires that the manufacturer is not a dominant supplier of an essential product that the retailer finds necessary. To establish that a dominant retailer is exercising its market power by coercing a manufacturer to act contrary to its economic interests requires, as noted, more than that the dominant retailer has informed the manufacturer that it will drop distribution (or stop active promotion) of the manufacturer’s products if the manufacturer does not stop retailer price discounting. If small retailers without any market power whatsoever are independently similarly complaining to the manufacturer about the adverse effects of retailer price discounting on their business and the likely future actions they expect to take if the price discounting does not stop, this would indicate that the dominant retailer’s communications, by themselves, are not a coercive exercise of its market power but part of normal business communications with regard to a voluntary economic transaction. Similar to small retailers, a dominant retailer need not provide its valuable distribution to a manufacturer if it can earn more distributing another manufacturer’s products. Rather than coercion, both large and small retailers of the manufacturer’s products are just informing the manufacturer about the market conditions they face and their likely competitive responses. A manufacturer’s decision to adopt resale price maintenance in reaction to retailer complaints about price discounting may merely be a response by the manufacturer to its genuine concern that retailers otherwise would not find it in their economic interests to distribute its products, or to distribute its products as effectively. If a manufacturer learns that extensive retailer price discounting is occurring and recognizes that it would ultimately lead retailers, including a dominant retailer, to unilaterally conclude that it no longer pays to devote their shelf space or other retail promotional efforts to the manufacturer’s products, it would be in the manufacturer’s independent economic interest to

8 

Monsanto Co. v. Spray-Rite Serv. Corp., 465 U.S. 752, 762 (1984).

Resale Price Maintenance of Online Retailing   297

terminate price discounting retailers. The manufacturer would stop the retailer price discounting even though the manufacturer has not received any warnings from its retailers because the manufacturer expects retailers to independently decide to stop carrying or promoting its products once they become aware of retail market conditions. Manufacturers may recognize that they receive benefits of short-term increased sales as a result of retailer price discounting, and therefore that they may bear a short-term cost of reduced sales when they terminate price discounting retailers. However, manufacturers also recognize that not terminating price discounting retailers would reduce the long-run demand for their products because of the resulting inability to obtain or maintain adequate retail distribution. If a manufacturer’s internal business documents indicate independent concern about the adverse effects of retailer price discounting on its retail distribution network through expected unilateral retailer reactions, and further indicate the necessity to terminate price discounting retailers to preserve their overall efficient retailer distribution network over the long term, this would be a clear indication that a dominant retailer’s communications with manufacturers regarding the adoption of resale price maintenance do not involve the exercise of retailer market power to force manufacturers to act contrary to their interests.

12.4.  Antitrust Regulation of Resale Price Maintenance in the United States and European Union 12.4.1.  The Antitrust Standard of Resale Price Maintenance Analysis in the United States and European Union The US Supreme Court in its landmark Leegin decision overturned the per se antitrust treatment of resale price maintenance in favor of a rule-of-reason analysis. Plaintiffs in the United States challenging a resale price maintenance arrangement therefore are now legally required to prove the existence of a likely anticompetitive effect, and defendants may present procompetitive business rationales. The Court made it clear that the fact that resale price maintenance may increase an individual product’s retail price in the short term, or even long term, should not, by itself, be considered evidence of such an anticompetitive effect (Leegin, at 895–97). Instead, US antitrust law post-Leegin requires demonstration of an anticompetitive market effect in terms of the particular anticompetitive theories of resale price maintenance described above (Klein 2012). In contrast to the US rule-of-reason standard that focuses on potential antitrust market effects, resale price maintenance is still categorized in the European Union under the European Commission Guidelines on Vertical Restraints (2010) as

298   Benjamin Klein a “hardcore restraint” that is “presumed to fall within Article 101 (1)” as “an agreement that appreciably restricts or distorts competition” and is unlikely to fulfill the minimum market share and other conditions for block exemption under Article 101 (3) (Guidelines at ¶¶ 47–48). However, similar to the US movement to a less stringent legal treatment of resale price maintenance, the European Commission now considers this anticompetitive presumption to be rebuttable if there are sufficient efficiencies associated with the resale price maintenance on an individual case basis so that the procompetitive benefits of the arrangement outweigh its anticompetitive effects (Guidelines at ¶ 223).

12.4.2.  Procompetitive Rationales for Resale Price Maintenance under US and EU Antitrust Law Given the heightened standard by which resale price maintenance is judged in the European Union compared to the United States, the legally accepted procompetitive rationales for resale price maintenance are more important under the EC Guidelines than under US antitrust law. Because antitrust liability requires demonstration of an anticompetitive market effect under US law, a relatively small single manufacturer’s adoption of resale price maintenance is unlikely to be illegal even when there is no obvious efficiency rationale. In contrast, under the EC Guidelines justification of a small firm’s use of resale price maintenance requires a sufficient efficiency rationale for the practice not to be considered an antitrust violation. Furthermore, the legally accepted efficiency rationales for resale price maintenance are significantly broader under US law than under the EC Guidelines. Specifically, US antitrust law does not rely entirely on the well-established procompetitive economic benefits of resale price maintenance in preventing free-riding. The US Supreme Court recognized in Leegin that resale price maintenance may serve as a way for competitive manufacturers to compensate retailers for retailing services as described in this chapter even when there is no free-riding (Leegin, 551 U.S. 877, 892 (2007), citing Klein and Murphy 1988 and Mathewson and Winter 1998). In contrast, the EC Guidelines and European Union national competition authorities do not recognize this procompetitive rationale for resale price maintenance. Instead, the Guidelines emphasize supplier-driven prevention of interretailer free-riding as the primary, if not sole, procompetitive efficiency justification that may possibly rebut the anticompetitive presumption and avoid antitrust liability of resale price maintenance arrangements. The focus in the EC Guidelines on the prevention of free-riding as the primary procompetitive justification for resale price maintenance is illustrated by statements in the Guidelines that resale price maintenance may be justified only for particular types of products, such as “experience or complex products” where it is essential for retailers to provide presale demonstration services (Guidelines at ¶ 225). The Guidelines also describe the free-riding problem more generally as applying to products that are “relatively new or technically complex or the reputation of the product must be a major

Resale Price Maintenance of Online Retailing   299

determinant of its demand” (e.g., products that rely on selective distribution to protect a brand name image such as high-end cosmetics) (Guidelines at ¶ 107). The emphasis on product characteristics where a potential free-riding problem may exist as the basis for a justification of resale price maintenance suggests that resale price maintenance cannot be justified for products that are not technologically “complex” or do not involve a brand name image that must be protected by selective distribution. For example, consider brand name products more generally, such as Nike sports shoes, which may have relatively little potential for free-riding under these criteria. Nike sports shoes are not technologically complex products that require substantial prepurchase demonstration services and are not products that require especially high-end retailing services to protect their image. However, competitive manufacturers of brand name products such as Nike earn a significant profit margin on their incremental sales and therefore may find it efficient to use resale price maintenance to compensate retailers for the display and selling efforts that are important for generating incremental sales. The EC Guidelines refer to the possibility that resale price maintenance may serve this procompetitive efficient retailer compensation role unrelated to the prevention of free-riding with regard to the introduction of new products. It is claimed that resale price maintenance may be used as a way to induce retailers to promote a new product during its introductory period by giving retailers the means to stock and increase their sales efforts (Guidelines at ¶ 225). But it is unclear why the EC Guidelines limit this efficient retailer compensation rationale for resale price maintenance solely to new products. The economic importance of resale price maintenance as a way to support a retail distribution network that includes a sufficient number and quality of retailers, both brick-and-mortar and online retailers, that stock and actively promote a manufacturer’s products also is relevant for manufacturers of established products. Manufacturers of established products may already have a large, effective retailer distribution network, where the retailers of their products supply point-of-sale display and other retail services that induce incremental manufacturer sales. However, the manufacturer must ensure its retailers that they will continue to earn a sufficient return on their retailing efforts, without which the manufacturer’s established distribution network will deteriorate. In fact, individual retailer price discounting of well-known products may have even greater interretailer demand effects and therefore would be more likely than new products to result in a reduction in the compensation received by retailers for providing retailing services in connection with the sale of the manufacturer’s products.

12.4.3.  Manufacturer Control of Online Retailing under US and EU Antitrust Law US antitrust law does not specify the particular conditions under which manufacturers may use resale price maintenance to control online retailing of their products. It is generally left up to manufacturers to choose the retail distribution arrangement with regard to online retailing that best fits their business needs as long as there is no demonstrable

300   Benjamin Klein anticompetitive market effect from the arrangement. And manufacturers in fairly wide circumstances have the ability under US antitrust law to use resale price maintenance to protect their brick-and-mortar retail distribution network. In contrast, the EC Guidelines adopt a regulatory approach that controls the types of restrictions a manufacturer may impose on the online retailing of its products. The EC has an additional stated policy aim, separate from antitrust considerations, of encouraging the expansion of online retailing. The EC believes that the growth of online retailing is a way to reduce intercountry price differences and thereby ensure that consumers receive the benefits of increased price competition and consumer choice from the single European Union market (EC 2009).9 The EC Guidelines implicitly acknowledge that there may be legitimate problems associated with uncontrolled online retailer discounting by recognizing that some forms of manufacturer control of online retailing to protect a brick-and-mortar retail distribution network are considered legally acceptable (Guidelines, ¶¶ 52–54). In particular, under the EC Guidelines (at ¶ 52(c)) a manufacturer may require authorized retailers to sell a certain absolute amount (in value or volume) of its products at brick-and-mortar retail outlets, thereby giving the manufacturer the ability to prevent online retailing “pure plays.” Manufacturers, however, cannot restrict the fraction of online sales an authorized retailer may make. This and other restrictions on manufacturer control of online retailing in the Guidelines are designed so that manufacturers cannot reduce retailer incentives on the margin to increase online sales. Most importantly, under the EC Guidelines a manufacturer is permitted to compensate retailers differentially for their brick-and-mortar sales relative to their online sales. The Guidelines, however, explicitly prevent a manufacturer from compensating brick-and-mortar retailers for supplying additional costly retailing services by charging retailers a lower wholesale price on their brick-and-mortar sales than on their online sales, or equivalently, by raising the wholesale price on online sales (Guidelines at ¶ 52(d)). While differential compensation is permitted, the EC Guidelines, once again, are concerned that the relative marginal costs of online retailing not be increased and online retailer incentives to discount not be reduced. Therefore, under the EC Guidelines manufacturers may provide brick-and-mortar retailers either with fixed per unit time payments to cover the higher costs of brick-and-mortar retailing or presumably with extra direct services, such as providing brick-and-mortar retailers with salespeople who are paid employees of the manufacturer. As the difficulties of resale price maintenance increase with the growth of online retailing, as described above, companies in the United States also may find these brick-and-mortar retailer compensation arrangements economic. But because the EC Guidelines are concerned that there be no increase in the relative marginal costs of 9 

To facilitate the growth of online retailing the EC Guidelines (at ¶¶ 51–53) also classifies most online retailing transactions as “passive,” that is, initiated by consumers who decide to visit the website of a retailer outside of the consumer’s area, rather than involving an online retailer actively entering a retailer’s established territory in violation of selective distribution agreements.

Resale Price Maintenance of Online Retailing   301

online retailing, the slotting fee payment or free services provided to brick-and-mortar retailers cannot be related to sales.10 Although the EC Guidelines clearly recognize that competitive manufacturers may reasonably wish to “purchase” a greater amount of brick-and-mortar retailing than would exist if retailer competition were uncontrolled, a relevant question is whether the Guidelines would permit manufacturers to use resale price maintenance to control online retailing so as to be able to maintain an overall effective retail distribution network when resale price maintenance is the most efficient form of retailer compensation. Under the EC Guidelines manufacturer payment of brick-and-mortar retailer compensation, in whatever form it takes, must be for brick-and-mortar retail services considered by the Guidelines to be economically legitimate. Therefore, brick-and-mortar retailer compensation is likely to be limited in practice to compensation for the point-of-sale demonstration of complex products or the maintenance of a high-end product image with expensive store rent and sales staff. These are the much too narrow brick-and-mortar retailer services that are considered economically legitimate based solely on the avoidance of free-riding by online retailers on brick-and-mortar retailers. The EC Guidelines do not recognize the inherent problem that uncontrolled online retailer price discounting may disrupt or even destroy a manufacturer’s valuable retailer network crucial for creating profitable incremental sales even when there is not a free-riding problem.

12.5. Conclusion The growth of online retailing has made it economically more important for manufacturers to be able to use resale price maintenance to preserve essential elements of their efficient retail distribution networks, both efficient brick-and-mortar and established online retailer distribution. In the United States this has been facilitated by the legal movement to a rule-of-reason standard under Leegin. In contrast, in the European Union, resale price maintenance remains a hardcore restraint, now rebuttable but with the recognized procompetitive role served by resale price maintenance largely limited to the avoidance of free-riding. This ignores the competitive use of resale price maintenance when there is not a consumer free-riding problem as a way to efficiently compensate retailers for the provision of interbrand point-of-sale retailing services that induce incremental manufacturer sales. Manufacturers in the European Union therefore are more likely to use other retailer compensation mechanisms, such as per unit time payments, to preserve efficient elements of their retail distribution networks.

10  The EC Guidelines at ¶¶ 203–6 separately set restrictive conditions on the use of up-front access payments, but such slotting payments are not considered a hardcore restraint.

302   Benjamin Klein

References Areni, Charles, Dale Duhan, and Pamela Kiecker. 1999. Point-of-Purchase Displays, Product Organization, and Brand Purchase Likelihoods. Journal of the Academy of Marketing Science 27:428–41. Binkley, Christina. 2012. Polo Puzzle: What Goes into a $155 Price Tag? Wall Street Journal, February 2. Comanor, William B. 1985. Vertical Price-Fixing, Vertical Market Restrictions, and the New Antitrust Policy. Harvard Law Review 98:983–1002. Butz, David. 1997. Vertical Price Controls with Uncertain Demand. Journal of Law and Economics 40:433–60. Carlton, Dennis W., and Judith A.  Chevalier. 2001. Free Riding and Sales Strategies for the Internet. Journal of Industrial Economics 49:441–61. Deneckere, Raymond, Howard P.  Marvel, and James Peck. 1996. Demand Uncertainty, Inventories, and Resale Price Maintenance. Quarterly Journal of Economics 111:885–913. European Commission. 2009. Report on Cross-Border E-Commerce in the EU. Commission Staff Working Document, SEC (2009) 283 final. European Commission. 2010. Guidelines on Vertical Restraints. 2010/C 130/01, May 19. Gould, J. R., and L. E. Preston. 1965. Resale Price Maintenance and Retail Outlets. Economica 32:302–12. Gundlach, Gregory T., Joseph P. Cannon, and Kenneth C. Manning. 2010. Free Riding and Resale Price Maintenance: Insights from Marketing Research and Practice. Antitrust Bulletin 55:381–422. Jacoby, Jacob, and David Mazursky. 1984. Linking Brand and Retailer Images: Do the Potential Risks Outweigh the Potential Benefits? Journal of Retailing 60:105–22. Klein, Benjamin. 1993. Market Power in Antitrust: Economic Analysis after Kodak. Supreme Court Economic Review 3:43–92. Klein, Benjamin. 2009. Competitive Resale Price Maintenance in the Absence of Free-Riding. Antitrust Law Journal 76:431–81. Klein, Benjamin. 2012. Assessing Resale Price Maintenance after Leegin. In Research Handbook on the Economics of Antitrust Law, ed. Einer Elhauge. Northampton, MA: Edward Elgar, 174–200. Klein, Benjamin, and Kevin M.  Murphy. 1988. Vertical Restraints as Contract Enforcement Mechanisms. Journal of Law and Economics 31:265–97. Klein, Benjamin, and Kevin M. Murphy. 2008. Exclusive Dealing Intensifies Competition for Distribution. Antitrust Law Journal 75:433–66. Klein, Benjamin, and Joshua D. Wright. 2007. The Economics of Slotting Contracts. Journal of Law and Economics 50:421–54. Marvel, Howard P., and Stephen McCafferty. 1984. Resale Price Maintenance and Quality Certification. Rand Journal of Economics 15:346–59. Mathewson, Frank, and Ralph Winter. 1998. The Law and Economics of Resale Price Maintenance. Review of Industrial Organization 13:57–84. O’Brien, Daniel P., and Greg Shaffer. 1992. Vertical Control with Bilateral Contracts. Rand Journal of Economics 23:299–308. Overstreet, Thomas R. 1983. Resale Price Maintenance:  Economic Theories and Empirical Evidence. Federal Trade Commission Bureau of Economics staff report, November.

Resale Price Maintenance of Online Retailing   303

Pitofsky, Robert. 1984. Why “Dr. Miles” Was Right. Regulation 8:27–30. Scherer, Frederic M. 1983. The Economics of Vertical Restraints. Antitrust Law Journal 52:687–707. Steiner, Robert L. 1991. Manufacturers’ Promotional Allowances, Free Riders, and Vertical Restraints. Antitrust Bulletin 36:383–411. Telser, Lester. 1960. Why Should Manufacturers Want Fair Trade? Journal of Law and Economics 3:86–105. Underhill, Paco. 2009. Why We Buy: The Science of Shopping. New York: Simon & Schuster. Van Baal, Sebastian, and Christian Dach. 2005. Free Riding and Customer Retention across Retailers’ Channels. Journal of Interactive Marketing 19:75–85. Winter, Ralph. 1993. Vertical Control and Price versus Nonprice Competition. Quarterly Journal of Economics 108:61–76.

CHAPTER 13

E XC LU SI V E DE A L I N G HOWARD MARVEL

13.1. Introduction Exclusive dealing is a vertical restraint that requires distributors of a supplier’s products to purchase most or all of the products or services that they offer for sale within a product category from that supplier, thereby excluding competing products of rival suppliers from those distributors. In its strictest form, exclusive dealing requires that dealers or distributors wishing to distribute a supplier’s products agree not to handle competing products. But more broadly, exclusive dealing can include arrangements that simply cause distributors to accord a supplier’s products preferential treatment. For example, a supplier can purchase enhanced access to a distributor’s customers using wholesale price discounts applicable at the margin, promotional fees, and payments for shelf space (slotting allowances) or featured placements, such as end cap displays. Under such contracts, rivals are not completely excluded, but may find their products placed in ways that do not generate as many sales as they could have achieved in an unconstrained distribution environment. Klein and Murphy (2008) have termed such restraints “partial exclusive dealing.” Exclusive dealing can also take the form of requirements contracts in which a distributor or ultimate consumer agrees to buy all of its requirements for a particular product from a single supplier. Even if we restrict attention to its simplest form—namely, a supplier requirement that a dealer or distributor not handle the products of competing suppliers—exclusive dealing is perhaps the most ubiquitous of vertical restraints. A survey of marketing managers for companies in SIC 35, industrial machinery/equipment, and 36, electronic and electric equipment, reported by Heide, Dutta, and Bergen (1998), found that exclusive dealing was employed by 46 of 147 respondents. More generally, exclusive dealing is common both for goods sold to consumers and for intermediate inputs as well.

Exclusive Dealing   305

Examples of products subject to exclusive dealing include racing tires,1 sandpaper,2 at-shelf coupon dispensers,3 bladder catheters purchased through a group-purchasing organization,4 pulse oximetry sensors and monitors,5 artificial teeth,6 condoms,7 photochromic lenses,8 petroleum-dispensing equipment (gas pumps),9 and beer (Sass 2005). Business format franchise contracts typically require complete exclusive dealing—a franchisee is required to adhere strictly to the specified format, which precludes offering additional products at its locations. In addition, such franchise agreements can also require that inputs be purchased exclusively from the franchisor, or from an approved supplier. Traditional franchising relationships also typically contain exclusivity clauses. Automobile dealers, for example, are often asked to agree not to engage in “dualing”— sharing facilities with multiple automobile makes.10 Similarly, gasoline stations offer only a single fuel brand.11 Exclusive dealing, by its very nature, excludes,12 at least from the portion of a market covered by the exclusive dealing contract. As a consequence of this exclusion, its treatment constitutes an important part of antitrust policy. An exclusive dealing agreement by which a supplier and its dealer agree to exclude rival products from a “quantitatively substantial” portion of the market (that foreclose rivals from access to a significant portion of the potential customer base) can be held to be a violation of Section 1 of the Sherman Act. With Section 3 of the Clayton Act, Congress identified exclusive dealing as a specific contractual provision requiring scrutiny to check whether its effect was significantly to lessen competition.13 A dominant supplier whose exclusive dealing requirement excludes rivals from access to an economically significant portion of market 1 

Race Tires Am., Inc. v. Hoosier Racing Tire Corp., 614 F. 3d 57 (3d Cir. 2010).

2 NicSand, Inc., v. 3M Co., 457 F. 3d 534 (6th Cir. 2006). 3 

Menasha Corp. v. News Am. Mktg. In-Store Inc., 354 F. 3d 661 (7th Cir. 2004). St. Francis Med. Ctr. v. C.R. Bard, Inc., 657 F. Supp. 2d 1069 (E.D. MO. 2009). 5  Allied Orthopedic Appliances, Inc. v. Tyco Health Care Group, LP, 592 F. 3d 991 (9th Cir 2010). 6  United States v. Dentsply International, Inc., 399 F. 3d 181 (3rd Cir 2005). 7  Church & Dwight v. Mayer Labs., Inc., U.S. Dist. Lexis 51770 (No. D. Cal., 2012). 8  In the Matter of Transitions Optical, Inc., Federal Trade Commission, Docket No. C-4289, 2010. 9  Omega Envtl. v. Gilbarco, Inc. 127 F. 3d 1157 (9th Cir. 1997). 10  For a recent example of an automobile supplier taking action to prevent “dualing,” see Jaguar Land Rover North America v. Manhattan Imported Cars, 2012 U.S. App. LEXIS 8260 (4th Cir. 2012). The controversial nature of automobile suppliers to “de-dual” their dealerships in order to obtain exclusive dealing is indicated by the existence of legislation (in this case, Maryland Code §15-207(d)) prohibiting de-dualing of existing franchised dealers. 11  But see Standard Oil Co. v. United States, 337 U.S. 294 (1949) (Standard Stations), holding that exclusive dealing requirements that covered 6.7% of sales in the seven western states in which the company sold were sufficiently quantitatively substantial enough to trigger condemnation under Section 3 of the Clayton Act. 12  Exclusion, however, is not equivalent to foreclosure. A manufacturer may be excluded from particular distributors but may still be able to reach customers through available alternatives. See, e.g., Omega Envtl. v. Gilbarco, Inc., 127 F. 3d 1157 (9th Cir. 1997). 13  In its Standard Stations opinion, the Supreme Court held that unlike tying, exclusive dealing was not a per se violation, but neither was a full rule-of-reason investigation required as long as a substantial volume of commerce was foreclosed. 4 

306   Howard Marvel distribution for the product in question can also be found to have engaged in monopolization and thereby to have violated Section 2 of the Sherman Act.14 Limits on the use of exclusive dealing extend beyond those derived from the Sherman and Clayton Acts. The Federal Trade Commission (FTC) has recently obtained a consent order against exclusive dealing under its broad authority to control unfair competition provided in Section 5 of the FTC Act.15 Since it is targeted at the products of rivals who could otherwise choose to have their products distributed through the channels that exclusive dealing restricts, the practice almost by definition raises rivals’ costs.16 The degree to which costs are raised depends on the availability and efficacy of alternative distribution channels open to rivals. Thus in many circumstances, prior to considering the costs and benefits of particular exclusive dealing restrictions, a threshold inquiry into the characteristics of alternative channels is required. The European Commission has a two-pronged approach to antitrust enforcement as applied to exclusive dealing that is similar to the US application of Sections 1 and 2 of the Sherman Act. Article 101 of the TFEU (Treaty on the Functioning of the European Union) prohibits agreements that restrict competition. The application of this article to vertical agreements is covered by Commission Regulation 330/2010/EU of April 20, 2010, which provides for a number of safe harbor exemptions that apply when market shares are low. Exclusive dealing is, however, explicitly excluded from the safe harbor for vertical restraints.17 Article 102 of the TFEU is supposed to prevent anticompetitive abuses by dominant firms. The focus of Article 102 enforcement is on whether distributors find it essential to carry the products of the dominant firm. If so, exclusive dealing forecloses rival producers. Thus according to the European Commission, The capacity for exclusive purchasing obligations to result in anti-competitive foreclosure arises in particular where, without the obligations, an important competitive constraint is exercised by competitors who either are not yet present in the market at the time the obligations are concluded, or who are not in a position to compete for the full supply of the customers. Competitors may not be able to compete for an individual customer’s entire demand because the dominant undertaking is an unavoidable trading partner at least for part of the demand on the market, for instance because its brand is a “must stock item” preferred by many final consumers or because the capacity constraints on the other suppliers are such that a part of demand can only 14  Thus Microsoft’s exclusive dealing agreements with its distributors were not found to have violated Section 1 of the Sherman Act since Microsoft’s rival, Netscape, had managed to distribute its browser to a substantial portion of the market, but Microsoft’s exclusive dealing was illegal under Section 2. 15  For a recent challenge to exclusive dealing brought under Section 5 of the FTC Act, see In the Matter of Transitions Optical, Federal Trade Commission, Docket No. C-4289, 2010. 16  By attracting customers who might otherwise have purchased from rivals, exclusive dealing shifts a supplier’s demand schedule out, increasing its power over price. See Krattenmaker and Salop (1986). But note that as Klein (2003, 119), quoting Judge Frank Easterbrook, points out, “injuries to rivals are byproducts of vigorous competition.” 17  The exemption does not apply to “any direct or indirect obligation causing the members of a selective distribution system not to sell the brands of particular competing suppliers.” Article 5, sec. 1(c).

Exclusive Dealing   307

be provided for by the dominant supplier. If competitors can compete on equal terms for each individual customer’s entire demand, exclusive purchasing obligations are generally unlikely to hamper effective competition unless the switching of supplier by customers is rendered difficult due to the duration of the exclusive purchasing obligation. In general, the longer the duration of the obligation, the greater the likely foreclosure effect. However, if the dominant undertaking is an unavoidable trading partner for all or most customers, even an exclusive purchasing obligation of short duration can lead to anti-competitive foreclosure.18

13.2.  Exclusive Dealing as Property Right Creation Vertical restraints generally are devices that create property rights to customers. Thus an exclusive territory creates a property right to benefit from customers who are drawn to purchase a particular product within a defined area. The right is typically created by the product’s producer, but assigned to distributors. The reason that the supplier establishes such a right is to provide an incentive for its distributors to promote its brands, bringing in those customers. The property right ensures that the distributor who provides promotional efforts receives the benefit of those efforts, and that rival distributors who might otherwise free ride on a dealer’s promotion are prevented from doing so. Resale price maintenance is a similar method for property right creation and promotion protection. So, too, is exclusive dealing, but it differs in an important manner from most other vertical restraints. A supplier that creates a property right for its dealers does so only to the extent that the right generates promotion for the supplier’s product that more than offsets the return that the right provides. That is, the supplier’s role is to structure its distribution so that distributors receive compensation for their competitive promotional activities, but no more. But with exclusive dealing, a supplier creates a property right with itself as the beneficiary. The supplier’s incentive to limit the right to match distributor compensation to the services they are expected to provide is absent, or at least attenuated.19 The mingling of the roles of creator of a property right and beneficiary of that right is the reason why exclusive dealing is normally considered more suspect than other vertical restraints. 18 

Communication from the Commission—Guidance on the Commission’s enforcement priorities in applying Article 82 of the EC Treaty to abusive exclusionary conduct by dominant undertakings, Official Journal C 045, 24/02/2009 P. 0007-0020, http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=CELE​ X:52009XC0224(01):EN:HTML. 19  The principal limitation on suppliers in promulgating exclusive dealing property rights lies in the requirement that they provide their dealers with returns sufficient to cover their opportunity costs of remaining with the brand demanding the exclusive.

308   Howard Marvel True exclusive dealing simply ensures that the supplier obtains the benefits of its own promotional activities. A supplier who brings customers to its dealers expects to charge a premium price reflecting the value of those customers. It could charge for customer leads directly, but if it did so, it would force onto its dealers a substantial amount of risk. Thus exclusive dealing is akin to a royalty agreement where the dealer (franchisee) pays its franchisor for the success of the franchisor’s design and promotion only when those activities yield sales. In the typical exclusive dealing setting, the royalty equivalent is a charge rolled into the wholesale price of the product. As with a revenue-based royalty, this charge is tied to actual sales rather than merely to customers who are drawn to the dealer by the manufacturer’s efforts. Alternatively, the manufacturer could impose a separate charge for its customer-generating activities, and indeed, incorporating a charge into completed sales clearly has the potential to distort retailer decisions through, for example, double marginalization. The advantage of linking promotion costs to sales is that doing so provides incentives to the supplier to design its product and promotion strategies with considerable care, as returns are contingent on those decisions. The dealer, too, need not monitor or approve the supplier’s customer generation, but can simply concentrate on converting those customers to sales. The problem with this approach is that the dealer can avoid the charge for customers that its supplier brings to its door by offering a rival’s product to customers, where the rival’s product does not incorporate a customer charge and accordingly has a lower wholesale price. The dealer can thereby execute a bait-and-switch scheme, baiting customers drawn by a supplier’s promotion but switching them to a product that carries a higher dealer margin. Both customers and dealers can benefit from this—the rival product may be functionally identical to that of its promoted competitor and the bait-and-switch will ordinarily require a discount for the lesser known rival product. Without exclusive dealing, however, the supplier may find its incentive to promote the product greatly attenuated. The following examples consider both the consequences of free riding and the possibility that the supplier may not alter its behavior should its exclusive dealing be prevented. The first provides a clear example of how a ban on exclusive dealing suppressed competition and harmed market efficiency. Final evaluation of the second example must wait until the effects of a different ban on exclusive dealing can be assessed.

13.2.1.  Hearing Instruments The above explanation for exclusive dealing appeared first in Marvel (1982) with illustrative examples provided by an investigation of insurer/agent relationships in the property/casualty insurance industry and by an analysis of the market for dress patterns as considering in the Standard Fashion case.20 The analysis was inspired by a different example of exclusive dealing, one that was the centerpiece of an investigation of the effects of the exclusive dealing clauses that hearing instruments manufacturers had 20 

Standard Fashion Co. v. Magrane-Houston Co., 256 U.S. 346 (1922).

Exclusive Dealing   309

demanded from their dealers.21 Hearing instruments were and are sold both through dealers and through audiologists. The latter channel obtains clients primarily from physician referrals. Hearing instruments sold through dealers, however, require active client generation. A supplier who wishes to promote its products to potential customers faces significant difficulties in identifying hearing-impaired individuals. Radio and television advertising has obvious drawbacks, and while local newspaper advertising is sometimes employed, the supplier must pay for messages based on the number of customers reached, a far greater number than the set of individuals who might potentially purchase a hearing aid. During the 1970s, suppliers favored periodicals targeted specifically at older consumers as advertising vehicles. A supplier would typically prepare a print advertisement with a mail-in postcard that an interested reader could fill out and submit in order to receive a hearing test. The supplier, upon receipt of the postcard, would assign it to a dealer in the relevant territory. The dealer was responsible for administering the test and, if warranted22 would suggest an appropriate hearing aid. Most customers would not have been familiar with the brands of hearing instruments available, so that a dealer dispatched by one supplier would have little difficulty substituting the product of a rival. Since the wholesale price of a hearing aid provided by a postcard (lead) supplying manufacturer carried a premium over “bargain” instruments that came without promotion, the incentive to switch customers was substantial.23 The FTC obtained consent orders against four of the largest five hearing instrument manufacturers outlawing their use of exclusive dealing. The consequence of this action was dramatic—within one year, all four of the manufacturers had either abandoned the dealer channel or had disappeared altogether.24 One remaining manufacturer, Beltone, refused to enter a consent order and ultimately prevailed at the Commission level.25 The company remains a successful competitor. But Beltone’s success in selling to dealers hearing instruments bundled with customer leads was not sufficient to counteract the adverse effects of the ban on exclusive dealing for the remaining manufacturers. In fact, the decline in promotion following the consent orders appears to have caused a very large drop in sales through the dealer channel with no corresponding increase in sales through the alternative audiologist channel. The hearing instruments case thus suggest that exclusive dealing can be a crucial element of a competitive strategy by which consumers are drawn to dealers of the brand requiring exclusive dealing. Banning the practice can impair or even terminate firms’ ability to compete and thereby has potentially dramatic consequences for competition. 21 

The results of that investigation appear in Marvel (1984). This work was not public at the time of the publication of Marvel (1982). 22  Or perhaps if not warranted. Investigations by both the FTC and FDA uncovered a number of sales that appear unlikely to have improved customer hearing. 23  One manufacturer who agreed to drop exclusive dealing stopped providing leads to dealers and reduced its wholesale price by $10, or roughly 10%. 24  The exit of these manufacturers came in spite of the lower prices at which they offered hearing aids to dealers. 25  In re Beltone Electronics Corp., 100 F.T.C. 68, 271 (1982).

310   Howard Marvel

13.2.2.  Photochromic Lenses Photochromic lenses darken in the presence of significant sunlight and become clear indoors or in darkness. They allow prescription eyewear consumers to obtain at least some of the benefits of sunglasses without the need to purchase a second pair of glasses. Photochromic coatings are applied to about 20% of all corrective lenses, and the market leader, Transitions Optical, has a market share in excess of 80% of photochromic lenses. It is also a heavy promoter of photochromic lenses, so much so that the term “transition” applied to such lenses verges on generic.26 It should be no surprise, then, that Transitions implemented exclusive dealing arrangements with downstream distributors of its products inhabiting a number of levels in the distribution process, including lens manufacturers (casters), wholesale labs, and retailers. In 2010, the Federal Trade Commission challenged Transitions’ use of exclusive dealing and negotiated a consent order that bans its use through 2030 (In the Matter of Transitions Optical, Docket No. C-4289, Decision and Order, US Federal Trade Commission, 2010, http://www.ftc.gov/os/caselist/091006​ 2/10042tranopticaldo.pdf). The circumstances of Transitions Optical seem to closely parallel those of the hearing instruments cases. Given Transitions’ intensive advertising of the benefits of its photochromic lenses,27 how did the FTC deal with the possibility that other photochromic coating manufacturers might free ride on the customers pulled to eyewear dispensers by Transitions’ advertising? The vast majority of Transitions’ promotional efforts are brand specific, reducing the significance of any free-riding concern. While Transitions’ marketing efforts may generate some consumer interest in the product category as a whole—and not just in Transitions’ own products—this is part of the natural competitive process. This type of consumer response does not raise a free-riding concern sufficient to justify the substantial anticompetitive effects found here.28

The FTC thus argued that Transitions had been successful in convincing the targets of its advertising to “accept no substitutes.” Hence the advertising expenditures redounded to Transitions benefit, with little impact on the sales of rival 26  See, for example, http://science.howstuffworks.com/innovation/science-questions/question412.​ htm. Sunglasses or prescription eyeglasses that darken when exposed to the sun were first developed by Corning in the late 1960s and popularized by Transitions in the 1990s. In fact, because of the extreme popularity of the Transitions brand, these lenses are usually referred to as transition lenses. 27  An article entitled Increasing Patient Awareness of Photochromic Lenses, appearing in Vision Care Product News, April 2008, catalogs extensive Transitions Optical advertising campaigns including cooperative advertising with lens labs. Advertising of Transitions Optical’s competitors is also described, but appears to be minimal in comparison to the volume of Transitions advertising. 28  Analysis to Aid Public Comment, In the Matter of Transitions Optical, Inc., File No. 091-0062, March 3, 2010, http://www.ftc.gov/os/caselist/0910062/100303transopticalanal.pdf, at 6 (footnotes omitted).

Exclusive Dealing   311

photochromic coatings producers. The alternative is, of course, that the Transitions advertising produced customers for photochromic lenses who would seek out distributors offering Transitions lenses but would accept recommendations of alternate brands from those distributors, given that they carry the Transitions product and therefore the brand’s imprimatur. If so, Transitions can be expected to reduce or at least reorient the focus of its promotions to deal with the threat of free riding. If the consent order is successful in opening Transitions’ distribution to competing coating manufacturers, the market’s experience will provide a useful empirical contrast between the free rider theory and the FTC’s analysis. If the FTC is correct, the share of lenses with photochromic coating should continue to increase and the intensity of Transitions’ advertising will be unaffected. Rivals, given access to broader distribution will not only gain share at Transition’s expense, but may also compensate for any reductions in Transitions’ advertising efforts. There will be a clear benefit to impact evaluation of this case.

13.3.  Theories of Exclusive Dealing 13.3.1.  The Role of Externalities Bernheim and Whinston (1998, 75) provide an elegant theoretical treatment of exclusive dealing that opens with a demonstration that without contractual externalities, exclusive dealing cannot be used profitably to foreclose a rival from a market. Because each manufacturer must effectively compensate the retailer to attract it to an exclusive deal, manufacturers internalize the retailer’s cost from the loss in product variety.

Bernheim and Whinston thus validate the argument offered by Bork (1978) that retailers must be compensated for forgone surplus when rival products are dropped. But the validation of the argument does not end the analysis, for as Bernheim and Whinston point out, while a ban on exclusive dealing cannot increase welfare in their basic model, something must be missing. In the basic model, exclusionary contracts are superfluous—even without a contract, rival suppliers will not be able to offer a deal sufficient to induce the retailer to carry their products in addition to the most efficient supplier.29

29 

Zenger (2010) notes that the Bernheim-Whinston model assumes that suppliers employ nonlinear contracts. In practice, exclusive dealing contracts often involve prices that are linear in output or royalties conditional on sales. The linear pricing in particular is unlikely to be accompanied by significant fixed fees. Ironically, Zenger’s point is less relevant in the setting he analyze, partial exclusive dealing. For example, retailers often receive fixed payments for display space (slotting and promotional fees), and when suppliers purchase additional customers from distributors, the purchase is often in the form of bundled discounts.

312   Howard Marvel The Bork approach of searching for payments from suppliers to retailers that compensate retailers for lost sales due to exclusion also faces a problem for classic exclusive dealing arrangements: the payments flow from retailer to supplier. As we saw above, the payments compensate suppliers for their demand-generation activities. The royalties paid by business format franchisees provide a particularly clear illustration of the payments that exclusive dealing protects. A franchisee who sold rival products to customers drawn to the franchisor’s trademark and format would avoid compensating the franchisor for those customers. The principal benefit of the Bernheim and Whinston analysis comes from its guidance concerning where to look for harmful exclusive dealing. They suggest the search for anticompetitive effects should focus on “noncoincident” markets, which they define to be markets other than the ones in which exclusive dealing is practiced. The survey of potentially harmful effects of exclusive dealing that follows indeed features a number of such situations. But in a sense, the Bernheim-Whinston approach places too high a hurdle before arguments asserting harm from exclusive dealing. Given that the use of exclusive dealing to assert a property right is an indication that fully specified contracts between suppliers and their dealers are infeasible (for otherwise, customer leads could be dealt with separately from the particular products to which those leads are attached), it will not be surprising that some instances of exclusive dealing assert a right to which distributors would not have agreed in a fully specified up-front contract. For example, a supplier who has cooperated with rivals to build a market for its products may at some point determine that it would benefit by asserting a right to the market as a whole, thereby harming its prior collaborators as well as consumers whose range of options is thereby reduced. But given Bernheim and Whinston’s results, it should not be surprising that it is difficult to develop equilibrium models in which exclusive dealing proves harmful to competition. Harm is often asserted to arise from exclusive dealing’s ability to raise the distribution costs of rival firms, but the raising-rivals’-costs analysis is more of a loose description of a potential problem than a full-fledged theory. Thus it is claimed that “if there are scale economies or other entry barriers in retailing, exclusive dealing arrangements can raise rivals’ costs of distribution” (Salop and Sheffman 1983, 267). But even if distributors face no entry barriers, a rival supplier faces higher costs to the extent that the supplier cannot free ride upon the promotion efforts of the supplier imposing exclusive dealing. Salop and Sheffman (1987) do not consider this possibility, as they treat market demand for the product in question as given, but exclusive dealing that protects promotion will, if successful, shift the demand schedule out. The raising-rivals’-cost theory simply assumes that demand is stable, and thus cannot distinguish the extent to which anticompetitive exclusive dealing operates in addition to the ubiquitous use of exclusive dealing to protect demand-increasing investments. The theory is useful, however, to the extent that it can be taken to provide necessary conditions for harmful effects of exclusive dealing. The harmful effects of exclusive dealing to raise rivals’ costs require that rivals find distribution alternatives to exclusive dealing retailers impracticable. Thus a demonstration of the availability of reasonably

Exclusive Dealing   313

efficient distribution alternatives should suffice to defeat accusations of anticompetitive exclusive dealing. As will see below, some courts, but not all, have adopted such a test.30 We turn next to additional formal theories of anticompetitive effects of exclusive dealing.

13.3.2.  Naked Exclusion as a Coordination Failure Under what circumstances can exclusion limited to a single product be expected to yield anticompetitive outcomes? A series of papers (Rasmusen, Ramseyer, and Wiley 1991, 2000; Segal and Whinston 2000) has considered circumstances under which distributors can be induced to sign contracts with a monopoly supplier that serve to entrench that supplier against the threat of entry by rivals requiring access to those same distributors in order to compete. The feasibility of successful exclusion stems from the presence of a high minimum efficient scale of production for the product that the monopoly supplier and potential rivals wish to market. That level of production can be denied to rivals if the incumbent supplier can lock up contractually a sufficient share of the available distributors. But this outcome can only occur if downstream purchasers are unable to coordinate. As Rasmusen, Ramseyer, and Wiley (2000) explain,31 Another condition is that the victims—customers or suppliers—must expect that the exclusionary tactic will succeed, and must be unable to coordinate their actions to defeat the tactic. An excluding firm in this situation can buy naked exclusion affordably because it can scare victims into selling cheaply; no single victim can stop the exclusion by itself, so no single victim has any bargaining power. At a theoretical limit, the excluding firm can gain the exclusionary rights for free.

There are several problems with this analysis. One is that exclusive dealing commitments may be required in order to permit the initial provision of what becomes the incumbent’s efficient scale facility. For example, contracts between farmers and elevator operators often contain provisions to prevent the farmers from defecting to a rival, should one be built (Frasco 1991). Thus if the marginal cost of adding a customer to a facility with substantial sunk capacity cost is low, customers who are located near to a new rival can expect to avoid their share of capacity costs due to competition, leaving the costs to be borne by those customers who turn out to be less felicitously located. But if the facility was sized initially based on the willingness of customers to contribute to capacity charges, and if those charges were not captured up front, the expectation of ex post opportunism will lead to underinvestment. Enforceable exclusive dealing contracts 30 Note that the converse does not hold. Granting access to an existing distribution channel to a firm

that has no feasible alternative will not increase welfare if the result of this access a substantial cutback in incumbent supplier actions to increase demand funneled through that channel. 31  See also Innes and Sexton (1994).

314   Howard Marvel can prevent this. Breach would be permitted under such contracts but only upon payment of a penalty tied to unrecovered capacity costs. Put differently, the coordination failure that would result is one that did not force customers into exclusive agreements, but rather one that caused customers to avoid such agreements, preferring to play off suppliers against one another. A more serious problem is that there appear to be few if any examples of the combination of minimum efficient scale and coordination failure that result in anticompetitive exclusive dealing—an absence that Rasmusen, Ramseyer, and Wiley (2000) emphasize.32 One possible example is provided by a recent exclusive dealing case, United Regional.33 In this case, the Department of Justice (DOJ) argued that while the demand for hospital services in the relevant market, Wichita Falls, Texas, came from Medicaid/ Medicare, the Blue Cross system insurers, other private third-party payers such as John Hancock, and uninsured individuals, only the third-party insurers were charged fees sufficient to contribute to the cost of medical facilities. United Regional, the dominant hospital, was the sole provider of must-have tertiary care facilities.34 Facing competition for other services, it offered low prices for those services to insurers that agreed to designate it their sole network supplier for the Wichita Falls market. This requirement was interpreted by the DOJ as tantamount to exclusive dealing, for consumption of out-of-network services would result in substantially copays for patients. The coordination problem was that had the insurers been able to cooperate, they could have commissioned one of United Regional’s rivals to expand to provide the necessary services.35 The United Regional requirement that insurers grant exclusive network provider status to United Regional in return for very substantial discounts on medical services that were available from other providers in the market was likely to exclude those other providers from selling their services to customers that paid the highest prices in the Wichita Falls market. The DOJ argued that the lack of access to such customers would slow expansion of competing providers, though the net result was much lower prices paid for the competitive services than without the exclusive dealing requirement. By granting exclusive network provider status to United Regional, an insurer increased significantly 32  In particular, they remark on “the lack of fit between [their naked exclusion] theory and the cases in which the U.S. Supreme Court has forged the law most relevant to exclusionary conduct” (310). 33  United States and State of Texas v. United Regional Health Care System, Case No.: 7:11-cv-0003 (N.D. Texas, 2011). 34  Tertiary care facilities provide specialty referral care in high-cost facilities are typically available only in the most comprehensive hospitals in a market. They are “must-have” facilities for insurers who wish to provide comprehensive medical care coverage to their insureds. 35  It should be noted, however, that private third-party payer insurers covered patients who represented only about 8% of United Regional’s patient volume, and the payments that United Regional received from these insurers accounted for approximately 30%–35% of United Regional’s profits in total. See “Competitive Impact Statement,” United States and State of Texas v. United Regional Health Care System, Case No.: 7:11-cv-0003 (N.D. Texas, 2011), http://www.justice.gov/atr/cases/f267600/267653.pdf (at 11). Note, however, that while third-party payers account for only about a third of United Regional’s profits, those profits derive primarily from services that competing hospitals doe not provide—the services that were subject to exclusive dealing requirements thus constituted a far smaller share of recorded profits.

Exclusive Dealing   315

the probability that a consumer would choose to have the procedure in question done by United Regional. The exclusive dealing arrangement thus delivered customers in return for prices that were substantially below those that the rival providers offered. Thus the exclusive dealing may have been a competitive device by which United Regional offered lower prices in return for the added customers that those discounts would bring. The task in this case is to determine whether the low prices were designed to exclude rivals from the market entirely by denying them the high-margin customers that they needed to justify facilities construction, or whether instead the low prices were a net benefit of the ability of United Regional to buy additional customers at low prices. This example thus focuses on the difficulty of distinguishing anticompetitive exclusive dealing from exclusive dealing that defines a property right to customers as a part of an effort to compete for those particular customers vigorously. United Regional’s competitors were in fact in operation, and thus not excluded. Competition from United Regional denied them customers who absent the exclusion might have generated revenues that would have supported expansion of the competing facilities. Those customers, however, would have paid more for their services than the price offered by United Regional. It is possible that the United Regional offer included prices for the must-have services that were sham prices inflated in order to make the discounts for marginal customers appear attractive, and therefore that United Regional’s scheme was an essential facilities tying arrangement. That is, absent the exclusive dealing scheme, the price of the essential facilities sold on their own would have been much less than the price quoted by United Regional. This is another case in which impact evaluation of the effects of the DOJ consent order requiring United Regional to drop its scheme would be very useful. But without such an assessment, combining this example with the reluctance of Rasmusen, Ramseyer, and Wiley to identify circumstances in which their anticompetitive explanation of exclusive dealing could apply suggests that considerable caution is warranted before concluding that exclusive dealing can be explained as a device to take advantage of coordination failures. Before turning to other anticompetitive theories of exclusive dealing, it is useful to consider experimental evidence for the naked exclusion theory provided by Landeo and Spier (2009, 2012). The first of these papers starts in an unpromising fashion, suggesting that exclusive dealing imposed by large brewers had caused craft brewers to “hit a brick wall” so that “analysts were predicting the demise of many small brewers.”36 The brick wall apparently crumbled, as craft beer sales rose from less than 1% of domestic sales in 1990 to 3% in 2000 and 5%in 2010.37 By 2011, craft brewing accounted for 5.7% of sales by volume and 9.1% of sales by dollars.38 36  Landeo and Spier (2009), at 1850, quoting “Amid Probe, Anheuser Conquers Turf,” Wall Street Journal, March 9, 1998. 37  See 2011 Draft Beer Industry Update, October 2011, www.beerinstitute.org. 38  Brewers Association statistics. See www.brewersassociation.org/bages/business-tools/​ craft-brewing-statistics/facts, visited January 22, 2013.

316   Howard Marvel The explanation for the failure of the “brick wall” of exclusive dealing is illuminating. Exclusive dealing excludes only if the dealers subject to exclusive dealing requirements represent the sole viable mechanism for new entrants to reach the market. This requires not only significant scale economics for the manufacture of the product in question, but also scale economies in distribution. The explosive growth of the craft beer segment of the beer market suggests that neither of these conditions was met. Landeo and Spier provide an experimental design that has a single incumbent seller facing two buyers. A seller who obtains a deal with just one of the two buyers is able to deter entry, because the potential entrant cannot obtain sales sufficient achieve efficient scale. The buyers do not compete with each other. If they did, the buyer who agreed to purchase from a more efficient potential entrant would capture the market from the incumbent’s dealer.39 Given the experimental design, Landeo and Spier show that allowing the supplier to discriminate in offers allows it to follow what they term a “divide-and-conquer” strategy. Buyer communication enhances the effectiveness of exclusionary practices when the seller can make discriminatory offers, but when discrimination is not permitted, communication increases both the generosity of the incumbent’s offers and the probability that the buyers will reject those offers, clearing the way for entry. A second paper by Landeo and Spier (2012) uses a similar design, but adds a potential entrant as a player. Communication with the entrant appears to make it easier for the buyers to agree on their preferred joint strategy of rejecting exclusive dealing contract offers from the incumbent seller, who responds with more generous offers to obtain exclusive contracts. The Landeo and Spier experimental results buttress the Rasmusen, Ramseyer, and Wiley theoretical finding that anticompetitive exclusion can happen when scale economies are very significant both in production and in distribution, but it remains an open question as to when, or whether, such exclusion occurs in the marketplace.

13.3.3.  Credible Commitments to Tax Entrant Surplus When a monopoly manufacturer distributing through independent dealers confronts the prospect of a competing, and more efficient entrant, that manufacturer may choose to offer exclusive dealing contracts that commit dealers to defecting to the rival only upon payment of a portion of the rents that the new entrant expects to obtain. Suppose that but for the exclusive dealing, all dealers would immediately defect to the new, more efficient entrant upon its appearance. The new entrant, succeeding the existing monopolist, can capture the incumbent monopolist’s rents plus the additional value of its efficiency advantage. Competitive dealers are no better off as a result of the switch in 39 Note, too, that if the buyers are dealers who then resell to their captive customers, the only welfare loss possible is due to prevention of a more efficient producer from supplanting the incumbent—there is no impact on consumers. The issue instead is the split of surplus between the supplier and its dealers.

Exclusive Dealing   317

suppliers, receiving merely a competitive return no matter which supplier is present. Is there a way that the incumbent manufacturer and its dealers can take advantage of their incumbency to tax the incremental rents that the new entrant will earn? Aghion and Bolton (1987) propose exclusive dealing contracts with liquidated damages as a device to extract such a tax. If dealers have been convinced to sign exclusive contracts with the incumbent producer, once a new, more efficient entrant appears, the deal it offers the new dealers must be sufficient to compensate them for the liquidated damages they must incur should they defect. The formation of contracts together with the payment of liquidated damages thus can transfer rents to the incumbent (who receives the liquidated damage payments) and the dealers (who must be compensated for signing the liquidated damages contracts). Whinston (2006) criticizes this theory because it works only if the new entrant indeed is able to surmount the contractual barrier. That is, the payoff to the incumbent for forming such a scheme is triggered when the new entrant actually appears and liquidated damages are paid. Exclusion of a more efficient entrant only occurs when the damages that an entrant must pay are less than the benefit accruing to the new entrant. But Whinston’s criticism is not completely compelling, for the exclusive dealing contracts designed to tax new entrants must be signed in anticipation of the appearance of a more efficient entrant. The greater the efficiency gain of the entrant, the higher the rents that can be extracted. but the exclusive dealing contracts need to be formed prior to the observation of the entrant’s efficiency. But since liquidated damages need to be specified in advance, a potential entrant that appears with small but positive efficiencies will choose not to enter. Thus exclusive dealing can be harmful, but its efficiency losses will be small. Perhaps a more serious objection is that the scheme requires both lock-in and a court system willing to enforce what is truly a penalty clause.40 Aghion and Bolton cite United States v. United Shoe Machinery Corporation41 as an example of a monopolist imposing lease terms that charged a penalty to customers wishing to switch show machinery suppliers. But Masten and Snyder (1993) investigate the relevant liquidated damages clauses and conclude that they were designed instead to protect against opportunism on the part of machinery customers. In particular, they find that damages declined with the remaining duration of the lease, where in contrast a scheme to extract rents from entrants would suggest a consistent penalty. On balance, it appears that the Aghion and Bolton analysis of exclusive dealing contracts as potentially anticompetitive does not support strong action against exclusive dealing contracts as the damages that such contracts cause will be limited in size to circumstances 40 

Liquidated damage clauses that bear some relationship to the true damages suffered by a firm in the event of a breach of contract are enforceable, but penalty clauses are not. The Aghion and Bolton theory requires enforceable penalty clauses, as the damages must be tied to the efficiency benefits expected to be achieved by a new entrant, not the damages actually suffered by the incumbent. See Lake River Corp. v. Carborundum Co., 769 F. 2d 1284 (7th Cir. 1985). Note also that the Landeo and Spier analysis discussed above shares the assumption that penalties are enforceable. 41  258 U.S. 451 (1922).

318   Howard Marvel where a new entrant offers only a modest advance in efficiency over the incumbent supplier and that simply refusing to enforce penalty clauses may well prove a more effective way to prevent anticompetitive harm compared to direct attacks on exclusive dealing.

13.3.4.  Ex Post Exclusive Dealing The role of scale economies is central to both the “naked exclusion” and entrant-tax-bycontract theories of anticompetitive exclusive dealing. With naked exclusion, the incumbent can keep rivals from amassing a market share essential to achieve efficient scale by tying up a substantial portion of existing distributors with exclusive dealing contracts. It is, of course, crucial that the distributors in question cannot be cloned readily, for otherwise new distributors would emerge that could undercut existing dealers. Thus in order for exclusive dealing to work, existing dealers must constitute an essential distribution component for both the incumbent and potential entrant suppliers. The Aghion-Bolton approach is different, in that it assumes that an entrant can offer to purchase existing distribution, paying distributors to breach their existing distribution contracts. But like the naked exclusion arguments, the requirement is that the entrant acquire existing distributors as opposed to relying on new entrant distribution for its product. Note, however, that while new distributors must find it difficult to enter and expand, existing distributors must also be sharply constrained. If not, just a few existing distributors who defected from the incumbent producer could use the low prices available to dealers of an entrant to grab a very substantial share of the market in question, thereby undoing any anticompetitive effects of exclusive dealing (Fumagalli and Motta 2006).42 The economic theories of exclusive dealing discussed thus far have grappled with Robert Bork’s (1978) claim that distributors must be compensated by their suppliers for accepting exclusive dealing contracts.43 But there is also a possibility that dealers have locked themselves into distribution arrangements with an incumbent, and that as a result, the incumbent can unilaterally demand exclusive dealing as a way of appropriating customers that distributors would prefer to market to rival manufacturers. This ex post opportunism has been recognized as an important impetus for exclusive dealing, as this passage from a leading antitrust law treatise indicates: [S]‌uppose an established manufacturer has long held a dominant position but is starting to lose market share to an aggressive young rival. A set of strategically planned 42  See, however, Simpson and Winkelgren (2007) for an argument that as long as some product differentiation is present, tough postentry competition can be avoided if an incumbent monopoly supplier cuts its dealers in on some of its rents in return for their loyalty. The strategy works only if the dealers anticipate truly brutal competition in the wake of successful entry. 43  While the Rasmusen, Ramseyer, and Wiley approach can lead to retailers who fear being trapped without access to a crucial product and thus sign exclusive dealing contracts that need not carry a premium, a supplier may well wish to offer small price benefits to dealers who sign up before the incumbent locks in a scale of operation that forecloses potential rivals.

Exclusive Dealing   319

exclusive-dealing contracts may slow the rival’s expansion by requiring it to develop alternative outlets for its product, or rely at least temporarily on inferior or more expensive outlets. Consumer injury results from the delay that the dominant firm imposes on the smaller rival’s growth. (Areeda and Hovenkamp 2002, ¶1802c, at 64)

Forced to choose which brand to deal with, a distributor may well choose to stay with an incumbent as opposed to defecting to a rival. With a fixed number of distributor outlets available, an incumbent can thus exploit its position simply by refusing to share its dealers with rivals. It will be very difficult to separate anticompetitive motivations for exclusive dealing in such cases from a supplier’s desire to protect the customers its brand name or promotional activity drives to its dealers, but it is possible that the desire to cripple rival entrants can be at least a part of the explanation for the imposition of exclusive dealing. The presence of locked-in dealers makes the use of exclusive dealing requirements less expensive to an incumbent supplier. Exclusive dealing can also be a problem when a supplier asserts its right to customers that it has not, in fact, generated, or, more generally, that the supplier has generated in cooperation with other firms selling complements. Consider a supplier of a core system product whose product cooperates with producers of complementary products to make its system more attractive to consumers. For example, a computer manufacturer will find its system demand enhanced when producers design complementary applications for its system. Suppliers will wish to cooperate with producers of complementary products, recognizing that their activities enhance its own sales. But what happens if a rival appears? If the cooperating manufacturers offer variants of their products designed to work with the new entrant, the new entrant’s system becomes more capable and demand for its product increases in consequence. To combat the competitive effect of making complementary products to rivals, the incumbent may copy the products of the complementary manufacturers and then insist that its distributors handle only its product. The benefit that the incumbent obtains will be of limited duration, and the incumbent risks making the new competitor more and not less attractive (since the new entrant obtains the products of the complementary suppliers), but it is possible that the incumbent can extend the life of its system for a time by this sort of exclusive dealing.44 If distributors are locked into a product, the product’s supplier can assert a property right to customers who wish to purchase the product even if at least a portion of the demand for the product ultimately derives from the efforts of cooperating manufacturers. The conditions for such an assertion of a property right to be profitable for a supplier are substantial. The supplier must have very substantial monopoly power over the system it 44 

See, for example, the complaint in RealNetworks v. Microsoft, http://news.cnet.com/pdf/ne/2003/​ complaint.pdf. RealNetworks complained that Microsoft’s agreements with its distributors (“PC Makers and others”) “had the effect of restricting the distribution, installation or promotion of RealNetworks’ or other competitors’ products, unreasonably restrained and substantially lessened competition in the relevant markets.” Microsoft ultimately agreed to cooperate with RealNetworks and paid $761 million to settle the suit. See Jonathan Krim, RealNetworks, Microsoft Settle Suit, Washington Post, October 12, 2005, http://www.washingtonpost.com/wp-dyn/content/article/2005/10/11/AR2005101100661.html.

320   Howard Marvel controls. Distributors must have no significant options to handling the supplier’s product and accepting ex post restraints that are not in their long-term interest. Rival suppliers must not have available alternative paths to the market. The monopoly supplier must be prepared to accept that its actions will discourage future demand-increasing investments by complementary suppliers who see their own property rights to innovations at increased risk. As Areeda and Hovenkamp suggest, this harmful exclusive dealing is most likely to occur when an incumbent faces changes in its position in the market. A supplier may have cooperated with its distributors or complementary suppliers to build a following for its product, but when that supplier’s position is threatened by entrants, as, for example, when a patent expires, it may try to prolong its position by unilaterally asserting a property right to consumers through exclusive dealing. Note the difference between this situation and the economic models of harmful exclusive dealing discussed above. The initial development of the supplier’s market may have occurred with the voluntary cooperation of its dealers and complementary suppliers, but the exertion of exclusive dealing is anything but voluntary. Thus this ex post opportunistic use of exclusive dealing contrasts sharply with exclusive dealing that protects ongoing investments in generating customers, and which does not require a monopoly position on the part of the supplier demanding exclusive dealing. In practice, however, it will be difficult to distinguish ex post opportunism from contracting to protect a stream of customers that the supplier demanding exclusive dealing generates by dint of its ongoing promotional efforts.

13.4.  Assessing Harm We have seen that the economic theories that yield anticompetitive effects of exclusive dealing are, as best, suggestive of where one might look for welfare reductions flowing from the practice, and, indeed, that harmful exclusive dealing may occur when a party has locked itself into an arrangement that puts it at risk of an inappropriate assertion of a property right by an exclusive dealing supplier. It should not be surprising, then, that courts have resorted to rough-and-ready rules of thumb to determine whether particular uses of exclusive dealing are anticompetitive. Consider, for example, this discussion of the requirements for such a finding: There is no set formula for evaluating the legality of an exclusive dealing agreement, but modern antitrust law generally requires a showing of significant market power by the defendant, substantial foreclosure, contracts of sufficient duration to prevent meaningful competition by rivals, and an analysis of likely or actual anticompetitive effects considered in light of any procompetitive effects. Courts will also consider whether there is evidence that the dominant firm engaged in coercive behavior and

Exclusive Dealing   321

the ability of customers to terminate the agreements. The use of exclusive dealing by competitors of the defendant is also sometimes considered.45

This approach, while far from precise, does represent an evolution from simply accepting that exclusive dealing is harmful when rivals are denied access to a substantial portion of distribution available in a market: In decades past, courts analyzed nonprice related exclusionary conduct simply by inferring competitive harm where a substantial percentage of the market was foreclosed to rivals. Today, courts take the analysis one step further. Rather than simply calculating the percentage of the market foreclosed, courts also examine how the exclusionary conduct affects competition and whether any competitive harm results from the exercise of market power, rather than from unrelated factors (e.g., consumer choice, inefficiency of competitors). Similarly, when market foreclosure occurs at the distribution level, courts assess whether competitors can simply circumvent the foreclosed distribution channels and reach end-users through alternative means (i.e., whether entry into the distribution of the product is easy). This trend toward a more probing analysis of competitive effects is not surprising with the advent of more sophisticated economic analyses. As noted by Jonathan Jacobson, the focus of the antitrust inquiry has moved from considering whether the conduct foreclosed competition, to whether “the foreclosure or other aspect of exclusion was imposed in a way designed to lead to an increase in prices or restriction of output in the market as a whole. (Jacobson and Sher 2006, 786)

If suppliers compete with one another to drive customers to their dealers, the net effect of their actions will be to increase output. Clearly in the case of hearing aids distributed through dealers, the suppliers who employed exclusive dealing did so to protect the streams of customers that their promotion generated. The hearing aids market was an easy case, particularly given the collapse of hearing aid manufacturers who dropped exclusive dealing. But for other products, it will be difficult to determine whether exclusive dealing protects ongoing investments in new customers, or whether it instead prevents rival suppliers from accessing a stable pool of customers by offering lower prices. For artificial teeth, a similar argument for efficiencies associated for exclusive dealing was rejected emphatically by the district court. Unlike for hearing aids, the market for artificial teeth was not expanding overall, as improvements in dentistry and fluoridation had resulted in decreased use of dentures. Competition in the artificial tooth market has instead focused on driving demand for better looking, longer wearing “premium” teeth. The district court in Dentsply rejected Dentsply’s argument that its exclusive dealing was designed to protect its investment in customers generated by its promotion, finding that 45  ZF Meritor, LLC v. Eaton Corp., 2012 U.S. App. LEXIS 20342, 2012-2 Trade Cas. (CCH) P78078 (3d Cir. Del. 2012) (citations omitted).

322   Howard Marvel Dentsply, with its market share in excess of 75% for premium teeth, “has not reacted with lower prices when others have not followed its price increases,” and that its exclusive dealing “has clearly been anticompetitive.” Dentsply’s business justification for its exclusive dealing was “merely pretextual.”46 Dentsply, by using its sales representatives to promote premium teeth to its customers, dental laboratories, had clearly driven customers to the higher end segment of the market. But the rapid growth of the higher end was not sufficient to convince the judge that Dentsply’s promotional efforts were sufficiently competitive to overcome the exclusion of premium teeth rivals from its dealers. Nonetheless, the district court did not condemn Dentsply’s exclusive dealing, as it was determined that exclusive dealing did not close distribution to rivals, and that Dentsply’s own distribution was not crucial to those rivals’ success. It is worth considering the findings in this very significant case in more detail. Dentsply was by far the largest supplier of artificial teeth to dental laboratories, where those teeth were fabricated into custom dental appliances such as complete dentures (replacements of all the teeth in one or both [upper and lower] dental arches), or partial dentures replacing a subset of teeth in an arch. Dentsply distributed its teeth to the laboratories through a network of dental supply dealers who maintained stocks of teeth of varying shapes and shades. Because the laboratories often ordered artificial teeth only when a customer’s existing teeth had been removed and a casting of the gum for the desired dental appliance had been made, labs placed a substantial premium on being able to obtain the necessary teeth in the appropriate shape and color very quickly. The tooth dealers also maintained tooth desks that provided advice on available tooth products. While Dentsply was the largest manufacturer of artificial teeth, rival tooth manufacturers also offered their teeth through dealers who were authorized to sell Dentsply teeth. The situation changed when Dentsply introduced “premium” lines of artificial teeth distinguished by improved material formations that resulted in more luminous and better wearing teeth. Dentsply’s smaller rivals did not follow in its path of introducing more aesthetic and longer lasting teeth, so that Dentsply’s efforts to promote these teeth were not threatened by the possibility of dealers recommending substitute products to lab customers who came to the dealers seeking Dentsply teeth. There were, however, rival producers of premium teeth, principally two large European firms, Vita Zahnfabrik and Ivoclar, who possessed both premium tooth product lines and sales forces that were in a position to promote those teeth to the dental laboratories. Their entree to the labs was a consequence of the large crown and bridge businesses that each possessed. In contrast to artificial teeth that are embedded in custom acrylic bases fitted to consumer gums, crown and bridge materials are sold to labs who then fabricate them into permanent fittings that are permanently implanted. Both European competitors had crown and bridge business that were larger than that of Dentsply, and accordingly, had substantial relationships with the bulk of the dental laboratories to whom Dentsply sold its teeth.

46 

United States v. Dentsply Int’l, Inc., 277 F. Supp. 2d 387 (D. Del. 2003).

Exclusive Dealing   323

Each of these competitors sought to distribute its products through Dentsply dealers. Dentsply responded by requiring any dealer who wished to stock Dentsply teeth to forgo the rival premium teeth products, though dealers were allowed to continue to distribute the lower-end lines that they had handled previously. Dentsply’s contracts did not lock in dealers, as they were terminable at will by either party. The rival premium tooth manufacturers complained to the Department of Justice that Dentsply’s exclusive dealing excluded them from the marketplace, and the Department, in turn, sued Dentsply to stop the practice. The case combined a very high market share for Dentsply, estimated to be about 80% for premium teeth, with no long-term dealer commitments. DOJ initially attempted to argue that the dealers were locked into Dentsply by their existing tooth stocks. But the one example of a dealer who added Ivoclar teeth did not support the allegation—far from being locked in, that dealer, cut off by Dentsply, was buying Dentsply teeth from authorized Dentsply dealers in order to maintain its access to Dentsply customers. This behavior was consistent with the theory that the dealer wished to convert customers who came to it for Dentsply teeth to those of the rival, and needed to continue to offer Dentsply’s teeth until the labs seeking Dentsply teeth could be converted, if ever. As we have seen, the district court rejected Dentsply’s explanation that it was attempting to prevent Ivoclar from poaching its customers, but nevertheless found for Dentsply. Its finding was based on Dentsply’s argument that rival tooth suppliers had efficient distribution alternatives, principally direct distribution that did not require dealer tooth stocks. Support for this conclusion was provided by Dentsply’s own distribution— while continuing to use dealers, it drop-shipped an increasing portion of its tooth sales directly to laboratories. Accordingly, the number of dealer tooth stocks was dropping rapidly as falling prices for overnight delivery reduced the importance of local tooth stocks. Dentsply lost on appeal, however, as a Third Circuit chose not to accept the district court’s finding of sufficient alternative distribution. The Third Circuit instead held that Dentsply possessed a market share so high that it must have been successfully excluding rivals, and that the exclusion must have been anticompetitive. That some manufacturers resort to direct sales and are even able to stay in business by selling directly is insufficient proof that direct selling is an effective means of competition. The proper inquiry is not whether direct sales enable a competitor to “survive” but rather whether direct selling “poses a real threat” to defendant’s monopoly. See Microsoft, 253 F.3d at 71. The minuscule 5% and 3% market shares eked out by direct-selling manufacturers Ivoclar and Vita, Dentsply’s “primary competitors,” FF26, 36, 239, reveal that direct selling poses little threat to Dentsply. (399 F.3d 181, 193)

Was this finding correct? The Dentsply case illustrates the value of follow-up impact evaluation of vertical restraints decisions. Ivoclar, given the opportunity to distribute through Dentsply dealers, chose instead to use the alternative channel that the district court had found was available to it. Doing so required that Ivoclar promote its teeth

324   Howard Marvel directly to the dental laboratories. This it managed by modifying its incentives to its sales representatives to concentrate not only on crown and bridge (its previous strategy), but to sell teeth as well. With its own demand in place, it chose not to employ the Dentsply dealers open to it. Its strategy has apparently been a success. In contract, Vita Zahnfabrik did employ dealers, but unsuccessfully. Indeed, while Dentsply dealers did add Vita’s teeth, no Dentsply dealer has continued to stock those teeth. The reason appears to be that Vita’s teeth were priced higher to dealers than comparable Dentsply teeth and so free riding was not feasible. The experience of Dentsply also suggests care in assessing the alternatives against which exclusive dealing distribution should be compared. Dentsply dealers handled much more than teeth, suggesting an economies-of-scale barrier to starting rival dealers was a concern. But in fact, the principal alternative to controlled dealer distribution turned out to be vertical integration, not tooth dealers.

13.5.  Exclusive Dealing and Dealer Property Rights We have thus far considered cases in which suppliers impose exclusive dealing to protect customers generated by a supplier’s own efforts. This need not be the case for all exclusive dealing arrangements. Instead, exclusive dealing may define customers generated by distributors, permitting access to such customers to be transferred to a supplier. That is, in return for a payment from a supplier to distributors, those distributors may agree to suggest to their customers that they purchase the supplier’s products. The suggestion may take the form of a recommendation or may simply involve featuring the product more prominently in the distributor’s assortment. Exclusive dealing ensures that the customers are delivered to the best of a distributor’s ability, while partial exclusive dealing transfers a defined portion of the distributor’s customers to the supplier.47 At one time, advertising, particularly successful advertising, was highly suspect from the standpoint of antitrust. Firms that were successful in attracting customers for their products were thought to possess monopoly power over those customers. It is now more common to interpret advertising as a competitive tool for attracting those customers from rivals. But recall that vertical restraints represent methods to coordinate the actions of firms at different stages of the production/distribution process. Thus a firm may chose to advertise its products directly to consumers, or, alternatively, it may farm out the customer acquisition to its distributors. Typically firms do both. Some customers may be drawn to the brand, and here exclusive dealing can prevent 47 Note that the decision to grant exclusive network provider status is an example of this kind of exclusive dealing.

Exclusive Dealing   325

dealers from switching those customers to rivals. But suppose that some customers are drawn to a dealer, and that that dealer can steer those customers to one or the other manufacturer. A manufacturer may be willing to pay for such customers, but is most likely to want to pay for only the portion of its customers that the dealer is able to steer in its direction. The result will be a payment for those customers, but only if the customers are actually delivered. How can property rights be specified to ensure that compensation for the customers actually generates the promised customers? A lump sum payment for a group of customers will handle this. The supplier offers a lower wholesale price in return for more sales—this is competitive compensation to the dealers who control those sales. A payment to a dealer or distributor for such customers thus serves a purpose similar to an advertising expenditure for a message designed to reach consumers directly. Payments by suppliers to dealers for customers driven to the supplier’s products ordinarily will be uncontroversial. But as with traditional exclusive dealing, when a supplier possesses a large share of a market, the purchase of even more customers may raise concerns about anticompetitive behavior. For example, some retailers prefer to rely on a supplier to organize the merchandising of a particular class of products. Suppliers, in turn, compete to be named the category manager for the product class in question. The supplier named category manager thereby obtains enhanced access to a retailer’s customers. The category manager will need to include rival products in a retailer’s assortment if those products are important to at least some of the retailer’s customers, but when consumers are indifferent between products of rival supplier, the category manager can include its own products to the exclusion of rival products and the retailer does not suffer as a result. Moreover, when rival products are included in a retail assortment, that assortment can still be arranged in ways that favor the category manager’s products. Like advertising, partial exclusive dealing steers customers to particular suppliers. It thereby harms rivals and by making customer acquisition tougher, can raise the rivals’ costs. But as Klein and Murphy (2008) emphasize, the ability to buy exclusive access to at least a portion of a distributor’s customers permits competition among suppliers for the rights to customers that distributors control access to. Whether the competition results in better options for consumers depends on the degree to which distributors must compete for the consumers they at least temporarily control. Anticompetitive category management would impair the distribution of the category manager’s rival products, increasing rivals’ costs and reducing the attractiveness of the category to a retailer’s customers. This was claimed to be the outcome of category management by U.S. Tobacco for sales of smokeless tobacco.48 Conwood, the second largest producer of smokeless tobacco, did not offer category management services, but complained that when U.S. Tobacco was appointed category manager for a retailer, Conwood’s products suffered as a result. A  jury found 48 

Conwood Corp, v. United States Tobacco Co., 290 F. 3d 768 (6th Cir. 2002).

326   Howard Marvel for Conwood, awarding $350 million (before trebling). But the focus on the effects of category management on Conwood, rather than on whether category management increased the success of retailer assortments, left unresolved the question of whether competition for the position of category manager benefited or harmed consumers.49 One firm’s purchase of customers from distributors will almost certainly increase the costs to rival firms of reaching those same customers; relying on the presence of such an effect on rivals threatens to commit the error of focusing on the effect of a practice on competitors, rather than on competition. A recent case suggests that Conwood’s finding that category management by a supplier with a high share is anticompetitive is not the final word on the issue. Church & Dwight (C&D) is the leading seller of condoms in the United States, with market share in excess of 75%, up from 67% in 2001. Its two leading rivals have shares of 14% and 10%, so that the three leaders account for in excess of 99% of the US market.50 A much smaller rival, Mayer Laboratories, challenged C&D’s widespread category management and its unwillingness to include Mayer’s competing products in retail assortments under its control. The district court ruled that market share losses could not be clearly attributed to a lessening of competition in the market overall. There was no showing of a reduced output or raised prices in the market as a whole.51

13.6. Conclusion Exclusive dealing will likely remain a controversial segment of antitrust policy. There are clearly strong procompetitive reasons for its use in a wide variety of settings. But since suppliers can create property rights to customers unilaterally, and, given locked-in distributors, can enforce those property rights with anticompetitive effects, the boundaries of efficient and therefore permissible exclusive dealing will continue to be fought over for the foreseeable future. Standards, at least in the United States, have become much more lenient over time, and this evolution may continue. But it is unlikely that any dominant firm will be able to adopt exclusive dealing without triggering charges that its actions harm competition.

49 

See Wright (2009) for an analysis of the case. See Church & Dwight v. Mayer Laboratories, 2012 U.S. Dist. LEXIS 51770; 2012-1 Trade Cas. (CCH) P77,863 (N.D. Cal. 2012). On a worldwide basis, the three firms together have a share of 62%, with C&D’s 11% share trailing those of each of its two large rivals. Thus it seems unlikely that the rivals’ existence is threatened by insufficient scale. 51  Church & Dwight v. Mayer, at 51. 50 

Exclusive Dealing   327

References Aghion, Philippe, and Patrick Bolton. 1987. Contracts as a Barrier to Entry. American Economic Review 77: 388–401. Areeda, Phillip E., and Herbert Hovenkamp. 2002. Antitrust Law. 2nd ed. New York: Aspen Publishers. Bernheim, B. Douglas, and Michael D. Whinston. 1998. Exclusive Dealing. Journal of Political Economy 106: 64–103. Bork, Robert H. 1978. The Antitrust Paradox. New York: Basic Books. Frasco, Gregg P. 1991. Exclusive Dealing: A Comprehensive Case Study. Lanham, MD: University Press of America. Fumagalli, Chiara, and Massimo Motta. 2006. Exclusive Dealing and Entry:  When Buyers Compete. American Economic Review 96: 785–95. Heide, Jan B., Shantanu Dutta, and Mark Bergen. 1998. Exclusive Dealing and Business Efficiency: Evidence from Industry Practice. Journal of Law and Economics 41: 387–407. Innes, Robert, and Richard J.  Sexton. 1994. Strategic Buyers and Exclusionary Contracts. American Economic Review 84: 566–84. Jacobson, Jonathan M. 2002. Exclusive Dealing, “Foreclosure,” and Consumer Harm. Antitrust Law Journal 70: 387–407. Jacobson, Jonathan M., and Scott A.  Sher. 2006. “No Economic Sense” Makes No Sense for Exclusive Dealing. Antitrust Law Journal 73: 779–801. Klein, Benjamin. 2003. Exclusive Dealing as Competition for Distribution “On the Merits.” George Mason Law Review 12: 119–62. Klein, Benjamin, and Kevin M. Murphy. 2008. Exclusive Dealing Intensifies Competition for Distribution. Antitrust Law Journal 75: 433–66. Krattenmaker, Thomas G., and Stephen C.  Salop. 1986. Anticompetitive Exclusion:  Raising Rivals’ Costs to Achieve Power over Price. Yale Law Journal 96: 209–93. Landeo, Claudia M., and Kathryn E. Spier. 2009. Naked Exclusion: An Experimental Study of Contracts with Externalities. American Economic Review 99: 1850–77. Landeo, Claudia M., and Kathryn E. Spier. 2012. Exclusive Dealing and Market Foreclosure: Further Experimental Results. Journal of Institutional and Theoretical Economics 168: 150–70. Marvel, Howard P. 1982. Exclusive Dealing. Journal of Law and Economics 25: 1–25. Marvel, Howard P. 1984. Vertical Restraints in the Hearing Aids Industry. In Ronald N. Lafferty, Robert H. Lande, and John B. Kirkwood, Impact Evaluation of the Federal Trade Commission Vertical Restraints Cases. Washington, DC: Federal Trade Commission, 270–384. Maston, Scott E., and Edward A. Snyder. 1993. United States versus United Shoe Machinery: On the Merits. Journal of Law and Economics 36: 33–70. Posner, Richard A. 2001. Antitrust in the New Economy. Antitrust Law Journal 68 (3): 925–43. Rasmusen, Eric B., J. Mark Ramseyer, and John Shepard Wiley Jr. 1991. Naked Exclusion. American Economic Review 81(5): 1137–45. Rasmusen, Eric B., J. Mark Ramseyer, and John Shepard Wiley Jr. 2000. Naked Exclusion: Reply. American Economic Review 90(1): 310–11. Salop, Steven C., and David T.  Scheffman. 1983. Raising Rivals’ Costs. American Economic Review 73: 267–71. Salop, Steven C., and David T. Scheffman. 1987. Cost-Raising Strategies. Journal of Industrial Economics 36: 19–34.

328   Howard Marvel Sass, Tim R. 2005. The Competitive Effects of Exclusive Dealing: Evidence from the U.S. Beer Industry. International Journal of Industrial Organization 23: 203–25. Segal, Ilya R., and Michael D. Whinston. 2000. Naked Exclusion: Comment. American Economic Review 90: 296–309. Simpson, John, and Abraham L.  Wickelgren. 2007. Naked Exclusion, Efficient Breach, and Downstream Competition. American Economic Review 97: 1305–20. Whinston, Michael D. 2006. Lectures on Antitrust Economics. Cambridge, MA: MIT Press. Wright, Joshua D. 2009. Antitrust Analysis of Category Management: Conwood v. United States Tobacco Co. Supreme Court Economic Review 17: 311–37. Zenger, Hans. 2010. When Does Exclusive Dealing Intensity Competition for Distribution? Comment on Klein and Murphy. Antitrust Law Journal 77: 205–11.

CHAPTER 14

T YING ARRANGEMENTS ERIK HOVENKAMP AND HERBERT HOVENKAMP

14.1.  Introduction; Identifying Ties Tying arrangements, sometimes known as “ties,” “tie-ins,” “tied-in sales,” or “bundles,” occur when a firm offers two separate products together, refusing to sell one of them without the other. Identifying when two things are really a single product—such as a shirt and its buttons or an automobile and its tires—has proven controversial. The dominant position looks at ordinary business practices in order to determine whether the products are commonly sold separately. In its Jefferson Parish decision (1984) the Supreme Court held that the separate product inquiry depends “not on the functional relation” between two goods, but rather “on the character of the demand for them.” This query requires the plaintiff to show simply that the two products are commonly sold separately under ordinary market conditions. As a result the “separate products” requirement serves mainly as a screening device to weed out frivolous cases involving goods (such as a right shoe and a left shoe) that are rarely sold separately (e.g., Reisner case [tying of car to drive train]; Areeda and Hovenkamp 2011, ¶¶1741–51). This definition is in contrast to a more substantive one that a single product should be found when there are “rather obvious economies of joint provision” (Jack Walters case).1 That test would merge a test for efficiency, otherwise a defense, into the definition of separate products. Ties can come in one-way and two-way varieties. For example, Baskin-Robbins as franchisor might insist that its franchisees sell its own Baskin-Robbins brand ice cream; 1  Four justices in Jefferson Parish preferred this view: “When the economic advantages of joint packaging are substantial the package is not appropriately viewed as two products, and that should be the end of the tying inquiry.” Jefferson Parish, 466 U.S. at 40–41.

330   Erik Hovenkamp and Herbert Hovenkamp however, it might willingly sell the ice cream alone to independent retailers or consumers. Two-way ties are sometimes referred to as “bundles.” In a one-way tie we can readily speak of a “tying product” (e.g., the Baskin-Robbins franchise) and a “tied product” (the ice cream). In a two-way tie each product serves both functions. Bundling arrangements often involve more than two products. For example, a patent “package license” may include several hundred patents (e.g., Automatic Radio case), and the “blanket license” agreements used in the music broadcasting industry include several thousand compositions (e.g., Broadcast Music case). So-called bundled discounts typically operate in both directions. For example, the seller may offer A and B individually at given prices, but a lower price for someone who takes an A + B bundle. In any event, the tying condition must be something that prevents a customer from purchasing the goods separately or else that creates a disincentive to do so. The orthodox tying form is the “contractual” tie, in which a contract requires the buyer to take the two products together. Legally, tying can also be inferred from an observed practice of refusal to make untied sales (Areeda and Hovenkamp, 2011, ¶¶1755–56). Tying can also be inferred from a discount that attaches when the buyer takes two or more products together (see Hovenkamp and Hovenkamp 2008; 2009), or from a technological design or interoperability requirement that effectively forces the customer to use the two products together even if she is not contractually required to do so. Well-known examples are Microsoft (2001), which condemned the blending of Internet Explorer code with the Windows computer operating system code, and Berkey Photo, which refused to condemn a camera and film design under which the two would work only with each other (Berkey Photo, 1979; see Bohannan and Hovenkamp 2012, 321–22; Evans and Salinger, 2005). If a customer is free without significant constraint to purchase two products separately as well as together, there is no tie.

14.2.  Statutory Coverage and Idiosyncratic “Per Se” Rule The Sherman Act does not explicitly reach tying arrangements, but rather agreements that restrain trade or actions by dominant firms that monopolize markets (15 U.S.C. §1 [contractual restraints on trade], §2 [monopolization]). In 1912 the Supreme Court confronted its first tying case in Henry (1912), which upheld the right of a maker of office mimeograph machines to require purchasers to use the machine exclusively with its own ink and stencils. The case was not brought by the buyer as an antitrust action, but rather by the seller as a patent infringement action for sales in violation of the patent license. The Court held that the tying of patented and unpatented products was not an unwarranted extension of the patent. Congress was not happy with the Henry decision. Instead of addressing the legal merits of the Supreme Court’s holding by modifying the Patent Act, however, Congress

Tying Arrangements   331

placed an antitying provision in the Clayton Act (15 U.S.C. §14), which expanded the antitrust laws to as to prohibit anticompetitive contractual tying. It changed patent doctrine only forty years later.2 Congress’s displeasure with Henry may also be read as its conclusion that a tie should be unlawful even if it threatened no monopoly in the secondary market, given that the stencils and ink in Henry were common commodities. Today the antitrust legality of tying is assessed under four different statutes. Most explicitly, Section 3 of the Clayton Act makes it unlawful for a firm to sell or lease a good subject to a condition or discount requiring the buyer not to purchase goods from a rival, and where the effect “may be to substantially lessen competition or tend to create a monopoly” (15 U.S.C. §14). The statute applies to both patented or unpatented “goods” or “commodities” but excludes services or other intangibles. In addition, Section 1 of the Sherman Act reaches agreements “in restraint of trade,” which includes anticompetitive tying and is not restricted to goods. Section 2 of the Sherman Act condemns tying when it is shown to be an exclusionary practice by a monopolist, or dominant firm. Finally, Section 5 of the Federal Trade Commission Act (15 U.S.C. §45(a)(1)) can also be used against tying, although that provision may be enforced only by the Federal Trade Commission. During the period from the Clayton Act’s passage through the 1970s Supreme Court doctrine was particularly hostile toward tying. The Court’s Standard of California decision (1949) declared that ties “serve hardly any purpose beyond the suppression of competition.” The Court developed a “per se” rule against them under Section 1 of the Sherman Act “whenever a party has sufficient economic power with respect to the tying product to appreciably restrain free competition in the market for the tied product and a ‘not insubstantial’ amount of interstate commerce is affected” (Northern Pacif. Rwy, 1958).3 This rule differs from the per se rules applied to such practices as naked price fixing because it requires a showing of (1) market power in the tying product and (2) at least minimal impact on competition. Further, it has been interpreted to permit a range of justifications or defenses that are not ordinarily available in per se antitrust cases. The result of these requirements is that, notwithstanding the per se label, most ties today are found to be lawful. In addition, ties can also be condemned under a rule of reason under either Section 3 of the Clayton Act or Section 1 of the Sherman Act. Further, certain unilateral monopolistic practices such as technological ties that do not meet the “agreement” requirements in these statutes can be condemned under the antimonopoly provision, Section 2 of the Sherman Act, which always requires a showing of actual or threatened market power and anticompetitive effects.

2  In the 1952 Patent Act Congress provided that the doctrine of contributory patent infringement could not be applied to the sale of a “staple article or commodity of commerce suitable for substantial noninfringing use.” Patent Act, ch. 950, 66 Stat. 811 (1952) (codified as amended at 35 U.S.C. § 271(c)). As a result, Henry’s sale of ink capable of other uses could not be the basis of an infringement action. 3  On the meaning of a “not insubstantial” volume of tied commerce, see Areeda and Hovenkamp, 2011, ¶1721.

332   Erik Hovenkamp and Herbert Hovenkamp

14.3.  Minimum Conditions for Competitive Harm; Market Power Almost everyone agrees that tying is harmless if the markets for the tying and tied product are both structurally competitive. If a competitive firm attempts tying that is undesirable to consumers, they will go elsewhere. As a result, any tying that exists in such markets must make consumers better off. The situation is more complex if a seller has market power in one or both the tying and tied markets. While market power in the tying product has generally been stated as a legal requirement for unlawful tying, the courts have not always taken the requirement very seriously and have inferred sufficient power from phenomena that indicate very little power at all. The Supreme Court has permitted sufficient power to be inferred from the tying product’s “uniqueness” (Loew’s, 1962), from the fact that the defendant was able to impose a “host” of tying arrangements (Northern Pacif. Rwy, 1958), and from the fact that purchasers may be “locked in” by a previous purchase and thus required to buy the seller’s aftermarket parts or service (Kodak, 1992). A few lower courts have even held that a long-term contract can impose the requisite power (e.g., Collins, 1997). Most of these rationales arose out of a deep suspicion as well as a poor economic understanding of ties, which presumed them to be anticompetitive in most cases and thus warranting per se condemnation. While none of these cases has been expressly overruled, their approaches to power are largely ignored or else construed very narrowly. The Kodak lock-in doctrine applies mainly when the defendant requires previous purchasers of its durable good to use its aftermarket parts or service. The firm may be nondominant in the primary market, as Kodak was in that case, but the effect of the lock-in doctrine is to create a relevant market for the firm’s own brand. The theory, quite simply, is that once a person has purchased a durable good such as a photocopier, “switching costs” are so high in relation to aftermarket value that the firm can be forced to pay monopoly prices for tied specialty products or service. The Supreme Court rejected the defense that a rational customer would engage in “lifecycle” pricing, attributing high aftermarket prices to the overall price. But the Court found two possible exceptions: first, there might be a significant group of “myopic” customers who are initially beguiled by a low price on the primary good and ignore the high aftermarket prices. Second, a firm might increase aftermarket prices late in a product’s lifecycle, perhaps sacrificing some prospective hardware sales but earning more from the “installed base” of customers who have previously purchased. As the quotation marks suggests, Kodak has produced its own vocabulary of specialty terms. Nevertheless, the doctrine has been sharply criticized and the courts generally construe it narrowly. For example, they do not apply it when the purchasers had adequate knowledge of downstream prices when they made their initial choice (e.g., Hack, 2001).

Tying Arrangements   333

Finally, the Supreme Court had previously held that the seller’s ownership of a tying product covered by a patent (International Salt, 1947) or copyright in the motion picture “block booking” cases (Loew’s, 1962; Paramount, 1948) created a presumption of sufficient power, and a few lower courts had extended this presumption to trademarks (e.g., Siegel, 1971). The Supreme Court overturned this presumption in its 2006 Illinois Tool Works decision, and today market power in tying products protected by intellectual property rights must be established through the ordinary tools for showing power. At this writing the Supreme Court has not overruled its legal conclusion that ties are unlawful “per se” when tying product power is proven and the tie affects a substantial volume of commerce. The existence of this per se rule has had one very unfortunate consequence: litigants in tying cases have not been required to make a record documenting the true effects of tying, given that those effects were simply presumed. A few passages in Illinois Tool Works suggest that the per se tying rule might be ripe for reconsideration. Market power is a necessary but hardly a sufficient condition for competitive harm from tying. Indeed, at least presumptively a tie does not cause any additional welfare harm beyond that which already exists in the untied monopoly market. Further, two common economic effects, price discrimination and control of double marginalization, are more likely to increase than to decrease welfare from the stand-alone monopoly level. Finally, various benefits of tying, including production or distribution economies or improvements in product quality, can apply to dominant firms and competitors alike.

14.4.  Motives and Effects of Tying 14.4.1.  Rise and Fall of the Leverage Theory The leverage theory, which originated in patent law in the nineteenth century, expressed a concern that a patent owner could use contracts or conditions to “leverage” more power than the patent itself granted. The best-known statement of the doctrine is Justice Brandeis’s conclusion in Carbice that a requirement that those using a firm’s patented refrigeration container also use its dry ice improperly permitted the seller “to derive its profit, not from the invention on which the law gives it a monopoly, but from the unpatented supplies [which are] wholly without the scope of the patent monopoly” (Carbice, 1931). The leverage theory has nothing to do with the exclusion of any rival (foreclosure). Rather, the harm is based on the threat that the tying monopolist will extract higher prices from consumers. This view was largely exploded by Ward S.  Bowman, who observed that when a purchaser wants a combination of two complementary products, the profit-maximizing price is computed by reference to the combination. See Bowman (1957). One can raise the price of the second product only by reducing the price of the

334   Erik Hovenkamp and Herbert Hovenkamp first product. For example, if shoes and laces are complements, a shoe monopolist can extract all available monopoly profits in the price of shoes, and cannot profitably extract more by tying otherwise competitive laces and charging a second monopoly price for them. Consumers attribute value to the combination of shoes plus laces. The leverage theory is clearly a fallacy in situations where the tying product is monopolized and the tied product is competitive. It is also incorrect in cases where both products are subject to the exercise of some market power, because in these situations the elimination of double marginalization is likely to produce lower rather than higher prices, as we shall see below. A type of leveraging may occur when a dominant firm ties sequential rather than simultaneous monopolies, but this is clearly something different than Justice Brandeis and other proponents of the leverage theory had in mind. For example, a firm such as Microsoft, seeing that a rival’s Internet browser with operating system capabilities poses a future threat to its own operating system, might tie its operating system and its browser in order to keep that threat from being realized (Microsoft, 2001). Even in this case, however, the competitive threat resembles foreclosure of rivals more than pure leveraging. See Bohannan and Hovenkamp (2012, 25–26, 269); Whinston (1990).

14.4.2.  Price Discrimination In a variable proportion tie, consumers purchase one unit of the tying product and amounts of the tied product that vary with their use of the tying product. For example, printer manufacturers will frequently tie their printers to their own ink cartridges. Consumers who do more printing must purchase more ink cartridges, but most users buy only one printer. In nearly all litigated cases these tying arrangements involve a price reduction from the nontied level on the tying product, but a markup on the price of the tied product. Indeed, often the tying product is priced at below cost or given away. See Hovenkamp and Hovenkamp (2010). This pricing strategy is frequently referred to as “metering.” A more technical account is presented in Schmalensee (1981), who finds that such arrangements may often be welfare increasing. Price discrimination occurs when the ratio of the average price to marginal cost varies among buyers of the same product. If marginal cost is the same for all customers, then price discrimination occurs whenever two consumers pay different unit prices for the same product. When demand for the tied product varies significantly among buyers of the tying product, variable proportion tying may be used to discriminate among buyers with different intensity levels. By shifting profits to the tied sales, the firm earns greater profits from higher intensity users. Moreover, by reducing the price of the tying good the firm can obtain a profit from lower intensity buyers who would not have purchased the product at all under separate provision. Because all consumers face the same price schedule, metering is an example of second-degree price discrimination. It is important to note that the object of this price discrimination is the combined use of the two products. Individually, both the tying and the

Tying Arrangements   335

tied products are sold at the same nominal price to all. For example, in a tie of printers and ink cartridges it is the price of printing that is discriminatory. Of course, the combined use of the tying and tied products will always have a price schedule of this form, even under separate provision. However, in that case no single firm can utilize this property to discriminate, and so the resulting prices differ from the optimal discriminatory prices. To illustrate suppose that a printer monopolist faces costs of 10 per unit and that the cost of cartridges is 2. Cartridges are sold competitively. The printer maker could sell the printer at a stand-alone profit-maximizing price of 14, earning 4 on each sale, for a return of 40%. It would sell the cartridges at the competitive price. Instead the monopolist sells the printers at its cost price of 10 and ties cartridges, for which it charges 4. The first thing that will happen is that more buyers will come into the market for printers. Output will not rise all the way to the competitive level in this illustration because each customer needs at least one cartridge, so even the lowest intensity user will end up paying 14 rather than the competitive price of 12. Returns (right column in the table below) are then as follows for customers requiring different numbers of cartridge (second column):

Number of Cartridge Total Printer price/cost cartridges price/cost price/cost Return 10/10 10/10 10/10 10/10 10/10 10/10 10/10

1 2 3 4 5 10 20

4/2 8/4 12/6 16/8 20/10 40/20 80/40

14/12 18/14 22/16 26/18 30/20 50/30 90/50

16% 28% 37.5% 44% 50% 67% 80%

Such differences are robust over all situations in which tying involves a transfer of part of the available monopoly overcharge from the tying to the tied product. In the above example, printing costs per page actually decline as usage increases, because the printer price is amortized over more copies while the cartridge cost is constant. However, the seller earns higher returns per customer as usage increases. The welfare effects of variable proportion tying are generally ambiguous. We can safely assume that total profits—the sum of tying and tied good profits—are higher under tying. Otherwise the seller would not tie. However, some consumers benefit from tying, while others are injured by it. As figure 14.1 illustrates, we can divide consumers into three groups: low, medium, and high intensity. Low-intensity consumers are those with relatively low demand who are unwilling to pay the higher tying good price that arises under separate provision. They purchase the goods only under tying and are thus clearly benefitted by the tie. Medium intensity consumers are those who buy the products under both pricing regimes, but who receive more surplus under tying. Because the marginal cost of consumption (the tied good price) increases, these buyers consume less under tying. But the reduction in the tying good’s price is sufficiently large to overcome

336   Erik Hovenkamp and Herbert Hovenkamp Consumer Surplus Separate Provision

Tying

Low FIGURE  14.1  Welfare

Medium

High

Consumer Intensity

Effects of Variable Proportion Tie

this decline, and the result is a net increase in consumer surplus. Finally, high-intensity consumers are those who achieve less surplus under tying. These consumers have relatively high demand, and the price of the tying good makes up a relatively small fraction of their total consumption expenditure. Unlike medium-intensity buyers, the reduction in the tying good price is insufficient to overcome the negative impact of an increased tied good price. The net effect of tying on consumer welfare is the aggregation of the effects on all consumer types. Given the varied surplus effects depicted in figure 14.1, the overall effect of tying on welfare generally depends on the distribution of consumer intensity levels within the population. However, three additional factors can strongly affect the welfare impact of variable proportion ties. These are the impact of tying on producer costs of the tying product, the degree of competition in the tied market, and economies of scale in the tying or both markets. If the marginal cost of the tying product is relatively high—meaning that it is close to the maximum willingness to pay among potential buyers of the tying good—they tying is more likely to increase consumer welfare. In this case, tying permits the seller to price the tying good near or below cost, which results in many more consumers entering the market. In fact, if this marginal cost is sufficiently high, then tying may be the only pricing arrangement in which sales of the tying product are profitable. For example, a device such as a smartphone may require a price of $400 if sold alone, but may be included without charge or at a nominal price when bundled with a two-year service contract. Returns on the phone will depend on consumer usage. The level of competition in the tied market also affects the impact of tying. In particular, consumer welfare under separate provision ordinarily decreases as tied market competition diminishes. Under tying, however, consumer welfare is largely unaffected by the level of tied market competition. The tying product cannot be used with competing

Tying Arrangements   337

versions of the tied product in any event, so the tying firm’s pricing decision is independent of tied market prices. Thus a lack of tied market competition will tend to ease the conditions under which tying improves consumer welfare. The third factor is the extent of scale economies in production of the tying product and, occasionally, the tied product. Because the price cut in the tying product results in greater tying product sales, more units of the tying product will be produced. If fixed costs are substantial, as they often are for manufactured products and particularly for those with a significant R & D component, then the output increase that results from tying will likely produce a lower profit-maximizing price quite aside from the pricing shift to the tied product. The output impact on the tied product is more ambiguous. On the one hand, high-intensity users consume less of the tied product because they pay more for it under tying. On the other hand, low-intensity users would not be in the market at all under separate provision. Any demand that tying creates among them constitutes an output increase. Further, the tie itself switches tied product output away from rivals and toward the tying firm. If the net result is an output increase, then economies of scale may reduce the costs of the tied product as well. In sum, while precise calculation may be impossible, there are good reasons for thinking that price discrimination ties increase aggregate consumer welfare as a general matter.

14.4.2. Foreclosure The dominant defense of antitrust tying doctrine today is that ties may unreasonably foreclose, or exclude, rivals, particularly in the tied product market. For example, once a major hospital enters an exclusive arrangement with an anesthesiologist, rival anesthesiologists will be denied the right to practice at that hospital (e.g., Jefferson Parish, 1984), or once Microsoft bundles its Windows operating system with its Internet Explorer browser, rival browser makers may find it more difficult to sell their product (Microsoft, 2001). Several writers, particularly from the Chicago School, have been highly skeptical of foreclosure claims. First, tying may require no more than a realignment of purchasing patterns. For example, once an independent anesthesiologist loses the right to practice anesthesiology at a particular hospital because of its tie, she will have to practice at a different hospital, but market competition need not be affected. Second, the critique of the leverage theory also applies in foreclosure cases: that is, even assuming that tying forecloses rivals, it still does not change the fact that the profit-maximizing price of the tying-tied combination is no higher than it was before. See Bork (1978, 231–45). In all events, foreclosure requires an assessment of the downstream market, something that cannot be inferred from upstream market share. For example, even if the patent monopolist of a salt-injecting machine for canned foods has a 90% market share in such machines, foreclosure in the salt market must be assessed against the full range of uses for salt (e.g., Int’l Salt, 1947). Use through the salt-injecting machine may represent only a small proportion. Many ties, particularly in the franchise and aftermarket parts

338   Erik Hovenkamp and Herbert Hovenkamp contexts, have involved relatively common commodities with a wide variety of uses unrelated to the tying product. Today the rise of transaction cost economics and greater sensitivity toward the cost of moving resources has forced relaxation of the strict Chicago School view. See Hovenkamp (2010). Even if tying does not lead to an immediate increase in prices, it can serve to deter or delay entry by rivals. Before this can occur, however, actual exclusion must be shown and entry or mobility barriers must be shown to be significant. For example, the OS/browser tie in Microsoft very likely delayed the development of independent browsers with OS capabilities. These barriers to entry or mobility can emanate from a variety of sources, including intellectual property rights or other government-created licensing restrictions, differential economies of scale in the markets for the tying and tied products, or reasons related to asset specificity and risk that are commonly associated with barriers to entry. For example, once excluded from practicing his profession by the dominant hospital’s anesthesiology tie, an independent anesthesiologist may be prevented from building his own hospital by government-imposed regulatory requirements, different scale economies as between hospitals and anesthesiological practice, or else the simple fact that a hospital is a costly specialized asset that cannot readily be redeployed in the event of failure. Of course, none of these explain why the hospital would be able to charge higher prices when the independent anesthesiologist had been excluded from the anesthesiologist market. Even a monopoly hospital would generally profit by procuring inputs of the highest quality and at the lowest cost, and once its profit-maximizing price is established it will not be able to extract more profits simply by tying one of these inputs. More realistic foreclosure fears arise in changing markets where single-level entry threatens to turn into two-level competition, thus eroding the dominant firm’s position. In Microsoft (2001) the D.C. Circuit found this to be such a case. Because the Netscape browser contained operating system features, Microsoft CEO Bill Gates feared that it could eventually “commoditize” the operating system market by creating competition there. The Internet Explorer tie was designed to switch customers away from Netscape in order to suppress that threat.4 In other situations tying may deprive rivals in the tied market of output that is necessary for efficient production, with the effect of raising prices in the secondary market. Finally, the use of dominant firm tying to create a two-level monopoly can raise entry barriers by requiring entry at two levels rather than one. This can be particularly burdensome when differential scale levels apply at the two levels. See Areeda and Hovenkamp (2011, ¶1705). 4  See 84 F.Supp.2d 9, 29 (D.D.C. 1999) (district court findings of fact noting Gates’s email to employees about Netscape threat “to commoditize the underlying operating system”). See Areeda and Hovenkamp (2011, ¶¶1704, 1706). The Ninth Circuit has held that tying claims alleged under the rule of reason should be dismissed in the absence of any allegation of foreclosure. Brantley v. NBC Universal, Inc., 675 F.3d 1192 (9th Cir. 2012).

Tying Arrangements   339

14.5.  Tying of Complements and Double Marginalization Horizontal double marginalization occurs when two or more complementary products are sold by separate firms and each one individually maximizes at prices above marginal cost. As a result, purchasing under separate provision also entails a separate markup on each good. Such arrangements frequently result in prices whose sum exceeds the price that would be set by a single firm who sells the products together. By eliminating such “double marginalization,” tying can benefit consumers. Further, total profits will tend to be higher under tying, as a single firm will internalize the (negative) pricing externality of each good. This provides a clear justification for tying that is independent of any production efficiency gains that might result from joint provision. Two firms can achieve similar results by coordinating their output and pricing, but this arrangement will necessarily involve tying in any event, as coordination will require joint sales. The elimination of double marginalization is always sufficient to justify tying when the goods in question are mutually perfect complements, meaning that each good is valueless without the other. For example, an automobile consists of many individual parts that are essential to the final product, but which are independently useless to most consumers. If consumers purchased these parts from separate firms, then each part would be marked up independently. The total markup would be considerably higher than the single markup by a manufacturer who “ties” all of the parts into a finished car. With joint provision, the firm does not distinguish between the profits earned on separate parts; its only concern is the total profit earned on the car. The tying of mutually perfect complements virtually always increases both seller profits and consumer welfare. These results are formalized in the following simple example. Consider two mutually perfect complements, A and B. Each good i is produced at constant marginal cost ci ≥ 0. Market demand for the pair (one unit of each good) is Q (P ) = α − βP , where P denotes the price of the pair and α, β >0. First consider a single firm who ties the two goods, but does not achieve any production efficiencies (i.e. the cost of producing the tie is c A + c B ). The firm sets a tie price PT to maximize profits, which are given by [PT − c A − c B ]Q (PT ) Solving the firm’s optimization problem yields an optimal tie price of



PT* =

α + β (c A + c B ) 2β



Now consider a situation in which A and B are produced by separate firms. Each firm i ∈{A, B} sets the price of i, denoted Pi , taking its rival’s price as given. Thus, given PB ,

340   Erik Hovenkamp and Herbert Hovenkamp firm A chooses PA to maximize [PA − c A ]Q (PA + PB ). Solving for the Nash equilibrium yields prices of



PA* =

α + β (2c A − c B ) 3β

,

PB* =

α + β (2c B − c A ) 3β



Comparing the prices of the A − B pair between the two scenarios, we find that PT* ≤ (PA* + PB* ) if and only if α ≥ β (c A + c B ), which is true if and only if PT* ≥ (c A + c B ). Thus, whenever a firm would actually choose to tie, it must be the case that tying leads to a lower price for the pair. This implies that consumer welfare increases. Also, it follows from the fact that PT* ≠ (PA* + PB* ) in general that total profits are higher under tying. Thus the tying of mutually perfect complements unambiguously increases total welfare. If a tie includes an imperfect complement—meaning that at least one good has value independently of the others—then the elimination of double marginalization may not be sufficient to infer a welfare improvement. However, this is not because double marginalization ceases to be inefficient in these cases. It is still true that tying reduces the price paid for the tied-up package of goods, and that total profits increase, benefiting both the seller and those purchasers who actually want the entire package. The difference in this case is that some consumers may wish to buy only the tying product alone—something that does not happen when the goods are mutually perfect complements. Tying may injure these consumers, as they may be forced to buy unwanted items in order to obtain the products they desire. The extent of this harm will tend to decrease as the level of complementarity among products increases. For sellers, situations of this type are more readily addressed by bundled discounts rather than tying. In a bundled discount those customers who want only one component pay the maximizing price of that item; but sellers who prefer the bundle can obtain a price that is lower than the summed prices of sales by two different firms. For this reason one might expect to see bundled discounts as an alternative to tying in situations where the bundled goods are often but not invariably used together. The price to the stand-alone buyers will depend on the demand elasticity the seller faces. In some cases the buyers who prefer the primary product alone may value it more highly than the bundle buyers, and in some cases less. As a result competitive harm cannot be inferred from the fact that the seller increases the price of the primary product when it introduces a bundled discount. The above double-marginalization arguments may not apply straightforwardly to ties involving products used in variable proportions, such as printers and ink cartridges. Joint maximization may still occur at a lower price than when sales are made by separate sellers. As developed previously, in variable proportion situations the seller is typically engaged in price discrimination by reducing the price for the fixed component (e.g., the printer) and increasing the price of the variable component (the ink). As a result, it is no longer the case that tying benefits all consumers who wish to buy all of the tie’s components. Rather, tying reduces the total price to consumers who desire

Tying Arrangements   341

relatively few variable units, while increasing the total price for higher intensity consumers. As a result, the double-marginalization effects of variable proportion ties are more complex.

14.6.  Economies of Joint Provision 14.6.1. Generally Economies are always relevant to tying law, although how they are used depends on one’s definition of the “separate products” test. Under one version of the test two goods will be treated as a single product if tied provision is cheaper than provision by two different firms. In that case the plaintiff would have to show separate products by proving the absence of economies. Under the dominant test that five justices of the Supreme Court approved in Jefferson Parish, however, proof of economies does not enter at the definitional stage but rather operates as a defense to the tie. This means that the burden of proving economies normally lies with the defendant. Virtually any type of cost savings can justify a tie, including manufacturing economies, distribution economies, economies in R & D, and purely transactional economies such as risk reduction or quality control. An example of the first is the Times-Picayune case (1953), a government challenge to a newspaper’s requirement that classified advertisements be placed simultaneously in its morning and evening editions. As the record in that case developed, the advertisements were typeset manually and under the tie could be set a single time for both newspapers. This cost savings could not be achieved by any mechanism other than requiring identical morning and evening advertising sections. Many of the economies produced by ties are economies of scope, or the reduced costs that attend producing two goods or services together rather than separately. The phrase “economies of joint provision” is useful, because not all economies of scope justify tying. For example, significant economies of scope justify producing beef and cowhide together, or lumber and sawdust. But once a firm has produced these two goods together, they can and typically are distributed separately. By contrast, the typesetting in the Times-Picayune advertising case could not produce the economies in question unless buyers were required to purchase advertising in both newspapers together. The joint provision of distinct products will frequently result in efficiency gains that make tying beneficial for both firms and consumers. These efficiencies typically involve reduced costs or improved product quality. Production cost reductions occur when the production of one good allows another to be produced more cheaply, or when there is a common fixed cost that can be spread across multiple production processes. Improvements in product quality typically arise when joint production allows for increased compatibility between the goods, making their combined use easier or more enjoyable.

342   Erik Hovenkamp and Herbert Hovenkamp

14.6.2.  Production Cost Efficiencies When joint production of two or more different goods is less costly than separate production of each good, we say that joint production exhibits economies of scope. Formally, consider two products and A and B , let C(q A , qB ) denote the total cost of jointly producing q A units of A and qB units of B , where C(∙,∙) is nonnegative-valued and increasing in both situations. Then joint production of A and B exhibits economies of scope at quantities (q A , qB ) if C(q A , qB ) < C(q A , 0) + C(0, qB ). This may arise because enabling production of one good reduces the marginal cost of producing another, or because total fixed costs are lower under joint production. Perhaps the most common source of scope economies are joint cost savings, which arise when there is a shared input or production process that can be “spread” across the production of different products. Suppose that θ is a costly input used in the production of both A and B, and let θ*(qA,qB) be the quantity of θ required to jointly produce qA units of A and qB units of B. In general, θ may be a fixed or variable input, and it can be either tangible (e.g., a plant or machine) or intangible (e.g., a patent). Then the joint production of A and B exhibits joint cost savings at (qA,qB) if θ*(qA,qB) < θ*(qA,0) + θ*(0,qB). For example, Internet and cable television can be delivered by the same wires, making the cost of joint provision much lower than if two companies offered the services separately and each one supplied its own wires. The classic production assembly line is another example. Adding components as a good travels down a single line may be much cheaper than subjecting the good to repeated trips down different lines. In many of these cases a tie is the only way in which a firm can capitalize on joint cost savings. That is, the savings achieved by tying could not be achieved by separate sales even if both goods are produced by a single firm. For example, consider a pharmaceutical firm that can produce capsules of a cough suppressant and a decongestant. See Evans and Salinger (2007). The per-unit cost of each medication is 10 cents; the cost of buying and filling a capsule is 50 cents. Under a tying arrangement for a multisymptom capsule the firm fills each capsule with both drugs, for a total cost of 70 cents per unit. Under separate production of single-symptom capsules, however, each drug is encapsulated separately, and the total cost of the drug combination is $1.20. Thus the production process exhibits joint cost savings, which occur because encapsulation is a shared cost. Moreover, these savings would not arise if a single firm sold the drugs separately unless they could form a production joint venture and make the multisymptom capsules.

14.6.3.  Quality Control Efficiencies Tying frequently produces quality improvements when achieving compatibility between products is costly and highly sensitive to production specifications. This occurs when the compatibility between products is (1) complex and subject to considerable variation; and (2) essential to the value of one or more products. In these

Tying Arrangements   343

situations, the direct effect of joint production is to reduce the costliness of ensuring compatibility; quality improvements arise indirectly in response to this cost shift. For example, computer operating systems typically include a number of independent programs or subroutines that are developed by the same manufacturer. Clearly the value of each tied program is highly dependent on its compatibility with the user’s operating system. By jointly producing these programs, the producer can capitalize on its familiarity with its own operating system. This is particularly important when product compatibility depends on technological specifications that are not fully observable by a producer’s rivals. A great many antitrust challenges to ties have arisen in the franchising industry, particularly for fast foods. Monopoly in these cases is almost never in prospect in either market. The tying markets for the franchised brands or systems are nearly always competitive, and the tied products are typically common commodities such as spices, cookies, pizza dough, or general purpose real estate rentals. As a result, anticompetitive foreclosure is virtually out of the question. The plaintiffs are almost invariably franchisees rather than rivals, and the claims are for overcharges in the tied product. One robust explanation for these ties is price discrimination, as outlined previously. In addition, however, franchise ties are important for quality control, as an alternative to self-distribution. Some franchisors such as Baskin-Robbins make their own ice cream and have chosen to sell it through independently owned franchisees. Others, such as McDonalds, produce very little product themselves. In both cases, however, the franchisor has a strong interest in maintaining high quality and uniformity across all of its stores, while individual franchisees may have incentives to cut costs by buying cheaper products, free riding on the strength of the franchise as a whole. This is particularly likely when the franchise as a whole depends on repeat business but individual franchisees serve transient customers that they are likely to see only once. See Blair and Lafontaine (2005); Klein and Saft (1985). The Collins v. Dairy Queen case (1999) involved DQ’s insistence that franchisees use Nabisco cookies in its “Blizzard” ice cream concoctions rather than cheaper cookies made by generic bakeries. In Queen City (1997) the Domino’s pizza franchisor required its franchised stores to use its particular pizza dough. Either price discrimination or quality control, or perhaps both, are likely explanations for these ties. Monopolies of either cookies or pizza dough was not in prospect.

14.6.4.  Package Licensing and Similar Intellectual Property Bundles Package licensing of patents creates significant opportunities for cost savings, particularly if the patents are complements. The transaction costs of negotiating license fees over groups of patents, of offering standardized packages, or of monitoring to ensure that licensees are using only technology that they have actually licensed can all be reduced by package licensing. See Bohannan and Hovenkamp (2012, 345–48). Relatedly,

344   Erik Hovenkamp and Herbert Hovenkamp “blanket” licensing of recorded music can enable radio stations and other broadcasters to achieve instant indemnified access to performance rights without individual negotiation. The value of such licenses would largely be lost, however, if individual licensees could willy-nilly designate small portions of them and obtain a pro rata price reduction. Thus in Broadcast Music (1982) the court rejected a tying challenge to a jukebox blanket licensing provision by a bar that wanted to play only country and western music at a reduced license price. The court found that determining which songs were included in the smaller license and monitoring to ensure that the bar did not cheat would actually cost more than providing the full blanket license.

14.7.  Self-Distribution Alternative to Vertical Integration; Risk-Sharing A great many tying claims arise in the context of franchise and dealership networks when the manufacturer forces a dealer or franchisee to take certain products for resale. An automobile manufacturer may require dealers to sell slower moving lines. A fast-food franchisee may require individual retailers to sell specific products supplied by the franchisor. In these cases the independent dealership or franchise arrangement is a substitute for the manufacturer’s own vertical integration into retailing. That is, instead of selling its own vehicles through wholly owned dealerships the manufacturer enters long-term arrangements with locally owned independent dealers. In these cases the arrangement plus the tying requirement acts as a substitute for self-distribution. Such arrangements also operate as a form of risk-sharing, converting the dealer from an employee to an entrepreneur with a significant investment in the local business. Or to say this somewhat differently, the contractual dealership arrangement plus the tie is nothing more than a substitute for a single firm. That observation does not necessarily end all legal inquiry, because the supplier may still be in a position to impose harmful consequences on its dealers, just as a harsh employer can impose such consequences on its employees. It very largely ends the antitrust inquiry, however, because it is difficult to see how the franchise arrangement can be more harmful to competition than outright ownership of retail outlets. Variable proportion tying may benefit businesses or consumers who are uncertain of the value of a particular good. Specifically, a consumer may worry that she will use the tying product too infrequently to justify its purchase. If the consumer is risk averse, then this risk may deter her from purchasing the tying product even if her expected surplus is positive. By reducing the price of the tying product, tying helps to mitigate this risk by increasing the surplus she receives in the event that her future demand for the tied product is low. Economically speaking, the tie has the effect of reducing the purchaser’s fixed cost investment (e.g., the durable and costly printer) and increasing variable costs (e.g., ink). By contrast, if her future demand turns out to be high, then she may achieve less

Tying Arrangements   345

surplus under tying. Thus tying serves to reduce the “spread” between the consumer’s possible surplus realizations. A risk-averse agent may prefer this trade-off. We can illustrate this effect using figure 14.1. Suppose that a consumer is unsure about her demand for a product such as a printer that uses ink cartridges. If she buys the tying product and turns out to be a low-intensity user, then she could receive negative surplus under separate provision (because of the higher printer price) and positive surplus under tying. She could have garnered zero surplus by not buying anything, and thus she regrets buying the tying product under separate provision, but not under tying. If she turns out to be a high-intensity user, then she receives more surplus under separate provision, but even her tying surplus is positive. That is, she does not regret purchasing the tying product in either case. If the agent is risk averse, then avoiding a regrettable situation is valuable in its own right, and so she may prefer tying even if her expected surplus is higher under separate provision. This illustrates a way in which variable proportion tying may improve consumer welfare that is independent of price discrimination effects.

14.8.  Tying in High-Technology Markets; “Misuse” Technology-rich markets are particularly prone to tying because networking and interconnection are frequently major components in such markets. See Bohannan and Hovenkamp (2012, ch. 2); Liebowitz, Stan, and Margolis (2007). As products become more complex, quality control issues loom larger. Further, as products are more specialized, aftermarket goods must be individually tailored as well. At the same time, however, technology can provide opportunities for ties that are anticompetitive because they limit competitor entry or expansion, or in some cases because they restrain innovation. The earliest tying cases in the federal courts were not antitrust challenges. Rather, they were patent infringement cases brought by sellers in which the defendant defended by arguing that the tying requirement was not enforceable under the Patent Act because the tie was an improper attempt to expand the scope of the patent “monopoly.” The Supreme Court rejected this defense in the Henry case (1912), but fairly consistently accepted it in the Motion Picture Patents (1917) and later decisions. The antitrust law of tying actually evolved out of these patent “misuse” cases, where the law was entirely judge-made with no clear statutory authorization. See Bohannan and Hovenkamp (2012, ch. 10). The full-blown modern doctrine of patent misuse did not emerge until the 1940s, principally in the Morton Salt case (1942). In that case patentee Suppiger owned a patented machine that injected salt into canned foods. Its license agreement required users of the machine to purchase their salt from Suppiger. When Suppiger brought a patent infringement suit against Morton for selling an infringing

346   Erik Hovenkamp and Herbert Hovenkamp machine, Morton raised the tie as a defense. The Supreme Court held that Suppiger lost the right to enforce its patent as long as it was tying, even though Morton was not injured by the tie. This loss of enforcement right should persist, the Court held, until the misuse was “purged”—that is, until Suppiger should stop assessing or enforcing the tying requirement. The decision did not rest on any finding of market power or exclusionary effects, but only on the proposition that the tying of unpatented staples improperly extended the patent’s boundaries. The Fourth Circuit extended this rule to copyrights in its Lasercomb decision (1990). As the doctrine of misuse evolved it became narrower than antitrust liability in some ways but broader in others. It is narrower in the very important sense that it operates almost exclusively as a defense to a patent or copyright infringement suit (e.g., Lasercomb, 1990).5 There is no affirmative cause of action for misuse, and thus no treble damages or attorneys’ fees such as successful antitrust plaintiffs can recover. Misuse is also broader, however, in that it has been held to reach substantively to practices that would not be antitrust violations, does not have a market power requirement except in the case of patent tying arrangements,6 and can provide the very draconian remedy of making a misused patent unenforceable against everyone, including people who are not harmed by the misuse. The issue of misuse law’s substantive coverage has proven quite problematic, and several decisions have tried to restrict its reach by limiting it to situations that would actually violate the antitrust laws (e.g., USM).7 Other decisions have found that misuse reaches further, although the modern trend is to apply misuse doctrine restrictively (e.g., Princo, 2010; see Bohannan and Hovenkamp 2011). Misuse is entirely a creature of IP policy, not antitrust law. While intellectual property law shares with antitrust a concern for maintaining competition, its principal concerns are with innovation and protection of the public domain, or the realm of ideas and expressions that are not covered by IP laws and for which access by future innovators is so essential. As a result, a good case can be made that the substantive coverage of misuse law should include antitrust violations but also unreasonable restraints on innovation or denials of access to the public domain that antitrust law does not reach. However, the ordinary remedy should be limited to an injunction against the misuse. See Bohannan and Hovenkamp (2012, ch. 10).

5  One exception to the rule that misuse is asserted as a defense to an IP infringement claim is Brulotte (1964), where the patentee brought a state court breach-o- contract action when the purchaser stopped paying royalties on use of a machine after its patents expired; the Supreme Court permitted a misuse defense. 6  Since 1988 the Patent Act has provided that a patent tie cannot be misuse unless the patentee has market power in the tying product. 35 U.S.C. §271(d)(5). 7  But see Assessment Techs., 2003 (Judge Posner, suggesting that copyright misuse might reach beyond antitrust).

Tying Arrangements   347

14.9.  Remedies, Mainly Damages Both the government and private plaintiffs are entitled to obtain an injunction against a tie that has been found unlawful. The more interesting question concerns antitrust treble damages, which are required by Section 4 of the Clayton Act to be based on the injury that the plaintiff sustained. Some tying plaintiffs are rivals in the tied product market who have been foreclosed by the tie. This was true, for example, in the Jefferson Parish case. In those cases damages for successful plaintiffs are based on lost profits or loss of business value, as in most antitrust cases with competitor plaintiffs (e.g., Moore, 1982). When the plaintiffs are consumers the damages inquiry is particularly complicated by the fact that many consumer ties are brought as class actions, and the analysis above makes clear that a tying arrangement can affect consumers in very different ways. For example, some purchasers of a variable proportion tie in the franchise or aftermarket product setting may be injured by the tie while others are benefited. Historically consumer damages for tying were based on overcharges on the tied product alone (e.g., Siegel, 1971).8 That number is clearly excessive, because ties almost always involve an increase in the tied product price accompanied by a reduction in the tying product price, sometimes all the way to zero. As figure 14.1 illustrates, lower intensity customers in such cases are benefited from the tie, sometimes a very great deal. As a result several courts have adopted an alternative suggested in Areeda and Hovenkamp (2007, ¶340c) that damages be based on the net overcharge (e.g., Kypta, 1982). For example, if tying results in a $3 price cut in the tying product and a $5 price increase in the volume of tied product that is purchased, the correct measure of damages should be $2, not $5. Tying may increase a buyer’s surplus, however, even if the cost of her original bundle (the bundle she purchases under separate provision) increases. We can see this in figure 14.2, which shows the effect of a tie that increases the tied product’s price from P to P + a. The quantities Q(P) and Q(P + a) denote the agent’s demand for tied units under separate provision and tying, respectively. Assuming tying reduces the price of the tying product by some amount b > 0, tying increases the cost of the agent’s original bundle by A + B + C − b (this will be negative for some consumers). However, the agent’s surplus falls by only A + B − b, because region C was not achieved as surplus under separate provision. This implies that the agent’s surplus falls by less than the increase in the cost of her original bundle. For example, if the cost of the agent’s optimal bundle is unchanged by tying, then she receives more surplus under tying. Similarly, it is possible that the cost of her original bundle increases slightly, and yet she achieves more surplus under tying. As such, the change in the cost of an agent’s original bundle is not a strong indication of 8  The court directed, however, that an offset should be provided for the cost of a reasonable franchise fee.

348   Erik Hovenkamp and Herbert Hovenkamp P

Marginal Utility

P+a C

A

B

P

Q(P+a) FIGURE  14.2  Relation

Q(P)

Q

Between Bundled Price and Consumer Harm

how her surplus is affected by tying. Damages measured by the net overcharge on the tying and tied product are therefore likely to be excessive as well. The plaintiff who claims that it was injured, not by an overcharge, but rather by the requirement that it purchase a product that it did not want at all, has not suffered antitrust harm. Although such a buyer’s welfare may be lower, that injury does not result from a reduction of competition unless an alternative product was actually foreclosed from the market.

14.10. Conclusion In the vast majority of cases tying arrangements increase welfare, whether measured under a general welfare or a consumer welfare test. Competitive harm is a threat in a very few situations involving actual market foreclosure or the use of ties to enable dominant firms to retain their market position as one technology rolls into the next. As a result, the so-called per se rule for tying is wrongheaded, and ties should be addressed under the rule of reason, with fairly substantial proof requirements on challengers.

Tying Arrangements   349

References Areeda, Phillip E., and Herbert Hovenkamp. 2007. Antitrust Law. New York: Aspen Publishers. Areeda, Phillip E., and Herbert Hovenkamp. 2011. Antitrust Law. 4th ed. New York: Aspen Publishers. Blair, Roger D., and Francine Lafontaine. 2005. The Economics of Franchising. New York: Cambridge University Press. Bohannan, Christina, and Herbert Hovenkamp. 2011. Concerted Refusals to License Intellectual Property Rights. Harvard Business Law Review Online 1:21–24. Available at http://papers.​ ssrn.com/sol3/papers.cfm?abstract_id=1710531. Bohannan, Christina, and Herbert Hovenkamp. 2012. Creation without Restraint: Promoting Liberty and Rivalry in Innovation. New York: Oxford University Press. Bork, Robert H. 1978. The Antitrust Paradox: A Policy at War with Itself. New York: Basic Books. Bowman, W.  S., Jr. 1957. Tying Arrangements and the Leverage Problem. Yale Law Journal 67:19–36. Evans, David S., and Michael Salinger. 2005. Why Do Firms Bundle and Tie? Evidence from Competitive Markets and Implications for Tying Law. Yale Journal on Regulation 22:37–89. Evans, David S., and Michael A. Salinger. 2007. Recent Developments in Antitrust: Theory and Evidence. Cambridge, MA: MIT Press. Hovenkamp, Erik, and Herbert Hovenkamp. 2008. Exclusionary Bundled Discounts and the Antitrust Modernization Commission. Antitrust Bulletin 53:517–53. Hovenkamp, Erik, and Herbert Hovenkamp. 2009. Complex Bundled Discounts and Antitrust Policy. Buffalo Law Review 57:1227–66. Hovenkamp, Herbert. 2010. Harvard, Chicago and Transaction Cost Economics in Antitrust Analysis. Antitrust Bulletin 55:613–62. Hovenkamp, Herbert. 2011. Federal Antitrust Policy: The Law of Competition and Its Practices. St. Paul, MN: West. Klein, Benjamin, and Lester F. Saft. 1985. The Law and Economics of Franchise Tying Contracts. Journal of Law and Economics 28:345–80. Liebowitz, Stan J., and Stephen E. Margolis. 2007. Bundles of Joy: The Ubiquity and Efficiency of Bundles in New Technology Markets. Journal of Competition Law and Economics 5:1–47. Schmalensee, Richard. 1981. Monopolistic Two Part Pricing Arrangements. Bell Journal of Economics 12:445–466. Whinston, Michael D. 1990. Tying, Foreclosure, and Exclusion. American Economic Review 80:837–59.

Cases Assessment Techs., LLC v. WIREdata, Inc., 350 F.3d 640, 647 (7th Cir. 2003). Automatic Radio Mfg. Co. v. Hazeltine Research, Inc., 339 U.S. 827 (1950). Berkey Photo, Inc. v. Eastman Kodak Co., 603 F.2d 263 (2d Cir. 1979).

350   Erik Hovenkamp and Herbert Hovenkamp Brantley v. NBC Universal, Inc., 675 F.3d 1192 (9th Cir. 2012). Broadcast Music, Inc. v. Moor-Law, Inc., 527 F.Supp. 758 (D.Del. 1981), aff ’d mem., 691 F.2d 490 (3d Cir. 1982). Brulotte v. Thys Co., 379 U.S. 29 (1964). Carbice Corp. v. Am. Patents Dev. Corp., 283 U.S. 27, 27–30 (1931). Collins v. International Dairy Queen, Inc., 980 F. Supp. 1252 (M.D. Ga. 1997); 59 F.Supp.2d 1312 (M.D.Ga. 1999). Eastman Kodak Co. v. Image Technical Servs., 504 U.S. 451, 460–62 (1992). Hack v. President and Fellows of Yale College, 237 F.3d 81 (2d Cir. 2000), cert. denied, 534 U.S. 588 (2001). Henry v. A.B.Dick Co., 224 U.S. 1 (1912). Illinois Tool Works, Inc. v. Independent Ink, Inc., 547 U.S. 28, 42 (2006). International Salt Co. v. United States, 332 U.S. 392 (1947). Jack Walters & Sons Corp. v. Morton Bldg., Inc., 737 F.2d 698, 703 (7th Cir.1984). Jefferson Parish Hosp. Dist. No. 2 v. Hyde, 466 U.S. 2, 19 (1984). Kypta v. McDonald’s Corp., 671 F.2d 1282, 1285 (11th Cir.), cert. denied, 459 U.S. 857 (1982). Lasercomb Am., Inc. v. Reynolds, 911 F.2d 970, 973 (4th Cir. 1990). Moore v. James H. Matthews, Inc., 682 F.2d 830, 836–37 (9th Cir.1982). Morton Salt Co. v. G.S. Suppiger Co., 314 U.S. 488, 62 S.Ct. 402 (1942). Motion Picture Patents Co. v. Universal Film Mfg. Co., 243 U.S. 502 (1917). Northern Pacific Rwy. v. United States, 356 U.S. 1, 6 (1958). Princo Corp. v Intl. Trade Com’n., 616 F.3d 1318 (Fed. Cir. 2010). Queen City Pizza, Inc. v. Domino’s Pizza, Inc., 124 F.3d 430 (3d Cir. 1997). Reisner v. GM Corp., 511 F.Supp. 1167 (S.D.N.Y. 1981), aff ’d 671 F.2d 91 (2d Cir.), cert. denied, 459 U.S. 858 (1982). Siegel v. Chicken Delight, Inc., 448 F.2d 43, 47 (9th Cir. 1971), cert. denied, 405 U.S. 955 (1972). Standard Oil Co. of Calif. v. United States, 337 U.S. 293, 305–6 (1949). Times–Picayune Pub. Co. v. United States, 345 U.S. 594 (1953). United States v. Loew’s Co., 371 U.S. 38, 45 (1962). United States v. Microsoft Corp., 253 F.3d 34, 65–67 (D.C. Cir.), cert. denied, 534 U.S. 952 (2001). United States v. Paramount Pictures, Inc., 334 U.S. 131, 158 (1948). USM Corp. v. SPS Techs., Inc., 694 F.2d 505 (7th Cir. 1982).

CHAPTER 15

V E RT IC A L R E S T R A I N T S AC R O S S J U R I SD IC T ION S EDWARD M. IACOBUCCI AND RALPH A. WINTER

15.1. Introduction In a very simple economy, sellers would set prices and buyers would chooses quantities. A manufacturer would sell to any distributor or retailer wanting to purchase its product. No restrictions would be placed on distributor pricing, territories, the other products that could be offered by distributors, or on customers to whom the distributor could sell. The entire property rights to a product would be transferred with the exchange of units of a product and it would be up to the distributor of the product to decide on how and where to resell the product. In reality, contracts struck by firms along a supply chain, from the providers of raw materials down to retailers selling to consumers, are more complex. Prices remain the principal means by which incentives are aligned along a supply chain, but we also observe more complex contract terms. The following types of payment terms are among those adopted in real-world contracts: • General nonlinear pricing schedules, including quantity discounts, block pricing, two-part pricing, minimum quantity contracts, and take-or-pay contracts • Royalty contracts, in which payment depends on quantity of a product resold by a downstream retailer, not the quantity purchased by the retailer • Loyalty contracts, in which (in one form) the buyer’s payment for an input depends on the proportion of the buyer’s needs that are met by the input

352   Edward M. Iacobucci and Ralph A. Winter • Slotting allowances, which are a fixed payment by a manufacturer for the right to be represented in a retail outlet and which may include the rights to a specified amount of shelf space or floor space • Buyback options under which the retailer can return unsold inventory • Consignment selling arrangements We also see contracts that impose restraints on buyers’ actions: • Price floors or ceilings imposed on retailers of a product (resale price maintenance) • Territorial restrictions on where a dealer may sell, where a dealer may actively sell, or where a dealer may locate. Territorial restrictions may also be placed on the upstream firm, such as a franchisor, in the location of future outlets • Exclusivity clauses, either constraining the downstream buyer not to purchase from other suppliers or constraining the supplier not to deal with other outlets • Tying restrictions, either in the form of bundling products for which separate markets could be or are established, or requirements tying, which stipulates that the buyer must purchase all of its requirements of an input B from the seller if the buyer is to purchase input A from the seller In this chapter, we offer an overview of the law and economics of the most important restraints—resale price maintenance, territorial restrictions, exclusivity clauses, and tying—within a unified framework. Our approach is applicable more broadly to the entire range of contracts. Vertical restraints are subject to considerably divergent antitrust policy across jurisdictions. Differences in antitrust policy toward vertical restraints have become even sharper recently, especially between the United States and the EU. These differences reflect in some measure a variation in the influence of economics on antitrust law. The US policy towards vertical restraints is now among the most liberal among developed nations, largely through the influence of the “Chicago school” of antitrust economics. Chicago scholars have emphasized that vertical restraints imposed by a manufacturer on prices, sales territories, and exclusivity and other decisions can often be explained as means of efficiently responding to the failure of simple price contracts alone to coordinate decisions along a supply chain. In other situations, vertical restraints, including restraints on final buyers, are means of price discrimination. Neither case yields a presumption of benefits from antitrust restrictions on the set of vertical contractual strategies available to a firm. Anticompetitive incentives cannot simply be assumed to explain vertical restraints even where, as is typical, these contracts suppress competition among dealers or distributors of a product or products sold by a manufacturer. While US antitrust policy towards vertical restraints has been influenced most by economic thinking, it would be a mistake to regard economists as having one voice in this area. Some economists emphasize the efficiency theories of vertical restraints, others the possibility that vertical restraints can dampen price competition or deter entry. Even among US policymakers there is a divergence in views. In September

Vertical Restraints across Jurisdictions   353

2008, the US Department of Justice released a 200-page report on single-firm conduct (U.S. Department of Justice 2008), largely dealing with vertical restraints. Three of the four sitting commissioners at the Federal Trade Commission objected to the report on the grounds that it offered insufficient protection to consumers against abuse of market dominance.1 President Obama’s new antitrust appointee then withdrew the report. Our focus in analyzing competition policy towards vertical restraints is on US and European law, not just because these are the two most important jurisdictions in the size of economic activity affected but because the differences in the law between the two areas provide us with points of strongest contrast for analyzing the law against the foundation of the underlying economics. In the next section of this chapter we provide a synthesis of the economic theories of vertical restraints, both efficiency theories and anticompetitive theories, and the types of evidence that have been brought to support the theories. Section 3 outlines the law on vertical restraints in the United States and the EU. In section 4 we discuss the gaps between the legal restrictions on vertical restraints and economics foundations underlying optimal policy.

15.2.  The Economics of Vertical Restraints As an empirical matter, vertical restraints are common, and arise in many settings where concerns about market power are minimal. Evidence on the frequency of vertical restraints is available primarily for resale price maintenance, the most popular restraint during the times when it has been legal. Vertical price floors have been imposed on retailers of a wide cross-section of products: clothing, skis, and other sports equipment, watches, jewelry, luxury goods of all kinds, candy, beer, bread, floor wax, furniture polish, milk, toilet paper, cereal, canned soup, books, shoes, mattresses, large appliances, and automobiles, to name a few (Overstreet 1983; Ippolito 1988; Ippolito and Overstreet 1996). Products in virtually every category have been subject to resale price maintenance at one time or another, and estimates of the proportion of retail sales that have been subject to resale price maintenance range as high as 25% in the UK and 4% to 10% in the United States (Scherer and Ross 1990, 549). In Canada, before the law prohibiting resale price maintenance was enacted in 1951, an estimated 20% of goods sold through grocery stores and 60% sold through drugstores were “fair-traded” (Overstreet 1983, 153, 155). 1 

Three of the four sitting Federal Trade commissioners called the report a “blueprint for radically weakened enforcement of section 2 of the Sherman Act” (FTC Press Release, September 8, 2008). Originally the report had been planned as a joint publication, and historically the Department of Justice and FTC have worked closely together in developing guidelines.

354   Edward M. Iacobucci and Ralph A. Winter Information is also available on the extent of exclusivity. Of retail sales through independent retailers, more than one-third were found to be subject to some form of exclusive dealing in a recent study (Lafontaine and Slade 2005, citing a US Department of Commerce Study from 1988). Some exclusivity restraints are clearly efficient and would attract no attention from antitrust law. A McDonalds franchisee must sell exclusively McDonalds’ hamburgers as opposed to Wendy’s or Burger King’s products. Any luxury jewelry retailer is restrained against placing cheap watches for sale in its store. Finally, tying, defined broadly, is ubiquitous. Shoelaces are sold with shoes. Gloves and shows are sold in pairs. Cars are sold with tires. Tying presents potential anticompetitive concerns in only a very small percentage of cases where it is observed. The task for competition policy in this area is to distinguish anticompetitive uses of vertical restraints from efficient or benign uses of vertical restraints. We offer below a general perspective on the economics of vertical restraints, then turn to specific theories of why the restraints are adopted.

15.2.1.  General Perspective on the Economics of Vertical Restraints The traditional theory of vertical restraints, or what could be termed the pre-economics theory, is that vertical restraints are imposed on downstream firms by a manufacturer with market power.2 Contractual restraints are not a reflection of joint wealth maximization under this theory and are instead explained by the power of the party to impose the restrictions on innocent downstream distributors.3 From an economist’s perspective, the traditional view is wrong. Aside from price discrimination motives, any contract is struck to align the incentives of individual parties to the contract with the collective interest of the parties to the contract. Under simple assumptions (symmetric information and transferable wealth), a contract must maximize the sum of wealth or expected wealth of the contract parties: if another contract were available with greater total profit, the parties could move to the new contract and share the gain in profits. A vertical restraint is not imposed in order to achieve an increase in profits for one party to the contract at the expense of the other party and total

2  See the discussion of the market power requirement in Jefferson Parish Hospital District No. 2 v. Hyde, 466 U.S. 2 (1984), for example. 3  The traditional, or pre-economic, view of vertical restraints is that the restraints are often imposed by manufacturers on retailers against the retailers’ interest. An expert witness in a Canadian antitrust case articulated the traditional view very clearly: “The dangers from contracting arise when one side of the market has the power to impose contract terms on the other. . . . If one contracting party is a monopolist . . . it can preserve its market power by insisting that its customers (or suppliers) sign long-term contracts,” and “buyers gain nothing from the . . . provisions in the contract [at issue in the case]. Hence, the very fact that nearly all buyers sign such contracts is evidence that Laidlaw has and exercises market power” (Government Expert Report in Director of Investigation and Research v. Laidlaw Waste Systems Ltd. ((1992), 20 C.P.R. (3d) 289 (Comp. Trib.), paras. 21 and 42).

Vertical Restraints across Jurisdictions   355

profits. Rather, it must maximize the combined profit of the contracting parties. In other words, contracts must be Pareto efficient across the contracting parties. This does not mean that all contracts are socially efficient. Information may not be symmetric, and wealth transfers may not be possible. Moreover, parties outside the contract may be harmed. An agreement to establish a cartel or to adopt vertical restraints to facilitate cartel pricing, for example, is Pareto improving for the members of the agreement but not for consumers, who are outside the agreement. A firm that signs exclusive contracts with all upstream suppliers of an essential input can monopolize a market to the detriment of downstream buyers. Inefficiencies, including anticompetitive effects, are the result of externalities on parties outside the contract. An assessment of the competitive impact of any vertical restraints contract must be based on the application of economic theory to the specific facts of the case at hand. As Cooper and coauthors (2005a, 2005b) have emphasized, however, decisions in a case must also be informed by one’s prior probability that a restraint is likely to be anticompetitive, this prior being based on evidence and general principles regarding the impact of vertical restraints. The single most important economic principle underlying contracts, general agreements, and mergers in competition policy is the following. Agreements between producers of substitute products (competitors) tend to be detrimental to social welfare; contracts between producers of complementary products tend to be welfare-enhancing. O’Brien (2008) refers to this proposition, dating back to Cournot ([1838 [1971), as “the fundamental theorem of antitrust.”4 The core example of an agreement between substitutes is a cartel agreement to raise prices. The most important example of complementary products is inputs in the same supply chain. Thus, at a broad level, economists’ prior is that agreements among suppliers of complementary inputs or products are efficient. A merger between producers of complementary products, each with market power, results in lower prices, because the positive externality that each firm had imposed on the other in any decision to lower price is internalized. Buyers, who are parties outside the merger, benefit from the lower price. Given that there are complementarities between suppliers at different stages of the distribution chain, contracts incorporating vertical restraints we can think of a partial merger or coordination of actions, and the same principle should apply. Thus, a purely vertical agreement on prices, for example, lowers prices by eliminating the double-marginalization effect that Spengler (1950) discussed.5 Efficiencies arise from agreements among firms with complementary assets for a set of reasons beyond those identified by Cournot or Spengler. Vertical agreements allow 4  O’Brien expresses the proposition as “Combining substitutes is bad, and combining complements is good, unless demonstrated otherwise” (emphasis in the original). William Baxter, the former assistant attorney general in charge of antitrust enforcement in the United States was a strong proponent of this principle. 5  If an upstream firm has market power, it will sell at a markup to a downstream firm; if the downstream firm has market power, it too will sell at a markup to buyers. This “double marginalization” results in higher prices, and greater distortions, than would be the case if there were only a single firm in the vertical distribution chain.

356   Edward M. Iacobucci and Ralph A. Winter closer coordination of investment and production plans, alignment of incentives, and so on. When a pure monopolist seeks to extend a monopoly market structure into other market levels, such an expansion is generally procompetitive, because there is typically no reason for a monopolist to expand into an upstream or downstream market unless it is more efficient than competitors. In summary, the economists’ prior—although one that has exceptions and can therefore be defeated by evidence in a case—is that agreements among firms operating in a supply chain are not anticompetitive. The exceptions to the principle that vertical restraints are likely to be benign arise where these restraints have horizontal impacts. Consider the following example. A dominant firm in a market faces a new threat of entry. Either product must be distributed through a limited set of distributors (e.g., prominent retailers in a city). In response to the new threat of entry the incumbent firm offers the main distributors a payment not to accept the entrant’s product, that is, to sell the dominant firm’s product exclusively. The firm can share some of its profits earned from its dominant position with the limited number of distribution outlets in order to sustain the position. The set of exclusive contracts profitably excludes the entrant. The contracts are all vertical, between producers of complementary products, yet the impact is anticompetitive. The example motivates the following more specific principle for guiding competition policy in this area. The principle follows from the basic theorem of antitrust, which we suggest fits the mainstream economic thinking: Antitrust scrutiny of vertical restraints should be limited to a focus on adverse effects on competition across supply chains. A firm should have the right to extract as much surplus as it can from its own supply chain, in the absence of horizontal effects across supply chains. In the absence of strategies that diminish competition, a manufacturer’s market power is determined by the extent to which its good is superior to other products, and the profit earned from this level of market power encourages innovation and development of new products. We do not assume in stating this generally accepted normative principle that each and every vertical restraint with a role of coordinating incentives along a supply chain, or with a role of extracting surplus from downstream firms, necessarily raises total surplus. Vertical restraints can be used to adjust the mix of price and nonprice competition (competition in the dimensions of service quality, promotion, sales effort, and so on), as we shall discuss. It is entirely possible, as a matter of economic theory, that the use of vertical restraints to raise service quality or retail promotion at the expense of higher prices, for example, results in lower consumer surplus and total surplus in a market. The argument for laissez-faire is not based on an assertion that market allocations are always optimal. It is based on the recognition that it is impossible or impractical to identify empirically the full set of circumstances under which a decision by a firm to increase quality or promotion reduces total surplus. We do not as a matter of public policy constrain firms’ mix of promotion and advertising decisions when these decisions are implemented directly; nor should we when the mix of price and nonprice decisions is implemented via vertical restraints. This is an extension of a principle applied to price discrimination that is generally accepted by policy economists: price discrimination in and of itself is not a basis for

Vertical Restraints across Jurisdictions   357

intervention. Economic theory does not provide a workable guide to determine when price discrimination does or does not harm total welfare. Firms’ strategies for surplus extraction can be placed along a spectrum from uniform pricing (no discrimination at all) at one end to perfect price discrimination at the other end. Firms in reality attempt to find strategies to move along the spectrum. We know that on average total surplus increases along this spectrum, since under perfect price discrimination total surplus is maximized. Examples tell us that the increase in total surplus along this spectrum is not necessarily monotonic—and therefore it is possible that a particular strategy decreases welfare. But theory offers us little practical guidance as to when this might be the case. Similarly, contracts more complex than uniform pricing are intended, for a profit-maximizing firm in the absence of horizontal anticompetitive effects on prices or exclusion, to either increase total surplus along the supply chain, including final consumers, or to increase the share of the surplus captured by the firm. The former is directly welfare increasing; the latter tends to be surplus increasing “on average.” This simple perspective is not enough to prove that vertical restraints are efficient, but it does present us with a prior expectation in approaching any case. In the absence of clear evidence, contracts cannot be assumed to decrease welfare. In summary, economics supports a laissez-faire policy towards vertical restraints unless there is evidence that the restraint is supporting horizontal anticompetitive effects in the form of higher prices between manufacturers or an exclusionary effect in the case of a dominant firm. The suppression of price competition may be at the retail level in the case of a vertical restraint with the sole purpose of supporting cartel pricing among retailers, to the detriment of the upstream manufacturer. An immediate implication of the principle is the rejection of two common approaches to thinking about vertical restraints policy. First, it is common to suggest a policy that attempts to assess a vertical restraint on the basis of a trade-off between negative effects on intrabrand competition and positive effects on interbrand competition is misguided. Consider the adoption of vertical restraints by a pure monopolist, facing no competitors or even the threat of competitors. A monopolist benefits from greater price competition and lower prices at the retail level, all else equal, because lower prices mean greater demand if demand curves are downward sloping. The observation that a monopolist is adopting vertical restraints therefore means that the vertical restraint must be increasing some nonprice retail activity that enhances demand, helps the monopolist extract surplus, or simply encourages outlets to carry the monopolist’s product. Retailers engage in many demand-enhancing activities, as we discuss below. The observation of a vertical restraint on competition means that the nonprice benefits to the monopolist of greater nonprice activity more than offset any negative impact on profits of higher prices on demand. No basis exists for presuming that the benefits to consumers of the encouraged activity do not offset any detrimental impact of higher prices. The vertical restraint should therefore be allowed in the case of a pure monopolist. Yet the approach of balancing increases in intrabrand competition against a decrease in interbrand competition would lead to the conclusion that the restraint should be prohibited in this case: there is no

358   Edward M. Iacobucci and Ralph A. Winter interbrand competition to promote, so the test of balancing of the positive impact of restraints in promoting interbrand competition against the decrease in intrabrand competition would suggest prohibiting the restraint. This is a conclusion without an economic basis. For this reason, we would reject the approach of the EU Guidelines on Vertical Restraints, which describe the reduction of intrabrand competition as a generally harmful competitive effect that may result from vertical restraints.6 This approach leads the Guidelines astray in a number of specific respects, perhaps most prominently in restricting resale price maintenance (RPM) because the direct impact of the practice is to raise price. This may be an impact of RPM in a particular case, but this says nothing about the effect of RPM on horizontal competition. The EU Guidelines’ principle that a reduction in intrabrand competition is harmful appears to be based on the idea that, other things held constant, an increase in price reduces welfare. This idea itself is well founded. The application of it is not. When a pure monopolist adopts a price floor, for example, other things are not held constant. If price were the only demand-influencing variable affected by the restraint, then profits would fall with the use of the vertical restraint. The monopolist would not adopt the restraint. A second approach to thinking about vertical restraints policy can also be rejected at this point, even before we have discussed specific theories of vertical restraints. The approach reflected in the European Guidelines, and commonly argued even among economists, is that the key piece of evidence in understanding whether a vertical restraint is efficient or anticompetitive is the degree of market power of the manufacturer. In a market with a competitive structure upstream and downstream, it is (correctly) argued that vertical restraints are presumptively procompetitive. But sometimes policymakers go a step further in arguing that a competitive structure is not only sufficient for a presumption of benign or procompetitive use of restraints, it is necessary as well. In other words, the use of restraints by a dominant firm should be presumptively anticompetitive, or at a minimum the firm should have the burden of proof of in demonstrating a procompetitive role for a restraint. The case of a pure, unthreatened monopolist shows this to be wrong; there is no presumption that the monopolist’s choice of vertical restraints to enhance a nonprice activity at the cost of higher prices is against the social interest. (Policy generally does not restrict a firm from raising its prices and spending the additional revenue per unit on product promotion; nor should it regulate the same trade-off when the trade-off is implemented indirectly through restraints.) The extent to which vertical restraints (or business practices in general) should attract antitrust scrutiny is nonmonotonic as a function of the market power of the firm adopting the practices. With a competitive market structure, restraints are presumptively efficient, and at the other extreme the use of vertical restraints by a pure, unthreatened monopolist cannot be presumed anticompetitive. Only when market concentration is high but not at the monopoly level can there be a significant chance a vertical restraint will have potentially anticompetitive consequences.

6 

European Commission, Guidelines on Vertical Restraints, 2010/C 103/01 at para. 100.

Vertical Restraints across Jurisdictions   359

Against the general background offered above, we turn next to analyzing the incentives for vertical restraint by a single firm.

15.2.2.  Efficiency Theories: The Incentive for Vertical Restraints on the Part of a Single Firm, Including Price Restraints Our overview of the economic theories of vertical restraints starts with a simple observation. Retailers do more than post prices. Retailers offer convenience, specifically a low time-cost of purchasing by providing staff, well-organized inventory, clear information, and even short cashier lines. Retailers invest in enthusiastic staff and sales effort in providing accurate and complete point-of-sale information. Retailers choose the level of postsales service of items that may need repair as well as the return policies, not just the specific written return policies but the general sense of either willingness or reluctance that they exhibit in accepting returns. Retail activity is multidimensional, with prices being only one component of the set of decisions. Consumers purchasing a specific product end up with the same physical product wherever they shop, but the surplus gained from purchasing depends upon their entire retail purchase experience. This is especially true for luxury or fashion items where the consumer value is particularly sensitive to product image. Expensive perfume purchased from Walmart or Carrefour is simply not the same product as an identical bottle of chemicals purchased from a luxurious perfume counter in an upmarket outlet with classical music playing in the background. Retailers add value to the final product purchased by consumers, whether by saving consumers’ time on the purchase of a routine item or by investing in enhancement of their customers’ shopping experience. We add another observation. The additional dimensions of retailer input, beyond price, are costly or even impossible to monitor in a contractual relationship with a manufacturer. The enthusiasm with which a retailer sells a pair of skis, the effort that the retailer puts into learning the intricacies of new electronic products, the expense that a retailer puts into maintaining well-organized inventory and displays, and the general investment that the retailer undertakes in improving the purchasing experience cannot be contractually specified—at least not precisely. The manufacturer’s contract with a retailer cannot dictate that the retailer express a level of enthusiasm of 8.43 on a scale of 10. Instead, retailers undertake their decisions with incentives provided by two sources: the contracts with manufacturers on dimensions that can be specified, and competition with other retailers. The efficiency theory of vertical restraints rests on two propositions. First, a simple price contract may leave retailers with inadequate incentives to provide sales effort in its various dimensions. Second, contracts with vertical restraints can restore or at least enhance these incentives. We begin by developing the first proposition, that is, explaining why the price system alone may fail to coordinate incentives along a supply chain by leaving retailers with inadequate incentives to exert the level of sales effort that would be specified in a hypothetical, ideal, complete contract that restricted all of the actions of the retailers. This

360   Edward M. Iacobucci and Ralph A. Winter is the issue of market failure in the Williamsonian (1975) sense. The failure of the price system to coordinate incentives, that is, to maximize the combined wealth of firms along the supply chain, opens up the potential for more complex contracts. We restrict the discussion for the moment to one dimension of sales effort, that is, demand-enhancing activity, and measure investment in this activity in dollars.7 Consider a manufacturer selling through a set of retailers that, for simplicity, adopt the same price and effort decisions in a symmetric retail equilibrium whatever the wholesale price set by the manufacturer. We denote the price and sales effort by p and e, respectively, and the wholesale price by w. The demand in the market, at a symmetric retail market equilibrium is denoted by Q(p, e). Finally, we denote the elasticity of demand with respect to price by εp and the demand elasticity with respect to sales effort by εe. The price and sales effort that would be set in an ideal, complete contract are at the levels, denoted by (p*, e*), that would maximize the combined profit of all parties to the contract, that is, the manufacturer and the retailers.8 In the setting in which a manufacturer has a fixed fee to collect profits from retailers, the wholesale price is freed from its role in collecting profits and is available to use as an instrument to elicit optimal decisions on the part of retailers, as in Mathewson and Winter (1984). The question is whether the single instrument, w, is enough to elicit the decisions (p*, e*). That is, as w is raised from marginal cost to the level w* that elicits p*, will the retailers offer the optimal sales effort, e*, or some effort level below or above e*? And if the optimum is not achieved, that is, there is a Williamsonian market failure, what are the sources of this market failure? As w is raised, p goes up, but (under normal assumptions) the retail margin, (p − w) will fall. The retail margin represents the marginal benefit that a retailer obtains from attracting one more unit of demand through sales effort, so the incentive to provide effort drops as w is raised. In addressing the market failure question, a very useful result is the Dorfman-Steiner theorem (Dorfman and Steiner 1954). Dorfman and Steiner considered a firm, facing demand Q(p, e) that depends on price and sales effort (effort being advertising or product quality, in their discussion). They showed that the firm will optimally devote a proportion of revenue to sales effort that is given by the ratio of the two elasticities of demand:

(Dorfman-Steiner )



7 

e / pQ = ε e / ε p



(15.1)

The discussion in this section is based on Mathewson and Winter (1984) and Winter (1993). In stating that a wholesale contract will maximize combined profits of the parties to the contract, we are assuming that the manufacturer has the ability to set a fixed fee in contracts with retailers. More generally, even if a simple fixed fee is impossible (e.g., because of limited wealth on the part of retail agents), manufacturers benefit from profits at the retail level in a number of ways. Higher retail profits encourage more outlets to carry a product, to the manufacturer’s benefit, and also provide incentives to encourage higher retail quality if this quality is being monitored by the manufacturer (Klein and Murphy 1988). The assumption that a wholesale contract maximizes the combined profit of the contract parties is surely a reasonable approximation. 8 

Vertical Restraints across Jurisdictions   361

In our context, the theorem describes the efficient (collective profit-maximizing) effort decision as determined in the hypothetical complete contract. But we can also use (15.1) to characterize the choice of effort and price by a single retailer, within a simple wholesale price contract, in exactly the same way—substituting the retailer’s elasticities for the market elasticities in the right-hand side of (15.1). Thus the market-failure question reduces to the following: why would the ratio of advertising to price elasticity differ for an individual retailer than for the market as a whole?9 The right-hand side of (15.1) is equal to the marginal rate of substitution between effort and price for the firm making the decision; the market failure question is, then, why the marginal rate of substitution between prices and sales effort differs between an individual retailer and the market as a whole. Our second proposition is that vertical restraint contracts can resolve the market failure or incentive distortions in retailers’ decisions on prices and other demand-enhancing activities. This proposition can also be posed within the Dorfman-Steiner framework before we proceed to addressing the questions. When the ratio εe / εp is smaller for the individual retailer than it is for the market as a whole (evaluated at the first-best p and e), then retailers are “biased” at the margin towards prices that are too low and effort levels that are too low. Retailers rely excessively (from the perspective of private efficiency) on attracting customers through low prices rather than high sales effort. In this case, to start with the most important vertical restraint, two different roles for vertical price floors arise. If the manufacturer maintains a price floor at p* and lowers the wholesale price w, it is increasing the incentive to provide effort (since the marginal benefit of attracting demand, the retail margin, is expanding). It can do so until e* is achieved. The price floor influences directly the incentives to exert sales effort under this mechanism. A second role for price floors arises when the manufacturer can implicitly contract for effort (e.g., maintaining a sufficiently high level of freshness of the product, or adequate servicing) but only at the cost of periodic monitoring of the level of effort. A price floor can act to protect retail profits against erosion from intensive price competition. Under this indirect mechanism, incentives for providing effort are improved because a retailer has more to lose in the event that it is caught shirking on effort (Klein and Murphy 1988).10 9  In a simple model of a symmetric retailer duopoly downstream, one can show that the individual firm elasticity of demand is equal to the sum of the market elasticity of demand and the cross-elasticity of demand. One can therefore pose the question as the following: why would the ratio of cross-elasticities between retailers differ from the ratio of own-elasticities? 10  Both the direct and the indirect mechanism work in the same direction: increasing sales effort at the expense of higher prices. If we extend the setting to one in which effort takes on multiple important dimensions, the single instrument of a price floor will not achieve first-best profits (unless the key elasticity ratio condition holds for all effort dimensions simultaneously). If one effort dimension is perfectly substitutable for price, in buyer preferences, then the price floor may be limited in its usefulness altogether. For example, when American Airlines tried to constrain travel agents against undercutting their price schedule, agents simply offered exceptionally low prices on hotel and car rental packages with the tickets (see discussion by Judge Frank Easterbrook in Illinois Corporate Travel v. American Airlines Inc. (CCH Trade Regulation Reports, P. 61,921)). Both the direct and indirect mechanisms break down when an increase in a noncontractable dimension of effort is a perfect substitute for a lower prices. There is also the question of whether price floors, as opposed to some other reward, are necessary to protect downstream incentives to provide effort.

362   Edward M. Iacobucci and Ralph A. Winter The economic explanation of why a manufacturer would benefit from imposing resale price maintenance in the form of a vertical price floor thus reduces to asking why retailer demand is relatively more sensitive to price, relative to sales effort, than market demand as a whole. We express this condition for the profitability of resale price maintenance below:

ε e / ε p ( for retailer ) < ε e / ε p ( for market )



(15.2)

We can now return to the question of why retailer incentives deviate from market incentives, and apply the framework to explaining observed contracts. We offer within the framework five theories of the incentive for resale price maintenance, although many other theories are available. The first is a correlation argument. Suppose that consumers most willing to shop among retail outlets for their preferred combination of price and sales effort are also those consumers for whom price matters relatively more than sales effort. This is a natural assumption because sales effort often reduces the time cost of purchasing; consumers end up with the same physical product wherever they purchase but at lower time cost when greater sales effort takes the form of shorter cashier lines and more highly trained sales staff and so on. Consumers willing to travel to shop at different stores are those with low costs of time, and these consumers are also willing to tolerate long cashier lines in favor of lower prices. This structure yields individual retailer demand that is relatively price elastic, with the inequality (15.2) being met. In other words, the consumers whom a retailer attracts away from other retailers are relatively more influenced by low prices than by high sales effort. Retailers set their sale strategy to attract consumers not just into the market, but also away from other retailers. Attracting consumers away from other retails involves setting low prices, since this is what attracts shoppers, but is a pure waste from the prospective of total market demand. The demand attracted away from other retailers does not increase demand for the product at all. Retailers are therefore biased in their strategies towards low prices and inadequate service. Resale price maintenance counters this inefficiency, altering incentives towards higher sales effort and constraining prices against falling (Winter 1993). A second theory is that retailer effort towards product promotion and greater product awareness may influence demand upwards in other outlets. Even prominent displays of a product on the store floor or in shop windows raises consumers awareness of the product and makes them more likely to purchase the product not just from the outlet undertaking the sales effort but from other outlets as well (if the consumer happens to find another outlet most convenient when the need for the product arises). This is particularly true if a retailer’s decision to display (or simply carry) a product conveys information about the quality of the product (Marvel and McCafferty 1984). If even only some potential consumers of other outlets are affected this way, the market demand for the product is relatively more sensitive to sales effort as opposed to price than the individual retailer demand, that is (15.2) is met. Resale price maintenance again counters the distortion in incentives.

Vertical Restraints across Jurisdictions   363

The third theory is similar. The quality of a product such as a photocopier machine, an appliance, or an automobile may depend not only on the quality decisions of the manufacturer but also on the quality of postsales service or installation by the dealer. If the consumer cannot perfectly distinguish between a failure of quality as between the dealer and the product itself, then greater efforts by the dealer towards quality enhance the reputation not just of the dealer but of the product itself. Again, the market demand will be relatively more sensitive to the effort expended by the dealer (relative to price) than the dealer’s own demand is. The condition (15.2) for the profitability of resale price maintenance is met. A fourth theory is the classic “free-riding” story (Telser 1960). Suppose that stereo equipment requires detailed information and a listening room experience in order for a consumer to decide which model of amplifier or speakers best fits her needs. Outlets provide this information. A new stereo store could open up selling equipment in boxes, with very low prices but no information provided at the point of sale. Consumers could avail themselves of the information provided by the informing stores, which charge higher prices to cover the cost of the information provided, then make their purchases at the low-priced store. Demand for the product as a whole will suffer from this type of free-riding. Resale price maintenance restores demand by preventing this kind of free-riding. Without the ability to attract consumers on the basis of low prices alone, outlets must provide the entire package of information and competitive prices.11 The free-riding story is far stronger than is necessary to explain resale price maintenance, as Klein (2009), Winter (2004), and a number of other authors have argued. Our framework makes this very clear. The free-riding story involves a positive impact of service by one retailer (the informing retailer) on the demand faced by another (the free-riding retailer). This is a positive cross-elasticity. This is sufficient for (15.2), but all that is necessary for the condition for resale price maintenance to be profitable is that demand cross-elasticity be relatively less sensitive for sales than it is for price. To generate an incentive for resale price maintenance by an individual manufacturer, it is not necessary that the sales cross-elasticity be positive (as it is in the classic free-riding story: greater information by the informing store increases demand at the free-riding store). That is, it is not necessary that some stores (noninforming stores) benefit from the sales effort supplied by other stores. If the sales cross-elasticity is negative but relatively low, resale price maintenance is profitable. 11  As an example, in Applewood Stoves v. Vermont Castings, Inc. USC.A 7th CC No. 86-2818, Judge Richard Posner writes:

As a new company, selling a somewhat complex product [wood-burning stoves], Vermont Castings . . . needs dealers who understand the product, can explain it to consumers and can persuade them to buy it in preference to substitute products. . . . These selling efforts, which benefit consumers as well as the supplier, cost money—money that a dealer can’t recoup if another dealer “free-rides” on the first dealer’s efforts by offering a discount to consumers who have shopped at the first dealer. . . . As one of Vermont Casting’s dealers explained in a letter to it, “The worst disappointment is spending a great deal of time with a customer only to lose him to Applewood because of price. This letter was precipitated by the loss of 3 sales of V.C. stoves today [to] people whom we educated and spent long hours with.” (CCH, Trade Regulation Reports, 1I58,344, p. 12)

364   Edward M. Iacobucci and Ralph A. Winter In addition to the set of theories under which a manufacturer may want to alter the mix of price and nonprice competition among its distributors, we would mention an additional explanation of resale price maintenance. The restraint may be necessary to encourage adequate inventory holdings by downstream distributors. As Deneckere, Marvel, and Peck (1996, 1997) have shown, the price system alone fails to convey adequate incentives for inventory investment by competitive retailers. Krishnan and Winter (2007) reach the same conclusion for retailers with market power, based on a framework similar to the one that we have adopted above. The evidence for the role of resale price maintenance in enhancing inventory incentives is strong. When resale price maintenance contracts were struck down as illegal in the early 1970s, manufacturers were suddenly constrained in their design of distribution systems. Retailer inventories collapsed for some products and the distribution of the products suffered.12

15.2.3.  Territorial Restraints The general framework that we have offered for the efficiency of price restraints is immediately applicable to territorial restrictions. Let us revisit the market failure question: why would a manufacturer not adopt a simple wholesale contract with only prices? When distributors are geographically differentiated, for example, they sell in different countries, the failures of the simple, nonexclusive price contract to align distributor incentives with supply chain efficiency are clear. Suppose that a manufacturer assigns each country to a particular national distributor, but that there are parallel imports into each country. The sales effort that a distributor in one country undertakes to establish a valuable brand name for the product in its home country benefits any foreign distributor that exports into that country. Higher quality of retail service, product promotion or any activity that adds to brand name capital or simply to demand will have spillover benefits to the foreign distributor. These positive spillovers lead to a positive cross-elasticity with respect to sales effort across national distributors; an increase in promotion in one country adds to the total sales of the foreign distributor.13 It is easy to show that a positive cross-elasticity is sufficient for condition (15.2) on the market failure of the simple price contract. Furthermore, in the context of geographical differentiated distributors, any efforts that a distributor undertakes specifically to attract demand away from a rival distributor, such as investing in a network to export to the rival’s territory, are a pure waste. Investments that merely transfer demand 12 

Corning Glass Works used this restraint from 1937 until it was prevented from doing so in a case brought by the Federal Trade Commission in 1975. In interviews ten years after the case, Corning executives indicated that one of the most important effects of the case was the loss of many of its smaller outlets. In another example, after legislation had ended an earlier era of resale price maintenance, the number of dealers selling Schick shavers fell from 35,000 to 7,000 in one year (Andrews and Friday 1960). 13  The one potential theoretical source of negative cross-elasticity is the attraction of foreign buyers by the increase in domestic promotion. But it is unlikely that many customers will travel to another country to take advantage of better service or promotion.

Vertical Restraints across Jurisdictions   365

from one distributor to another add nothing to total product demand.14 The price system alone fails to coordinate incentives. It is natural for a manufacturer to impose territorial restraints in order to eliminate the wasteful activity. Territorial restraints, in short, eliminate the wasteful effort to divert demand away from rival distributors. In fact, territorial restraints are a more powerful resolution of incentive distortions than vertical price restraints in two respects. When effort (any nonprice determinant of demand) is multidimensional, “first-best” resolution of incentive distortions with price restraints requires that the demand elasticity with respect to effort be identical across all dimensions. (Recall our discussion that if one dimension of sales effort is perfectly substitutable with price decreases, resale price maintenance is useless.) This is not true of territorial restraints. Under the simplest set of assumptions, for example, territorial restraints are accompanied by variable prices equal to marginal cost so that the downstream distributor becomes a residual claimant on profits earned from its territory. As such, it takes privately efficient decisions in all dimensions. In addition, resale price maintenance involves an upstream manufacturer imposing the same price floor across all retailers, instead of allowing the price to vary across retailer locations (and over time) in response to information about demand that may be available only at the retail level. Territorial restraints, by supporting residual claimancy contracts, avoid distortions that result from this rigidity of prices.

15.2.4.  Exclusivity Restraints and Tying Exclusivity contracts extend beyond territorial exclusivity. Exclusivity restraints may be imposed on buyers or retail intermediaries (“exclusive dealing”), prohibiting them from purchasing from other sellers. Or these restraints may be imposed on sellers. The economics of exclusivity are reviewed in the chapter by Douglas Bernheim and Randall Heeb and the chapter by Howard Marvel in this volume. Here we outline two of the most important efficiency explanations of the contracts. Marvel (1982) has explained exclusive dealing as a remedy to potential free-riding not on retailer investments (like the free-riding explanation of resale price maintenance) but on manufacturer investments. Manufacturers generate customers for their products through advertising and investment in their brands. A manufacturer is essentially providing to a retailer a tied product—the physical product itself plus investment in generating demand for the product. If the dealer carries a low-priced, nonadvertised product, then the dealer can sell the lower priced product to customers attracted to the dealer by the promotion expenditure on the part of the manufacturer. The dealer can gain a reputation for offering similar products at lower prices. The problem with this situation is that the manufacturer then 14  This argument assumes that the impact of investment in exporting channels is mainly diversion of demand away from the distributor located in the destination country, rather than bringing more demand to the market.

366   Edward M. Iacobucci and Ralph A. Winter has the incentive to cut back on productive, brand-enhancing investment. Exclusive dealing is efficient when used to prevent free-riding. Exclusivity in contracts can be used to prevent holdup. Suppose that a franchisor sells to a franchisee the right to open an outlet in a small town. Once the franchisee has built up the market for the product by maintaining a high level of quality and through promotion, the franchisor can sell a share of the market to a new franchisee— unless the first franchise has territorial exclusivity rights. In any contract, exclusivity can serve to protect against inefficient investment. Contracts are inevitably incomplete in that they do not specify the rights and obligations to parties for all time; some negotiations are necessary even after a contract has been signed. In a contract without exclusivity, a party may have the incentive to invest in options to leave the contract and transact with other parties—not with the intention necessarily of exercising the option but rather for the purpose of enhancing its threat point in bargaining with the contractual partner. The investment in these options would be inefficient, since it has the impact of simply implementing a transfer between contractual parties (Segal and Whinston 2000). Exclusivity may be reciprocal, with the buyer and seller each constrained to deal with the other, in an arrangement that is sometimes referred to as quasi-vertical integration. Tying, another kind of exclusivity, may also have an efficiency explanation. The most common efficiency explanation of tying is that there are economies in jointly providing two “products.” For example, the vast majority of consumers prefer to buy a pair of shoes and not just one. Selling shoes in pairs reduces the cost of provision. The law in the EU and the United States appropriately accounts for this efficiency explanation providing that a tying restriction must involve two separate products, which is not true unless there is significant independent demand for each possible product. But even if there are two clearly independent products, tying may be efficient. For example, when a consumer cannot distinguish failures in service quality from failure of the product itself, then increased efforts by service providers exert a positive externality between the service providers and the product manufacturer. Because of this externality, an independent supplier of service may have weak incentives to provide quality service: the service provider does not internalize the reputational benefits of high quality. A tying arrangement may require buyers to buy high-quality service from the manufacturer and thus efficiently protect the manufacturer’s reputation for quality.15 Tying may also be imposed as a price discrimination mechanism, which, as we discussed above, may increase surplus. For example, tying aftermarket products to original equipment may allow the manufacturer to meter demand for its product. Buyers with relatively high demand for original equipment, such as a printer, may have relatively high demand for aftermarket products, such as ink. Tying the aftermarket product to

15  Iacobucci (2004) shows that, if buyers anticipate the externality between independent service providers and manufacturers, there may or may not be a need to tie to protect a reputation for quality. The incentives for tying depend on the technological relationship between service and product quality.

Vertical Restraints across Jurisdictions   367

the original equipment, and imposing a price-cost markup on the tied good, allows the seller to extract more surplus from high-demand buyers than it would otherwise.

15.2.5.  Restrictions on Internet Distribution Because the European Guidelines adopt a number of restrictions against limiting online sales, it is useful to consider the possible efficiency roles of vertical restraints on online sales. The Guidelines’ rules are complex, but include the limits on the proportion of overall sales made by a distributor on the Internet. Why would a manufacturer want to limit the proportion of sales done over the Internet? Consider the case of luxury good sales, which is one of the most important sectors affected by the European Commission’s restrictions on Internet sales (Buettner et al. 2009a and 2009b). Image is a vital component of luxury products, in the sense that many consumers reveal a preference for products on which suppliers have invested substantially in image. Suppliers of perfume, for example, spend much more on creating the image of the product than on the chemicals in the physical product itself. A supplier of perfume that simply sold its product in bulk, with no advertisement or other investment in image (perhaps at a discount to a consumer who provided her own container) would not succeed in this market. The product itself is not just the liquid perfume. Consumers value not just the scent but the image of the product. Manufacturers of perfume are not restricted—nor should they be—in their expenditures on image. Investment in image is undertaken not just by manufacturers but by bricks-and-mortar retailers. Consumers are attracted to expensive luxury goods by the retail environment, the retail store prestige and by promotion at the retail outlet. Outlets like Tiffany or Saks Fifth Avenue come to mind, but any upmarket retailer can both add directly to the demand for a product as well as to its image through investment in prestige and an attractive shopping experience. This experience can include point-of-sale information such as product sampling in the case of perfume, but retail investment in these dimensions is much more general. When outlets are allowed to sell unlimited amounts through the Internet, however, then these outlets benefit from the investment in images by bricks-and-mortar outlets without contributing anything to the investment themselves. This free-riding reduces the incentives for outlets to invest in the product image or to carry the product at all, and this reduces demand for the product. A manufacturer limits sales through the Internet to protect the incentives of bricks-and-mortar retailers to offer investment in image. Manufacturers may benefit to some degree from Internet sales through expansion of its market or from the low prices that result. A manufacturer that constrains the percentage of sales through the Internet is simply adjusting the mix of its price and nonprice demand variables (image investment, for example) just as if it were adjusting these variables directly. Restrictions to limit sales through the Internet are not part of a scheme to limit competition among manufacturers and should not be regulated.

368   Edward M. Iacobucci and Ralph A. Winter

15.3.  Anticompetitive Uses of Vertical Restraints Vertical restraints have a host of efficiency explanations. There are, however, anticompetitive explanations as well. We next turn to these anticompetitive explanations.

15.3.1.  Resale Price Maintenance Resale price maintenance can serve to facilitate an upstream cartel among manufacturers. Coordinating wholesale prices would be difficult for members of an upstream cartel because these prices are not posted and may be part of more complicated contracts. Coordinating an upstream cartel via the monitoring of retail prices without vertical restraints would also be difficult because of the variation or “noise” that enters the relationship between a wholesale price and the set of retail prices charged in different locations. Retail price floors allow upstream cartel members to agree on prices and to monitor prices. Telser (1960) used this theory to explain the adoption of resale price maintenance by GE and Westinghouse in the market for large lamps. Jullien and Rey (2007) formalize this argument. RPM can also be used to support a downstream retail cartel. When retailers sell multiple products and require collectively that all manufacturers in a product market engage in RPM, then the retailer cartel can effectively implement cartel pricing across products. This explanation is of some historical importance in explaining, for example, the success of traditional drugstores in delaying the development of discount drugstores in North America (Overstreet 1983, 143). The US Supreme Court in Leegin discussed a third potential anticompetitive theory of resale price maintenance: that the practice results in exclusion at the retail level. A manufacturer can use resale price maintenance to protect rents at the retail level, as in the Klein-Murphy theory, but the retailer “performance” that is contingent upon continued receipt of the rents is not provision of adequate service but rather agreeing to refrain from carrying the products of a new entrant. A dominant firm or small set of collectively dominant firms can thus protect their dominant position against entry by sharing the rents from the dominance with retailers. Retailers know that once entry is allowed, rents will be much reduced in the market and, as agents who share in these rents, retailers will suffer. John Asker and Heski Bar-Isaac (2011) offer a thorough analysis of this theory (see also Paldor 2008). The theory of resale price maintenance as exclusionary requires that there be a small number of distributors, as in the case of the American Sugar Refining Company, discussed by Asker and Bar-Isaac (see also Zerbe 1969 and Marvel and McCafferty 1985). With many retailers, it is unlikely that a new entrant could not gain toehold entry by sharing profits from entry with one or a small number of retailers.

Vertical Restraints across Jurisdictions   369

15.3.2.  Territorial Restraints It is possible as a matter of theory that territorial restraints dampen interbrand competition. Rey and Stiglitz (1988) show that territorial restraints can have the impact of dampening competition between manufacturers. Recall that in the simplest theory of territorial restraints, a manufacturer sets variable price equal to marginal cost and charges downstream retailers a fixed fee in exchange for their right to sell exclusively in an area. Under a territorial restraint, with variable wholesale prices set above marginal cost, a manufacturer acts as a less aggressive competitor.16 The Rey-Stiglitz theory has, however, not been applied to specific cases of which we are aware and would be difficult to distinguish from industry-wide use of the restraints for efficiency purposes.17 Exclusive territories may support price discrimination by allowing the upstream seller to partition buyers downstream into different classes, each facing different prices. Such a scheme might arise where buyers in different territories have different elasticities of demand for a product. While it is conceivable that price discrimination would lower welfare, price discrimination may also increase welfare. Indeed, as noted above, moving from a single price set by a monopolist along the spectrum of increasing surplus extraction to the limit of perfect price discrimination will, on average, increase total welfare. Given that there are no practical means for determining whether price discrimination in any given case would lower welfare, we join the consensus of economists that would not legally forbid price discrimination. We therefore also would reject the establishment of limits on exclusive territories out of concerns about price discrimination.

15.3.3.  Exclusivity Restraints and Tied Sales It had, until recently, been argued that an exclusivity restriction would not be accepted by a distributor unless the distributor were offered some offsetting benefit to the restriction distributor under the restraint (Bork 1978). Contracts maximize the sum of benefits to the contractual parties, and if an exclusivity restriction is added it must be surplus-increasing. Aghion-Bolton (1987) showed this to be false. If a contract imposes externalities on parties outside the contract, then the contract may be anticompetitive as it benefits parties to the contract. The most important formulation of this theory 16 

More precisely, when manufacturers compete in prices, vertical restraints combined with above-marginal-cost wholesale prices have the effect of shifting reaction curves upwards. 17  An additional theory of territorial restraints as anticompetitive would build on the Asker and Bar Isaac (2011) theory of resale price maintenance as facilitating an exclusionary equilibrium. Suppose that the only feasible means of entry into an industry by an entrant is to choose a particular location, then build up gradually from there. If territorial exclusivity is used by the incumbent, then a distributor that is the first to accommodate the entrant would quickly find its profit eroded by the competition upstream. The condition for profitably refusing an entrant would be weaker (more general) than in the Asker–Bar Isaac model.

370   Edward M. Iacobucci and Ralph A. Winter involves an entrant with fixed costs that can be covered only if the entrant secures enough buyers (Rasmusen, Ramseyer, and Wiley 1991, Segal and Whinston 2000). If the incumbent offers each buyer a small price reduction in exchange for an exclusive contract, then acceptance by all buyers is a Nash equilibrium because each buyer knows that refusing the offer of an exclusivity contract (given that other buyers are accepting the contract) will not be enough to induce entry.18 Even a more efficient entrant can be inefficiently deterred from the market by exclusive contracts. A parallel theory applies to the possibility that tied sales can be anticompetitive (Whinston 1990). The following example of Whinston’s theory is offered by Carlton and Heyer (2008, p. 14) (and attributed to Rob Gertner): Consider the case of a hypothetical island on which there is a monopoly hotel serving many tourists. Natives live on the island. The hotel operates a restaurant, which competes for diners, both tourists and natives, in competition with local restaurants. By tying meals to lodging, the hotel can so diminish the number of tourists dining at local restaurants that, in the extreme, lack of scale prevents any local restaurants from surviving. The hotel thus acquires a monopoly over natives in the provision of restaurant services.

Whinston’s theory shows that market power in one market can be profitably extended to market power in two markets provided that (1) not all consumers in the second market use the product in the two markets jointly; (2) in the absence of tying, there is some market power, for example a duopoly, in the second market; and (3) a rival or rivals incur some fixed costs, such as costs of product development or continual product updating, in the second market. Whinston notes, however, that even in the simple models that he develops, which set aside various efficiency explanations of tying in order to focus on exclusion, the welfare implications are unclear. In summary, contracts with exclusivity constraints or tying restrictions can be anticompetitive but there are many theories available under which these restrictions serve efficiency or price discrimination roles. The application of theories to a particular case, to understand whether the evidence in the case supports an efficiency explanation or suppression of competition either in the form of higher prices or exclusion, is a challenging exercise because even theories supporting exclusion can yield ambiguous welfare implications of a practice. It would appear that the vast majority of contracts with these restrictions are efficient, since such restrictions are seen across such a wide variety of markets, including markets with competitive structures. We turn next to a systematic review of empirical evidence on vertical restraints.

18 

Note that with small bribes, buyers would be each be better off in another Nash equilibrium: only (n − 1) accept, where n is the minimum number of free buyers that will induce entry. Segal and Whinston adopt a refinement of Nash equilibrium that rules out the dominated equilibrium and find that the essential argument remains valid.

Vertical Restraints across Jurisdictions   371

15.3.4.  Empirical Evidence on Vertical Restraints Our review of empirical evidence on vertical restraints is in the form of a summary of summaries. Cooper and coauthors (2005a and 2005b) reviewed 24 empirical papers published between 1984 and 2004 on the effects of vertical integration, RPM, and exclusive territories. LaFontaine and Slade (2005) reviewed 23 empirical papers on vertical restraints and vertical integration. The samples of papers reviewed in the two studies overlapped. The Lafontaine and Slade sample include 15 papers on RPM, exclusive territories, and vertical integration. Of these papers, 13 conclude that the restraints under study either benefitted consumers or did not harm them. Two papers found evidence of higher prices under exclusive territories, but the higher prices may simply be a trade-off accepted for higher sales effort, as Lafontaine and Slade point out. Lafontaine and Slade conclude in summarizing their evidence that when manufacturers’ choose contracts with vertical restraints “not only do they make themselves better off, but they also typically allow consumers to benefit from higher quality products and better service provision.” Similarly, Cooper and coauthors reach the conclusion that the literature offers little support for the proposition that vertical restraints or integration are likely to harm consumers. Some papers reviewed find evidence consistent with both anticompetitive and procompetitive incentives, but none finds that the hypothesis of procompetitive incentives can be rejected in favor of the alternative. O’Brien (2011) reviews an additional set of recent papers in the area that are again consistent with the overall conclusion. The empirical evidence reviewed is not, of course, a random sample of contracts from markets across the economy. Some industries such as beer distribution and gasoline distribution are overrepresented. And the overall thrust of the evidence, that vertical restraints tend to be procompetitive, does not rule out the possibility of valid cases against particular uses of vertical restraints. By analogy, less than 2% of mergers raise competitive concerns, yet merger policy is an active and well-founded area of competition policy. But the evidence supports our conclusion that competition policy towards vertical restraints in any case must be approached with a prior expectation that the restraints are very likely to be efficient. There is no place in competition policy for placing the burden of proof on defendants to show that restraints are efficient. With this perspective in mind, we review the law of vertical restraints in the EU and United States in the next section, and offer a critique in the final section.

15.4.  The Law of Vertical Restraints in the United States and the EU The differences between the US and EU approaches to competition law on vertical restraints are interesting as a purely academic matter, and critical as a practical matter in light of the

372   Edward M. Iacobucci and Ralph A. Winter potentially severe financial consequences that follow from a finding of a legal violation in either jurisdiction. Significant financial penalties can follow violations of vertical restraints law in either jurisdiction. In the United States, those adjudged to have improperly lessened competition through the use of vertical restraints face the threat of publicly imposed fines, or treble damages as a consequence of private actions. The EU does not, at present, allow private actions against anticompetitive behavior of any kind, including improper vertical restraints, but has imposed very large public fines for anticompetitive conduct. In 2007, for example, the European Commission imposed a fine of 497 million euros on Microsoft for tying Windows Media Player to Windows Operating System.19 In what follows, we review the key principles of law in each jurisdiction on RPM, exclusive territories, tying, and exclusive dealing. We focus in each case only on the “federal” law, examining EU and US law and setting to the side potentially important deviations from the pan-jurisdictional approach at the state, or member state, level. As an introductory matter, note that the central sources of law in both jurisdictions are statutes that are phrased in broad, general terms. The US Sherman Act has a provision on anticompetitive agreements, Section 1, as well as Section 2 on monopolization. Other relevant statutes include the Clayton Act, Section 3 of which prohibits anticompetitive tying and exclusive dealing. In Europe, the Treaty on the Functioning of the European Union (“TFEU”) also has a provision on anticompetitive agreements and concerted practices, Article 101, and one on abuse of dominance, Article 102. These provisions set out standards, not rules, and judicial and administrative interpretations are consequentially very important in the application of the law. Such an approach contrasts with other jurisdictions, such as Canada, where the relevant competition statute is codified to a much greater extent, with, for example, practices such as RPM dealt with in specific sections.20 One cannot examine the statutes in order to understand the prevailing approach to a particular vertical restraint, but rather must go to the case law and administrative pronouncements, the latter being particularly important in this area in the EU.

15.4.1.  Resale Price Maintenance US antitrust law historically banned minimum RPM. The leading case for decades was Dr. Miles Medical Co. v. John D. Park & Sons Co.21 In this case, the Supreme Court held that a manufacturer of medicines could not impose minimum price restrictions on downstream resellers. Just as horizontal agreements between sellers of a product are per se illegal, which is to say that they violate antitrust law regardless of their competitive effects in any given case, so too were agreements between sellers and buyers of a product. Equating horizontal and vertical agreements for antitrust analysis makes little economic sense, but such thinking has been influential in the United States. 19 

Microsoft Corp. v. Commission of the European Communities (2007) CFI. See e.g., Canadian Competition Act, Section 76. 21  220 U.S. 373 (1911). 20 

Vertical Restraints across Jurisdictions   373

Judicial decisions retreated somewhat from the harshness of the per se illegality standard by creating exceptions, such as an exception for a seller that stipulates resale prices and unilaterally refuses to deal with downstream firms that do not adhere to these prices; in such case, there is no agreement, and hence no per se illegal agreement.22 Moreover, statutory exemptions to federal antitrust laws were passed, such as the Miller-Tydings Act in 1937 and the McGuire Act in 1952, that allowed state fair trading laws, which permitted RPM, to apply in states that had adopted such statutes. The McGuire Act and the Miller-Tydings Act were repealed by the Consumer Goods Pricing Act of 1975. Two recent cases altered the RPM legal landscape significantly. First, State Oil v.  Khan23 changed the law with respect to maximum resale prices. Recognizing that there is little reason to conclude that a price cap would hurt competition or consumers, the Supreme Court rejected per se illegality for maximum RPM and instead adopted a rule of reason approach: a plaintiff would be required to show a harm to competition from RPM to succeed. Despite the venerable status of the Dr. Miles precedent, the Supreme Court abandoned per se illegality in Leegin Creative Leather Products, Inc. v. PSKS Inc.24 In a decision replete with reference to economic thinking on the matter, the Supreme Court rejected per se illegality for RPM, holding that it should be considered under the rule of reason instead. The Court suggested that RPM may be more suspect where there is evident market power on the part of sellers or buyers, and where RPM may be used in a manner that is harmful to horizontal competition, by supporting a cartel, or by inducing downstream firms to exclude the upstream firm’s rivals. The notion that an upstream firm and downstream firm are fixing prices with necessarily harmful effects when agreeing to RPM is no longer accepted in US federal antitrust law.25 But it would be wrong to say that the law on RPM is approaching per se legality. The onus of proof to demonstrate various factors that support the legality of RPM rests with the defendant.26 In practice, as of mid-2012 lawyers were still designing “Colgate programs” for their clients to ensure that the clients’ distribution practices fall within the Colgate exception to Dr. Miles.27

22 

U.S. v. Colgate & Co. 250 U.S. 300 (1919). 522 U.S. 3 (1997). 24  551 U.S. 877 (2007). 25  State laws on RPM vary, with some states effectively banning RPM under their own antitrust laws. 26 Following Leegin, the law on RPM, as interpreted by the FTC in granting a petition to Nine-West in 2008 to modify an earlier FTC order, is that a manufacturer bears the burden of proving that the use of RPM is procompetitive unless it can prove the absence of all of the so-called Leegin factors: (1) that dealers, not the manufacturer, were the impetus for RPM; (2) that the use of RPM was ubiquitous in the industry; (3) that there is a dominant manufacturer or dealer. 27  Several attempts have been made in Congress to reimpose the per se ban on RPM after Leegin by drafting statutes to repeal Leegin. Senators Herb Kohl, Joseph Biden, and Hillary Clinton introduced the Discount Pricing Consumer Protection Act immediately following the decision. This bill proposed to revise Section 1 of the Sherman Act to make RPM per se illegal. Although this bill never made it out of committee, similar bills in the House and the Senate are in process as of the writing of this chapter but have not been passed. 23 

374   Edward M. Iacobucci and Ralph A. Winter The EU, on the other hand, continues to disapprove of minimum or fixed RPM. In 2010, the EU adopted Regulation 330/2010, which exempts a range of vertical agreements from the TFEU Article 101 ban on anticompetitive arrangements, and offers guidance on its approach to remaining restraints. The Regulation will be in place until 2022. The Regulation creates safe harbors that automatically protect certain “non-hardcore” vertical agreements from antitrust sanctions (though failing to fall within a safe harbor does not create a presumption of illegality for non-hard-core restrictions). The general structure is to exempt all vertical non-hard-core restrictions where the market share of seller and buyer28 in each of their respective markets is below 30%. Hard core restrictions, on the other hand, are subject to a presumption of illegality. Minimum or fixed RPM is a hard core restriction under the Regulation. Maximum RPM, on the other hand, is not a hard core restriction under the Regulation, and thus can qualify under the safe harbor. The Regulation cites a number of reasons for the treatment of RPM as a hard core restriction.29 Some are similar to the concerns about horizontal competition that the US Supreme Court pointed out in Leegin: RPM may facilitate collusion at the upstream or downstream distributor levels; or may be used to induce retailers not to carry a rival’s brand. Other reasons are more cryptic. For example, the Regulation suggests that even absent collusion, RPM may “soften” competition between manufacturers and/or retailers. Others parts seem to stress the importance of protecting intrabrand competition. For example, the Regulation notes that minimum or fixed RPM prohibits price-cutting, and thus the direct effect of RPM is a price increase. The classification of a practice as a hard core restriction means that there is a presumption that the practice is contrary to Article 101 of the TFEU, and that safe harbors are unavailable. Minimum or fixed RPM is not illegal per se, however, as firms may be able to invoke the efficiencies defense under Article 101(3), which allows practices even if they are anticompetitive as long as they generate efficiencies and these efficiencies are passed on to the consumer. The Regulation acknowledges that RPM may be efficient, as where, for example, a seller introduces a new product and seeks to expand its distribution network. But with a presumption of illegality, the onus will be on the firms to prove the efficiency effects to justify RPM.

15.4.2.  Exclusive Territories Just as the law on RPM has changed considerably over time in the United States, so too has the approach to exclusive territories. But while change in the law on RPM took many decades, the Supreme Court’s precedent on exclusive territories oscillated dramatically in a period of only fifteen years. The first Supreme Court case on the matter was White Motor Co. v. U.S.30 In this case, a truck manufacturer sought to impose territorial 28 

Previously, only the seller’s market share was relevant. See discussion in Regulation 330/2010 at 63. 30  373 U.S. 253 (1963). 29 

Vertical Restraints across Jurisdictions   375

restrictions on its distributors. The district court accepted the government’s argument that such restrictions were analogous to horizontal market division, and to RPM, and thus should be treated as illegal per se. The district court granted summary judgment to the government. The Supreme Court, however, overturned, holding that summary judgment was inappropriate. The majority of the Court did not reach a conclusion on the proper treatment of territorial restrictions, holding that a full trial, with consideration of all the economic evidence, was warranted. Four years later, the question of the appropriate treatment of exclusive territories was before the Court again in two cases, U.S. v. Sealy Inc.,31 and U.S. v. Arnold, Schwinn & Co.32 The Court was decisively against the practice in these cases. Sealy involved a number of licensees of the Sealy brand that manufactured mattresses. Conditions to license the brand included a number of vertical restraints, including RPM and territorial exclusivity. The Court held that the exclusive territorial restraints were tainted by the aggregate nature of the nature of the restraints in the case, which involved other legally suspect practices including RPM, and illegal. Schwinn did not involve such an aggregation of restraints, but again the Court expressed disapproval. In particular, making reference to doctrines against restraints on alienation, a majority of the Court condemned as unreasonable the imposition by a manufacturer of restrictions on the downstream sales of products after title to the products had passed to a downstream distributor. Schwinn met with considerable hostility and did not survive long. Continental TV Inc. v. GTE Sylvania Inc.33 involved a manufacturer of televisions agreeing with its franchised dealers that they would not be permitted to sell Sylvania products if they were to relocate without Sylvania’s approval. While not strictly speaking an agreement to establish exclusive territories, the arrangement had a similar effect. Stressing the importance of economic analysis in applying antitrust law, and the role of territorial restrictions in potentially stimulating interbrand competition, a majority of the Supreme Court rejected the per se approach to exclusive territories, concluding that they should be considered instead under the rule of reason. Since then, as with most practices governed by the rule of reason, territorial restrictions have rarely been challenged successfully. Exclusive territories law in the EU raises interesting trade-offs. On the one hand, the case law has long recognized the potential economic benefits of such restraints; on the other hand, in addition to concerns about intrabrand competition, exclusive territories raise concerns about the greater integration of the EU itself.34 The trade-off has manifested in the case law and Commission guidelines over the years, most recently in the EU 2010 Guidelines that contain a general disapproval of exclusive distribution, but with fairly broad exceptions. Under the Guidelines, strict exclusive territories are classified as a “hardcore” restriction on competition, which as set out above not only removes them from the possibility 31 

388 U.S. 350 (1967). 388 U.S. 365 (1967). 33  433 U.S. 36 (1977). 34  See, e.g., Consten and Grundig. 32 

376   Edward M. Iacobucci and Ralph A. Winter of falling within a safe harbor, but moreover establishes a presumption of illegality.35 There are, however, important exceptions to this treatment. For example, block exemptions are available for restrictions on the buyer’s place of establishment as long as they do not also contractually confine sales to a particular territory. Perhaps the most significant exception concerns the distinction between “active” and “passive” sales. Active sales are sales that result from action by the downstream firm, while passive sales are unsolicited. Reflecting trade-offs between economic and integration concerns, the Guidelines allow upstream firms to impose restrictions on active sales by downstream firms outside of designated territories, but disallow a restriction that prevents downstream firms from making passive sales. This is typical of the compromises that EU law on exclusive territories has incorporated over the years, but there was an important addition in the 2010 Guidelines. The Guidelines treat Internet sales as passive sales and, thus, territorial restrictions cannot prevent sales outside a designated territory on the Internet. Other exceptions to hardcore treatment of territorial restrictions include restrictions that prevent wholesalers from selling to end users, from selling to unauthorized distributors, and from selling components to competitors of the upstream firm. In summary, the EU is again stricter than the United States: exclusive territories are subject only to rule-of-reason review in the United States, but may be presumptively problematic under EU law.

15.4.3. Tying The treatment of tying in the United States is simple to state, but much more complex to apply. The basic rule is that tying is per se illegal. But there are preconditions for tying to be per se illegal that undermine significantly the importance of the per se label. First, there must be two separate and distinct products that the seller ties together. The second requirement is that the seller must have market power. The requirement of two products is in some respects a logical prerequisite for a tying agreement, but is not straightforward to apply. Products have multiple features, and distinguishing two features of a single product from two distinct products may be difficult in practice. Cars come with tires; are these elements of a single product, or is this a tie (bundle)? Shoes come with laces; a single package may contain several oranges; a cable television package may have several channels—are these tying arrangements? The Supreme Court offered some guidance on the question in the leading case of Jefferson Parish Hospital District No. 2 v. Hyde.36 In this case a hospital entered into an agreement with a designated professional corporation that it would rely only on its anesthesiologists in providing services to patients. Dr. Hyde, an anesthesiologist who was prevented by the arrangement from providing services to the hospital’s patients, challenged the arrangement as an unlawful Section 1 agreement, and the case found its way to the 35  36 

See Guidelines, supra at para. 50. 466 U.S. 2 (1984).

Vertical Restraints across Jurisdictions   377

Supreme Court. A majority of the Court upheld the per se illegality of tying, but ultimately held that the arrangement at issue was not per se illegal. In so finding, the Court found the existence of two separate products, setting out a test that asked not whether it was feasible for the seller to sell the items separately, but rather whether there existed sufficient demand for each item independently such that they could be considered separate products for the purposes of tying law. Here there was sufficient evidence of distinct demand for anesthetic services independently from hospital services that there were two products, and thus there was a tying arrangement. The market power requirement is also far from straightforward, with courts taking more hawkish and more dovish views at different points in time. The leading case is again Jefferson Parish, which, in holding that the tying arrangement in question was not per se illegal, stressed the importance of the market power requirement. Some indication of market imperfections that might lead to some degree of market power on the part of the seller, such as the presence of insurance in the health services market and thus price-insensitive buyers, is not sufficient; rather, the majority stated that the seller must use its market power to force the buyer into buying a product that the buyer did not want at all, or that the buyer might have preferred to buy from a different seller. Such market power was not present in Jefferson Parish, and the tie was not per se illegal. Given the significant degree of market power required under Jefferson Parish, the later case of Eastman Kodak Co. v. Image Technical Services, Inc. took many by surprise. In this case Kodak ceased to supply parts to independent service operators (ISOs) that worked on its photocopy equipment. ISOs launched an antitrust complaint under Sections 1 and 2 of the Sherman Act, contending, among other things, that Kodak tied the sale of service parts to the sale of its parts. Kodak responded by arguing, among other things, that it did not have market power in the original equipment market, and thus could not have market power in “aftermarkets” for service and parts even if its market share were 100%. The Supreme Court held that summary judgment in Kodak’s favor was not appropriate. Information problems on the part of buyers who may not anticipate high prices in aftermarkets, or problems resulting from buyers being “locked in” to Kodak aftermarkets given past investments in Kodak equipment, could result in the necessary market power that would support a per se tying claim. The “post–Chicago School” approach to market power in Kodak suggests greater scope for successful per se tying claims than Jefferson Parish might imply. In Europe, tying is not treated as a hard core restriction, and thus is eligible to fall within the safe harbor if both seller and buyer have market shares below 30% in both the tying and the tied product markets. Outside the safe harbor, tying arrangements will be assessed on the merits and not subject to any kind of presumption of illegality. The 2010 Guidelines outline the key considerations in a review of the tying arrangement. As in the United States, the determination of whether there are two distinct products, a logical prerequisite for a tie, depends on demand-side considerations.37 If there is sufficient demand for the

37 

Guidelines, supra at para. 215.

378   Edward M. Iacobucci and Ralph A. Winter sale of the items independently of each other, then there are two products. Laces are not independent from shoes, for example, and thus selling laces with shoes is not tying. The analysis of harms to competition from tying turns importantly on the presence of market power. If the seller faces effective competition, the Guidelines conclude that “no anti-competitive effects can be expected . . . unless other suppliers are applying similar tying.”38 Interestingly, however, in reviewing the potential for tying to increase prices, the Guidelines observe that customers may not always be able to calculate the consequences of tying where aftermarkets are involved.39 This would seem to cast doubt on the importance of market power in original equipment markets, as in Kodak. While the requirements of two products and market power are similar to the US approach to tying, the EU diverges from the US approach by requiring an analysis of competitive effects before condemning a tying arrangement. Foreclosure is a central concern, with the Guidelines calling for an examination of the percentage of sales that are covered by the tying arrangement. The Guidelines observe that a tying arrangement may foreclose entry by forcing entrants to provide both tying and tied goods, or by denying scale to independent sellers of tied goods. The Guidelines also express concern about higher prices in aftermarkets, as noted above, and the use of tying as price discrimination: the tied good may meter demand for the tying good, allowing the seller to extract greater surplus from buyers. The focus on high prices, and concern about price discrimination, suggest that tying may be problematic even without competitive effects if the practice helps a seller exploit its market power. The leading case on tying in recent years in the EU is the Microsoft case.40 In this case, Microsoft was found by the Commission to have unlawfully tied Windows Media Player to Windows Operating System. The Court of First Instance found that these were two separate products, and that Microsoft had market power in operating systems. The Court also found that there was the risk of foreclosure associated with the tie; thus there were the requisite competitive effects to make an order against the practice, as well as a fine of 497 million euros. The analysis of foreclosure did not engage in a far-ranging analysis of what the economic benefits to Microsoft would be from extending a monopoly in operating systems to media players, but rather emphasized that the tie prevented competition in media players given Windows’ share of operating systems, and thus was unlawful. To summarize, the United States has a per se rule against tying, but requires that there be two products and that there be market power in the tying market. The EU also requires two products and market power, and indeed establishes a safe harbor for tying arrangements involving less than 30% of the tying and tied product markets up- and downstream. To show that tying is unlawful, the Commission must show negative competitive effects from the practice, though the Guidelines (with their contemplation of price discrimination as a problematic motivation, for example) and the case law (most prominently, Microsoft) do not impose a high hurdle for such a finding. 38 

Ibid. at para. 220. Ibid. at para 217. 40  Supra. 39 

Vertical Restraints across Jurisdictions   379

15.4.4.  Exclusive Dealing As with other practices, the law on exclusive dealing in the United States has undergone a number of twists and turns with the adoption of different tests over time. The earliest leading case is Standard Fashion Co. v. Magrane-Houston Co.41 In this case, a supplier of paper patterns for women’s and children’s clothing entered into exclusive dealing arrangements with a large number of retail outlets. As a result of a contractual dispute, Magrane-Houston, one of the retailers, accused Standard Fashion of violating Section 3 of the Clayton Act by foreclosing a large number of potential outlets for dress patterns. The Supreme Court observed that Section 3 condemned arrangements that “tended” to create a monopoly, but held that this required that an arrangement probably lessens competition or creates an actual tendency to monopoly. The threshold for such a finding was not high, however. The Court concluded that Standard Fashion’s large market share, and the fact that 40% of potential outlets had agreed to exclusivity, would foreclose competition, and consequently there was a violation of Section 3. The strict posture toward exclusive dealing became even more aggressive in Standard Oil of California v. U.S.42 Standard Oil contractually required independent service stations that collectively had 6.7% of the market to buy gasoline exclusively from it. The government contended that the exclusive dealing agreements violated the Sherman and Clayton Acts. The Supreme Court in this case observed that exclusive dealing should be treated more leniently than tying, stating, “Tying arrangements serve hardly any purpose beyond the suppression of competition,” while exclusive dealing may be economically advantageous to both sellers and buyers.43 At the same time, however, Justice Frankfurter expressed skepticism that courts would be able to engage in the economic analysis necessary to apply a full rule-of-reason approach. In the result, the Court concluded that exclusive dealing was unlawful where a substantial share of the line of commerce that was involved. It thus failed to overturn the finding of the district court that foreclosing 7% of retail outlets with exclusive dealing contracts was illegal per se. The Supreme Court took the economic merits of exclusive dealing more seriously in Tampa Electric Co. v. Nashville Coal Co.44 The Court rejected lower court findings that an agreement between a coal supplier and an electric company that the latter would buy coal only from the former violated Section 3 of the Clayton Act. The lower courts examined the volume of commerce affected by the contract, a very large volume in absolute terms, and concluded that exclusivity must lessen competition by foreclosing access to the electric company by other coal suppliers. The Supreme Court redefined the relevant geographic area that was covered by the arrangement to a much larger area than that considered by the lower courts. As a consequence of this redefinition, the foreclosed commerce was less than 1% of the relevant market. This was sufficient to find for the 41 

258 U.S. 346 (1922). 337 U.S. 293 (1949). 43  Ibid. at 305–6. 44  365 U.S. 320 (1961). 42 

380   Edward M. Iacobucci and Ralph A. Winter defendant, but the Court went on to set out a test for substantial foreclosure that invited lower courts to follow a rule-of-reason, and not a per se, approach in the future: To determine substantiality in a given case, it is necessary to weigh the probable effect of the contract on the relevant area of effective competition, taking into account the relative strength of the parties, the proportionate volume of commerce involved in relation to the total volume of commerce in the relevant market area, and the probable and immediate effects which pre-emption of that share of the market might have on effective competition therein.45

As Hovenkamp points out, following Tampa Electric, lower courts have generally followed the rule of reason in assessing exclusive dealing (Hovenkamp). In the EU, exclusive dealing, or “single branding,” is judged on what is effectively the rule of reason: the arrangement must lessen competition to be unlawful. Single branding is not a hard core restriction and thus is eligible for a safe harbor under the Guidelines. For this practice, there is both a market share and a temporal requirement for the safe harbor: if seller and buyer have less than 30% of their respective markets and the exclusive arrangement is less than five years long, then the practice will not be challenged. Arrangements that fall outside the safe harbor are not presumed wrongful, but rather will be subject to a competition test. The Guidelines recognize that if there is competition for the entirety of each customer’s business, then exclusive dealing raises little competitive concern. The test turns on foreclosure. The greater the seller’s market share, and the longer the duration of the agreement, the stronger the probable impact of the arrangement on foreclosure. The Guidelines also stress the importance of entry barriers, indicating that if entry is easy, single branding is unlikely to pose a problem. The Guidelines contemplate the potentially cumulative anticompetitive effects of single branding arrangements, holding that even if each supplier is covered by the block exemption (safe harbor), withdrawal of the exemption may be appropriate where there are such cumulative effects. The Guidelines state that where the largest market share of a supplier is below 30% and the market share of the five largest suppliers is below 50%, there is unlikely to be a single or cumulative anticompetitive situation. The Guidelines do not, however, indicate any market share threshold for concluding that such a cumulative anticompetitive effect is probable. To summarize, both the United States and EU consider exclusive dealing under the rule of reason. A complainant must establish that there are negative competitive effects from the practice. One of the prominent forms of legal restrictions on distribution contracts in the EU Guidelines is on vertical restraints over Internet distribution. The set of regulations is complex, distinguishing “passive sales,” which are transactions in which consumers’ search effort plays an important role, from “active sales,” which are transactions in which a distributor expends effort to reach consumers in a specific territory or a specific 45 

Ibid. at 329.

Vertical Restraints across Jurisdictions   381

customer group. A distributor website is an example of passive sales, and territory-based banners on third-party websites are a form of active sales in the territory in which the banners are shown. With some exceptions, vertical restraints on passive online sales by a distributor are a hard core restriction, illegal without a market-share-based safe harbor. Active sales by a distributor in support of a (legal) distribution system with exclusive territories or customers is allowed.

15.5.  Conclusion: Where the Law and Economics Diverge In the absence of evidence that vertical restraints are being used to aid collusion or to aid in the exclusion of firms from a market, the restraints should be legal. A manufacturer adopting vertical restraints on intrabrand competition must be doing so for a reason: the restraint on price competition is a cost to the manufacturer since higher retail prices reduce demand, and this cost would not be incurred without a benefit. Many theories are available as to why retailers would not provide enough sales effort (defined broadly) under simple price contracts—and why vertical restraints can resolve the incentive problems. We showed that vertical restraints can be explained by a difference between the marginal rates of substitution over price versus nonprice demand-enhancing activities. Many different specific sources of incentives for vertical restraints fall within this general framework. The welfare effect of vertical restraints is sometimes discussed in terms of a trade-off between reduced intrabrand competition and greater interbrand competition—or as an instrument by which an upstream firm can better compete.46 Certainly in European law dominance on the part of a manufacturer creates suspicion about vertical restraints, and such dominance appears to be a requirement in the US approach to per se illegal tying, but many academic contributions also point to the trade-off, or the importance of upstream competition, in assessing vertical restraints. We disagree. Even in the case of a pure monopolist upstream, without even the threat of competition, there is no presumption that trading off higher prices in exchange for demand-enhancing activities is harmful to welfare. For this reason, we would reject the approach of the EU Guidelines that describes the reduction of intrabrand competition as a generally harmful competitive effect that may result from vertical restraints.47 This approach leads the Guidelines astray in a number of specific respects, perhaps most prominently in restricting RPM

46  Rey and Vergé (2008) write, “Even when vertical restraints eliminate intra-brand competition, if there is sufficient competition from other structures [supply chains] this will not decrease economic welfare since the structure will be unable to exercise market power.” 47  Guidelines, supra at para. 100.

382   Edward M. Iacobucci and Ralph A. Winter because the direct impact of the practice is to raise price. This may be an impact of RPM in a case, but this says nothing about the effect of RPM on horizontal competition. The law currently places too much emphasis on free-riding as an explanation of vertical restraints. For example, in his dissent in the important Leegin case, Justice Breyer of the US Supreme Court writes: Petitioner and some amici have also presented us with newer studies that show that resale price maintenance sometimes brings consumer benefits. Overstreet 119–129 (describing numerous case studies). But the proponents of a per se rule have always conceded as much. What is remarkable about the majority’s arguments is that nothing in this respect is new. . . . The one arguable exception consists of the majority’s claim that, even absent free riding, resale price maintenance may be the most efficient way to expand the manufacturer’s market share by inducing the retailer’s performance and allowing it to use its own initiative and experience in providing valuable services . . . Ante, at 12. I cannot count this as an exception, however, because I do not understand how, in the absence of free-riding (and assuming competitiveness), an established producer would need resale price maintenance. Why, on these assumptions, would a dealer not expand its market share as best that dealer sees fit, obtaining appropriate payment from consumers in the process? There may be an answer to this question. But I have not seen it. (551 U.S. 15 (2007) Breyer, J., dissenting [emphasis added])

The European Commission Guidelines on Vertical Restraints also place undue emphasis on free-riding as an evidentiary requirement for the proposition that resale price maintenance is efficient for established firms.48 In explaining why the practice may be used to induce retailers to provide additional presales service, the Guidelines describe the traditional free-riding argument (involving consumers’ obtaining services at one outlet and then purchasing from a low price outlet), and then state (at paragraph 225): The parties will have to convincingly demonstrate that the RPM agreement can be expected to not only provide the means but also the incentive to overcome possible free riding between retailers on these services and that the pre-sales services overall benefit consumers as part of the demonstration that all the conditions of Article 101(3) are fulfilled.

Like Justice Breyer’s opinion in Leegin, the apparent theory is that the efficiency role of resale price maintenance is limited to the free-riding argument. Our position is that manufacturers use vertical restraints to readjust the mix of price and nonprice competition among retail distributors of their products for a host of reasons. The traditional free-riding theory is only one, narrow explanation. For example, retail distributors design their strategies to attract consumers not just away from other products, but also 48  The Guidelines also recognize limited roles for resale price maintenance for new products and for short-term price promotions.

Vertical Restraints across Jurisdictions   383

away from other retailers. Designing strategies for the latter purpose is wasteful for the supply chain as a whole, and is likely to involve low prices and less sales effort. A manufacturer can correct the distortion (in terms of achieving maximum profits) with vertical restraints. Our view that competition policy should examine only horizontal effects on competition across supply chains is not based on the assumption that the adjustment of price and nonprice competition such as service quality or promotion by a manufacturer will necessarily increase consumer surplus or total surplus. Our position is that it may; that isolating the evidentiary requirements for identifying whether it does or not is impractical or impossible; and that the burden of proof in antitrust restrictions contracts should lie on the side of intervention. We do not regulate a manufacturers’ mix of prices or promotion when these decisions are made directly, for example, in a vertically integrated firm. Nor should we regulate these decisions when they are implemented indirectly through vertical restraints.49 Turning to specific restraints, in the United States, following Leegin, resale price maintenance is no longer per se illegal, but the burden of proof on the firm adopting the practice to establish various conditions is misguided. A fortiori, the continuing approach of the EU to treat price maintenance as a hard core vertical restriction that is subject to a presumption of illegality is also a mistake. While both the United States and EU are relatively permissive of exclusive territories, the EU’s insistence that passive extraterritorial sales, including Internet sales, be permitted creates the kind of externalities that motivate exclusivity in the first place. This will be especially true as Internet sales grow in importance over time. It is not fair to judge the European approach to exclusive territories based only on economics given the political motive that influences the law in this area: the EU is concerned about economic integration as a political matter, and may sacrifice economic gains in order to promote it. But the EU approach does sacrifice efficiency. That said, the incentive problems that exclusive territories address are most worrisome for upstream suppliers where one downstream distributor actively pursues another distributor’s customers. Restrictions on active, out-of-territory sales are permitted in Europe, so European law does permit an especially beneficial version of exclusive territories. This is not a justification for Europe’s insistence on passive sales across territories, but is at least a positive feature of the Guidelines. Vertical restraints on online passive sales by a distributor, including restraints on the proportion of sales online, are prohibited under the EU Guidelines. This is another clear case where competition law in Europe is at odds with economics. Vertical restraints against excessive online sales are not part of a scheme to lessen horizontal competition among manufacturers, but can instead be simply explained as a mechanism to correct distortions in distribution systems in which online sellers can free ride on the investments in product image and other point-of-sale demand enhancement such as information or sampling. 49  This is not to say that resale price maintenance is always efficient. It may be used to suppress horizontal competition, as we discuss below.

384   Edward M. Iacobucci and Ralph A. Winter The areas of exclusive contracts and tying present a more difficult challenge. While these restraints will ordinarily benign, they will on occasion allow dominant firms to exclude rivals. Both the EU and United States appropriately require dominance before challenging these practices, but dominance is merely necessary and far from sufficient to conclude that the restraints are likely to lessen competition. While it is qualified in important respects, we do not believe that even a nominally per se approach to tying in the United States is appropriate. Rather, each case should be examined on its merits, and only in clear cases should such a restraint be regarded as anticompetitive. Because of the difficulty in determining the economic impact of exclusivity and tying, we are also concerned about the relatively expansive approach of the EU to foreclosure. In the Microsoft media player case, for example, the EU was seemingly content to observe that tying a media player to a dominant operating system would give Microsoft a clear advantage in the media player market. What the EU did not clearly analyze, however, was how Microsoft would gain economically from doing so: if Microsoft was a monopolist in operating systems, why was it not content to reap its profits in that market? How did tying the media player increase Microsoft’s profits? How did tying the media player hurt consumers? In our view questions like these should be at the forefront of investigations into exclusivity and tying arrangements, with the burden on those objecting to the restraints. Yet under the modified per se approach to tying in the United States, and the expansive approach to foreclosure in the EU, these questions are relegated to a minor role. The EU and United States provide a useful contrast for the exposition of different approaches to vertical restraints. They are also the most important jurisdictions in practice, with many multinational distribution chains likely to be affected by American and/ or European law. But of course many distribution chains will be narrower in scope and confined to a single jurisdiction governed by a different approach to vertical restraints. Given the absence of international consensus on optimal policy, even on such matters as hard core price-fixing cartels (for example, should they be criminalized?), it is not surprising that there is considerable worldwide variation in the approach to vertical restraints. Canada, for example, is a jurisdiction that has tended to move in recent years in the direction of the American, more permissive approach to vertical restraints. Its statute, which provides a much more complete code of antitrust rules on particular practices than that in either the EU or United States, was amended in 2009 to move the law from treating RPM as a per se illegal criminal offence to requiring that the practice have an “adverse effect on competition” to be subject to civil remedies, not criminal. Tying, exclusive territories, and exclusive dealing are also subject to a competition test. In each case, the practice must substantially lessen competition to be subject to an order. But this liberalizing trend is not true across different matters and different jurisdictions. For example, many significant jurisdictions, such as Germany and Australia, continue to treat RPM as per se illegal. Given that rules such as the per se illegality of RPM fail to account for the broad range of efficiency motivations that exist for many vertical restraints, and moreover are often based on an ill-conceived concern for intrabrand competition, the law on vertical restraints in many jurisdictions is due for reform.

Vertical Restraints across Jurisdictions   385

References Aghion, Philippe, and Patrick Bolton. 1987. Contracts as a Barrier to Entry. American Economic Review 77: 388–401. Andrews, P. W. S., and Frank A. Friday. 1960. Fair Trade: Resale Price Maintenance Re-examined. London: Macmillan. Bork, Robert. 1978. The Antitrust Paradox. New York: Basic Books. Buettner, Thomas, Andrea Coscelli, Thibauld Vergé, and Ralph A.  Winter. 2009a. Selective Distribution by Luxury Goods Suppliers. European Competition Journal 5(1): 201–26. Buettner, Thomas, Andrea Coscelli, Thibauld Vergé, and Ralph A.  Winter. 2009b. Selective Distribution by Luxury Goods Suppliers: A Response to Kinsella et al. European Competition Journal 5(2): 613–21. Carlton, Dennis W., and Ken Heyer. 2008. Appropriate Antitrust Policy towards Single Firm Conduct. US Department of Justice Working Paper, Economic Analysis Group Discussion Paper No. EAG 08-2. Cooper, James, Luke Froeb, Daniel O’Brien, and Michael Vita. 2005a. Vertical Restrictions and Antitrust Policy: What about the Evidence? Competition Policy International 1(2): 45–63. Cooper, J., Luke Froeb, Daniel O’Brien, and Michael Vita. 2005b. Vertical Restraints as a Problem of Inference. International Journal of Industrial Organization 23: 639–64. Cournot, A. [1838] 1971. Mathematical Principles of the Theory of Wealth. New York: Augustus M. Kelley. Deneckere, R., H. Marvel, and J. Peck. 1996. Demand Uncertainty, Inventories, and Resale Price Maintenance. Quarterly Journal of Economics 109: 885–913. Deneckere, R., H. Marvel, and J. Peck. 1997. Demand Uncertainty and Price Maintenance:  Markdowns as Destructive Competition. American Economic Review 87(4): 619–41. Dorfman, R., and P. O.  Steiner. 1954. Optimal Advertising and Optimal Quality. American Economic Review 44: 826–36. Ippolito, P. 1988. Resale Price Maintenance: Economic Evidence from Litigation. Journal of Law and Economics 34: 263–94. Ippolito, P., and T. Overstreet. 1996. Resale Price Maintenance:  An Economic Assessment of the Federal Trade Commission’s Case against the Corning Glass Works. Journal of Law and Economics 39: 285–328. Jullien, Bruno, and Patrick Rey. 2007. Resale Price Maintenance and Collusion. Rand Journal of Economics 38: 983−1001. Klein, Benjamin. 2009. Competitive Resale Price Maintenance in the Absence of Free-Riding. Antitrust Law Journal 76(2): 431–81. Klein, Benjamin, and Kevin Murphy. 1988. Vertical Restraints as Contract Enforcement Mechanisms. Journal of Law and Economics 31: 265–98. Lafontaine, F., and M. Slade. 2005. Empirical Assessment of Exclusive Contracts. In Paolo Buccirossi, Handbook of Antitrust Economics. Cambridge, MA: MIT Press. Marvel, Howard P. 1982. Exclusive Dealing. Journal of Law and Economics 25: 1−25. Marvel, Howard P. and Stephen McCafferty. 1984. Resale Price Maintenance and Quality Certification. Rand Journal of Economics 15: 340–59. Mathewson, G. F., and R. A. Winter. 1984. An Economic Theory of Vertical Restraints. Rand Journal of Economics 15: 27–38.

386   Edward M. Iacobucci and Ralph A. Winter O’Brien, Daniel P. 2008. The Antitrust Treatment of Vertical Restraints:  Beyond the Possibility Theorems. In Arvid Fredenberg, The Pros and Cons of Vertical Restraints, Stockholm: Konkurrensverket, Swedish Competition Authority, 40–101. Overstreet, T. 1983. Resale Price Maintenance:  Economic Theories and Empirical Evidence. Washington, DC: Federal Trade Commission. Paldor, Ittai. 2008. RPM as an Exclusionary Device. Unpublished. Rasmusen, Eric B., J. Mark Ramseyer, and John Shepard Wiley Jr. 1991. Naked Exclusion. American Economic Review 81: 1137–45. Rey, Patrick, and Joseph Stiglitz. 1988. Vertical Restraints and Producers’ Competition. European Economic Review 32(2–3): 561–68. Rey, Patrick, and Thibaud Vergé. 2008. Economics of Vertical Restraints. In Paolo Buccirossi, Handbook of Antitrust Economics, 353–90. Cambridge, MA: MIT Press. Salop, S.  C., and D. T.  Scheffman. 1983. Raising Rivals’ Costs. American Economic Review 73: 267–71. Scherer, F. M., and D. Ross. 1990. Industrial Market Structure and Economic Performance. 3rd ed. Chicago: Rand McNally. Segal, Ilya, and Michael Whinston. 2000. Naked Exclusion: Comment. American Economic Review 90: 296−309. Spengler, J. 1950. Vertical Integration and Antitrust Policy. Journal of Political Economy 53: 347–52. Telser, L. 1960. Why Should Manufacturers Want Fair Trade? Journal of Law and Economics 3: 86–105. U.S. Department of Justice 2008, Competition and Monopoly; Single Firm Conduct under Section 2 of the Sherman Act. Whinston, Michael D. 1990. Tying, Foreclosure, and Exclusion. American Economic Review 80(4): 837–59. Williamson, Oliver. 1975. Markets and Hierarchies:  Analysis and Antitrust Implications. New York: The Free Press. Winter, R. A. 1993. Vertical Restraints and Price versus Nonprice Competition. Quarterly Journal of Economics 108: 61–76. Zerbe, Richard. 1969. The American Sugar Refinery Company, 1887–1914: The Story of a Monopoly. Journal of Law and Economics 12(2): 339–75.

CHAPTER 16

F R A N C H I SI N G A N D E XC LU SI V E D I S T R I BU T ION Adaptation and Antitrust FRANCINE LAFONTAINE AND MARGARET E. SLADE

16.1. Introduction Vertical restraints are contract clauses that one link in a vertical chain imposes on another. Most often vertical restraints arise in the context of distribution and in retail settings, with the upstream firm or manufacturer restricting its downstream distributors or retailers’ choices. For example, a manufacturer might limit its retailer’s product line or geographic market, or set the retail price. In this chapter, we examine vertical restraints in franchising and distribution agreements. First, we give a brief outline of the history of franchising and present some data on franchising in the United States as of 2007, the year of the most recent economic census. We then discuss the extent to which franchise relationships rely on specific types of vertical restraints and the evidence regarding their efficiency or lack thereof. Next, we describe how changes in the antitrust treatment of some practices are affecting franchise contracts and franchise relationships today. Although we pay particular attention to the United States, we also discuss the evolution of antitrust rules as they apply to franchise contracts in the European Union.1 Finally, we conclude with some comments on the contractual nature of distribution and franchise relationships, and the need to reach a 1  For an international comparison of the legal treatment of franchising and the interface with antitrust more generally, see WIPO (2011). For a general treatment of vertical restraints and antitrust, see Iacobucci and Winter, c­ hapter 15 in this volume.

388   Francine Lafontaine and Margaret E. Slade consensus concerning the role of, and harmonize the treatment of, restrictive contract clauses in distribution and retailing.

16.2.  Franchising: History and Extent in the United States Today Franchising is a form of business organization in which an upstream firm, the franchisor, enters into contractual relationships with downstream firms, the franchisees, who operate under the franchisor’s trade name and usually with the franchisor’s guidance. In some sense, franchising is as old as commerce itself. However, some attribute the origin of modern franchising to early nineteenth-century German brewers, who granted pubs and taverns the right to sell their products and use their name. According to Dicke (1992), franchising in the United States can be traced back to the mid-1800s when the McCormick Harvesting and Singer Sewing Machine Companies granted exclusive territories to sales agents in exchange for a franchise fee.2 Initially, like other firms using agents at the time, these two companies imposed few restrictions on, and required limited qualifications from, their agents. Over time, however, both found that they needed more control if they were to protect their reputations and brands. The McCormick Harvesting Machine Company proceeded to establish company-owned branch houses throughout the United States and Canada, branch houses that were given oversight responsibilities for the sales agents in their territories. This allowed McCormick to systematize procedures and communications with its agents, thereby transforming them into what we would now call “dealers.” The Singer Sewing Company addressed the need for control by converting many of the independent agencies into company outlets. It also devised a series of recommendations for the remaining agents as to how the offices should be run and, for the first time, required detailed financial reporting from them. The contracts and methods of control that Singer developed at the time are often described as the forerunners of the modern franchise agreement. Coca-Cola began developing its franchise bottling system in 1899, and car manufacturers and oil companies followed suit soon after.3 The specific form of organization used by these firms, whereby the franchisor is a manufacturer that sells finished or semifinished products to its dealers/franchisees, has since come to be known as “traditional” 2  See Dicke (1992) for more on the history of franchising in the United States, including a detailed account of its evolution at these two companies. See also Marx (1985) on the development of franchising in automobile retailing in the United States. 3  Starting with the arrangements between German brewers and retailers, franchising has always been a common method of selling beer in Europe. For a discussion of exclusivity in UK beer sales, see Slade (1998 and 2011). This arrangement, however, is illegal in most states in the United States. Nonetheless, contractual relationships between beer producers and their distributors are often exclusive and have several of the characteristics of franchise agreements.

Franchising and Exclusive Distribution   389

franchising. The franchisees in these systems sell their franchisor’s products to consumers or to other firms in the distribution chain. The franchisor’s profit from its dealer network flows from the markups that it earns on the products sold to franchisees. In contrast to business-format franchising, as described below, traditional franchisees do not pay running royalties. Martha Matilda Harper is said to have created the first true business-format franchise (Plitt 2000). And indeed, this entrepreneur grew her network of Harper Beauty Shops from the early 1890s onward using a business model that included all the components of a business format as described by the US Department of Commerce today. To be sure, her network of hair salons was developed around the distribution of a product, which she manufactured in Rochester, New York, similar to traditional franchising. But sales of the tonic were only a small component of the business concept. To operate a Harper salon, a woman—all Harper franchisees were women—had to go through rigorous training and agree to abide by a set of rules that guaranteed customers the same quality of service at any Harper salon. Although there were more than 500 Harper salons in the United States, Canada, and Europe by the mid-1920s, Matilda Harper’s business did not survive her, and ultimately did not leave a visible mark.4 Many business-format franchisors, such as A&W and Howard Johnson restaurants, established their franchise networks in the 1920s and 1930s. However, it was not until the 1950s, with the advent of chains such as Burger King and McDonald’s and the economic boom of the postwar era, that business-format franchising blossomed in the United States and Canada and, gradually, throughout much of the rest of the world. Business-format franchising today encompasses a large number of firms that provide a wide array of goods and services. In exchange for the business format, or trademark and way of doing business, franchisees typically pay a relatively small lump-sum fee at the beginning of the contract period as well as running royalties that are usually a fixed percentage of the franchisee’s revenues. They also often contribute an additional fraction of their revenues to a separate advertising fund. Presumably, the advertising carried out with these funds benefits all franchisees as well as the franchisor. Although often made, the distinction between traditional and business-format franchising is somewhat arbitrary. Dnes (1992) and Klein (1995), for example, argue that there is little economic difference between the two in terms of the type of agreements that franchisors rely on, the type of support that they provide, and the control that they exert. Moreover, when constructing theoretical models, researchers often make no distinction between the two types of franchising, and their models apply equally well to both. On the other hand, from a descriptive standpoint the distinction can be important because many countries outside the United States do not include traditional franchising in their data on franchising activities. The distinction also can be relevant from an antitrust perspective, particularly in relation to tying, as discussed below.

4 

However, one can still find the building that housed the Harper Laboratory in Rochester, New York.

390   Francine Lafontaine and Margaret E. Slade Table 16.1  Traditional versus Business-Format Franchising Establishments of Franchising firms, Traditional total franchising firms No. of Establishments Sales ($000) Employment Payroll ($000) Employment per establishment Sales per establishment Payroll per establishment

453,326 1,288,171 7,881,975 153,680 17.39 2,842 339.00

66,223 867,559 1,480,265 56,144 22.35 13,100 847.80

% 15 67 19 37

Business-format franchising firms % 387,103 420,612 6,401,710 97,535 16.54 1,087 251.96

85 33 81 63

Source: 2007 Economic Census, Core Business Statistics Series: Summary Statistics by Franchise Status for the United States and authors’ calculations.

Historically, the US government did not collect data on franchise activities in its economic censuses. In 1971, the US Department of Commerce began publishing reports, “Franchising in the Economy,” that provided data on franchising, which were obtained via franchisor questionnaires. Unfortunately, this initiative was terminated in 1986. With the 2007 Economic Census, the Census Bureau began a new effort to collect data on franchising, resulting in information on the franchising status of more than 4.3 million establishments across 295 six-digit NAICS industries. The results of this effort, which were released in the fall of 2010, include data on sales, employment, payroll, and the number of establishments of franchised chains. As most franchised chains rely on dual distribution, operating both company-owned and franchised establishments, the census data also quantify the level of activities of franchised and company-owned establishments separately within each sector. We summarize some of the information generated by this effort in tables 16.1 and 16.2.5 In table 16.1, we give an overview of the extent of franchising by showing the number of employer establishments, and total sales, total employment, and total payroll of the establishments of franchised firms. We also present information about the size of establishments, in terms of average sales, employment, and payroll per establishment. The data are presented separately for traditional and business-format franchising. To give a better sense of the sectors where franchising occurs, and of the extent of dual distribution, table 16.2 shows the number of employer establishments across detailed six-digit NAICS industries for those industries that rely on franchising the most (per these data). The top panel of the table pertains to traditional franchising, whereas the lower panel is for business-format franchising. The data in this table are further broken down between establishments that are owned by franchisees and those that are owned by franchisors. 5 

For more on this, see Kosová and Lafontaine (2011).

Franchising and Exclusive Distribution   391

Table 16.2  US Establishments by Franchise Status and Industry, Six-Digit NAICS Franchisee-owned Franchisor-owned establishments establishments

Establishments

NAICS

Sector

312111 Soft-drink manufacturers 441110 Automobile dealers, new 447110 Gasoline stations with convenience stores 447190 Gasoline stations, no convenience store

Sector total

# in % in % of % of franchised franchised sector sector chains chains Number total Number total 16.10

11

2.33

24,888

100

0

0

34.86

17,194

17.63

16,797

17.23

7,257

34.15

4,811

22.64

2,446

11.51

91,893

47.05

72,357

37.05

19,536

10.00

211,313 124,898 220,089 30,130 48,108 22,585

59.11 13.69 46.95

97,262 20,643 20,655

46.03 9.38 42.94

27,636 9,487 1,930

13.08 4.31 4.01

109,472

22,009

20.10

19,667

17.97

2,342

2.14

50,491

16,721

33.12

15,413

30.53

1,308

2.59

31,919

9,082

28.45

8,593

26.92

489

1.53

53,814 81,632 25,510 13,149

6,569 6,502 5,896 5,776

12.21 7.97 23.11 43.93

5,833 6,179 3,249 375

10.84 7.57 12.74 2.85

736 323 2,647 5,401

1.37 0.40 10.38 41.08

Other business-format 3,295,252 sectors

111,265

3.38

80,721

2.45

30,544

0.93

Total business-format 4,140,749 361,433 franchising sectors Total franchising 4,336,046 453,326

8.73

278,590

6.73

82,843

2.00

350,947 77.42 102,379

22.58

Total traditional franchising sectors 722211 Fast-food restaurants 722110 Full service restaurants 721110 Hotels and motels (except casinos) 531210 Offices of real estate agents and brokers 722213 Snack and nonalcoholic beverage bar 713940 Fitness and recreational sports centers 561720 Janitorial services 812112 Beauty salons 445120 Convenience stores 446130 Optical goods stores

472

87

18.43

76

24,888

24,888

100

97,508

33,991

21,248 195,297

10.45

Source: 2007 Economic Census, Core Business Statistics Series: Summary Statistics by Franchise Status for the United States and authors calculations.

These tables establish a number of important facts about franchising. First, table 16.1 shows that, in 2007, franchising firms operated more than 450,000 employer establishments, or about 10.45% of all employer establishments in these sectors. Also, sales through these amounted to almost 1.3 trillion dollars, or 9.2% of GDP of 14.061 trillion

392   Francine Lafontaine and Margaret E. Slade dollars. These firms, moreover, accounted for 7.9 million jobs in the US economy. By comparison, it should be noted that US manufacturing accounts for a total of 13.4 million jobs. In other words, franchising accounts for more than half as many jobs as does all manufacturing activity in the United States. Clearly, the results of the 2007 census confirm that franchising plays a vital role in the US economy.6 As basically all franchising involves operations under exclusive trademarks, that is, exclusive dealing, and much of it entails the use of other vertical restraints as well, and since such restraints are not limited to this context, as discussed further below, these figures also confirm that much business activity is carried out under some form of vertical restraint in the US economy. Table 16.2 shows that about 23% (or 102,379) of the 453,326 employer establishments of franchised companies are operated directly by franchisors. Put differently, while establishments of US franchisors represent about 10.5% of total employer establishments in the economy, 77% of these are owned by franchisees, while the other 23% are owned and operated directly by franchisors. Except for new automobile sales, where state laws prevent car manufacturers from owning dealerships (Lafontaine and Scott Morton 2010), all of the sectors listed are characterized by dual distribution—both up- and downstream ownership. This is not the result of aggregation across firms that use and do not use franchising—though not shown in this table, most individual franchisors operate both types of establishments. This fact has led a number of authors to explore factors that can explain the variation in the proportion of franchised outlets across franchised firms.7 For our purposes here, however, the important point is that under upstream ownership, there are no vertical restraints: instead, the firm controls its distribution via vertical integration. When considering the legal treatment of vertical restraints, this implies that the alternative of vertical integration is a very viable option for most franchisors. Thus changes in the legal status of vertical restraints almost surely will result in changes in the extent to which these firms rely on vertical integration. In other words, an important concern in considering the rules that govern the use of vertical restraints is that the imposition of strict rules may lead to a reduction in the extent to which firms rely on franchising, and hence in the set of opportunities available to potential franchisees.

6  Like most data in the economic censuses, these data ignore nonemployer establishments. This is an important omission in the context of franchising, and so all these estimates should be viewed as conservative. 7  See Blair and Lafontaine (2005) for data at the individual franchisor level, and Lafontaine and Slade (2007) for a survey of the empirical work on the propensity to franchise. Note that Lafontaine and Shaw (2005) use panel data approaches to show that mature franchisors operate a stable proportion of their outlets corporately, and that those with higher valued brands operate a greater proportion of such outlets. They argue that this is consistent with the franchisor’s desire to protect its investment in the value of the brand from franchisee free riding. They also find, as prior empirical analyses found, that more geographically dispersed chains rely on franchising more, a result consistent with agency-theoretic arguments for franchising.

Franchising and Exclusive Distribution   393

16.3.  Antitrust and Its Role in Franchising and Distribution From an antitrust perspective, the main concern with both traditional and business-format franchising is that they rely extensively on vertical restraints of various forms. The vertical restraints that have received the most attention in the literature include the following:8 • Exclusive dealing—where a manufacturer requires that a retailer sell only her products—is perhaps the most common form of restraint. Indeed, basically all franchising involves some form of exclusive-dealing arrangement. For studies of exclusive dealing in, for example, beer distribution, see Asker (2004), Sass (2005), and Slade (2011). • Exclusive territories—where a manufacturer assures a downstream firm that it will be the exclusive reseller of its brand in a geographic region—often accompany exclusive dealing. This restraint is common in franchising, and also in beer distribution in the United States (see Jordan and Jaffe 1987, Culbertson and Bradford 1991, and Sass and Saurman 1993). • Tying occurs when a manufacturer requires that its customers purchase product B (the tied product) as a condition for obtaining what they really want, namely product A (the tying product). Tying was common, for example, in movie distribution (see Hanssen 2000). • Resale price maintenance (RPM)—where the upstream firm exerts control over the price that the downstream firm can charge—can take the form of setting a specific price or a price floor or ceiling. Franchisors have been known to impose mostly downward pressure on franchisee prices, that is, maximum RPM (see Blair and Lafontaine 2005). Antitrust authorities have been concerned with all of these restraints as applied to franchising and distribution: exclusive dealing because of its potential to foreclose and exclude, exclusive territories due to their potential to create local monopoly power, tying because of the possibility of extending monopoly power from one sector to another, and RPM due to its similarity with horizontal price fixing. Although the intent of antitrust law is to protect competition, a perhaps unintended consequence of that law is that it gives franchisees some incentives to make antitrust claims against their franchisors—where they can request treble damages—even when the fundamental dispute is, in essence, contractual (see, e.g., Joseph 2011, 3). Unfortunately, from a policy point of view, the theoretical welfare effects of vertical restraints are almost always ambiguous. In the remainder of this section, we discuss 8 

See ­chapters 12 to 15 in this volume for more on these restraints.

394   Francine Lafontaine and Margaret E. Slade some evidence regarding the use of each of these restraints in franchising and distribution, and what is known about the effects of that use. In the next section, we explore the extent to which changes in the legal treatment of vertical restraints have affected franchise and distribution contracts.

16.3.1.  Exclusive Dealing Historically, antitrust authorities have treated exclusive dealing more leniently than other forms of vertical restraints.9 Were that not the case, franchising could not have grown so rapidly and would not have assumed such an important position in modern economies. The tendency to treat exclusive dealing leniently is perhaps surprising. Indeed, with respect to vertical restraints, one of the major concerns of regulators has been their potential to foreclose competitors and to raise rivals’ costs, and exclusive dealing clauses have greater potential than other vertical restraints to have such effects. In spite of this concern, exclusive dealing has a long history in franchising. For example, in the early 1920s, Sinclair already required its licensed dealers to sell only Sinclair gasoline from pumps bearing its trademark.10 Exclusive dealing is also used in other contexts, including movie and beer distribution. Heide, Dutta, and Bergen (1998), moreover, document that 46 of the 147 manufacturers they surveyed—all of them in the industrial machinery and equipment or electronic and electric equipment sectors—used exclusive dealing clauses in their contracts with distributors. Given its prevalence, it is not surprising that the reasons for employing exclusivity clauses have been assessed empirically. Heide, Dutta, and Bergen (1998), for example, examined this question via survey data collected from manufacturers. They found that manufacturers who were more concerned that their promotional efforts, training, or general support of distributors might benefit their competitors were more likely to adopt such exclusivity. On the other hand, when it was costly for manufacturers to assess whether their dealers sold other manufacturers’ products (i.e., when monitoring the behavior of dealers was difficult ex post), or when manufacturers perceived that their customers had a preference for multiproduct distribution, they were less likely to rely on exclusive dealing. These results are consistent with the type of efficiency or principal-agent arguments that are proposed in the organization economics literature to explain the use of exclusivity restraints. A number of other empirical studies have attempted to understand the reasons for exclusive dealing by examining the consequences of its usage. Those studies again find considerable support for efficiency-enhancing effects. For example, in the US beer market, Asker (2005) finds that brewers that employ exclusive dealing arrangements have lower cost distributors than their competitors. Similarly, Sass (2005) finds that exclusive distribution is associated with greater brand and market sales, and Chen (2012) finds that exclusive distribution is probably adopted to prevent the erosion of efforts to promote 9 

For a more complete discussion of exclusive dealing, see Marvel, c­ hapter 13 in this volume. Federal Trade Commission v. Sinclair Refining Co., 261 U.S. 463 (1923).

10 

Franchising and Exclusive Distribution   395

individual brands that can occur when distributors handle large numbers of brands. In other words, empirical studies in the United States have uncovered little evidence of anticompetitive harm due to exclusive dealing (see Cooper et al. 2005 and Lafontaine and Slade 2008 for more on this). However, in the context of car retailing in Belgium, Nurski and Verboven (2011) find some evidence that exclusive dealing restrictions may be detrimental to consumers. We briefly revisit this study in our discussion of VRs in the EU in Section 16.3.5.

16.3.2.  Exclusive Territories The 1890 Sherman Act made allocating exclusive territories per se illegal in the United States, and this situation persisted until the Supreme Court’s 1977 decision in Continental T.V. v. GTE Sylvania Inc. Since that time, exclusive territories have been subject to a rule-of-reason standard. Moreover, examples where manufacturers provide exclusive territories to their distributors are easy to find. For example, wholesale beer distributors and retail automobile dealers often operate under exclusive territorial clauses that protect them from entry of sellers of the same brand in their geographic selling areas. As a matter of fact, state laws governing new car sales in almost all US states make it a requirement for manufacturers to demonstrate “need” to establish a new dealership in a dealer’s “Relevant Market Area,” as defined in the statute (rather than the territory the manufacturer might have defined).11 In other words, state laws impose a requirement that antitrust laws deem potentially detrimental to consumers. An early example of the use of territorial restrictions in the United States—exclusivity in the distribution of Sealy mattresses—is the focus of a study by Mueller and Geithman (1991). Sealy introduced territorial restrictions in 1926, and those restrictions were found to be illegal under US antitrust law by the Supreme Court in 1967. However, the first ban proved to be ineffectual, as Sealy found ways to circumvent the law. As a consequence, a private antitrust suit was brought against Sealy and the new territorial restrictions were banned in 1975. Mueller and Geithman conclude that the restraints indeed had anticompetitive effects, as they led to local monopoly power and reduced output.12 From an efficiency perspective, however, exclusive territories can be used to provide some security to distributors and franchisees, or at least clarify how the manufacturer or franchisor’s future plans might affect downstream operations. Table 16.3 shows the frequency with which franchisors in different industries offered exclusive territories to their franchisees according to the International Franchise Association (IFA) and Frandata’s (1998) analysis of franchise disclosure documents. In this table, any form of exclusive territory, described by geography, population, miles, or number of vehicles, is counted as a yes. The table confirms that the majority of franchisors offered some form of territorial protection to their franchisees. Moreover, Azoulay and Shane (2001) found that a contractual guarantee of an exclusive territory significantly increased the likelihood that new franchised chains survive 11  12 

See Lafontaine and Scott Morton (2010) for more on this. However, see Eckard (1994) for a criticism of their methodology.

396   Francine Lafontaine and Margaret E. Slade Table 16.3  Exclusive Territories Sector Automotive Baked goods Building and construction Business services Children products and services Education products and services Fast food Lodging Maintenance services Personnel services Printing Real estate Restaurants Retail: Food Retail: Nonfood Service businesses Sports and recreation Travel Total

Number of Number with franchisors exclusive territories 89 39 70 57 27 21 197 39 77 35 21 39 99 60 130 105 37 14 1156

62 28 61 42 24 20 136 13 49 33 14 26 79 27 101 90 31 8 844

% 70 72 87 74 89 95 69 33 64 94 67 67 80 45 78 86 85 57 73

Source: IFA Educational Foundation and Frandata Corp. (1998), as reproduced in Blair and Lafontaine (2005), c­ hapter 8.

beyond their first few years in business. They interpret this to mean that territorial protection is so important to franchisees that young franchise systems that fail to offer such protection from encroachment are unable to attract franchisees, which leads to their failure. Interestingly, 83.5% of the 170 new franchisors in Azoulay and Shane’s (2001) data offered territorial protection. By contrast, in a 1993 study of the largest 50 restaurant franchisors, the Frandata Corporation found that only 26 of them offered some form of territorial exclusivity.13 This is much lower than the proportion in Azoulay and Shane, or than the 80% reported in table 16.3 for the sit-down restaurant industry, or even the 69% reported for the fast-food industry. It appears then that large franchisors, for whom claims of territorial encroachment are more likely to be an issue, have a lower than average propensity to offer territorial protection. Anecdotal evidence from Love (1986), who mentions how Ray Kroc, the founder of the McDonald’s franchise system, offered exclusive territories to early franchisees, but then reduced the size of the territories over time and eliminated them entirely by 1969, suggests that the tendency of new franchisors to offer this protection while established franchisors do not do so as frequently, is a within-firm effect. 13  Study results mentioned in Encroachment Issues in the Restaurant Industry, 2nd Quarter Franchise Update 14, 1994.

Franchising and Exclusive Distribution   397

While one might expect that concerns over antitrust may be the reason that franchisors eschew the use of exclusive territories as they become large, or more dominant in their market, Azoulay and Shane (2001) mention an alternative motive for eliminating territorial protection as a franchisor becomes better established. They report that the main reason that franchisors give for not offering exclusive territories is the concern that “exclusivity would allow franchisees to hold them up through underdevelopment.” In the words of one franchisor, “If they (franchisees) can’t afford new stores and they don’t operate well, they will slow down our growth if we can’t put someone else in the area” (353). Hence, development can be postponed due to franchisee liquidity constraints or, even worse, because a particular franchisee is not talented or ambitious enough to operate more stores. Such concerns reduce the willingness of franchisors to provide territorial guarantees.14 This problem is, of course, not insurmountable. In particular, grants of exclusive territories could be made contingent on some objective measures of franchisee performance. Such contingency clauses are common in master franchise agreements. Indeed, master franchise contracts not only provide territorial guarantees to the master franchisee, they also typically stipulate a fee to be paid for the rights to develop the franchise in the territory as well as a development schedule within the territory in question. The franchisor then evaluates the franchisee’s performance in terms of number of outlets opened, and the territorial guarantee is predicated on this number reaching specific target values over time. The drawback to this approach is that the targets may be unrealistic, causing the master franchise agreement to fail.15 More generally, it is good practice for franchisors to state explicitly what they consider to be reasonable sales and profit levels per outlet, and for franchisees to know upfront that new outlets will be added when outlet sales in the region go above those levels. These types of safeguards for franchisors and franchisees go much beyond the simple grant of a territory, and are more closely aligned with other types of practices that franchisors currently use to minimize tension over geographic and other forms of expansion. As mentioned briefly above, in addition to federal antitrust laws that govern exclusive territories, state regulations have resulted in territorial protection for new car dealers in the United States.16 All states require that car dealers be licensed in the state, thereby 14  See also Mathewson and Winter (1994) for a model emphasizing the role of exclusive territories in setting the starting point of future renegotiation processes in franchise relationships. 15  See Kalnins (2004) on the frequency of failure of master franchise agreements in international markets, which he attributes to development schedules that are too aggressive. 16  According to Dicke (1992), early dealer contracts at Ford, for example, also involved territorial exclusivity, even before the Sealy contracts. He notes that the standard yearly contracts introduced in 1908 by the company were simple, but covered such matters as pricing, quantity forcing, and territorial protection. Specifically, “a dealer agreed to purchase a set number of cars each month at a discount of 15% to 25% and to sell them only at prices established by Ford” (Dicke 1992, 64). Dealers were also bound to keep an inventory of parts and make repairs on all cars made by Ford, whether they had made the sale or not. “In exchange for meeting the terms of his contract, the dealer received the exclusive right to sell Ford products in a clearly defined area. . . . Ford reserved the right to fine and/or cancel any dealer who engaged in cross-selling, that is selling to customers who lived outside the dealer’s territory, or ‘bootlegging,’ as the wholesaling of cars to unauthorized dealers was called” (65).

398   Francine Lafontaine and Margaret E. Slade preventing manufacturers from retailing cars directly, or through other channels. In addition, most states protect auto dealers against what they refer to as encroachment by requiring that a car manufacturer demonstrate a “need” to establish a new dealership. Combined, these restrictions prevent manufacturers from adjusting dealer networks to match changing demand patterns. They have also been a major impediment to the development of, for example, Internet distribution of new cars. Finally, in almost all states it is illegal for manufacturers to require that franchisees purchase vehicles that they have not ordered, which amounts to a prohibition against quantity forcing. Unfortunately, a dealer who has an exclusive territory might exercise market power in that region, with the result that prices are too high and quantities sold too low from the manufacturer’s point of view. In the absence of regulation, the manufacturer could turn to quantity forcing to lower prices, increase total surplus, and reduce deadweight loss. As should be clear from the description of the state laws however, car manufacturers are not able to address the problem via quantity forcing, opening new dealerships, or selling cars directly to customers. The consequence, according to the few studies that have considered the effect of these state laws, is that cars sell at higher prices than they would absent the laws. For example, Smith (1982) finds that the territorial exclusivity for car dealers has led to higher prices and lower service levels, that is, fewer hours of operation. The Petroleum Marketing Practices Act at the federal level, and state laws governing liquor distribution, also provide territorial protection to dealers. Contrary to the findings for autos, however, in the beer industry Sass and Saurman (1993) conclude that territorial exclusivity, whether mandated or privately chosen, leads to higher demand through its effect on promotion, which is procompetitive or at least not anticompetitive. Most recently, a few authors have assessed the effects, both potential and actual, of the change in policy that resulted in a ban of exclusive territories in new car retailing in the EU. Brenkers and Verboven (2006) find that, if a manufacturer’s capacity to grant exclusive territories and to choose dealers selectively resulted in double margins, the change to a system where manufacturers could no longer grant exclusive territories would greatly benefit car buyers via reduced prices. In contrast, Zanarone (2009), who examined the effect of banning exclusive territories on the other clauses found in the franchise contracts of 19 car dealer networks in Italy, found that prior to the ban manufacturers relied on a mix of exclusive territories and quantity floors. After the ban, they introduced standards on verifiable marketing and service inputs, such as floors on advertising and the number of salespeople. He interprets this as evidence that manufacturers used exclusive territories to prevent free riding and induce desired dealer services. Thus, once exclusive territories were prohibited, they switched to other contractual devices to achieve the same goals, presumably at higher cost. In sum, the evidence that relates to the motives for adopting, and the effects of using, exclusive territories is somewhat mixed. Often, other restraints are used along with exclusive territories, and conclusions reached in the literature are sensitive to the set of other restraints used or available to manufacturers. Thus much more work will be needed before a consensus on the effects of particular restraints such as exclusive territories can be reached.

Franchising and Exclusive Distribution   399

16.3.3. Tying Among vertical restraints, tying has one of the most varied legal histories.17 For one thing, unlike exclusive dealing and exclusive territories, tying is still per se illegal under US law.18 However, for it to be so, certain conditions must be met, which seems like a contradiction in terms. In addition to the existence of a tie, which would be sufficient in other per se circumstances, the following conditions must apply: First, the two products or services must be separate, which eliminates tying claims for products such as left and right shoes. Second, the seller must have sufficient market power in the tying good market, which eliminates cases where the buyer has many alternative trading partners in that market. And third, the arrangements must affect a “not insubstantial” amount of commerce, which eliminates arrangements that have little economic importance. In practice, these conditions mean that, with respect to tying, per se illegality is not very different from a rule-of-reason standard. The use of tying contracts to ensure input quality and thereby protect the value of a manufacturer’s or franchisor’s trademark has a long history. It has also been subject to various legal restrictions under US antitrust laws and EU competition policy. A major early test of tying in distribution occurred in the United States in 1917 when the Motion Picture Patents Company, the first great film trust, was sued for tying the sale of movie projectors to the purchase of films. The defendant lost the case, but on the basis of patent, not antitrust law. Nevertheless, this case gave rise to the leverage theory of tying (Gift 2008). Other landmark decisions involved major oil refiners in the United States, all of whom required that lessee dealers operating under their brands sell only their gasoline. Prior to the 1960s, most of them extended that requirement to include tires, batteries, and accessories, allegedly to ensure the quality of all products and services offered at their dealerships. Those requirements, however, were found to be in violation of US antitrust laws.19 As to open supply, that is, the notion that dealers might be allowed to sell gasoline from different sources under their refiner’s trade name, Marvel (1995) notes that “The Lanham Act, governing trademarks, was interpreted to require retailers to keep gasoline of different brands physically separate, stored in separate tanks and dispensed through separate and clearly marked pumps. Few dealers availed themselves of the opportunity to operate multibrand outlets” (213). Tying similarly became an issue in later years for business-format franchising. In its 1971 Siegel v.  Chicken Delight decision, the court held that the requirement that franchisees in a fast-food chain purchase paper products and other inputs from their franchisor was an illegal tie. The court reasoned that in this type of business, the franchisor could ensure product and input quality, and thus consistency, by providing its franchisees with a set of approved suppliers rather than requiring that the franchisee 17 

For a more general discussion of tying, see Hovenkamp and Hovenkamp, ­chapter 14 in this volume. Tying became per se illegal with the International Salt Co. v. United States decision in 1947. 19  See Marvel (1995) for a detailed discussion and references to relevant court cases. 18 

400   Francine Lafontaine and Margaret E. Slade purchase the input from the franchisor (who obtained it in turn from the manufacturer). This decision has been interpreted to mean that a business-format franchisor’s trademark and method of doing business is a tying product that, for antitrust purposes, should be considered separate from food or other items that franchisees are required to purchase from their franchisors. Later, in Krehl v. Baskin-Robbins Ice Cream Company (1982), a franchisee alleged that Baskin-Robbins illegally tied its ice cream to the Baskin-Robbins trademark. Although ice-cream stores are often considered a type of business-format franchise, the court nonetheless noted that “the franchised outlets serve merely as conduits through which the trademarked goods of the franchisor flow to the ultimate consumer.” The court thus found that Baskin-Robbins ice cream was not a separate product from the Baskin-Robbins trademark, and as such there was no tie-in. In other words, as in traditional franchising, the court found that input purchase requirements in the subset of franchised chains that amount to distribution franchises were not tying arrangements. The practical result of this jurisprudence has been that most business-format franchise contracts in the United States today rely on approved suppliers rather than input purchase requirements for purposes of quality control, except in the special cases where the input is proprietary or indistinguishable from the trademark in customer’s minds. This can be seen in table 16.4, which shows the frequency with which franchisors required that franchisees buy some inputs from them in 1988 and 1989, the only years for which information on such requirements is available for a large number of franchisors (see Blair and Lafontaine 2005, ch. 6). The data indicate that franchisors involved in various types of retail activities frequently impose input purchase requirements. Franchisors that provide mostly services, on the other hand, typically do not require that franchisees buy inputs from them. What may seem surprising, however, is the frequency of purchase requirements in the restaurant and fast-food sectors. Michael (2000) examined the use of tying in these sectors, using the franchise disclosure documents of 100 different restaurant and fast-food franchisors. He found that 30 of them required franchisees to purchase some product from the franchisor. Among the 30 firms with purchase requirements, on average the value of required purchases represented about 8.4% of all franchisees’ wholesale purchases. Moreover, consistent with the arguments above, firms with the largest proportions of purchase requirements tended to sell only proprietary products, such as batter for pancake houses or ice cream for family restaurants (e.g., Brigham’s and Howard Johnson respectively). Given that a consequence of the Chicken Delight decision was to rule out one of the “per-unit” type of mechanism by which franchisors could extract franchisee rent, it would be expected to lead to an increase in reliance on royalty rates and advertising fees by those franchisors who are constrained by this ruling. And indeed, the data in Blair and Lafontaine (2005, ch. 3) show that royalty rates are lower for business-format franchisors involved in retail compared to services, which again is consistent with the idea that in distribution franchises, franchisors extract some profits via markups on the products they sell to franchisees in a manner similar to traditional franchising.

Franchising and Exclusive Distribution   401

Table 16.4  Mandatory Purchase Requirements 1988 Sector Automotive Business Contractors Cosmetic Education Fast food Health and fitness Home furnishings Hotels and motels Maintenance Personal services Real estate Recreation Rental Restaurants Retail food Retail other Total

1989

Number of franchisors

Number with requirement

Percent

Number of franchisors

Number with requirement

Percent

56 96 18 14 8 119 20 14 9 45 52 15 11 15 37 22 92 643

13 19 7 3 1 51 11 4 1 12 14 1 4 2 16 13 22 194

23.2 19.8 38.9 21.4 12.5 42.9 55.0 28.6 11.1 26.7 26.9 6.7 36.4 13.3 43.2 59.1 23.9 30.2

54 93 20 17 12 112 12 14 5 47 56 17 11 18 39 22 81 630

14 14 7 4 5 45 6 4 1 16 14 5 5 3 18 12 24 197

25.9 15.1 35.0 23.5 41.7 40.2 50.0 28.6 20.0 34.0 25.0 29.4 45.5 16.7 46.2 54.5 29.6 31.3

Source: Blair and Lafontaine (2005), c­ hapter 6.

Like the separate products aspect of tying, market definition issues also have proved controversial. In particular, a spate of franchise tying suits that occurred in the 1990s can be traced to a notion of market definition spawned by the Supreme Court’s 1992 decision in Eastman Kodak v. Image Technical Services. This case involved the tie of one aftermarket product (repair services) to another (repair parts). Specifically, for years, Kodak sold its proprietary repair parts to independent service organizations (ISOs) so they could provide maintenance and repair services to owners of Kodak copiers and micrographic equipment. When Kodak changed its policy and refused to sell its repair parts to the ISOs, Kodak became the only source of maintenance and repair services to its end customers. Image Technical Service and 17 other ISOs sued Kodak, alleging that it had illegally tied its service to its repair parts. Kodak argued that since it sold its equipment in competition with other manufacturers, it could not have appreciable economic power in the aftermarket for repair parts (tying good). In essence, Kodak argued that the relevant product market for antitrust analysis is the primary market for photocopying and micrographic equipment, rather than the aftermarket for repair parts. The Supreme Court, however, rejected Kodak’s reasoning and ruled that the relevant market for determining market power is based on “the choices available to Kodak equipment owners.” Accordingly, it held that the

402   Francine Lafontaine and Margaret E. Slade relevant market was the market for replacement parts for Kodak equipment. Moreover, since the repair parts were proprietary, Kodak was found to have a market share of virtually 100%. The Supreme Court concluded that since repair parts for Kodak equipment are unique (i.e., neither IBM nor Xerox repair parts will fit a Kodak copier), the owners of Kodak copiers were now “locked in” to Kodak services if they wanted to keep their copiers operational. The only alternative was to switch to another brand of copier. But the Court reasoned that the high costs associated with switching prevented this option from being economically viable and provided Kodak with the power to exploit the locked-in owners.20 Franchisees employed the lock-in argument to support their claims of illegal tying in the franchise tying cases of the 1990’s, simultaneously asserting that the market for the tying product was restricted to the market for the defendant’s franchise, and the market for the tied product was the market for supplies to a particular franchise system. The cases also focused on the franchisor’s refusal to approve or allow alternate sources of supply, or restrictive policies of approving alternate suppliers. Among these cases, Queen City Pizza, Inc. v. Domino’s Pizza, Inc. (1996, 1997) was the most influential. Queen City Pizza and 10 other Domino’s Pizza franchisees, along with the International Franchise Advisory Council, filed suit against Domino’s, alleging various antitrust violations, including two tying claims. First, they alleged that Domino’s illegally tied other ingredients and supplies to the purchase of fresh pizza dough, and second, they alleged that Domino’s illegally tied ingredients and supplies to the franchise license itself. Integral to these complaints were the allegations that Domino’s unreasonably withheld its approval of alternative sources of supply and provided incomplete product specifications, thereby eliminating viable competitive sources of supplies to the franchise system. The net effect was that Domino’s franchisees purchased approximately 90% of their ingredients and supplies from the franchisor even though Domino’s typically did not produce these goods; rather, it purchased them from third parties and resold them to its franchisees. Interestingly, the court concluded that Kodak was not applicable in the context of franchise tying suits, and the Queen City litigation ultimately was resolved in favor of the franchisor. The district court noted that “[t]‌he economic power DPI [Domino’s Pizza Inc.] possesses results not from the unique nature of the product or from its market share in the fast food franchise business, but from the franchise agreement.” The Third Circuit affirmed the lower court’s reasoning, and further noted: “If Domino’s Pizza, Inc. acted unreasonably when, under the franchise agreement, it restricted plaintiffs’ ability to purchase supplies from other sources, plaintiffs’ remedy, if any, is not under the antitrust laws.” In other words, the court made a distinction between a franchisor’s pre- and postcontractual market power. This, in our view, is correct.

20  For an analysis of the reasons for vertical restraints by durable good producers, see Blair and Herndon (1996). It should also be noted that switching to an IBM copier merely locks one in to a different repair parts supplier. Thus, it does not solve the customer’s problem.

Franchising and Exclusive Distribution   403

Barkoff (2008) notes that this and other decisions since that time make it unlikely that franchisees today would be successful in a franchise tying case. And indeed, there is some evidence that a few business-format franchisors, including Domino’s, have now put in place extensive distribution systems with the goal of supplying most of the inputs that their franchisees need. Thus the effect of Chicken Delight seems to have dissipated in the last decade or two. Although the move from antitrust to contract law has many advantages, in particular the removal of treble damages, it is not without problems. As Grimes (1999) points out, a franchisor could collude with a designated vendor (named in a sourcing requirement), allowing the vendor to charge a high price in return for a rebate. This could lead the franchisor to search for the highest rebate rather than the best quality/price package.21 This problem, however, is not resolved under the standard remedy of approved supplier programs proposed in, and largely adopted after, the Chicken Delight decision unless franchisors are required to approve several suppliers for each input in each market, a task that is not only demanding, but also unlikely to yield the kind of purchasing economies that franchisees often seek from their affiliation with a franchisor.

16.3.4.  Resale Price Maintenance Resale price maintenance, or RPM, has an even more checkered antitrust history in the United States.22 Indeed, in 1911 the US Supreme Court declared minimum RPM to be per se illegal, and in 1968 the Court extended the minimum RPM per se rule to include maximum RPM. However, in 1997 the Court overruled its 1968 extension to maximum RPM, and in 2007 it overturned its ban on minimum RPM.23 Nevertheless, since many states have their own RPM laws, the legal status of minimum RPM in the United States is far from clear today. Comparing the United States to the EU, the treatment of maximum RPM was more stringent in the United States after the Albrecht v. Herald Company decision in 1968. In its 1997 State Oil Company v. Khan decision, however, the Court returned the antitrust treatment of maximum resale price fixing to a rule-of-reason standard. This decision, which for all intents and purposes, permits maximum resale price restraints, has made the treatment of maximum RPM in the United States closer to that of the EU. On the other hand, the 2007 Supreme Court decision in Leegin Creative Products, Inc. v. PSKS, Inc., also moved minimum resale price maintenance to a rule-of-reason standard. This has widened the gap in policy between the two regions, since minimum RPM remains a hard core offense in the EU. The Leegin case involved a manufacturer of leather goods and accessories (Leegin) and one of its retail distributors (PKSK). Leegin had a policy of refusing to sell its 21 

Note that input prices are not stipulated in franchise contracts. For a more general discussion of RPM, see Klein, ­chapter 12 in this volume. 23  Note that setting a specific price is both minimum and maximum RPM. 22 

404   Francine Lafontaine and Margaret E. Slade products to retailers who priced below its suggested retail prices. When Leegin learned that PKSK was discounting, it suspended sales to that retailer. In response, PKSK sued Leegin, alleging illegal RPM. However, PKSK lost its case. The Leegin decision made it no longer per se illegal for manufacturers and franchisors to impose minimum retail prices on goods that they sell to franchisees for resale. Nevertheless, individual state antitrust laws, as well as the hostility of some state attorneys general towards minimum RPM, can make it difficult for such firms to control downstream prices.24 Furthermore, since business-format franchisors “sell” intellectual property such as trademarks, rather than goods, the legacy of Leegin for that sector is unclear. In general, franchise RPM cases in the United States suggest that franchisors have been constrained more often by rules against maximum resale price maintenance than those against minimum RPM. In other words, when franchisors have tried to control their franchisees’ prices, they have exerted pressure for their franchisees to reduce rather than increase their prices (see, e.g., Blair and Lafontaine 2007, ch. 7). There are a number of possible reasons for franchisors’ concern with lower prices. For one, franchisees with exclusive territories can exercise market power and thus charge prices that are too high from the franchisor’s perspective per the usual double-margins argument. Moreover, maintaining a maximum price level might also be a way to overcome demand externalities. Indeed, when brand loyalty is important, franchisees who charge high prices can damage the brand and thus reduce sales at other establishments as well as their own. Unlike franchisors, franchisees do not capture this externality. Finally, in some circumstances, royalties on sales might lead upstream firms to care more about downstream revenues than downstream profits, which would cause a similar divergence in desired prices. As with the other vertical restraints, the theoretical effects of RPM are ambiguous. Unfortunately, given the historically harsh treatment of such practices, it has also been difficult to examine the effects of RPM, maximum or minimum, empirically. Ippolito (1991), however, is a rare exception. She examined the population of all 203 reported cases of resale price maintenance in the United States between 1975 and 1982, a period during which a fairly broad interpretation of what constitutes RPM was adopted by the courts, and during which, she argues, the courts adhered quite strictly to the per se standard. She shows first that vertical restraints are often used together. Firms simultaneously relied on other vertical restraints in 122 of the RPM cases, most frequently territorial, tying, or customer restrictions (49, 31, and 32 of the cases respectively). Cases of RPM also often involved other charges, in particular horizontal price fixing in 30, and refusal to deal in 40 of the cases. In addition, she finds evidence that a nontrivial portion of RPM cases, namely 65% of all private, and 68% of all public cases in her data, arose in contexts where products can be classified as complex, new, or infrequently purchased, which are the types 24  Even though, in contrast to federal law, the fair trade laws that were adopted by many states after the depression allowed minimum RPM, in recent years some states (e.g., Maryland in 2009) have passed laws that permit a per se challenge to minimum RPM. For a discussion of federal/state conflicts, see Joseph (2010) and Barkoff (2010) and the references therein. The effect of the change in Maryland has been analyzed by Bailey and Leonard (2010) and found to be minimal for video games.

Franchising and Exclusive Distribution   405

of products where the special services theory for RPM is most likely to hold. Another largely overlapping segment of both private and public cases arose in resale contexts where dealers can influence the quality of the final good or the customer’s experience in important ways. Here again, manufacturer controlled pricing can alleviate the fundamental principal-agent problem that efficiency motives and organizational economics emphasize. Yet another set of mostly franchising cases seems well explained by concerns over vertical sales-effort externality problems. She concludes that collusion was not the primary explanation for the RPM practices that were prosecuted during this period. To the best of our knowledge, the more recent, changing stance regarding the role of RPM in franchise and distribution has not led to significant changes in the contracting practices of franchisors and manufacturers per se. However, there are signs that some franchisors are trying to assume more control over the prices that their franchisees set. For example, McDonald’s has removed a statement from its website—a statement that was quoted in Blair and Lafontaine (2005, ch. 7)—to the effect that franchisees are free to choose retail prices. For its part, Burger King recently settled a dispute with its franchisees over the maximum prices of some menu items. We believe that, before Khan, the franchisees in that dispute would have had a stronger hand.

16.3.5.  Contract Adaptation in the EU Despite changes in the antitrust treatment of tying and resale price maintenance over the last couple of decades, large changes in the form and content of franchise or distribution agreements have not occurred in the United States. However, some practices and clauses related to territories and other sources of competition have been expanded and/ or clarified. Similarly, some franchisors have increased the extent to which they provide inputs to their franchisees. During the same period, however, more dramatic changes have occurred in the antitrust treatment of vertical restraints in Europe. After a series of decisions regarding franchise relationships in the late 1980s, the European Commission (EC) adopted a Block Exemption Regulation on franchise agreements, which took effect on February 1, 1989, and remained in effect until May 2000. It was then superseded by a new 2000 Block Exemption Regulation (BER) on Vertical Restraints. Finally, the 2000 BER and Guidelines for Vertical Restraints were replaced by new 2010 versions, which are expected to remain in force until June 2022. This is in sharp contrast to the United States, where there are no comprehensive guidelines on vertical restraints.25 In what follows, we briefly summarize the effects of the EU legal changes on franchising and contrast the treatment of restraints in the EU with that in the United States. The original Block Exemption on franchise agreements gave franchisors a great deal of flexibility to impose or rely on exclusive territories, exclusive dealing, tying, and some 25  After the 1985 Vertical Restraints Guidelines were rescinded by the Clinton administration, no new comprehensive policy document has appeared.

406   Francine Lafontaine and Margaret E. Slade price restraints—but minimum RPM was still not allowed. The BER and Guidelines on Vertical Restraints are in general stricter, but they do provide a safe haven under which small firms are exempt. In practice, because most franchisors fall within the safe heaven, that is, they have less than 30% of the market, as do their franchisees, the effects of the new Guidelines are likely to be felt most extensively in distribution franchises, such as car, beer, and gasoline retailing. On the other hand, the goal of opening competition across markets within the EU led to rules against territorial restrictions that are more stringent than they used to be for franchise firms, and more stringent than in the United States. In addition, restrictions on Internet presence and distribution, which are discussed extensively in the latest Guidelines, are now viewed as more strict in the EU. Finally, the EU Vertical Restraints Regulation includes rules related to post-term noncompete clauses, or noncompete obligations imposed on a buyer after the termination of a contract with a seller. The guidelines require that they be “indispensable to protect know-how transferred by the supplier to the buyer,”26 limited to only the point of sale used during the contract, and no longer than one year. In the United States, there is no consistent noncompete policy, with some states enforcing, while others do not enforce, such restrictions. The automobile industry has received particularly close scrutiny from EU competition authorities. In 1995 the European Commission passed a regulation whose stated goals were to strengthen the competitiveness of dealers and improve market access to spare parts for both producers of competing parts, and independent repairers.27 In 2002 it enacted Regulation 1400/2002, a regulation that, among other things, eliminated exclusive territories and sales targets or quantity forcing. With the exception of certain hard core practices, however, the 2002 automotive block exemption also created safe havens for car manufacturers, distributors, and repair shops. In 2010, a new automotive BER (Regulation No 461/2010) was passed that superseded the 2002 version. This new BER is generally considered to be less restrictive. In particular, as of 2013, new car sales, like business-format franchising contracts, were subject only to the general BER on Vertical Restraints. Repairs, maintenance, and spare parts—the aftermarket—have been subject to the new automotive BER since 2010, and are treated somewhat more harshly by said BER. The distinction between the two markets was made because the primary market was thought to be workably competitive, whereas problems were perceived to remain in the after market. We highlight some of the most important differences between the old and new regimes.28 In the new car market, manufacturers will be able to impose exclusive dealing restrictions that cover up to 80% of a dealer’s requirements instead of the former limit of 30%. In addition, the market share thresholds will be 30% for both up- and downstream firms, whereas formerly only the manufacturer’s share was considered. Also, as 26  European Commission, Guidelines on Vertical Restraints (2010/C130/01) in Official Journal of the European Union No. C 130, p. 1, of May 19, 2010. 27  See European Commission, General Directorate IV—Competition, Distribution of Motor Vehicles (Regulation (EC) No. 1475/95 published in the Official Journal, L 145 of 29 June 1995), Explanatory Brochure. 28  For more on the new block exemption for autos, see Zuehlke and De Stefano (2010).

Franchising and Exclusive Distribution   407

long as the market share thresholds are met, it will be possible to impose clauses that prevent dealers from opening new dealerships. Exclusive territorial clauses are still prohibited. In the aftermarket, authorized repair networks are unapt to benefit from the block exemption because markets are treated as brand specific, and thus market shares more easily exceed the thresholds. Moreover, technical information must now be made available to independent repairers; manufacturers are not allowed to use warranties to prevent owners from using independent repair service providers; and spare parts must be freely available to all service facilities. Not surprisingly, given these changes in policy, the automobile sector also has been the subject of empirical studies. It should come as no surprise, for example, that franchisors and dealers have adapted their behavior in response to changes in their environment. Of particular interest, we note again that public policy towards vertical restraints in franchise contracts can have unforeseen consequences. This is true because franchisors have many tools at their disposal to control franchisees, and so banning one practice can lead to the adoption of others. Thus Zanarone’s (2009) study of how some franchisors adapted to changes in the 2002 EU law towards vertical restraints shows that, once exclusive territories became illegal, the number of car manufacturers that imposed price ceilings, required dealers to abide by a variety of explicit standards, and required dealers to contribute to an advertising fund that the manufacturer controlled, went up significantly. He explains the latter two changes as direct responses to reduced dealer incentives to advertise and provide presales services. As for price ceilings, he argues that they may have become necessary to prevent dealers from circumventing quantity floors— something he suggests they could do by selling aggressively outside their territories while maintaining supranormal prices in their own, perhaps isolated, markets. Nurski and Verboven (2011) studied the likely effect of the change in regulation in the new car market in Belgium. They found that a ban on exclusive dealing would benefit consumers in this market in that it would favor smaller entrants and provide buyers with increased spatial availability. They therefore concluded that the European Commission’s 2010 decision to facilitate exclusive dealing might not have been warranted. Unfortunately, their analyses abstract away from the type of dealer incentive and cost issues that efficiency arguments for exclusive dealing typically focus on. Finally, although the EU controls competition policy that affects more than one member state, individual member states deal with those practices that affect competition within national boundaries. In that context, the UK Office of Fair Trading (OFT) recently undertook a major review of the newspaper and magazine distribution industry. In 2009, however, it announced that it would not refer the sector to the UK Competition Commission. In other words, it did not find sufficient evidence of competitive harm to pursue a case. A key reason for this was that developments benefiting consumers had come about from the review and self-assessment of distribution agreements undertaken by members of the industry. However, the sector was not given a clean bill of health, and the OFT did not rule out the possibility of further review. The OFT ruling is consistent with the empirical findings of Ferrari and Verboven (2011) concerning Belgian magazine distribution. In particular, those authors find evidence

408   Francine Lafontaine and Margaret E. Slade suggesting that a government ban on restrictive licensing or on uniform wholesale prices would have little effect on upstream profits. They conclude that the rationale for those practices is unlikely to be anticompetitive, and should be sought elsewhere.

16.4. Conclusion The extent to which federal antitrust law in the United States is binding on the nature of franchise contracts is surely less today than it was in the 1970s and 1980s. To illustrate, Barkoff (2008) notes that in the mid-1970s: The need for the franchise lawyer to be educated in matters of antitrust law had been exacerbated by the antitrust decisions in the 1960’s . . . today one would not be wrong to describe antitrust law as being almost irrelevant with respect to the franchise community. A slight exaggeration? Yes, but not that far off. (Barkoff 2008, 1)

In particular, with the 1997 State Oil v. Kahn decision, maximum RPM—the type of price restraint that franchisors have tended to impose—has been returned to the rule of reason. Since, in a competitive environment, retail price is no higher under maximum RPM than it would be otherwise,29 most economists believe that this restraint should not be per se unlawful. Similarly, in the years since the 1971 Chicken Delight tying decision, there has been a clear trend away from the logic in that case. In particular, in one important subsequent decision, the Queen City Pizza v. Domino’s Pizza Inc., which was specific to franchising, the US Court of Appeals for the Third Circuit rejected the logic proposed in the then recent Kodak case. Instead it ruled that as long as the input purchase requirements imposed by a franchisor were explicit, described in its franchise contract, such requirements would not be considered a violation of antitrust laws. While US federal antitrust regulation seems to have much less impact on franchise contracts today, state antitrust laws may still be relied upon in some contexts. Furthermore, state laws governing the distribution of goods such as cars, gasoline, and beer continue to affect the types of vertical restraints that can, or must, be applied to these relationships. The conflict between federal and numerous state antitrust laws, as well as cross-state variation in the standards of enforcement, are problems that continue to inhibit the rational development of franchising in the United States. This is the case because many franchisors operate nationally, and are thus subject to regulation by numerous jurisdictions. Given the division of power in the United States, we are not optimistic that federal/state harmonization will occur in the near future.30 29 

In an oligopoly, however, maximum RPM can actually raise retail prices (see Perry and Besanko 1991). Of course, many US franchisors also operate internationally, and harmonization across countries is even less likely. 30 

Franchising and Exclusive Distribution   409

In the EU, meanwhile, the new block exemption and vertical restraints guidelines put in place in 2000 and then in 2010 have modified the treatment of franchise contracts, especially clarifying issues surrounding the use of exclusive territories and Internet distribution for firms operating in the Union. But here again, individual country antitrust laws can affect what can be included in distribution and franchise contracts, or even the existence of certain types of contracts,31 and how these relationships work. Experimentation is part and parcel of the evolution of any mode of organization, and franchising is no exception. As a business organization, franchising has evolved, and will continue to evolve. Presumably, what works well is retained and refined and what does not work well is discarded in an evolutionary manner. The changing legal environment has been one of the factors leading to changes and refinements. For example, the strategy initially adopted by Chicken Delight, which was to earn income by requiring franchisees to purchase inputs exclusively from the franchisor instead of relying solely on a sales-based royalty or fixed fee, was found in violation of US antitrust laws. As a result of this decision, business-format franchisors moved away from input purchase requirements toward approved supplier clauses and adopted approaches where the franchisor is available as a potential, but not a required, source of supplies. This same decision also implied that business-format franchisors would move toward franchise fees and royalties on sales as their main sources of income. However, in recent years, the increased tolerance of antitrust authorities towards requirements stipulated in franchise contracts has led at least some franchisors to increase the role they play in the sourcing of inputs for their franchisees. Similarly, disputes over territorial rights in the quick-service restaurant sector and in hotel chains in the 1990s led many of the firms in those industries to include much more detailed territorial definitions in their contracts. Franchisors also have developed more systematic review processes for new sites and adopted policies for allocating new units to owners of nearby outlets, thereby resolving at least part of the conflict. In industries where encroachment through alternative channels (e.g., Internet sales) has been more problematic, a number of franchisors have worked on ways to channel part of the sales or profits to local franchisees, while others have opted to stay out of the alternative channels. In all cases, franchise contracts, and hence franchisor-franchisee relationships, have been modified as the institutional and legal environments have evolved. This process is a slow one, but we believe that the trend towards treating vertical issues within the context of contract rather than antitrust law is here to stay.32 We view this as a positive development, one that allows disputes between franchisors and franchisees, that is, parties to a contractual relationship, to be brought forth and decided in the United States by reference to actual, rather than treble, damages. Nevertheless it is not a panacea as some postcontract issues may still raise legitimate antitrust concerns (see also Barkoff 2010). 31  See Slade (1998) for an analysis of a UK decision that limited dual distribution in beer retailing, a common practice in Europe that is illegal under post-Prohibition state laws in the United States. 32  In some sense the treatment of vertical restraints has come full circle: for example, the first tying case was decided under patent, not antitrust law. See Hovenkamp and Hovenkamp (2010).

410   Francine Lafontaine and Margaret E. Slade

Acknowledgments This chapter borrows from some of our joint work, especially Lafontaine and Slade (2008 and 2013), as well as from Blair and Lafontaine (2005).

References Asker, John. 2005. Diagnosing Foreclosure Due to Exclusive Dealing. Department of Economics, Harvard University. http://www.johnasker.com/Paper1_ExclusiveForeclosure.pdf Azoulay, Pierre, and Scott Shane. 2001. Entrepreneurs, Contracts, and the Failure of Young Firms. Management Science 47: 337–58. Bailey, Elizabeth M., and G. K. Leonard. 2010. Minimum Resale Price Maintenance:  Some Empirical Evidence from Maryland. B.E. Journal of Economic Analysis and Policy: Contributions 10: Article 17. Barkoff, Rupert M. 2008. Antitrust Law Becoming Irrelevant to Franchise Lawyers? New York Law Journal, July 28. Barkoff, Rupert M. 2010. Antitrust Law Becoming Irrelevant to Franchise Lawyers? A Second Look. New York Law Journal, June 30, 243. Blair, Roger D., and Jill Boylston Herndon. 1996. Restraints of Trade by Durable Good Producers. Review of Industrial Organization 11: 339–353. Blair Roger D., and Francine Lafontaine. 2005. The Economics of Franchising. Cambridge: Cambridge University Press. Brenkers, Randy, and Frank Verboven. 2006. Liberalizing a Distribution System: The European Car Market. Journal of the European Economic Association 4: 216–51. Chen, Chia-Wen. 2012. Does Exclusive Dealing Matter? Evidence from Distribution Contract Changes in the U.S. Beer Industry. National University of Taipei. http://web.ntpu.edu. tw/~cwzchen/AB_InBev.pdf Cooper, James C., Luke M. Froeb, Daniel O’Brien, and Michael G. Vita. 2005. Vertical Antitrust Policy as a Problem of Inference, International Journal of Industrial Organization 23: 639–664. Culbertson, W.  Patton, and David Bradford. 1991. The Price of Beer:  Some Evidence from Interstate Comparisons. International Journal of Industrial Organization 9: 275–89. Dicke, Thomas S. 1992. Franchising in the America: The Development of a Business Method, 1840–1980. Chapel Hill: University of North Carolina Press. Dnes, Antony W. 1992. Franchising: A Case-Study Approach. Aldershot: Ashgate Publishing. Eckard, E. Woodrow. 1994. An Empirical Test of the Free Rider and Market Power Hypotheses: A Comment. Review of Economics and Statistics 76: 586–89. Ferrari, Stijn, and Frank Verboven. 2012. Vertical Control of a Distribution Network:  An Empirical Analysis of Magazines. Rand Journal of Economics 43: 26–50. Gift, P. 2008. The Tie-In Decision. Graziadio Business Review 11: 1–9. Grimes, Warren. 1999. Franchising: The Vacuum of Antitrust Leadership. American Antitrust Institute, FTC Watch No. 531.

Franchising and Exclusive Distribution   411

Hanssen, Andrew. 2000. The Block-Booking of Films Re-examined. Journal of Law and Economics 43: 395–426. Heide, John B., Shantanu Dutta, and Mark Bergen. 1998. Exclusive Dealing and Business Efficiency: Evidence from Industry Practice. Journal of Law and Economics 41: 387–407. Hovenkamp, Erik, and Herbert Hovenkamp. 2010. Tying Arrangements and Antitrust Harm. 2010. Arizona Law Review 52: 925–976. Available at http://works.bepress.com/ herbert_hovenkamp/10. IFA Educational Foundation and Frandata Corp. 1998. The Profile of Franchising. 3  vols. Washington, DC: International Franchise Association. Ippolito, Pauline M. 1991. Resale Price Maintenance:  Empirical Evidence from Litigation. Journal of Law and Economics 34: 263–94. Jordan, W. John, and Bruce L. Jaffe. 1987. The Use of Exclusive Territories in the Distribution of Beer: Theoretical and Empirical Observations. Antitrust Bulletin 32: 137–64. Joseph, Robert T. 2011. Antitrust Law, Franchising, and Vertical Restraints: How Did We Get Here? Where Are We Going? Franchise Law Journal 31: 3–13. Klein, Benjamin. 1995. The Economics of Franchise Contracts. Journal of Corporate Finance 2: 9–37. Kosová, Renáta, and Francine Lafontaine. 2011. Understanding Franchising: A New Look at U.S. Franchisor Activities. Economies et Sociétés, série Dynamique Technologique et Organisation, No. 13. Lafontaine, Francine, and Fiona Scott Morton. 2010. Markets: State Franchise Laws, Dealer Terminations, and the Auto Crisis. Journal of Economic Perspectives 24: 233–50. Lafontaine, Francine, and Kathryn L.  Shaw. 2005. Targeting Managerial Control:  Evidence from Franchising. Rand Journal of Economics 36: 131–50. Lafontaine, Francine, and Margaret E. Slade. 2008. Empirical Assessment of Exclusive Contracts. In Paolo Buccirossi, ed., Handbook of Antitrust Economics, 391–414. Cambridge, MA: MIT Press. Lafontaine, Francine, and Margaret E. Slade. 2013. Inter-Firm Contracts: The Evidence. In R. Gibbons and J. Roberts, eds., Handbook of Organizational Economics, 958–1013. Princeton, NJ: Princeton University Press. Love, John F. 1986. McDonald’s: Behind the Arches. New York: Bantam Books. Marvel, Howard. 1995. Tying, Franchising, and Gasoline Service Stations. Journal of Corporate Finance 2: 199–225. Marx, Thomas G. 1985. The Development of the Franchise Distribution System in the U.S. Automobile Industry. Business History Review 59: 465–74. Mathewson, Frank, and Ralph Winter. 1994. Territorial Restrictions in Franchise Contracts. Economic Inquiry 32: 181–92. Michael, Steven C. 2000. The Extent, Motivation, and Effect of Tying in Franchise Contracts. Managerial and Decision Economics 21: 191–201. Mueller, Willard F., and Frederick E. Geithman. 1991. An Empirical Test of the Free Rider and Market Power Hypotheses. Review of Economics and Statistics 73: 301–8. Nurski, Laura, and Frank Verboven. 2011. Exclusive Dealing as a Barrier to Entry? Evidence from Automobiles. University of Leuven. http://ideas.repec.org/p/cpr/ceprdp/8762.html Perry, Martin K., and David Besanko. 1991. Resale Price Maintenance and Manufacturer Competition for Exclusive Dealerships. Journal of Industrial Economics 39: 517–44. Plitt, Jane R. 2000. Martha Matilda Harper and the American Dream. New York: Syracuse University Press.

412   Francine Lafontaine and Margaret E. Slade Sass, Tim R. 2005. The Competitive Effects of Exclusive Dealing: Evidence from the U.S. Beer Industry. International Journal of Industrial Organization 23: 203–25. Sass, Tim R., and David S.  Saurman. 1993. Mandated Exclusive Territories and Economic Efficiency:  An Empirical Analysis of the Malt-Beverage Industry. Journal of Law and Economics 36: 153–77. Slade, Margaret E. 1998. Beer and the Tie: Did Divestiture of Brewer-Owned Public Houses Lead to Higher Beer Prices? Economic Journal 108: 565–602. Slade, Margaret E. 2011. Competition Policy towards Brewing: Rational Response to Market Power or Unwarranted Interference in Well Functioning Markets? In J. Swinnen, ed., The Economics of Beer, 173–95. Oxford: Oxford University Press. Smith, Richard L., II. 1982. Franchise Regulation: An Economic Analysis of State Restrictions on Automobile Distribution. Journal of Law and Economics 25: 125–57. World Intellectual Property Organization (WIPO). 2011. Survey on Measures to Address the Interface between Antitrust and Franchising Agreements. CDIP/4/4 Rev./Study/INF/4. Zanarone, Giorgio. 2009. Vertical Restraints and the Law:  Evidence from Automobile Franchising. Journal of Law and Economics 52: 691–700. Zuehlke, Susanne, and Gianni De Stefano. 2010. EC Motor Vehicle Block Exemption Reform: Are You Ready for the New Regime? European Competition Law Review 3: 93–97.

Cases Referenced Albrecht v. Herald Co., 390 U.S. 145 (1968). Eastman Kodak Co. v. Image Technical Services Inc., 504 U.S. 451, 489 (1992). Federal Trade Commission v. Sinclair Refining Co., 261 U.S. 463 (1923). Eastman Kodak Co. v. Image Technical Services Inc., 504 U.S. 451 (1992). Krehl v. Baskin-Robbins Ice Cream Company, 664 F.2d 1348, 1354 (9th Cir. 1982). Leegin Creative Leather Products, Inc. v. PSKS, Inc., 551 U.S. 877 (2007). International Salt Co. v. United States, 332 U.S. 392 (1947). Queen City Pizza, Inc. v. Domino’s Pizza, Inc., 922 F. Supp. 1055 (E.D.Pa. 1996). Queen City Pizza, Inc. v. Domino’s Pizza, Inc., 124 F.3d 430, reh’g denied, 129 F.3d 724 (3d Cir. 1997). Siegel v. Chicken Delight, 448 F. 2d 43 (9th Cir. 1971). State Oil Co. v. Khan, 522 U.S. 3 (1997).

PA R T I I I

C OL LU SION A M ON G O ST E N SI B L E C OM P E T I TOR S

CHAPTER 17

C A RT E L S A N D C OL LU SION Economic Theory and Experimental Economics JAY PIL CHOI AND HEIKO GERLACH

17.1. Introduction The essence of market competition is to attract customers by delivering better and less costly products, and the goal of antitrust laws is to preserve competitive market environments. Collusion is an agreement—tacit or explicit—among horizontal competitors to suppress competition by coordinating their prices and quantities. This coordination is intended to raise prices and earn higher profits at the expense of consumers. Given this intention, collusion is deemed to be per se illegal in the United States, EU, and virtually every country with antitrust laws, and it is considered the top priority in antitrust enforcement.1 For instance, Section 1 of the Sherman Act in the US states: Every contract, combination in the form of trust or otherwise, or conspiracy, in restraint of trade or commerce among the several States, or with foreign nations, is declared to be illegal.

Similarly, Article 101 of the Treaty on the Functioning of the European Union (2009) declares all agreements between undertakings that “have as their objective or effect the prevention, restriction or distortion of competition” to be “incompatible” with the European Union’s internal market. This chapter provides a selective review of economic theory and experimental evidence on cartels and collusion. In particular, we highlight 1  US Supreme Court justice Antonin Scalia designates collusion as “supreme evil of antitrust” (Verizon Communications v. Law Offices of Curtis V. Trinko, 540 U.S. 398, 408 (2004)) and the European commissioner for competition Mario Monti calls cartels “cancers on the open market.”

416   Jay Pil Choi and Heiko Gerlach the role of incentives in collusion and cartel formation, and we identify conditions that are conducive to collusive behavior.2 We then discuss enforcement against collusion with a particular focus on the recent work on leniency programs and cartel screening. Finally, we look at experimental evidence on cartel formation and the effects of competition policy.

17.2.  Theory of Collusion As emphasized by Whinston (2006), one implication of the illegality of a collusive agreement is that it is not enforceable in the court. When a party to the agreement does not honor the contract and deviates, other parties cannot rely on the court because such an attempt puts them at the risk of being prosecuted. Thus, any (tacit or explicit) collusive agreement should be self-enforcing. That is, it should induce cartel members to adhere to the agreement without the intervention of the court. The following illustrates a canonical model of self-enforcing collusion in repeated games using a very simple stylized Bertrand price competition model. This model is then used as a benchmark to discuss the factors that facilitate or hinder cartel formation. Suppose that there are two firms in the market that produce homogenous products with a constant marginal cost of c. If they set the same price, firms share the market demand equally at the given price. If one firm sets a price lower than the other, the firm with a lower price attracts all consumers. In a static, one-off competition between firms, it is easy to show that the only equilibrium is for both firms to set their prices at the common marginal cost of c. However, if the game is played repeatedly over time, firms can potentially sustain collusion by punishing any deviating firms with price wars. To see this, consider the following strategy in an infinite horizon model with the stage game described above being played in each period.3 Both firms set the monopoly price pm and each receives half of the monopoly profit Πm in every period t as long as pm has been played by both firms up to that time period. Otherwise, they set the price equal to the static Nash equilibrium price of c that leads to a profit of zero.4 Now consider

2 

The economics literature has been very successful in characterizing the set of possible agreements that are sustainable as a collusive outcome, but it is more or less silent on how a particular agreement is reached. 3  Collusion requires an infinite horizon or stochastic termination of the game. Otherwise, the logic of backward induction implies that we have a competitive outcome in each period as long as the stage game has a unique Nash equilibrium. 4  This punishment is credible in the sense that given the other firms enact the punishment, an individual firm has no incentive to deviate. However, firms might have a collective interest to avoid costly punishment and renegotiate the agreement towards a Pareto-superior continuation of the game. If such renegotiation takes place, the initial threat of punishment is eroded and the strategy is not sustainable. Farrell and Maskin (1989) elaborate this point and propose a renegotiation-proof equilibrium instead. In such an equilibrium firms use asymmetric punishment strategies that are immune to Pareto-improving renegotiation. In this vein, McCutcheon (1997) argues that cartels are more likely to

Cartels and Collusion   417

conditions under which this strategy can be sustained as a subgame perfect equilibrium.5 Each firm maximizes the present discounted value of its profits. Let δ denote the discount factor. The optimal deviation strategy is one that just undercuts the monopoly price, thereby capturing the whole monopoly profit in the current period, but receives zero profit from the next period on. The incentive compatibility (IC) condition that such a deviation is unprofitable and that each firm charges pm, assuming that the other one also plays according to the putative equilibrium strategy, is given by



Πm Πm + (δ + δ 2 + δ 3 + ⋅⋅⋅) ≥ ∏ m + 0(δ + δ 2 + δ 3 + ⋅⋅⋅) 2 2

(IC)

This condition can be rewritten as



∏ LL =

δ ∏ (δ + δ + δ + ⋅⋅⋅) = 2 1− δ 2 m

2

3

m

∏ ≥

2

m

= SG .



LL represents the discounted long-term loss from a deviation due to the breakdown of the cartel. SG represents the short-term gain from attracting 50% of the total demand at the monopoly price. The IC condition is satisfied if and only if δ ≥ ½. That is, if firms are sufficiently patient and care about future profits enough, the LL outweighs the SG and firms are not willing to deviate from collusion.

17.3.  Factors Facilitating Collusion There is a rich literature extending the canonical model above to explore conditions under which collusion is facilitated. In the following we present a selective discussion with a focus on more recent contributions to this literature.

form if the cost of negotiation is neither too high nor too low. For low values, firms are unable to sustain a renegotiation-proof equilibrium, whereas for high values, the initial cartel meeting would be too costly. 5  The folk theorem states that any feasible payoffs between zero and the fully collusive outcome can be sustained as an equilibrium in this game if the discount factor is sufficiently close to 1. We focus on the most collusive (symmetric) equilibrium. See Fudenberg and Maskin (1984) for more details on the folk theorem.

418   Jay Pil Choi and Heiko Gerlach

17.3.1.  Number of Firms, Symmetry, and Concentration We can easily generalize the basic model to explore the effects of the number of firms on collusion incentives. Let there be n firms in the market with the number of firms representing the degree of market concentration. Then, the IC condition becomes

∏ LL(n) =

(n − 1)∏ (n − 1)∏ δ ∏ (δ + δ + δ + ⋅⋅⋅) = ≥ ≥ n 1− δ n n n



m

2

3

m

m

m

= SG(n).



This implies that the discount factor required for collusion is given by δ ≥ δ* (n) = n / (n − 1). Observe that δ*(n) is increasing in n; it is more difficult to sustain collusion with a larger number of firms. More cartel members reduce the collusive profit per firm. As a result, the temptation to deviate is higher because the short-term gains from deviation SG(n) increases while the long-term loss from deviation LL(n) decreases. So far we have just assumed that there is an agreement in place and focused only on incentives to carry out the agreement among participants. If all firms are symmetric, the most natural agreement point is a symmetric outcome that would maximize the industry profit. However, if firms are different in important dimensions (such as capacities, cost structures, and product characteristics), it may not be clear how the market should be divided among them, which may make a mutually acceptable agreement difficult to achieve. In addition, even if they can agree on an outcome, asymmetry restricts the set of sustainable collusive outcomes due to differential incentives to deviate. Compte, Jenny, and Rey (2002), for instance, consider a model of collusion in which firms differ only in their capacities. They show that the firm with the largest capacity has the highest incentive to deviate from collusion, while the ability of firms with small capacities to punish such deviation is limited. A more equal distribution of capacities makes firms’ incentives more aligned and facilitates collusion.6

17.3.2.  Demand Conditions We have considered a stationary demand that does not change over time. The evolution of demand can have effects on the sustainability of collusion. If firms are operating in

6 

Vasconcelos (2005) derives a similar result showing that collusion is easier when capital is transferred from the largest firm to the smallest firm. See also Bos and Harrington (2010), who analyze the effect of firm asymmetry while endogenizing the composition of a cartel rather than assuming an all-inclusive cartel. They show that there exists a stable cartel that involves only the largest firms, while firms with sufficiently small capacity are not cartel members. In such a case, cartel members set a price that serves as an “umbrella,” with non-cartel members undercutting the cartel price and supplying at their capacity.

Cartels and Collusion   419

a growing market, it is easier to sustain collusion because future profits become more important in such a market. Mathematically, the effect of a growing market is equivalent to an increase in the discount factor if demand expansion is modeled as a scale parameter of the demand function. Similarly, collusion becomes more difficult to sustain if the market demand is declining over time. Rotemberg and Saloner (1986) show that if demand is fluctuating over time, it is more difficult to sustain collusion in high-demand states. They consider a model with independent and identically distributed demand shocks. The implication of this assumption is that the loss from deviation is the same regardless of the current demand state. However, the short-term gain from deviation is higher when the demand state is higher. As a result, to maintain IC in the high-demand states, it may be necessary to collude at a price less than the full monopoly price. Haltiwanger and Harrington (1991) extend the analysis of fluctuating demand to allow for cyclical demand pattern. With a cyclical demand pattern, the severity of punishment for deviation depends on whether the current period is in the up or down phase of the demand cycle, whereas the temptation to cheat depends on the current level of demand. This implies that for a given level of demand, collusive prices tend to be higher in booms than in recessions.7

17.3.3.  Multimarket Contact If the same set of firms competes in several markets, they may recognize their interdependency better and refrain from vigorous competition. The idea that contact across markets can affect the degree of sustainable collusion was first proposed by Edwards (1955, 337). He reasoned that firms that compete against each other across multiple markets “may hesitate to fight local wars vigorously because the prospects of local gain are not worth the risk of general warfare.” Bernheim and Whinston (1990) formalize Edwards’s idea that the multiplicity of contacts among conglomerate firms may induce “mutual forbearance” and “blunt the edge of their competition.” They show that multimarket contacts can be used as a mechanism to pool the incentive constraints across markets. When there is slack in the incentive constraint in one market, the pooling of the incentive constraints allows the slack to be transferred to the other market where the constraint is binding, thereby aiding collusion in the market with the binding constraint. To illustrate the idea of multimarket contact, consider two firms, 1 and 2, competing in two markets, A and B that are unrelated in other aspects. Let SGi be the short-term

7 

Bagwell and Staiger (1997) define a boom as a sequence of periods of high demand growth and a recession as a sequence of periods of low demand growth. They show that collusive prices are weakly procyclical when growth rates are positively correlated through time and countercyclical when there is a negative correlation.

420   Jay Pil Choi and Heiko Gerlach gains from deviating from a collusive agreement and LLi be the long-term losses from such a deviation in market I = A, B. Then, a necessary and sufficient condition for a collusive agreement to be sustainable in market i is

LLi ≥ SGi .

Now consider a situation in which the following conditions hold: LLA > SGA and LLB < SGB. This implies that if these two markets are considered separately, a collusive agreement is sustainable in market A, but not in B. Suppose firms realize the interdependency and punish in both markets when there is a deviation in any market. Such a punishment strategy entails that the optimal deviation is to cut prices in both markets. In turn, the overall IC condition can be satisfied if we pool the two individual IC conditions together, that is,

LLA − SGA ≥ SGB − LLB .

Collusion in both markets is sustainable if the slack in the incentive constraint of market A (LLA – SGA) outweighs the gains from deviating in market B (SGB − LLB). In other words, multimarket contact allows shifting slack collusive rents from the more collusive market to the less collusive market B.8 This logic also implies that if both markets are symmetric, multimarket contact is irrelevant. Several papers have identified further benefits of multimarket contact, which do arise with symmetric markets. Spagnolo (1999) shows that when market profits enter a concave objective function, multimarket contact may be beneficial, as it reduces the short-run gains from deviation and increases the value of future losses compared. Matsushima (2001) points out that with imperfect monitoring multimarket contact might have a procollusive effect as it makes the detection of deviation easier. Choi and Gerlach (2013) consider demand linkages between markets and derive that collusion in one market is easier (harder) to sustain when products are substitutes (complements).

17.3.4.  Imperfect Observability and Monitoring Successful collusion requires that noncompliance is observed and punished. When each firm’s pricing or output decisions are transparent and observable, any firm’s deviation from the agreed strategy can be easily detected and disciplined appropriately. In the absence of perfect monitoring, collusive agreements tend to break down due to firms’

8  If the pooled IC constraint is not satisfied, the optimal strategy is to treat each market separately and collude in only one market where the IC constraint is satisfied. See Bernheim and Whinston (1990) for more details.

Cartels and Collusion   421

incentives to engage in secret price cuts (Stigler 1964). Green and Porter (1984) formalize this problem in a setting with imperfect public monitoring.9 While firms are unable to directly observe the industry demand level and the quantities set by their rivals, the resulting market price is publicly available. In such a scenario, firms face an inference problem when confronted with low market prices. A low market price can arise due to either a negative demand shock or a quantity expansion by a cartel member. If firms attribute low prices to quantity deviations and engage in a grim strategy of permanently reverting back to competitive Cournot behavior, they run the risk of being stuck with low prices even if no one has engaged in secret price cuts in the event of a negative demand shock. On the other hand, if firms give the “benefit of the doubt” and attribute low demand to a negative demand shock, each firm will have an incentive to engage in secret cartel deviations, and collusion will not be sustainable. Green and Porter (1984) show that the optimal collusive strategy may entail the use of a trigger price to which firms compare the current price before choosing their quantity. If the market price falls short of this trigger price, firms revert to competitive quantities for a fixed amount of time before resuming collusion until the next trigger and so on.10 The optimal collusive strategy prevents deviations by balancing the short-run gains from overproducing with the expected future loss from triggering price wars (periods of low prices due to competitive quantities). On the equilibrium path, price wars occur in low-demand periods although all firms adhere to their collusive strategy. This is in sharp contrast to the perfect monitoring case where a credible threat of punishment is sufficient to deter deviation from the collusive equilibrium. In Green and Porter (1984), costly price wars constitute a necessary evil for the cartel to sustain collusion when direct information on its members’ strategies is unavailable.11 In turn, the occasional breakout of price wars should not be construed as a sign of unsustainability of collusion. Sannikov and Skrzypacz (2007) consider a similar framework where market price information arrives continuously and depends on the total supply level plus a random shock. They show that collusion with imperfect public monitoring is impossible when 9 

In models with public monitoring, firms receive public signals generated by price or output choices. Hence, the continuation play can always be constructed as an equilibrium of the repeated game, and the dynamic programming technique of Abreu, Pearce, and Stacchetti (1986, 1990) and Fudenberg, Levine, and Maskin (1994) can be applied to establish folk theorems. If, by contrast, firms receive private signals, this recursive structure is destroyed. In models with private monitoring, communication can generate a publicly observable history on which the continuation play can be conditioned and the recursive structure recovered (Kandori and Matsushima 1998; Compte 1998). 10  Green and Porter (1984) considered a restricted set of these “trigger strategy” profiles with a predetermined T-period punishment phase. Abreu, Pearce, and Stacchetti (1986) provide a more general analysis that eschews the restriction to trigger strategy profiles. They characterize optimal pure strategy symmetric equilibria, in which the most collusive solution is described by two regions in the signal space and two actions. Thus, the duration of the punishment phase is stochastic rather than fixed at T as in Green and Porter (1984). 11  Green and Porter (1984) use strongly symmetric equilibria, in which firms only recur to symmetric stage game action profiles. By contrast, Fudenberg, Levine, and Maskin (1994) show that if the cartel can play asymmetric equilibria, the cost of inefficient price wars can be reduced. Indeed, the Folk Theorem applies and monopoly profits are attainable in the limit as the firms become more patient.

422   Jay Pil Choi and Heiko Gerlach firms’ interactions become too frequent (or, equivalently, firms’ production becomes more flexible). More generally, they derive that the frequency of market interaction has a nonmonotonic effect on collusion; first it increases and then decreases collusive payoffs. As the time between actions increases, the gains from deviation in a given period go up, which hurts collusion. By contrast, when frequency increases (and time between actions decreases), firms look at noisier information to infer deviation and trigger price wars too often. Hence, collusive payoffs first increase and then decrease with the frequency of interaction. Harrington and Skrzypacz (2007) consider a model in which firms are unable to observe other firms’ prices and individual demand shocks whereas market shares (out of the inelastic total demand) are publicly available. They consider market share trigger strategies and show, among other things, that collusion cannot be sustained if firms recur to symmetric punishments. The intuition is that if one duopolist cuts his price, he increases the probability that his market share exceeds the trigger value, but at the same time this reduces the probability that the rival’s market share triggers the punishment phase. By contrast, collusive equilibria with asymmetric punishments, such as buy-in or buy-back agreements, are sustainable. Harrington and Skrzypacz (2011) further consider the effectiveness of such asymmetric punishments in an informational setting in which both firms’ prices and quantities are private information, but players engage in costless communication on their sales and make side payments based on the reports in the form of guaranteed buy-ins or buy-backs. The properties of the equilibrium they characterize fits broadly with the collusive mechanisms adopted in a number of prominent recent cartel cases such as in the citric acid, lysine, and vitamins markets.

17.3.5.  Incomplete Information and Communication A recent related strand of the literature analyzes cartel formation and organization when firms hold asymmetric information but are able to observe rivals’ strategies. The main focus of these papers is to characterize the optimal cartel organization with respect to pricing and punishment, as well as to discuss the role of communication among the cartel members. Athey and Bagwell (2001) consider a repeated game duopoly with inelastic demand in which firms’ costs can either be high or low, with independent draws in each period. Each firm knows its own cost realization but not the cost level of its rival. Athey and Bagwell characterize an asymmetric perfect public equilibrium that implements first-best profits and induces firms to truthfully communicate their cost level. Productive (firm) efficiency is achieved by allocating reporting high-cost firms a higher future market share. In a similar setup, Athey, Bagwell, and Sanchirico (2004) consider a continuum of cost types and show that the optimal strongly symmetric perfect public equilibrium sacrifices productive efficiency by using a rigid, nonsorting price scheme in order to deter high-cost firms from mimicking low-cost types. Athey and Bagwell (2008) consider persistent cost shocks to firms and derive some interesting results. When the initial distribution is log-concave and shocks are perfectly consistent, then the equilibrium that maximizes the cartel’s ex ante profits is again pooling. That is, firms

Cartels and Collusion   423

share the market at the consumer’s reservation price. When log-concavity does not hold, equilibria with an initial learning phase and lower prices might be better than pooling. By contrast, when persistence is not too high relative to firms’ patience, firms achieve the first-best equilibrium with productive efficiency as in Athey and Bagwell (2001). Hanazono and Yang (2007) and Gerlach (2009) consider optimal collusion when firms receive private information about the industry demand level. The first paper shows that in the absence of communication, the optimal cartel organization might imply price rigidity. The second paper allows for communication among firms and demonstrates that a collusive scheme in which firms only communicate in high-demand periods approaches the first-best cartel organization when the signal noise goes to zero.

17.3.6.  Vertical Mergers and Restraints The nature of vertical relationships and contracts between upstream and downstream firms can have important effects on collusion. In fact, many cases of collusion have involved intermediate goods industries, while the bulk of the theory on collusion concerns firms selling to final consumers. To address this issue, Nocke and White (2007) examine collusion among upstream firms selling to strategic downstream firms that are competitors to each other. In particular, they investigate the impact of vertical mergers on the upstream firms’ ability to collude. They identify several channels through which vertical merger affects collusion in a setting where upstream firms compete in two-part tariffs for the business of downstream firms. First, vertical mergers give rise to an outlets effect: the deviation profits of cheating unintegrated firms are reduced as the downstream affiliates of their integrated rivals internalize the effects of their purchase decisions on their own upstream firms’ profits and refuse to accept the offer of a deviating firm. As a result, collusion is facilitated with vertical merger. However, there is a countervailing punishment effect: the merged entity suffers less from punishment when it deviates compared to a stand-alone upstream firm, because its downstream firms may earn positive profits in the punishment phase. Thus, the net effect of vertical merger on collusion by upstream firms depends on the relative magnitude of these two countervailing forces. However, Nocke and White show that the outlets effect in general dominates the punishment effect and vertical merger tends to facilitate collusion.12 Vertical restraints that limit competition among retailers can have a facilitating effect on upstream collusion. For instance, Rey and Stiglitz (1995) show how exclusive territories can serve as a mechanism to reduce interbrand competition at the upstream level. In the context of a repeated game, Jullien and Rey (2007) formalize the idea that resale price maintenance (RPM) can help upstream firms collude. They consider a market environment that 12 

The baseline model of Nocke and White (2007) assumes that upstream two-part tariff offers and downstream retail prices are set simultaneously. They also consider an alternative timing in which downstream firms set their retail prices after they have observed upstream offers, and identify two additional effects that further facilitate collusion: the reaction effect and the lack-of-commitment effect.

424   Jay Pil Choi and Heiko Gerlach is subject to local demand and cost shocks, which requires retail prices to respond to retailers’ information about the relevant shocks for efficiency. RPM thus entails efficiency loss and increases the short-run gains from deviation, but makes the detection of deviations easier. Jullien and Rey characterize conditions under which RPM can facilitate collusion and show that it reduces total welfare whenever firms choose to adopt.

17.4.  Cartels and Antitrust Enforcement Given the substantial body of literature on identifying factors that facilitate or hinder the formation of cartels, it is somewhat surprising that until recently there has been very little research devoted to the effects of an active competition policy against cartels. The two main weapons of antitrust authorities in the fight against cartels are leniency programs and surprise inspections (“dawn raids”) (OECD 2003). Leniency programs promise a reduction in government fines in exchange for direct evidence from cartel participants. Direct evidence includes statements or documents explicitly detailing the agreement and identifying the parties to it. Such direct evidence is typically sufficient to convict the cartel. Surprise inspections are based on cartel screening, which involves the collection of circumstantial evidence from market observation or buyers’ complaints. When the antitrust authority reaches an evidence level that justifies further investigation, it can trigger dawn raids to search firms’ headquarters for more evidence to convict the cartel. In most jurisdictions, circumstantial evidence is not sufficient to prosecute a cartel but can be successfully used if there is additional evidence that firms communicated or reached an agreement otherwise. In what follows, we first review the recent theoretical literature on the effects and effectiveness of leniency programs. We then canvas different strands of literature related to cartel screening and cartel behavior in the presence of an antitrust authority.

17.4.1.  Leniency Programs A corporate leniency program offers reduced penalties to cartel members in exchange for revealing direct evidence and cooperating with the antitrust authority during the prosecution phase. In practice, the designs of leniency programs differ with respect to (1) whether leniency is given before and/or after the launch of an investigation, (2) the amount of leniency for the second and third firm to come forward, (3) minimum evidence disclosure standards, (4) exclusion of cartel ringleaders, and (5) the existence of individual leniency programs.13 13  See Spagnolo (2008) for a comprehensive survey and a detailed discussion of institutional differences across different leniency programs.

Cartels and Collusion   425

The analysis of leniency programs in antitrust follows and builds on the literature on self-reporting and general law enforcement. Kaplow and Shavell (1994) and Malik (1993) analyze the optimal design and the efficiency of self-reporting schemes. An individual violator has an incentive to self-report his harmful act if the authority proposes a reduced sanction that equals (or is slightly less than) the expected value of not reporting, which is the (full) stipulated sanction times the probability of apprehension and prosecution. This policy ensures that individuals who commit the act self-report in equilibrium, whereas the authority audits nonoffenders with a positive probability. Thus, for a given audit probability, self-reporting permits the authority to save on costly auditing, while the individual’s expected return from the act and deterrence are the same as without a self-reporting policy. The use of self-reporting thus strictly increases welfare.14 The most important difference between this literature and the analysis of leniency policies in antitrust is that cartels, by definition, are multiagent crimes and require self-enforcing cooperation to prevent opportunistic deviation by individual firms. Hence, the analysis of leniency programs in antitrust requires a dynamic or repeated game framework that accounts for (1) the cartel members’ incentives to adhere to the cartel strategy, (2) the incentives to self-report, and (3) the interaction between the two sets of incentives.15 The antitrust literature has identified three basic effects of leniency programs on cartel sustainability and self-reporting incentives of cartel members. To illustrate these effects, consider the following simple example. Suppose a two-firm cartel has broken up and both firms hold incriminating direct evidence they could present to the antitrust authority. The antitrust authority has a simple leniency policy in place. The first firm to present evidence that leads to the prosecution of the cartel receives a reduced fine F − L, whereas the other cartel member pays the full fine F > 0. The policy parameter L ≥ 0 measures the amount of leniency granted by the authority. If both firms apply simultaneously for leniency, each firm receives the reduced fine with probability ½. If no firm applies, the antitrust authority uses its circumstantial evidence and is able to convict the cartel with probability p, in which case firms pay the full fine F. Independent of the leniency program, firms occur a cost of D if and only if they are convicted. This cost represents damage payments to customers, criminal sanctions for the management, and/or reputational loss. The strategic situation for the firms can be represented in the normal form shown in ­figure 17.1.

14  Innes (1999a, 1999b) extends the framework and shows that self-reporting has a positive effect on ex post violator remediation. Innes (2001) allows violators to engage in avoidance activities and demonstrates that self-reporting is beneficial since it reduces the wasteful cost of avoidance. Gerlach (2013) considers the case where the authority is unable to commit to an investigation effort and shows that leniency programs are efficient if and only if the harm of the act is not too large. 15  Leniency programs are also related to the literature on plea bargaining and in particular multilateral plea bargaining (Kobayashi 1992). Plea bargaining occurs after the start of an investigation, whereas leniency policies aim to induce spontaneous self-reports of cartels. Leniency programs also guarantee automatic leniency, whereas the amount of amnesty in plea bargaining is uncertain and depends more closely on the evidence held by the antitrust authority.

426   Jay Pil Choi and Heiko Gerlach Firm 2 apply

not apply

apply

–D–F+L/2, –D–F+L/2

-D-F+L, –D–F

not apply

–D–F, –D–F+L

–p(D+F), –p(D+F)

Firm 1

FIGURE  17.1  The

Self-Reporting Game with Leniency

Expected sancation

p(D+F)

D+F-L/2

3

2 D+F-L

1 p 1-L/(D+F) FIGURE  17.2  The

1-L/[2(D+F)]

Effects of Leniency on Self-Reporting and Expected Sanctions

Note that (apply, apply) is always a Nash equilibrium of this game. Given that the rival applies for leniency and the cartel is convicted, not applying would mean that the firm forgoes the opportunity to race for the fine reduction. There is a second Nash equilibrium (not apply, not apply) if p(D + F) ≥ D + F − L or

L ≤ (1 − p)(D + F ) or p ≤ 1 − L / (D + F ).



(LP)

Firms can sustain an equilibrium with no self-reporting if and only if the fine reduction proposed by the leniency program is less than the expected fine savings from concealing in the hope that the antitrust authority is unable to convict the firms. Figure 17.2 compares the expected sanctions of the two pure-strategy equilibria (apply, apply) and (not apply, not apply) as a function of the probability of conviction p. Clearly, whenever the (not apply, not apply) equilibrium exists and condition (LP) holds, firms expect lower sanctions in this equilibrium compared to the self-reporting equilibrium. Hence, equilibrium selection based on Pareto dominance suggests that firms are

Cartels and Collusion   427

not self-reporting if and only if condition (LP) is satisfied. This leads to a discontinuity (a jump) in the expected sanctions at parameter values where condition (LP) holds with equality. For low values of p, the antitrust authority holds little circumstantial evidence and the firms do not self-report in equilibrium. Figure 17.2 also shows that for intermediate values of p in [1 − L / (D + F), 1 − L / (2(D + F))], where self-reporting is the unique equilibrium, firms would be better off in the equilibrium without self-reporting. However, each firm has a unilateral incentive to deviate and self-report the cartel. In this case, the leniency program turns the self-reporting game into the classic prisoner’s dilemma where apply is the dominant strategy. What are the effects of increasing leniency on the expected sanctions and the incentives to self-report? First, raising L makes it more attractive for a firm to deviate from (not apply, not apply) and condition (1) becomes harder to satisfy. The cutoff for the self-reporting equilibrium in figure 17.2 shifts to the left and there is self-reporting for lower values of p. This is the source of the Deviator Amnesty effect.16 This effect increases cartel detection with direct evidence and is strictly positive for the antitrust authority. Second, more leniency induces firms to shift from the (not apply, not apply) equilibrium to the equilibrium in which both firms rush to be the first firm to come forward in order to receive leniency from the antitrust authority. The switch in equilibrium regime discontinuously increases the expected sanctions for firms. This Race to the Courthouse effect is also positive for the antitrust authority. From an ex ante point of view, leniency increases the expected sanctions on the collusive equilibrium path and thus reduces expected cartel profits and increases deterrence. Finally, there is a third effect for parameter values at which an increase in leniency has no effect on the equilibrium outcome since firms would self-report no matter what. In figure 17.2 this holds for values p > 1 − L(D + F). Here an increase in leniency simple reduces the expected sanctions for firms. This raises ex ante collusive profits and reduces cartel deterrence. This negative effect of leniency is referred to as the Cartel Amnesty effect. In order to evaluate the net impact of these effects on deterrence and characterize an optimal leniency policy, a dynamic theory of cartel formation, prosecution, and self-reporting is necessary. Harrington (2008) sets up a repeated oligopoly model of collusion, which incorporates all three effects in one framework. In each period, the antitrust authority (AA) opens an investigation with an exogenous probability. During an investigation the AA holds circumstantial evidence, which would—absent any self-reporting—allow a successful prosecution with probability p. This conviction probability is randomly drawn from a continuous distribution function and observed by all firms. Firms then simultaneously decide whether to adhere to the cartel’s market strategy and whether to self-report. Once the cartel has been convicted, firms are unable to form a cartel again. The AA’s leniency program is similar to the simple model above. The first firm to come forward pays a reduced fine, whereas all other convicted firms pay the full fine. The amount of leniency is the only policy variable in this model. 16 

The nomenclature of these effects follows Harrington (2008).

428   Jay Pil Choi and Heiko Gerlach The analysis focuses on cutoff strategies of the following type. If in a given period p is sufficiently small, firms collude. For intermediate values, firms compete, and for high values of the conviction probability, firms compete and report their past cartel activity. Deviations are punished by infinite reversion to the static Nash equilibrium. The sustainability of cartel is thus a function of the endogenously determined cutoff values and the cartel members’ expected profits in each of these equilibrium regimes. The existence of the three effects described above depends on the amount of leniency given. For high levels of leniency, only the Deviator Amnesty effect and the Cartel Amnesty effect are at work. Here the former effect dominates, which means that at least locally (for a generous policy) more leniency increases ex ante deterrence. By contrast, if leniency is sufficiently small, there only exist the Race to the Courthouse effect and the Cartel Amnesty effect. In this case either effect can dominate. Overall, the optimal corporate leniency policy is either one that waives all penalties or one with moderate levels of leniency. Two further results in Harrington (2008) are noteworthy. First, within this framework it is always optimal to give leniency only to the first firm to come forward. Giving amnesty to the second firm does not increase the incentive of an individual to deviate from an equilibrium in which firms are supposed not to apply. As in our simple model above, the Deviator Amnesty effect is only affected by the leniency the first firm receives. However, ex ante, more leniency for the second firm reduces expected sanctions along the collusive equilibrium path and leads to more cartel formation. Furthermore, the analysis shows that it might be optimal to condition admission to the leniency program on the amount of circumstantial information that the AA possesses. In fact, a feature of several corporate leniency programs is that leniency is only awarded if the information significantly contributes to successful prosecution of the cartel. To see this, consider again our model above. For values p ≥ 1 − L / (2(D + F)), the unique equilibrium is for firms to report, which yields lower expected sanctions compared to a situation without a leniency program. In other words, if the conviction probability is high, the leniency program shields the firms from higher sanctions. Hence, excluding firms from the leniency program reduces the negative Cartel Amnesty effect and increases ex ante expected deterrence. Motta and Polo (2003) is the first paper to discuss the effects of leniency programs in a dynamic model. This paper differs in two key aspects (and some other minor features). Once an investigation into the industry starts (which occurs with some exogenous probability), the conviction probability can take exactly one value p > 0 (as opposed to a continuous random variable in Harrington 2008). This implies that the cartel behavior after the launch of an investigation along the equilibrium path is ex ante determined. Firms choose one of three options: collude and report, collude and conceal, or do not collude at all. As a consequence, the amount of leniency has no impact on the expected profits on the collusive equilibrium path without self-reporting, and the Run to the Courthouse effect does not exist in this framework. The model also assumes a different stage game timing. First, firms choose simultaneously their market strategy. If a firm deviates from the cartel strategy, it is impossible for the AA to convict the firm. Then, if an investigation opens, the firms decide whether to self-report for a given market strategy. Hence,

Cartels and Collusion   429

an equilibrium without self-reporting only exists if the incentive constraint at the market strategy stage and at the self-reporting stage is satisfied. Since the amount of leniency does not affect the former constraint, it follows that the Deviator Amnesty effect is only present for parameter values at which the self-reporting constraint is more restrictive. This is the case when the conviction probability is sufficiently high. By contrast, the third effect, the Cartel Amnesty effect, is always present in this model. An interesting aspect of the analysis of Motta and Polo (2003) is that the AA has a richer policy instrument set than in Harrington (2008). An AA chooses the amount of leniency and the investment in ex ante monitoring (which positively affects the probability that an investigation starts) and ex post investigation (which increases the probability of successful conviction p). In a first step, it is shown that full leniency is optimal only if it induces firms to collude and self-report in equilibrium. If firms collude and conceal or if no cartel forms, it is optimal to minimize the Cartel Amnesty effect and not to award any leniency. In a second step, the optimal enforcement policy is determined as a function of the resources of the AA. Deterring collusion (and not offering leniency) is optimal if the authority has sufficient resources. With fewer resources available, the AA optimally chooses between two collusion regimes: a regime without reporting and no leniency or a regime with reporting and full amnesty. In order to provide self-reporting incentives the AA has to commit to a high prosecution effort p, which is costly. Thus, a leniency program is optimal if the AA’s resources are intermediate. By contrast, if the AA’s budget is small, it is optimal not to give leniency and induce firms to collude without self-reporting. Spagnolo (2004) and Aubert, Kovacic, and Rey (2006) both consider frameworks in which firms have to apply for leniency before an investigation is opened. In each period there is an exogenous probability that an investigation opens. Once the investigation is launched it leads with probability 1 to the conviction of the cartel. This approach thus focuses on the effect of leniency on cartel deterrence. In fact, leniency programs are not used in equilibrium as it is not optimal for firms to collude and self-report the cartel immediately afterwards. Hence, of the three effects identified above, only the Deviator Amnesty effect is at work. Consequently, more leniency always reduces cartel stability, and both papers advocate not only fine reductions for leniency applicants but rewards for the first firm to come forward. They argue that positive rewards provide more leverage and might be necessary to induce self-reporting and deter cartel formation.17 Additionally, Spagnolo (2004) shows that a program that rewards the first firm with a fines-financed bounty can completely deter cartels at a finite level of fines and without any prosecution cost for the AA. Another interesting feature of this paper is the discussion of the idea that leniency programs spread mistrust among cartel members and introduce strategic risk. To capture this effect, the risk dominance criterion is used to compare the two pure strategy equilibria (apply, apply) and (not apply, not apply). It is demonstrated that moderate leniency programs (stipulating reduced fines rather than rewards) increase the riskiness of the (not apply, not apply) equilibrium and thus raise 17  In our simple model above, rewards (that is, L > F) are necessary to induce self-reporting if and only if the conviction probability is small relative to fines and damages, i.e. P < D / (D + F).

430   Jay Pil Choi and Heiko Gerlach cartel deterrence. This mistrust effect is stronger if the leniency program only admits the first firm to come forward. An interesting feature of Aubert, Rey, and Kovacic (2006) is their discussion of how individual leniency programs might be able to complement corporate leniency program. In situations where corporate leniency programs are not able to induce self-reporting from firms, AAs can offer a bounty and individual amnesty to employees reporting incriminating direct evidence. In order to prevent employees from reporting, cartel members have to increase their pay to informed employees. This makes collusion less profitable and increases ex ante deterrence. Cartel members thus have an interest in minimizing the number of informed employees. Finally, Choi and Gerlach (2012b) analyze cartel formation and self-reporting incentives when firms operate in several geographical markets and face antitrust enforcement in different jurisdictions. They use the framework of Harrington (2008) with a fixed conviction probability like in Motta and Polo (2003). When international antitrust prosecution is uncoordinated, multimarket contact allows firms to reduce self-reporting in equilibrium and sustain cartels more effectively. The reason is that partial self-reporting strategies allow firms to pool cartel stability constraints and transfer slack from one market to prevent self-reporting in the other market. The effectiveness of cooperation and information sharing among antitrust authorities is a function of how much and which type of information is shared. Sharing cartel leads and evidence gathered by the antitrust authority unambiguously raises global welfare. By contrast, sharing information obtained from leniency applicants can increase or decrease the incentives to self-report. In the absence of confidentiality agreements, leniency applicants optimally apply in both jurisdictions or not at all (rather than in one jurisdiction only). If antitrust penalties and conviction probabilities are relatively low, information sharing induces firms not to report any cartel under investigation. This analysis thus warrants a cautious approach of antitrust authorities with respect to sharing information from self-reporting companies.

17.4.2.  Screening and Detection Screening refers to the collection of circumstantial evidence that can be used to (1) identify suspect markets for further investigation, (2) increase the likelihood of conviction in order to induce cartel members to self-report using a leniency program and (3) raise general cartel deterrence. Circumstantial evidence consists of economic evidence and “communication” evidence. Economic evidence can be structural, relating to industry characteristics that are more conducive to cartel formation (as discussed in section 17.2), or behavioral. Behavioral evidence identifies firm conduct that suggests that a cartel agreement has been reached, such as parallel price movements or a surprising increase (or decrease) in prices. Economic evidence in itself is not sufficient to convict a cartel. However, cartel cases have been brought when the authorities held economic and communication evidence. Communication evidence is evidence that cartel operators met or otherwise communicated, but does not describe the substance of their communications.

Cartels and Collusion   431

This includes meetings at trade associations or internal documents that demonstrate detailed knowledge of competitors’ pricing strategies.18 Behavioral screening by buyers or antitrust authorities exposes cartel members to detection and prosecution and introduces additional constraints on cartel behavior. A cartel does not only have to deter individual members’ defections but also choose a cartel strategy to avoid detection and minimize the risk of antitrust enforcement. The earlier literature on cartel screening focused on avoiding detection, taking internal cartel stability as given. This includes papers like Block, Nold, and Sidak (1983) and Feinberg (1980, 1984). They analyze situations where the probability of conviction is an exogenous function of the price-markup of the cartel. Thus, cartels optimally reduce the price below the monopoly level to avoid antitrust enforcement and penalties. A degenerate form of this setup, where full collusion results in an exogenous conviction probability, has been extensively used in the literature on leniency programs and by some more recent papers on antitrust enforcement with multimarket contact (Choi and Gerlach 2012a, 2013). Choi and Gerlach (2012a) analyze cartel formation and international antitrust enforcement when multinational firms operate in several jurisdictions with local antitrust authorities. Trade and arbitrage create a negative demand linkage between the local markets, and the sustainability of collusion in one market is affected by the existence of collusion in the other market. This interdependence leads to externalities in local antitrust enforcement, and there might be an over- or underprovision of cartel prosecution from a global welfare point of view. The paper also derives conditions under which information sharing among antitrust authorities can achieve the globally first-best enforcement outcome. Besanko and Spulber (1989) is the first paper to endogenize the functional relationship between the cartel’s price behavior and the detection probability in a static principal-agent model of antitrust enforcement under asymmetric information. The antitrust authority is unable to observe the cost level of the cartel and conditions its audit (or dawn raid) probabilities and antitrust fines on the observed price in the industry. Faced with this policy, the cartel decides whether to set competitive or collusive prices as a function of its cost type. The optimal policy has the following features. It is optimal to audit only when prices are high. This might result in erroneously auditing competitive, high-cost industries in order to prevent more severe price collusion in low-cost industries. Moreover, it is always optimal to allow some minor degree of collusion in low-cost industries instead of recurring to costly audits to deter collusion entirely. While the analysis yields some interesting insights, its use is somewhat limited by the fact that often antitrust authorities do not hold sufficiently precise information on industry prices. This informational constraint is less of a concern in models of private antitrust lawsuits where buyers screen for cartels. In the models of Salant (1987) and Baker (1988) buyers observe the cartel’s production cost and strategically increase their demand in anticipation of being awarded damages, which are a function of the 18 

See section 4 of Marshall and Marx (2012) for a comprehensive discussion of the detection of cartels using economic evidence.

432   Jay Pil Choi and Heiko Gerlach cartel’s cost overcharge. This perverse incentive offsets the increase in the cartel’s price due to the expected increase in penalties. As a consequence, the cartel produces exactly the unconstrained monopoly quantity, and private cartel enforcement is neutral. By contrast, in Besanko and Spulber (1990), consumers are unable to observe the cartel’s cost level. In this case a higher price does not necessarily imply that the antitrust violation is more severe. Hence, buyers are less inclined to strategically increase demand in the hope of collecting treble damages in a private antitrust lawsuit. Asymmetric information thus entails that private antitrust enforcement can have a strictly positive welfare effect.19 A first step towards a dynamic theory of cartel pricing and screening is made in Harrington (2004, 2005). Harrington (2005) characterizes the optimal cartel price path from the noncollusive to its steady-state level in the presence of buyer detection and (accumulating) damage awards. The buyer detection process is assumed to be exogenous.20 In a given period, the detection probability increases in the size of the cartel’s price change from the previous period. In other words, big price movements make buyers suspicious that a cartel is operating. Upon detection, the cartel has to award damages to the buyers, which are a function of the current and past price levels. The impact of past overcharges on damages decreases over time. Furthermore, it is assumed that the discount factor is sufficiently high such that the usual incentive constraints for cartel stability are satisfied. The cartel’s optimal price path is shown to exist for fairly general functional forms of the detection probability and the damage awards. In the transition to the steady-state level, the cartel price is nondecreasing over time. The steady-state cartel price is strictly below the monopoly level when penalties include (price dependent) damage awards. When penalties are price independent, the cartel reaches the monopoly price in the long run. The reason for this result is that in the steady state, marginal price changes have a negligible effect on detection and only affect penalties through the damage awards. If penalties are not sensitive to the cartel’s price, the monopoly price is achieved at no cost. The paper also reports an interesting comparative static result for the steady-state price. The damage awards are based on the difference between cartel price and the (but for) competitive price. If the damages are calculated at a more punishing, that is, lower but-for price, the cartel’s optimal steady-state price goes up. The reason for this is that instead of reducing the price to lower the damage per unit sold, it is optimal to increase the price and reduce the number of units subject to damage awards. Thus, contrary to the static models of buyer detection discussed above, damage awards can create perverse incentives for the cartel even in the presence of nonstrategic

19 

Spiller (1986) achieves a similar result by introducing limited liability for the cartel’s damage awards. 20  In a related paper, Harrington and Chen (2006) model the buyers’ belief formation process more explicitly. Buyers are unable to anticipate the optimal cartel price path but they react to structural breaks in the pricing pattern by inferring collusion statistically from past prices. Numerical solutions of the cartel’s optimal pricing problem show that there is a transitory and a stationary cartel phase. In the stationary cartel phase, collusion reduces the price variance generated by the underlying cost shocks.

Cartels and Collusion   433

buyers. Harrington (2004) considers the same framework and analyzes the case where the discount factor is lower and the incentive constraints for cartel stability (ICC) may be binding. Numerical analysis shows that if the ICC is binding, the optimal cartel price might be first increasing and then decreasing. The author explains this result by decomposing the effects of the two dynamic forces at work. If penalties are independent of the price level but the detection probability increases in the absolute value of price changes, then the optimal price path satisfying the ICC is always nondecreasing. Deviations from a planned price hike reduce the detection probability. Hence, if the firms have no incentive to deviate from a planned price increase, then there is also no incentive to deviate when prices remain stable. A reduction in cartel prices is thus never necessary to maintain cartel stability. At the other extreme, if the detection probability is independent of prices and prices only affect damage awards, then a nonincreasing continuation price path is optimal. The payment of damages is more likely on the equilibrium path. Since damage awards with an active cartel grow faster over time, the cartel needs to reduce the equilibrium cartel price in order to satisfy the ICC and prevent deviations. The paper also shows that—given the assumed screening technology—antitrust policy might have the undesired effect of making cartels more stable since deviations increase the probability of detection.

17.5.  Experimental Evidence Due to the illicit nature of collusion, it is difficult to test the theory of collusion with empirical data.21 In response, there has been a stream of studies over the last two decades testing some of the theoretical models in experimental settings. First we briefly discuss studies that relate to conditions that facilitate or hinder collusion. Then we have a more detailed look at the emerging experimental literature on collusion with an active antitrust authority.

17.5.1.  Experimental Evidence on Collusion Feinberg and Husted (1993) test the effect of the discount factor on the extent of collusion in a laboratory setting by running Cournot sessions with high and low continuation probabilities (as proxy for discount factors). They confirm the theory by documenting that collusion emerges as an equilibrium only with the high probability of continuation. Similarly, Dal Bo (2005) finds that cooperation increases with the probability of 21  See ­chapter 18 in this volume by Levenstein and Suslow for empirical evidence on cartels and collusion. This section is also complementary to c­ hapter 10 in volume 1 of this Handbook by Müller and Normann, “Experimental Economics in Antitrust.”

434   Jay Pil Choi and Heiko Gerlach continuation and that infinitely repeated games result in higher levels of cooperation than finitely repeated ones of the same expected length.22 Huck, Normann, and Oechssler (2004) study the impact of the number of firms in an industry on the competitiveness of Cournot markets, and they show that a collusive outcome is indeed only possible with a small number of firms. In particular, they find that collusion is possible with two firms, the typical Nash outcome results with three firms, and there is never a collusive outcome with four or more firms. One well-established and robust outcome in experiments is that preplay communication and announcement significantly enhances the ability to collude (Davis and Holt 1990, Cason and Davis 1995). Since these communications are essentially a cheap talk, they cannot affect the incentives to sustain collusion. However, they help in coordinating on a particular collusive outcome since the folk theorem tells us that there is a continuum of equilibria in infinitely repeated games with a sufficiently high discount factor. Fonseca and Normann (2011) also explore the impact of communication on the ability to collude in experimental markets. They run experiments with n ∈{2,4,6,8} firms and find strong evidence that communication facilitates collusion for any number of firms. Nevertheless, it turns out that the gain from communication is nonmonotonic in the number of firms, with the largest gain being found when the market has four firms. In addition, they find that tacit collusion is common with two firms, but very rare with four or more firms. By contrast, the experimental evidence on ex post sharing of information is not yet fully settled. Theory predicts that the publication of actions and profits of market participants makes monitoring easier and the punishment of deviation more effective, which facilitates collusion. The experimental results indeed confirm that sharing of information on market conditions such as demand and cost leads to a more collusive outcome (Feinberg and Snyder 2002). More surprisingly, Huck, Normann, and Oechssler (2000) show that sharing of firm-specific information about action and profits of the firms leads to a more competitive outcome, a puzzling result that contradicts the theory. They attribute this outcome to imitative behavior by less successful firms. In a similar vein, Rojas (2012) finds that information about demand might facilitate collusion more than monitoring. Phillips and Mason (1992) test the idea of mutual forbearance via multimarket contact in experimental conglomerate markets in an attempt to provide experimental evidence of the theory of multimarket contact formalized by Bernheim and Whinston (1990). They consider two markets with different demand and cost conditions. However, they do not find support for the notion that conglomeration causes more cooperation in both markets. Rather, they find that compared to single-market outcomes, multimarket contact induces more cooperative behavior in a more competitive market than in an isolated market, whereas it induces more competitive behavior in a relatively cooperative market under controlled single-market experiments. In 22  Dal Bo and Frechette (2011) further show that cooperation in repeated games does not necessarily increase as players gain more experience.

Cartels and Collusion   435

addition, they find that the relative gain from more cooperation in the more competitive market outweighs the relative loss from less cooperation in the other, which results in higher overall profits. Mason, Phillips, and Nowell (1992) investigate the effect of firm asymmetry on collusion. They compare market outcomes under symmetric cost conditions with those under asymmetric cost conditions and find that in asymmetric markets it is more difficult to sustain collusion, and it takes longer to reach equilibrium compared to in symmetric ones.

17.5.2.  Experimental Evidence on Leniency Programs The fact that is hard to observe the population of active cartels also means that it is difficult to assess the success of policy measures against cartels. Does an increase in cartel detection mean that the policy is indeed working, or is the policy ineffective and the number of cartels has increased?23 This fundamental data problem favors experimental studies as a potentially useful tool to shed some light on the potential effects of antitrust policy against cartels. The first experimental study to investigate the effectiveness of policies against cartels is Apesteguia, Dufwenberg, and Selten (2007). They set up a static game of cartel formation and self-reporting in three stages. In the first stage, players can decide whether to form a cartel or not. If all three players in a market agree, the cartel forms and the players are allowed to discuss their pricing strategy. In the second stage, they set prices along a discrete demand function, which reflects a homogenous good Bertrand tripoly. Finally, the players learn all prices and can decide whether to self-report the cartel or not. A cartel is only prosecuted if at least one firm self-reports. The authors look at four different treatments. “Ideal” is the competitive benchmark, in which players are not allowed to form cartels and only the second stage is played. “Standard” allows players to form cartels and self-report. However, there is no leniency as all firms of a self-reported cartel are fined 10% of their revenues. In “Leniency,” if one firm reports the cartel, it receives full amnesty, whereas all other firms pay 10% of their revenues. If two (three) firms report, then each of these firms pays a 5% (6.667%) fine. Finally, in “Bonus” whistleblowers are not only exempt from fines but share the total collected fines as reward. The overall setup of this experiment, in particular the fact that there is no detection by the antitrust authority and that firms observe prices before self-reporting, is somewhat geared towards the idea that self-reporting might be used as a threat to enforce collusive pricing. This is also borne out by the results. The Standard treatment yields the highest prices, significantly above those in 23 

Harrington and Chang (2009) is a first step towards addressing this issue from a theoretical point of view. They develop a model of cartel formation and dissolution, which allows them to endogenously generate the populations of active cartels and detected cartels. Policy changes affecting the population of detected are then used to trace changes to latent, active cartels.

436   Jay Pil Choi and Heiko Gerlach the competitive benchmark Ideal. The prices in Leniency are just above the competitive benchmark, whereas Bonus performs badly (in allocative terms) with high prices close to the Standard treatment. Subsequent studies use a simulated repeated game framework and are closer to the theoretical models discussed above. Hinloopen and Soetevent (2008) directly build on Apesteguia, Dufwenberg, and Selten (2007) and use their three-phase model as the stage game of a repeated game experiment with uncertain horizon. They introduce two useful modifications. Leniency is given as a function of the order of reports. The first firm receives full amnesty, the second firm gets a 50% reduction, while the last firm pays the full 10% of current revenues. Furthermore, communication makes firms liable and, in the absence of self-reporting, entails a 15% chance of being detected by the antitrust authority. The main results are as follows. The average prices in the cartel treatments with and without antitrust enforcement do not significantly differ from the competitive benchmark. The leniency treatment, however, yields considerably lower prices. The authors decompose this price effect of leniency programs. First, compared to the antitrust treatment without leniency, leniency programs more than halve the percentage of cartels formed from 27% to 13%. This is the cartel amnesty effect from the theoretical models discussed above. The second factor is that cartels that do form in the leniency treatment are not as successful in raising the price as cartels in other treatments. It is shown that the agreed-upon price is undercut in the leniency treatment by one or more cartel members in 97% of the cases compared to 75% in other treatments. Bigoni and coauthors (2012) investigate similar research questions but propose several modifications to the design setup. The most important one is that they allow firms to self-report at two points in the game. Firms can either self-report at the price-setting stage or after firms have set prices. This assumption introduces the Deviator Amnesty effect of leniency programs into the experimental setup. Moreover, the authors use a discrete price-setting duopoly with differentiated goods and consider a different design for the uncertain horizon assumption.24 Antitrust fines are fixed and independent of current revenues. Four main treatments are considered. “Laissez-Faire” is the setup where discussing prices and forming cartels is legal and self-reporting is not available. In “Fine” cartel members can report the cartel, but there is no leniency and all firms pay the full fine. “Leniency” considers a policy program that rewards the first firm to come forward; if both firms apply, they both get 50% of the fine reduction. Finally, there is a treatment “Reward,” in which the reporting firm is awarded the fine paid by the other firm. Several results are noteworthy. First, in the Leniency and the Reward treatment, as expected, many subjects combined price deviation with self-reporting. At the same time, there was also a high rate of self-reporting after the price stage in cases where price deviators did not self-report. Hence, self-reports were used as a punishment for price deviation, as in the previous studies. The authors argue that these reports are more likely to be altruistic 24  At the end of each round of play, subjects are assigned a new competitor with a probability of 15%. After 20 periods, the game ends if there has been a rematch.

Cartels and Collusion   437

punishments.25 While interesting, this argument casts some doubt on how important this strategic punishment might be in practice when the stakes for cartel members are significantly higher. Second, as predicted, antitrust enforcement in all of its three treatments increases cartel deterrence relative to Laissez-Faire. Moreover, Leniency and Rewards drastically increase cartel detection due to higher rates of self-reporting. The welfare ranking of the four treatments in terms of average prices reveals that Fine yields the highest average price, higher than the Laissez-Faire treatment without any antitrust enforcement. By contrast, the average price in Leniency is significantly lower than in Fine but not significantly different from Laissez-Faire. Overall, Rewards is the only treatment that significantly reduces average prices relative to Laissez-Faire and Leniency. These results are due to the fact that average prices when cartels actually form are significantly higher in all policy treatments relative to Laissez-Faire. In turn, it is likely that the welfare ranking is mainly driven by the possibility of strategic (possibly altruistic) punishment via self-reporting in the policy treatments. Finally, the paper gives some interesting results on the postconviction behavior of cartelists. In the Leniency and Rewards treatment recidivism, that is, the probability that cartels reform after conviction, is significantly higher than in Fine. This is consistent with the view that self-reporting destroys trust and makes future collusion harder to sustain. This could also explain the observation that when cartels are not reforming, the postconviction price in Leniency remains low, whereas in Fine firms seem to reach a tacit agreement and prices rise gradually. Using the same experimental framework, Bigoni and coauthors (2010) consider how the legal framework of antitrust enforcement affects the incentives to collude. In particular, they look at four different treatments that vary the probability of conviction in the absence of self-reports, that is, p, and the level of the antitrust fines F. The study shows that, for the same expected penalty from the antitrust authority (that is, holding pF constant), the probability of communication in order to coordinate prices is lower when F is higher (and p lower). This holds for the Fine and the Leniency treatment alike. At the same time, as in the companion study, the probability of communication is higher in Fine. The authors also argue that subjects appear to perceive past fine payments as salient events, which may have an undue impact on the decision to reform a cartel after conviction. The experimental findings show that the fines paid by a subject in previous periods due to detections as well as the number of detections itself, have a significant and substantially negative effect on the propensity to communicate in Fine but not in Leniency. The body of literature on experimental cartel formation and antitrust enforcement is still very small. While it seems too early to draw any conclusions from the available evidence, the early papers all suggest that leniency programs are an effective weapon in the arsenal of antitrust authorities. Leniency increases cartel deterrence and lowers the prices that are sustained on the cartel’s equilibrium path. However, more experimental work is needed to distinguish effects that are relevant to the formation of cartels in practice and those effects that are specific to the laboratory setting. 25  To make this point, the authors run an additional treatment where subjects are rematched each period. In this setup there is no direct gain from self-reporting, but the self-reporting rates are even higher.

438   Jay Pil Choi and Heiko Gerlach

References Abreu, Dilip, David Pearce, and Ennio Stacchetti. 1986. Optimal Cartel Equilibria with Imperfect Monitoring. Journal of Economic Theory 39: 251–69. Abreu, Dilip, David Pearce, and Ennio Stacchetti. 1990. Toward a Theory of Discounted Repeated Games with Imperfect Monitoring. Econometrica 58(5): 1041–63. Apesteguia, Jose, Martin Dufwenberg, and Reinhard Selten. 2007. Blowing the Whistle. Economic Theory 31(1): 143–66. Athey, Susan, and Kyle Bagwell. 2001. Optimal Collusion with Private Information. Rand Journal of Economics 32(3): 428–65. Athey, Susan, and Kyle Bagwell. 2008. Collusion with Consistent Cost Shocks. Econometrica 76(3): 493–540. Athey, Susan, Kyle Bagwell, and Chris Sanchirico. 2004. Collusion and Price Rigidity. Review of Economic Studies 71(2): 317–49. Aubert, Cecile, Patrick Rey, and William Kovacic. 2006. The Impact of Leniency and Whistle-Blowing Programs on Cartels. International Journal of Industrial Organization 24: 41–1266. Baker, Jonathan B. 1988. Private Information and the Deterrent Effect of Antitrust Damage Remedies. Journal of Law, Economics and Organization 4: 385–408. Bagwell, Kyle, and Robert W. Staiger. 1997. Collusion over the Business Cycle. Rand Journal of Economics 28: 82–106. Bernheim, Douglas, and Michael D.  Whinston. 1990. Multimarket Contact and Collusive Behavior. Rand Journal of Economics 21: 1–26. Besanko, David, and Daniel F. Spulber. 1989. Antitrust Enforcement und Asymmetric Information. Economic Journal 99: 408–25. Besanko, David, and Daniel F. Spulber. 1990. Are Treble Damages Neutral? Sequential Equilibrium and Private Antitrust Enforcement. American Economic Review 80: 870–87. Bigoni, Maria, Sven Olof Fridolfsson, Chloe Le Coq, and Giancarlo Spagnolo. 2010. Trust, Salience and Deterrence. SSE/EFI Working Paper Series in Economics and Finance No. 696. Bigoni, Maria, Sven Olof Fridolfsson, Chloe Le Coq, and Giancarlo Spagnolo. 2012. Fines, Leniency and Rewards in Antitrust: An Experiment. Rand Journal of Economics 43(2): 368–90. Block, Michael K., Frederick C. Nold, and Joseph G. Sidak. 1983. The Deterrent Effect of Antitrust Enforcement. Journal of Political Economy 89(3): 429–45. Bos, Iwan, and Joseph E. Harrington, Jr. 2010. Endogenous Cartel Formation with Heterogeneous Firms. Rand Journal of Economics 41: 92–117. Cason, Timothy N., and Douglas D.  Davis. 1995. Price Communications in Laboratory Markets: An Experimental Investigation. Review of Industrial Organization 10: 769–87. Choi, Jay Pil, and Heiko Gerlach. 2012a. International Antitrust Enforcement and Multimarket Contact. International Economic Review 53(2): 635–57. Choi, Jay Pil, and Heiko Gerlach. 2012b. Global Cartels, Leniency Programs and International Antitrust Cooperation. International Journal of Industrial Organization 30(6): 528–40. Choi, Jay Pil, and Heiko Gerlach. 2013. Multi-market Collusion with Demand Linkages and Antitrust Enforcement. Journal of Industrial Economics 61(4): 987–1022. Compte, Olivier. 1998. Communication in Repeated Games with Imperfect Private Monitoring. Econometrica 66(3): 597–626.

Cartels and Collusion   439

Compte, Olivier, Frederic Jenny, and Patrick Rey. 2002. Capacity Constraints, Mergers and Collusion. European Economic Review 46:1–29. Dal Bo, Pedro. 2005. Cooperation under the Shadow of the Future: Experimental Evidence from Infinitely Repeated Games. American Economic Review 95(5): 1591–1604. Dal Bo, Pedro, and Guillaume Frechette. 2011. The Evolution of Cooperation in Infinitely Repeated Games: Experimental Evidence. American Economic Review 101(2): 411–29. Davis, Douglas, and Charles Holt. 1990. The Effects of Non-binding Price Announcements in Posted-Offer Markets. Economics Letters 34: 307–10. Edwards, Corwin. D. 1955. Conglomerate Bigness as a Source of Power in Business Concentration and Price Policy. National Bureau of Economic Research Conference Report. Princeton, NJ: Princeton University Press. Farrell, Joseph, and Eric Maskin. 1989. Renegotiation in Repeated Games. Games and Economic Behavior 1: 327–60. Feinberg, Robert M. 1980. Antitrust Enforcement and Subsequent Price Behavior. Review of Economics and Statistics 62: 609–12. Feinberg, Robert M. 1984. Strategic and Deterrent Pricing Responses to Antitrust Investigations. International Journal of Industrial Organization 2: 1241–66. Feinberg, Robert M., and Thomas A. Husted. 1993. An Experimental Test of Discount Rate Effects on Collusive Behaviour in Duopoly Markets. Journal of Industrial Economics 41: 153–60. Feinberg, Robert M., and Christopher Snyder. 2002. Collusion with Secret Price Cuts:  An Experimental Investigation. Economics Bulletin 3: 1–11. Fonseca, Miguel Alexandre, and Hans-Theo Normann. 2011. Explicit vs. Tacit Collusion: The Impact of Communication in Oligopoly Experiments. Available at http://ssrn.com/​ abstract=1937803. Fudenberg, Drew, David Levine, and Eric Maskin. 1994):  The Folk Theorem with Imperfect Public Information. Econometrica 62: 997–1040. Fudenberg, Drew, and Eric Maskin. 1986. The Folk Theorem in Repeated Games with Discounting or with Incomplete Information. Econometrica 54: 533–54. Gerlach, Heiko. 2009. Stochastic Market Sharing, Partial Communication and Collusion. International Journal of Industrial Organization 27(6): 655–66. Gerlach, Heiko. 2013. Self-Reporting, Investigation and Evidentiary Standards. The Journal of Law and Economics 56: 1061–90. Green, Edward, and Robert Porter. 1984. Non-cooperative Collusion under Imperfect Price Information. Econometrica 52: 87–100. Haltiwanger, John, and Joseph E. Harrington Jr. 1991. The Impact of Cyclical Demand Movements on Collusive Behavior. Rand Journal of Economics 22: 89–106. Hanazono, Makoto, and Yang, Huanxing. 2007. Collusion, Fluctuating Demand, and Price Rigidity. International Economic Review 48: 483–515. Harrington, Joseph E., Jr. 2004. Cartel Pricing Dynamics in the Presence of an Antitrust Authority. Rand Journal of Economics 35: 652–74. Harrington, Joseph E., Jr. 2005. Optimal Cartel Pricing in the Presence of an Antitrust Authority. International Economic Review 46: 145–69. Harrington, Joseph E., Jr. 2008. Optimal Corporate Leniency Programs. Journal of Industrial Economics 56: 215–46. Harrington, Joseph E., Jr., and Joe Chen. 2006. Cartel Pricing Dynamics with Cost Variability and Endogenous Buyer Detection. International Journal of Industrial Organization 24: 1185–212.

440   Jay Pil Choi and Heiko Gerlach Harrington, Joseph E., Jr. and Andrzej Skrzypacz. 2007. Collusion under Monitoring of Sales. Rand Journal of Economics 38(2): 314–31. Harrington, Joseph E., Jr., and Andrzej Skrzypacz. 2011. Private Monitoring and Communication in Cartels: Explaining Recent Collusive Practices. American Economic Review 101: 2425–49. Hinloopen, Jeroen, and Adrian Soetevent. 2008. Laboratory Evidence on the Effectiveness of Corporate Leniency Programs. Rand Journal of Economics 39(2): 607–16. Huck, Steffen, Hans-Theo Normann, and Jorg Oechssler. 2004. Two Are Few and Four Are Many:  Number Effects in Experimental Oligopolies. Journal of Economic Behavior and Organization 53: 435–46. Innes, Robert. 1999a. Remediation and Self-Reporting in Optimal Law Enforcement. Journal of Public Economics 72: 379–93. Innes, R. 1999b. Self-Policing and Optimal Law Enforcement When Violator Remediation Is Valuable. Journal of Political Economy 107(6): 1305–25. Innes, R. 2001. Violator Avoidance Activities and Self-Reporting in Optimal Law Enforcement. Journal of Law, Economics and Organization 17(1): 239–56. Jullien, Bruno, and Patrick Rey. 2007. Resale Price Maintenance and Collusion. Rand Journal of Economics 38: 983–1001. Kandori, Mitchihiro, and Hitoshi Matsushima. 1998. Private Observation, Communication and Collusion. Econometrica 66(3): 597–626. Kaplow, Louis, and Steven Shavell. 1994. Optimal Law Enforcement with Self-Reporting of Behavior. Journal of Political Economy 102: 583–606. Kobayashi, Bruce H. 1992. Deterrence with Multiple Defendants: An Explanation for Unfair Plea Bargains. Rand Journal of Economics 23(4): 507–17. Malik, Arun. 1993. Self-Reporting and the Design for Regulating Stochastic Pollution. Journal of Environmental Economics and Management 24: 241–57. Marshall, Robert C., and Leslie M. Marx. 2012. The Economics of Collusion: Cartels and Bidding Rings. Cambridge, MA: MIT Press. Mason, Charles F., Owen R. Phillips, and Clifford Nowell. 1992. Duopoly Behavior in Asymmetric Markets: An Experimental Evaluation. Review of Economics and Statistics 74: 662–70. Matsushima, Hitoshi. 2001. Multimarket Contact, Imperfect Monitoring and Implicit Collusion, Journal of Economic Theory 98: 158–78. McCutcheon, Barbara. 1997. Do Meetings in Smoke-Filled Rooms Facilitate Collusion? Journal of Political Economy 105(2): 330–50. Motta, Massimo, and Michele Polo. 2003. Leniency Programs and Cartel Prosecution. International Journal of Industrial Organization 21: 347–79. Nocke, Volker, and Lucy White. 2007. Do Vertical Mergers Facilitate Upstream Collusion? American Economic Review 97: 1321–39. Organization for Economic Cooperation and Development (OECD) 2003. Hard Core Cartels: Recent Progress and Challenges Ahead. Paris: OECD. Phillips, Owen R., and Charles F.  Mason. 1992. Mutual Forbearance in Experimental Conglomerate Markets. Rand Journal of Economics 23: 395–414. Rey, Patrick, and Joseph E.  Stiglitz. 1995. The Role of Exclusive Territories in Producers’ Competition. Rand Journal of Economics 26: 431–51. Rojas, Christian. 2012. The Role of Demand Information and Monitoring in Tacit Collusion. Rand Journal of Economics 43(1): 78–109. Rotemberg, Julio J., and Garth Saloner. 1986. A Supergame-Theoretic Model of Business Cycles and Price Wars during Booms. American Economic Review 76: 390–407.

Cartels and Collusion   441

Salant, Steven W. 1987. Treble Damage Awards in Private Lawsuits for Price Fixing. Journal of Political Economy 98: 1326–36. Sannikov, Yulij, and Andrzej Skrzypacz. 2007. Impossibility of Collusion under Imperfect Monitoring with Flexible Production. American Economic Review 97(5): 1794–823. Spagnolo, Giancarlo. 1999. On Interdependent Supergames: Multimarket Contact, Concavity, and Collusion. Journal of Economic Theory 89: 127–39. Spagnolo, Giancarlo. 2004. Divide et Impera: Optimal Leniency Programs. CEPR Discussion Paper 4840. Available at www.cepr.org and www.ssrn.com. Stigler, George J. 1964. A Theory of Oligopoly. Journal of Political Economy 72: 44–61. Vasconcelos, Helder. 2005. Tacit Collusion, Cost Asymmetries, and Mergers. Rand Journal of Economics 36: 39–62. Whinston, Michael. 2006. Lectures on Antitrust Economics, Cambridge, MA: MIT Press.

CHAPTER 18

C A RT E L S A N D C OL LU SION Empirical Evidence MARGARET C. LEVENSTEIN AND VALERIE Y. SUSLOW

18.1. Introduction The incentive to coordinate behavior in order to increase profits is a powerful one. Despite the now widespread legal prohibitions on explicit coordination, firms continue to form cartels to restrict output or set prices. Cartels occur in a wide range of industries and they engage in a range of behaviors in their efforts to increase profits. They may set prices, rig bids, allocate markets or customers, make side payments to one another, and even develop elaborate structures for implementing and hiding their activities. In this chapter, we discuss the variety of techniques that cartels use to increase prices and profits. While we emphasize that cartels are pervasive across a variety of industries and are more stable than one might expect, they also require fairly elaborate techniques to achieve these goals. This suggests that tacit collusion, without communication, may be harder than one might otherwise presume. Empirical research has informed our understanding of the determinants of the formation, prevalence, and duration of cartels. Studies of national and international markets across the twentieth century find cartels in a variety of products and services, and these cartels typically last between five and eight years. The most important determinant of cartel breakup is effective antitrust policy. While it has often been presumed that cartels’ demise results from cheating by member firms tempted by short-term profits, empirical analysis suggests that cheating rarely destroys cartels. The potential profits from collusion provide sufficient incentives for cartels to develop creative ways to limit the temptations that inevitably arise.

Cartels and Collusion   443

While scholars and policymakers have often been concerned that business cycle downturns are associated with cartel formation, the evidence we review here does not suggest strong cyclical effects. There is evidence that cartels are formed during periods of falling prices, but these are more likely to be the result of market integration or an increase in competitive intensity than macroeconomic fluctuations. Similarly, cartel breakup does not evidence strong cyclicality. In the sections that follow we address cartel formation and breakup, how cartels raise prices, and the effects their actions have. We focus on recent empirical analyses that constitute the current state of knowledge on these topics. See Levenstein and Suslow (2006a) for a more comprehensive overview of the earlier literature. We conclude with a discussion of the effectiveness of antitrust and leniency policy.

18.2.  Cartel Prevalence Perhaps the most surprising thing about cartels is how pervasive they are. Over a century after the United States first adopted laws making price-fixing a felony, and two decades after the United States, the European Commission, and competition authorities around the world reached consensus that hard core cartels would not be tolerated, cartels continue to form. They form in local markets with relatively small firms, and they form in highly concentrated global markets dominated by multiproduct multinationals. Connor (2007) assembles evidence on the operation of cartels in 279 markets between 1888 and 2005. These include at least 57 that were operating legally and over 100 that had international membership.1 The Great Britain Board of Trade (1976) studied 125 cartels active in Britain prior to World War II. Many of these were international cartels. As is the case today, they operated primarily in sophisticated intermediate goods industries, including electrical and other manufacturing machinery, chemicals, coal and steel, textiles, paper, glass, and nonferrous metals (91–92). George Symeonidis found that in the UK in the 1950s, 36% of industries self-reported having collusive agreements; another 26% self-reported some form of coordination.2 A  comparable study of Finnish cartels found that about a third of all industries self-reported having a cartel.3 Fölster and Peltzman (2010) examine similar data for registered cartels in Sweden: “Around 1990, 1  Connor (2007) lists all the cartels in his sample along with industry, country, and some details regarding the cartel’s operation and prosecution (74, table 2; 136–53, table A1). 2  For his classification of individual industries, see Symeonidis (2002), table 3.1 and 69–71. For his mechanism for classifying industries as collusive, competitive or ambiguous, see 61–65. 3  Hyytinen, Steen, and Toivanen (2011) use a sample of legal cartels in Finland from 1951 to 1990 to estimate the prevalence of cartels during this period of government-sanctioned price-fixing. The sample covers 234 manufacturing industries: in 109 of these industries there was at least one cartel of national scope during the sample period. For the remainder of the sample, no cartel was registered, but it is unknown whether a cartel existed. They present a Hidden Markov Model that “consists of a hidden process and an observation process that reveals information on the hidden state of the industry for some periods, but not for others” (2).

444   Margaret C. Levenstein and Valerie Y. Suslow there were over one thousand cartel agreements registered, affecting about 15% of total sales of goods and services” (255). Audretsch (1989) reports that there were “more than 300 legal cartels in West Germany as of January 1987” (580). Posner’s (1970) landmark study of antitrust enforcement in the United States catalogued 989 cases involving a horizontal conspiracy between 1890 and 1969. Posner, like others who have since studied US price-fixing, used the Commerce Clearing House Trade Regulation Reporter (CCH) to draw a picture of the breadth of cartel activity in the United States. The CCH reports all antitrust convictions by the US Department of Justice (DOJ). Using CCH data from 1955 to 1997, Gallo et al. (2000) reports 688 horizontal per se violations. Bryant and Eckard (1991) use CCH data to estimate the probability of cartel detection; they report 184 price-fixing indictments between 1961 and 1988.4 Calculating a rough average of the number of domestic cartel cases per year, we find a range of 12 to 16 cases per year (using Posner’s and Gallo’s broader definition) and about 6.5 per year (from Bryant and Eckard). From 1992 to 2013 there were approximately 1040 DOJ cartel convictions, or about 47 convictions per year.5 It is important to note that in most cases a cartel results in more than one conviction, so this average of raw data from contemporary US cases is not comparable to those of earlier studies. However, it does indicate that cartels are still active in the US market despite increased enforcement. On average, cartels in the Bryant and Eckard sample lasted a little over five and a half years.6 Using comparable CCH data, the average duration of collusion of US price-fixing cases over the last 50 years is a little longer, about 6 years. While the mid-twentieth-century cases include some convictions of large, well-known firms, as in the famous GE-Westinghouse electrical equipment conspiracy of the 1950s, many of the convicted firms were relatively small and in geographically well-defined markets, such as road paving and school milk.7 While these markets are still prone to collusion, cartels consisting of multinational firms in global markets now capture headlines. International cartels are not a new phenomenon. Liefmann (1927), for example, documented 40 international cartel agreements from the late nineteenth century (129). Regarding the early twentieth century, Joel Davidow (1983) states: “In the pre–World War II era, cartels were as popular at the international level as they were at the national. Various studies estimated that during the 1930s, 30 to 40% of all world trade in manufactured goods was cartelized” (351). International cartels were probably less pervasive in the late twentieth century. Levenstein and Suslow (2004) estimate that $190 billion of global trade was affected by international cartels in 1997, representing less than 4% of all global trade (815). 4 

“We treat nolo contendere (“no contest”) pleas as equivalent to guilty pleas and convictions in constituting evidence of a conspiracy.” Bryant and Eckard (1991), 532. 5  Authors’ calculations from individual cases listed on US Department of Justice website. Antitrust Case Filings, US Department of Justice, accessed April 4, 2014, http://www.justice.gov/atr/cases. 6  Calculated from data provide by Bryant and Eckard (1991). 7  Lanzillotti (1996), Lee (1999), Porter and Zona (1999), and Scott (2000) all study cartels in public provision of school milk. Feinstein, Block, and Nold (1985), Porter and Zona (1993), and Gupta (2001) study price fixing in highway construction. For more on the electrical equipment conspiracy, see Bane (1973) and Baker and Faulkner (1993).

Cartels and Collusion   445

Many countries provide or have provided a mechanism for industries to register cooperative export agreements. These agreements sometimes, but not always, explicitly provide for cooperative price setting. They often include coordination and sharing of marketing and transportation costs.8 While these registries have largely disappeared along with their explicit exemptions from antitrust laws, hundreds of industries participated in such agreements during the twentieth century.9 For example, the US Webb-Pomerene Act of 1918 gives an antitrust exemption to firms that create voluntary associations for cooperation in export activities. Andrew Dick’s (1996a) study found 125 Webb-Pomerene Associations were created between 1920 and 1965. In 1975, Australia reported 69 registered export associations, Japan had 180 in 1973, and Germany had 227 in 1972.10 The United States had only approximately 30 export associations registered under the Webb-Pomerene Act by the 1970s. The Export Trading Company Act of 1982 expanded the scope of the permissible activities and limited potential liability, generating new interest in export exemptions. In 2002, there were 150 companies registered under this Act in the United States.11 Cartels are not only widespread across time and geography, they occur across a variety of products and sectors. Levenstein and Suslow (2011) assemble a sample of 81 international cartels prosecuted by the US Department of Justice or the European Commission between 1990 and 2010. Most of these were in intermediate manufactured goods and services and span a variety of industries: Forty percent are in chemicals, especially food additives. Another quarter are in a variety of other manufacturing industries, with multiple cartel convictions in steel, carbon and graphite products, plastics, and paper industries. Cartels were also found in specialized services, such as fine arts auctions and specialized tanker shipping. The only major sector that does not appear in the sample is final consumer goods. (Levenstein and Suslow 2011, 462–63)

There are some industries that seem particularly prone to collusive activity, across space and time. Levenstein and Suslow (2006a, table 3) discuss a number of historical examples of industries in which there are repeated episodes of collusion. Industries as varied as shipping and sugar, which are characterized by relatively high fixed costs, rarely seem to maintain stable competition. There are also firms that seem drawn to cooperative behavior.12 This may reflect firm or industry culture. For example, Spar (1994) argued 8  Dick (1996b) analyzes the difference between Webb-Pomerene Associations that focused on price and those that focused on joint marketing activities. 9  Levenstein and Suslow (2005) discuss changes in the laws underlying these registrations and explicit exemptions in 55 countries (801–17). 10  Levenstein and Suslow (2005), 793. Jacquemin, Nambu, and Dewez (1981) examine 545 Japanese export cartels between 1960 and 1970. See also Levenstein and Suslow (2006a), 50n. 11  Levenstein and Suslow (2005), table 1, 818–19. 12  For example, Degussa was convicted of price fixing by the European Commission for participation in cartels in methionine (1986–1999), methacrylate (1997–2002), organic peroxides (1971–1999), hydrogen peroxide and perborates (1994–2000), and vitamin B3 (1992–1998). See Levenstein and Suslow (2010), 34n.

446   Margaret C. Levenstein and Valerie Y. Suslow that the culture of cooperation in the early diamond industry facilitated collusion. In the early twentieth century United States, intense competition was sometime considered unethical because it deprived others of their livelihood (Levenstein 2012, 8). There are also contemporary industries whose culture and business practices appear to facilitate or encourage collusion. For example, in a post-2008 financial crisis municipal bond market case, industry observers claim that bid rigging was considered normal business practice: But the reason no one was whispering isn’t that their actions weren’t illegal—it’s because the bid rigging was so incredibly common the defendants simply forgot to be ashamed of it. “The tapes illustrate the cavalier attitude which the financial community brought toward this behavior,” says Michael Hausfeld, a renowned class-action attorney. . . . “It became the predominant mode of transacting business.” (Taibbi 2012, 82)

Recidivism may also be evidence of learning on the part of managers and firms. Participants in a cartel in one market may transfer their experience to other markets.

18.3.  Cartel Formation Economists’ theoretical models provide limited insight into why an industry that could profitably collude might not do so.13 The empirical literature on cartel formation is equally sparse.14 In general cartel formation is not publicly observed, and therefore is harder to measure than breakup. There are relatively few natural experiments that a researcher can use to test hypotheses about conditions under which cartels may be more likely to form. Changes in enforcement offer one such opportunity. There is evidence that changes in antitrust policy affect the incentive to form a cartel. For example, scholars have argued that the temporary suspension of antitrust enforcement during the period of the National Recovery Act in the United States (1933–1935) encouraged cartel formation (Alexander 1994; Krepps 1997). Conversely, increases in enforcement may discourage cartel formation. The most prominent example of the latter is the restructuring of leniency programs in the mid-1990s in the United States and the subsequent adoption of similar programs in numerous other countries around the world. The impact of leniency is nuanced and depends on policy design. Chen and Harrington (2007) demonstrate that while full 13  See, for example, Filson et al. (2001), Schmitt and Weder (1998), Bos and Harrington (2010), Eaton and Eswaran (1998), Hviid (1992), Kuipers and Olaizola (2008), Prokop (1999), and Thoron (1998). 14  See, for example, Filson et al. (2001) on formation of federally sanctioned agricultural cooperatives (“marketing orders”), Schmitt and Weder (1998) on cartel formation and sunk costs, Alexander (1994) and Krepps (1997) on the National Industrial Recovery Act, and Dick (1996a), testing a model of cartel formation on a Webb-Pomerene cartel data set.

Cartels and Collusion   447

leniency decreases the incentive to form cartels, since it makes defection more profitable, partial leniency combined with imperfect detection can provide cartels with a tool to punish defectors. It follows that partial leniency can increase the profitability of collusion and the likelihood of cartel formation.15 The essential question regarding cartel formation is why an industry that was not colluding yesterday would start colluding today. The literature often frames this as a question of the relationship between the business cycle and cartel formation: do firms that were not colluding start doing so during economic downturns? While cartels seem to form during periods of falling prices, the existing evidence suggests that formation occurs in response to increases in the intensity of competition. That is, what matters is the nature of competition in a particular market, rather than declines in the aggregate price level.16 Although data are limited, one can further explore this question of the possible link between cartel formation and business cycles. Levenstein and Suslow (2006a) include a meta-analysis of case studies of cartels that formed between 1857 and 1950 (67, table 10). These studies vary in their methodology and data; in order to have some consistency, we focus on 12 US cartels and compare to NBER business cycle reference dates for this subsample.17 We find that eight (67%) began in years during which there was a recession.18 Alone, one might draw the conclusion from this that cartels are much more likely to form during recessions. However, 63% of the years between 1857 and 1950 experienced recessions. Cartels in this subsample started disproportionately in years with recessions, but not excessively so, given that there were many recession years during this period.19 A similar analysis of the 71 cartel episodes in Suslow (2005)’s interwar sample (1918–1938) finds that 55% of the cartels began during months that were recessions. Over this period, 43% of the months were recession months. As in the sample of 12 case studies, cartels started disproportionately during recessions, but not overwhelmingly so. 15  Spagnolo (2000; 2008) makes a similar point that partial amnesty can increase the set of collusive equilibria. Thus, it is possible that leniency policies that reduce liability, but do not entirely eliminate it, facilitate collusion. See Aubert, Rey, and Kovacic (2006) for a proposal to improve leniency programs. 16  “Many studies report that a cartel was formed during a period of falling prices, but this is not always, or even usually, associated with falling demand (either for the particular product or in the general economy). Instead, falling prices were often the result of entry or the integration of previously distinct markets” (Levenstein and Suslow 2006a, 67). See also Levenstein and Suslow (2011). 17  It is important to note that cartel formation is often measured with error. We use the start date of the first known agreement between any two cartel members as reported by the competition authority in its enforcement action. In some cases, the cartel may have begun earlier, but the authorities do not have evidence of explicit communication. 18  NBER business cycle reference dates, from 1857 to the present, can be found at US Business Cycle Expansions and Contractions, National Bureau of Economics Research, September 20, 2010, www.nber.​ org/cycles.html. 19  The NBER dates the beginning of a recession to a particular month. For most cartels, we only know the year of cartel formation. Any cartel that began during a year when there was a recession is counted as “starting during a recession year” even though it may have started before or after the months of the recession.

448   Margaret C. Levenstein and Valerie Y. Suslow Turning to Bryant and Eckard’s (1991) sample of US horizontal price-fixing conspiracies that formed between 1932 and 1985, we find that 39% formed during years in which there was a recession.20 Thirty-nine percent of these years are designated by the NBER as experiencing a recession. This evidence does not indicate any relationship between cartel formation and business cycle downturns. For about a third of the sample, we also know the month or quarter in which the conspiracy began. For this subsample, about 25% of the cartels formed during months in which the US economy was in recession. NBER designates about 21% of the months between 1932 and 1985 as recessionary. Again, this admittedly rough characterization of the data does not suggest that cartels are more likely to form during recessions.21 Finally, we turn to Levenstein and Suslow’s (2011) sample of contemporary international cartels prosecuted by the US DOJ or the European Commission. For this chapter, we have updated the sample to include all convictions through the end of 2011. We find that only 11% of the cartels started during recessionary months. This undoubtedly reflects in part the dampening of the business cycle during the period 1971 to 2004, when these cartels were formed: a mere 13% of the months within this period were recessionary.22 For this sample in particular, it is important to recognize that start dates are negotiated with the authorities. The only cartels included in the data set are those that have been selected for us by the legal process; this is not likely to be a random selection of cartels. Because earlier reported start dates increase fines, firms have an incentive to negotiate for a later reported start date. Unless the authorities have conclusive evidence of cartel activity, they are more likely to settle on a later date and a guilty plea than risk a trial over an earlier start date. Because cartels do not form all at once, and because their formation is almost always furtive, reported start dates are almost surely later than actual start dates, on average, but it is unlikely that there is any systematic bias in measuring their formation relative to the business cycle. NBER business cycle reference dates capture economy-wide recessions, not sector-specific price fluctuations. Cartels do often form in response to falling prices for particular goods, but these are not necessarily linked to falling aggregate price levels associated with downturns in overall economic activity. Firms may have been responding to 20 

We are extremely grateful to Bryant and Eckard for their willingness to share their data with us. The finding is comparable to Andrew Dick’s (1996a) study of the formation of Webb-Pomerene Associations between 1920 and 1965. He finds no association between the timing of the formation of export association and fluctuations in export demand (215, table 4). He defines downturns as periods of declining prices in the broad sector in which the cartel is operating. As he explains in Dick (1996b), “I define business cycle timing based on U.S. export price movements. To assure exogeneity with respect to a cartel’s actions, business reference cycles are measured using one of four broad export price indexes-foods, crude materials, semifinished manufactures, or finished manufactures, depending on the cartelized product. . . . I define a dummy variable CYCLE that equals one if the cartel’s sector was in a peak-to-trough period (a downturn) . . . and equals zero if the sector was in a trough-to-peak period (an upturn)” (270–71). 22  Suslow’s (2005) and Levenstein and Suslow’s (2011) samples both consist of international cartels. NBER business cycle reference dates are for the US economy. These are likely highly correlated with the business cycles of the relevant countries, but certainly not perfectly. During the 1920s the US economy was probably stronger than many European economies, thus our estimate of the proportion of cartels formed during downturns might be an underestimate. 21 

Cartels and Collusion   449

falling prices, but prices were often falling because of events in a specific market. A common “explanation” for cartel formation is that some industry participant became more aggressive, cutting prices or going after others’ customers. In some cases this reflects an internal change in strategy on the part of an incumbent firm, such as a change in leadership. But in many cases, there is a clear underlying exogenous economic reason for this change in behavior such as entry, fluctuating exchange rates, or market integration. For example, European integration and broader globalization increased competitive intensity in many markets. In one illustrative case, hydrogen peroxide producers reported to European Commission investigators that their cartel formed following the expansion of Nordic and former East German producers.23 Similarly, the European Commission reported:  “In early 1986 Rhone-Poulenc and Degussa contacted Nippon Soda and Sumitomo because they felt that the Japanese [methionine] producers were encroaching on ‘their’ home markets. . . .”24 In a third example, ferry operators shipping freight across the English Channel formed a cartel after the United Kingdom withdrew from the European Monetary System. The depreciation of the pound led to falling real revenues for Continental firms, “requiring” a coordinated response on both sides of the Channel.25 There are also industries where cartels seem to form repeatedly. Following the meeting of the 2012 International Competition Network, Phil Evans wrote: “The cartel discussions . . . all appeared to point to a factor in cartel formation that is rarely talked about—the fact that some industries appear to have cultures of behaviour that appear to more easily step over into cartelisation, than others. This is one of the ‘softer issues’ that emerged during discussions that need exploring, but that the agencies themselves are unlikely to be in a particularly strong position to do so.”26 It would be useful to have research that analyzes characteristics of business or national culture that make cartel formation more likely. This could provide insight into the challenges to cartel formation, building on research on cartel formation and maintenance costs as discussed in Bradburd and Over (1982) and Alexander (1994). Such an analysis might also provide the basis for effective outreach and educational tools for competition authorities attempting to change business norms and encourage a culture of competition.

18.4.  How Do Cartels Set Prices? What kinds of agreements do cartels come to in order to raise prices? Sometimes they simply agree not to undercut one another. Other times they agree to specific prices, 23 

Commission Decision 2006/903/EC (Hydrogen Peroxide and Perborate), 2006 O.J. (L 353) 54. Commission Decision 2003/674/EC (Methionine), 2003 O.J. (L 255) 1. 25  Commission Decision 97/84/EC (Ferry Operators-Currency Surcharges), 1996 O.J. (L 026) 23. 26  Phil Evans, Reflections on ICN (Rio) 2012, Competition Online Forum, May 7, 2012, http://groups.​ yahoo.com/group/CompetitionOnlineForum. 24 

450   Margaret C. Levenstein and Valerie Y. Suslow which the authorities and the conspirators often refer to as target prices. They may only agree to a minimum price, so that it is permissible for firms to charge different prices. Under such an agreement, firms may still negotiate prices with their own customers. Cartel members may agree to specific percentage increases; in other cases, the agreements coordinate prices to limit secular declines in price. The latter occurs both for products that are at the end of the product life cycle (e.g., fax paper, videotapes) and products that have been affected by aggregate fluctuations in demand. Cartel members may agree to coordinate the timing of price announcements. In some cases, the timing of price announcements is intentionally manipulated to disguise collusive activity, with the initial price announcement rotating among cartel members (e.g., electrical carbon) or is intentionally staggered (e.g., MCAA). Cartels may engage in coordinated price discrimination, agreeing to vary prices across customers, geography, distribution channels, and the product portfolio (e.g., sorbates). Some cartels focus on the spread between prices in different markets in order to limit arbitrage. For example, cartels may respond to exchange rate fluctuations by adjusting prices across markets. While much of this behavior is similar to or the same as what one would predict for a monopoly firm, some is specific to cartels. In particular, cartels attempt to hide their coordination from customers and competition authorities. They also make agreements to engage in specific behaviors that facilitate monitoring of one another’s pricing by, for example, making public price announcements. Successful collusion requires extensive communication, both in private and via public signals.27 Cartels set parameters over the many dimensions in which they might compete. At times they agree to restrict or coordinate rebates, discounts, surcharges, transportation costs, or the provision of credit. The cartel may also agree to set the length of permissible contracts. The cartel may also agree to limit the length of permissible contracts. This prevents the temptation for a firm to undercut other firms by promising to supply at the current price for a longer period than the competition is offering. Where the product has multiple components that are combined to make a customer-specific product, collusion may be more difficult. This too can be overcome.28 As was the case in the 1950s US electrical equipment conspiracy, such cartels may publish a price list or a price book from which the collusive price can be constructed. For example, the electrical and mechanical carbon and graphite products cartel attempted to set prices in a market where there were “literally thousands of varieties. . . . Most often they are customer-designed.”29 They resolved this by 27  For more discussion of how cartel conspirators communicate see Harrington (2006) and Levenstein and Suslow (2006b). 28  In the 1920s, the US Federal Trade Commission encouraged firms to adopt uniform cost-accounting systems in order to facilitate tacit collusion in markets like these where complicated products were believed to lead firms to charge prices below their own costs. It was hoped that using similar cost accounting systems to price their products would dampen competition. Levenstein (1998), 35. 29  Commission Decision 2204/. . ./EC (Electrical and Mechanical Carbon and Graphite Products), 2003 http://ec.europa.eu/competition/antitrust/cases/dec_docs/38359/38359_36_1.pdf.

Cartels and Collusion   451

Table 18.1  What Do Cartels Do? Cartel mechanism

Frequency

Shares, regions, or customers were explicitly assigned to cartel members

80%

Sales information was exchanged for monitoring purposes

79%

Trade association facilitated agreement

31%

Cartel members agreed to and implemented compensation scheme

33%

Cartel took actions to exclude nonmembers Cartel took retaliatory actions against members who cheated

36% 19%

Source: Adapted from Levenstein and Suslow (2011), 471, table 3.

establishing a “basic material price” and adding charges for additional machining or components: [E]‌ach member would arrive in principle at exactly the same price increase, in relative terms, for each additional cubic centimetre of carbon used or for each additional screw or other tooling added. Then they multiplied by a “currency co-efficient” and “quantity co-efficient.” (Commission Decision, Electrical and Mechanical Carbon and Graphite Products)

In most cases, agreement on price is not sufficient. Thus, cartels generally assign output levels or market shares.30 In Levenstein and Suslow (2011), 80% of international cartels had some sort of market allocation (table 18.1). They may allocate geographic markets to members, but increasingly, in a twenty-first-century world of global supply chains, cartels allocate customers. This may make it harder to detect and enforce collusion on a national scale and highlights the importance of international coordination among antitrust authorities. Where necessary, cartels buy output from one another in order to maintain these quantity agreements. This is similar to the behavior of an efficient monopolist who reallocates production among plants. One difference is that while more efficient cartel members may be able to extract larger quotas from their coconspirators, there is rarely any systematic

30  Harrington and Skrzypacz (2011) provide a theoretical explanation of this common practice. Setting price and quantity is redundant in a world of perfect information, but not in the world of uncertainty in which these cartels operate.

452   Margaret C. Levenstein and Valerie Y. Suslow attempt to allocate production or output according to firm costs (Athey and Bagwell 2001). In Levenstein and Suslow (2011), about one-third of the cartels used a mechanism to compensate one another when one firm’s sales exceeded its cartel allocation (table 1). These agreements do not cover all dimensions along which cartel members could compete, but they certainly discuss and come to agreement on many of them. In some cases, where the cartel restricts sales in one geographical area, firms respond by competing in other parts of the world. Analogously, if a cartel restricts price and output, firms may compete in other dimensions, such as offering preferred terms or higher quality. Some of these other terms are much harder for cartels to limit and police. Cartels address these challenges by sharing information and monitoring one another, which virtually all cartels do (80% in table 18.1). In order to protect the cartel’s market share, they may agree to acquire competitors or to refrain from sharing technology with cartel outsiders (36% engaged in activity to exclude or eliminate noncartel producers, table 18.1). Agreement alone is not sufficient. Cartels engage in three types of behavior to ensure that their members adhere to the agreement. They enhance “the information that firms have about one another and the market; [compensate] one another when firms’ sales vary from assigned quotas due to factors outside of their control, such as random fluctuations in demand; and [punish] firms when violations do occur” (Levenstein and Suslow 2012, 458–59). We have found that compensation schemes are very effective—much more so than punishments—at keeping cartels together. Nineteen percent of the cartels in table 18.1 punished one another for violating cartel agreements, but these cartels tended to be less stable (Levenstein and Suslow 2012, 476). Harrington and Skrzypacz (2011) demonstrate that in markets with private quantities and prices, monitoring and compensation arrangements allow cartels to arrive at a collusive equilibrium. They do so by creating incentives for truthful reporting. These schema also create evidentiary trails that deserve attention from competition authorities. In particular, information sharing and interfirm sales, while having potential procompetitive justification, are shown both theoretically and in practice to be very useful collusive tools (Levenstein and Suslow 2006b). The most successful cartels do not simply agree to a certain set of parameters. They create organizations or governance structures that allow them to address challenges that arise sequentially, expand the scope of the agreement, and provide flexibility in changing economic conditions. For example, Genesove and Mullin (2001) describe collusion among sugar producers in the 1920s and 1930s: “meetings embodied a governance structure for the agreement, ensuring its adaptation to (typically endogenous) changing circumstances” (380). For a discussion of the variety of organizational structures that cartels used for self-governance in the late twentieth century, see Levenstein and Suslow (2006b). In the most sophisticated cartels, top-level management sets overall cartel strategy, but the inherent uncertainty of the economic environment requires ongoing communication and decision-making among operational employees.31

31  See Levenstein and Suslow (2006b), table 1. For rich descriptions of the inner workings of these cartels see Connor (2008) and Harrington and Skrzypacz (2011).

Cartels and Collusion   453

18.5.  How Do Cartels Affect Prices? While much attention has been given to the dynamic effects of collusion and cooperation, the price effects associated with collusion can have a sizable economic impact that should not be ignored. Unfortunately, objective estimates of the effects of cartels on prices (not to mention deadweight loss) are relatively few. In 1989 Cohen and Scheffman wrote that “there is a sparse . . . economics literature attempting to estimate the markup in price-fixing cases” (344).32 The situation has not changed dramatically in the intervening decades. Science has often been stymied because the only researchers with access to the data necessary to estimate counterfactual prices that would allow one to make objective estimates of price effects are researchers who are hired in contested court cases. These estimates are, quite reasonably, influenced by their intended purpose, but this undermines their usefulness in establishing the overall effects of cartels. For example, Cohen and Scheffman (1989) attempt to back out price effects from Block, Nold, and Sidak’s (1981) study reporting a 2.87% average settlement in seven class action cases in the mid-1970s (Cohen and Scheffman 1989, 345). They suggest that the 2.87% should be divided by three (for treble damages) and the number of years of each case, resulting in a low estimate of the price effects of collusion. But it is odd to use settlements to estimate price effects, rather than the other way around. In particular, the size of settlements may depend more on what is possible to prove in court than on economic effects. Werden (2008) provides a useful summary of the research on price effects. Much of this research has focused on bid-rigging in which there is repeated price formation and the winning (and sometimes losing) bids are often public information (see also Porter and Zona 1999). Werden cites six studies of bid-rigging conspiracies, which found that collusion raised prices between 6.5% and 30% (436). Froeb, Koyak, and Werden (1993) estimate an 18% to 30% price increase resulting from a conspiracy that targeted the US Defense Department’s purchases of frozen seafood. Bid-rigging seems to be particularly common in public procurement, in part because government transparency and anticorruption rules can make cartel enforcement easier. Transparency makes for good government, but it also makes it easier for cartel participants to detect, and therefore deter, defection. The problem is sufficiently pervasive that the US Justice Department initiated a training program for government officials to “detect and report collusive and fraudulent conduct” during the implementation of the 2009 American Recovery and Reinvestment Act program (Hammond 2009). Several competition agencies have been actively using econometric screens to detect price fixing. For example, Brazil used 32 

Cohen and Scheffman (1989), 344–45, report two studies estimating cartel effects: the corrugated container cartel allegedly raised prices between 7.8% and 26% (Finkelstein and Levenbach 1983) and the Washington baker cartel at least 15% (Federal Trade Commission 1967). Newmark (1988) argues that the bakers’ effect was closer to zero.

454   Margaret C. Levenstein and Valerie Y. Suslow screening to identify which of several hundred gasoline markets was most likely to have been the victim of collusive pricing (Ragazzo 2012).33 While noting that “studies of the effects of price fixing are far less common” (2008, 436), Werden concludes that “cartels prosecuted as felonies in the U.S. typically had substantial effects—at least 10% on average” (436). Connor and Lande (2005) argue that the 10% figure, which is used by the US Justice Department in calculating damages, underestimates the impact.34 Connor (2008) examines the price effects of several global cartels in food additives and related markets. He finds prices increases of between 12% and 28%.35 Levenstein, Sivadasan, and Suslow (2011) examine price effects for seven international cartels and find significant declines in prices following the breakup of each of the cartels selected for analysis.36 However, these cartels were selected for further analysis precisely because of the decline in the price of the product at the time of the cartel’s demise. Thus, one cannot extrapolate from these case studies to cartels generally. Levenstein, Sivadasan, and Suslow (2011) illustrate a more general point about research on cartels. Given the multiplicity of equilibria in oligopoly models, it is often most useful to generate stylized facts from meta-analysis of multiple case studies of individual markets. This is particularly true when studying price effects. It is impossible to estimate meaningfully the effects of collusion on price cross-sectionally. Sutton (1997, 67–68) makes this point and suggests alternative econometric techniques based on establishing “bounds” around possible outcomes consistent with particular forms of competition. The obvious limitation of this line of research is the lack of necessary data for most cartels. Most civil cases end with settlement agreements that protect the confidentiality of the data used to estimate cartel effects. Criminal cases in the United States are almost all resolved with a plea bargain with virtually no information made public. The European Commission has published a great deal of descriptive information on how cartels organize and operate. Unfortunately, competition authorities are not charged with making data on prices, quantities, or costs available for scientific research. Measurement of price effects is also complicated by the difficulty in estimating a counterfactual price. This is not only because prices are often hidden; It is also that cartels influence entry, the pace of technological change, and other characteristics of the market that 33  See also Mena-Labarthe (2012), who writes that “the Mexican Competition Commission has made some efforts to use screening to detect collusion and to prioritize investigation resources” (3). For advances in screening techniques, see Abrantes-Metz et al. (2006). For an overview of the potential of screens to detect explicit collusion, see Abrantes-Metz and Metz (2012). See also Bolotova, Connor, and Miller (2008). 34  See also Connor and Bolotova (2006), which examines 800 cartels over the last 250 years. They find an average price effect of 29% (1134). OECD (2002) notes the general lack of any measure of cartel harm: “Real-world data on actual harm is sparse” (9). The OECD does provide “estimates of harm” for 14 cartels for which such an estimate could be determined. These estimates range from 3% to 65% (9, par. 21; annex A). A useful and accessible summary of the effects of cartels can be found in Whinston (2008), 20–38. 35  Connor (2008) reports ranges for estimates of overcharges for vitamins (339), lysine (235), and citric acid (165). 36  The price is defined as the ratio of trade value to trade quantity for each specific product. The average price reported is for each product, year, and bilateral trade pair.

Cartels and Collusion   455

in turn affect price. Levenstein and Suslow (2011) found that over a third of contemporary international cartels engaged in behavior designed to restrict entry. For example, “members of the electrical carbon cartel refused to supply any graphite to an East German competitor that had entered the international market after unification” (472). Thus a counterfactual price, absent explicit collusion today, but given today’s industry structure, might not be the right counterfactual. Symeonidis (2002) is the exception that proves the rule in that his analysis does examine the endogeneity of industry structure. He is able to do this because there was a specific historical episode with an exogenous change in competition rules that allows him to identify the counterfactual industry structure. He uses this to infer the effects of collusion on productivity, advertising and research and development expenditures.37 But in even in this case, it is not possible to estimate counterfactual prices.

18.6.  What Causes Cartel Breakup? Since Stigler (1964), economists have focused on cheating as the most significant challenge to cartels, but historically, bargaining, coordination, and entry have been the most frequent causes of cartel breakup (Levenstein and Suslow 2006a, 45). With the emergence of a “hard core” cartels and the adoption of the leniency policies discussed above, antitrust action is responsible for most cartel breakup today: about 80% of the cartels in Levenstein and Suslow (2011) were ended by antitrust intervention. The precipitating cause of cartel breakup in this sample of contemporary international cartels is reported in table 18.2. There are several lessons to draw from this sample about the economic causes underlying cartel breakup. Our statistical analysis suggests two primary factors that increase the likelihood of cartel breakup: the financial fragility of a cartel member or the entry of a new producer (Levenstein and Suslow 2011, 485).38 Cartels are fairly savvy about the challenges that they face, including the oft-discussed incentive of members to cheat on cartel agreements. We find that cartels that have to punish their members are relatively unstable cartels. Many cartels suffer from a “little” cheating; this cheating does not result in punishment, let alone cartel death. Cartels use compensation mechanisms to limit the incentive to cheat as well to respond to variation in demand—which they know will occur. Cartels that use such compensation mechanisms are more likely to endure. Cartels that punish are frequently suffering from fundamental disagreements about how to divide markets or set 37 

For details on the impact on R & D and advertising see Symeonidis (2002). For productivity, profitability and wages, see Symeonidis (2007) and Symeonidis (2008). 38  This is consistent with the repeated game literature on cartels, but suggests that firm-specific discount rates are more relevant than market interest rates in determining cartel stability.

456   Margaret C. Levenstein and Valerie Y. Suslow Table 18.2  What Causes Cartel Breakup? Cause of breakup Antitrust death   Follow-on investigation   Customer complaint  Other sources (including whistle-blowers) Natural death   Amnesty application  Cheating   Growing fringe   Unknown cause of breakup All cartels

Number of cartels

Average duration (years)

13 7 29

8.8 4.0 8.2

17 6 7 2 81

10.3 7.7 6.4 4.5 8.1

Source: Adapted from Levenstein and Suslow (2011), 468, table 2.

prices. Pervasive and repeated violations of the terms of a cartel agreement do result in retaliatory punishments, but these punishments do not save the cartel. They undermine trust and lead to cartel breakup. (485)

As discussed above, substantial attention has been given to cartel stability over the business cycle. In Levenstein and Suslow (2011) we test for the impact of the business cycle on cartel breakup among international cartels. We find that “[n]‌either common knowledge measures of business cycles nor measures of unexpected shocks to demand appear to have any significant effect on cartel stability” (483). This contrasts with studies of earlier cartels that found significant effects of the Great Depression on cartel duration. This may reflect the greater severity and depth of business cycle fluctuations before World War II. However, Levenstein and Suslow (2006a) report that case studies, even during this earlier period, rarely find any role for macroeconomic fluctuations on cartel stability (67). Since 1993, when the DOJ revised and expanded its amnesty policy to offer automatic amnesty to the first cartel member to confess voluntarily, leniency policies have been the primary tool of cartel detection and breakup.39 The European Commission adopted a similar policy in 1996 and strengthened it in 2002.40 Both the DOJ and the EC have increased the severity of penalties, with multiple fines above the DOJ’s older 39 

The DOJ has had a corporate amnesty program since 1978, but the earlier program was ambiguous and ineffective. See Corporate Leniency Policy, Department of Justice Antitrust Division, August 10, 1993, http://www.justice.gov/atr/public/guidelines/0091.htm; Individual Leniency Policy, Department of Justice Antitrust Division, August 10, 1994, http://www.justice.gov/atr/public/guidelines/0092.htm. Hammond (2009) states: “The revised Corporate Leniency Program has resulted in a surge in amnesty applications. Under the old policy, the Division obtained roughly one amnesty application per year. Under the new policy, the application rate has jumped to roughly two per month” (10). 40  Leniency Policy, European Commission, last updated April 16, 2012, http://ec.europa.eu/​ competition/cartels/leniency/leniency.html. There are other conditions required for full immunity,

Cartels and Collusion   457

statutory limit of $10 million and more frequent prison sentences in the United States. Multiple jurisdictions have expanded the opportunities for customers to undertake private actions, suing for damages caused by cartels. The number of cartel prosecutions has increased dramatically since the adoption of these policies. Despite almost two decades of more active enforcement, the rate of cartel discoveries seems not to have abated.41 As discussed above, theory suggests that leniency may have positive, insignificant, or even perverse negative effects on the probability of cartel breakup. Spagnolo (2000, 2007) demonstrates that partial amnesty can be used as a threat to prevent cheating, which could facilitate collusion.42 Thus, the existence of private damage suits limits the effectiveness of leniency policies. While these suits bring sizable private resources to bear in the anticartel fight, those resources are primarily directed toward known cartels, not new cartel detection. These private damage suits may increase deterrence by increasing expected penalties, but they may create perverse effects if they allow cartels to use the threat of reporting a cartel to authorities as a tool of cartel discipline. This suggests that deterrence might be better enhanced by increasing individual penalties, including jail terms, rather than creating or expanding private antitrust enforcement. Harrington and Chang (2009) distinguish between the impact of leniency on detection of existing cartels and deterrence of future cartels. The continuing stream of confessions suggests that while leniency policies may have increased deterrence, many firms still find it profitable to form cartels. There is a small but growing empirical literature aimed at evaluating the impact of leniency policy. Levenstein and Suslow (2011) find that the average duration of international cartels ending with amnesty application is 10.3 years, somewhat longer than average cartel duration (see table 18.2). Disentangling the effects of leniency policy is methodologically very challenging, but those studies that have attempted it have found that leniency is effective. Miller (2009) examines all US Department of Justice cartel indictments between 1985 and 2005. He finds “that leniency enhances deterrence and detection capabilities [leading to] a 59% lower cartel formation rate and a 62% higher cartel detection rate.  . . .”43 Even a successful leniency program is likely to discover the least profitable cartels. For example, when Rhône-Poulenc requested amnesty for its participation in the methionine cartel, it continued to participate and hide its participation in the methylglucamine cartel. The methionine cartel was at risk of collapse due to entry from a large, such as immediately ending the infringement and not serving as the cartel ring leader. See also Directorate-General for Competition, European Commission (2003), 32nd Report on Competition Policy 2002, pars. 19–20, Belgium: Office for Official Publications of the European Communities, http://​ ec.europa.eu/competition/publications/annual_report/2002/en.pdf. 41  See, for example, Joseph Wayland’s 2012 speech, “Litigation in the Antitrust Division,” reporting that the US Department of Justice has “obtained more than $2 billion in criminal fines and more than 88,000 days of jail time for criminal defendants since the start of 2009” (6). 42  For further discussion of strategic responses to amnesty, see Harrington and Chang (2009) and Miller (2009). 43  Brenner (2009) examines 53 EC cartel cases from 1990 to 2003 and concludes that “the program provides incentives to reveal information on criminal activities,” but he was unable to show conclusively whether leniency increased cartel breakup or reduced the incentive to collude (639).

458   Margaret C. Levenstein and Valerie Y. Suslow well-established firm. The methylglucamine cartel was a successful duopoly with no imminent threat to its stability.44 This suggests that while leniency is a useful tool, it is not sufficient. The most profitable cartels will not be destabilized. While the optimal level of deterrence is presumably not perfect deterrence, the continuing stream of cartel confessions suggests that current policy should be strengthened. This is particularly challenging in current economic conditions. If anything, there has been pressure on competition authorities to reduce fines so as not to bankrupt fragile firms (Levenstein and Suslow 2010). It may be that punishments that target individual managers with both fines and imprisonment or removal from corporate leadership (e.g., industry or leadership disbarment) will be more effective than simply increasing corporate fines (Ginsburg and Wright 2010). Other tools targeting corporate behavior and structure, such as consent agreements that provide ongoing monitoring or enhanced scrutiny of postcartel mergers, may also be necessary.45 It is possible that leniency generates so many confessions that it consumes the resources of competition authorities to the neglect of other methods for detecting cartels. If it is the most successful and stable cartels that leniency misses, it is important that the authorities have both the resources and the strategic vision to continue active detection beyond leniency. This may be particularly challenging in a resource-constrained environment when it appears that leniency is doing all that is necessary.

18.7. Conclusion As with examinations of any kind of criminal activity, empirical research on cartels is challenging. Cartel members hide their activities. Legal authorities, balancing numerous demands in deciding how much information to make public, suppress much of what they learn about cartels. Despite this, creative research strategies have been employed to establish some basic stylized facts about cartels. On average cartels last between five and eight years. The variance in cartel duration is high, but this average holds across time, space, and industries. The incentive to cheat, rather than causing cartel breakdown, leads to extensive communication and the adoption of sophisticated countervailing organizational mechanisms. Cartels are pervasive. They are probably less pervasive than in previous centuries, but they still operate in many markets. Cartels form in response to declining prices, but do not form disproportionately during business cycle downturns. Similarly, cartel breakup is remarkably free of cyclicality, especially since the Great Depression.

44 

For details regarding these two cartels, see Levenstein and Suslow (2011), 466–67. There is an emerging literature examining post-cartel mergers. For example, see Davies, Ormosi, and Graffenberger (2014), Hüschelrath and Smuda (2012), and Marx and Zhou (2014). 45 

Cartels and Collusion   459

Firms attempt to collude along as many dimensions as firms attempt to compete. These include prices, quantities, market shares, contract duration, credit terms, transportation costs, and access to technology. Understanding how cartels actually set prices can inform the design of more effective techniques for detecting collusion. For example, some cartels set customer-specific prices. This implies that looking for uniform prices may be an unsatisfactory screen for collusion in markets with large or concentrated customers. On the other hand, some cartels strategically reduce price variance across markets, so changes in price spreads may be a meaningful signal of active cartel efforts. While some individual cartels do not succeed in raising prices, virtually all studies find that, on average, they raise prices to some extent. Because the US DOJ has long used 10% as a guideline in its sentencing, many studies implicitly or explicitly ask whether this is a good estimate of cartel effect. Most studies find effects at least that large, but there are surprisingly few rigorous studies of price effects because data availability is severely limited. Antitrust is the most important force leading to cartel breakup. The adoption of amnesty and leniency policies, combined with dramatic increases in the largest fines, has led to the demise of many cartels that had negatively affected global markets. However, there are limitations to the effectiveness of these policies as currently designed. Cartels continue to form, suggesting that deterrence may still be insufficient. Leniency policy is also more likely to catch cartels that are generating limited profits at the time they are revealed. Given scarce resources, it is important not to let leniency policies crowd out other anticartel enforcement tools.

References Abrantes-Metz, Rosa, Luke M.  Froeb, John F.  Geweke, and Christopher T.  Taylor. 2006. A Variance Screen for Collusion. International Journal of Industrial Organization 24:467–86. Abrantes-Metz, Rosa, and Albert Metz. 2012. How Far Can Screens Go in Distinguishing Explicit from Tacit Collusion? New Evidence on the LIBOR Setting. Antitrust Chronicle 3(1): 1–9. Alexander, Barbara. 1994. The Impact of the National Industrial Recovery Act on Cartel Formation and Maintenance Costs. Review of Economics and Statistics 76:245–54. Athey, Susan, and Kyle Bagwell. 2001. Optimal Collusion with Private Information. Rand Journal of Economics 32:428–465. Aubert, Cėcile, Patrick Rey, and William E.  Kovacic. 2006. The Impact of Leniency and Whistle-Blowing Programs on Cartels. International Journal of Industrial Organization 24:1241–66. Audretsch, David B. 1989. Legalized Cartels in West Germany. Antitrust Bulletin 34:579–600. Baker, Wayne E., and Robert R. Faulkner. 1993. The Social Organization of Conspiracy: Illegal Networks in the Heavy Electrical Equipment Industry. American Sociological Review 58:837–60. Bane, Charles A. 1973. The Electrical Equipment Conspiracies:  The Treble Damage Actions. New York: Federal Legal Publications.

460   Margaret C. Levenstein and Valerie Y. Suslow Block, Michael Kent, Frederick Carl Nold, and Joseph Gregory Sidak. 1981. The Deterrent Effect of Antitrust Enforcement. Journal of Political Economy 89:429–45. Bolotova, Yuliya, John M. Connor, and Douglas J. Miller. 2008. The Impact of Collusion on Price Behavior: Empirical Results from Two Recent Cases. International Journal of Industrial Organization 26:1290–307. Bos, Iwan, and J. E.  Harrington Jr. 2010. Endogenous Cartel Formation with Heterogeneous Firms. Rand Journal of Economics 41:92–117. Bradburd, Ralph, and Mead Over Jr. 1982. Organizational Costs, “Sticky Equilibria,” and Critical Levels of Concentration. Review of Economics and Statistics 64:50–58. Brenner, Steffen. 2009. An Empirical Study of the European Corporate Leniency Program, International Journal of Industrial Organization 27:639–45. Bryant, Peter G., and E. Woodrow Eckard. 1991. Price Fixing: The Probability of Getting Caught. Review of Economics and Statistics 73:531–36. Chen, Joe, and Joseph E. Harrington Jr. 2007. The Impact of the Corporate Leniency Program on Cartel Formation and the Cartel Price Path. In Vivek Ghosal and Johan Stennek, The Political Economy of Antitrust, 59–80. Boston: Elsevier. Cohen, Mark A., and David T.  Scheffman. 1989. The Antitrust Sentencing Guideline:  Is the Punishment Worth the Costs? American Criminal Law Review 27:331–66. Connor, John M. 2008. Global Price Fixing. 2nd ed. Berlin: Springer. Connor, John M. 2007. Price-Fixing Overcharges: Legal and Economic Evidence in Research. In Richard O. Zerbe and John B. Kirkwood. Law and Economics, vol. 22, 59–154. Boston: Elsevier. Connor, John M., and Yuliya Bolotova. 2006. Cartel Overcharges: Survey and Meta-Analysis. International Journal of Industrial Organization 24:1109–37. Connor, John M., and Robert H. Lande. 2005. How High Do Cartels Raise Prices? Implications for Optimal Cartel Fines. Tulane Law Review 80:513–69. Davies, Stephen, Peter L. Ormosi, and Martin Graffenberger. 2014. Mergers after Cartels: How Markets React to Cartel Breakdown. Centre for Competition Policy Working Paper 14-1. Available at http://competitionpolicy.ac.uk/documents/107435/107587/14-1+complete+v2. pdf/3661a080-30bd-490e-8bc0-ea6e0c1fd62d. Davidow, Joel. 1983. Cartels, Competition Laws and the Regulation of International Trade, New York University Journal of International Law and Politics 15:351–76. Dick, Andrew R. 1996a. Identifying Contracts, Combinations and Conspiracies in Restraint of Trade. Managerial and Decision Economics 17:203–16. Dick, Andrew R. 1996b. When Are Cartels Stable Contracts? Journal of Law and Economics 39:241–83. Eaton, Curtis, and Mukesh Eswaran. 1998. Endogenous Cartel Formation. Australian Economic Papers 37:1–13. Federal Trade Commission. 1967. Economic Report on the Baking Industry. Washington, DC: For sale by the Supt. of Docs., U.S. Govt. Print. Off. Feinstein, Jonathan S., Michael K. Block, and Frederick C. Nold. 1985. Asymmetric Information and Collusive Behavior in Auction Markets. American Economic Review 75:441–60. Filson, Darren, Edward Keen, Eric Fruits, and Thomas Borcherding. 2001. Market Power and Cartel Information: Theory and an Empirical Test. Journal of Law and Economics 44:465–80. Finkelstein, Michael O., and Hans Levenbach. 1983. Regression Estimates of Damages in Price-Fixing Cases. Law and Contemporary Problems 46(4): 145–69. Fölster, Stefan, and Sam Peltzman. 2010. Competition, Regulation, and the Role of Local Government Policies in Swedish Markets. In Richard B.  Freeman, Birgitta Swedenborg,

Cartels and Collusion   461

and Robert Topel, Reforming the Welfare State: Recovery and Beyond in Sweden, 253–84. Chicago: University of Chicago Press. Froeb, Luke M., Robert A. Koyak, and Gregory J. Werden. 1993. What Is the Effect of Bid-Rigging on Prices? Economics Letters 42:419–23. Gallo, Joseph C., Kenneth Dau-Schmidt, Joseph L.  Craycraft, and Charles J.  Parker. 2000. Department of Justice Antitrust Enforcement, 1955–1997:  An Empirical Study. Review of Industrial Organization 17:75–133. Genesove, David, and Wallace P. Mullin. 2001. Rules, Communication, and Collusion: Narrative Evidence from the Sugar Institute Case. American Economic Review 91(3): 379–98. Ginsburg, Douglas, and Joshua Wright. 2010. Antitrust Sanctions. Competition Policy International 6(2): 3–32. Great Britain Board of Trade. 1976. Survey of International Cartels and Internal Cartels, 1944 and 1946. London: Central Library, Department of Industry. Gupta, Srabana. 2001. The Effect of Bid Rigging on Prices: A Study of the Highway Construction Industry. Review of Industrial Organization 19:453–67. Hammond, Scott D. 2009. Follow the Money: An Update on Stimulus Spending. Transparency and Fraud Prevention, Statement, US Senate Committee on Homeland Security and Governmental Affairs, Washington, DC. http://www.justice.gov/atr/public/testi​ mony/250274.htm. Harrington, Joseph E., Jr. 2006. How Do Cartels Operate? Foundations and Trends in Microeconomics 2(1): 1–105. Harrington, Joseph E., Jr., and Myong-Hun Chang. 2009. Modeling the Birth and Death of Cartels with an Application to Evaluating Competition Policy. Journal of the European Economic Association 7:1400–1435. Harrington, Joseph E., Jr., and Andrzej Skrzypacz. 2011. Private Monitoring and Communication in Cartels: Explaining Recent Collusive Practices. American Economic Review 101:2425–49. Hüschelrath, K., Smuda, F., 2012. Do Cartel Breakdowns Induce Mergers? Evidence from EC Cartel Cases. European Competition Journal 9, 407-429. Hviid, Morten. 1992. Endogenous Cartel Formation with Private Information. Canadian Journal of Economics 25:972–82. Hyytinen, Ari, Frode Steen, and Otto Toivanen. 2011. Cartels Uncovered. Working Paper, Department of Managerial Economics, Strategy and Innovation, KU Leuven, Belgium. March 11. http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1793665. Jacquemin, Alexis, Tsuruhiko Nambu, and Isabelle Dewez. 1981. A Dynamic Analysis of Export Cartels: The Japanese Case. Economic Journal 91:685–96. Krepps, Matthew B. 1997. Another Look at the Impact of the National Industrial Recovery Act on Cartel Formation and Maintenance Costs. Review of Economics and Statistics 79:151–54. Kuipers, J., and N. Olaizola. 2008. A Dynamic Approach to Cartel Formation. International Journal of Game Theory 37:397–408. Lanzillotti, Robert F. 1996. The Great School Milk Conspiracies of the 1980s. Review of Industrial Organization, 11:413–58. Lee, In K. 1999. Non-cooperative Tacit Collusion, Complementary Bidding and Incumbency Premium. Review of Industrial Organization 15:115–34. Levenstein, Margaret C. 1998. Accounting for Growth: Information Systems and the Creation of the Large Corporation, Palo Alto, CA: Stanford University Press. Levenstein, Margaret C. 2012. Escape from Equilibrium:  Thinking Historically about Firm Responses to Competition. Enterprise and Society. Published online, doi: 10.1093/es/khs025.

462   Margaret C. Levenstein and Valerie Y. Suslow Levenstein, Margaret C., Jagadeesh Sivadasan, and Valerie Y.  Suslow. 2011. The Effect of Competition on Trade Patterns:  Evidence from the Collapse of International Cartels. IPC Working Paper Series No. 114, International Policy Center, Gerald R. Ford School of Public Policy, University of Michigan, Ann Arbor. Levenstein, Margaret C., and Valerie Y.  Suslow. 2005. The Changing International Status of Export Cartel Exemptions. American University International Law Review 20:785–829. Levenstein, Margaret C., and Valerie Y. Suslow. 2006a. What Determines Cartel Success? Journal of Economic Literature 44(1): 43–95. Levenstein, Margaret C., and Valerie Y. Suslow. 2006b. Cartel Bargaining and Monitoring: The Role of Information Sharing. In The Pros and Cons of Information Sharing. Stockholm: Lenanders Grafiska AB. Levenstein, Margaret C., and Valerie Y. Suslow. 2010. Constant Vigilance: Maintaining Cartel Deterrence during the Great Recession. Competition Policy International 6(2): 145–62. Levenstein, Margaret C., and Valerie Y. Suslow. 2011. Breaking Up Is Hard to Do: Determinants of Cartel Duration. Journal of Law and Economics 54:455–92. Liefmann, Robert. 1927. International Cartels. Harvard Business Review 5:129–47. Marx, Leslie M. and Zhou, J., 2014 The Dynamics of Mergers among (Ex)Co-Conspirators in the Shadow of Cartel Enforcement (March 26, 2014). Available at SSRN: http://ssrn.com/ abstract=2410948 or http://dx.doi.org/10.2139/ssrn.2410948 Mena-Labarthe, Carlos. 2012. Mexican Experience in Screens for Bid-Rigging. Antitrust Chronicle 3(1): 1–8. Miller, Nathan H. 2009. Strategic Leniency and Cartel Enforcement. American Economic Review 99:750–68. Newmark, Craig M. 1988. Does Horizontal Price Fixing Raise Price? A Look at the Bakers of Washington. Journal of Law and Economics 31:469–84. Organization for Economic Cooperation and Development (OECD). 2002. Report on the Nature and Impact of Hard Core Cartels and Sanctions against Cartels under National Competition Laws. DAFFE/COMP(2002)7. Porter, Robert H., and Douglas J. Zona. 1993. Detection of Bid Rigging in Procurement Auctions. Journal of Political Economy 101:518–38. Porter, Robert H., and Douglas J. Zona. 1999. Ohio School Milk Markets: An Analysis of Bidding. Rand Journal of Economics 30:263–88. Posner, Richard. 1970. A Statistical Study of Antitrust Enforcement. Journal of Law and Economics 13(2): 365–419. Prokop, Jacek. 1999. Process of Dominant-Cartel Formation. International Journal of Industrial Organization 17:241–57. Ragazzo, Carlos. 2012. Screens in the Gas Retail Market: The Brazilian Experience. Antitrust Chronicle 3(1): 1–8. Schmitt, Nicolas, and Rolf Weder. 1998. Sunk Costs and Cartel Formation: Theory and Application to the Dyestuff Industry. Journal of Economic Behavior and Organization 36:197–220. Scott, Frank A., Jr. 2000. Great School Milk Conspiracies Revisited. Review of Industrial Organization 17:325–41. Spagnolo, Giancarlo. 2000. Self-Defeating Antitrust Laws:  How Leniency Programs Solve Bertrand’s Paradox and Enforce Collusion in Auctions. FEEM Working Paper No. 52.2000, Department of Economics and Institutions, University of Rome. http://dx.doi.org/10.2139/​ ssrn.236400.

Cartels and Collusion   463

Spagnolo, Giancarlo. 2008. Leniency and Whistleblowers in Antitrust. In Paolo Buccirossi, Handbook of Antitrust Economics, 259–304. Cambridge, MA: MIT Press. Spar, Debora L. 1994. The Cooperative Edge: The Internal Politics of International Cartels. Cornell Studies in Political Economy. Ithaca, NY: Cornell University Press. Sutton, John. 1997. Game-Theoretic Models of Market Structure. In David M. Kreps and Kenneth F. Wallis, Advances in Economics and Econometrics: Theory and Applications. Seventh World Congress, vol. 1, 66–86. Cambridge: Cambridge University Press. Symeonidis, George. 2002. The Effects of Competition: Cartel Policy and the Evolution of Strategy and Structure in British Industry. Cambridge, MA: MIT Press. Symeonidis, George. 2007. Price Competition, Innovation and Profitability:  Theory and UK Evidence. In M. C. Levenstein and S. W. Salant, Cartels. The International Library of Critical Writings in Economics Series, vol. 2, 612–36. London: Edward Elgar. Symeonidis, George. 2008. The Effect of Competition on Wages and Productivity: Evidence from the United Kingdom Review of Economics and Statistics 90:134–46. Taibbi, Matt. 2012. The Scam Wall Street Learned from the Mafia: How America’s Biggest Banks Took Part in a Nationwide Bidrigging Conspiracy—until They Were Caught on Tape. Rolling Stone, July 5, 80–88. Thoron, Sylvie. 1998. Formation of a Coalition-Proof Stable Cartel. Canadian Journal of Economics 31:63–76. Wayland, Joseph H. 2012. Litigation in the Antitrust Division. Remarks as Prepared for the Georgetown Law Sixth Annual Global Antitrust Enforcement Symposium, Washington DC. http://www.justice.gov/atr/public/speeches/287117.pdf. Werden, Gregory J. 2008. Assessing the Effects of Antitrust Enforcement in the United States. Economist 156:433–51. Whinston, Michael D. 2008. Lectures on Antitrust Economics. Cambridge, MA: MIT Press.

CHAPTER 19

TAC I T C O L LU SIO N I N OL IG OP OLY EDWARD J. GREEN, ROBERT C. MARSHALL, AND LESLIE M. MAR X

19.1. Introduction In this chapter, we examine the economics literature on tacit collusion in oligopoly markets and take steps toward clarifying the relation between tacit collusion in the economics and legal literature. Economists distinguish between tacit and explicit collusion. Lawyers, using a slightly different vocabulary, distinguish between tacit coordination, tacit agreement, and explicit collusion. In hopes of facilitating clearer communication between economists and lawyers, in this chapter we attempt to provide a coherent resolution of the vernaculars used in the economics and legal literature regarding collusion.1 Perhaps the easiest place to begin is to define explicit collusion. It is our understanding that both the economics and legal professions use the term “explicit collusion” to mean an agreement among competitors that relies on interfirm communication and/or transfers to suppress rivalry.2 In the United States, Section 1 of the Sherman Act makes explicit collusion illegal, forbidding agreements that unreasonably restrain competition and affect interstate commerce.3 1 

It is important to note that there is substantial debate within the legal profession on the meaning of these terms, as most recently noted by Kaplow (2011). See also Kaplow and Shapiro (2007) and Kaplow (2013). In this chapter, in order to take steps toward clarifying communication between economists and lawyers, we opt for specific definitions of terms but recognize that there remain both substantial debate among legal scholars and latitude for interpretation by courts regarding these definitions. 2  As we discuss legal concepts it should always be understood that it is our interpretation as economists, not as legal scholars, since we are not the latter. 3  See Turner (1962). The Sherman Act states: “Every contract, combination in the form of trust or otherwise, or conspiracy, in restraint of trade or commerce among the several States, or with foreign nations, is declared to be illegal” (15 USC § 1).

Tacit Collusion in Oligopoly   465

As described in the seminal work of Stigler (1964), the key problem faced by firms attempting to collude is the need to deter secret deviations. The successful suppression of rivalry, elevating prices and restricting output relative to what it otherwise would be, creates incentives for secret price cutting by the firms. Thus, as described by Stigler, in order to successfully collude, firms must put in place collusive structures to govern the interaction among the colluding firms and between the colluding firms and other market participants. These necessary structures include pricing, allocation, and enforcement structures.4 One would typically expect that the establishment and implementation of these structures would require communication and possibly transfers among the colluding firms, in which case the conduct would fall under the heading of explicit collusion. However, firms in an oligopoly can be expected to recognize their mutual interdependence in the market. Each firm realizes that its profits depend not only on its own actions, but also on the actions of its rivals. It is possible that firms, each possessing this insight and understanding that its competitors all possess it, might be able to succeed in the implementation or even the establishment of a collusive agreement without communication. There has been broad agreement in principle that monopoly conduct can arise spontaneously in highly concentrated markets that satisfy some other (possibly restrictive) conditions. But, because economists have not yet been able to characterize those conditions with full confidence and precision, there has been room for courts to vary from one another. Since the 1980s, legal scholars have couched essentially this point in the language of an “extra ingredient” of centralization. As stated in Kovacic and coauthors (2011): In highly concentrated markets, the recognition of interdependence can lead firms to coordinate their conduct simply by observing and reacting to their competitors’ moves. In some instances, such oligopolistic coordination yields parallel behavior (e.g., parallel price movements) that approaches the results that one might associate with a traditional agreement to set prices, output levels, or other conditions of trade. The line that distinguishes tacit agreements (which are subject to section 1 scrutiny) from mere tacit coordination stemming from oligopolistic interdependence (which eludes section 1’s reach) is indistinct. The size of the safe harbor that Theatre Enterprises recognized depends on what conduct courts regard as the extra ingredient of centralized orchestration of policy which will carry parallel action over the line into the forbidden zone of implied contract and combination. [Schwartz, Flynn, and First, 1983] Courts enjoy broad discretion to establish the reach of section 1 by defining this extra ingredient broadly or narrowly. (Kovacic et al., 2011, 405)

4  For elaboration on collusive structures, see Marshall and Marx (2012, c ­ hapter 6). Levenstein and Suslow (in this volume) state, “While it has often been presumed that cartels’ demise results from cheating by member firms tempted by short-term profits, empirical analysis suggests that cheating rarely destroys cartels. The potential profits from collusion provide sufficient incentives for cartels to develop creative ways to limit the temptations that inevitably arise.”

466   Edward J. Green, Robert C. Marshall, and Leslie M. Marx In this chapter, we offer a way to distinguish between the legal profession’s use of tacit coordination and tacit agreement as in the above quote and to reconcile those with notions from the economics literature. In order to parse the language of the economics profession and the language of the legal profession, it will be useful to recognize that in order to collude, firms have to solve two broad problems: how to initiate a collusive arrangement and how to implement that arrangement. As described by Isaac and Plott (1981), “First an opportunity for conspiracy must exist. . . . The opportunity to conspire must be followed by an attempt to conspire. The attempt to conspire must be followed by an actual conspiracy and the resulting conspiracy must have an impact on the market” (Isaac and Plott, 1981, 1). The problem of initiating collusion involves coming to agreement on what the collusive structures required to deter secret deviations will be. This includes coming to agreement (reaching at least mutual knowledge, and perhaps common knowledge) regarding the mechanism for elevating prices, how the rents from collusion will be split among the firms, and how deviations will be detected and deterred. The problem of implementation involves managing the ongoing operation of the collusive arrangement, including the implementation of the collusive structures. The pricing structures may require coordination of price increases, the allocation structures may require transfers among the firms to achieve the agreed division of the collusive gain and to ensure incentive compatibility of the arrangement, and the enforcement structures may require information collection to stay in sync about the environment and to maintain compliance. Solving the two broad problems of initiation and implementation may require explicit communication among the firms. This is particularly true in environments with strategic buyers. As noted by Isaac and Plott (1981, 2), “Markets have two sides, and those on the other side and not a party to the conspiracy may not passively acquiesce to the establishment of such a conspiracy. The market reactions might be such that the efforts of the conspirators are rendered ineffectual.” Collusion without communication at the implementation stage would require that the firms establish during the initiation stage a contingent agreement specifying reactions to observable outcomes during implementation,5 in particular specifying what outcome would trigger reversion to noncollusive behavior (or, on a temporary basis, to even more severe retaliation). Depending on the product, market, and industry, the set of contingencies may be too numerous or too complicated for there to be the absence of communication during implementation. However, in some environments, which we discuss below, the problem of initiation or implementation may be solvable without explicit communication or transfers. Table 19.1 illustrates our interpretation of how the economics and legal professions would classify coordination depending on whether there were communication or transfers and either the initiation or implementation stage. It is important to note that this is 5  In the context of illegal agreements, there is no distinction between verifiable and observable, unless we consider a third-party cartel organizer with quasi-judicial powers (see Marshall and Marx, 2012, ­chapter 6.6).

Tacit Collusion in Oligopoly   467

Table 19.1  An Interpretation of Economics versus Legal Terminology Communication/transfers Initiation

Implementation Economics

Law

1.

Yes or noa

Yes

Explicit collusion

Explicit collusion

2.

Yes

No

Explicit collusion

Tacit agreement

3.

No

No

Tacit (implicit) collusion

Tacit coordination

Unifying terminology Strong explicit collusion Weak explicit collusion Tacit coordination

a

An industry with a history of collusion that has resumed collusive conduct after a hiatus may not need communication at the initiation phase.

one interpretation, and that both some legal scholars and some courts may interpret the terms in different ways.6 As described in table 19.1, communication or transfers at the implementation stage is sufficient for either the economics or legal profession to classify the conduct as explicit collusion.7 The economics literature would also classify conduct as explicit collusion if there were only communication or transfers at the initiation stage; however, that is likely the case that the legal literature would term tacit agreement.8 With no communication or transfers at either the initiation or implementation stage,9 the economics literature would refer to the conduct as tacit collusion or implicit collusion (in what follows we use only tacit collusion), while the legal profession would likely refer to this 6 

“All concur that express agreements are a subset of interdependent behavior that counts, that is, triggers liability. There is, however, no sharp consensus either on the boundaries of this subset or on whether other subsets, such as one including tacit agreements, also suffice. These and other questions are interrelated. For example, if express agreements are defined broadly, to include what others might deem to be tacit agreements, then express agreements might be viewed as exhausting the space of interdependent behavior that suffices, supposing that the only other candidate behavior involves tacit agreements. Many, including the Supreme Court in both earlier decisions and its most recent. . ., do in fact state that tacit agreements are sufficient, yet it is hard to know what to make of these proclamations given the great ambiguity of the term” (Kaplow, 2011, 700). 7  Direct interaction that is disguised through the use of an intermediary, such as a trade association, would be included in the definition of direct interaction. “The communication among colluding sellers needed to insure successful price reporting may be indirect, in that it all proceeds through a trade association or statistical service, but it is none the less explicit” (Kaysen, 1951, 266). On this point, Kaysen (1951) cites the Maple Flooring Manufacturers Association discussed in 268 U.S. 563 (1925). 8  Kaplow (2011, 700) notes that there is no clear consensus in the law as to the illegality of a “tacit agreement.” It is our impression that this speaks to the lack of agreement regarding the definition of “tacit agreement” in the law. Again, we are posing one interpretation to take steps forward in clarifying communication between lawyers and economists. 9  With the understanding that transfers would require communication, in some cases we limit our description to requiring only an absence of communication.

468   Edward J. Green, Robert C. Marshall, and Leslie M. Marx as tacit coordination—it appears that tacit coordination in the legal profession refers to long-run recognized mutual interdependence among oligopolists that generates outcomes that exceed those that would be realized under myopic interaction (static Nash equilibrium) without any direct interfirm communication or transfers. Adding to the complexity of parsing terminology, some economics literature focuses only on the implementation stage of a collusive arrangement. In those cases, the outcome may be described in the economics literature as tacit collusion, even though it may be clear that communication would be required in order to initiate the arrangement. In these cases, the label “tacit” applies only to the implementation stage, not to the collusive arrangement as a whole. This was recognized by Green and Porter (1984). Green and Porter analyze a collusive arrangement that does not require communication at the implementation stage, but they recognize the need for communication at the initiation stage, saying: It is logically possible for this agreement to be a tacit one which arises spontaneously. Nevertheless, in view of the relative complexity of the conduct to be specified by this particular equilibrium and of the need for close coordination among its participants, it seems natural to assume here that the equilibrium arises from an explicit agreement. (Green and Porter, 1984, 89 n. 5)

In the remainder of the chapter, we continue this discussion and provide examples. In section 19.2, we consider when the economic environment might be such that collusive profits can be achieved without communication and, thus, when tacit coordination is sufficient to elevate profits versus when strong or weak explicit collusion would be required. In section 19.3, we discuss the evolution of the theory of explicit collusion in the economics literature, particularly as related to the question of under what circumstances explicit coordination among sellers would be required during the operation of the cartel in order to achieve prices above a competitive level. In section 19.4, we focus on the issue of coordination during the initiation stage. We conclude with section 19.5, which comments on the role of tacit coordination in antitrust litigation.

19.2.  Tacit Coordination versus Strong and Weak Explicit Collusion Antitrust practitioners turn to the industrial organization economics literature for guidance regarding the issues concerning initiation and implementation of collusive agreements among firms in an industry. In this light, we provide comments regarding the key economic results and link these results to the issues confronted by antitrust practitioners as they evaluate a product/market/industry for a potential collusive agreement.

Tacit Collusion in Oligopoly   469

In the economics literature, a standard folk theorem characterizes the set of equilibria of a repeated oligopoly game and shows that for sufficiently patient firms (or for sufficiently short delay between repetitions of the game), the set of equilibria includes strategy profiles that generate the monopoly outcome.10 A folk theorem says that, in some environment, the problem of operating the cartel can be solved without setting up any ongoing, centralized mechanism of coordination—there is no need for explicit accounting, settlement, and enforcement.11 However, folk theorems deal with the implementation of collusion, and have nothing to say about its initiation. The folk theorem itself does not address whether firms would choose to play the strategies that generate the monopoly outcome nor how firms might coordinate on those strategies. As stated in Ivaldi and coauthors (2003, 6), “While economic theory provides many insights on the nature of tacitly collusive conducts, it says little on how a particular industry will or will not coordinate on a collusive equilibrium, and on which one.” The economics literature addressing collusion without communication at the implementation stage typically considers repeated interaction among firms that allows the firms to maintain higher prices with the (possibly implied) threat that deviation would trigger retaliation. Retaliation mechanisms must be both credible and sufficient (see Ivaldi et al., 2003). For sufficiency, the reduced payoffs from the retaliation must be sufficiently large to deter deviations in the first place, and in order to be credible, it must be in the best interest of the firms to follow through on the retaliation following the observation of a deviation. The folk theorem literature typically assumes away the problem of cartel initiation by characterizing the set of Nash equilibria without communication in the implementation stage (see, for example, Fudenberg and Maskin, 1986). In a Nash equilibrium each player chooses a best response to the strategies of the other players, which means identifying the set of Nash equilibria means essentially identifying outcomes that could arise in the implementation stage if firms were to coordinate on strategies that support that outcome in the initiation stage. The gas station example of Carlton, Gertner, and Rosenfield (1997) fits within this framework—two firms compete by setting prices, where those prices are perfectly observable and can be adjusted instantaneously, and profit of each firm is determined by the two prices and a fixed demand curve. Carlton, Gertner, and Rosenfield (1997) 10  The first folk theorem (cf. Friedman, 1977) assumed perfect information and perfectly patient players. The next generation (e.g., Fudenberg and Maskin, 1986; Radner, 1986) assumed near perfect patience (i.e., a discount factor asymptotic to 1) and perfectly correlated information. The current state of the art (cf. Mailath and Samuelson, 2006) makes much less stringent assumptions about information, but continues be phrased in terms of theorems about the asymptotic approach to perfect patience. Current folk theorems strongly support the idea that a collusive arrangement can be operated in an environment that is not informationally rich if producers are sufficiently patient. Econometric studies (e.g., Porter’s [1983b] study of the JEC cartel), as well as information disclosed in court proceedings, show that various groups of producers have been sufficiently patient for successful operation. 11  For a discussion of the canonical model of weak explicit collusion in a repeated-game model of Bertrand price competition, see Choi and Gerlach in this volume.

470   Edward J. Green, Robert C. Marshall, and Leslie M. Marx note, in this environment, that one would not be surprised to find that tacit coordination (involving no communication) could support the monopoly outcome. Any deviations from monopoly pricing would be immediately observed and met by a response from the other firm. However, such equilibria can be sufficiently complex that it is difficult to believe that firms could coordinate on a particular outcome without communication.12 As described in Stigler (1964), the central problem facing a cartel is secret price cutting by cartel members, so effective collusion requires that firms establish collusive structures, including pricing, allocation, and enforcement structures, in order to avert secret deviations (see also Marshall and Marx, 2012, ­chapter 6). Thus, one would expect that collusion would at least require communication at the initiation stage in order to establish the necessary collusive structures, although it is possible the conspirators might not need further communication once the collusive structures are established.13 In the gas station example of Carlton, Gertner, and Rosenfield (1997), no round of preliminary communication was used before the implementation stage, but it is a rare circumstance when that could be effective. For example, there may be several equilibria that provide payoffs in excess of the static Nash equilibrium, and formation stage communication may be needed to coordinate on a given equilibrium. Given the absence of implementation phase communication, the firms may need to select an equilibrium that is not payoff maximizing in order to avoid secret deviations, such as hidden loyalty rebates. Retail gas stations in Canada have been accused of strong explicit collusion in recent years, suggesting that the incremental elevation of prices at the pump required interfirm communication at the implementation stage. Excerpts from the announcements of the price-fixing conspiracy investigations by the Canadian Competition Bureau are as follows: During the investigation, the Bureau uncovered evidence of agreements between competitors to fix the price at the pump where gasoline was sold to consumers. The evidence indicated that participants carried out the conspiracy mainly by phoning one another to agree on the price of gasoline and about the timing of price increases, contrary to the conspiracy provision, section 45 of the Competition Act.14 12  The equilibria in question include those in which certain market outcomes trigger a Nash-reversion punishment phase as in Porter (1983a) and Green and Porter (1984), and other equilibria involve more sophisticated stick-and-carrot strategies as in Abreu (1986) and Abreu, Pearce, and Stacchetti (1986, 1990). Athey, Bagwell, and Sanchirico (2004) assume firms observe the realized prices of their rivals. As long as this information is available without communication, these equilibria would be viewed as weak explicit collusion. 13  Levenstein and Suslow (in this volume) state, “The most successful cartels do not simply agree to a certain set of parameters. They create organizations or governance structures that allow them to address challenges that arise sequentially, expand the scope of the agreement, and provide flexibility in changing economic conditions. . . . In the most sophisticated cartels, top- level management sets overall cartel strategy, but the inherent uncertainty of the economic environment requires ongoing communication and decision-making among operational employees.” 14  Competition Bureau Announces New Price-Fixing Charges in Quebec Gasoline Cartel, July 15, 2010, http://www.competitionbureau.gc.ca/eic/site/cb-bc.nsf/eng/03261.html.

Tacit Collusion in Oligopoly   471

Today’s criminal charges and guilty pleas are the result of an extensive Bureau investigation that found evidence that gas retailers or their representatives in these local markets phoned each other and agreed on the price they would charge customers for gasoline. The Bureau’s investigation into potential price-fixing in the retail gasoline market continues in the Southeastern Ontario market.15

These investigations do not imply that there was no elevation of price relative to static Nash from tacit coordination, or even weak explicit collusion, prior to the explicit interfirm communication during the implementation phase. However, the investigations highlight that strong explicit collusion was thought to be incrementally profitable by the colluding firms relative to the tacit coordination or weak explicit collusion that they had functioned under prior to the use of implementation phase communication. Furthermore, the buyers in the Carlton, Gertner, and Rosenfield (1997) gas station example are not players in the game and so have no ability to take actions that might disrupt the ability of the two firms to maintain their tacit coordination. When buyers are players, they have an incentive to pursue strategies that disrupt equilibria that allow the sellers to capture supracompetitive profits. Buyer resistance limits the ability of firms to maintain collusive prices through only tacit coordination because buyer resistance exploits the lack of communication, monitoring, and enforcement characterizing tacit coordination. If prices are not observable and demand has at least a small random component, then one enters the environment where the combination of a need for equilibrium path punishments with asymmetry of information about market outcomes has the consequence that perfect collusion is not possible without communication at the operation stage. Other changes in the environment reinforce the need for implementation-stage communication, including moving away from posted prices to, for example, competitive procurements and allowing buyers to be true players in the game. Collusive equilibria in games with repeated interaction that are supported by reversion to noncooperative behavior or some other equilibrium punishment, theoretically speaking, do not require communication among the firms in the implementation stage. Equilibria such as the collusive equilibrium constructed in Green and Porter (1984) are discussed in the law literature as “oligopoly pricing” or as consciously parallel decisions of a few dominant sellers in an industry to maintain the same high noncompetitive price.16 The idea is that this type of behavior might arise “without overt communication or agreement, but solely through a rational calculation by each seller of what the consequences of his price decision would be, taking into account the probable or virtually certain reactions of his competitors” (Turner, 1962, 661). In the environment of Green and Porter (1984), demand uncertainty prevents firms from being able to monitor perfectly the quantity choices of their rivals. Because of this, 15  Gasoline Companies Plead Guilty to Price-Fixing in Kingston and Brockville, Ontario, March 20, 2012, http://www.competitionbureau.gc.ca/eic/site/cb-bc.nsf/eng/03447.html. 16  Posner (1976, 40) refers to this as “tacit collusion.”

472   Edward J. Green, Robert C. Marshall, and Leslie M. Marx punishment periods are triggered in equilibrium. In this environment, even in the optimal collusive mechanism without implementation-stage communication, there will be punishment periods with low profits for the firms, and the periods of high profits may not yield profits as high as under strong explicit collusion. In an environment with quantity competition and demand uncertainty, the choice of the quantity targets in cooperative periods must balance the cartel’s desire to decrease those targets to increase payoffs in cooperative periods and the cartel’s desire to increase those targets to reduce the frequency of punishment periods, where payoffs are low. The more aggressive are the cartel’s quantity restrictions, the greater is the incentive for unilateral deviations, which, in this model, can only be prevented by making it more likely that such deviations will trigger a punishment period. The analysis of collusive mechanisms without implementation-stage communication in theoretical environments can provide valuable insights into collusive behavior. However, in few real-world environments would the informational requirements be met for there to be tacit coordination equilibria or weak explicit collusion equilibria that provide firms with the same level of profits as through strong explicit collusion, where communication occurs at both the initiation and implementation phase.17 As described in Marshall and Marx (2012, ­chapter  1.4), when purchases are sufficiently large and infrequent, or demand is sufficiently uncertain, or buyers are strategic, even though firms are engaged in repeated interaction over time, they may not be able to accomplish payoff-maximizing collusive outcomes.18 With enough lumpiness or randomness or difficulty interpreting responses from buyers, a firm often cannot rely on repeated play to discipline its rivals. Without communication and transfers, the sellers are left in a tough spot in terms of trying to achieve substantially elevated prices and profits. Secret deviations induced by self-interested profit maximization will creep into their conduct, and joint profits will fall short of monopoly levels.19 Strong explicit 17  “Oligopolists behaving in a legal and consciously parallel fashion could achieve high and rising prices, even as costs remain stable, by engaging in price leadership. The odds that they could achieve a price and profit increase and maintain incredibly high incumbency rate—that is, maintain the very same distribution of municipal contracts year after year—are miniscule, however, unless the oligopolists were communicating with one another.” City of Tuscaloosa v. Harcros Chem., Inc., 158 F.3d 548, 565 (11th Cir. 1998). The clear presumption of the court in this case was that buyers, the municipalities in Stateplace, Alabama, would conduct competitive procurements and push back against price increases by the sellers to the effect that incumbency rates would be volatile. It was not the presumption of the court that the buyers were passive. Rather, the court presumed that the buyers were players in the game, and thus much different from the customers of the gas stations in Carlton, Gertner, and Rosenfield (1997). 18  See Ivaldi et al. (2003) on how the feasibility of weak explicit collusion is affected by the number of firms, asymmetries among firms, entry barriers, market transparency, demand growth, innovation, product differentiation, multimarket contact, and other factors. See Choi and Gerlach (in this volume) on how the feasibility of weak explicit collusion is affected by the number of firms, symmetry and concentration, demand conditions, multimarket contact, imperfect observability and monitoring, incomplete information and communication, vertical mergers and restraints, and the presence of antitrust leniency programs. 19  The difficulties associated with sustaining successful tacit collusion when buyers make purchases infrequently or demand is uncertain can be seen formally in Tirole’s (1988, c­ hapter 6.7.1, 262–65)

Tacit Collusion in Oligopoly   473

collusion gives the firms the additional tools of communication and interfirm transfers that may allow them to achieve the joint profit-maximizing price.20

19.2.1. Illustration In what follows, we describe the distinction between static competition, tacit coordination, weak explicit collusion, and strong explicit collusion using a series of examples. These examples are drawn from ­chapter 1 of Marshall and Marx (2012).

19.2.1.1.  Example 1: Observable Posted Prices Consider a small town where, on opposing street corners, there are two firms that sell the same product.21 These firms are the only two firms that sell this product in the town. Other firms who sell products exactly like this are at least 150 miles away from this town. The firms can post a price at their location for everyone to see. Each firm can post a price exactly once a day at exactly the same time. Consumers buy from the vendor offering the best value. Absent any nonprice competition between the firms, if the two firms post the same price, then they each get half of market demand.

19.2.1.1.1.  Scenario 1: Gas Stations—Static Environment We begin by assuming the firms are two gas stations, and we label them as A  and B. Suppose that each one of these gas stations buys its gasoline for $2 per gallon and that, because demand is noisy, each knows that a price somewhere between $2.90 and $3 per gallon maximizes their combined profits. In a static environment, where no firm was concerned about future profits, rivalry between the firms implies that the price would be the marginal cost, or $2.01 if pricing is done in cent increments.

19.2.1.1.2.  Scenario 2: Gas Stations—Dynamic Environment, Uncertain Demand Now change this environment to allow more dynamic interaction. Suppose that the future matters almost as much as the present to each of the gas station owners. It is natural to think of the folk theorem, which suggests that the monopoly outcome should be possible without implementation-stage communication, but without communication, what price would be selected between $2.90 and $3.00?22 Perhaps the two firms can settle quickly on a price such as $2.90, but with a small amount of communication at the initiation phase they may be able to agree to higher price, say $3. In this case, weak explicit collusion would allow the two firms to achieve a bigger payoff than tacit coordination. exposition of Green and Porter’s (1984) model of tacitly collusive behavior (Marshall and Marx, 2012, ­chapter 1.10). 20  Levenstein and Suslow (2011, table 3) show that approximately one-third of the cartels they consider used a mechanism to arrange transfers when a cartel member’s sales exceeded its cartel allocation. 21  This example is based on Carlton, Gertner, and Rosenfield (1997). 22  For empirical evidence of collusion in retail gasoline, see Borenstein and Shepard (1996). See also Slade (1992).

474   Edward J. Green, Robert C. Marshall, and Leslie M. Marx

19.2.1.1.3.  Scenario 3: Gas Stations—Dynamic Environment, Uncertain Demand, Nonprice Competition But suppose there are nonprice dimensions for interfirm rivalry such as loyalty rebates, cleanliness of facilities, service for cars, and a variety of items available in the service station shop. If both firms are charging $2.90 through tacit coordination, but firm A starts to experience an erosion of its market share, then firm A would infer that firm B is competing on nonprice dimensions. Firm A can then either cut its price or increase nonprice offerings. In this kind of environment, with nonprice competition available to each firm, it may not be possible for each firm to experience an increase in profits from an increase in price—each firm may invest in the nonprice competition to an extent that no incremental profit is left from the price increase. But the two firms could communicate at the initiation of collusive play and commit to the absence of nonprice competition. If the market share of any firm falls significantly below 50%, then they know there was a breach in that the other firm was offering nonprice enhancements, implying that the agreement will end and prices will revert to marginal cost. Each firm may recognize that without communication during implementation it will not be possible to maintain a price of $3—there will be just too much temptation for nonprice competition. But they may determine a lower price, say $2.15, where they are each relatively confident that the initiation-stage communication regarding assurances of nonprice incremental enhancements will be honored, since the payoff to a secret deviation is just too low (relative to a price of $2.90). There may also be a lower price, such as $2.10, that the firms need no initiation stage communication to achieve. Specifically, the payoff to investment in nonprice enhancements is just too low when the price is $2.10 for either firm to undertake it—a slightly higher price would induce such investments in the absence of initiation-stage communication. In this example, tacit coordination at $2.10, weak explicit collusion at $2.15, and strong explicit collusion at $3 are each conceptually possible. The price increment that can be achieved with weak explicit collusion depends on the magnitude of the investment each firm would need to make in secret nonprice actions to deviate without direct detection. If the investment is small, then the price increment above $2.10 that can be achieved with weak explicit collusion is small. Similarly, if the investment is large, then the price increment will be much larger.

19.2.1.2.  Example 2: Unobservable Bid Prices We now change things again. Suppose that the two firms are not gas stations but instead are manufacturers of plastic industrial bags used for packaging a number of different types of finished products. If you think about a plastic bag that is used for packaging some types of processed vegetables, fertilizer, or ready-mix concrete, there will be printing on the bag. Many firms that package and ship their product, or are engaged in the business of shipping,

Tacit Collusion in Oligopoly   475

buy plastic industrial bags. There are standard bags constructed of standard materials with no printing. An example might be an 18 × 12 inch 3ml thick bag. However, these kinds of bags are typically a small percentage of sales for an industrial bag plant. Rather, most sales are to manufacturers or shipping companies that buy large numbers of bags of specific sizes, where the material used in the bag construction, the thickness of the bag, and the printing on the bags are specified by the buyer. Almost all of the purchases of such bags will be done by procurement. Specifically, a buyer will invite bag plants to submit bids, where the details of the bag size/thickness, material used, and printing will be part of the bid solicitation. The bag plants may post a price for the aforementioned 18 × 12 inch standard bag and may offer such bags at the posted price in small volumes at each firm’s office location, but the bulk of sales for each bag plant will be through procurements. The posted price for a standard bag has no effect on the bid submitted by any bag plant in a procurement, and the buyers know that. Each bag plant has a reasonably good idea of the costs of its rival. Also, unlike the gas station example, where no consumer is going to drive 150 miles each way to buy gas from a cheaper nonlocal gas vendor, it is economically viable to buy bags from a bag plant that is 150 miles away, although such a bag plant has additional transport costs. The bid by any bag plant in any procurement is not revealed by the buyer to the competing firm— a bag plant only knows if it won or lost any given procurement. For this example, we assume that each bag plant values the future highly. If we consider an environment with just large buyers, the two bag plants can potentially agree on bids to submit at the initiation phase to make sure that the work is shared in a relatively equal way. If one bag plant has won a recent large contract, then the other bag plant can win the next one. The two bag plants can potentially develop initiation phase contingent plans about the bids each will submit. Thus, the behavior can be supported with weak explicit collusion.

19.2.1.2.1.  Scenario 4: Bag Plants—a Small Number of Large Strategic Buyers, Uncertain Demand To answer the question as to why strong explicit collusion might be required above and beyond weak explicit collusion, suppose that buyers are small in number, large in size, and strategic. In addition, assume demand is uncertain. Suppose that the large buyers conduct procurements infrequently and at irregular intervals. Suppose that each bag plant expects to receive, in total, one bid solicitation each quarter. Furthermore, suppose that these large buyers may choose to extend their current bag contract terms for a year or more without reconducting a procurement. There are not many contract awards each year—on average, only four per year. If one bag plant receives only one or even none of the awards, then it will be in dire straits. For a given procurement, suppose that a bag plant submits a bid that is close to the joint monopoly price and observes that it loses the award. This could be from one or more of four causes: (1) random bad luck, (2) its competitor undercut it in the bidding, (3) there was a negative demand shock in the market that affected both bag plants but the shock was unknown as such to either bag plant, and (4) the buyer acted strategically by offering

476   Edward J. Green, Robert C. Marshall, and Leslie M. Marx the contract award to a competitor with extended award length if the bid price were lowered by that firm. Suppose the bag plant in question loses a second procurement. Now the bag plant that has not won a contract in six months is really feeling the financial pinch. Strong explicit collusion can resolve the conundrum for the two firms. By communicating and monitoring one another during the implementation phase the firms are able to eliminate certain conjectured causes for the lack of an award and directly address buyer resistance. Weak explicit collusion is much more difficult to implement. The number and complexity of contingencies that would need to be specified at the initiation phase is substantial. Creating contingent plans to address the entire range of potential buyer resistance would be like creating a contingent plan for a chess game. The results of Green and Porter (1984) imply that even a minor incremental payoff from tacit coordination would not be possible. When a procurement environment has this much noise from a number of sources, direct interfirm communication at both the initiation and implementation phase is needed to suppress interfirm rivalry.

19.2.1.2.2.  Scenario 5: Bag Plants—Many Nonstrategic Buyers, Uncertain Demand Consider another scenario where there are many small manufacturers buying bags with great regularity. Suppose that these buyers are all in the same business. For example, they make frozen French fries. They need bags to ship their frozen French fries. But suppose that frozen French fry demand for these local manufacturers is highly unstable, and the instability affects each of them in a similar way. Suppose that the bag plants cannot observe frozen French fry demand conditions. Furthermore, each bag plant cannot observe how much production is occurring at its rival’s plant. In normal demand conditions, each bag plant expects to receive 500 bid solicitations each quarter. But a given frozen French fry manufacturer may decide not to make any award in a given quarter if demand conditions are sufficiently depressed. Suppose that a bag plant submits bids over the course of a quarter that are at the joint profit-maximizing level. If it observes that it has won approximately 250 of 500 solicitations, then it will continue to hold to that price. But, if it observes that it has won 125 of 500 solicitations, then the bag plant has some serious questions to address. Are the frozen French fry manufacturers experiencing a negative demand shock, or has the other bag plant undercut the joint profit-maximizing price to obtain more awards, or is some of both occurring? By assumption, the bag plant in question is unable to determine the cause for the reduced number of contract awards. As a consequence, a bag plant will reduce its bid below the joint profit-maximizing price and move it toward the marginal cost of production. To see this, note that if the bag plant does not react based on the belief that bad outcomes are attributable to negative demand shocks and not the conduct of its rival, then the rival would take advantage of this belief by undercutting the high-priced bag plant. In this case, strong explicit collusion can elevate profits, probably to the joint profit-maximizing level. But weak explicit collusion can produce a payoff above static

Tacit Collusion in Oligopoly   477

Nash albeit below strong explicit collusion. Initiation-stage communication can resolve the price that each firm will bid at the procurements. The weak explicit collusion described here is the same as that in Green and Porter (1984).

19.2.2.  Context for Merger Reviews The issues of tacit and explicit collusion also arise in the context of merger reviews. The 2010 Horizontal Merger Guidelines (Guidelines) of the US Department of Justice and Federal Trade Commission discuss concerns about “express collusion,” by which they presumably mean strong explicit collusion, and also about “coordinated interaction.”23 According to the Guidelines (24–25): Coordinated interaction includes a range of conduct. Coordinated interaction can involve the explicit negotiation of a common understanding of how firms will compete or refrain from competing. Such conduct typically would itself violate the antitrust laws. Coordinated interaction also can involve a similar common understanding that is not explicitly negotiated but would be enforced by the detection and punishment of deviations that would undermine the coordinated interaction. Coordinated interaction alternatively can involve parallel accommodating conduct not pursuant to a prior understanding. Parallel accommodating conduct includes situations in which each rival’s response to competitive moves made by others is individually rational, and not motivated by retaliation or deterrence nor intended to sustain an agreed-upon market outcome, but nevertheless emboldens price increases and weakens competitive incentives to reduce prices or offer customers better terms. Coordinated interaction includes conduct not otherwise condemned by the antitrust laws.

Thus, it appears that the Guidelines intend for the term “coordinated interaction” to encompass conduct ranging from tacit coordination to strong explicit collusion and apparently more. For an attempt to disentangle the notion of parallel accommodating conduct, which is introduced in the 2010 revision of the Guidelines, from existing notions of collusion, see Harrington (2012).

19.3.  Evolution of the Theory of Explicit Collusion Before the arrival of formal dynamic game theory, ideas on collusion without communication relied on two theories: focal points and price leadership. The theory of focal 23 

The Guidelines are available at http://www.justice.gov/atr/public/guidelines/hmg-2010.pdf.

478   Edward J. Green, Robert C. Marshall, and Leslie M. Marx points stated in Schelling’s (1960) The Strategy of Conflict played an important role in the justification of tacit coordination as shown in Scherer (1970, 192): “Even a price that has no particular uniqueness or compulsion in its own right may become a focal point simply by virtue of having been quoted repeatedly.” On price leadership, when Bain (1968) analyzed conventions and agreements to set a price leader, he argued: Evidence of such direct consensual action not being found (and it seldom is), it is more usual to recognize price leadership as a form of tacit collusion, resulting from the existence of an unspoken agreement. The notion of tacit agreement, however, is itself somewhat nebulous, and it seems perhaps equally appropriate to designate the conduct pattern in question as one of interdependent seller action without basis in agreement. (Bain, 1968, 312)

From this starting point evolved the analysis of collusion in terms of repeated and dynamic games using the tools of game theory.24 The focus was on the tension described by Bagwell and Staiger (1997, 82): “Collusion is a balancing act. Each colluding firm balances the short-term temptation to cut its price against the expected long-term cost of the price war that such an act might instigate.” In this literature, the threat of punishment of deviations is the force driving collusive outcomes. Naturally, the ability of firms to detect deviations plays an important role. At the same time as one community of economists was developing the relatively informal theories of focal points and price leadership, another community was developing explicit, formal theories of equilibrium in repeated games. However, after having flourished for more than a decade, by the mid-1960s game theory was viewed by many economists as a sterile mathematical field that was too highly schematic to be applied successfully to collusion or to other concrete issues of economics and social science. Large credit for superseding that negative view is due to James W. Friedman. Both through his own research and through his lucid presentation of the game-theoretic ideas in Oligopoly and the Theory of Games (1977), Friedman educated “mainstream” IO economists about the way of thinking and the specific results and techniques developed by game theorists in the 1960s and 1970s. One of his important, specific contributions was to frame repeated games in a discounted-payoff context, rather than using the limit-of-payoffs criteria that had dominated preceding research. By doing so, he brought into focus the crucial role of the trade-off between short- and long-term costs and benefits that Bagwell and Staiger’s summary emphasizes. The main limitation of Friedman’s theory of collusion without implementation-stage communication was its restriction of attention to an environment with perfect monitoring. Abreu (1986) provided the capstone of that research program by characterizing the maximum profits that can be sustained without implementation-stage communication or transfers in terms of the most severe punishments for defection that colluders could impose on one another. 24  See Choi and Gerlach (in this volume) for a review of experimental evidence on collusion and antitrust leniency programs.

Tacit Collusion in Oligopoly   479

The literature advanced to consider the possibility of collusion without implementation-stage communication, that is, weak explicit collusion in our terminology, in ever more complex environments. For the case of imperfect monitoring, Green and Porter (1984) show that some degree of collusion can be achieved in a quantity-setting supergame even when firms are not able to perfectly observe their opponents actions. Abreu, Pearce, and Stacchetti (1986) study the same issues as Abreu (1986) but for the imperfect monitoring case.25 Brock and Scheinkman (1985) and Lambson (1994) study weak explicit collusion in price-setting supergames. The first shows that some degree of price elevation can be achieved and study how this is affected by the number of firms in the industry. The second studies the maximum profits achievable through weak explicit collusion and shows that the structure of the most severe (optimal) punishments is much simpler in price-setting supergames than in quantity-setting supergames. Finally, Benoit and Krishna (1987) study weak explicit collusion in a model in which firms choose both quantities and prices. Additional complexities are considered in literature that considers environments with incomplete information. In environments with incomplete information, belief-free equilibria are important for private monitoring. If players observe different signals—for example, an oligopolistic market in which each firm observes only its profits but no prices, quantities, or other firms’ profits—then beliefs about what other players are observing complicate the equilibrium analysis. But still, given that the signals are informative enough, it is possible to sustain some level of collusion with a subset of subgame perfect equilibria in which beliefs about what the other players have observed play no role, the belief-free equilibria (see Ely and Valimaki, 2002; Ely, Hörner, and Olszewski, 2005). In environments with incomplete information about the profitability of a market, recent work shows that collusive payoffs may still be possible without communication at the implementation stage. See, for example, Yamamoto (2014) and Schenone (2012).26 In this literature, in some cases, the necessary conditions for folk theorem type

25  Abreu, Pearce, and Stacchetti (1990) rely on a different method to solve for the equilibria of dynamic games, using the notion of self-generation, which allows applications to asymmetric games. 26  The results of Yamamoto (2014) imply that under some conditions a cartel can self-enforce some degree of collusion even if firms do not know how profitable the market is. The environment is a two-player game with a public randomization device and private monitoring. The state of the world is chosen by nature at the beginning of play and it influences the distribution of signals and/or payoff functions of the stage game. Yamamoto’s (2014) solution concept is belief-free ex post equilibria (BFXE). In a BFXE a player’s best response does not depend on her beliefs about the state of the world or her beliefs about her opponents’ private information. A folk theorem for this environment/solution is not obtained, but some degree of collusion is sustainable. Schenone (2012) considers an environment different from Yamamoto’s. First, instead of private monitoring, Schenone’s game is one of perfect monitoring. Second, each player knows its own payoff, and the state of the world can be uniquely identified by pooling together the information of all players. Schenone finds conditions that are both necessary and sufficient for a perfect Bayesian equilibrium folk theorem. The conditions are also necessary and sufficient for a belief-free equilibria folk theorem to hold. In this case, the solution concept is perfect type-contingent ex post equilibria, which is equivalent to a belief-free equilibrium under perfect monitoring.

480   Edward J. Green, Robert C. Marshall, and Leslie M. Marx results are restrictive, telling us that a folk theorem does not hold for a wide range of games. Additional challenges to the legal interpretation of firm behavior and enforcement of antitrust laws are raised by more recent literature such as Hörner and Jamison (2007), which presents a model in which almost no information is required to sustain full collusion. The model consists of an infinitely repeated price-setting game with inelastic demand. At each period, each firm draws a per unit cost that can be either high or low. Costs are private information. At each period, the firm with the lowest price gets the whole demand and if more than one firm set the lowest price, a randomly selected firm gets the whole demand. A firm only observes its own price and whether it sells. In this model, full collusion is understood as an equilibrium in which in almost all periods a low price gets the sale, and the firm charges a price close to the monopoly price. The main result states that if firms are patient enough, they can get arbitrarily close to the monopoly outcome without implementation-stage communication.

19.4.  The Initiation Phase of Collusion Prior to this point, our discussion has focused primarily on the question of under what circumstances explicit coordination among sellers would be required during the operation of the cartel, in order to be successful in maintaining prices above a competitive level. However, we have emphasized that coordination might also be necessary prior to the cartel beginning to operate, in order to agree on a mechanism by which subsequent collusion would be enforced and on how the rents from successful collusion would be distributed among the cartel’s members. We now turn to the questions: under what conditions would such explicit coordination/communication ex ante be required, and under what conditions would it even be helpful, in order for a viable cartel to be born? Throughout most of our investigation, we maintain the view taken in the several preceding sections that industry conduct is a Nash equilibrium outcome of strategic interaction between agents in the market environment.

19.4.1.  Nash Equilibrium as Self-Enforcing Agreement A widely held interpretation of a Nash equilibrium is that it is a self-enforcing agreement among the players of a game (Myerson 1991). Taking this interpretation literally, a Nash equilibrium could not be tacit coordination, since an agreement cannot be reached without communication. A less restrictive interpretation is that players conduct themselves as though they had previously reached an agreement among themselves, and moreover the counterfactual

Tacit Collusion in Oligopoly   481

agreement would enforce itself. That is, the prior meeting to reach and ratify an agreement can be treated as a parable rather than as an actual historical event, much as the political theorists interpret the “social contract.” Aumann and Brandenburger (1995, 1162–63) have made explicit what such a counterfactual account of agreement entails: Suppose that the game being played [by two players] (i.e., both payoff functions), the rationality of the players, and their conjectures [about the probabilities with which actions their respective opponents will take] are mutually known. Then the conjectures constitute a Nash equilibrium. . . . [Moreover, suppose that there are more than two players, that] their payoff functions and their rationality are mutually known, and that their conjectures [about one another’s actions] are commonly known. Then [their conjectures coincide and constitute] a Nash equilibrium.27

To say that something is mutually known just means that everyone knows it. To say that it is commonly known means that everyone knows it, everyone knows that everyone else knows it, and so forth. People generally do not need to meet, in order to have mutual knowledge of some aspect of their shared environment. For example, if ten people read this sentence, then they have mutual knowledge that it begins with a three-letter word. But, unless they are all read the sentence in one another’s presence (in which case, they are meeting) or they subsequently discuss the sentence after having read it (in which case they also meet, subsequently to their reading), presumably none would know that the others know this fact. Moreover, although direct observation (without communication) by all competitors is sufficient to establish mutual knowledge of an observable fact, it is not sufficient to establish mutual knowledge of one another’s intentions. In particular, given that repeated games of the sort considered above have both competitive and collusive equilibria, one competitor cannot observe which equilibrium another expects everyone else to play (and so will play himself). Thus, in any cartel situation—even a duopoly—it is difficult to see how prospectives would achieve mutual knowledge of each other’s intent to collude, unless they were to communicate those intentions to one another. Common knowledge, and mutual knowledge about other players’ mental states (e.g., intentions, beliefs) is called higher-order knowledge. So, let’s formulate a two-part Need to Meet (NTM) principle:



1) In general, players cannot arrive at a Nash equilibrium by a process of reasoning, unless they attain higher-order knowledge of one another’s conjectures. In particular, they cannot “reason their way” to a collusive Nash equilibrium without attaining higher-order knowledge. 2) Such higher-order knowledge cannot be attained, in practice, without having an explicit meeting.

27  Because we will not discuss it here, we have omitted a common-prior assumption of Aumann and Brandenburger’s result.

482   Edward J. Green, Robert C. Marshall, and Leslie M. Marx Let us acknowledge, before going further, that the NTM principle is not a logical consequence of Aumann and Brandenburger’s result. They have framed a sufficient condition for Nash equilibrium, while the NTM principle envisions common knowledge, and hence occurrence of a meeting, as a necessary condition. They write (1163) that “It is always possible for the players to blunder into a Nash equilibrium ‘by accident,’ [but a higher-order-knowledge assumption] cannot be . . . significantly weakened.”28 But, attaining such higher-order knowledge is virtually a necessary prerequisite for players (e.g., sellers in a market) to reason their way to any particular type of Nash conduct, collusive or otherwise. Henceforth we will treat it as being necessary. That is, we stipulate clause (1) of the NTM condition. If clause (1) of the NTM condition holds, then clause (2) is tantamount to an assertion that higher-order knowledge of competitors’ intent to collude must always be reached explicitly, rather than tacitly. In the next two sections, we assess clause (2).

19.4.2.  What Does “Arrive by Reasoning” Mean? We have identified a situation in which Aumann and Brandenburger’s higher-order-belief conditions are satisfied, and in which there is a self-confirming conjecture that a particular Nash equilibrium will be played, with a situation in which the players have arrived by reasoning at the equilibrium. This situation is contrasted with the one in which, in Aumann and Brandenburger’s words, the players have blundered into an equilibrium by accident. But to make this identification is of no help for assessing whether or not clause (2) is reasonable. To make that assessment, we need an explicit account of the process that players are envisioned to follow. That account is provided by the theory of rationalizable play. A player is rational if he plays an action that is a best response to his conjecture (represented by a probability measure over opponents’ possible actions) of what actions other players may take. The first step of his reasoning process is to eliminate actions that would not be a best response to any conjecture about what others might do. Then, he would put himself in the shoes of other players, whom he knows to be rational. He would realize that, at a minimum, each of them would also be eliminating any never-a-best-response actions. At that point, the player would review his actions that have survived being eliminated in the first round. Has any of those actions been retained solely because it would be a best response to some opponent’s action that is never a best response for the opponent? In that case, he knows that the opponent will not play the action in question, so he should eliminate the action that he retained on account of it. When each player has gone through this second round of eliminating actions, and has again put himself in the shoes of other players who have done likewise, there will be a third round, and so on. A player’s actions that survive elimination in every round, forever, are rationalizable for that player. Pearce (1984) and 28  Aumann and Brandenburger (1995, sec. 5) show that their result would fail if any of the conditions were weakened significantly.

Tacit Collusion in Oligopoly   483

Bernheim (1984) formulated the concept of rationalizability, and they showed that all actions played (with positive probability) in a (mixed-strategy) Nash equilibrium must be rationalizable. Tan and Werlang (1988) showed that a profile of actions in a game is rationalizable for the respective players if and only their being played is consistent with it being common knowledge that all players are rational.29 That is, the iterative process of reasoning by which players determine their rationalizable actions (a) never eliminates any action that is ever played in any Nash equilibrium, and (b) always eliminates any action that would be inconsistent with common knowledge of rationality. In view of these results, rationalizability is evidently the right concept to make precise the idea of “arrival by reasoning” at Nash play in a game.

19.4.3.  The (Im)plausibility of Arriving at Collusion by Reasoning The standard, intuitive view of tacit coordination in the IO/legal literature seems to be as follows. At the beginning of the story, each of prospective colluders is pondering whether to act competitively or monopolistically. These agents do not communicate with one another. Nevertheless, it is common knowledge that all of them want to maximize their respective profits, and it is also common knowledge that there is a self-enforcing profile of actions such as has been discussed above in section 19.2.1, which would solve the problem of maximizing joint profits and distributing them equitably (according to some standard that they commonly know that they all embrace) subject to the constraint that collusion must be self-enforcing. In this situation, according to the intuitive view, each agent should conjecture that the other agents will play as specified in that profit-maximizing equilibrium, and therefore should also play according to the equilibrium because the equilibrium is a self-enforcing arrangement. This statement of the intuitive view is ambiguous between two versions. One version—let’s call it the convergent-elimination view—assumes that, bootstrapping from just the facts that have been stipulated to be common knowledge, the iterative elimination procedure will zero in on precisely the profit-maximizing Nash equilibrium. That is, that equilibrium profile of actions will be the unique profile of actions that survives the iterative-elimination process. The other version—let’s call it the focal-point view— recognizes that the rationalizable actions are those that could be chosen by agents with arbitrary beliefs (subject only to having certainty of the facts that are stipulated to be common knowledge) and makes a further assumption that all of the agents’ beliefs place very high prior probability on the profit-maximizing equilibrium being played. On this view, the role of the iterated-elimination process is to condition probabilities repeatedly on higher-order-belief conditions, raising agents’ probability assessments from the

29  Again here, we are giving a brief incomplete statement of a result in order to avoid digression from the topic of the chapter.

484   Edward J. Green, Robert C. Marshall, and Leslie M. Marx initial, very high levels to virtual certainty that the profit-maximizing equilibrium will be played. The explication of “reasoning to an equilibrium” as rationalizability shows that the convergent-elimination view is completely untenable. To begin, the game among prospective colluders always has a competitive Nash equilibrium, as well as possibly having a collusive one. Property (a) of rationalizability is that no action consistent with any Nash equilibrium is ever eliminated. Thus, at most, rationalizability might establish that some Nash equilibrium or other might be played, but it could not establish that the collusive equilibrium specifically would be played. Further analysis is even less favorable than that to the convergent-elimination view. Specifically, rationalizability is such a weak solution concept that it is consistent with it being common knowledge that no player will play a Nash strategy.30 Convergence to Nash equilibrium in general, let alone to a specific Nash equilibrium of a multiple-equilibrium game, cannot be achieved entirely from the “Archimedean lever” of common knowledge of rationality. The focal-point view is logically coherent, because it adds a strong hypothesis about the prospective colluders’ prior beliefs to the minimal hypothesis of common knowledge of rationality. The focal-point view is as plausible, but only as plausible, as is that hypothesis. The gist of the argument for the hypothesis is that, when players of a game share a common background, and when that common background is common knowledge among them, then their conjectures are much closer to being coincident than would be the case if they lacked such shared experience.31 Consider the plausibility of the focal-point view with respect to the spectrum of cases of collusion described in section 19.2.1. Three questions should be considered: (1) Is it plausible that all prospective colluders would conjecture that some Nash equilibrium would be played? (2) If so, then is it plausible that they would all conjecture that some collusive equilibrium (that is, one that would achieve an above-competitive level of profit) would be played? (3) If so, would their conjectures all be (nearly) identical, so that they collectively would succeed in playing a specific, collusive equilibrium?

30  Bernheim (1984, 1011–12) provides an example of a game that has a unique Nash equilibrium. It is a pure-strategy equilibrium in which each player plays one specific action with certainty. However, that Nash action is not the player’s only rationalizable action. To the contrary, there is a set of profiles of actions, in each of which all players play non-Nash actions, such that rational players can have common knowledge of those profiles as conjectures. 31  The focal-point view was first set forth by Schelling (1960). Schelling envisioned two people who have common knowledge that they need to meet in Manhattan, but have not had an opportunity to communicate about where they should meet. He suggested that two tourists would both go to the Empire State Building, while two NYC natives would both go to Grand Central Station. That is, a common background makes a particular Nash equilibrium to be focal, but which equilibrium were to become focal would depend on specifics of the background.

Tacit Collusion in Oligopoly   485

Let’s stipulate for now that the answer to questions (1)  and (2)  are affirmative. Certainly the predominant economic modeling strategy in IO is to assume Nash play. We will further consider question (2) below, but at the very least, a collusive equilibrium is a natural conjecture in a context where it is common knowledge that everyone wants to maximize profits. In our basic example of gas-station duopolists, affirmative answers to questions (1) and (2) strongly suggest an affirmative answer to question (3), as well. This implication reflects our assumptions in the example, such as that gasoline is an undifferentiated product, that competitors have no private information about their respective costs, and that price discrimination of any sort is impossible for any seller to implement. It is not quite true that there is a unique profit-maximizing equilibrium in this setting, but there is a unique price-maximizing equilibrium path, and it is a deterministic path. It does not matter much, then, what kind of punishments the colluders believe would be imposed off the equilibrium path. In fact, as long as each colluder conjectures that he would be punished with sufficient severity if he were to break the cartel, the equilibrium path of play will be realized. This outcome is robust to that type of heterogeneity of conjectures. But the farther one moves away from the basic example, the less plausible it is that industry participants would share a focal point. Consider three representative reasons why coordination would be unlikely to be achieved in various situations: • In a differentiated-product oligopoly, joint profit maximization is no longer salient. Rather, there is a Pareto frontier of profit-maximizing price profiles. Because of imperfect substitutability, a seller can be better off to maintain a price somewhat above its competitor’s price than to sell at the competitive price. A firm that charges a high price, but one that is slightly lower than its competitor’s price, gets the lion’s share of the rents. The division of rents would be determined by the competitors’ respective bargaining power against one another. If each competitor believes that it is in the strongest bargaining position, for example, then the various competitors’ conjectures will correspond to different points on their Pareto frontier of profitability, and coordination on a focal Nash equilibrium will fail. • Even if competitors agree about what is their desired imputation of rents, there are a plethora of ways to achieve it. Consider, for example, the notorious cartel of electrical-generator manufacturers in the 1950s. Bidding in procurement auctions was rigged by a scheme that selected a winner according to the phase of the moon on the day bids were requested. As long as purchasers remained unaware of the arrangement, the lunar phase would be random, so the selection of the winning bidder would be equitable. But, certainly it could not have been common knowledge, prior to any communication, that this was the scheme that would be adopted. That is no more a focal point than using the last two digits of the number (mod 4) of shares traded on the New York Stock Exchange on the day prior to the opening of bidding. Or, one month prior to the opening of bidding. Or, the high temperature (mod 4) in Chicago on that day, as reported by the US National Weather Service. The number of equally salient candidates for a coordination

486   Edward J. Green, Robert C. Marshall, and Leslie M. Marx device is huge. To implement a Nash equilibrium, it is not sufficient for cartel participants to agree that they will split the market in a way that provides equal shares in the long run. They must have a unanimous conjecture about precisely how the splitting will be done. • In an environment with private information, the equilibrium path will be stochastic. If “punishment phases” are randomly triggered by outcomes such as steep decline in demand at the cartel price, for which a deviation from collusion has a high likelihood-ratio statistic (even though deviation is not believed to have actually occurred), then colluders face a coordination problem. If the “punishment” is not going to last forever, then all colluders must abandon it simultaneously. The reason is that, while anyone acts as prescribed in that phase, prices tend to remain low. Thus, if someone continues that conduct for longer than the others conjecture, then they will perceive the resulting low price to be occurring during the active-monopoly phase of the cartel and will initiate the punishment phase. With that phase being in force most of the time, the cartel will not be profitable. But there are many equilibria corresponding to various rules about when to re-enter the active-monopoly phase, none of which is an obvious candidate to be a focal point. One candidate, the profit-maximizing equilibrium, involves successive punishment phases of independent, random duration. (Abreu, Pearce, and Stacchetti, 1986). Analogously to the problem of random selection of the respective winning bidders in a sequence of rigged auctions, there is indeterminacy in the selection of the randomizing device, and coordination will fail unless all participants adopt the same device. To summarize, there are two versions of the idea that sellers in a market might coordinate on a specific pattern of collusive conduct without the need to communicate plans with one another. One version, that such coordination would occur simply as a result of the sellers following the logic of optimization in their shared situation—that is, the idea of “reasoning to an equilibrium”—does not survive careful game-theoretic analysis. The other version—the focal-point view—is logically sound. Nevertheless, because most market environments are sufficiently complex that there are numerous possible ways to collude, none of which will work unless it is adopted by all of the significant market participants, that view suggests that it is difficult, and probably rare, for successful collusion to obtain in the absence of explicit communication. The exception to this generalization is a market that is so simple and transparent—such as the gas-station-duopoly example discussed earlier—that there is a unique candidate for the optimal collusive agreement.

19.4.4.  Risk Dominance as an Obstacle to Collusion Let’s further consider an infinite-horizon version of the gas-station duopoly. Suppose that the stations are identical, having a fixed cost F of operation and selling gas at

Tacit Collusion in Oligopoly   487

constant marginal cost. Moreover, suppose that there are only two prices that could be charged: a high pH and a low pL . Let’s make an assumption that the gas stations will continue to charge pH forever if they both charge that price initially, but that they will both charge pL subsequently if they do not both charge pH the first time. This assumption expresses, in stark form, the idea that initial success in collusion establishes a successful long-term arrangement, but that any convergence to collusion after an initial failure to collude would be slow enough that the present discounted value of eventual collusive profits would be low. If both sellers charge pH, then each receives half of the monopoly profit M in each period, and the cartel continues forever. Thus, the discounted present value of participating in the cartel is ρM , where ρ > 1.32 If they both charge the low pL, then each earns zero profit.33 If they start out making opposite choices, then the high-price station initially sells no gas but bears its fixed cost, losing F, while the low-price station initially earns revenue pL D( pL ) = 2F and consequently earns positive profit F net of its fixed cost. Subsequently to the initial period, by the assumption above, both stations charge pL and earn zero profit. These payoffs are summarized by the following matrix. pH pH pL

pL ρM / 2

ρM / 2

F −F

−F F

0 0

If the interest rate is low (so that ρ is large), M is large, and F is small, then ρM / 2 > F , and both symmetric-price profiles are Nash equilibria, with the collusive one (that is, for both sellers to charge pH) being the more profitable one. Profitability seems to make collusion a focal point. However, there is a countervailing argument. A seller who sets price pH will suffer a loss if, contrary to his conjecture, the other seller charges pL. In contrast, a seller who sets price pL will make an unanticipated profit if, contrary to his conjecture, the other seller charges pH. In this sense, setting pL is the less risky choice. Indeed, if a seller conjectured that the other seller was equally likely to charge pL or pH, then he would charge pL in order to maximize his expected profit. The technical name for this property of the low-price equilibrium is risk dominance. It was introduced by Harsanyi and Selten (1988), who proposed the tracing procedure as an account of how rational players choose among multiple Nash equilibria of a game that 32 

If the duopolists face 1-period interest rate r, then ρ = (1 + r ) / r . So, if D(p) is market demand at price p, then pL D( pL ) / 2 = F . Note that, as pointed out by Edgeworth, this price would not be a Bertrand equilibrium (and, with positive F, no Bertrand equilibrium would exist) if a seller could marginally lower its price and capture the entire market. 33 

488   Edward J. Green, Robert C. Marshall, and Leslie M. Marx they are playing.34 Roughly speaking, their proposal was first to discard all but the Pareto frontier of the Nash equilibrium set, and then to use risk dominance to select among the remaining equilibria. But Aumann (1990) subsequently argued, in effect, that such a lexicographic priority for Pareto efficiency was misguided, and that risk dominance is actually the more compelling criterion. Harsanyi (1995) accepted Aumann’s argument and formulated a new equilibrium-selection theory based solely on a version of risk dominance. The case for the risk-dominant equilibrium being played is even further strengthened by research of Carlsson and van Damme (1993). They consider games such as the one above, to which they add the realistic consideration that players’ information about one another’s payoffs is accurate but not perfectly precise. Imagine, for example, that two gas-station owners are franchisees of different chains and are contractually obligated to use their respective franchisers as their sole suppliers of gasoline. Each supplier sets its price at any date in a way that closely reflects the market price of crude petroleum, and each gas station owner can read the price of crude in the newspaper but cannot directly observe the wholesale refined-gas price of his competitor’s supplier. Thus, each owner is highly confident that the other purchases gas at a price very close to his own price, but does not know his competitor’s price—or, consequently, his competitor’s payoff function—with complete accuracy. The generalization of Nash equilibrium to such a situation is Bayesian Nash equilibrium, and Carlsson and van Damme prove that the risk-dominant Nash equilibrium of the complete-information game corresponds to the unique Bayesian Nash equilibrium of the incomplete-information version of the game, even as the precision of information tends to certainty.35 Carlsson and van Damme’s formal proof is a piece of mathematical analysis that may seem to be a magician’s hat trick, but they provide a discussion that makes it seem much more intuitive. Recall the distinction between mutual knowledge and common knowledge that was introduced above. Something is mutual knowledge among a group of competitors if each knows it. A synonym for mutual knowledge is first-order knowledge. Define the competitors to have second-order knowledge of some fact if each of them both knows the fact itself and knows that the others know it. Third- and higher-order knowledge are defined successively in this way. A fact is common knowledge if there is first-order and second-order and . . . nth-order and . . . knowledge of it. When information is sufficiently precise, it can be nth-order knowledge (for very high n) that rational competitors will charge the monopoly price (that is, will play the payoff-dominant but risk-dominated strategy), but that is never common knowledge unless information about competitors’ payoffs is absolutely precise. Rubinstein (1989) has shown that no finite level of belief is sufficient to make it rational for everyone to play the payoff-dominant Bayesian Nash equilibrium. Only common knowledge will do. This same insight is the economic content of Carlsson and van Damme’s result. 34  See Harsanyi and Selten (1988) for a formal definition of risk dominance and discussion of its significance. 35  Actually, Carlsson and van Damme prove an even stronger conclusion than this.

Tacit Collusion in Oligopoly   489

Two rejoinders to the foregoing arguments why the risk-dominant equilibrium would be played are available. One rejoinder, specific to Carlsson and van Damme’s argument, is that the technical assumptions of their theorem are more restrictive than their account of the assumptions’ intuitive content would suggest, and that those technical assumptions are unlikely to hold exactly. This rejoinder can be rebutted, in turn, by the observation that assumptions of scientific theories invariably fail to correspond precisely to the truth. In a specific form, the rejoinder simply expresses an antiscientific attitude, and it is not worthy of being taken seriously. At this point, there is nothing further to be said on either side. Probably neither party to the discussion will have persuaded the other, and their views will be as far apart as when they began. Just as it is not likely to produce scientific agreement, this line of argument is also unlikely to produce consensus about practical questions of competition policy or jurisprudence. A much more compelling and interesting rejoinder is based on an argument made by Aumann (1990). To begin to set forth this rejoinder, let’s step back for a moment and put Carlsson and van Damme’s argument in context. Someone who uses that argument to argue that monopolistic outcomes cannot arise from tacit coordination intends, presumably, to contrast tacit coordination with weak explicit collusion involving overt negotiation of a cartel arrangement, and to assert that the cartel members could possibly reach common knowledge of their collusive intent through communication. But Aumann claims to refute that assertion.36 According to Aumann, a player can credibly communicate to an opponent that he would like the opponent to play the payoff-dominant strategy, but he cannot credibly communicate to the opponent that he will play the payoff-dominant strategy himself. The upshot is that the risk-dominant (that is, nonmonopolistic) equilibrium is the only one that could rationally be played, even if there is unlimited opportunity for explicit communication among prospective colluders. The conjunction of Aumann’s conclusion (that explicit collusion cannot be more feasible than tacit coordination is) and Carlsson and van Damme’s conclusion (that tacit coordination is not feasible) flies in the face of the incontrovertible evidence that explicit collusion has sometimes occurred and been stable through long intervals of time.37 Thus, if the soundness of Aumann’s argument is conceded, then Carlsson and van Damme’s conclusion must be denied. Again, though, there is a rebuttal to the rejoinder. The rebuttal is that there is experimental evidence against Aumann’s view. Specifically, Charness (2000) reports an experiment in which a high proportion of subjects played a payoff-dominant, risk-dominated equilibrium after having had opportunity for explicit communication, but in which that equilibrium was rarely played by subjects who had not had such an opportunity.38 On 36  Moreover Harsanyi, a game theorist of unparalleled subtlety and insight, was persuaded by Aumann’s argument. 37  On the prevalence of cartels, see Levenstein and Suslow (in this volume). 38  Charness cites some previous research that also suggests that conclusion. See also Choi and Gerlach (in this volume).

490   Edward J. Green, Robert C. Marshall, and Leslie M. Marx the basis of this evidence, someone might decide to reject Aumann’s conclusion (even if he could not pinpoint where the logic of Aumann’s argument for it goes wrong), and could consequently hold the view that explicit collusion is likely to occur but that tacit coordination is implausible.

19.4.5.  Initial Capital Investment as Communication So far in this section, we have supposed that industry conduct begins immediately with decisions about output or pricing, with capital investment implicitly having been determined previously and not being subject to adjustment. That is, investment (or capacity) is not a strategic choice. The argument of Aumann that has just been discussed, in particular, is formulated in that context. Now, let’s reconsider that argument in the context of an industry with a two-stage life. In the first stage, an incumbent with large production capacity is joined by an entrant who bears an investment cost of acquiring capacity. (Whatever cost the incumbent may have borne in the past to build capacity, is now “sunk cost” and will not figure in the analysis that we make under these assumptions.) In the second stage, the firms will be quantity-setting competitors who choose their respective outputs at every date (0, 1, 2, . . .) subject to their respective capacity constraints. These constraints are at the exogenous (but nonbinding) level for the incumbent and the level determined by stage-1 investment for the entrant, and cannot be adjusted during stage 2. At each date, firms produce perfect-substitute goods and the market price is perturbed by a demand shock that the firms cannot observe, as in the model of Green and Porter (1984). As in the preceding discussion we simplify the analysis by supposing that only finitely many—in this case, three—output levels are feasible. One is the static Cournot level. Call this output q2 . The other two feasible levels are parameters of the joint-profit-maximizing equilibrium of the Green-Porter environment that Abreu, Pearce, and Stacchetti (1986) (APS) have characterized. In that equilibrium, each firm produces output q1 < q2 , which is approximately the static monopoly output, in the “cooperative” phase of equilibrium; and each produces a maximin output q3 > q2 during reversionary episodes. Thus we specify {q1 , q2 , q3 } to be the set of feasible output levels. Given this assumption about feasible output levels, it is natural to specify also that, in stage 1, the entrant must choose a capacity level in {q1 , q2 , q3 }. Now, the question is, can the entrant credibly signal collusive intent and also motivate the incumbent to collude by choosing some particular level of capacity. Of course, the entrant could do more than signaling—he could commit always to produce the collusive output—by choosing capacity q1 . But, if the entrant did that, then he would lose the capability to punish the incumbent for not reciprocating in collusion. The incumbent would make strictly higher profit in any period by producing q2 than by producing q1 , and so would never produce q1 . The resulting industry conduct would not be collusive.

Tacit Collusion in Oligopoly   491

Nevertheless, the entrant would succeed in both signaling collusive intent and also motivating the incumbent to collude as well, if he were to choose capacity q3 . The rational incumbent should understand the implication of that choice in terms of a thought process that game theorists call forward induction. He should think as follows: If the entrant conjectured that I  were going to play static Cournot, then he would be certain that he would never benefit (either directly or indirectly, by setting incentives for me) by producing more than q2 . Therefore, he would be sacrificing profit by making more costly investment to acquire capacity q3 . Since he is a profit maximizer, I deduce (by modus tollens) that his conjecture is that we are going to play the APS equilibrium rather than the static Cournot equilibrium in stage 2. Consequently, if I were to produce more than q1 at the initial production date, then with high probability, the market price would be lower than the trigger price, and he—with certainty that I am behaving identically—would switch to the reversionary output level q3 , which would be bad for my profitability. That is, my conjecture is that he is going to produce the monopoly level of output as long as the market price stays above the trigger price, and my Nash response to that conduct is to do likewise. The foregoing analysis follows closely the analysis of forward induction first proposed by van Damme (1989) for equilibrium of the burning-the-dollar game. That is a two-stage game, in which some particular single-stage game with several Nash equilibria is to be played in stage 2. In stage 1, one of the players has the opportunity publicly to burn a dollar bill. This action has no implication whatsoever for the feasibility of strategies or for the payoffs to any player of strategy profiles in stage 2. Nevertheless, the two-stage game has a unique equilibrium consistent with forward induction: the player does not burn the dollar, and even though he abstains from that action, the Nash equilibrium most favorable to him is played in stage 2. The analogy between the oligopoly game discussed here and the burning-the-dollar game is not exact, because the analogue of burning a dollar bill would be to invest in even more capacity than is required to play the APS equilibrium, and we have specified that option not to be available. Nevertheless, the logic of the incumbent’s thought process described above is precisely that spelled out in the analysis of burning the dollar. Alternately, consider an industry such as production of a good that has just been invented by a nonproducer (e.g., by a government laboratory), so that all prospective producers make simultaneous capacity choices in stage 1. Does forward induction also select the APS equilibrium in this environment? The answer is negative. Ben-Porath and Dekel (1992) have shown that, if both players of a single-stage game have the opportunity to publicly burn a dollar beforehand, then all equilibria of the single-stage game can be consistent with forward induction. The upshot is that there is a theoretical model in which capacity choice functions as communication of collusive intent in the context of one, specific, initial configuration of an industry, but that the model does not show investment invariably to succeed to function in that way. In the current state of its development, game theory does not provide any more concrete guidance about how general may be the circumstances in which tacit coordination might arise via forward-induction reasoning.

492   Edward J. Green, Robert C. Marshall, and Leslie M. Marx

19.4.6.  Blundering into Tacit Coordination In the introductory section of this chapter, we stated, “There has been broad agreement in principle that monopoly conduct can arise spontaneously in highly concentrated markets that satisfy some other (possibly restrictive) conditions.” Building largely on the analysis provided by Aumann (1990) and Aumann and Brandenburger (1995) of Nash equilibrium as an outcome of rational interaction among players, we have found on the whole that the conditions for monopoly conduct to arise spontaneously as a result of such rational interaction are indeed restrictive. But Aumann and Brandenburger provided sufficient conditions for players’ conjectures to be a Nash equilibrium, not necessary ones. They mentioned the possibility that players might “blunder into an equilibrium by accident,” even in a situation where their sufficient conditions for Nash equilibrium were not satisfied. To assess this possibility requires a complement to the theory of outcomes of rational interaction under strong assumptions about higher-order knowledge of other players’ rationality: a theory of equilibrium as a long-run outcome of “blundering” by (predominantly) rational agents. Those agents can be modeled as making “optimal” decisions based on understanding their environment and on knowing what actions other players are using (on average, at least), but without consideration of what other agents might be thinking and of how their future actions might change as a result of their own optimization/reasoning processes. This complementary theory is the topic of evolutionary game theory. Like other theories, this one is actually a family of related models. Virtually all evolutionary models have the implication that behavior almost surely approaches some Nash equilibrium in the long run. The models that make the most precise predictions introduce a low level (made asymptotic to zero, in formal analysis) of irrational behavior that corresponds to mutation in a biological population. The canonical models of this sort were formulated by Young (1993) and by Kandori, Mailath, and Rob (1993). Analysis of those models provides a formal foundation for the informal Aumann-Harsanyi argument that when one Nash equilibrium is payoff dominant and another one is risk dominant, play will converge to the risk-dominant equilibrium. As they relate to our analysis in section 19.4.4, those models suggest that firms will not blunder into tacit coordination. However, by relaxing only slightly the assumptions of those models about the optimality of nonmutant actions, Fudenberg and Imhof (2008) formulate a model in which the efficient outcome can be selected in repeated prisoner’s dilemma games that are the abstract analogue of the repeated-interaction environments discussed earlier in this chapter. Thus, Fudenberg and Imhof ’s model lends some support to a favorable view toward the possibility of competitors blundering into tacit coordination. Sabourian and Juang (2008) ably survey general issues regarding use of evolutionary game theory to select the most plausible among the Nash equilibria in an environment. As they mention, the predictions of particular evolutionary models are sensitive to the specific assumptions that are incorporated in those models, and that might not be persuasive to a skeptic. Bergin and Lipman (1996) have shown that, if “mutations” are

Tacit Collusion in Oligopoly   493

modeled as being state dependent, then any strict Nash equilibrium (that is, equilibrium in which each player’s Nash action is strictly preferred to all of his other actions) can be selected. The gist of Bergin and Lipman’s research is that, until economists develop intuitively compelling justifications for specific mutation processes, evolutionary game theory will remain an insightful but inconclusive framework for reasoning about whether “blundering” will lead to payoff-dominant (that is, tacitly coordinated) outcomes, or to risk-dominant (that is, noncollusively oligopolistic) ones.

19.5.  Tacit Coordination in Antitrust Litigation A potentially collusive pattern of prices is a violation of the first section of the Sherman Act only if it is shown that there is an agreement among competitors. Unfortunately, there is no recipe to identify agreements from economic circumstantial evidence if the environment is such that collusive prices may arise without communication. Posner (2001) states that in some circumstances competing sellers might be able to coordinate their pricing without conspiring in the usual sense of the term—that is, without any overt or detectable acts of communication. This is the phenomenon that lawyers call “conscious parallelism” and some economists term “oligopolistic interdependence,” but which I prefer to call tacit collusion in contrast to explicit collusion of the formal cartel or its underground counterpart. (Posner, 2001, 52–53)

In the same chapter, Posner (2001) tells the history of the price-fixing criminalization including the possibility of tacit coordination. He argues that just after the Sherman Act, judges and lawyers based their cases in the mere fact of explicit collusion and not in the economic consequences, probably because lawyers were more comfortable with conspiracy doctrine that with an economic theory of pricing. Posner (2001) points out the inadequacy of considering the existence of overt communication as the only decisive factor in collusion prosecution. Once the economic effects are taken into account, one must consider the possibility that a seemingly collusive price may have been reached without communication or even an agreement but from the understanding of the strategic interdependence. Posner (2001) approaches the issue of what constitutes a tacit agreement by referring to Turner (1962). For Turner (1962), whether a pattern of prices was reached through an agreement or not could be “considered purely as a problem in linguistic definition.” Posner (2001) then discusses when “plus factors” may allow the inference of explicit collusion from economic evidence.39 39 

See Kovacic et al. (2011) on identifying the strength of various plus factors.

494   Edward J. Green, Robert C. Marshall, and Leslie M. Marx Baker (1993) also reached the conclusion that distinguishing an “agreement” among competitors when the evidence is “entirely circumstantial” is difficult from both a legal and an economic perspective: Courts and commentators have debated for decades whether parallel price changes by oligopolists who recognize their interdependence provide a sufficient basis for a court to infer an unlawful horizontal agreement under Sherman Act 1, and if not what additional circumstantial evidence is required to prove a conspiracy. (Baker, 1993, 144)

Baker (1993) concludes that regardless of the requirements to distinguish among coordination and agreement, antitrust policy should prevent practices that facilitate oligopoly coordination, which is the approach of the European Commission. Turning to Europe, Garces-Tolon, Neven, and Seabright (2009) and Ivaldi and coauthors (2007) analyze the evolution of tacit coordination in Europe from 1990, year in which merger control became a responsibility of the European Union. Ivaldi and coauthors (2007) point out that tacit coordination, or collective dominance as it is called by the European Commission, has been a controversial issue. According to the article, the interpretation of collective dominance in litigation used to have no structural links and it was not until recent years that it was more precisely specified. Consistency is up to some point due to a regulation in 2004 in which the Commission declared “incompatible with the EC treaties mergers that create or strengthen a dominant position as a result of which effective competition would be significantly impeded” (Ivaldi et al., 2007, 217– 18). In this sense, the European Commission is relying on game theory, mainly results on subgame perfect equilibrium in repeated games, to determine whether or not a concentration or a merger is strengthening a dominant position. An analysis of how results in repeated games have been used in litigation can be found in Garces-Tolon, Neven, and Seabright (2009). Ivaldi and coauthors (2007) also analyze the impact of game theory on European merger policy. We hope this chapter facilitates communication between lawyers and economists on these issues by providing a common language and framework for discussion. It is clear that for years to come antitrust enforcement authorities will wrestle with the identification of what types of conduct violate antitrust laws.

Acknowledgments The authors thank the Human Capital Foundation (http://www.hcfoundation.ru/en/), and especially Andrey Vavilov, for financial support. The paper benefited from discussions with Roger Blair, Louis Kaplow, Bill Kovacic, Vijay Krishna, Steven Schulenberg, and Danny Sokol. We thank Gustavo Gudiño for valuable research assistance.

Tacit Collusion in Oligopoly   495

References Abreu, Dilip. 1986. Extremal Equilibria of Oligopolistic Supergames. Journal of Economic Theory 39: 191–225. Abreu, Dilip, David Pearce, and Ennio Stacchetti. 1986. Optimal Cartel Equilibria with Imperfect Monitoring. Journal of Economic Theory 39(1): 251–69. Abreu, Dilip, David Pearce, and Ennio Stacchetti. 1990. Toward a Theory of Discounted Repeated Games with Imperfect Monitoring. Econometrica 58(5): 1041–63. Athey, Susan, Kyle Bagwell, and Chris Sanchirico. 2004. Collusion and Price Rigidity. Review of Economic Studies 71(2): 317–49. Aumann, Robert J., Nash Equilibria Are Not Self Enforcing. In J. J. Gabszewicz, J.-F. Richard, and L. Wolsey. 1990. Economic Decision Making: Games, Econometrics, and Optimisation: Essays in Honor of Jacques Drèze. Amsterdam: Elsevier Science Publishers, 201–6. Aumann, Robert J., and Adam Brandenburger. 1995. Epistemic Conditions for Nash Equilibrium. Econometrica 63: 1161–80. Bagwell, Kyle, and Robert W. Staiger. 1997. Collusion over the Business Cycle. Rand Journal of Economics 28: 82–106. Bain, J. 1968. Industrial Organization. New York: Wiley. Baker, Jonathan B. 1993. Two Sherman Act Section 1 Dilemmas: Parallel Pricing, the Oligopoly Problem, and Contemporaneous Economic Theory. Antitrust Bulletin 38: 143–219. Benoit, Jean-Pierre, and Vijay Krishna. 1987. Dynamic Duopoly: Prices and Quantities. Review of Economic Studies 54: 23–35. Ben-Porath, Elchanan, and Eddie Dekel. 1992. Signalling Future Actions and the Potential for Sacrifice. Journal of Economic Theory 57: 36–51. Bergin, James, and Barton L.  Lipman. 1996. Evolution with State-Dependent Mutations. Econometrica 64: 943–956. Bernheim, B. Douglas. 1984. Rationalizable Strategic Behavior. Econometrica 52: 1007–28. Borenstein, Severin, and Andrea Shepard. 1996. Dynamic Pricing in Retail Gasoline Markets. Rand Journal of Economics 27: 429–51. Brock, William A., and Jose A.  Scheinkman. 1985. Price Setting Supergames with Capacity Constraints. Review of Economic Studies 52: 371–82. Carlsson, Hans, and Eric van Damme. 1993. Global Games and Equilibrium Selection. Econometrica 61: 989–1018. Carlton, Dennis W., Robert H. Gertner, and Andrew M. Rosenfield. 1997. Communications among Competitors: Game Theory and Antitrust. George Mason Law Review 5: 423–40. Charness, Gary. 2000. Self-Serving Cheap Talk:  A  Test of Aumann’s Conjecture. Games and Economic Behavior 33: 177–94. Ely, J., J. Hörner, and W. Olszewski. 2005. Belief-Free Equilibria in Repeated Games. Econometrica 73(2): 377–415. Ely, J., and J. Valimaki. 2002. A Robust Folk Theorem for the Prisoner’s Dilemma. Journal of Economic Theory 102(1): 84–105. Friedman, James W. 1977. Oligopoly and the Theory of Games. New York: North Holland. Fudenberg, Drew, and Lorens A.  Imhof. 2008. Monotone Imitation Dynamics in Large Populations. Journal of Economic Theory 140: 229–45. Fudenberg, Drew, and Eric Maskin. 1986. The Folk Theorem in Repeated Games with Discounting or Incomplete Information. Econometrica 54: 533–54.

496   Edward J. Green, Robert C. Marshall, and Leslie M. Marx Garces-Tolon, E., D. Neven, and P. Seabright. 2009. The Ups and Downs of the Doctrine of Collective Dominance: Using Game Theory for Merger Policy. In B. Lyons, Cases in European Competition Policy:  The Economic Analysis, 349–82. Cambridge:  Cambridge University Press. Green, Edward J., and Robert H. Porter. 1984. Noncooperative Collusion under Imperfect Price Information. Econometrica 52(1): 87–100. Harrington, Joseph E., Jr. 2012. Evaluating Mergers for Coordinated Effects and the Role of “Parallel Accommodating Conduct.” Working paper, Johns Hopkins University. Harsanyi, John C. 1995. A New Theory of Equilibrium Selection for Games with Complete Information. Games and Economic Behavior 8: 91–122. Harsanyi, John C., and Reinhard Selten. 1988. A General Theory of Equilibrium Selection in Games. Cambridge, MA: MIT Press. Hörner, Johannes, and Julian Jamison. 2007. Collusion with (Almost) No Information. Rand Journal of Economics 38: 804–22. Isaac, R. Mark, and Charles R. Plot. 1981. Price Controls and the Behavior of Auction Markets: An Experimental Examination. American Economic Review 71(3): 448–59. Ivaldi, Marc, Bruno Jullien, Patrick Rey, Paul Seabright, and Jean Tirole. 2003. The Economics of Tacit Collusion. Final Report for DG Competition, European Commission. Ivaldi, M., B. Jullien, P. Seabright, and J. Tirole. 2007. The Economics of Tacit Collusion:  Implications for Merger Control. In V. Ghosal and J. Stennek, The Political Economy of Antitrust, 217–39. Boston: Elsevier. Kandori, Michihiro, George J. Mailath, and Rafael Rob. 1993. Learning, Mutation, and Long-Run Equilibria in Games. Econometrica 61: 29–56. Kaplow, Louis. 2011. On the Meaning of Horizontal Agreements in Competition Law. California Law Review 99(3): 683–818. Kaplow, Louis. 2013. Competition Policy and Price Fixing. Princeton, NJ: Princeton University Press. Kaplow, Louis, and Carl Shapiro. 2007. Antitrust. In A. Mitchell Polinsky and Steven Shavell, Handbook of Law and Economics, vol. 2, 1073–225. Amsterdam: North Holland. Kaysen, Carl. 1951. Collusion under the Sherman Act 1. Quarterly Journal of Economics 65(2): 263–70. Kovacic, William E., Robert C. Marshall, Leslie M. Marx, and Halbert L. White. 2011. Plus Factors and Agreement in Antitrust Law. Michigan Law Review 110(3): 393–436. Lambson, Val E. 1994. Some Results on Optimal Penal Codes in Asymmetric Bertrand Supergames. Journal of Economic Theory 62: 444–68. Levenstein, Margaret C., and Valerie Y. Suslow. 2011. Breaking Up Is Hard to Do: Determinants of Cartel Duration. Journal of Law and Economics 54(2): 455–92. Mailath, George J., and Larry Samuelson. 2006. Repeated Games and Reputations. New York: Oxford University Press. Marshall, Robert C., and Leslie M. Marx. 2012. The Economics of Collusion: Cartels and Bidding Rings. Cambridge, MA: MIT Press. Myerson, Roger B. 1991. Game Theory, Analysis of Conflict. Cambridge, MA: Harvard University Press. Pearce, David G. 1984. Rationalizable Strategic Behavior and the Problem of Perfection. Econometrica 52: 1029–50. Porter, Robert H. 1983a. Optimal Cartel Trigger Price Strategies. Journal of Economic Theory 29(2): 313–38.

Tacit Collusion in Oligopoly   497

Porter, Robert H. 1983b. A Study of Cartel Stability: The Joint Executive Committee, 1880–1886. Bell Journal of Economics 14: 301–14. Posner, Richard A. 1976. Antitrust Law:  An Economic Perspective. Chicago:  University of Chicago Press. Posner, Richard A. 2001. Antitrust Law. 2nd ed. Chicago: University of Chicago Press. Radner, Roy. 1986. Repeated Partnership Games with Imperfect Monitoring and No Discounting. Review of Economic Studies 53: 43–57. Rubinstein, Ariel. 1989. The Electronic Mail Game: Strategic Behavior under “Almost Common Knowledge.” American Economic Review 79: 385–91. Sabourian, Hamid, and Wei-Torng Juang. 2008. Evolutionary Game Theory: Why Equilibrium and Which Equilibrium. In S. Bold, B. Lowe, T. Rash, and J. van Benthem, Foundations of the Formal Sciences V: Infinite Games. London: College Publications, 187–222. Schelling, Thomas C. 1960. The Strategy of Conflict. Cambridge, MA: Harvard University Press. Schenone, P. 2012. A Necessary and Sufficient Condition for a PBE Folk Theorem. Innocenzo Gasparini Institute for Economic Research Job Market Paper, January 23. Scherer, F. 1970. Industrial Market Structure and Economic Performance. Chicago:  Rand McNally. Schwartz, Louis B., John J.  Flynn, and Harry First. 1983. Free Enterprise and Economic Organization: Antitrust 439. 6th ed. Mineola, NY: Foundation Press. Slade, Margaret E. 1992. Vancouver’s Gasoline-Price Wars: An Empirical Exercise in Uncovering Supergame Strategies. Review of Economic Studies 59: 257–76. Stigler, George J. 1964. A Theory of Oligopoly. Journal of Political Economy 72: 44–61. Tan, Tommy C.-C., and S. R. da Costa Werlang. 1988. The Bayesian Foundations of Solution Concepts of Games. Journal of Economic Theory 45: 370–91. Tirole, Jean. 1988. The Theory of Industrial Organization. Cambridge, MA: MIT Press. Turner, Donald F. 1962. The Definition of Agreement under the Sherman Act:  Conscious Parallelism and Refusals to Deal. Harvard Law Review 75(4): 655–706. van Damme, Eric. 1989. Stable Equilibria and Forward Induction. Journal of Economic Theory 48: 476–96. Yamamoto, Y. 2014. Individual Learning and Cooperation in Noisy Repeated Games. Review of Economic Studies 81(1): 473–500. Young, Peyton. 1993. The Evolution of Conventions. Econometrica 61: 57–84.

CHAPTER 20

AU C T IO N S A N D B I D R IG G I N G KEN HENDRICKS, R. PRESTON MCAFEE, AND MICHAEL A. WILLIAMS

20.1.  Introduction and Overview 20.1.1. Auctions An auction is “a market institution with an explicit set of rules determining resource allocation and prices on the basis of bids from the market participants” (McAfee and McMillan, 1987).1 Auctions are among the oldest market institutions; “marriage auctions” for brides were used in Babylon as early as 500 BC. Products commonly bought and sold using auctions include art and jewelry, US Treasury Bills, used cars, the assets of bankrupt companies, and radio spectrum. Auctions are also commonly used in the government procurement of goods and services. Online auctions comprise a significant portion of the rapidly growing electronic commerce marketplace. Auction theory provides an explicit model of price formation. Traditional models of competition that furnish equilibrium prices given demand and supply characteristics, such as Cournot and Bertrand models of competition, do not explain the price-formation process in terms of buyer-seller interactions. In contrast, auction mechanisms explicitly lay out rules that govern the formation of prices. A typical auction consists of a single auctioneer, responsible for selling an object, and a number of bidders who wish to buy the object. The auctioneer may announce a reserve price, that is, the

1  A number of survey papers on auction theory and empirical work have been published. See, e.g., McAfee and McMillan (1987), Wilson (1992), Klemperer (1999), and Hendricks and Porter (2007). In addition, Klemperer (2008) analyzes the use and misuse of bidding markets in antitrust economics.

Auctions and Bid Rigging   499

lowest price at which the object would be sold. There are four basic types of single-item auctions analyzed in the economics literature: (1) the English auction, (2) the Dutch auction, (3) the first-price sealed-bid (FPSB) auction, and (4) the second-price sealed bid (SPSB) auction (also called a Vickrey auction). These four types of auctions are characterized by the following two properties: (1) whether bids are open or closed (i.e., sealed) and (2) whether the winning bidder pays an amount equal to the winning bid or the highest losing bid. An English auction is an open-bid auction in which bidders sequentially raise the price of the item until no bidder is willing to raise it further. A Dutch auction is an open-bid auction that proceeds in the opposite direction: The seller begins the auction at a particular high price, and lowers the price until a bidder is willing to purchase the item. In both FPSB and Dutch auctions, the winning bidders pay an amount equal to the winning bid. In a closed auction, each bidder may submit only a single, sealed bid (that is, no bidder may observe another bidder’s bid). The seller then considers all submitted bids simultaneously. In FPSB and SPSB auctions, bids are submitted in a sealed fashion and are considered simultaneously by the seller to determine the winning bidder. In a FPSB auction, the highest bidder wins the item and pays an amount equal to the highest bid, that is, the winning bid. In a SPSB auction, the highest bidder wins the item and pays an amount equal to the second-highest bid, that is, the highest losing bid. In all four auction formats, the seller does not sell the item if the price determined by the auction is below the reserve price. In the theoretical literature on auctions, such as Milgrom and Weber (1982), English auctions are usually approximated by Japanese auctions for the sake of analytical convenience. In a Japanese auction, the seller starts the auction with the price of the item at a minimum level. Each bidder presses and holds down a button while the price of the item is raised continuously. Bidders drop out of the auction by releasing their button when the price of the item exceeds their willingness to pay, that is, their valuation. Each bidder can observe all of the bids. The auction ends when the second-highest bidder drops out, so the winning bidder is the last remaining bidder. The continuous increase in price in a Japanese auction simplifies the theoretical analysis and makes it transparent that the winning bid in an English auction is approximately the second-highest valuation.

20.1.2.  Standard Auction Models The economic literature on auctions typically distinguishes two extreme types of information environments: private value (PV) and common value (CV). In a PV auction, each bidder knows her own value but not those of other bidders. For example, an art auction with art collectors who do not intend to resell the item would be analyzed using the PV framework. Each bidder’s valuation is a personal characteristic that does not depend on other bidders’ valuations. In a CV auction, the value of the item up for auction is the same for all bidders, but that value is unknown to the bidders before the auction. Prior to

500   Ken Hendricks, R. Preston McAfee, and Michael A. Williams bidding, bidders receive different private signals about the true value of the object. For example, in an offshore oil tract auction, bidders conduct their own geologic research to form their private estimates of the value of an offshore tract. The standard economic model of an auction consists of a single seller and n bidders, i = 1, 2, ..., n, who bid at an auction to purchase a single indivisible good. Bidder i’s valuation for the good is Vi , a real-valued random variable distributed according to the cumulative probability distribution Fi . The value of V is known only to bidder i , but Fi is known to the seller and the remaining bidders. In the symmetric independent private value model (IPV), the valuations of bidders are assumed to be independent draws from the same distribution F. In the CV model, the valuations of bidders are informative signals about a common value V unobservable to the bidders. In this case, Vi is typically modeled as an independent draw from a conditional distribution, H (.| V ) , known to the bidders. Bidders in a common value auction are susceptible to the “winner’s curse.” A bidder tends to win when she overestimates V and wins rarely when she underestimates V . The PV and CV frameworks also have been merged to study auction models in which bidders’ information consists of both private and common value components; see, for example, Milgrom and Weber (1982) or Goeree and Offerman (2003). Auctions are generally modeled as a Bayesian game, and bidders have Bayesian-Nash equilibrium bidding strategies. In a FPSB IPV auction, the unique equilibrium with n risk-neutral, symmetric bidders is the collection of bidding functions x n −1 bi (vi ) = vi − [( F ( s ) ds ) / F (x )n −1 ] for each i = 1, 2,…, n . So each bidder bids less than ∫ −∞ her value, and the markdown factor decreases with the number of bidders. In contrast, in a SPSB PV auction, each bidder has a weakly dominant strategy to bid her valuation, regardless of the number of bidders. In a FPSB CV auction, each bidder shades her bid relative to her expected value of the object conditional on winning. However, the shading factor typically increases with the number of bidders due to the winner’s curse. In SPSB CV auctions, the equilibrium is not in dominant strategies and not unique without some refinement. One equilibrium involves bidding more than the average value (specifically bidding the value conditional on a second bidder having the same signal). In this case, the actual bid will exceed the willingness to pay, although the second-highest bid, the price, will not. A fundamental (and remarkable) result in auction theory is the revenue equivalence theorem for IPV auctions. An auction mechanism (e.g., an English or Dutch auction) is efficient if the mechanism allocates the object to the bidder with the highest valuation (Dasgupta and Maskin, 2000). The revenue equivalence theorem essentially states that all efficient, symmetric IPV auctions yield the same expected revenue. Thus, for IPV auctions, the expected winning bid is the same whether the auction mechanism is English, Dutch, FPSB, or SPSB. A necessary assumption for the theorem to hold is that the bidder with the lowest feasible valuation expects zero surplus. A first version of this result was obtained by Vickrey (1961) and later generalized by Myerson (1981) and Riley and Samuelson (1981). The revenue equivalence result does not hold, in general, for common-value auctions.

Auctions and Bid Rigging   501

20.1.3.  Collusion in Auctions: Bid Rigging In a bid-rigging scheme, bidders agree to collude to limit competition and obtain the good for a lower price (or higher in the case of procurement) than would result in the absence of such explicit collusion.2 A group of bidders who collude in an auction is called a bidding ring. Bidding rings may be all-inclusive, that is, include every bidder participating in an auction, or incomplete, consisting only of a subset of the set of all bidders. As Marshall and Marx (2012, p. 175) summarize: “Colluding bidders suppress their rivalry through the elimination of meaningful bids by all colluding bidders except for the ring bidder with the highest value. All other details and logistics of bidder collusion flow from this foundational principle of the ring.” Bid rigging is generally illegal in the United States, the EU, China, and many other countries.3 Bid rigging is the most common form of explicit collusion in auction markets (the other, less common, form being explicit collusion between one or more bidders and the auctioneer). This should be expected since noncooperative behavior is not jointly optimal for bidders. Bidders are collectively better off colluding and transferring gains from trade from the seller to the ring. Rings differ from cartels insofar as they focus on individual auctions rather than a broad cross-section of product and geographical markets. The focus simplifies the task of a ring: it needs to coordinate the bids submitted by bidders. However, in doing so, rings face some of the same challenges as cartels: detection by the antitrust agencies or by the seller, internal enforcement, entry, and private information about the gains from trade. The ability of rings to solve these problems, and the nature of the solution, depends on the type of auction the seller uses. One solution is to have all bidders submit identical bids, preferably at or near the reserve price, and let the seller randomly select the winning bidder. This scheme wastes a lot of the potential surplus from collusion because the good is unlikely to go to the bidder with the highest valuation. It is also difficult to enforce. Individual bidders have a strong incentive to bid slightly more than the agreed-upon price and win the good for sure. It is also easy to detect. Not surprisingly, this simple form of collusive bidding has become much less common after the US Federal Trade Commission prosecuted numerous cases of identical bidding in the mid-twentieth century. A second solution is to have 2  There is a substantial economic literature on noncooperative, tacit collusion in auctions. For theoretical analyses, see, e.g., Brusco and Lopomo (2002), Fabra (2003), and Blume and Heidhues (2006). For empirical analysis, see, e.g., Cramton and Schwartz (2000) and Ishii (2009). For experimental analyses, see, e.g., Sherstyuk (1999), Sherstyuk (2002), Kwasnica and Sherstyuk (2007), Li and Plott (2008), Sherstyuk and Dulatre (2008), Boone et al. (2009), Haan, Schoonbeek, and Winkel (2009), Potters (2009), Phillips and Menkhaus (2009), Brown, Plott, and Sullivan (2009), Hinloopen and Onderstal (2010), and Müller and Normann in volume 1 of this Handbook. In addition, there is also a substantial economic literature on corruption in auctions, i.e., bidders colluding with the auctioneer. See, e.g., Compte, Lambert-Mogiliansky, and Verdier (2005). We do not analyze these literatures here. 3  An interesting exception in the United States is the legality of explicit bid rigging in hostile takeovers of publicly traded companies. See McAfee et al. (1993).

502   Ken Hendricks, R. Preston McAfee, and Michael A. Williams bidders pay kickbacks to each other as payment for either refusing to bid or for submitting “phantom” bids, that is, intentionally losing bids. In order to avoid detection, bidders coordinate on the amount of the “serious” bid, and then “nonserious” bidders place phantom bids. This technique requires more communication and explicit discussion of prices, but makes collusion more difficult to detect by antitrust authorities using statistical methods (see discussion below). The side payments among bidders can also solve the problem of private information. The gains from trade captured by the ring from the seller, called the collusive surplus, are maximized when the serious bid is submitted by the ring member with the highest valuation. But in order to make this assignment, the members of the ring need to be incentivized to reveal their valuations. In one-shot auctions, rings typically solve this issue by holding a knockout auction, either before or after the seller’s auction, for example, Marshall and Marx (2012). The bidder with the highest valuation typically wins the knockout auction. The ring members share a sum of money equal to the difference between the price in the ring’s knockout auction and the price in the original auction. How the collusive surplus is allocated among ring members depends on the nature of the ring. If the ring members are symmetric, for example, their valuations are drawn from the same distribution, the ring may agree to allocate equal shares of the collusive surplus to each ring member. Alternatively, if some members of the ring are economically stronger than others, for example, have lower costs or higher valuations, the ring may allocate the collusive surplus based on those characteristics; see Graham, Marshall, and Richard (1990). For example, Asker (2010) analyzes how members of a bidding ring for the purchase of stamps in the United States allocated the collusive surplus among ring members depending on their bids in a knockout auction. Ring members shared each increment between bids in the knockout auction, provided their bids were above the auction price in the original auction. Half the increment was kept by the winner of the knockout auction, with the balance shared equally between those bidders who bid equal to or more than the incremental bid in the knockout auction. In some auctions, for example, procurement contracts and corporate takeovers, a bid-rigging ring may allocate the collusive surplus by using subcontracts between ring members, by bringing “losing” ring bidders back into the deal as equity owners, or by splitting the takeover target into pieces (see McAfee et al., 1993). In repeated auctions, the ring has more scope for meeting the challenges of collusion. The simplest assignment is for ring members simply to take turns, with each being the sole bidder in a given auction. More sophisticated bid rotation schemes involve bidders communicating with each other regarding items they would like to win. The ring can solve the incentive problem without knockout auctions or side payments by agreeing on future allocations (Athey and Bagwell, 2001). Bidders can keep a “tally sheet” recording each bidder’s winnings to ensure the bidders’ wins approximately balance out over time. This type of scheme is more efficient than simple bid rotation schemes, but requires more coordination and communication, thus increasing the likelihood the collusion is discovered and prosecuted by antitrust authorities.

Auctions and Bid Rigging   503

The enforcement issue is also more easily solved by rings in repeated auctions. To deter bidders from cheating on a collusive agreement in repeated auctions, the ring can punish members who cheat by reversion to non-cooperative bidding. A “bidding war” ensues in which all firms pay high prices and earn low profits until the collusive agreement can be reestablished (see, e.g., Porter, 1983 and 2005 in the context of a railroad cartel). In what follows, we will survey the literature on the theory and practice of bidding rings in one-shot auctions and in repeated auctions. The main theme is how the type of auction, whether it is first-price or second-price, sealed bid or oral, affects the incentive of bidders to collude and the way in which they collude.

20.2.  Bid Rigging: Theory 20.2.1.  The Incentive to Collude in Different Auction Formats Economic theory shows that the incentive of bidders to collude in an auction depends in part on the auction format (see, e.g., Robinson, 1985, Waehrer, 1999, and Marshall and Marx, 2007). In an ascending bid auction, for example, an English auction, the ring bids up to the highest valuation of its members. This bidding strategy is the same as if the ring member with the highest valuation had bid when there was no collusion. Thus, if in the presence of the bidding ring a non-ring member wins the auction, that non-ring member pays the same price that she would have paid in the absence of the ring. As Marshall and Marx (2012, p. 176) discuss, this has an important implication: “if a ring wins, and there are gains to their collusion, then the ring captures all of those gains; however, if the ring loses, there are no gains from the collusion for non-ring bidders. The collusion is self-contained in this sense.” In contrast, at a sealed-bid auction, for example, a FPSB auction, the ring submits a bid lower than the amount that the ring member with the highest valuation would have bid in the absence of the collusion. This shading of the ring’s bid can sometimes result in a non-ring member winning the collusive auction when that non-ring member would not have won the auction in the absence of the collusion. As Marshall and Marx (2012, p. 176) note: “the extra bid shading by the highest-valuing colluding bidder opens the possibility that the ring does not capture all the gains to its collusive conduct. The non-colluding bidders are beneficiaries, in expected terms, from the collusion.  . . .The leakage of some of the collusive gain to the outside bidders, which is absent at the ascending-bid auction, means the incentives for suppression of rivalry through collusion are typically weaker at the sealed-bid auction than at the ascending-bid auction.” An interesting corollary to this discussion relates to the incentives of non-ring members to join a ring. In the case of ascending bid auctions, non-ring members have no

504   Ken Hendricks, R. Preston McAfee, and Michael A. Williams incentive to remain outside of the ring, since the ring captures all the collusive gains. In contrast, in the case of FPSB auctions, non-ring members may find it more profitable on average to remain outside the ring since the collusive gains leak out to non-ring members.

20.2.2.  Bid Rigging in One-Shot versus Repeated Auctions Given an auction format, what is the optimal, incentive-compatible collusive mechanism? Is the collusive mechanism efficient? Is it unique? Economic theory attempts to answer these questions. A primary issue in the theoretical analysis of bid rigging is stability. A collusive scheme cannot be sustained in the absence of an appropriate enforcement device, such as side payments or punishments. In the absence of such a device, bidders have an incentive to lie about their valuations or otherwise deviate from the collusive scheme. Theoretical studies of bid rigging in auctions fall into two broad areas: collusion in one-shot auctions and collusion in repeated auctions.

20.2.2.1.  Bid Rigging in One-Shot Auctions Graham and Marshall (1987) develop models of bidding rings in SPSB and English IPV auctions. They assume that all bidders are ex ante identical in the sense that their valuations are drawn from the same distribution. They describe a collusion mechanism in which, prior to the main auction, each ring member submits a sealed “reported bid” to a risk-neutral ring center. The ring center determines the two highest reported bids and selects the member with the highest reported bid to act as the sole bidder in the actual auction. The ring center specifies that the bid submitted in the actual auction should equal the highest reported bid submitted to the ring. Other ring members are instructed to bid zero or not submit a bid. If the ring member wins the main auction, she pays the auctioneer an amount equal to the second-highest bid of all bids submitted in the main auction. She also must pay the ring center the difference between (1) the second-highest reported bid to the ring and (2) the second-highest bid of all bids submitted in the main auction, if this amount is positive. Graham and Marshall show that the auctioneer’s best response to the formation of a bidding ring is to set a reserve price that increases as a function of the number of bidders in the ring. For any given number of bidders in a ring, each ring member’s payoff to collusion decreases as the reserve price increases. Thus, for any given reserve price, the expected payoff to ring members increases as the number of bidders in the ring increases. Therefore, the Nash equilibrium is characterized by (1) a ring that includes all bidders and (2) a reserve price optimal for such a ring. From an ex ante perspective, the ring center’s expected payments to ring members equal the center’s expected revenues from the ring member who submits the highest reported bid to the ring. In this sense, the ring center is ex ante budget balancing. However, a difficulty with the Graham and Marshall collusion mechanism is that the

Auctions and Bid Rigging   505

ring center is not ex post budget balancing. That is, ex post the center’s expected payouts exceed its expected revenues. Mailath and Zemsky (1991) relax the identical-bidder assumption in Graham and Marshall (1987) and show that an ex post budget balancing efficient collusion can be achieved in a second-price auction. Furthermore, this outcome is possible even for a proper subset of bidders. They construct an explicit mechanism that implements this ex post efficient collusive result. With this efficient mechanism, the collusive surplus can always be divided up in such a way that, not only will every bidder wish to participate, but every subset of bidders will also wish to participate. As a consequence, no subset of bidders can do better by colluding among themselves and excluding the other bidders. That is, there exist allocations of the collusive surplus achieved by the ring coalition that make each subring better off than it could be in the absence of the ring. In contrast to the study of bidding rings in SBSP and English auctions, McAfee and McMillan (1992) analyze bidding rings in a FPSB IPV game.4 They define two types of cartels: weak cartels in which cartel members cannot make side payments, and strong cartels in which side payments are permitted. For weak cartels, they prove that the optimal collusive scheme (for a large class of valuation distribution functions) is identical bidding. All bidders with valuations above the reserve price bid the reserve price, and those with valuations below the reserve price submit bids equal to zero. They show that the optimal collusive scheme is inefficient, since the winning bidder is picked at random by the auctioneer out of all the bidders submitting identical bids. Thus, the winning bidder is not necessarily the bidder with the highest valuation for the object. For strong cartels, McAfee and McMillan (1992) demonstrate that an efficient, optimal mechanism exists for all-inclusive rings. They show this mechanism can be implemented by having the bidders hold a first “knockout” FPSB auction, and then having the winner of this knockout round participate in the actual auction if that winner’s bid in the knockout round exceeds the reserve price. The winning cartel member then pays each of the losers an equal share of the difference between her bid in the prior, knockout auction and the reserve price. McAfee and McMillan (1992) characterize optimal collusive mechanisms when members of the bidding ring report their valuations to a direct, incentive-compatible mechanism, for example, a “center.” McAfee and McMillan assume the center specifies and enforces the bids. Marshall and Marx (2007) and Lopomo, Marx, and Sun (2011) extend the model of McAfee and McMillan (1992) by assuming the center cannot control the bids ring members submit at the auction, but the center can enforce side payments between ring members. Lopomo, Marx, and Sun (2011) show in this case that no collusive mechanism exists that improves bidders’ payoffs relative to noncooperative 4  McAfee and McMillan (1992) focus on private value auctions because the optimal ring mechanism in the pure common value case is trivial. Efficiency is attained regardless of which member gets the right to bid in the seller’s auction. Thus, an all-inclusive ring can use some exogenous method to allocate the right to one of its members, such as a random allocation with equal probability weights, and ask each bidder to report her valuation.

506   Ken Hendricks, R. Preston McAfee, and Michael A. Williams bidding even if side payments that depend only on ring members’ reported valuations are allowed. Lopomo, Marshall, and Marx (2005) show that in ascending-bid auctions, ring members can have an incentive to bid in ways that can cause the ring to win the item even though a non-ring bidder has the highest valuation. The ring mechanism may cause a ring member’s reported valuation to increase her payment if another member of the ring wins the auction. In this case, ring members have an incentive to report valuations in excess of their actual valuations. This causes the ring’s highest valuation to increase, so that the ring may win the auction even though a non-ring bidder has the highest valuation. Similarly, if the ring uses a knockout auction, ring members may have an incentive to bid in excess of their valuations if their payment in the knockout auction depends on their bids. Lopomo, Marshall, and Marx’s inefficiency result depends on two assumptions. First, there is no preauction communication among the ring members regarding their valuations. Second, the ring must ex post balance its budget. If the first assumption does not hold, then the results of Mailath and Zemsky (1991) show that an efficient explicit collusion mechanism exists. If the second assumption does not hold, then a modified version of the collusive mechanism in Graham and Marshall (1987) yields efficient explicit collusion without preauction communication. Finally, Hendricks, Porter, and Tan (2008) analyze whether efficient collusion is possible in FPSB common value auctions. Assuming a ring forms, efficiency is trivially obtained since the value of the item by definition is the same, that is, common, to all bidders. So which particular ring member acquires the item is irrelevant from the perspective of the efficiency of the collusion. However, Hendricks, Porter, and Tan show that the ring may be unable to form in the first place, even if the cartel is legal. Hendricks, Porter, and Tan show that in a common value auction, buyers who receive high signals regarding the true value of the object may prefer not to join a bidding ring. In the absence of a bidding ring, a buyer’s only expected payment is to the seller in the main auction. In the presence of a bidding ring, a buyer’s expected payment is the sum of two components: (1) the payment to the seller in the main auction and (2) the payment to other buyers in the knockout round. The expected payment of a buyer with a high signal to the seller in the main auction is lower with the bidding ring. However, with the bidding ring, low-signal bidders, free from worry regarding the winner’s curse, bid aggressively in the knockout round. Thus, the expected sum of (1) the payment to the seller in the main auction in the presence of a bidding ring and (2) payments to other buyers in the knockout round may be greater than the expected payment to the seller in the main auction in the absence of a bidding ring. In this case, a buyer who receives a high signal regarding the true value of the object prefers not to join a bidding ring. In summary, the literature on collusion in one-shot auctions establishes that bidders can collude efficiently in an IPV environment if they can engage in preauction communication and make side transfers to each other. Communication is essential because members of the ring have to reveal their private values in order to allocate the object to

Auctions and Bid Rigging   507

the member with the highest value. Side transfers are needed to ensure that the members tell the truth and bid accordingly.

20.2.2.2.  Bid Rigging in Repeated Auctions The theory of repeated games suggests that repeated play makes it easier for bidders to collude, since they can condition their behavior on bids and enforce collusive outcomes by threatening to respond to deviations with competitive bidding. This form of collusion is known as tacit collusion, and it is not illegal. The celebrated folk theorem establishes that sufficiently patient players can construct a self-enforcing scheme in which they act as would a single firm and, thereby, achieve first-best collusive profits. However, this result assumes that players do not have private information. Therefore, an important question addressed in the literature on collusion in repeated auctions is whether an all-inclusive ring in an IPV environment can earn first-best collusive profits when members do not communicate or make side transfers to each other. Before proceeding, we define efficient collusion in repeated auctions. A  collusive scheme is efficient if, in each auction, the ring (1) bids if and only if the highest valuation of its members exceeds the reserve price; (2) never pays the seller more than the reserve price; and (3) assigns the object to the member with the highest valuation. Efficient collusion is basically a (random) bid rotation scheme in which each member gets to win whenever her valuation is the highest and exceeds the reserve price. By definition, efficient collusion maximizes the ex post surplus of the ring and generates first-best collusive profits. In a seminal contribution, Skrzypacz and Hopenhayn (2004) establish that tacit collusion cannot achieve first-best collusive profits when bids are private information. They consider a model in which bidders bid repeatedly for identical objects sold sequentially over time. The seller publicly announces whether the object is sold and the identity of the winner but does not provide any information on the bids or the identities of the losing bidders. The bidders’ private valuations are distributed independently and identically across bidders and auctions. The auction format can be any one of the standard first-price or second-price auctions. Bidders cannot communicate or make side transfers to each other. In this environment, the authors derive an upper bound on the profits that a ring can obtain and show that this upper bound is substantially smaller than the profits in an efficient collusive scheme. The authors also characterize the types of collusive schemes that can achieve the upper bound on profits. They first note that any scheme in which, following every public history, bidders adopt symmetric bidding strategies does no better than a simple bid rotation scheme. This result corresponds to the weak cartel result of McAfee and McMillan (1992). That is, symmetric continuation values imply that the bidders share equally in future punishments and rewards so there are no transfers and, in that case, the best the ring can do is bid rotation. The schemes that do better than bid rotation involve treating bidders differently based on their history of wins and losses. The idea is to punish bidders that have more wins with a lower probability of winning, and even possible exclusion, in future auctions. As a result, winners have lower continuation values than

508   Ken Hendricks, R. Preston McAfee, and Michael A. Williams losers, and these differences work like transfers. But, despite these transfers, the lack of communication reduces collusive profits below the efficient level. Athey, Bagwell, and Sanchirico (2004) explore this trade-off between efficiency and profits in an infinitely repeated Bertrand game with inelastic demand. The firms’ costs are independently and identically distributed across firms and periods. The stage game is equivalent to a procurement auction since the firm that sets the lowest price wins the entire market. Consequently, the model is essentially the same as Skrzypacz and Hopenhayn except that bids are publicly observed. The authors characterize the set of symmetric equilibria and obtain two striking results. First, they show that, in any collusive equilibrium in which bidders use monotone bid functions, the ring achieves efficiency (due to the sorting of types) but earns profits that are no better than those generated by the noncooperative equilibrium. Second, the symmetric scheme that yields the highest profit to the ring involves rigid pricing; that is, bidders bid the reserve price and the seller randomly chooses a winner. This result is similar to the result obtained by Skrzypacz and Hopenhayn and generalizes the weak cartel result of McAfee and McMillan (1992) to repeated auctions with observable bids. If tacit collusion is inefficient, can communication among the members of the ring regarding their private information help them achieve first-best collusive profits? And, if so, what does the collusive scheme look like? These questions have been addressed in several papers. Athey and Bagwell (2001) consider a simplified version of the infinitely repeated Bertrand game discussed above. There are only two firms and their costs in each period are identical and independent Bernoulli random variables: “high” with probability η and “low” with probability 1 − η. As noted above, their model is equivalent to a repeated procurement auction. The firms explicitly, albeit illegally, collude. The authors assume that (1) the firms can communicate with each other in each period regarding their current costs, but (2) they cannot make side payments (so as to reduce the probability of detection by the antitrust authorities). Prices, that is, bids, are public information. The main result of their paper is that a sufficiently patient ring can achieve the first-best collusive profits. The efficient collusive scheme is similar to one that Skrzypacz and Hopenhayn construct to show that a ring can do better than bid rotation. In each period, the high-cost bidder does not bid and the low-cost bidder bids the reserve price. However, the high-cost bidder is favored in future auctions with a higher probability of being selected in the event that both bidders have same cost. These future “market-share” favors have no impact on efficiency since they only occur in periods in which the firms’ realized costs are the same. But the market-share favors mean that the continuation value of the low-cost bidder is lower than the continuation value of the high-cost bidder, and this difference acts like a transfer. The only issue is whether the magnitudes of the feasible market-share favors are sufficient to incentivize bidders to cooperate and report their costs truthfully. The authors show that, if the firms do not discount future profits too heavily, the answer is yes. Aoyagi (2003) analyzes a model of infinitely repeated auctions consisting of two symmetric bidders who bid every period on a single indivisible good. The auction format can

Auctions and Bid Rigging   509

be either first-price or second-price. The private signals of the bidders are real-valued and identically and independently distributed across bidders and auctions. The bidders report their private signals to a communication device, called a “center,” to coordinate their bids in each auction. After receiving the reported signals, the center subsequently instructs the bidders on how to bid in the auction. A “collusion scheme” consists of the center’s choice of instruction rule in every period as a function of the reports and the public histories. The public history is the sequence of instruction rules chosen by the center in past auctions and the bids in those auctions. A collusion scheme is an equilibrium if, for each bidder, telling the truth about her valuations is incentive compatible, and it is rational to adhere to the instructions. Aoyagi studies a class of dynamic bid rotation schemes with “grim trigger” punishments. In these schemes, the players begin in the collusion phase in which only one bidder is instructed to bid in a given stage auction. If a player deviates from the instructions, a punishment phase is triggered in which the one-shot Nash equilibrium results. Aoyagi’s dynamic bid rotation scheme works as follows. In the first phase, the center uses the efficient instruction rule. That rule instructs the bidder with the higher valuation (based on the bidders’ reported valuations to the center) to bid the reserve price if and only if the valuation exceeds the reserve price. The center instructs the other bidder not to bid. The difficulty is that the efficient instruction rule is not incentive compatible. Each bidder has an incentive to overstate their reported valuations to the center in hopes of winning the auction at the reserve price. To solve this problem, the bid rotation scheme has a second phase in which the payoff to the bidder with the highest reported valuation is reduced relative to the other bidder with some positive probability. The instruction rule in this phase is incentive compatible. The bid rotation scheme proceeds in this second phase for a fixed number of periods before reverting to the first stage. The collusive profits from Aoyagi’s scheme are not first-best but, as in Skrzypacz and Hopenhayn’s model, they are higher than the profits the ring would obtain from the scheme in which the bidders take turns winning at the reserve price independently of their values. The main difference between his model and that of Athey and Bagwell is the cardinality of the set of valuations. In the Athey and Bagwell model the set is finite, but in Aoyagi’s model the set is the unit interval. Thus, the kind of “market-share” favors that Athey and Bagwell use in their optimal collusion scheme is not possible in Aoyagi’s model, since the probability of both bidders having the same valuation is zero. In a subsequent paper, Aoyagi (2007) uses the same model but with a finite type space to show that a ring can achieve the first-best collusive profits. The optimal collusion scheme is fully efficient, that is, each bidder’s equilibrium payoff is close to what she would get if the object were allocated at the reserve price to the highest-valuation bidder in every stage auction. Using a similar dynamic bid rotation scheme to that in his 2003 paper, which allows for bidder communication, Aoyagi identifies conditions under which an equilibrium collusion scheme is fully efficient. One conclusion of the literature on repeated auctions is that bidders do not have to use side payments to implement the efficient bid rotation scheme. Strategies in which current losers are rewarded in future play can provide sufficient incentives for bidders to cooperate. A second conclusion is that sellers can make it difficult for bidders to achieve

510   Ken Hendricks, R. Preston McAfee, and Michael A. Williams first-best profits by not making bids public. In this case, bidders have to communicate to earn the full benefits from collusion, which makes a ring illegal and easier to detect. A third conclusion is that, when bidders are sufficiently patient, communication is sufficient to achieve first-best collusion profits.

20.3.  Bid Rigging: Empirical Studies 20.3.1.  Empirical Studies of Bid Rigging in Auctions Porter and Zona (1993) examine bidding for state highway construction procurement auctions in Long Island, New York, from 1979 to 1985.5 The New York Department of Transportation (“DOT”) awarded approximately $120 million in 186 separate highway contracts in this period. The DOT used the FPSB auction format in its highway construction procurement auctions. Porter and Zona first evaluate whether the characteristics of these procurement auctions would tend to facilitate collusion. The first characteristic of the auctions is the public nature of certain information. Prior to a given procurement auction, the DOT made public a “Plan Buyers List” that listed the firms that purchased the plans for that highway construction project. Thus, cartel members had knowledge of the set of potential bidders against whom they would bid before each procurement auction. On the day the winning (low) bidder was selected, the DOT publicly announced all bids and the identity of each bidder. As Porter and Zona note, this information allowed cartel members to detect deviations from an agreement. The second characteristic of the auctions is the DOT’s inelastic demand. Out of the 186 highway procurement auctions, 185 were ultimately funded and awarded to the low bidder. Thus, any increase in the winning (low) bid caused by successful collusion was captured as profits by the cartel. The third characteristic of the auctions is the structure of the market for highway construction services. Porter and Zona note that on the 25 largest construction jobs, the four largest firms accounted for 45% of the bids. The fourth characteristic of the market is the opportunity for the firms to communicate. Most of the bidders belonged to the same local trade associations, and joint bidding was allowed. The fifth characteristic was that the DOT tended to run its procurement auctions for larger jobs on a regular basis, at the beginning of each year. As Porter and Zona note, the regularity of the auctions would tend to make calculating the discounted present value of continued collusion more predictable. The final characteristic was that the highway construction firms were relatively homogeneous. They used the same technologies and purchased inputs from the same suppliers. 5 

See also Doane et al. in this volume for a discussion of detecting explicit collusion.

Auctions and Bid Rigging   511

Porter and Zona next identify five firms as possible members of a highway construction cartel. One of the firms was convicted in federal court of bid rigging on a highway construction job in Long Island in 1984, and the other four firms were unindicted coconspirators. The same five firms had been named as participants in bid-rigging conspiracies in other antitrust or racketeering suits in New York. Using this information, Porter and Zona divide their data into two sets: (1) bids from firms other than the five possible cartel firms (called bids from “competitive firms”) and (2) bids from the five possible cartel firms. They also restrict their data set to 75 road-paving jobs that had at least two bids from competitive firms. Porter and Zona first estimate a regression model in which the dependent variable is the log of a firm’s bid. The independent variables consist of measures of a firm’s capacity and capacity utilization, and dummy variables indicating whether the firm’s headquarters was on Long Island, and for competitive firms, whether that firm previously had won a highway construction auction on Long Island. Using first the bids from competitive firms, Porter and Zona conclude that the estimated regression fits the data reasonably well and the coefficient estimates have the expected signs. In contrast, using bids from the cartel firms, many of the estimated regression coefficients do not have the expected signs. A Chow test for equality of the estimated coefficients in the two regressions rejects the null hypothesis that the coefficients in the two regressions are equal. On the basis of these results, Porter and Zona conclude that they can reject the null hypothesis of no bid rigging. Finally, Porter and Zona estimate a second regression model that uses the same independent variables, but changes the dependent variable to the rank of a firm’s bid in a given auction, rather than the dollar value of its bid as used in their first regression model. They first estimate this model using bids from competitive firms. They estimate three regressions, one using all the bids, one using just the low bids, and another using all bids other than the low bid. Comparing the estimated coefficients in the three regressions, they cannot reject the null hypothesis of no model misspecification. On this basis, they cannot conclude that the bids from competitive firms are generated by different processes depending on whether the bids are low or not. Porter and Zona then estimate the same regression model using bids from the cartel firms, again estimating three regressions, one using all the bids, one using just the low bids, and one using all bids other than the low bid. Comparing the estimated coefficients in the three regressions, they reject the null hypothesis of no model misspecification. On this basis, they find that they can conclude that the bids from cartel firms are generated by different processes depending on whether the bids are low or not. They reject the null hypothesis of no phantom bidding and conclude that these five firms engaged in bid rigging. Baldwin, Marshall, and Richard (1997) examine the winning bids and characteristics for 108 oral auctions for timber sold by the Forest Service in the Pacific Northwest from 1975 to 1981. Baldwin, Marshall, and Richard described several salient features of the auctions. First, the Forest Service’s reserve prices in the auctions are very low. Second, old-growth timber is quite heterogeneous. Bidders invest substantial resources to

512   Ken Hendricks, R. Preston McAfee, and Michael A. Williams develop their own valuations of specific forest tracts. Third, second-growth or young timber is relatively homogenous and of lower quality than old-growth timber. Fourth, despite the fact that timber is extremely heavy per dollar of value, firms with mills within 100 miles of a given forest tract have approximately the same costs of cutting and transporting the logs. Fifth, the mills differ substantially in terms of their efficiency, which is measured by the quantity of merchantable board feet that can be extracted from a given log. Logging firms closely guard information regarding the efficiencies of their mills. Finally, prior to each oral auction, the Forest Service publicly releases the number of qualified bidders, and after each auction, the Forest Service makes public the quantity of timber purchased by each firm. In contrast, private sellers of timber do not make such information public. Given the importance of private information held by logging firms regarding the efficiency of their mills, Baldwin, Marshall, and Richard use an IPV model. In order to reduce the heterogeneous nature of the product, they restrict their data set to auctions for second-growth timber. Baldwin, Marshall, and Richard attempt to determine whether, after controlling for demand conditions, variations in winning (low) bids are better explained by collusion or by variations in the supply of timber. They estimate several models: (1) the noncooperative (i.e., noncollusive) model without considering supply effects; (2) the collusive model without supply effects; (3) the noncooperative model with supply effects, and two nesting models that include both collusion and supply effects. They conclude that their results strongly suggest that variation in the winning bids is best explained by collusion and not by noncooperative behavior or changes in supply conditions. They estimate that collusion reduces the auction revenues received by the Forest Service by approximately 7.9% across all the auctions. However, for a subset of 13 auctions that had particularly low winning bids, the loss in revenues to the Forest Service was approximately 52.9%. Athey, Levin, and Seira (2011) examine Forest Service auctions that occurred in Idaho-Montana and California between 1982 and 1990. (See also Athey and Levin, 2001, for analyses of Forest Service auctions.) Since the Forest Service used both open and sealed bids, Athey, Levin, and Seira are able to test how prices and revenues vary between the two auction formats. They show that sealed bid auctions attract more small bidders and tend to yield higher revenues to the Forest Service. Bidders may be loggers (i.e., small firms without manufacturing capacity) or mills (i.e., larger firms with manufacturing capacity). They estimate a structural IPV model that allows for entry on the part of bidders, and then they use that model to predict prices and revenues in Idaho-Montana and California with open or sealed bid auctions. They find that average open auction sales prices and average sealed bids in California are close to their predicted values, assuming firms behave competitively. This competitive result also holds for average sealed bids in Idaho-Montana. However, average open auction sales prices in Idaho-Montana are statistically different than both predicted competitive and collusive prices. Athey, Levin, and Seira (p. 251) conclude that “mildly cooperative behavior on the part of participating mills appears to provide a better match than either the competitive or fully collusive extremes.” To further test for the presence of collusion in

Auctions and Bid Rigging   513

Idaho-Montana auctions, the authors compare average sales prices in both open and sealed bid auctions as a function of the number of mills. When the number of mills is zero or one (recall that bidders can be loggers or mills), their predicted competitive prices are close to the actual prices. However, when the number of mills is two or more, their predicted competitive prices are substantially above the actual prices. They conclude this finding is consistent with collusive bidding. Given these results, Athey, Levin, and Seira investigate the welfare consequences of the Forest Service using either open or sealed bidding exclusively. They use two alternate specifications of mill bidding behavior: competitive behavior and collusive behavior 18% of the time, that is, the frequency of collusive behavior that best matches the observed open auction prices in Idaho-Montana. With the assumption of competitive behavior, they find that average revenues to the Forest Service are nearly the same with open or sealed bid auctions. However, with the more realistic assumption of collusive behavior (18% of the time), they find that the average revenues to the Forest Service would be substantially higher with sealed bid auctions than with open auctions. This result follows from their finding that sealed bid auctions encourage more participation from smaller bidders. Porter and Zona (1999) study procurement auctions conducted by Ohio school districts for milk. The State of Ohio charged 13 dairies with bid rigging in the period 1980– 1990. Porter and Zona prepared expert reports on behalf of the plaintiff in State of Ohio v. Louis Trauth Dairies, Inc. et al. Their article analyzes whether the behavior of three of the alleged conspirators located near Cincinnati is more consistent with competition or collusion. Using data collected in the litigation, they create a sample of bids for a control group of firms not accused of bid rigging and a sample of bids for the three firms accused of bid rigging. They first estimate a probit regression to calculate the probability that a control firm will submit a bid as a function of the distance between a relevant school district and that firm’s nearest milk processing plant. They find that the probability falls as distance increases. For example, the probability that a control firm with a plant very near a school district will bid in a milk procurement auction run by that district exceeds 50%, but falls essentially to zero when the distance exceeds 75 miles. They next estimate an OLS regression to calculate how a control firm’s bid changes as distance increases. They find that, all else equal, a control firm’s bid increases by approximately 0.5 cents per half pint when the distance increases by 50 miles. Given that the average price per half pint was approximately 13 cents in this period, this amounts to approximately a 4% increase in the control firms’ bid. Comparing these results for the group of control firms to the three alleged bid-rigging firms, Porter and Zona find that the alleged conspirators bid more frequently than the control group model predicts at longer distances, for example, in excess of 60 miles. They also find that two of the three alleged conspirators actually bid less the further away their plants are to the relevant school district. Porter and Zona interpret these results as showing an “inverted price umbrella,” with higher bids in school milk auctions close to the conspirators’ plants and lower bids at more distant locations. They conclude

514   Ken Hendricks, R. Preston McAfee, and Michael A. Williams the evidence is consistent with a local conspiracy in the Cincinnati area and inconsistent with competitive behavior. Finally, their empirical results show that the bid rigging increased the prices paid for milk by school districts by approximately 6.5%. Bajari and Ye (2003) use auction theory and empirical methods to distinguish competitive from collusive bidding behavior. They develop a general procurement auction model with asymmetric bidders, for example, bidders with different costs. They then derive a series of necessary and sufficient conditions for a distribution of bids to be consistent with bids generated by a model with competitive bidding. For example, one condition is that, given publicly available information that affects all firms’ bids, competitive bidding necessarily implies that the bids of any two firms must be independently distributed. Bajari and Ye use this condition to evaluate whether the bids made by construction firms in the Midwest are more consistent with competition or collusion. They regress firms’ bids on several explanatory variables, for example, distance from a firm’s location to the job site; the minimal distance of its rivals to the job site; the firm’s capacity; and the maximum free capacity of its rivals. They test whether the regression residuals for any pair of firms are correlated. If they reject the null hypothesis of no correlation, then the two firms’ bids are not independently distributed and so cannot be consistent with competition. They reject the null hypothesis for one pair of firms that bid against each other a large number of times. Bajari and Ye derive a second condition stating that competitive bidding necessarily implies that each firm’s estimated coefficient for any given explanatory variable in the bid regression should be the same. Their test results identify a specific pair of firms that fails this “exchangeability” test. Interestingly, one of the two firms in this pair is also one of the two firms in the pair that fails the “conditional independence” test. Bajari and Ye conclude that these three firms constitute a candidate cartel. However, they also emphasize that their test results could be explained by conduct other than collusion. They conclude that an economist finding such results should consult with industry experts to make sure the inconsistencies with the competitive bidding model do not result from ignorance of the industry cost structure. Rather than attempting to distinguish firms engaged in bid rigging from firms acting competitively, Pesendorfer (2000) analyzes the characteristics of two documented school milk cartels in Florida and Texas from 1980 to 1991. All the firms in his study pled guilty to price-fixing allegations brought by the US Department of Justice. Pesendorfer first shows that the market shares of firms in the Florida cartel fluctuated more than the market shares of firms in the Texas cartel. He examines several noncollusive explanations for this fact: (1) less potential competition in Texas; (2) some cartel members in Texas may bid on only a small subset of the school milk contracts; (3) a decline in the number of contracts per school district in Florida; and (4) an individual cartel member in Florida may have had costs that were correlated over time (e.g., high cost realizations in one year followed by low realizations the next year) that would cause its market share to fluctuate substantially over time. Pesendorfer finds that the empirical evidence does not support any of these noncollusive explanations. He concludes that the best explanation for the larger market share fluctuations in Florida is that cartel members in that

Auctions and Bid Rigging   515

state used side payments, rather than a market allocation agreement, to operate their cartel. That is, they operated a strong cartel. In contrast, the Texas cartel operated by dividing the state into different geographic regions and specifying which firms should win in which regions. That is, they operated a weak cartel. Pesendorfer also analyzes theoretically the efficiency of weak cartels. Recall that an auction mechanism is efficient if the mechanism allocates the object to the bidder with the highest valuation. Or alternatively, a cartel mechanism is efficient if it designates the member with the lowest cost to submit the lowest cartel bid. Pesendorfer shows that weak cartels cannot be efficient as long as the number of procurement contracts offered simultaneously by the buyer is finite. However, he shows that a weak cartel can use a mechanism (the “Ranking Mechanism”) in which each cartel member announces a ranking of the procurement contracts according to her costs. The member who ranks a contract highest (assigns the highest preference) will be the sole cartel bidder for that contract. If several cartel members rank a contract at the same position, the sole bidder can be determined by a coin toss. Pesendorfer shows the Ranking Mechanism is incentive compatible (i.e., cartel members will truthfully reveal their costs). Moreover, the mechanism converges to the optimal, efficient outcome as the number of contracts increases. Finally, Pesendorfer notes that 136 contracts are awarded every year in Texas, which leads him to conclude that the weak Texas cartel is likely almost efficient.6

20.3.2.  Experimental Studies of Bid Rigging in Auctions Isaac and Plott (1981) is the first experimental study of explicit collusion in an auction format. Their study addresses two primary questions. First, antitrust laws presume that opportunities to conspire, attempts to conspire, and successful conspiracies are closely linked. Is this presumption correct when the organizational costs of conspiracy are low? Second, which economic model of the price and output effects of price-fixing conspiracies best explains the observed behavior in their experiments? Isaac and Plott perform seven experiments using a double-auction mechanism in which buyers and sellers both make price offers. They first run three auctions in which neither buyers nor sellers have the opportunity to collude explicitly. They then run two auctions in which sellers are allowed to collude explicitly, without the knowledge of buyers. Finally, they run two auctions in which buyers are allowed to collude explicitly, without the knowledge of sellers. The explicit collusion was restricted in that participants could not reveal their private valuations or discuss side payments. Their basic finding is that, compared to the no-collusion auctions, both the seller- and buyer-collusion auctions are less competitive. Regarding their first question, Isaac and Plott reach several conclusions. First, in an environment with low costs of explicit collusion, buyers and sellers attempt to reach an agreement. Second, the attempts to reach an 6  See Abrantes-Metz et al. (2006) for a discussion of the use of econometric screening methods to detect collusion in the retail gasoline industry.

516   Ken Hendricks, R. Preston McAfee, and Michael A. Williams agreement result in a coordinated pricing strategy. Third, the pricing conspiracies have substantial, anticompetitive effects on equilibrium prices and quantities. With respect to price dynamics observed in the experiments, Isaac and Plott note that price changes in the three noncollusive experiments were always in the direction of the predicted equilibrium price. In contrast, prices in both the seller- and buyer-collusion auctions were more erratic, often moving away from the predicted cartel price. The average price change from period to period was also larger in the collusive experiments than in the noncollusive experiments. Finally, regarding their second question, Isaac and Plott conclude that the textbook cartel model best describes the behavior observed in the collusive experiments. However, the prices in these experiments generally did not converge to the predicted cartel price, as participants had difficulty maintaining agreed-upon prices. In follow-on research, Isaac, Ramey, and Williams (1984) show that the double-auction mechanism used by Isaac and Plott (1981) tends to make the enforcement of price-fixing agreements more difficult than posted offer prices. Isaac and Walker (1985) study sealed-bid auctions experimentally. They investigate two questions. First, in a series of FPSB auctions for identical items, are winning bids affected by revealing the losing bids from prior auctions? Second, does the ability of buyers to collude explicitly reduce the winning bids? They conducted a total of 30 experiments. In 10 of the experiments, explicit collusion was not allowed and buyers were not informed of losing bids in prior auctions. In 10 of the experiments, explicit collusion was not allowed and buyers were informed of losing bids in prior auctions. In the final 10 experiments, explicit collusion was allowed. In five of the collusive auctions buyers were not informed of losing bids in prior auctions, and in the other five collusive auctions buyers were informed of the losing bids. With respect to their two questions, Isaac and Walker find that the revelation of losing bids reduces the prices paid by buyers in subsequent auctions. They also find that the ability to collude explicitly reduces winning bids. However, the collusive prices were not affected by whether losing bids were revealed. Kwasnica (2000) evaluates experimentally the strategies used by auction participants to collude. Using a series of simultaneous, first-price auctions, he addresses three empirical questions: (1) Do bidders form cooperative agreements in simultaneous first-price auctions? (2) If so, what types of strategies do they utilize? (3) What effect do these strategies have on the outcome of the auction? The auction format consisted of 10 separate experiments. In each experiment a total of five objects were sold to five buyers in five simultaneous single-unit first-price auctions. This format was chosen to facilitate collusion. The auctions ran for 17–22 periods. In periods 1–5 participants were not allowed to communicate, but they were allowed to communicate thereafter. The communication was restricted in that participants could not reveal their private valuations or discuss side payments. In six of the 10 experiments the bidders had symmetric valuations, and in the remaining four experiments they had asymmetric valuations. In the latter, one bidder’s valuations were substantially higher than the symmetric valuations of the other four bidders. Finally, in six of the auctions, participants were informed of the winning bid and the identity of the winning bidder, while in the remaining four auctions participants were only informed of the winning bid.

Auctions and Bid Rigging   517

Consistent with Issac and Walker (1985), Kwasnica finds that bidders formed cooperative agreements in all 10 experiments. More interestingly, he finds that in seven of the 10 experiments participants used a bid rotation mechanism. They were able to obtain approximately 90% of the total surplus in the auctions. In other words, they achieved approximately 90% of the maximum possible efficiency, which would occur if the winning bidder had the highest valuation in each auction. Kwasnica concludes that the Ranking Mechanism proposed by Pesendorfer (2000) best describes the behavioral strategy chosen by the bidders. Finally, in three of the experiments, Kwasnica finds that bidders deviated from incentive compatible strategies. Instead, they selected a linear bid reduction strategy; for example, they agreed to bid 1% of their valuations. Participants used a linear bid reduction strategy only in experiments in which they had symmetric valuations and knew the identity of the winning bidder. Although a linear bid reduction strategy was not incentive compatible, under the experimental framework such a strategy led to higher profits than the incentive compatible, bid-rotation strategy. Kwasnica concludes that asymmetry of bidders’ valuations and less information leads bidders to revert to incentive compatible strategies.7

20.4.  Deterring Bid Rigging In theory, explicit collusion in auctions always can be prevented. The difficulty is that the information required to do so is essentially impossible to obtain. Che and Kim (2006, 2009) show how to convert any given auction mechanism into a collusion-proof auction, that is, an auction in which a bidding ring cannot earn profits. In such a collusion-proof auction, the seller’s expected total revenue will be the same as in a revenue-maximizing auction in the absence of collusion. However, as discussed by Marshall and Marx (2012), the Che and Kim collusion-proof mechanism is impractical, as it requires the following:



1. The mechanism requires all bidders, including losing bidders, to make payments to the auctioneer. However, the auctioneer may cheat by demanding higher-than-required payments from the bidders. This means that all the bidders must be able to verify that the auctioneer has implemented the mechanism correctly. 2. To establish the collusion-proof mechanism, the auctioneer must know the number of bidders and the set of types of the bidders, that is, their valuations, as well as the identities and types of two of the bidders in the ring. In Che and Kim (2006), bidders must agree to participate in the mechanism prior to the ring-formation game. In addition, the identities of the bidders that will have the

7  Other experimental studies of explicit collusion in auctions include Davis and Wilson (2002) and Phillips, Menkhaus, and Coatney (2003).

518   Ken Hendricks, R. Preston McAfee, and Michael A. Williams



opportunity to form a ring (only one ring is allowed to form) is fixed prior to the bidders agreeing or not to participate in the collusion-proof mechanism. As shown in Che and Kim (2009), under certain conditions, similar results hold when bidders first decide whether to collude and then decide whether to participate in the auctioneer’s collusion-proof mechanism. 3. The bidding ring must submit a bid for every ring member, despite possible incentives for the ring to suppress some members’ bids.

Given these practical drawbacks to the Che and Kim collusion-proof mechanism, economists have investigated other ways to deter explicit collusion in auctions. The literature offers several general suggestions (Klemperer, 2002; Marshall and Marx, 2009; Marshall and Marx, 2012). First, the use of FPSB auctions instead of English auctions should reduce collusion. Recall that bid-rigging agreements are stable in English auctions since no bidder has an incentive to cheat on the agreement since the cartel will bid up to the highest valuation of its members. In contrast, in FPSB auctions, the cartel must reduce its bid below the highest valuation of its members in order to earn positive expected profits. This reduction in the bid provides ring members the incentive to cheat on the agreement by outbidding the cartel. Hu, Offerman, and Onderstal (2011) provide experimental results that suggest using the Amsterdam second-price auction, which combines aspects of both FPSB and English auctions, to reduce collusion.8 Second, the auctioneer should limit the information provided to bidders regarding the number of bidders, their identities, their bids, including their losing bids if a similar auction will occur in the future. Finally, the auctioneer should not hold auctions at regular intervals and for relatively small items; instead the auctioneer should hold auctions at longer, irregular intervals and for relatively large items, for example, packages of separate items. This increases the costs of creating and maintaining a bidding ring by raising the gains to cheating.

20.5.  Conclusions and Open Questions In his classic article on collusion, Stigler (1964) observed that “the system of sealed bids, publicly opened with full identification of each bidder’s price and specifications, 8  The Amsterdam second-price auction has two phases (Goeree and Offerman, 2004). In phase 1, the auctioneer raises the price successively and bidders drop out when the price exceeds their individual valuation. This process continues until two bidders remain. The price at which the third-highest bidder drops out defines the endogenous reserve price or “bottom” price for the second phase. In phase 2, both remaining bidders independently submit sealed bids, which must be at least as high as the bottom price. The highest bidder wins and pays a price equal to the second-highest sealed bid. Both bidders in the phase two receive a premium, which equals a fraction 0 < α < 0.5 of the difference between the second-highest sealed bid and the bottom price.

Auctions and Bid Rigging   519

is the ideal instrument for the detection of price-cutting” and argued that collusion is always more effective when the market is more transparent. Open auctions are often more transparent than sealed bid auctions because bidders learn about the participation and bids of their rivals in real time. But, even in sealed bid auctions, the seller can choose how much information to make public before and after the auction. The less information the seller reports, the harder it is for the ring to achieve the gains from collusion. Indeed, the conclusion of the literature surveyed in this paper is that a seller should use a first-price, sealed bid auction and reveal as little information as possible. This conclusion, however, is based mainly on analyses of private value, single-item auctions. In common value auctions or multi-item auctions, the lack of transparency in the price-formation process can lead to inefficient allocations and lower revenues. For example, in common value auctions, bidders bid more aggressively in ascending bid auctions than in sealed bid auctions because they are able to learn about their rivals’ information, mitigating the effects of the winner’s curse. In auctions of spectrum licenses or oil and gas leases where spatial complementarities are important, bidders can more easily acquire their desired bundles and face less risk doing so in a multiround ascending bid auction than in a single round of sealed bids. Thus, in most situations, the choice of auction mechanism is likely to involve an important trade-off. On the one hand, more transparent auction mechanisms can yield greater efficiency and revenues when bidders behave competitively; on the other hand, they are also more vulnerable to collusion. A greater understanding of this trade-off is one of the big open questions in auction design. The theory literature has focused almost exclusively on the case of an all-inclusive ring.9 However, in practice, most rings involve subsets of bidders—not all bidders. Extending the theory to include partial rings is an important, although difficult, topic. Partial rings introduce asymmetries into the auction, and characterizing equilibrium bidding when the bidder types are drawn from asymmetric distributions is complicated. Finally, an important policy issue is to examine the rules under which the federal government in the United States operates as a buyer or seller. The Federal Acquisition Regulations, the Freedom of Information Act, and other regulations mandate transparency. The constraints that these rules impose on the choice of auction design may be quite costly.

9  Two exceptions are McAfee (1994) and McAfee and McMillan (1992), who show that, in first-price environments, the equilibrium ring in a cartel formation game generally does not include all of the bidders.

520   Ken Hendricks, R. Preston McAfee, and Michael A. Williams

References Abrantes-Metz, R., L. Froeb, J. Geweke, and C. Taylor. 2006. A Variance Screen for Collusion. International Journal of Industrial Organization 24: 467–86. Aoyagi, M. 2003. Bid Rotation and Collusion in Repeated Auctions. Journal of Economic Theory 112: 79–105. Aoyagi, M. 2007. Efficient Collusion in Repeated Auctions with Communication. Journal of Economic Theory 134: 61–92. Asker, J. 2010. A Study of the Internal Organization of a Bidding Cartel. American Economic Review 100: 724–62. Athey, S., and K. Bagwell. 2001. Optimal Collusion with Private Information. Rand Journal of Economics 32: 428–65. Athey, S., and J. Levin. 2001. Information and Competition in U.S. Forest Service Timber Auctions. Journal of Political Economy 109: 375–417. Athey, S., J. Levin, and E. Seira. 2011. Comparing Open and Sealed Bid Auctions: Evidence from Timber Auctions. Quarterly Journal of Economics 126: 207–57. Bajari, P., and L. Ye. 2003. Deciding between Competition and Collusion. Review of Economics and Statistics 85: 971–89. Baldwin, L., R. Marshall, and J-F. Richard. 1997. Bidder Collusion at Forest Service Timber Sales. Journal of Political Economy 105: 657–99. Blume, A., and P. Heidhues. 2006. Private Monitoring in Auctions. Journal of Economic Theory 131: 179–211. Boone J., R. Chen, J. Goeree, and A. Polydoro. 2009. Risky Procurement with an Insider Bidder. Experimental Economics 12: 417–36. Brown, A., C. Plott, and H. Sullivan. 2009. Collusion Facilitating and Collusion Breaking Power of Simultaneous Ascending and Descending Price Auctions. Economic Inquiry 47: 395–424. Brusco, S., and G. Lopomo. 2002. Collusion via Signalling in Simultaneous Ascending Bid Auctions with Heterogeneous Objects, with and without Complementarities. Review of Economic Studies 69: 407–36. Che, Y.-K., and J. Kim. 2006. Robustly Collusion-Proof Implementation. Econometrica 74: 1063–107. Che, Y.-K., and J. Kim. 2009. Optimal Collusion-Proof Auctions. Journal of Economic Theory 144: 565–603. Compte, O., A. Lambert-Mogiliansky, and T. Verdier. 2005. Corruption and Competition in Procurement Auctions. Rand Journal of Economics 36: 1–15. Cramton, P., and J. Schwartz. 2000. Collusive Bidding: Lessons from the FCC Spectrum Auctions. Journal of Regulatory Economics 17: 229–52. Dasgupta, P., and E. Maskin. 2000. Efficient Auctions. Quarterly Journal of Economics 115: 341–88. Davis, D., and B. Wilson. 2002. Collusion in Procurement Auctions:  An Experimental Examination. Economic Inquiry 40: 213–30. Fabra, N. 2003. Tacit Collusion in Repeated Auctions: Uniform versus Discriminatory. Journal of Industrial Economics 51: 271–93. Goeree, J., and T. Offerman. 2003. Competitive Bidding in Auctions with Private and Common Values. Economic Journal 113: 598–613. Goeree, J., and T. Offerman. 2004. The Amsterdam Auction. Econometrica 72: 281–94.

Auctions and Bid Rigging   521

Graham, D., and R. Marshall. 1987. Collusive Bidder Behavior at Single-Object Second-Price and English Auctions. Journal of Political Economy 95: 1217–39. Graham, D., R. Marshall, and J.-F. Richard. 1990. Differential Payments within a Bidder Coalition and the Shapley Value. American Economic Review 80: 493–510. Haan, M., L. Schoonbeek, and B. Winkel. 2009. Experimental Results on Collusion: The Role of Information and Communication. In J. Hinloopen and H-T Normann, Experiments and Competition Policy, 9–33. Cambridge: Cambridge University Press. Hendricks, K., and R. Porter. 2007. An Empirical Perspective on Auctions. In M. Armstrong and R. Porter, Handbook of Industrial Organization, vol. 3: 2073–143. Amsterdam: Elsevier. Hendricks, K., R. Porter, and G. Tan. 2008. Bidding Rings and the Winner’s Curse. Rand Journal of Economics 39: 1018–41. Hinloopen, J., and S. Onderstal. 2010. Collusion and the Choice of Auction: An Experimental Study. Tinbergen Institute Discussion Paper. Hu, A., T. Offerman, and S. Onderstal. 2011. Fighting Collusion in Auctions: An Experimental Investigation. International Journal of Industrial Organization 29: 84–96. Isaac, R., and C. Plott. 1981. The Opportunity for Conspiracy in Restraint of Trade. Journal of Economic Behavior and Organization 2: 1–30. Isaac, R., V. Ramey, and A. Williams. 1984. The Effects of Market Organization on Conspiracies in Restraint of Trade. Journal of Economic Behavior and Organization 5: 191–222. Isaac, R., and J. Walker. 1985. Information and Conspiracy in Sealed Bid Auctions. Journal of Economic Behavior and Organization 6: 139–59. Ishii, R. 2009. Favor Exchange in Collusion: Empirical Study of Repeated Procurement Auctions in Japan. International Journal of Industrial Organization 27: 137–44. Klemperer, P. 1999. Auction Theory: A Guide to the Literature. Journal of Economic Surveys 13: 227–86. Klemperer, P. 2002. What Really Matters in Auction Design. Journal of Economic Perspectives 16: 169–89. Klemperer, P. 2008. Competition Policy in Auctions and “Bidding Markets.” In P. Buccirossi, Handbook of Antitrust Economics, 583–624. Cambridge, MA: MIT Press. Kwasnica, A. 2000. The Choice of Cooperative Strategies in Sealed Bid Auctions. Journal of Economic Behavior and Organization 42: 323–46. Kwasnica, A., and K. Sherstyuk. 2007. Collusion and Equilibrium Selection in Auctions. Economic Journal 117: 120–45. Li, J., and C. Plott. 2008. Tacit Collusion in Auctions and Conditions for Its Facilitation and Prevention: Equilibrium Selection in Laboratory Experimental Markets. Economic Inquiry 47: 425–48. Lopomo, G., R. Marshall, and L. Marx. 2005. Inefficiency of Collusion at English Auctions. Contributions in Theoretical Economics 5, article 4. Lopomo, G., L. Marx, and P. Sun. 2011. Bidder Collusion at First-Price Auctions. Review of Economic Design 15: 177–211. Mailath, G., and P. Zemsky. 1991. Collusion in Second Price Auctions with Heterogeneous Bidders. Games and Economic Behavior 3: 467–86. Marshall, R., and L. Marx. 2007. Bidder Collusion. Journal of Economic Theory 133: 374–402. Marshall, R., and L. Marx. 2009. The Vulnerability of Auctions to Bidder Collusion. Quarterly Journal of Economics 124: 883–910. Marshall, R., and L. Marx. 2012. The Economics of Collusion:  Cartels and Bidding Rings. Cambridge, MA: MIT Press.

522   Ken Hendricks, R. Preston McAfee, and Michael A. Williams McAfee, R. P., and J. McMillan. 1987. Auctions and Bidding. Journal of Economic Literature 25: 699–738. McAfee, R. P., and J. McMillan. 1992. Bidding Rings. American Economic Review 82: 579–99. McAfee, R.P., D. Vincent, M. Williams, and M. Havens. 1993. Collusive Bidding in Hostile Takeovers. Journal of Economics and Management Strategy 2: 449–82. Milgrom, P., and R. Weber. 1982. A Theory of Auctions and Competitive Bidding. Econometrica 50: 1089–122. Myerson, R. 1981. Optimal Auction Design. Mathematics of Operations Research 6: 58–73 Pesendorfer, M. 2000. A Study of Collusion in First-Price Auctions. Review of Economic Studies 67: 381–411. Phillips, O., and D. Menkhaus. 2009. Maintaining Tacit Collusion in Repeated Ascending Auctions. Journal of Law and Economics 52: 91–109. Phillips, O., D. Menkhaus, and K. Coatney. 2003. Collusive Practices in Repeated English Auctions: Experimental Evidence on Bidding Rings. American Economic Review 93: 965–79. Porter, R. 1983. A Study of Cartel Stability: The Joint Executive Committee 1880–1886. Bell Journal of Economics 14: 301–14. Porter, R. 2005. Detecting Collusion. Review of Industrial Organization 26: 147–67. Porter, R., and J. Zona. 1993. Detection of Bid Rigging in Procurement Auctions. Journal of Political Economy 101: 518–53. Porter, R., and J. Zona. 1999. Ohio School Milk Markets: An Analysis of Bidding. Rand Journal of Economics 30: 263–88. Potters, J. 2009. Transparency about Past, Present, and Future Conduct: Experimental Evidence on the Impact on Competitiveness. In J. Hinloopen and H.-T. Normann, Experiments and Competition Policy, 81–104. Cambridge: Cambridge University Press. Riley, J., and W. Samuelson. 1981. Optimal Auctions. American Economic Review 71: 381–39. Robinson, M. 1985. Collusion and the Choice of Auction. Rand Journal of Economics 16: 141–45. Sherstyuk, K. 1999. Collusion without Conspiracy: An Experimental Study of One-Sided Auctions. Experimental Economics 2: 59–75. Sherstyuk, K. 2002. Collusion in Private Value Ascending Price Auctions. Journal of Economic Behavior and Organization 48: 177–95. Sherstyuk, K., and J. Dulatre. 2008. Market Performance and Collusion in Sequential and Simultaneous Multi-object Auctions:  Evidence from an Ascending Auctions Experiment. International Journal of Industrial Organization 26: 557–72. Skrzypacz, A., and H. Hopenhayn. 2004. Tacit Collusion in Repeated Auctions. Journal of Economic Theory 114: 153–69. Stigler, G. 1964. A Theory of Oligopoly. Journal of Political Economy 72: 44–61. Vickrey, W. 1961. Counterspeculation, Auctions, and Competitive Sealed Tenders. Journal of Finance 16: 8–37. Waehrer, K. 1999. Asymmetric Private Values Auctions with Application to Joint Bidding and Mergers. International Journal of Industrial Organization 17: 437–52. Wilson, R. 1992. Strategic Analysis of Auctions. In R. Aumann and S. Hart, Handbook of Game Theory, vol. 1: 227–79. Amsterdam: Elsevier.

 CHAPTER 21

S C R E E N I N G F O R C OL LU SION A S A PROBLEM OF INFERENCE MICHAEL J. D OANE, LUKE M. FROEB, DAVID S. SIBLEY, AND BRIJESH P. PINTO

21.1. Introduction In 1993, in the course of doing other research, Bill Christie and Paul Schultz noticed that market makers were avoiding odd-eighth quotes on some of the most heavily traded NASDAQ stocks. They tested and rejected a number of cost-based explanations for the practice, and concluded that it was the result of collusion. The obvious benefit of oddeighth avoidance is that it raises the bid-ask spread, a measure of the “price” that market makers charge for selling and buying stocks, to at least $0.25, over the regulatory minimum of $0.125. At the time, the cost of trading stocks on NASDAQ was about twice the cost of trading stocks on the NYSE, which used a more competitive auction-like mechanism (Morgenson, 1993). When the Los Angeles Times reported on the Christie and Schultz (1994) research, the NASDAQ market makers suddenly (within a day) stopped avoiding odd-eighth quotes, and the bid-ask spread on many stocks was cut in half. Professors Christie, Schultz, and Harris (1994) then published a companion piece titled “Why Did NASDAQ Market Makers Stop Avoiding Odd Eighth Quotes?” which reached the same conclusion as the original article. Together, the two papers lead to a raft of private suits and investigations by the US Department of Justice and the Securities and Exchange Commission. Although not originally designed as such, their research is arguably one of the most successful collusive screens ever, and settlements to the antitrust suits have changed the way that NASDAQ market makers trade stocks.

524   Michael J. Doane, Luke M. Froeb, David S. Sibley, and Brijesh P. Pinto On October 12, 2012, US representative Edward Markey wrote a letter to the chairman of the US Federal Trade Commission (FTC) asking him to investigate whether oil companies and refiners were “manipulating” gasoline prices in New England (Markey, 2012; FTC, 2012). His request was triggered by the observation that while oil prices had been falling, gasoline prices had not fallen by as much. Partially in response to these kinds of requests, a decade ago, the FTC began regularly screening gasoline prices for collusion. Although the screening has yet to uncover any conspiracies, screening seems to be catching on, as countries like Brazil and Mexico have started similar programs (Levenstein and Suslow, in this volume; Abrantes-Metz and Froeb, 2008; Abrantes-Metz and Bajari, 2010; Harrington, 2008). The question motivating this chapter is “How well do screens for collusion work?” It is hard to answer by looking at successful follow-on prosecutions. If a screen finds no collusion, it could be that the screen is not working or that there is no collusion to be found. If a screen does find collusion, one has to look carefully at the resolution of the legal process to determine exactly what happened. Even the NASDAQ market makers never admitted colluding, and there is some uncertainty as to how the conspiracy actually worked. Despite evidence of explicit communication, there were weak incentives for market makers to offer better quotes (even without colluding) because better quotes did not necessarily capture more order flow (Morgenson, 1993). Perhaps because of this uncertainty, the Department of Justice chose to file a civil antitrust suit, which has a lower burden of proof, rather than a criminal one. So rather than looking at follow-on prosecutions, we address the issue by modeling screening as a problem of inference whose essential elements include a competitive null hypothesis, a collusive alternative, and an empirical indicator to differentiate between them. For example, Christie and Schultz (1994) were implicitly using a screen based on a null hypothesis that costs were determining prices, a collusive alternative that odd-eighth avoidance was increasing the “price” of trading, and a variety of empirical indicators to rule out cost-based explanations and to show that odd-eighth avoidance was increasing price. Likewise, Representative Markey was implicitly using a screen based on the null hypothesis that competitive behavior results in a higher pass-through rate from wholesale cost to retail price, the collusive alternative that cartels have a lower pass-through rate, and estimated pass-through rates from oil prices to retail gasoline prices as an empirical indicator to differentiate between the two. Broken down into these elements, it becomes easy to critique his screen. In this case, Congressman Markey’s hypothesis that collusion has lowered pass-through rates is at odds with simple models of price-setting behavior (Yde and Vita, 1996; Froeb, Tschantz, and Werden, 2005). If these models characterize behavior in the gasoline industry, then the screen is likely to be a poor indicator of collusion. In this chapter, we use the taxonomy to better understand the conditions under which screens are likely to succeed or fail. The first is that the empirical indicator may be too “weak” to differentiate between H0 and H1 (e.g., Kovacic et al., 2011). The second is that the null hypothesis is not indicative of competition or that the alternative is not indicative of collusion. For example, Representative Markey’s implicit

Screening for Collusion as a Problem of Inference   525

screen may work in some theoretical models, or may characterize some actual cases of collusion. But absent evidence that the collusion or competition takes the assumed forms, the screen’s conclusions will be noisy indicators of collusion, at best. We call this the problem of “model fit.” The third reason that a screen may fail is that the world is neither H0 nor H1. In this case, differentiating between H0 and H1 is not useful if the world is really H2. This third criticism is similar to the second and stems from the same issue of model fit. Before using a screen to differentiate between collusion (H0) and competition (H1), it is important to try to rule out the existence of H2. This idea is illustrated by the NASDAQ screen. Christie and Schultz (1994) and Christie, Harris, and Schultz (1994) rule out several different cost-based explanations for odd-eighth avoidance, like adverse selection (informed traders are more likely to trade at favorable quotes) and the cost of holding inventory until an order arrives. Essentially, they reject a variety of hypotheses in favor of H1.1 In what follows, we formally draw the analogy between screening and Bayesian hypothesis testing. Next, we review and discuss the theoretical and empirical literature, breaking down each screen into its essential elements. We conclude by offering advice to academics designing screens, and practitioners using them: if you know what form competition (H0) and collusion (H1) are likely to take, and have a strong empirical indicator to distinguish between the two, then a screen is likely to be successful. If, however, you do not know the forms that competition and collusion are likely to take, or cannot rule out the likelihood of alternate behavior (H2), then the screen is more likely to be plagued by type I or type II errors. We choose to ignore the problem of strategic behavior on the part of the colluders to evade a known that can be thought of as a type of deterrence. We do this not because it is not an important problem (every successful screen has this shortcoming) but rather because screening is so young. We have to know whether screening is an efficient way to detect conspiracies before we address the more difficult question of whether screens can also deter them.

21.2.  Screening as a Problem of Inference 21.2.1.  Bayesian Hypothesis Testing To study screens, we use the metaphor of statistical inference by specifying the competing hypotheses, competition (H0) versus collusion (H1), and a variable X that can

1  Alternatively, H0 can be thought of as a composite hypothesis. For example, if H2 is another competitive (noncollusive) hypothesis, then the researcher tries to differentiate between H0 or H2 vs. H1.

526   Michael J. Doane, Luke M. Froeb, David S. Sibley, and Brijesh P. Pinto differentiate between them. Each hypothesis is related to the variable by a distribution function,

H 0 : X ~ f0



H1 : X ~ f1



and the problem of inference is how to determine which hypothesis generated the observable data x = {x1 , x2 , …, xn } . Each data point xi could be an individual observation, for example, on price or on margin, or a vector of what Kovacic and coauthors (2011) call “indicator variables” or “plus factors.” Although much of our analysis could be done using the metaphor of classical hypothesis testing, we choose to work within the Bayesian framework because it can accommodate both “structural” screening, based on prior beliefs, and “behavioral” screening, based on data (Harrington, 2006). Our goal is to compute the posterior probability of collusion given the data



P ( H1 | x ) =

P ( x | H1 ) P ( H1 )

P ( x | H 0 ) P ( H 0 ) + P ( x | H1 ) P ( H1 )

.

To do this, we have to specify prior beliefs, P (H 0 ) and P (H1 ), which are informed by structural indicators such as evidence of past collusion or product homogeneity (e.g., Stigler, 1964) or whether the government is the potential conspiracy victim (e.g., Froeb, Koyak, and Werden 1993). A  “neutral” or “uninformative” prior would have a prior belief characterized by P (H 0 ) = P (H1 ) = 0.5 . These assumptions imply a prior belief about the data generating process,

f prior ( x ) = γ f 0 ( x ) + (1 − γ ) f1 ( x ) ,

where γ = P (H 0 ) represents our prior assessment that the industry is competitive. After seeing the data, we construct the likelihoods of the competing hypotheses, n



L0 ( x ) = ∏ f 0 (xi ) i =1



n



L1 ( x ) = ∏ f1 ( xi ) . i =1



Screening for Collusion as a Problem of Inference   527

They tell us how likely it is that each hypothesis generated the data. As written, the likelihood equations imply that each piece of data, or each indicator variable, is independent and drawn from the same continuous distribution (iid), but this is not necessary. The likelihoods could be constructed from a vector of correlated factors, or from non-iid or discrete data. We update our prior beliefs using the likelihood ratio or “Bayes Factor,”



 γ   L0 ( x )  θ = , 1 − θ  1 − γ   L1 ( x ) 



which says that the posterior odds equals the prior odds times the likelihood ratio. The posterior belief about the data-generating process is

f posterior ( x ) = θf 0 ( x ) + (1 − θ) f1 ( x ) ,



where θ = P (H 0 | x ) and (1 − θ) = P (H1 | x ) is the posterior probability that H0 and H1 are true. This represents our beliefs after updating them with data. The final element to the screen is a way to turn the posterior belief into action. Here we could imagine that a decision-maker (presumably a competition agency) could decide to investigate an industry further if

P ( H1 | x ) > α ,



which says that the probability of collusion is higher than some threshold. Some have characterized α = .5 as “preponderance of evidence” standard and α = .95 as a “beyond all reasonable doubt” standard. Alternatively, the decision-maker could optimally choose α to minimize an expected loss function based on the costs of enforcement errors (CI , CII ) as in Cooper and coauthors (2005). In this case, investigate further if expected cost of type I error (failure to investigate when collusion is present) is greater than the expected cost of type II error (investigation when no collusion is present),

P ( H 0 | x ) CI > P ( H1 | x ) CII ,

or P (H 0 | x )

P ( H1 | x )

>

CII . CI



Easterbrook (1984) argues that the market will more readily correct false acquittals than false prosecutions, or that CII > CI , which would imply a higher optimal threshold for investigation.

528   Michael J. Doane, Luke M. Froeb, David S. Sibley, and Brijesh P. Pinto Even though few screens are characterized as a problem of inference, implicitly this is what each is trying to do. Formally classifying them in this way allows us to understand their elements, and more importantly to understand the conditions under which one would expect them to succeed or fail. The first reason for failure is that X is what Kovacic and coauthors (2011) call a “weak” indicator. In this case, the likelihood ratio will be near 1, and the posterior odds will be close to the prior odds. In other words, the likelihood does not help us to distinguish between the two hypotheses. In contrast, a “strong” indicator variable will have a likelihood ratio away from 1, which would shift the posterior odds towards one of the two competing hypotheses. The second reason for failure is that the null hypothesis is not indicative of competition or that the alternative is not indicative of collusion. So even if the data can tell us whether H0 or H1 is more likely to have generated the data, this doesn’t tell us very much about whether collusion is present. To see how this leads to errors, imagine that H0 could be consistent with either competition or collusion, but that H1 is consistent only with collusion. If the likelihood favors the competitive hypothesis such that P (H1 | x ) < α, no further investigation is taken. In this case, the screen is likely to miss collusive behavior. Conversely, imagine that H0 is consistent only with competition, but H1 is consistent with either competition or collusion. If the likelihood favors the collusive hypothesis such that P (H1 | x ) > α, further investigation is taken. In this case, the screen will allocate prosecutorial resources to an investigation with a low probability of finding collusion. When both H0 and H1 are consistent with both competition and collusion, the probability that the screen will commit both types of errors is high. This mistake occurs when the theoretical models or assumptions on which the screens are based do not fit the significant features of competition in the industry being studied. If the theoretical models do not adequately characterize behavior in the industry to which the screen is being applied, or if the collusive or competitive outcomes depend on factors that cannot be observed or verified, then the screen is likely to work poorly. We refer to this problem as one of model “fit.” Some of the screens are based on instances of past collusion. For example, as discussed more fully below, Abrantes-Metz and coauthors (2006) note that when a frozen fish conspiracy collapsed, the price went down by over 16%, but the standard deviation of price (over time) went up by over 200%. Using this empirical analogy, they construct a screen based on the coefficient of variation of price, and apply it to retail gasoline. However, unless collusion in the gasoline industry takes a form similar to collusion in frozen fish sold by auction to the government, then the screen is unlikely to work very well. This is the same problem of fit, but here the “fit” is to a past instance of collusion, not a theoretical model.

21.2.2.  Screening versus Testing One obvious question is what happens if neither H0 nor H1 generated the data? Our Bayesian framework can be easily modified to admit a third alternative, which we call H2:

Screening for Collusion as a Problem of Inference   529



H 0 : X ~ f0



H1 : X ~ f1



H 2 : X ~ f2 .

If H2 is known, this becomes what is known as a problem of “model selection,” where we choose between three alternatives. As above, we begin with a prior belief about the data-generating process,

f prior ( x ) = γ 0 f 0 ( x ) + γ 1 f1 ( x ) + (1 − γ 0 − γ 1 ) f 2 ( x ) ,



which we update with the likelihood in an analogous way. If we don’t know about the existence of H2, the results of the previous section still hold; that is, we can still compute the relative odds of H0 versus H1, but the computation of the posterior probabilities will be wrong. In this case, we will mistakenly compute



P ( H1 | x ) =

P ( x | H1 ) P ( H1 )

P ( x | H 0 ) P ( H 0 ) + P ( x | H1 ) P ( H1 )



instead of



P ( H1 | x ) =

P ( x | H1 ) P ( H1 )

P ( x | H 0 ) P ( H 0 ) + P ( x | H1 ) P ( H1 ) + P ( x | H 2 ) P ( H 2 )

.

We find that this is a useful way to think about the difference between what we call testing and screening. If you are “testing” for collusion, you know that the world is either H0 or H1, but if you are “screening” for collusion, you recognize that there may be other alternatives, like H2. In other words, when “testing,” all you have to worry about is “sampling uncertainty.” With enough data, or with good enough indicator variables, you can estimate the posterior probability with precision and take appropriate action. In contrast, when you are unsure about the possibility of other alternatives (H2), the posterior probability will reflect “model uncertainty” in addition to sampling uncertainty. This is closely related to the problem of model fit discussed above. If you know that the world is either H0 or H1 (and that H0 implies competition and that H1 implies collusion), with a strong enough indicator variable, you can distinguish collusion from competition. However, if

530   Michael J. Doane, Luke M. Froeb, David S. Sibley, and Brijesh P. Pinto the world can also be H2, then distinguishing between H0 and H1 is not likely to tell you very much about collusion or competition.

21.3.  Theoretical Screens The purpose of this section is to review some theoretical models of cartels and collusive behavior that have been used to construct screens. By reviewing the literature, we highlight the significant features of competition that each model is designed to capture, in addition to the maintained assumptions on which the models are built. We elucidate these features and assumptions for the purpose of highlighting the problem of model fit. Unless the model on which the screen is based captures the significant features of competition in the industry being screened, the screen is likely to perform poorly. The literature is summarized in table 21.1. Green and Porter (1984) show that periodic episodes of sharp drops in colluding firms’ prices and profits may be indicative of cartel self-enforcement. Departing from Stigler’s (1964) notion that these episodes reflect the instability of cartels, Green and Porter show instead that cartels may use price wars as a self-policing device. In their model, demand is uncertain, and colluding firms engage in collusive behavior or Cournot behavior depending on whether or not the market price is above an agreed-upon “trigger” price. The alternative hypothesis of collusion is that output levels follow a switching process triggered by drops in the market price. Green and Porter suggest that the American rail freight industry in the 1880s is an example of an industry that satisfies their assumptions on industry structure, and exhibits the kind of collusion that they model. Rotemberg and Saloner (1986) analyze an infinite-horizon dynamic model with observable demand shocks in which firms compete either in Cournot or in Bertrand fashion. In their model, a colluding firm’s incentive to cheat by undercutting the collusive price is greater when demand is high. They show that in periods of high demand, colluding firms are likely to behave more competitively. During these periods, the marginal benefit gained by a colluding firm from “cheating” on the cartel is greater since the share of market captured is larger. In response, the cartel reduces the collusive price to reduce the incentive to cheat. This behavior generates countercyclical price and margin movements; that is, the collusive price and margin is lower when demand is high and higher when demand is low (H1). The competitive hypothesis (H0) of competition is that price and margins are procyclical with respect to observable demand shocks. Athey, Bagwell, and Sanchirico (2004) develop a theory linking collusion and price rigidity. They show that if firms are sufficiently patient and the distribution of firms’ costs is log-concave, then optimal symmetric collusion in equilibrium is characterized both by price rigidity and by the absence of price wars. Athey, Bagwell, and Sanchirico analyze the equilibrium of an infinite-horizon, repeated game model in which the stage game is one of Bertrand competition among symmetric firms. Each firm possesses

Screening for Collusion as a Problem of Inference   531

private information about its marginal cost of production, and cost levels are independently and identically distributed across firms. As discussed below, several empirical studies cite this paper as suggesting that price rigidity (H1) can serve as a screen of collusive behavior in an industry that satisfies their structural assumptions. Under the null hypothesis of competition (H0), price varies more closely with cost. Marshall, Marx, and Raiff (2008) analyze price announcements in the vitamins industry, with a view to detecting collusion in the industry after 1985 They model public price announcements in an industry with homogenous products and capacity constraints as a multiperiod game and show that in the absence of explicit collusive behavior (H0) (1)  there exist equilibria in which none of the firms makes price announcements; (2)  there exist equilibria in which the larger firm leads a joint announcement, or makes a singleton announcement—the larger firm makes an announcement in an earlier period, while the smaller firm does not make an announcement but sets its price in a later period; and (3) no equilibria exist in which the smaller firm makes a singleton announcement or leads a joint announcement. Marshall, Marx, and Raiff observe frequent joint price announcements in the collusive post-1985 period. They also find that, relative to the pre-1985 period, announcements were made well in advance of effective dates in the collusive period. Furthermore, they observe that the timing of price announcements in the collusive period is consistent with regularly scheduled cartel meetings. For these reasons, the authors conclude that the empirical implications of their model are largely consistent with the absence of explicit collusion in the vitamins industry prior to 1985, but consistent with the presence of explicit collusion after 1985. Thus, according to Marshall, Marx, and Raiff the model suggests collusive behavior in the vitamins industry after 1985. We describe their empirical analysis in section 21.4. There is also a literature on optimal cartel pricing when detection is a possibility. Harrington and Chen (2006) study a dynamic model of oligopoly with stochastic costs, in which a firm that forms a cartel is detected with some probability. They assume that firms have a common constant marginal cost of production and, essentially, that a firm’s cost in a given period equals the sum of its cost in the previous period and a normally distributed cost shock. Further, “buyers have the maintained hypothesis that price is an affine function of cost and cost changes are normally distributed but do not know the coefficients to the pricing function or the moments of the cost distribution” (p. 1191). Their analysis, while admittedly exploratory and relying on strong maintained assumptions, is an attempt to enrich our understanding of cartel pricing behavior. Simulating their model produces cartel price paths with a transition phase followed by a stationary phase in which collusion (H1) reduces price variation (X). The simulated cartel price paths resemble the price paths of real cartels, and they take this as an encouraging sign that their model may help develop empirical tests for cartel detection. We take away two things from this brief review of theory. First is that these models of collusive behavior are all somewhat stylized, based on a number of assumptions that may not comport well with reality. What is important for our purposes is whether these models can capture the significant features of competition in the industries to which

Table 21.1  Collusion Theories Study Green & Porter (1984)

Model issue Maintain cartel outputs given imperfect information about demand.

Assumptions

H1: Collusion

x: Indicator variable

(1) Firms choose outputs; (2) market prices depend on market output and the realization of an unobserved demand shock; (3) firms choose outputs before the realization of the demand shock; infinite time horizon.

Firms each choose a set cartel outputs unless market price falls below an agreed “trigger” level; at that point each firm reverts to its Cournot output given its expectations about the demand shock.

Output level: Output follows a switching process triggered by a low market price.

Rotemberg & Saloner (1986)

Cartel pricing with observable demand shocks that are iid over time.

(1) Firms can be either Cournot or Bertrand; (2) market demand curve depends on prices (outputs) and the realization of an observable demand shock.

Cartel prices are chosen to make it unprofitable to defect from them, given the known state of demand and expectations about the future.

Price level: Cartel prices are countercyclical:  when current demand is high relative to expected future demand, the temptation to cheat is greatest, meaning that cartel prices must fall to deter cheating; and vice versa when current demand is low.

Harrington & Chen (2006)

Optimal cartel pricing when detection is a possibility

(1) Firms have identical marginal cost, (2) cost is a specific stochastic process, (3) buyers believe price is a random walk, (4) buyers do not know collusive pricing function.

Cartel price paths exhibit a transition phase and a stationary phase. During transition phase, price is independent of cost. During stationary phase, price responds to cost shocks.

Price level: cartel price path exhibits a transition and a stationary phase.

Marshall, Marx, & Raiff (2008)

Inferring collusion from the absence or existence of preannounced price changes

(1) Predictions generated by a benchmark model of price announcements; (2) the price announcement model assumes fixed capacities and homogeneous products; (3) the subgame perfect equilibrium of this game involves a price leader (the largest firm) announcing price change before smaller firms; (4) under collusion, firms preannounce price increases to convince strategic buyers that the price increases reflect general cost increases, and that strategic buying will not be profitable; (5) absent collusion, individual sellers do not have this concern, and do not preannounce price changes.

(1) In cases of known collusion, did the pattern of price announcements differ from in the same industry during noncollusive period? (2) advance announcements of price change under collusion should be public, rather than private; (3) price announcements should come just before existing contracts are renewed, so as to convince a strategic buyer that there is no reason to switch sellers; they should not reflect cost or demand changes, but rather the length of time since the most recent expiration date of a contract between a cartel seller and its customer.

(1) Likelihood of price announcements: The likelihood of price announcements does not depend on costs of demand, but on the length of time since the last set of price announcements. (2) Frequency of price announcements: The frequency of price announcements increases greatly relative to a “clean” period.

Athey, Bagwell, & Sanchirico (2004)

Can a cartel subject to costs shocks display rigid pricing, in accordance with much collusion folklore?

(1) Bertrand stage game, in which the low-priced firm(s) get all the business; (2) each firm gets an iid. cost shock each period, which is private information; (3) each firm has identical beliefs about future cost shocks; (4) each firm reports its current period cost value to the cartel, along with the price implied by that cost; (5) each firm can credibly precommit to this price mechanism; (6) transfers are possible between cartel members; (7) each firm maximizes expected present value over infinite time horizon; (8) sequential equilibrium concept for cartel.

Result is that cartel pricing will display rigidity if cartel firms are patient enough and if the distribution function of future cost shocks is log-concave. Would need to test for these, which would be difficult. If one could verify these conditions, then H0 is that the stage game is Bertrand and H1 is that you would never see sticky prices. If you did, this rules out the stage game, implying collusion.

Price level: Cartel prices will display rigidity.

534   Michael J. Doane, Luke M. Froeb, David S. Sibley, and Brijesh P. Pinto screens based on them are applied. This is another way of articulating the problem of “model fit.” Second, there are a variety of ways to collude. The models reviewed illustrate a variety of ways that firms can fix prices, but firms can also rig bids, allocate customers, agree to shut down capacity, refrain from advertising or price discriminating, or even collude on organizational form (Shor and Chen, 2009) as ways to reduce competition. When using a screen based on one of these models, it is important to first figure out how competition and collusion work and whether the screen would help you distinguish one from the other.

21.4.  Empirical Screens The empirical literature on collusive screens may be categorized broadly into the following categories: (1) screens based on the first and second moments of the price distribution;2 (2) screens based on structural modeling; and (3) screens based on various cartel markers derived from models of firms’ pricing behavior. Table 21.1 presents a sample of recent or significant papers that fall into these categories. We use our earlier distinction between “screens” and “tests” for collusion in that screens are designed to find unknown episodes of collusion, while tests are designed to test whether indicator variables (X) can correctly identify known or suspected episodes of collusion. Sometimes the distinction is not so clear or useful, but it is designed to capture the idea that if you know the form of the conspiracy (H1) and how it differs from competition (H0), then it should be easier to find a variable that correctly identifies the conspiracy because you are more confident that your empirical model “fits” the form of the collusion. Screens, on the other hand, are more exploratory in nature because the form of competition or collusion is not known, or because there may exist alternate hypotheses (H2) not contemplated by the screen.

21.4.1.  Screens Based on First and Second Moments of Price Distribution Empirical screens for collusive conduct have generally involved the examination of pricing patterns that might be indicative of collusion. Along these lines, the US Department of Justice has identified five pricing patterns as possible indicators of explicit collusion: (1) prices remain identical for long periods of time; (2) prices start to become identical after being different previously; (3) price increases that appear not to be explained 2  A variation of this theme include studies that have examined changes in the first and second moments of market shares and price-cost margins. See, for example, Genosove and Mullin (2001).

Screening for Collusion as a Problem of Inference   535

by exogenous cost increases; (4) elimination of price discounts in a market where discounts historically were given, and (5)  local customers pay higher prices than more distant customers (DOJ, n.d.). Empirical screens found in the academic literature have tended to follow this approach with much research focused on examining changes in the first and second moments of the price distribution. Below we discuss papers that are representative of this body of research. Froeb, Koyak, and Werden (1993) propose a method for estimating the price effects of bid-rigging and price-fixing conspiracies based on the level of prices that controls for cost differences and other factors that may be responsible for observed price differences between collusive and competitive periods.3 They apply their estimator to a known conspiracy involving bid rigging in the sale of frozen seafood to the Defense Personnel and Support Center (“DPSC”) in Philadelphia.4 Based on an inspection of quantity-weighted bid prices for frozen perch filets in procurements from July 10, 1984, through September 27, 1989, the authors divide their data into three distinct periods: a preconspiracy period characterized by relatively constant prices despite big seasonal swings in fresh fish prices, September 19, 1986 to July 20, 1988; a “transition” period during which the conspiracy collapsed characterized by a rapid decline in the price of frozen perch while cost rose, mid-July 1988 to mid-September 1988; and a postconspiracy period, September 28, 1988 to September 27, 1989. The null hypothesis (competition) is that the price of frozen perch was not statistically lower (on average) in the postconspiracy period than the preconspiracy period,5 controlling for cost differences. To estimate “but-for” conspiracy prices, the authors use weekly time series data in the postconspiracy period and fit a regression model of frozen perch log price as a function of current and lagged costs, as measured by fresh perch log prices. The model is used to backcast but-for conspiracy price in two earlier periods that preceded the collapse of the bid-rigging scheme. They find the price during the conspiracy period is significantly above the but-for predicted price in every auction, with an average cartel markup in the range of 23.1% to 30.4%, depending on the period chosen to define the conspiratorial conduct. By examining the first and second moments from the same conspiracy collapse, Abrantes-Metz and coauthors (“AFG&T”) (2006), propose a screen based on the standard deviation of price normalized by its mean, or the coefficient of variation. They find that while the mean price decreased by 16%, the coefficient of variation increased by 332% following the conspiracy collapse. The mean and standard deviation of the cost is also higher under competition, but not by nearly enough to account for the increase in price. Thus, the authors conclude that during the postconspiracy period, price began to covary more closely with cost, and thus exhibited larger variation (over time). 3  They note that prior studies finding that profits or prices fell after antitrust enforcement action are unreliable because they ignore cost differences and other factors (see, for example, Erickson, 1976 and Parker, 1969). 4  Five companies and associated individuals pleaded guilty to separate felonies and were fined. 5  The authors assume that the gap between the two periods represents a transition from collusion to competition.

536   Michael J. Doane, Luke M. Froeb, David S. Sibley, and Brijesh P. Pinto Based on this finding, the authors hypothesize that other conspiracies would exhibit similar patterns, and apply their “variance screen” to retail gasoline stations in Louisville, Kentucky in 1996–2002. The Louisville area is considered a good candidate for collusion because gasoline in Kentucky at both wholesale and retail is moderately concentrated and uses a unique formation of gasoline, that is, reformulated gasoline. They note that while a cartel the size of a city would be costly to organize and police, there may be a degree of market power conferred by the elimination of localized competition. The variance screen would identify a potential cartel as a cluster of gasoline stations located close to one another exhibiting lower price variation and higher prices relative to other stations in the city. They find no such clusters and conclude competitive conduct is a more plausible explanation than collusion. In contrast, Bolotova, Connor, and Miller (2008) find mixed evidence that price variance during collusion is lower than that observed prior to the formation of a cartel. The study considers two well-documented episodes of collusion: the lysine cartel (1991–1995) and the citric acid cartel (1992–1995). They hypothesize that the mean price is higher and the variance of the price is lower during collusive than noncollusive periods. Using monthly average contract citric acid prices and average monthly lysine prices, their statistical analysis is based on extensions of the traditional autoregressive conditional heteroskedasticity (“ARCH”) and generalized ARCH models. They find that mean prices are higher in both cartels during the collusive period. However, they find price variance is lower during the lysine cartel, but higher during the citric acid conspiracy than it was during more competitive periods. The authors suggest that foreign competition might account for this outcome. They conclude that the variance screen may be a useful tool to detect conspiracies that do not significantly raise price but tend to homogenize business practices, which may raise profits and also reduce variance. Abrantes-Metz and coauthors (2012) examine manipulation of the US dollar one-month LIBOR rate. Their paper is motivated by a May 29, 2008, Wall Street Journal article suggesting that several global banks were reporting LIBOR quotes significantly below those implied by prevailing credit default swap (“CDS”). To test this claim, they compare LIBOR with other short-term borrowing rates, analyze individual bank quotes, and compare these individual quotes to CDS spreads during three periods: January 1, 2007–August 8, 2007 (Period 1), August 9, 2007–April 16, 2008 (Period 2), and April 17, 2008–May 30, 2008 (Period 3). The three periods are separated by two dates in which major news events occurred: (1) on August 9, 2007, there were related press releases on (a) a coordinated intervention by the European Central Bank, the Federal Reserve Bank, and the Bank of Japan; (b) AIG’s warning that defaults were spreading beyond the subprime sector; and (c) BNP Paribas’s suspension of three mortgage-backed funds; and (2) on April 17, 2008, the British Banding Association announced its intent to investigate its LIBOR-setting process. Their screening methodology proposes three tests for collusion. First, they examine the relationship between LIBOR and other major benchmarks, assuming those to be nonmanipulated at the time they are used as benchmarks. Specifically, they test whether the LIBOR rate is manipulated downward during Period 2, relative to Periods 1 or

Screening for Collusion as a Problem of Inference   537

3. They cannot reject the null hypothesis that LIBOR rates are the same as the predicted rate. Thus, they conclude that the evidence on the average level of the LIBOR rate is consistent with the absence of a material manipulation. The authors also examine the pattern of individual LIBOR quotes by 16 participating banks. In particular, they examine how likely it is it that a large number of banks will submit identical LIBOR quotes without coordinating. To that end, they first examine the intraday variance of these individual quotes. They also calculate the frequency with which each bank appears in the “deciding group,”6 and identify banks that tend to be in the deciding group most often. They hypothesize that ‘‘manipulative’’ banks should cluster together in nonrandom patterns. To test this, the authors compute pairwise correlations between all possible bank-pairs and calculate the frequency with which each bank appears in the deciding group, and identify a group of banks that tend to be in the deciding group very often. From this analysis, the null hypothesis of nonrandom clustering patterns is rejected, raising the possibility of manipulation. The third collusion test follows from an analysis of the relationships between individual LIBOR quotes and proxies for individual borrowing costs as determined by CDS spreads. In particular, they examine whether banks with relatively low CDS spreads are also banks with relatively low LIBOR quotes. They detect several banks whose ordinal positions in LIBOR quotes are unrelated to their ordinal positions in CDS spreads, and raise the possibility of manipulation. Abrantes-Metz and Addanki (2007) propose a variance (over time) screen to detect manipulation of commodity prices based on an analysis of the Hunt Brothers’ silver manipulation episode of 1979–1980. Their screen is based on the notion that manipulation provides an informational advantage to manipulators over the rest of the market (i.e., market participants are fooled). This implies that when market participants form expectations on the likely levels of future prices, these forecasts are systematically wrong more often than they would otherwise have been in the absence of manipulation. Thus, the variable of interest is the variance in the forecasting error of future spot prices. If the collusion is effective in reducing prices, this will result in more negative forecast errors. Using daily futures and spot prices for silver from Comex from February 1975 through April 2004, their analysis examines the relationship between the futures contract price at maturity date T and the realized spot price on that date. They find that the coefficient of variation of the forecast errors is larger during the manipulation period, and that this result holds when regression analysis is used to control for changes in macroeconomic conditions. Abrantes-Metz and Metz (2012) attempt to determine how far screens can go in distinguishing explicit from tacit collusion. In doing so, they consider evidence from the LIBOR setting. Their purpose is to determine whether movements in LIBOR rates are best explained by noncooperative behavior, tacit collusion, or explicit collusion. Their analysis is based on an inspection of the coefficient of variation (across banks) in daily 6  LIBOR is established as the simple average of the middle set of eight quotes that are submitted by the 16 participating banks. These eight banks comprise the “deciding group.”

538   Michael J. Doane, Luke M. Froeb, David S. Sibley, and Brijesh P. Pinto LIBOR quotes for a cross section of 16 participating banks. They find that the coefficient of variation is near zero from early August 2006 through early August 2007, becoming abruptly positive thereafter. They note that if all banks were submitting unique quotes each day (which happened to average to the same level day after day), the coefficient of variation would be larger. On the other hand, if all banks submitted essentially the same quote, the coefficient of variation would be low, and if the middle eight quotes were identical, it would be zero. Their test for noncooperative behavior is whether the observed convergence across LIBOR quotes from the 16 participating banks is explained by identical borrowing costs. If so, they note that the triggering of the financial crisis would have affected all banks equally. Since the banks considered differ significantly in terms of their characteristics and borrowing costs (i.e., they have asset portfolios of varying risk, varying liability structures, and participate to different degrees in different market segments), they reject the noncooperative hypothesis. To distinguish between tacit and explicit collusion, the hypothesis is that the data would be more consistent with tacit collusion if banks were “learning” from the strategic reaction of the other banks. In this setting, a transition period would be observed in which the variation of intraday quotes would be decreasing towards zero. Since an abrupt transition to zero is observed, they conclude the data are inconsistent with tacit collusion. The authors note that the tacit collusion hypothesis is further complicated by the fact that individual bank quotes are sealed and are made public only after the LIBOR is computed. Thus, if the banks submit the same quote day after day, and other banks were learning and converging toward that common quote, the data may be described as consistent with tacit collusion. On the other hand, if many banks submit a common sealed quote one day, and a common but different sealed quote the next day, tacit collusion is a less likely explanation. Based on an inspection of individual bank quotes, they conclude that the tacit collusion (or learning theory) hypothesis does not fit the data well, leaving explicit collusion as the explanation that is most consistent with the observed LIBOR quotes. Eruthku and Hildebrand (2010) employ a differences-in-differences approach to determine whether a public announcement of an antitrust investigation (which triggered the collapse of a cartel) may be used to detect a price-fixing conspiracy. In May 2006, the Canadian Competition Bureau publically announced its investigation of retail gasoline markets in some local markets in Quebec. The authors’ test for collusion is whether retail prices fell following the announcement. The study employs weekly retail and wholesale prices for the period May 31, 2005–May 22, 2007, providing 52 weeks of data before and after the announcement. In a difference-in-difference regression analysis, Sherbrook is the “treatment” city and Montreal and Quebec City are “control” cities. They find that the announcement decreased the price differential between Sherbrook and Montreal by 1.75 cents per liter. This reduction is statistically significant at the 5% level. Interestingly, the authors find that the variance of retail prices for all cities in their sample increased after the announcement, but not significantly so.

Screening for Collusion as a Problem of Inference   539

Jiménez and Perdiguero (2011) apply a price variance (over time) screen to the retail gasoline market in the Canary Islands, Spain. The Canary Islands were selected because retail gasoline markets on different islands have different market structures. On some islands the markets have more than one supplier, while each of the remaining two islands is monopolized by a single supplier, the DISA company. The authors find that the coefficient of variation of the companies on oligopolistic islands is between 1.06% and 8% higher in the oligopolistic islands. From this, they conclude that a monopolistic firm follows a more rigid price path than do retail outlets on the oligopolistic islands. However, we note that the null hypothesis that the difference in means between the two groups is zero is not rejected using a 95% level of significance. The authors conclude that analyzing the results of the variance screen without comparing them to a benchmark would not enable them to draw any definitive conclusions. Note also that the use of the monopoly islands as a collusive benchmark (H1) may not be a good metaphor for kind of price rigidity that might occur when multiple firms try to coordinate pricing. The authors also employ a screen to try to find pockets of retailers with higher prices and lower coefficients of variation, similar to the screen used by AFG&T (2006). Their competitive benchmark is a retail gasoline supplier that traditionally competes more aggressively on price: PCAN. The study tests whether PCAN stimulates competition using two different approaches. The first involved examining prices and the coefficient of variation for prices in towns with and without PCAN gas stations. They find prices are always higher and the coefficient of variation smaller if PCAN is not present. The second test is whether the presence of an independent retailer within a half-mile radius is correlated to lower and more variable gasoline prices. They find both. Based on the variance screens, they conclude that the average performance of the gas retail outlets (excluding those run by PCAN) is very close to that shown on a monopolistic island, and further from CPAN. Genesove and Mullin (2001) review the rules of the Sugar Institute and meetings/ notes among its members. The trade association was formed in December 1927 following several years of declining margins and excess capacity. The trade association’s 14 members comprised nearly all the cane sugar refining capacity in the United States. The cartel did not directly fix output or set prices but instead homogenized business practices, thereby facilitating members’ ability to detect secret price cuts. The study is based on the average weekly prices of refined sugar and its primary input cost, raw sugar, in the United States from 1914 to 1941. Using these data, the authors calculate the yearly price-cost margin for sugar refining in the United States for three periods: before, during, and after the cartel period. The authors use a Lerner Index equal to 11% as a benchmark for monopoly pricing. This was the margin in 1892 when US domestic refiners’ share and margin reached their highest level. The authors’ hypothesis is that price-cost margins are higher and the variance of margin lower in collusive period than competition. The authors find that the conspiracy raised price-cost margins to about 75% of the monopoly level and the variance in this margin dropped by nearly 100% while the cartel remained active.

540   Michael J. Doane, Luke M. Froeb, David S. Sibley, and Brijesh P. Pinto As noted above, the use of variance screens has become rather widespread in empirical work. For example, Brannon (2003) summarizes research performed for the Joint Economic Committee, US Congress on the effects of resale price maintenance laws on petrol prices and station attrition in Wisconsin. The study concluded that the State of Wisconsin enacted a policy that implicitly taxes its citizens for the benefit not just of small independent stations but also for the large multinational oil companies that operate in Wisconsin. The article calculates the average margin and the variance for two markets that were affected by this legislation, as well as a control group market. The results show that the average margin was higher in the collusive markets, while the findings as regards to variance were not particularly conclusive. In another example, Abrantes-Metz and Pereira (2007) analyzed the mobile phone sector in Portugal before and after the entry of a new operator, Optimus.7

21.4.2.  Screens Based on Structural Models Christie and Schultz (1994) provide evidence to show that in 1991, NASDAQ dealers avoided odd-eighth quotes for 70 of 100 large, actively traded NASDAQ securities, including Apple Computer and Lotus development. The avoidance of odd-eighth quotes ensured an inside spread of at least $0.25 for these securities. The authors compare the dollar spreads for the NASDAQ and NYSE/AMEX stocks and find a systematic difference in the frequency of odd-eighth spreads in the two pools. For example, odd-eighth spreads of $0.125 and $0.375 occur nearly three times as often for NYSE/AMEX securities as for NASDAQ securities, where the spreads are mostly in even-eighths, such as $0.25, $0.50, and $0.75. The authors test which of three hypotheses best explains this odd-eighth avoidance: (1) coarse pricing increments to lower negotiation costs, (2) cost determinants of the spread, and (3) tacit collusion among dealers. Finding that, in the data, larger trades are far more likely to occur on odd-eighth spreads than smaller trades, the authors rule out the first hypothesis of negotiation costs. Next, the authors use logistic regression analysis to determine that economic determinants such as volume, volatility, market capitalization, and stock price do not play a role in predicting the probability that stocks are quoted using odd eighths. By elimination, the authors conclude that the only remaining hypothesis explaining the absence of odd-eighth spreads for NASDAQ securities is tacit collusion among market makers. Bresnahan (1987) attempts to explain a dramatic increase in US automobile production that occurred in 1955. In 1955, relative to the two surrounding years, automobile production was higher by 45% and quality-adjusted prices lower. The paper tests the hypothesis that there was a supply shock of a specific form: a one-year increase in the competitiveness of conduct in the industry. A structural model of supply and demand is estimated to test the fit of various firm behavioral hypotheses. Nonnested (Cox) 7  Other papers that discuss or apply the approach include Esposito and Ferroro (2006), Muthusamy et al. (2008), Abrantes-Metz (2011), and Abrantes-Metz and Bajari (2011).

Screening for Collusion as a Problem of Inference   541

hypothesis tests find that the collusive outcome fits the data best in 1954 and in 1956, while the competitive outcome fits the data best in 1955. Bresnahan’s paper is regarded as one of the few empirical studies finding evidence of tacit collusion. Rapson (2009) attempts to replicate the findings from Bresnahan’s 1987 study to determine whether the conclusion of tacit collusion hinged on restrictive assumptions about demand. In particular, he examines the independence of irrelevant alternatives assumption (“IIA”) which makes demand a function only of the price and quality of adjacent models on the quality spectrum, rendering nonadjacent price changes irrelevant to predicted demand. Rapson notes that the restrictive demand specification could cause intrafirm pricing strategies consistent with differentiated product Bertrand to be mistakenly interpreted as collusion. To allow for more realistic demand behavior, he employs a random coefficients logit model. He finds that for no year can Bertrand competition be rejected in favor of tacit collusion. However, his results also indicate that firms were not maximizing profit during this period. He concludes that a firm-competition model is rejected in favor of brand competition in 1955, and is a better fit (though insignificantly) in 1954 and 1956. Structural models have also been developed to analyze bidder collusion in auctions. In an early paper commenting on the paucity (at the time) of economic literature on the subject of collusion in auctions, Hendricks and Porter (1989) pointed out that the characteristics of a collusive scheme in an auction depend on the rules of the auction and the nature of auctioned object. They also expressed surprise that the empirical literature on collusion in auctions was lacking even as detailed data sets were publicly available. Porter and Zona (1993) develop an econometric test to detect “phantom bidding” in procurement auctions, based on the differences in bidding between cartel members and nonmembers. Their test exploits information on who was and was not in the cartel, and the form of the conspiracy (cartel members used phantom bidding). Using procurement auction data from the New York State Department of Transportation (“DOT”), Porter and Zona apply their test to Nassau and Suffolk county DOT contract lettings from 1979 to 1985, and find that the behavior of the collusive firms—one convicted in federal court of bid rigging and four unindicted coconspirators—was statistically different from that firms that did not belong to cartel. In a related paper, Porter and Zona (1999) develop a test to detect bid rigging in school milk procurement auctions. They list several characteristics of the school milk procurement market that facilitate their analysis: an auction design that was well understood, a relatively homogenous product, a fixed set of potential bidders in the short run, and a straightforward production process. In 1994, the State of Ohio charged 13 dairies with bid rigging in school milk procurement auctions from 1980 to 1990. They compare the bidding behavior of three defendant firms to that of nondefendant firms and find that each defendant firm’s bidding function is statistically different from the bidding functions of nondefendant firms. Porter and Zona also estimate the increase in price resulting from the collusion. They determine that bid rigging resulted in a 6.5% increase in the price of milk paid by the school districts.

Table 21.2  Empirical Tests/Screens Study

Screen? H1: Collusion

H0: Competition

X: Variable

Data

Conclusion

Comments

Category 1 Froeb, Koyak, & Werden (1993)

Test

DOJ investigation

Postcollusion period identified by drop in price

Price level

Weekly time series, 1984–89. Aggregate weekly bid prices and cost of fresh perch.

Collusion raised price by 23%.

Cost of fresh fish used as regressors (current and lagged) to control for cost

Test

Profit margins higher and variance of margin lower

No change in profit Annual profit margins margin or variance in calculated from prices profit margin of refined and raw sugar

Average weekly prices of refined and raw sugar in US, 1914–41; output of Atlantic refiners; imports of refined sugar; and domestic sugar beet production

Conspiracy raised price cost margin to about 75% of the monopoly level; rivals outside of collusive agreement responded to price increase by increasing output; variance in profit margin dropped by nearly 100% during conspiracy.

Paper reviewed the rules of the Sugar Institute and meetings and communications among members.

Genesove & Mullin (2001)

AbrantesMetz et al. (2006)

AbrantesMetz & Addanki (2007)

Test

Same as Froeb, Koyak, & Werden (1993)

Same as Froeb, Koyak Price coefficient of & Werden (1993) variation

Screen

Geographic clusters of gas No such clusters stations with significantly lower price variance

Test

Future spot prices are harder to predict using current spot prices in presence of collusion in spot market.

Future spot prices are easier to predict using current spot prices in presence of competition in spot market.

Conspiracy did not fix prices or output but homogenized business practices to make price cutting more transparent.

Same as Froeb, Koyak & Collusion raised price Werden (1993) coefficient of variation by 332%.

No control for cost

Price coefficient of variation

Daily time series, 1996–2002. Price data at 279 gas stations in Louisville.

No clusters of significantly lower price variance found

Imputed missing price data using AR(1) process. No control for cost.

Forecasting error of future spot prices. Control variables include monthly interest rate, 10-yr. T- bill rate, S&P 500 Index, monthly exchange rates for a basket of commodities.

Daily data on futures and spot prices for silver from Comex, 2/1975–4/2004.

Apply approach to the Hunt Brothers silver manipulation episode of early 1980s, and find that under manipulation the forecasting error is more volatile than under nonmanipulation, controlling for market fundamentals

Some evidence that spot prices are more volatile when they are above futures prices, than when they are below.

Bolotova, Connor, & Miller (2008)

Test

Mean price is higher and variance in price is lower.

No support for hypothesized changes in first and second moments of price distribution

Price level and variance. Analysis based on extensions of the autoregressive conditional heteroskedasticity (ARCH) model and generalized ARCH (GARCH).

Monthly average contract citric acid prices (2/1990– 4/1997) and average monthly lysine (1/1990– 6/1996)

Mean prices are higher in both citric acid and lysine cartels; price variance during the lysine conspiracy was lower, whereas variance during the citric acid conspiracy was higher than it was during more competitive periods.

“Variance is a useful tool for detecting conspiracies that do not significantly raise price but tend to control the price variance by homogenization of business practices, which may raise profits.”

AbrantesMetz & Metz (2012)

Screen

One-month LIBOR rate nearly constant from 8/2006–8/2007. Almost no variation in the middle eight bank quotes during same period. Tacit collusion:  if banks were “learning” from strategic interaction, there would be a transition towards convergence to an identical rate. Explicit collision: banks sealed bids move simultaneously to same bid from day to day.

Noncooperative outcome:  convergence across LIBOR quotes due to identical borrowing costs; triggering of financial crisis would affect all banks equally.

Coefficient of variation for individual one-month LIBOR quotes for 16 banks. Comparisons of LIBOR rate to T-bill rate and federal funds effective rate

Daily one-month LIBOR quotes for 16 participating banks surveyed by the British Banking Association, 2006–8.

Reject noncooperative hypothesis because banks do not have identical borrowing costs and would be affected by financial crisis in different ways. Reject tacit collusion because convergence was immediate and sealed bids for individual banks moved to same bid from day to day. Explicit collusion is the remaining explanation.

There are particular circumstances under which screens can go one extra step in helping to distinguish explicit from tacit collusion. They correspond to situations in which one can observe the dynamics of collusion; in this case, the quoting patterns in order to identify how the bids/quotes may have become identical. (Continued)

Table 21.2 Continued Study

Screen?

H1: Collusion

H0: Competition

X: Variable

Data

AbrantesMetz et al. (2012) (cont.)

Screen

Manipulated LIBOR rates are lower.

1: LIBOR rates are higher.

Spread comparisons of one-month LIBOR rate to one-month T-bill and the Federal Fund effective rate. Comparison of actual and predicted LIBOR rates in manipulation period based on regression analysis in clean period, LIBOR rate = f(federal funds rates).

Screen

Individual LIBOR quotes of “manipu­lative” banks are clustered together in nonrandom patterns.

2: Individual LIBOR quotes of “manipulative” banks are not clustered in nonrandom patterns.

Coefficient of variation Same as above for individual one-month LIBOR quotes for 16 banks

Individual quotes in Period 1 are more tightly clustered than in other periods. Reject null.

Screen

Banks with relatively low (high) collateral debt swap (CDS) spreads are also banks with high (low) LIBOR quotes.

3: No negative correlation between a bank’s LIBOR rate and its CDS spread

Correlations of individual bank LIBOR quotes and CDS spreads and bank market capitalization

Some outlier banks ordinal position in LIBOR quotes is unrelated to their ordinal position in CDS spreads.

Rates analyzed for 3 periods: 1/1/2007– 8/8/2009 (Period 1); 8/9/2007–4/16/2008 (Period 2); and 4/17/2008– 5/30/2008 (Period 3)

Same as above

Conclusion

Comments

Actual LIBOR rates are not Cost control is statistically different from Federal Funds Rate predicted rate. Cannot reject null.

Jiminez & Perdiguero 2012)

Screen

Higher price coefficient of variation relative to a competitive benchmark

Price variation not different from competitive benchmark

Price coefficient of variation

Prices of petrol 95 Gas stations in an oligopoly and diesel, weekly have a higher coefficient of time series, September variation. 2008 to April 2009 (24 weeks); 420 petrol 95 stations and 391 diesel stations

Author’s note: Missing data filled in using interpolation.

Author notes that findings are robust against alternative specifications.

Category 2: Structural models Bresnahan (1987) Test

Firms set prices to maximize their joint profits.

Each firms sets prices to maximize its own profit, taking the prices of other firms’ products as given.

Equilibrium prices and quantities calculated by solving demand systems and firm first-order conditions. Nonnested (Cox) hypothesis tests.

Automobile price, quantities, and characteristics, 1954–56.

Porter & Zona (1993)

Phantom bidding

No phantom bidding

Log of a firm’s bid on a particular job

Subset of all Nassau Cartel firms’ bids are and Suffolk county DOT statistically different from contract lettings from competitive firms’ bids. April 1979 through March 1985; 575 bids were submitted on 116 projects.

Avoidance of odd-eighth quotes is attributable to tacit collusion among bidders.

Avoidance of Equals 1 if a security is odd-eighth quotes quoted in odd-eighths is attributable (logistic regression) to economic determinants of spread such as volume, volatility, market capitalization, and stock price, and collusion is absent.

Test

Christie & Schultz Test (1994)

Inside quotes for 100 large, actively traded NASDAQ securities and for 100 comparable NYSE/AMEX securities, for the year 1991

1954 and 1956 generate a superior model fit under collusion hypothesis.

Avoidance of odd-eighth quotes is not attributable to economic determinants of spread. Tacit collusion among market makers is the only explanation of avoidance of odd-eighths.

Earlier, the authors rule out that odd-eighth avoidance results from coarse price increments to lower negotiation costs.

(Continued)

Table 21.2 Continued Study

Screen?

Porter & Zona (1999)

Test

Bajari & Ye (2003) Screen

Screen

Screen

H1: Collusion

H0: Competition

X: Variable

Data

Conclusion

Bidding behavior is inconsistent with that of control group of nondefendants.

Bidding behavior is consistent with that of control group of nondefendants

Equals 1 if bid is submitted in a district (probit regression)

School milk procurement data for 509 districts in Ohio from 1980 through 1990

Collusion raised market prices by about 6.5%.

Bids are independent after controlling for all information about costs.

Bids are not conditionally independent.

For a given firm and project, the ratio of the amount bid by the firm on the project to the engineer’s cost estimate for the project

Among 23 pairs with at least four simultaneous bids, the null hypothesis cannot be rejected except for four pairs of firms. In only one pair did bidders bid against each other more than a handful of times.

Bids are exchangeable: costs alone should determine how firms bid. Holding cost information constant, a firm’s bid shouldn’t depend on the identities of its competitors.

Bids are not exchangeable.

For a given firm and project, the ratio of the amount bid by the firm on the project to the engineer’s cost estimate for the project.

Detailed bidding information for almost all the public and private road construction projects conducted in Minnesota, North Dakota, and South Dakota during the years 1994–98. Data set contains nearly 18,000 procurement auctions.

Probability of collusion is low.

Firm’s private cost (estimated), used to calculate marginal likelihood of each model, in turn used to calculate posterior probabilities that equilibrium model is competitive/collusive

Only one pair (different from the four above) fails the exchangeability test. Also, the bidders in the pair bid against each other more than a handful of times. Pair 1 failing C.I. and pair 2 failing exch were models 2 and 3, model 1 was competition. Posterior probability of competitive model is highest.

Comments

Eruthku & Hilderbrand (2010)

Screen

Canadian Competition Bureau investigation of retail gasoline pricing in Sherbrook, Canada

Postinvestigation announcement decline in retail gasoline pricing

Weekly retail gasoline pricing. Difference-indifference regression model using prices in three cities: Sherbrook, Montreal, and Quebec.

Weekly data in cities during period 5/31/2005– 5/22/2007; 52 weeks before and after announcement

Announcement of conspiracy triggered a 1.75 cents per litre decline in prices.

Cost control is wholesale gasoline price. Stiglerian theory of collusion leads to ambiguous conclusion regarding likely collusion in Sherbrook.

1980–2007 annual time-series data for five German industries. German cement industry is the test market for a cartel (lasted for 21 years, 1981–2002). Four competitive German industries were considered— automobile, electrical, chemical, and mechanical industries.

The System of Collusive Markers indicates that the Cement industry was collusive during the relevant time period, and that all other industries were competitive.

Author’s note: SCM is more suitable for the detection of long-term cartels.

Category 3: Collusive market other than first and second moments of price distribution Blanckenburg & Geist (2009)

Screen

(2) Low level of capacity utilization, (2) slackness of price adjustments to exogenous shocks, (3) excess rates of return, (4) nearly constant capacities, (5) lower variance of price changes, and (6) lower variance of capacity growth rate

(1) No permanent excess capacities, (2) positive correlation, (3) no permanent excess rate of return, (4) positive correlation between rate of return difference and capacity growth rate changes, (5) higher variance of price changes, and (6) higher variance of capacity growth D15 rate changes

(1) Differenced indicator that measures the difference between supply and demand quantities, (2) change in nominal price level, (3) difference between a rate of return representing all firms prevalent in the market under consideration and a broader industry class comparison rate of return, (4) capacity growth rate

548   Michael J. Doane, Luke M. Froeb, David S. Sibley, and Brijesh P. Pinto Bajari and Ye (2003) develop a screen to detect bidder collusion in procurement auctions where bidders are asymmetric with respect to their ex ante costs. They posit two conditions that are necessary and sufficient for competition in their model: conditional independence and exchangeability. Under conditional independence, bids are independent after controlling for publicly observable cost information. Exchangeability implies that only costs, not the identities of other bidders, determine how a firm bids. Thus, Bajari and Ye’s null hypothesis of competition is that bids satisfy both conditional independence and exchangeability. Bajari and Ye apply their screen to the market for asphalt contracts in the 1994–1998 period, a market with a known history of collusion in the 1980s. They identify two pairs of firms that fail both the conditional independence and the exchangeability tests. Next, they use a Bayesian framework to choose among three alternative models of industry equilibrium: (1) competitive bidding by all firms, (2) collusive bidding by the first pair and competitive bidding by all other firms, and (3) collusive bidding by the second pair and competitive bidding by all other firms. They find that the posterior probability of the competitive model is the highest, and conclude that the bidding in the market is competitive. Marshall, Marx, and Raiff (2008) perform an empirical analysis to determine whether the timing of price announcements (X) after 1985 in the vitamins industry is the result of collusion rather than the result of demand and cost phenomena (H0). They construct a data set comprising the dates and number of public price announcements in the United States during the time period 1970–2001 for 10 vitamin products. The authors estimate a logit model for the probability that firms announce a price change in a given month as a function of the number of months elapsed since the previous announced price change, and demand and supply effects. The estimated coefficients of the logit regression show that for the two time periods prior to 1985, the time elapsed from the previous announcement does not have a significant effect on the probability of a new price announcement. However, the time elapsed from the previous announcement has a significant effect on the probability of a new price announcement for the two periods after 1985. Among demand and supply factors, the only variable with a significant coefficient for any time period is the exchange-rate, with a significant coefficient in the first time period after 1985. The authors interpret the results of the logit regression as implying that after 1985, the timing of public price announcements is tied to the timing of cartel meetings. Marshall, Marx, and Raiff conclude that the probability of new price announcements provides a potentially valuable tool for the identification of explicit collusion. The alternative hypothesis of collusion (H1) is that there is a much higher probability of price announcements (X) relative to a “clean” period.

21.4.3.  Screens Based on Collusive Markers Other Than Price Screens Collusive markers other than price have been have been used to try to detect cartels. Economic theory predicts that certain industry characteristics manifest differently

Screening for Collusion as a Problem of Inference   549

under the presence of collusive behavior than under that of competitive behavior. These characteristics have to do with industry processes such as market clearing, technological innovation, rate of return, and product innovation among others (see Lorenz 2008). The outcomes of these processes are different under competitive and collusive behavior, and, thus, serve as a means to detect cartels. Blanckenburg and Geist (2009) postulate these six markers as a tool for testing the “workability of markets.” These markers are (1) utilization rate of capacities, (2) correlation between the utilization rate of capacities and price changes, (3) difference between the rate of return in the industry and a broader comparison rate of return, (4) correlation between the rate of return difference and capacity growth rate changes, (5) variance of price changes, and (6) variance of capacity growth rate changes. Table 21.2 shows the null and alternative hypotheses associated with these markers. While Blanckenburg and Geist admit that cartel formation is only one of many reasons for the nonworkability of markets, they suggest that the system of collusive markers can be used by antitrust authorities as an aid in the detection of long-term cartels. They also apply their screen as an empirical test using annual time-series data on five German industries from 1980 to 2007. Among the five German industries they consider is the German cement industry, which is known to have had a cartel that lasted from 1981 to 2002. The other four industries serve as competitive benchmarks. Blanckenberg and Geist’s system of markers correctly identifies the cement industry as collusive and the others as competitive.

21.5. Conclusion Collusion screens (and tests) have been used in an informal way to help determine how to allocate scarce prosecutorial resources. Give their history, some might object to our formal characterization of screens by analogy to Bayesian hypothesis testing. But all screens are designed to help us make better decisions, so understanding exactly how they do this allows us to reduce a collusion screen to its essential elements. We can compare and contrast the growing number of screens that have been proposed by academics or used by practitioners in table 21.3. Formally modeling screening also allows us to better understand the conditions under which screens are likely to fail. The first is that the empirical indicator may not be able to differentiate between competition (H0) and collusion (H1). The second is that the null hypothesis is not indicative of competition or that the alternative is not indicative of collusion. The third is that the world is neither H0 nor H1. The second and third, often underappreciated reasons for failure, are due to what we call the problem of “model fit”: the screen is based on assumptions that do not fit the industry to which the screen is being applied. This criticism is similar to that leveled against the use of structural oligopoly models for merger screening (Werden et  al., 2004). Before a structural model is used

550   Michael J. Doane, Luke M. Froeb, David S. Sibley, and Brijesh P. Pinto Table 21.3  Summary of Findings Test Type

Study

Product / Industry

Conclusion

First Moment

Froeb, Koyak, and Werden (1993) Genesove and Mullun (2001) Bolotova, Connor, and Miller (2008) Bolotova, Connor, and Miller (2008) Abrantes-Metz et al. (2012)

Frozen fish Refined and raw sugar Lysine Citric acid LIBOR

Bid rigging Collusion Collusion Collusion No collusion

Second Moment

Genesove and Mullun (2001) Abrantes-Metz and Addanki (2007) Bolotova, Connor, and Miller (2008) Bolotova, Connor, and Miller (2008) Abrantes-Metz and Metz (2012) Abrantes-Metz, Kraten, Metz, and Seow (2012) Jiménez and Perdiguero (2012)

Refined and raw sugar Commodity prices Lysine Citric Acid LIBOR LIBOR

Collusion Market manipulation Collusion Competition Manipulation Explicit collusion

Petrol 95 and diesel

Oligopoly pricing

Structural Model

Bresnahan (1987) Porter and Zona (1993)

Automobile Highway construction contracts NASDAQ securities School milk supply contracts Seal coating contracts Vitamins Automobile Gasoline (Retail)

Tacit collusion Phantom bidding

Christie, Harris, and Schultz (1994) Porter and Zona (1999) Bajari and Ye (2003) Marshall, Marx, and Raiff (2008) Rapson (2009) Eruthku and Hildebrand (2010)

Collusion Bid rigging Competition Explicit collusion Competition Price-fixing conspiracy

to predict whether mergers will raise price, enforcers should make sure that the model is able to characterize the significant features of observed competition within the industry. The analogy to criticism of the use of structural models for collusion screening is obvious. Less obvious is the analogy to screens based on indicator variables chosen from past episodes of collusion. Here the assumption on which the screen is based is that the industry being screened exhibits collusive and competitive behavior similar to that of the industry in which collusion occurred. Before the screen is used, practitioners should make sure that the industry “fits” this often implicit assumption. As a final note of caution, it is important to remember that all models, both theoretical and empirical, are abstractions away from reality. This means that one can always find some feature of the industry being screened that is at odds with the model assumptions. But the art of model building is to determine whether this omission is likely to bias the screen’s predictions, or whether the screen works well despite the omission.

Screening for Collusion as a Problem of Inference   551

Acknowledgments We acknowledge useful comments from Joe Harrington, David Sappington, Danny Sokol, Bart Victor, and seminar participants at Vanderbilt University and the University of Florida.

References Abrantes-Metz, R. 2011. Design and Implementation of Screens and Their Use by Defendants. Competition Policy International Antitrust Chronicle 2: 1–11. Abrantes-Metz, R. 2012. Why and How to Use Empirical Screens in Antitrust Compliance? Competition Policy International Antitrust Chronicle 1: 1–8. Abrantes-Metz, R., and S. Addanki. 2007. Is the Market Being Fooled? An Error-Based Screen for Manipulation. Working paper. Abrantes-Metz, R., and P. Bajari. 2010. Screens for Conspiracies and Their Multiple Applications. Competition Policy International 6: 129–44. Abrantes-Metz, R., and L. Froeb. 2008. The Agencies Are Screening for Collusion: What Are They Likely to Find? ABA Newsletter. Abrantes-Metz, R., L. Froeb, J. Geweke, and C. Taylor. 2006. A Variance Screen for Collusion. International Journal of Industrial Organization 24: 467–86. Abrantes-Metz, R., M. Kraten, A. Metz, and G. Seow. 2012. LIBOR Manipulation? Journal of Banking and Finance 36: 136–50. Abrantes-Metz, R., and A. Metz. 2012. How Far Can Screens Go in Distinguishing Explicit from Tacit Collusion? New Evidence on the LIBOR Setting. Antitrust Chronicle 3: 1–9. Abrantes-Metz, R. M., and P. Perrira. 2007. The Impact of Entry on Prices and Costs. Working paper. Athey, S., K. Bagwell, and C. Sanchirico. 2004. Collusion and Price Rigidity. Review of Economic Studies 71: 317–49. Bajari, P., and L. Ye. 2003. Deciding between Competition and Collusion. Review of Economics and Statistics 85: 971–89. Blanckenburg, K., and A. Geist. 2009. How Can a Cartel Be Detected? International Advances in Economic Research 15: 421–36. Bolotova, Y., J. Connor, and D. Miller. 2008. The Impact of Collusion on Price Behavior: Empirical Results from Two Recent Cases. International Journal of Industrial Organization 26: 1290–307. Brannon, I. 2003. The Effects of Resale Price Maintenance Laws on Petrol Prices and Station Attrition: Empirical Evidence from Wisconsin. Applied Economics 35(3): 343–49. Bresnahan, T. 1987. Competition and Collusion in the American Automobile Industry: The 1955 Price War. Journal of Industrial Economics 35: 457–82. Christie, W. G., J. H. Harris, and P. H. Schultz. 1994. Why Did NASDAQ Market Makers Stop Avoiding Odd-Eighth Quotes? Journal of Finance 49(5): 1841–60.

552   Michael J. Doane, Luke M. Froeb, David S. Sibley, and Brijesh P. Pinto Christie, W. G., and P. H. Schultz. 1994. Why Do NASDAQ Market Makers Avoid Odd-Eighth Quotes? Journal of Finance 49(5): 1813–40. Cooper, J., L. Froeb, D. O’Brien, and M. Vita. 2005. Vertical Antitrust Policy as a Problem of Inference, International Journal of Industrial Organization 23: 639. Easterbrook, F. 1984. The Limits of Antitrust. Texas Law Review 63(1): 1–39. Erickson, W. B. 1976. Price Fixing Conspiracies: Their Long-Term Impact. Journal of Industrial Economics 24: 189–202. Esposito, F., and Ferrero, M. 2006. Variance Screens for Detecting Collusion: An Application to Two Cartel Cases in Italy. Italian Competition Authority, Working Paper Eruthku, C., and V. Hildebrand. 2010. Conspiracy at the Pump. Journal of Law and Economics 53: 223–37. Froeb, L., R. Koyak, and G. Werden. 1993. What Is the Effect of Bid-Rigging on Prices? Economics Letters 42: 419–23. Froeb, L., S. Tschantz, and G. Werden. 2005. Pass-Through Rates and the Price Effects of Mergers. International Journal of Industrial Organization 23: 703–15. FTC. 2012. Jon Leibowitz’s Reply to Letter from Representative Markey, Chairman, US Federal Trade Commission, October 25. Genesove, D., and W. Mullin. 2001. Rules, Communication and Collusion: Narrative Evidence from the Sugar Institute Case. American Economic Review 91: 379–98. Green, E., and R. Porter. 1984. Noncooperative Collusion under Imperfect Price Information. Econometrica 52: 87–100. Harrington, J. 2006. Behavioral Screening and the Detection of Cartels. European Competition Law and Policy Workshop/Proceedings: 1–17. Harrington, J. 2008. Detecting Cartels. In P. Buccirossi, Handbook of Antitrust Economics, : 213– 58. Cambridge, MA: MIT Press. Harrington, J., and J. Chen. 2006. Cartel Pricing Dynamics with Cost Variability and Endogenous Buyer Detection. International Journal of Industrial Organization 24: 1185–212. Jiménez, J. L., and J. Perdiguero. 2011. Could Transport Costs Be Lower? The Use of a Variance Screen to Evaluate Competition in the Petrol Market in Spain. International Journal of Transport Economics 38(3): 265–84. Keane, M. 2010. Structural vs. Atheoretical Approaches to Econometrics. Journal of Econometrics 156: 3–20. Kovacic, W., R. Marshall, L. Marx, and H. White. 2011. Plus Factors and Agreement in Antitrust Law. Working paper. Lorenz, C. 2008. Screening Markets for Cartel Detection:  Collusive Markers in the CFD Cartel-Audit. European Journal of Law and Economics 26(2): 213–32. Markey, Edward J. 2012. Letter to the Hon. Jon Leibowitz, Chairman, US Federal Trade Commission, October 12. Marshall, R., L. Marx, and M. Raiff. 2008. Cartel Price Announcements: The Vitamins Industry. International Journal of Industrial Organization 26: 762–802. Morgenson, G. 1993. Fun and Games on NASDAQ, Forbes (16 August 1993). Muthusamy, K., C. McIntosh, Y. Bolotova, and P. Patterson. 2008. Price Volatility of Idaho Fresh Potatoes: 1987–2007, American Journal of Potato Research 86(6): 438–444. Parker, R. 1969. The Banking Industry. Antitrust Law and Economics Review 2: 111–22. Porter, R., and D. Zona. 1993. Detection of Bid Rigging in Procurement Auctions. Journal of Political Economy 101: 518–38.

Screening for Collusion as a Problem of Inference   553

Porter, R., and D. Zona. 1999. Ohio School Milk Markets: An Analysis of Bidding. Rand Journal of Economics 30: 263–88. Rotemberg, J., and G. Saloner. 1986. A Supergame-Theoretic Model of Price Wars during Booms. American Economic Review 76: 390–407. Shor, M., and H, Chen. 2009. Contemporary Accounting Research. 26(2): 581–604. Stigler, G. 1964. A Theory of Oligopoly. Journal of Political Economy 72(1): 44–61. US Department of Justice, Antitrust Division (DOJ). N.d. Price Fixing, Bid Rigging, and Market Allocation Schemes: What They Are and What to Look For. Available at http://www.justice.​ gov/atr/public/guidelines/211578.pdf. Werden, G., L. Froeb, and D. Scheffman. 2004. A Daubert Discipline for Merger Simulation. Antitrust 18(3): 89–95. Yde, P. and M. Vita. 1996. Merger Efficiencies:  Reconsidering the “Passing-On” Requirement. Antitrust Law Journal 64: 735–47.

CHAPTER 22

C OM P E T I T IO N P OL IC Y F OR I N D U S T RY S TA N DA R D S RICHARD GILBERT

22.1. Introduction Standards are all around us. Screws fit into nuts thanks to standards. There are standards for weights and measures, professional certifications, time zones, money, wireless communications, plumbing fixtures, electrical components, and green buildings, to name just a few of the very many categories of economic activity impacted by standards. Despite the vast number of standards that impact commerce and daily life, standards often impress by their absence. Fire engines responding to the 1991 firestorm in Oakland, California, carried hoses that were incompatible with the local fire hydrants (Seek and Evans, 2004). The charger that refreshes my mobile phone is unlikely to work with yours, my word processor may not be fully compatible with your word processor (or with an earlier version of my own word processor), and if we travel abroad we should bring adapter plugs—lots of them. This chapter surveys competition issues raised by the development of industry standards, whether accomplished through a formal standard-setting committee structure or the unilateral conduct of a single sponsor. The focus is on the trade-off between the benefits from standards and possible costs that standards and the activity of standard development may impose on consumers. A  particular focus is on the consequences of intellectual property rights for standards. The topics are large and this chapter does not attempt a comprehensive survey, but rather complements several other surveys of standards and their economic impacts, examples of which include Farrell and Saloner (1987), David and Greenstein (1990), Katz and Shapiro

Competition Policy for Industry Standards   555

(1994), Matutes and Regibeau (1996), Shapiro (2001), Lemley (2002), ABA (2005), and Simcoe (2006).

22.1.1.  Types of Standards Standards have different functions. Compatibility standards (also called interoperability standards) are specifications that assure that one component will function as intended with another component. For example, the Small Computer Systems Interface (SCSI) specifies how small and medium-size computer processors communicate with external devices such as disk storage devices, printers, and scanners. Interoperability standards have mushroomed with the need for computers and telecommunications devices to function smoothly with complementary components and with advances in communications between devices and over the Internet. Biddle, White, and Woods (2010) concluded that a modern laptop computer has components that implement at least 251 different interoperability standards. The sources of these standards vary with 90 developed by formal standard development organizations, 112 developed in less formal consortia, and the remainder sponsored by individual companies. Quality and safety standards inform consumers about important product characteristics and may limit products that may be sold. The green building certification informs commercial tenants about the building’s energy performance and environment impacts. The classification scheme established by the Gemological Institute of America informs consumers about a diamond’s color, clarity, imperfections, and weight. The State of California requires petroleum refiners to produce and sell gasoline that meets the California reformulated gasoline emissions standard. Many standards help to coordinate economic activity. Interoperability standards have an explicit coordination function by ensuring that components work together. Standardized time and geographic time zones coordinate activity by allowing people to agree about the time of day. Currency standards, such as the $20 bill, help to coordinate economic exchange.

22.2.  The Development of Standards The International Organization for Standardization defines a standard as a “document, established by consensus and approved by a recognized body, that provides, for common and repeated use, rules, guidelines or characteristics for activities or their results, aimed at the achievement of the optimum degree of order in a given context.”1 A broader 1  ISO/IEC Directives Part 2, Rules for the Structure and Drafting of International Standards, Edition 6.0 2011-04.

556   Richard Gilbert definition embraces product or interface specifications developed through coordination or by independent action. Coordination can occur in a standard development organization (SDO) that holds regular meetings and follows defined procedures to specify and promulgate standards. Less formal organizations called standard-setting organizations (SSOs) also develop standards. As the line between SDOs and SSOs is often blurred, I use the term SSO to refer to both formal and less formal organizations that specify and promulgate standards. SSOs can be agencies of a government, organizations that have some affiliation with a government, or private. Compliance with a standard may be mandatory, meaning that parties engaged in commercial activity covered by the standard must employ the standard. Examples of mandatory standards include analog and digital broadcast standards, fuel emission standards, state and local building codes, and reflectivity standards for highway signs. These are de jure standards, although that term is sometimes applied to any standard developed by a formal standard development organization. More commonly, compliance with a standard is voluntary. Private parties, acting individually, often develop specifications or protocols that are widely adopted and become de facto industry standards. Examples include Microsoft’s Office Open XML file format or Hewlett-Packard’s PCL protocol for computer printers. De facto standards sometimes become adopted as formal standards. The Portable Document Format developed by Adobe is an example. Another dimension of standards and the standard development process is the degree to which the standard or the process is “open” (Krechmer, 2009). Standard development is open if the process is transparent, participation is available to any interested party, and consensus determines the choice of a standard. The standard itself is open if it is freely available and unencumbered by intellectual property rights such as patents, copyrights, and trademarks, although, according to some a standard can be open and protected by intellectual property rights if those rights are licensed to any interested party at reasonable and non-discriminatory terms.

22.2.1.  International Standard-Setting Organizations Three organizations headquartered in Geneva, Switzerland, operate at an intergovernmental level and develop standards intended for international application. ISO (the International Organization for Standardization) is a nongovernmental organization that is the world’s largest developer and publisher of international standards. The International Telecommunications Union (ITU) is an agency of the United Nations responsible for standards development in information and communication technologies. The third Geneva-based agency that develops international standards is the International Electrotechnical Commission (IEC), which provides a forum for companies, industries, and governments to develop international standards for electrical, electronic, and related technologies.

Competition Policy for Industry Standards   557

22.2.2.  National and Regional Standard-Setting Organizations Several organizations set standards for adoption at a regional level. In the European Union, the European Telecommunications Standards Institute (ETSI) is responsible for standardization of information and communication technologies within Europe and is the European counterpart to the ITU. Its accomplishments include the GSM technology for mobile communications. Nearly every country has a recognized standards body that acts as the country’s representative to the ISO. In the United States that body is the American National Standards Institute (ANSI). ISO representatives from other countries include the British Standards Institution, the German Institute for Standardization, the French Association for Standardization, the Standardization Administration of China, the Japanese Industrial Standards Committee, and more than 100 others.

22.2.3.  ANSI-Approved and Other US SSOs The American National Standards Institute (ANSI) coordinates the US voluntary consensus standards system by providing a forum for the development of policies on standards issues. ANSI also accredits organizations that meet its requirements for openness, balance, consensus, and due process in the development of voluntary national standards. ANSI’s members include more than 1,000 companies, government agencies, institutions, and international organizations. At the end of 2006 about 200 organizations were accredited by ANSI for standards development. The total number of American National Standards at that time exceeded 10,000. Numerous other organizations engage in standards development but are not formally accredited by ANSI either because they have not sought accreditation or because ANSI has not verified that the organizations conform to its requirements for openness, balance, consensus, and due process. Nonetheless, these organizations may develop and publish standards that obtain ANSI certification and may obtain certification by one of the international standardization organizations. These organizations include influential standard-setting consortia such as the Internet Engineering Task Force and the World Wide Web Consortium.

22.3.  Competition Principles for Standard-Setting Organizations Standards benefit consumers in many ways. Quality and safety standards provide valuable information about product characteristics and can prevent the sale of

558   Richard Gilbert shoddy or potentially hazardous products. Compatibility standards benefit consumers and producers by facilitating product adoption and promoting compatible solutions. Furthermore, cooperation to choose a standard can avoid or mitigate a costly “standards war” in which parties promote incompatible solutions. Competition to adopt a standard can be particularly intense and costly in markets characterized by strong network effects that support “winner-takes-all” or “winner-takes-most” market outcomes (Besen and Farrell, 1994, Shapiro and Varian, 1999). Familiar examples of rival standards in markets with network effects include the Betamax and VHS video tape formats, competing DVD formats (Dranove and Gandal, 2003), Blu-ray and HD high-definition recording media, and WiMax versus LTE and other fourth-generation mobile communications standards. Some standards wars splinter the market to such an extent that no winner emerges. An example is quadraphonic sound (Postrel, 1990). Standards can impose costs on consumers by reducing the diversity of products on the market, by restricting entry of suppliers, and by foreclosing or raising barriers to alternative implementations. Some exclusion is inevitable because the procompetitive purpose of standards is to focus industry efforts on one solution at the expense of others or to impose conditions that all suppliers must satisfy. Requiring surgeons to be board certified after years of training reduces the likelihood of being practiced upon by a charlatan, but the process of setting a quality standard can be manipulated to restrict the supply of surgeons in order to increase their compensation. Furthermore, a certification requirement that applies to all surgeons may be too restrictive for some types of medical procedures that can be performed with lesser skill sets. Standard development can provide a venue for competitors to act jointly to exclude rivals or fix prices. For example, practicing members of occupations can lobby legislatures for standards that limit competition from unlicensed suppliers, and many service occupations have succeeded in doing so (Kleiner, 2000; Edlin and Haw, forthcoming). At a more subtle level, a standard can facilitate collusion by making the supply of products more homogeneous, thereby making it easier for firms to monitor their rivals’ prices. Product homogeneity from standardization also can soften price competition by discouraging firms from offering competitive alternatives that depart from the standard. A standard that requires refrigerator manufacturers to meet or exceed an efficiency standard can increase costs, raise barriers to entry, make it easier for manufacturers to collude on higher prices, and prevent competition from cheaper units that do not meet the efficiency standard. Of course any adverse effects would have to be balanced against the benefits from lower energy usage. The timing of standard development also can have social costs if the process moves too slowly and delays procompetitive solutions or if the process moves too quickly and prematurely focuses market activities on an inferior technology, such as occurred in Japan for high-definition television (Farrell and Shapiro, 1992).

Competition Policy for Industry Standards   559

22.3.1.  Standard Setting as a Venue for Price Fixing and Exclusion Standard setting is likely to encounter antitrust concerns if the practice directly limits price competition. For example, in National Society of Professional Engineers v. United States, a professional society’s canons of ethics prohibited members from submitting competitive bids for engineering services. The Supreme Court affirmed an injunction prohibiting the association from adopting the rules or similar guidelines.2 The US Federal Trade Commission challenged an agreement by the National Macaroni Manufacturer’s Association concerning the relative proportions of wheat grains used in macaroni products, which the Commission argued was intended to moderate price competition for wheat grains that were in short supply. A court of appeals held that the actions of the Association constituted a combination to unlawfully fix prices in violation of Section 5 of the Federal Trade Commission Act. Several cases have alleged anticompetitive exclusion of rival suppliers or technologies by members of an SSO. In Wilk v. American Medical Ass’n, the plaintiff challenged a resolution adopted by the American Medical Association that labeled chiropractors “an unscientific cult” and implicitly made it unethical for a physician to associate with a chiropractor. The court required the AMA to retract this resolution, finding that the AMA failed to show that it was justified by a concern for patients. In TruePosition Inc. v. LM Ericsson Telephone Company, et al., the plaintiff alleged that defendants conspired to favor their own technologies and to block the adoption of TruePosition’s technology for enabling emergency responders to more accurately locate mobile phone users into new standards being developed for fourth-generation wireless networks. The defendants petitioned the court to dismiss the case, but the court refused, allowing the case to move forward.

22.3.2.  Standard Setting as a Joint Venture An organization that brings representatives of independent firms together to develop a standard is a joint venture. Joint ventures can create benefits, but also can impose costs on consumers. The participants in the SSO often include competitors and if the standard is successful, it is likely to have an effect on competition and prices. Firms that employ the standard in their products and the consumers that buy them are likely to benefit, while firms and consumers that are stranded with incompatible products may suffer. 2 

Case citations are listed at the end of this chapter. The discussion of antitrust law in this chapter is limited to US case decisions and is intended for illustration and not as a survey of the relevant law.

560   Richard Gilbert With few exceptions, antitrust authorities have accepted the choices determined by standard-setting organizations if standards were developed following an open and transparent process, participation in standard setting was unrestricted and non-discriminatory, and compliance with the standard is voluntary. Yet unlike the outcome of market competition that generally benefits consumers by resulting in lower prices, there is no corresponding “invisible hand” principle that standard-setting bodies will coordinate on the best standard, even if the process of standard setting is open and transparent. To illustrate the problem, suppose there are two alternative standards X and Y and three types of SSO participants affected by the standard, labeled A, B, and C. For example, A, B, and C may represent producers of different products that implement the standard. To keep the numbers simple, suppose each type has one vote in a standard-setting organization. Table 22.1a shows the value to each SSO participant from standard X and standard Y. Table 22.1a  Values That SSO Participants Place on Alternative Technology Choices Technology Agent

X

Y

A B C

5 3 4

6 2 3

In table 22.1a, technology X has greater total value. It has a total value of 12 and technology Y has a total value of 11. Agents B and C would vote for technology X, and it will prevail if the SSO follows a majority (or a two-thirds majority) voting rule. Table 22.1b differs from table 22.1a only in that agent A places a value of 10 on technology Y, rather than 6. As a result technology Y is the superior choice because it has a total value of 15, compared to 12 for technology X. However, technology X will prevail if the SSO follows a majority (or a two-thirds majority) voting rule.

Table 22.1b  Values That SSO Participants Place on Alternative Technology Choices Technology Agent

X

Y

A B C

5 3 4

10 2 3

Competition Policy for Industry Standards   561

The point of this simple illustration is that due process with open representation does not ensure that standard-setting organizations will make choices that maximize total welfare. Arrow (1950) established this result as a theorem that voting rules generally cannot assure outcomes that satisfy reasonable conditions for rational choice. This does not imply that consensus rules cannot be efficient. If the members of a standard-setting organization fairly represent the interests of all the parties that may benefit from the standard, the participants in the SSO may have opportunities to steer votes toward socially desirable outcomes. In the example corresponding to table 22.1b, agent A may offer two units of value to agent B (or C) in return for a vote in favor of technology Y. But the opportunities for such trades may be limited. Standard-setting organizations develop and promulgate thousands of standards that constrain the desired business strategies of firms and the choices of consumers, including firms and consumers that do not participate in the development of the standards. Although courts have been sensitive to antitrust issues that may arise in the context of collaboratively set standards, they generally have focused on circumstances involving the manipulation of the standard-setting process or the improper use of the resulting standard to gain competitive advantage over rivals rather than on the merits of alternative standards. In this respect the courts’ actions are consistent with monopolization cases in which the focus is typically on practices that may interfere with competition to become dominant or on practices that maintain a dominant position in a market rather than on dominance itself. In U.S. v. Microsoft, the district and appellate courts did not dwell on whether a different personal computer operating system and browser platform would be superior to the Windows/Internet Explorer platform. Rather, the courts’ focus was whether Microsoft interfered with the competitive process to maintain its market dominance.

22.3.3.  Corruption of the Standard-Setting Process Courts have intervened when there is evidence that the standard-setting process has been manipulated to favor particular interests (Anton and Yao, 1995). ANSI accreditation requires that standards developers adhere to principles of consensus, due process, and openness.3 According to ANSI, any organization, company, government agency, or individual with a direct and material interest in the activities of the standards body should have a right to participate in the development and selection of the standard. Consensus does not mean unanimity, but rather an absence of sustained opposition to substantial issues (Farrell and Simcoe, 2012).

22.3.3.1.  Vote Stacking A potential abuse of the standard-setting process involves “vote stacking,” in which members of a standards body recruit participants to vote affirmatively for their interests. 3  ANSI Standards Activities Overview, available at http://www.ansi.org/standards_activities/​ overview/overview.aspx?menuid=3, accessed May 3, 2012.

562   Richard Gilbert Allied Tube & Conduit Corp. v. Indian Head concerned approval by the National Fire Prevention Association (NFPA) of a standard for the use of plastic conduit for electrical wiring. The NFPA publishes the National Electrical Code, which establishes requirements for the design and installation of electrical wiring systems. Indian Head, a manufacturer of plastic conduit, submitted a proposal to extend the Code to approve the use of plastic as well as the conventional steel conduit. The proposal was approved by one of the NFPA’s professional panels, and under the Association rules could be adopted into the Code by a simple majority of the members attending the Association’s annual meeting. Before the meeting was held, Allied, the largest producer of steel conduit in the United States, members of the steel industry, other steel conduit manufacturers, and independent sales agents collectively agreed to recruit new Association members whose only function was to vote against the proposal to approve the use of plastic conduit. Their recruiting efforts were successful and the proposal was defeated. In response Indian Head sued, alleging that Allied and others had unreasonably restrained trade in the electrical conduit market in violation of Section 1 of the Sherman Act. A jury delivered a verdict that the conduct violated the antitrust laws. The case subsequently endured reversals and appeals, largely focusing on whether the conduct was permissible activity to influence legislation, because the National Electrical Code was routinely adopted into law by a substantial number of state and local governments. The Supreme Court ultimately sustained the jury verdict and agreed that the proper standard was whether the anticompetitive effects of Allied’s actions outweighed any procompetitive benefits of standard setting. The Court stated that “What [Allied] may not do (without exposing itself to possible antitrust liability for direct injuries) is bias the process by, as in this case, stacking the private standard-setting body with decision-makers sharing their economic interest in restraining competition.” The Court, however, did not define “stacking,” nor did it explain why vote stacking should not survive a rule-of-reason analysis. While the recruiting of participants in a standard-setting organization with instructions to support or oppose proposed standards appears improper on its face, participation in standard-setting meetings is driven by economic interests, and firms with more at stake in the standards process tend to supply more participants to relevant standard-setting organizations (Simcoe, 2012). Vote stacking could be defined as recruiting participants who are not employees of interested firms, but that distinction has little substance for large firms that can tap a large labor pool as potential representatives. Furthermore, it is not clear without further analysis that vote stacking results in inferior economic outcomes compared to more limited representation in a standard-setting organization. If vote stacking is defined by recruiting more participants than are necessary to evaluate the merits of proposed standards, this activity clearly incurs social costs that include the recruiting efforts and the opportunity costs of the recruited participants. It is likely that firms will incur recruiting costs in proportion to the private benefits they expect from their preferred standard relative to alternatives. A firm that expects to benefit from a particular standard has an incentive to incur costs to achieve that standard.

Competition Policy for Industry Standards   563

Moreover, private values more generally motivate efforts to influence the adoption of standards, whether or not those efforts may include packing standard-setting committees with friendly voters. Despite its costs, vote stacking could have plausible benefits compared to a majority or supermajority voting rule, which does not necessarily select the best standard. Vote stacking is similar to what is called an “all-pay auction.” In an all-pay auction the bidders pay the amounts they bid regardless of whether their bids are successful. Examples of conduct with characteristics of an all-pay auction include election campaigns, political lobbying, and research and development competition for a patent or for a technology with strong network effects. There is a single victor in these examples but all participants who compete to be the winner incur irreversible costs. In the vote stacking analogy the “bid” is the cost of recruiting participants in the standard-setting organization with instructions to vote for or against a proposal on behalf of a party with interests in the outcome. To further evaluate the costs and potential merits of vote-stacking as an all-pay auction, suppose two firms participate in the standard-setting process and choose to support either of two technology standards, denoted by X and y . Let X1 be the value of standard X to firm 1 and X2 the value of standard X to firm 2. Standard Y has value Y1 to firm 1 and Y2 to firm 2. Firm 1 prefers standard X if X1 > Y1 and firm 2 prefers standard y if Y2 > X 2 . Let V1 = X1 − Y1 and V2 = Y2 − X 2 and suppose both V1 and V2 are positive, so that firm 1 strictly prefers standard X and firm 2 strictly prefers standard Y. Furthermore, assume that V1 > V2 , which implies that standard X has a higher total value to both firms than does standard Y (that is , X1 + X 2 > Y1 + Y2 ). Firm 1 is willing to pay up to V1 to get its preferred standard and firm 2 is willing to pay up to V2 for its preferred standard. These are the private values of standards to each participant and not their social values. Nonetheless, vote stacking might result in desirable outcomes if the private values of alternative standards are proportional to their social values.4 Further analysis, however, shows that even with this favorable assumption, vote stacking does not necessarily result in the certain selection of the best standard. To see this, suppose that a simple majority vote is sufficient to adopt a standard, a firm’s costs increase with the number of recruited voters, both firms incur costs to recruit voters, and firm 1 recruits more voters than firm 2. This cannot be a profit-maximizing outcome for firm 2. Firm 2 will not prevail in the contest for the desired standard, so it may as well cut its losses by not recruiting any voters. But if firm 2 does not recruit voters, then firm 1 can save by recruiting fewer voters. However, if firm 1 recruits fewer voters, firm 2 might then increase its recruiting efforts to prevail in the standard selection. This circular argument shows that there is no “pure strategy equilibrium” of this competition in which each firm knows how many voters will participate in the standard-setting process. A profit-maximizing equilibrium exists, however, if the firms play a “mixed strategy” in which the firms recruit a random number of voters. 4  That would be the case if the ratio of producer to consumer surplus is the same for the alternative standards.

564   Richard Gilbert An implication of random recruiting is that each firm has some positive probability of obtaining its desired standard, for otherwise it would not incur the cost of recruiting voters. Hillman and Riley (1989) analyze an all-pay auction that is similar to this example. They consider an all-pay auction with one item for sale and two bidders. Bidder 1 has a value of V1 for the item and bidder 2 has a value of V2 . They show that when the bidders use equilibrium bid strategies in which they bid amounts with different probabilities, if V1 > V2 , the probability that bidder 1 wins the auction is: 1 − (1 / 2)[V2 / V1 ]. Applying this result to the choice of a standard, the “item” for sale is the right to choose the technology standard, which has value V1 to one of the participants in the standard-setting activity and value V2 to the other participant. The bid is the effort to pack the standard-setting committee to vote for the preferred standard. Suppose that V2 is only half as large as V1 . Following the bid strategies in Hillman and Riley (1989), the probability that vote stacking will result in a win for the technology with the higher value is only 75%; there is a 25% probability that the inferior technology will emerge as the standard. This analysis shows that vote stacking incurs social costs and does not necessarily result in the selection of a standard with the greatest private or social benefit relative to alternatives. The analysis is only an example. Modifying the voting rule for adopting a standard from a simple majority to a supermajority introduces complications because it is possible that no technology wins a sufficient plurality to gain adoption as a standard. The incentive to recruit randomly could disappear if technology values and the correspondence between votes and recruiting effort are uncertain; however, this does not ensure that socially desirable standards necessarily will emerge from a process that allows parties to recruit participants in standards committees to vote on their behalf. Despite the fact that the example in this section is highly simplified, it provides cause to doubt whether allowing parties in a SSO to recruit members on their behalf would lead to better technology adoption decisions than would occur with restrictions on vote stacking. Furthermore, packing standard-setting committees with disinterested participants can frustrate negotiations that could lead to adoption of a socially desirable standard. These are reasons why the recruitment of voters in standard-setting committees should potentially fail a rule-of-reason test, as in Allied Tube v. Indian Head. Of course standard selection committees do not necessarily converge on the standard with the greatest social benefit even if interested parties do not pack committees with friendly voters. Furthermore, it is difficult to know how to draw the line between vote stacking and the supply to a standard selection committee of numerous participants that represent a party’s private interests.

22.3.3.2.  Disparagement of Alternative Proposals Another example of alleged subversion of the standard-setting process is in American Society of Mechanical Engineers v. Hydrolevel Corp. The plaintiff Hydrolevel was a manufacturer of low-water cutoff valves for boilers. Hydrolevel alleged that a member of the American Society of Mechanical Engineers (ASME) subcommittee employed by a competitor improperly induced the subcommittee chairman to issue a letter indicating that Hydrolevel’s product did not meet the ASME’s safety standards. A jury found

Competition Policy for Industry Standards   565

that “ASME has conspired with others to restrain trade unreasonably . . . by disparaging Hydrolevel’s cut-off through a misrepresentation of the Boiler and Pressure Vessel Code.” The Supreme Court affirmed the verdict, noting in its opinion the potential for competitors to abuse the process of standard setting in order to gain improper advantage in the marketplace. As in Allied Tube, here a competitor expended effort to influence the outcome of the standard-setting process by means unrelated to the merits of the alternatives. In this case the effort was lobbying to obtain a false opinion about the characteristics of the rival product.

22.4.  Competition Principles for Unilaterally Determined Standards Standards that are sponsored by a single entity sometimes achieve broad market acceptance without the imprimatur of a standard-setting organization. Examples include the Intel x86 microprocessor architecture, Microsoft’s Windows application programming interfaces, Apple Quicktime, and Adobe Flash. These standards are typically proprietary, meaning that the firm that sponsors the standard (and possibly others) owns intellectual property rights that cover use of the standard. Proprietary standards may be licensed separately or bundled with a product. Firms often have strong incentives to promote the use of a standard, particularly in markets with network effects that reinforce demand. Sponsors of alternative standards may compete to become the dominant standard by subsidizing sales in the short run or by engaging in activities such as bundling a product employing the standard for free with other products that increase the opportunity to gain market acceptance for the standard. Sometimes firms are successful in promoting a standard and enjoying the benefits of network effects while maintaining ownership of the standard. Other times firms choose to relax control of a proprietary standard, trading off the advantage of control for broader market acceptance, despite the increase in competition that may result if the standard is not proprietary. Adobe licensed its Portable Document Format standard without charge before it was released as an open standard published by the ISO. Cisco licenses its Telepresence Interoperability Protocol for videoconferencing on a royalty-free basis. Several cases have alleged unlawful unilateral conduct among sponsors of alternative standards in the competition to become a dominant standard, although assessing when such competition represents competition on the merits and when it crosses the line to become exclusionary is often difficult. In U.S. v. Microsoft, the complaint initially included allegations that Microsoft distributed its Internet Explorer browser for free, but the Department of Justice did not pursue a predatory pricing theory at trial and the district court opinion noted that Microsoft’s free distribution of Internet Explorer benefited consumers. Relatedly, the Department has not challenged premature announcements

566   Richard Gilbert of new products in an attempt to capitalize on network effects by attracting a base of customers through a bandwagon effect or in an attempt to maintain an installed base of customers by encouraging customers to forestall purchases of competing products (Farrell and Saloner, 1986). In the contest to promote a desired standard a firm may take actions that disadvantage a competing standard. Conduct by a firm with monopoly power that excludes rivals by foreclosing access to efficient distribution channels or inputs can raise antitrust concerns if there are no offsetting benefits. The complaint in U.S. v. Microsoft is an example of such an allegation, focusing on conduct by Microsoft to deny distribution opportunities to Netscape’s Navigator web browser. The Netscape browser did not compete with the Microsoft Windows operating system, and indeed was a complement to the operating system and increased its value. However, the complaint alleged an “applications barrier to entry” to competitors of the Windows platform that could be surmounted by providing Netscape Navigator as an alternative platform for applications. By foreclosing access to Netscape’s browser, Microsoft sought to maintain the applications barrier to entry for the development of competing platforms. When a firm controls a compatibility standard, the firm may profit by promoting the supply of interoperable products and services by other parties in some situations, while in other situations the firm may profit by discouraging the supply of competing interoperable products and services. Interoperability embraces a wide range of activities including physical compatibility, software interfaces, digital rights management (such as Apple’s Fairplay technology for iTunes), and communications protocols, and is a matter of degree. A firm that controls a component of a system can benefit from competition that lowers the prices and increases the diversity of complementary components, which makes the system more attractive to consumers and increases their willingness-to-pay for the component. However, the owner of a component that is necessary to achieve interoperability also may choose to limit intraplatform or interplatform compatibility to pursue a strategy of price discrimination or to raise barriers to new competition (Farrell, 2007). A firm can use control over a complementary component such as replacement printer cartridges to meter demand and charge higher prices to more intensive users. The Microsoft case illustrated the control of compatibility to raise barriers to entry. Sponsors of compatibility standards can control interoperability through trade secrets, access to copyrights and patents, contractual restrictions, and frequent design changes (Samuelson, 2009). While antitrust policy frowns on conduct by a firm with monopoly power that excludes rivals with no offsetting benefits, a firm generally has no obligation to promote its rivals’ products or a rival standard (Garza et al., 2007). The economic welfare effects of interoperability are complex and depend on particular market circumstances. Interoperability often promotes consumer welfare by increasing the diversity of complementary products and by encouraging competition among suppliers of those products. This can promote a virtual cycle that results in lower costs and greater consumer benefits if there are scale economies or positive network effects. However, compatibility also can incur costs by limiting technological alternatives and can impede competition in some circumstances.

Competition Policy for Industry Standards   567

Sellers of incompatible systems can have greater incentives to innovate than sellers of compatible components. A firm that sells a system comprised of several components has greater sales to amortize research and development expenditures compared to a firm that sells only an individual component (Farrell and Weiser, 2004). Network effects reinforce the expectation of a high reward because a single product or standard is likely to dominate a market when network effects are large. The expectation of a high reward motivates firms to compete aggressively to become dominant. While consumers may pay a price for market dominance, they also benefit from the intense competition that occurs in the battle for dominance (Shapiro and Varian, 1999, Farrell and Klemperer, 2007). When markets can support different incompatible systems, competition among those systems can benefit consumers relative to a market with compatible components. When a firm sells a system whose components are incompatible with the components of other systems, the sale of another unit earns a margin on all of the firm’s system components. This margin is an inducement to cut prices to win the sale. In contrast, when a firm sells a compatible component, some of the benefit of a price cut is lost to other sellers of compatible components by increasing the demand for those products (Matutes and Regibeau, 1988). When firms sell incompatible systems, the increased incentive for price competition benefits consumers, although compatibility also benefits consumers by increasing the diversity of available products and by facilitating competition among independent suppliers of complementary components. Thus there is no simple conclusion that “compatibility” is always superior to “incompatibility.”

22.5.  Intellectual Property The focus of this section is on patents that are essential to comply with a standard, although much of the discussion also applies to other intellectual property rights such as copyrights and trademarks. Standard-setting organizations differ in their policies with respect to developing standards that require access to proprietary intellectual property rights. The American National Standards Institute patent policy states, “There is no objection in principle to drafting an American National Standard (ANS) in terms that include the use of an essential patent claim (one whose use would be required for compliance with that standard) if it is considered that technical reasons justify this approach” (ANSI, 2008). Other standard-setting organizations, such as the World Wide Web Consortium, are less receptive to proprietary technologies. With some exceptions, the World Wide Web Consortium’s patent policy requires that licenses for essential patents be made available on a royalty-free basis. When firms own intellectual property that is necessary to employ a standard, there is a potential tension between what Simcoe (2006) calls value creation and value capture. The cost of proprietary intellectual property rights limits the adoption and use of a standard and interferes with the potential value that the standard can create. Owners

568   Richard Gilbert of intellectual property rights may accept this limitation if those rights enable the owners to appropriate a greater share of the value created by the standard. This trade-off, however, omits an additional tension, which is that intellectual property rights provide incentives for innovation and for innovators to participate in the development of standards.

22.5.1.  The Potential for Patent Holdup Firms and consumers often make investments that are specific to the use of a particular standard, such as applications for the Windows operating system, interface designs for computer memories, or a library of DVDs. Switching to a different technology, such as a different computer operating system or different optical recording medium, may require new investments. Firms and consumers are “locked in” to a standard if it is not economically feasible to incur the costs of switching to a new standard. Even if irreversible investment costs are not large, the difficulty of coordinating substitution to an alternative technology can pose a barrier to switching, particularly when technologies exhibit networks effects, as is often the case for interoperability standards. When firms and consumers are locked into the standard, owners of proprietary rights that are essential to comply with the standard can have the economic power to charge high royalties and impose burdensome licensing terms. The term “patent holdup” is sometimes used to characterize attempts to charge such high royalties or impose burdensome terms “ex post,” after firms and consumers have made investments that are specific to a standard. If participants in a standard-setting organization are aware of the existence of proprietary rights that cover a proposed standard and the likely royalties that may be charged for those rights “ex ante,” before firms and consumers make investments that are specific to a proposed standard, they can make informed decisions that account for the risk of holdup. A minimum requirement for informed decision-making is disclosure of patents that are essential to make or use products that comply with a standard. The term “patent ambush” is sometimes used to describe a high royalty demand after a participant in a standard-setting organization fails to disclose to the SSO’s members the existence of the patent or patent application when the standard was first established. Standard-setting organizations differ in their disclosure policies (Lemley, 2002, Bekkers and Updegrove, 2012), and competition authorities have been reluctant to impose rigid disclosure rules. An obligation to identify individual patents and patent applications that are essential to a standard is complex and expensive, particularly for technologies in a crowded patent space. MPEG LA lists more than 900 essential patents and 150 essential patent families for the MPEG-2 standard.5 Participants in the ETSI standard-setting organization declared more than 750 US patents as essential 5 MPEG LA, MPEG-2 Patent Portfolio License Briefing, August 4, 2010, at 4. A patent family is a set of patents filed in different countries for a particular invention.

Competition Policy for Industry Standards   569

to W-CDMA, a third-generation mobile telecommunications standard (Bekkers, Bongard, and Nuvolari, 2011), and more than 500 unique patent families as essential to the fourth-generation LTE cellular standard (Blind et al., 2011). Note that a declaration of an essential patent is not a confirmation that the patent is actually essential to a standard. Furthermore, a disclosure requirement for planned patent applications may force firms to expose valuable trade secrets, potentially jeopardizing future patent protection by contributing inventions to the common pool of knowledge, and may discourage firms from participating in the standard-setting process. An alternative to mandatory disclosure is a blanket commitment in which SSO participants agree to license any essential claims at reasonable terms without disclosing patents in advance. While a default commitment to license any standard-essential patent at RAND (reasonable and non-discriminatory) terms may address the problem of patent ambush, it leaves uncertain whether particular patent claims cover a particular proposal. This is less of a concern when SSOs, such as the World Wide Web Consortium, require royalty-free licensing commitments.

22.5.2.  Reasonable and Non-discriminatory Licensing Commitments If a patented technology is essential to comply with a standard, most standard-setting organizations, and all ANSI-accredited standard development organizations, require the patent owner to make a license available for the purpose of implementing the standard either royalty-free or under reasonable terms that are free of any unfair discrimination. The latter condition is called a RAND (reasonable and non-discriminatory) or FRAND (fair, reasonable and non-discriminatory) licensing obligation, and the former is called RF (for royalty-free) or F/RAND-z, for zero royalty.6 The policies of standard-setting organizations differ on many issues, such as whether standard-essential patents (SEPs) are limited to patents with claims that are technically necessary to comply with a standard or may include patents that are commercially necessary, the obligations of patent owners to disclose patents and patent applications (or other intellectual property) that they represent to be essential to a proposed standard, and whether licensing obligations remain in force if a member withdraws from the SSO. Moreover, SSO policies typically are silent on critical issues such as the specifics and contractual form of RAND licensing obligations, whether licensing obligations travel with the patent if they are sold or assigned to another entity, and the consequences for failing to abide by the SSO’s policies. As a consequence it has been the job of antitrust enforcement agencies and the courts to fill in the missing dimensions of RAND commitments. 6  I make no distinction between “RAND” and “FRAND.” Note that a royalty-free license is not the same as an agreement not to assert a patent right, because a license may be royalty-free, but may impose other terms and conditions.

570   Richard Gilbert

22.5.2.1.  RAND Royalties Competition authorities and a number of scholars have endorsed the principle that a “fair and reasonable” royalty should reflect its incremental value relative to the next-best alternative, assessed before firms and consumers make investments that are specific to the technology. The Federal Trade Commission (2011) concludes that “Courts should cap the royalty at the incremental value of the patented technology over alternatives available at the time the standard was chosen.” Swanson and Baumol (2005) suggest that standard-setting organizations should conduct an auction in which sponsors of patented technologies submit bids that describe their intended licensing terms and imply that bids will reflect technologies’ incremental values. While the incremental value principle has theoretical appeal (see, e.g., Farrell et al., 2007), it does not have general acceptance as the interpretation of fair and reasonable licensing terms (Patterson, 2002). Some commentators have argued that a licensing agreement is fair and reasonable if it is the outcome of arm’s-length bargaining, even if that bargaining occurs after firms and consumers have made investments that are specific to a standard (see, e.g., Geradin and Rato, 2007). Furthermore, no SSO has yet offered a workable definition of RAND licensing terms and, as of this writing, US courts have not converged on a particular methodology to assess when royalty terms fail to be fair or reasonable. The determination of a fair and reasonable royalty for a patent is particularly difficult when the patent is one of many that are essential to practice a standard, as is often the case for compatibility standards. Every essential patent has a potential claim to the entire value of the product for which it is essential, because by definition the product cannot be made, sold, or used without access to each essential technology. However, this yields a contradiction when there is more than one essential patent. Approaches such as assigning equal values to essential patents are arbitrary, as patents may differ in quality, in their values relative to alternative technologies that were available when the standard was being developed, the probability that the patent is valid, and in the extent to which they are essential for different products that comply with a standard.7 The participants in the development of standards include patent owners and implementers of patented technologies who often have different business models that affect their assessments of a reasonable royalty. Implementers are firms that sell products that embody the patented technologies. These firms may be content to earn revenues primarily from the sale of products rather than from patent licenses. Other firms, including those that specialize in technology innovation, have business models that depend on earning revenues from patent licensing. A technology implementer may be unaccustomed to charging or paying significant royalties, while a technology innovator may deem a royalty to be unreasonable if it does not provide an adequate return on investment. These contrasting perspectives cloud the meaning of “fair and reasonable” licensing terms and explain why most standard-setting organizations have demurred from

7  For illustrations of royalty allocations for complementary technologies under different assumptions, see Layne-Farrar, Padilla, and Schmalensee (2007) and Salant (2007).

Competition Policy for Industry Standards   571

offering specific definitions of RAND commitments in their intellectual property policy statements. Enforcement agencies and many commentators agree that RAND commitments should prevent a patent owner from “holding up” a standard by demanding high royalties after firms and consumers have made investments that are specific to the standard. Such holdup occurs if the costs to firms and consumers of switching to a different standard are large and royalties capture these costs. There is, however, an additional concern that patent owners may demand royalties that exceed any reasonable allocation of value to their patents. The result can be a total royalty “stack” that harms the adoption and use of products that employ the patents (Lemley and Shapiro, 2007), even if the cumulative royalties do not reach a level that may constitute holdup. If some patent owners are content to charge low royalties, this increases the net value available to other patent owners. Suppose there is a technology for implementing secure wireless communications that is worth $2 more than the next best alternative for every handset that employs the technology. Furthermore, suppose that ten patents are essential to make, sell, or use that technology, all are valid, and do not differ in quality or available alternatives. An equal allocation of the $2 value would allocate $0.20 to each patent, ignoring any other inputs that may be necessary to achieve the technology’s value. But suppose that five of the patent owners would be content to license their patents at a zero royalty. That leaves the full $2 value for allocation to the remaining five patents. With an equal allocation rule they could each charge a “reasonable” royalty of $0.40. When multiple patents are essential to use a standard and some technology implementers are content to charge zero or low royalties, this leaves more “headroom” for other patentees to charge relatively higher royalties. It is open question whether these higher royalties would be considered fair and reasonable. For a patent owner that is accustomed to licensing technology without charge or for less than its apportioned value of $0.20, allowing other owners of essential patents to charge more than $0.20 may appear unreasonable. Moreover, if all patent owners move to a business model in which they seek their share of the value of the patented technologies, the result can be a total royalty burden that is so large that it impedes the adoption and use of the standard.

22.5.2.2.  RAND and Cross-Licenses Firms often enter into cross-licensing arrangements of large patent portfolios (Grindley and Teece, 1997). These arrangements give the parties the ability to operate free of patent disputes in a broad area of economic activity. Parties may enter into cross-licensing arrangements that involve patents with RAND commitments. However, insistence on a cross-license as a condition for a license to a patent with a RAND commitment complicates the determination of the royalty in the transaction because the royalty depends on the values of the patents that are cross-licensed. For illustration, putting aside the ambiguities of RAND, suppose there is agreement that the RAND royalty for a standard-essential patent is $0.50 per unit. Instead of offering the patent for license at a royalty of $0.50 per unit, suppose the owner of the patent insists on a royalty-free cross-license to the licensee’s patents. If the licensee’s

572   Richard Gilbert patents are also worth $0.50 per unit, this is equivalent to a transaction in which the standard-essential patent is licensed at its RAND value of $0.50 per unit. But if the licensee’s patents are worth $1 per unit, the implicit royalty for the standard-essential patent is $1, which exceeds its RAND value. Allegations of cross-licensing demands that violate RAND commitments play at center stage in the smartphone patent wars. Motorola and several other suppliers of smartphones own patents that have been declared essential for communications employing cellular and wireless network standards such as the third-generation UMTS standard and the 802.11 Wi-Fi family of standards, which are subject to RAND licensing commitments. Apple owns patents on desirable smartphone features such as pinch-tozoom and text recognition. These technologies have not been declared essential to a standard and are not subject to RAND licensing commitments, and until recently Apple owned no standard-essential patents.8 Motorola accused Apple of violating its RAND-encumbered SEPs (Apple v. Motorola) and offered to license these patents to Apple, but at a royalty that Apple considered to be excessive.9 Alternatively, Apple represented that Motorola offered to cross-license its SEPs if Apple would enter into a cross-license to its smartphone patent portfolio.10 This offer would be a Hobson’s choice if the cross-license effectively values the RAND-encumbered patents at a royalty similar to or larger than Motorola’s initial royalty offer for its SEPs.

22.5.2.3.  RAND and Injunctions A patent gives its owner the right to exclude others from making, using, or selling the article protected by the patent, but that right is not unconditional, as the Supreme Court noted in its eBay v. MercExchange decision. The right to exclude is a powerful tool for a patent owner even if it is not exercised because the patent owner can use the threat of exclusion to negotiate more favorable licensing terms (Farrell and Shapiro, 2008). The exclusion threat is particularly powerful when firms and consumers make investments that are specific to the standard. The opportunities for a technology user to avoid infringement by designing around a patent are severely limited when the patented technology is essential to comply with a standard. If the patent covers a feature in the 802.11 Wi-Fi standard, eliminating that feature could make a product inoperable for wireless connectivity. The power of the exclusion threat is compounded when the patent covers one of many technologies that are essential to make or use a product (Gilbert, 2010). Returning

8  Apple acquired rights to numerous standard-essential patents through its participation in the acquisitions of patent portfolios previously owned by Novell and Nortel. 9 In Apple v. Samsung, the court concluded that Motorola did not establish that its royalty demand based on a portfolio value equal to 2.25% of Apple’s product revenues was consistent with the requirements of a RAND royalty. In a separate case involving different products, an administrative law judge concluded that Motorola’s offer to Microsoft of a 2.25% royalty “could not possibly have been accepted by Microsoft” (In the Matter of Certain Gaming and Entertainment Consoles). 10  The European Commission reported that “[A]‌ccording to Apple, Motorola Mobility has insisted that Apple cross-license its full non-SEP portfolio in exchange for Motorola Mobility’s SEPs.” GOOGLE/ MOTOROLA MOBILITY, Case No. COMP/M.6381, para. 89 (Feb. 2012).

Competition Policy for Industry Standards   573

to the example of ten patents that are essential for a technology that has a value of $2, a single patent owner can demand the full $2 if it can exclude others from its patent even if nine other patents are also essential to the technology.11 If a RAND commitment requires a patentee to forgo the threat of an injunction, this has profound implications on licensing negotiations because it puts courts in the position of determining a reasonable royalty instead of possibly granting an injunction if parties cannot agree to licensing terms. Arguably, the commitment to license is the RAND promise’s core function and meaning (Miller, 2007). It is not, however, settled that a RAND commitment forgoes the opportunity for a patentee to seek injunctive relief. Some argue that the RAND commitment is by definition an obligation to grant a license and therefore is plainly inconsistent with a demand for injunctive relief,12 while others maintain that an injunction is warranted if the patent owner makes a reasonable licensing offer that is refused. A middle ground limits injunctions to circumstances in which the owner of a patent with a RAND commitment has made a licensing offer that a court or agreed-upon arbitrator has validated as being reasonable, the offer was refused, and continued infringement of the patent would create injury that could not be compensated. Patent owners are testing the limits of injunctive relief for patents subject to RAND commitments by bringing infringement suits to the International Trade Commission. The ITC provides a venue to adjudicate infringement of intellectual property through the importation and/or sale of imported goods protected by US patents under 19 U.S.C. §1337 (commonly referred to as “Section 337”). The ITC has the power to prohibit imports from entering US borders (an exclusion order) or to prohibit parties from distributing or selling infringing articles from existing US inventories (a cease-and-desist order) (Yeh, 2012). The Commission has some flexibility to tailor the scope of the order and to delay its onset, but unlike the courts, the ITC lacks the authority to award monetary damages for patent infringement. Since many products are manufactured abroad, the ITC is often available as a venue for patent holders to challenge alleged infringers. In 2011 every major smartphone manufacturer was a party to a patent infringement action at the ITC (Chien and Lemley, 2012). As a form of injunctive relief, the threat of an exclusion order at the ITC may allow owners of standard-essential patents with RAND commitments to negotiate royalties that are inconsistent with those RAND commitments, because the infringer has little power to bargain if excluded by the ITC from the US market.13 In deciding whether to issue an exclusion or cease-and-desist order for patent infringement the Commission has scope to consider the public interest, which includes findings on the public health 11 In Apple v. Motorola, a witness for Motorola claimed that a single essential patent can command

40%–50% of the value of the entire portfolio, even if the portfolio includes 100 essential patents. The court, however, rejected this claim. 12  See the submission of Commissioner Rosch in Third Party United States Federal Trade Commission’s Statement On The Public Interest, In The Matter Of Certain Wireless Communication Devices, Portable Music And Data Processing Devices, Computers And Components Thereof, United States International Trade Commission, Inv. No. 337-TA-745. 13  See Third Party United States Federal Trade Commission’s Statement On The Public Interest, In The Matter Of Certain Gaming And Entertainment Consoles, Related Software, And Components Thereof,

574   Richard Gilbert and welfare, competitive conditions in the US economy, the production of like or directly competitive articles in the United States, and US consumers.14 The ability to hold up an industry by refusing to license standard-essential patents with RAND commitments at reasonable terms has important consequences for competitive conditions in the US economy and for US consumers. In past cases the Commission has only very rarely invoked these considerations to permit infringing imports (Chien and Lemley, 2012). The Commission has notified requests for statements about public interest impacts in its proceedings relating to patents with RAND licensing obligations; however, the extent to which the Commission will consider these effects has yet to be determined.

22.5.2.4.  Sales and Assignments of RAND-Encumbered Patents Another factor that is relevant to the RAND commitment is whether it binds the patent or only its owner. If it is the latter, the RAND commitment may be lost if the patent is sold or assigned to a party that has not made a commitment to license its patents on RAND terms. Patents are easily sold or assigned. Reneging on a RAND commitment does not fit neatly into traditional antitrust paradigms that focus on collusion or the acquisition or maintenance of monopoly power. Nonetheless, it is clear that a RAND commitment is toothless if the commitment does not remain with a patent that is sold or assigned to a different party. The US Federal Trade Commission alleged that Negotiated Data Solutions (N-data) reneged on a RAND commitment when it acquired rights from National Semiconductor for a patent that was declared essential to the IEEE’s Ethernet local area networking standard. The Commission challenged N-data’s conduct as an unfair method of competition and unfair acts or practices under Section 5 of the Federal Trade Commission Act and negotiated a consent order in which N-data agreed to comply with the RAND commitment.15

22.5.3.  Antitrust Concerns from Failure to Disclose Standard-Essential Patents The failure of a patent owner involved in standard setting to disclose the existence of a patent or patent application that covers a proposed standard may give rise to a monopolization claim under the Sherman Act or to a claim of unfair competition or practices under Section 5 of the Federal Trade Commission Act. Such “patent ambush” antitrust United States International Trade Commission, Inv. No. 337-TA-752; Submission of 19 Economics and Law Professors, In the Matter of Certain Wireless Communications Devices, Portable Music and Data Processing Devices, Computers, and Components Thereof, United States International Trade Commission, Inv. No. 337-TA-745; and Prepared Statement of The Federal Trade Commission before the US Senate Committee on the Judiciary Concerning “Oversight of the Impact on Competition of Exclusion Orders to Enforce Standard-Essential Patents,” Washington, DC, July 11, 2012. 14  15 

19 USC § 1337—Unfair practices in import trade. In re Negotiated Data Services LLC, FTC File No. 051 0094, Decision and Order (Jan. 23 2008).

Competition Policy for Industry Standards   575

claims have been brought by private parties and by government antitrust enforcement agencies. In Dell Computer Corp. the FTC complained that Dell falsely certified to the Video Electronics Standards Association (VESA) that it did not own patents that covered a standard for communicating information between a computer’s central processor and peripheral devices (a computer “bus”). Dell sought to enforce its patent rights after adoption of the VESA-approved bus. Dell and the FTC entered into a consent decree prohibiting Dell from enforcing its patent rights on the VESA bus design. Patents also were at stake in a complaint brought by the FTC against the Union Oil Company of California (Unocal), alleging failure by Unocal to disclose that it owned patent rights covering a proposed standard for reformulated gasoline that was subsequently adopted by the California Air Resources Board. Unocal agreed not to enforce its relevant patents as part of a consent order for its acquisition by Chevron.16 In a case with a long history involving government and private plaintiffs, the Federal Trade Commission charged Rambus with failing to disclose patent applications covering standards for dynamic random access memory devices (DRAMs) developed by the Joint Electron Device Engineering Council (JEDEC). Rambus participated in the JEDEC committee for DRAM standards for a period of time during which it did not disclose that it had patent applications that covered standards being developed for DRAMs. The FTC alleged that Rambus’s silence and subsequent enforcement of its patents was a patent ambush in violation of Section 5 of the Federal Trade Commission Act. An FTC administrative law judge dismissed the Rambus complaint, concluding that JEDEC policies did not obligate Rambus to disclose its patent applications and that its conduct was not anticompetitive. The Commission, however, reversed the ALJ’s decision. Rambus appealed the Commission verdict and the court of appeals ultimately decided in favor of Rambus. The court noted that Rambus had no patent applications on file when it left JEDEC and that JEDEC policies did not clearly require disclosure of unfiled but planned applications for patents that may be necessary to make or use standard-compliant products. Furthermore, the court concluded that the Commission did not prove that JEDEC’s members would not have chosen the Rambus technologies even if they were aware of its patent applications. The lack of disclosure had no competitive effect if JEDEC’s members would have included the Rambus technologies in its DRAM specification without insisting on a RAND commitment. The fact that JEDEC’s members might have imposed a RAND licensing condition if they had been aware of the Rambus patent applications did not sway the court. The court reasoned that even a monopolist has latitude to determine its prices without triggering antitrust liability. The court drew a distinction between the unlawful acquisition of a monopoly and conduct that allows an otherwise legitimate monopolist to increase its price. The court reasoned that to the extent that Rambus had a monopoly it was a consequence of its 16 US Federal Trade Commission, “Dual Consent Orders Resolve Competitive Concerns About Chevron’s $18 Billion Purchase of Unocal, FTC’s 2003 Complaint Against Unocal,” June 10, 2005.

576   Richard Gilbert lawful patents, and given that monopoly it was not an antitrust violation for Rambus to set a high price. Indeed, the court noted that a higher price encourages more competition from alternative technologies and therefore does not harm the competitive process. The court did not weigh the probability that JEDEC either would have excluded the Rambus technologies from the DRAM standards or would have imposed a RAND condition if its members had been aware of Rambus’s intended patent applications. If JEDEC would have excluded the Rambus technologies, it is possible that the failure to disclose its plans enabled Rambus to acquire a monopoly. If JEDEC would have included the Rambus technologies in the standard but imposed a RAND obligation, then it is possible that the failure to disclose allowed Rambus to achieve more monopoly power (i.e., a higher price) than it otherwise would have been able to exercise. The distinction between achieving a monopoly and achieving more monopoly power has little economic significance. A monopoly does not confer the ability to set any price. Even a monopolist faces competition if its price is high enough—if only for a share of consumers’ wallets—and a profit-maximizing monopolist sets a price high enough to make its demand at least partially elastic, since otherwise it would have an incentive to increase its price. Hence the difference between acquiring a monopoly and acquiring more monopoly power is a matter of degree and not a fundamental economic distinction. Of course significant competitive effects could be absent if the standard does not achieve market acceptance or if the patent that is allegedly concealed is not essential to comply with the standard. Rights holders have not fared so well in other standards-related cases alleging failure to comply with patent disclosure and RAND licensing obligations. In Broadcom v. Qualcomm a court denied a motion to dismiss antitrust allegations that Qualcomm violated a commitment to license wideband CDMA technology at FRAND terms. The court held that a patent holder’s intentionally false promise to license essential proprietary technology on FRAND terms, coupled with a standard-setting organization’s reliance on that promise when including the technology in a standard, and the patent holder’s subsequent breach of that promise, is actionable anticompetitive conduct. In a separate case the Court of Appeals for the Federal Circuit sanctioned Qualcomm for failing to disclose patents to a standard-setting organization related to the H.264 video standard. The court applied the legal principles of waiver, equitable estoppel, and patent misuse, leading some scholars to conclude that these doctrines are more likely to capture conduct before standard-setting organizations that may harm consumers (Hovenkamp, 2007, Merges and Kuhn, 2009, Contreras, 2011). However, if the conduct actually harms consumers, this leaves unanswered why the appropriate doctrine to evaluate such conduct should not be antitrust law. In the Rambus and Qualcomm cases it was the patentee’s conduct that was the focus of potential antitrust liability. It is conceivable that a standard-setting organization as well as its members could be found liable for violating the antitrust laws if the SSO permits standards to be developed with no limitations, the standard succeeds in

Competition Policy for Industry Standards   577

establishing market power, and its members abuse this latitude to develop standards that promote their joint interests at the expense of consumers (Skitol, 2005). However, any evaluation of such conduct would have to recognize that standard setting is critical for a well-functioning economy and therefore must give considerable deference to the activities of SSOs (Carrier, 2003).17 A clear RAND requirement for standard-essential patents and patent applications may appear to be the solution to the risk of ex post holdup, but that oversimplifies the trade-offs involved in developing standards that invoke proprietary intellectual property rights. Participation in a standard-setting organization is voluntary. If SSOs impose licensing obligations that deny patent owners an opportunity to earn a reasonable return on their inventions, they will choose not to participate in standard-setting activities at these organizations. Patent owners that do not participate in the standard development process generally are not subject to RAND licensing obligations and therefore would have greater latitude to charge high royalties after firms and consumers make investments that are specific to the standard. The competitive dynamics within and among standard-setting organizations are complex. Technology sponsors compete to have their technologies adopted in a standard. Standard-setting organizations compete with other SSOs and with unilateral efforts of firms to develop de facto standards. A consequence of this competition is that sponsors of technologies that have few competitive alternatives are unlikely to accept restrictive licensing terms and SSOs may adopt measures such as vague RAND commitments that are relatively favorable to these rights holders in order to encourage their participation (Lerner and Tirole, 2006, Chiao, Lerner, and Tirole, 2007). Sponsors of weaker technologies are willing to participate in SSOs and accept stronger licensing commitments in order to obtain greater influence in the development of a standard. The RAND obligation is a quid pro quo for this influence, but the value of influence is diminished if participation in standard-setting activities triggers RAND obligations that leave little opportunity to monetize inventions.

22.5.4.  Overly Broad Declarations of Essential Patents, Overly Broad Standard Specifications, and Package Licensing of Standard Essential Patents Failure to disclose essential patents raises concerns about potential holdup, but concerns also can arise if firms disclose too many patents. Excessive disclosure can occur

17 Moreover, Teece and Sherry (2003) argue that total welfare is maximized if SSO members make technology choices based solely on performance and real resource costs without regard to licensing costs. However, high royalties can create deadweight losses from underutilization of a technology that can offset performance benefits or resource savings. Furthermore, royalties and other licensing terms are relevant because they affect incentives for innovation and rent-seeking (Patterson, 2003).

578   Richard Gilbert if firms incorrectly claim patents to be essential to comply with the standard. A study of patents declared essential to third-generation cellular standards concluded that only about 21% of the patents were technically essential to make, sell or use products that comply with the standards (Goodman and Myers, 2005), although a larger number may be commercially essential because they cover the best implementation of the standard. Excessive disclosure also can occur if a standard specification is too broad, in the sense that it includes patented technologies that are not necessary to achieve the procompetitive objectives of the standard. In addition, patent owners can increase the number of patents they declare as essential to a standard by arbitrarily subdividing patent claims into multiple patent applications. Patents that are declared to be essential to a standard are sometimes involved in cross-licensing arrangements or licensed as a package by their owners or by a group of owners acting through a licensing agent, as in a formal patent pool. Overstating the number of essential patents can benefit a party to a cross-licensing arrangement by making it appear that the party has a stronger patent portfolio. Including non-essential patents in a package license can raise issues related to licensing costs and the allocation of licensing revenues if the patents are licensed jointly. A per-patent royalty would reward all patents, even if they were not essential to comply with the standard. Patent pools sometimes address this concern by allocating licensing revenues in proportion to an assessment of patent values (Lerner, Strojwas, and Tirole 2007). In addition, the inclusion of non-essential patents in a package license can shelter competition that would otherwise occur between patent owners if the patented technologies are substitutes for each other, as the Federal Trade Commission alleged in its complaint against Summit Technology and VISX. The Court of Appeals for the Federal Circuit addressed alleged exclusionary effects from package licenses that included non-essential as well as standard-essential patents in Princo v. International Trade Comm’n. The case concerned patents related to a standard for recordable and rewritable compact discs. Philips and Sony developed two different methods to encode information on a disc so that a CD writer could maintain proper alignment while writing data to the disc. Both methods were patented. The parties adopted the Philips method in the compact disc standard. Sony and Philips licensed their patents for the CD standard as a package, which included the Sony and well as the Philips patents for the two different methods for encoding information on the disc. Princo challenged the package license as a tied sale that obligated the licensee to purchase rights to patents that it did not require, namely, the patents on the Sony method, and required licensees to pay for the Philips encoding technology even if it did not intend to use that technology. The Federal Circuit opined that a package license was not a tied sale because it did not obligate licensees to use a particular technology but merely gave its licensee the option of using the patents in the package. The court’s opinion endorsed the efficiencies from package licensing and recognized that for a particular application there is a single profit-maximizing royalty for a package that includes at least one patent that is essential for that application, even if the patent includes other patents

Competition Policy for Industry Standards   579

that are not essential (Gilbert, 2010).18 The court also found that Princo failed to show that competition would have been viable using the Sony technology. A separate potential concern from overly broad claims of essential patents is the manipulation of standards to impose RAND licensing obligations on patented technologies when those technologies are commercially desirable, but not technically essential, to achieve benefits from standardization such as interoperability. Including a technology in a formal standard generally imposes obligations to license the technology at RAND terms if the patent owner participates in the development of the standard, which then allows implementers to obtain access to the technology at favorable terms. This concern may be theoretical because the patent owner can avoid RAND licensing obligations by refusing to participate in the standard development process, but the owner could have other interests that mandate participation. If owners of technologies that are commercially desirable but not technically essential to a standard are compelled to offer the technologies at RAND terms, whether such obligations would promote economic welfare depends on many factors, including whether the technologies are genuinely essential for commercial feasibility and the effects of involuntary RAND obligations on incentives for innovation.

22.5.5.  Ex Ante Bargaining as an Alternative to RAND Commitments Some standard-setting organizations have considered measures to obtain greater certainty about potential royalty terms for patents that are essential to proposed standards. In the past, SSOs have shunned discussions of royalty terms out of concern for potential antitrust exposure. A 2007 policy statement by the IEEE addressed whether prices could be discussed in meetings of IEEE standards associations. The unequivocal answer was no (IEEE Standards Association, 2010). More recently, some SSOs have reconsidered this position and have received encouraging guidance from the US Department of Justice about proposals that would require patent owners to disclose maximum royalties and the most restrictive terms under which rights holders would license their technologies, prior to the development of a standard.19 Ex ante disclosure risks the exercise of monopsony power by members of standard-setting organizations. Members who have a common interest in obtaining low royalties may act in concert to impose low royalties on owners of patents that are 18  Allowing patents to be licensed separately as well as in a package addresses both the concern that the package license may exclude competition and the concern that the package would raise royalty payments by combining substitute patents (Lerner and Tirole, 2004). 19  See Business Review Letter from Thomas O. Barnett, Assistant Att’y Gen., to Robert A. Skitol, Drinker, Biddle & Reath, LLP (Oct. 30, 2006), available at http://www.justice.gov/atr/public/​ busreview/219380.pdf and Business Review Letter from Thomas O. Barnett, Assistant Att’y Gen., to Michael A. Lindsay, Dorsey & Whitney LLP (Apr. 30, 2007), available at http://www.justice.gov/atr/​ public/busreview/222978.pdf.

580   Richard Gilbert essential to make or use products that comply with the standard. Both the Department of Justice and the Federal Trade Commission have acknowledged this risk, but also recognized that ex ante disclosure of maximum royalties and most restrictive licensing terms can provide valuable information to avoid exposure to potential holdup. The agencies have indicated that they would apply a rule-of-reason framework to evaluate the relative benefits and risks of joint negotiation of licensing terms. The tolerance for coordinated conduct by members of an SSO that may exercise buyer market power should depend on the likelihood and magnitude of ex post holdup. If ex post holdup is unlikely, coordinated conduct to establish licensing terms ex ante has little benefit and may distort incentives for innovation by shifting the terms of patent licenses to favor technology adopters. If standard-related holdup is likely and substantial, joint negotiation of licensing terms by the members of an SSO before a standard issues can help fill the void left by vague RAND commitments to limit possible opportunistic conduct. However, a plausibly less restrictive alternative is to rely on independent bilateral negotiations between potential licensees and rights holders along with a clear non-discrimination requirement (Gilbert, 2011). Preventing undue discrimination between similarly situated licensees ensures that firms will gain the benefits of licensing terms negotiated by early adopters before these early adopters and consumers make investments that are specific to a standard. This alternative policy requires SSOs or the courts to convey the meaning and requirements of non-discrimination in a technology-licensing context, but it does not require a precise determination of fair and reasonable royalties.

22.6.  Summary Remarks The development of industry standards is a cooperative activity among representatives of firms that often compete to supply products covered by the standards. Cooperative activity by rivals potentially raises antitrust concerns. The decision to endorse a particular standard can have significant implications for the cost and performance of the products they sell and for competition from products that do not comply with the standard. The potential for participants in a standard-setting venue to harm competition by favoring one standard over another or by imposing conditions that affect prices of products covered by the standard cannot be ignored. Nonetheless, courts and antitrust agencies generally defer to the choices made in standard-setting organizations when those decisions are made in a consensus process that is transparent and open to interested parties. The justification for the deferential treatment of standard setting is not based on a general principle akin to the “invisible hand” that guides market outcomes. There is no guarantee that the members of a standard-setting organization will choose the “best” standard. Instead, courts acknowledge the likely benefits from standards and implicitly admit that litigation is unlikely

Competition Policy for Industry Standards   581

to result in better standards. Antitrust challenges to standard setting, when they have occurred, often have focused on coordinated conduct that distorts the process to fix prices or to favor entrenched interests by excluding or imposing costs on rivals. Technologies that are candidates for standards often require access to proprietary intellectual property rights, particularly in the information and communications technology industries for which hundreds of patents can be essential to comply with a standard. Most setting-setting organizations do not prohibit the development of standards that invoke patented technologies, but require assurance that licenses to these patents be available either royalty-free or on “reasonable and non-discriminatory” (RAND) terms. Standard-setting organizations have not clearly delineated the extent of disclosure obligations, the meaning of reasonable or non-discriminatory, or the conditions under with the owner of a patent with a RAND commitment may seek an injunction to prevent the use of a technology covered by the patent. Courts also have been slow to define these obligations, and recent developments at the International Trade Commission will test whether RAND commitments prevent patentees from pursuing exclusion orders at that venue. The resolution of the dimensions and extent of RAND commitments, or the failure to resolve these characteristics, will have important consequences for the benefits from industry standards.

Acknowledgments I am grateful for helpful discussions with Jorge Contreras, Bret Dickey, Joe Farrell, Amy Marasco, Jim Ratliff, Carl Shapiro, and Tim Simcoe. The author has consulted for Apple and Qualcomm on matters related to standards and standard-essential patents, and has testified adverse to Rambus on issues related to the disclosure of applications for standard-essential patents.

References ABA. 2005. Handbook on the Antitrust Aspects of Standards Setting. American Bar Association Section of Antitrust Law Monograph. American National Standards Institute (ANSI). 2008. ANSI Patent Policy: Inclusion of Patents in American National Standards. Anton, James J., and Dennis A.  Yao. 1995. Standard Setting Consortia, Antitrust, and High-Technology Industries. Antitrust Law Journal 64:247–65. Arrow, Kenneth J. 1950. A Difficulty in the Concept of Social Welfare. Journal of Political Economy 58:328–46. Besen, Stanley M., and Joseph Farrell. 1994. Choosing How to Compete: Strategies and Tactics in Standardization. Journal of Economic Perspectives 8:117–31.

582   Richard Gilbert Bekkers, Rudi, R. Bongard, and A. Nuvolari. 2011. An Empirical Study on the Determinants of Essential Patent Claims in Compatibility Standards. Research Policy 40:1001–15. Bekkers, Rudi, and Andrew Updegrove. 2012. A Study of IPR policies and Practices of a Representative Group of Standards Setting Organizations Worldwide. Working paper. Biddle, Brad, Andrew White, and Sean Woods. 2010. How Many Standards in a Laptop? And Other Empirical Questions. Working paper. Available at SSRN:  http://ssrn.com/ abstract=1619440. Blind, Knut, et al. 2011. Study on the Interplay between Standards and Intellectual Property Rights (IPRs). Tender No ENTR/09/015 (OJEU S136 of 18/07/2009). Final Report, April. Carrier, Michael A. 2003. Why Antitrust Should Defer to the Intellectual Property Rules of Standard-Setting Organizations: A Commentary on Teece and Sherry. Minnesota Law Review 87:2019–37. Chien, Colleen V., and Mark A Lemley. 2012. Patent Holdup, the ITC, and the Public Interest. Cornell Law Review 98:1–46. Chiao, Benjamin, Josh Lerner, and Jean Tirole. 2007. The Rules of Standard-Setting Organizations: An Empirical Analysis. Rand Journal of Economics 38:905–30. Contreras, Jorge L. 2011. Equity, Antitrust, and the Reemergence of the Patent Unenforceability Remedy. Antitrust Source, October, 1–8. David, Paul A., and Shane Greenstein. 1990. The Economics of Compatibility Standards: An Introduction to Recent Research. Economics of Innovation and New Technology 1:3–41. Dranove, David, and Neil Gandal. 2003. The DVD-vs.-DIVX Standard War: Empirical Evidence of Network Effects and Preannouncement Effects. Journal of Economics and Management Strategy 12:363–86. Edlin, Aaron, and Rebecca Haw (forthcoming), Cartels By Another Name:  Should Licensed Occupations Face Antitrust Scrutiny? University of Pennsylvania Law Review. Farrell, Joseph. 2007. Should Competition Policy Favor Compatibility? In Shane Greenstein and Victor Stango, Standards and Public Policy, 372–88. Cambridge: Cambridge University Press. Farrell, Joseph, and Paul Klemperer. 2007. Coordination and Lock-In:  Competition with Switching Costs and Network Effects. In M. Armstrong and R. Porter, Handbook of Industrial Organization, vol. 3, 1970–2072. Amsterdam: North-Holland. Farrell, Joseph, and Garth Saloner. 1986. Installed Base and Compatibility: Innovation, Product Preannouncements and Predation. American Economic Review 76:940–55. Farrell, Joseph, and Garth Saloner. 1987. Competition, Compatibility and Standards:  The Economics of Horses, Penguins and Lemmings. In L. Gabel, Product Standardization and Competitive Strategy, 1–21. Amsterdam: North-Holland. Farrell, Joseph, and Carl Shapiro. 1992. Standard Setting in High Definition Television. Brookings Papers: Microeconomics 1992:1–77. Farrell, Joseph and Carl Shapiro. 2008. How Strong Are Weak Patents? American Economic Review 98:1347–69. Farrell, Joseph and Timothy Simcoe. 2012. Choosing the Rules for Consensus Standardization. Rand Journal of Economics 43(2):235–52. Farrell, Joseph, and Philip Weiser. 2004. Modularity, Vertical Integration, and Open Access Policies: Towards a Convergence of Antitrust and Regulation in the Internet Age. Harvard Journal on Law and Technology 17:85–135. Farrell, Joseph, et  al. 2007. Standard Setting, Patents, and Hold-Up. Antitrust Law Journal 74:603–70.

Competition Policy for Industry Standards   583

Garza, Deborah A., et  al. 2007. Antitrust Modernization Commission Report and Recommendations. April. Geradin, Damien, and Miguel Rato. 2007. Can Standard-Setting Lead to Exploitative Abuse? A Dissonant View on Patent Hold-Up, Royalty Stacking and the Meaning of FRAND. European Competition Law Journal 3:101. Gilbert, Richard J. 2010. Ties That Bind: Policies to Promote (Good) Patent Pools. Antitrust Law Journal 77:1–45. Gilbert, Richard J. 2011. Deal or No Deal? Royalty Negotiations by Standard Development Organizations. Antitrust Law Journal 77:855–88. Goodman, David J., and Robert A.  Myers. 2005. 3G Cellular Standards and Patents. IEEE WirelessCom. Grindley, Peter C., and David J.  Teece. 1997. Managing Intellectual Capital:  Licensing and Cross-Licensing in Semiconductors and Electronics. California Management Review 39:8–41. Hillman, Arye L., and John G. Riley. 1989. Politically Contestable Rents and Transfers. Economics and Politics 1:17–39. Hovenkamp, Herbert. 2007. Standards Ownership and Antitrust. Boston College Law Review 48:87–109. IEEE Standards Association. 2010. Promoting Competition and Innovation: What You Need to Know about the IEEE Standards Association’s Antitrust and Competition Policy. Approved March 22, 2007, updated August 24, 2010. Katz, Michael L., and Carl Shapiro. 1994. System Competition and Network Effects. Journal of Economic Perspectives 8:93–115. Kleiner, Morris M. 2000. Occupational Licensing. Journal of Economic Perspectives 14:189–202. Krechmer, Ken. 2009. Open Standards: A Call for Change. IEEE Communications Magazine, May, 88–94. Layne-Farrar, Anne, A. Jorge Padilla, and Richard Schmalensee. 2007. Pricing Patents for Licensing in Standard-setting Organizations:  Making Sense of FRAND Commitments. Antitrust Law Journal 74:671–706. Lemley, Mark A. 2002. Intellectual Property Rights and Standard-setting Organizations. California Law Review 90:1889–990. Lemley, Mark A., and Carl Shapiro. 2007. Patent Holdup and Royalty Stacking. Texas Law Review 85:1991–2049. Lerner, Josh, and Jean Tirole. 2004. Efficient Patent Pools. American Economic Review 94:691–711. Lerner, Josh, and Jean Tirole. 2006. A Model of Forum Shopping. American Economic Review 96:1091–113. Lerner, Josh, Marcin Strojwas, and Jean Tirole. 2007. The Design of Patent Pools: The Determinants of Licensing Rules. Rand Journal of Economics 38:610–25. Matutes, Carmen, and Pierre Regibeau. 1988. Mix and Match: Product Compatibility without Network Effects. Rand Journal of Economics 19:221–34. Matutes, Carmen, and Pierre Regibeau. 1996. A Selective Review of the Economics of Standardization. European Journal of Political Economy 12:183–209. Merges, Robert, and Jeffrey Kuhn. 2009. An Estoppel Doctrine for Patented Standards. California Law Review 97:1–50. Miller, Joseph. 2007. Standard Setting, Patents and Access Lock-In: RAND Licensing and the Theory of the Firm. Indiana Law Review 40:351–95.

584   Richard Gilbert Patterson, Mark R. 2002. Inventions, Industry Standards, and Intellectual Property. Berkeley Technology Law Journal 17:1043–83. Patterson, Mark R. 2003. Antitrust and the Costs of Standard-Setting: A Commentary on Teece & Sherry. Minnesota Law Review 87:1995–2018. Postrel, Steven R. 1990. Competing Networks and Proprietary Standards:  The Case of Quadraphonic Sounds. Journal of Industrial Economics 39:169–85. Salant, David J. 2009. Formulas for Fair, Reasonable and Non-discriminatory Royalty Determination. International Journal of IT Standards and Standardization Research 2009:67. Samuelson, Pamela. 2009. Are Patents on Interfaces Impeding Interoperability? Minnesota Law Review 93:1943–2019. Seek, Momar D., and David D. Evans. 2004. Major U.S. Cities Using National Standard Fire Hydrants, One Century after the Great Baltimore Fire. NISTIR 7158, US Department of Commerce, August. Shapiro, Carl. 2001. Setting Compatibility Standards:  Cooperation or Collusion? In Rochelle Dreyfuss, Diane Zimmerman, and Harry First, Expanding the Bounds of Intellectual Property, 83–101. New York: Oxford University Press. Shapiro, Carl, and Hal Varian. 1999. Information Rules:  A  Strategic Guide to the Network Economy. Boston: Harvard Business School Press. Simcoe, Timothy. 2006. Open Standards. In Henry Chesbrough, Wim Vanhaverbeke, and Joel West, Open Innovation: Researching a New Paradigm, 161–83. New York: Oxford University Press. Simcoe, Timothy. 2012. Standard Setting Committees:  Consensus Governance for Shared Technology Platforms. American Economic Review 102:305–36. Skitol, Robert A. 2005. Concerted Buying Power: Its Potential for Addressing the Patent Holdup Problem in Standard Setting. Antitrust Law Journal 72:727. Swanson, Daniel G., and William J. Baumol. 2005. Reasonable and Nondiscriminatory (FRAND) Royalties, Standards Selection, and Control of Market Power. Antitrust Law Journal 73:1–58. Teece, David J., and Edward F. Sherry. 2003. Standards Setting and Antitrust. Minnesota Law Review 87:1913–94. US Department of Justice and Federal Trade Commission. 2007. Antitrust Enforcement and Intellectual Property Rights: Promoting Innovation and Competition. April. US Federal Trade Commission. 1996. Anticipating the 21st Century: Competition Policy in the New High-Tech, Global Marketplace. US Federal Trade Commission. 2011. The Evolving IP Marketplace: Aligning Patent Notice and Remedies with Competition. Yeh, Brian T. 2012. Availability of Injunctive Relief for Standard-Essential Patent Holders. Congressional Research Service, 7-5700, September 7.

Case Citations Allied Tube & Conduit Corp. v. Indian Head, Inc., 486 U.S. 492 (1988). American Society of Mechanical Engineers v. Hydrolevel Corp., 456 U.S. 556 (1982). Apple, Inc. et al. v. Motorola, Inc., U.S. District Court Northern District Of Illinois, Eastern Division, No. 1:11-Cv-08540, Opinion and Order of June 22, 2012. Broadcom Corp. v. Qualcomm Inc., 501 F.3d 297, 314 (3d Cir. 2007).

Competition Policy for Industry Standards   585

Dell Computer Corp., 121 F.T.C. 616 (1996). eBay Inc. v. MercExchange, L.L.C., 547 U.S. 388 (2006). Hydrolevel Corp. v. Am. Soc’y of Mech. Eng’rs, 635 F.2d 118, 124 (2d Cir. 1980). In re Union Oil Co. of California, 2004 FTC LEXIS 115 (FTC 2004). In the Matter of Certain Gaming and Entertainment Consoles, Related Software, and Components Thereof, International Trade Commission, Initial Determination, May 2012. National Macaroni Manufacturers Association v. F.T.C., 345 F.2d 421 (7th Cir., 1965). National Society of Professional Engineers v. United States, 435 U.S. 679 (1978). Princo Corp. v. International Trade Comm’n, 616 F.3d 1318, (Fed. Cir., Aug. 30, 2010). Rambus, Inc. v. Federal Trade Commission, 522 F.3d 456, (D.C. Cir. 2008). TruePosition Inc. v.  LM Ericsson Telephone Company, et  al., No. 11-4574, District Court, Eastern District of Pennsylvania (2012). U.S. v. Microsoft Corp., 253 F. 3d 34 (D.C. Cir. 2001). U.S. Federal Trade Commission, In the Matter of Negotiated Data Solutions LLC, Docket No. C-4234, Decision and Order (September 22, 2008). U.S. Federal Trade Commission, In the Matter of Summit Technology and VISX, Docket No. 9286, Complaint (March 24, 1998). Wilk v. American Medical Ass’n, 895 F.2d 352 (7th Cir. 1990).

CHAPTER 23

A N T I T RU S T C O R P O R AT E G OV E R NA N C E A N D C OM P L IA N C E ROSA M. ABRANTES-METZ AND D. DANIEL SOKOL

23.1. Introduction The study of cartel detection has been an important part of antitrust scholarship and policy for some time. Most of the development of the literature on cartel detection has focused at the firm level. This should not be surprising since industrial organization economics focuses on firms and markets. Antitrust scholarship has not focused as much on compliance and corporate governance within a given firm. Understanding the internal workings of firms would allow for closer to optimal deterrence, as this understanding would allow for the calibration of policy around the incentives within a given firm, its subunits, and individuals who work therein, to comply with antitrust law. Both theoretical and empirical work in a number of different fields, including economics, accounting, finance, organizational theory, and sociology provide important insights indicating that a firm is not merely a single entity in its actions. Rather, a firm is made up of a number of various components, each of which has its own incentives that shape firm behavior. This chapter reviews both the antitrust and the non-antitrust literatures on compliance and corporate governance to provide a clearer picture of the extant literature and the theoretical and empirical gaps within the antitrust literature to better inform antitrust policy on detecting cartels. This chapter explores the scholarship both within and outside of antitrust to better understand internal detection of wrongdoing and improved compliance in the antitrust cartel context.

Antitrust Corporate Governance and Compliance   587

23.2.  Organizational Environment Cartels are somewhat distinct from other types of white-collar crime such as accounting or tax fraud because cartel activity requires coordination across firms. Like other types of white-collar crime, cartels require that illegal activity be undertaken within a given firm, which implicates issues of compliance and corporate governance. What makes cartels distinct from all other types of white-collar crime is that for a cartel to succeed, there needs to be coordination both within a firm (vertical relations given levels of management and oversight) as well as across firms (horizontal relations among the cartel members). This chapter focuses exclusively on cartel vertical relations within a firm. Chapters in this volume by Choi and Gerlach; Levenstein and Suslow; Hendricks, McAfee, and Williams; Doane and coauthors; and Green, Marshall, and Marx address the horizontal issues of cartels. A firm’s environment and the amount of individual discretion affect decision-making for the entire organization and may constrain the decision-making of individuals within it (Finkelstein and Hambrick 1990). Understanding organizational structure and incentives may illuminate how to better structure closer to optimal corporate compliance to police against antitrust violations. We begin with an analysis of agency costs to better understand incentives within the firm. A core part of non-antitrust literature in both economics and finance is the concept of agency costs (Jensen and Meckling 1976). Within a firm, the agent may have incentives that differ from those of management. Firms work to reduce this misalignment through improved monitoring. Strong monitoring can reduce this divergence but might deter agents from risk taking, which in turn could benefit the firm. Sometimes even if at the board of directors level the firm wants to comply with antitrust law, its agents may not (Kaplow 2011). Corporate crime is an agency cost (Alexander and Cohen 1999). Agency costs and the ability of the firm to effectively monitor rogue agents also play into the importance of focusing antitrust cartel detection at the individual level rather than just at the firm level. For example, a manager may ignore risks because she believes that she will not get caught for cartel behavior. The inability of antitrust agencies to detect existing cartels (relative to individuals within firms), the relatively low levels of punishment for individuals who are discovered engaging in cartel behavior (Combe and Monnier 2011; Connor and Lande 2012), and the low level of cartel detection (Connor 2012; Ormosi 2011) may affect the risk/reward calculation of a particular manager in participating in a cartel. The broader non-antitrust literature suggests that many firms behave illegally and consider many factors in their decision to comply or not to comply, based on the relative costs and benefits of compliance (Helland 1998). If cartel behavior goes undetected, both the individual and the company may benefit (and individuals can justify their involvement as somehow saving jobs in the company), because the harms are externalized

588   Rosa M. Abrantes-Metz and D. Daniel Sokol (Siltaoja and Vehkaperä 2010) and both individual and company benefit from greater stock price based on the supracompetitive pricing of the cartel (Spagnolo 2008). There is some general work on managerial incentives regarding managerial effort and industry competition, such as Scharfstein 1988 and Schmidt 1997. The work on managerial incentives specific to cartels is relatively sparse. Aubert 2007 analyzes the interaction of managerial incentives to collude in which collusion may be a substitute for increased managerial effort. The model shows that, due to incentives to collude, firms may behave inefficiently, which in turn creates even more incentive to collude, as firms cannot easily control the actions of their employees. Spagnolo (2000) provides a binding contract from the firm to the manager to support tacit collusion. Similarly, in another work Spagnolo (2005) suggests that a firm might hire a manager with a preference for income smoothing, thus creating an incentive to “meet the numbers” via a cartel. Two recent works offer some policy solutions. Angelucci and Han (2010) suggest that for a cartel compliance program to be effective, managerial sanctions need to be low. Another solution may be to penalize shareholders through a dilution of shares so that if managers undertake cartel activity through poorly aligned incentives, shareholders will be punished but not to the point that the firm would go bankrupt due to high fines (Buccirossi and Spagnolo 2008). The empirical work to date of managerial incentives and cartels is thin. One paper analyzes managerial compensation among 796 Japanese firms from 1968 to 1992 (Joh 1999) and another examines a sample of German firms involved in coal cartels from 1881 to 1913 (Burhop and Lubbers 2009). Often bonuses are tied to performance, and successful collusion can reward individuals for this reason. The opposite might, however, be even more important—that the fear of being fired for poor performance may incentivize a midlevel manager who fears losing his job to participate in collusion. Adding to the profit motive for collusion, there might be a contractual undertaking by firms (where this is not outlawed) to pay the penalties levied on individual employees for their participation in cartel activity. This may be in the form of informal contracting for jurisdictions where such formal contracts may be illegal. One recent empirical study reviews the specific pay mechanisms and other incentives used by firms to reward individual cartel individual members. In terms of stock options, cartel individual members are more likely to exercise stock options more rapidly than individuals in other firms (Gonzalez et al. 2013). Work has yet to be done that examines actual contractual incentives of individual members of cartels. It may be that one reason that such studies have not yet been undertaken is due to the nature of what evidence gets collected for criminal cases. Antitrust enforcers and private plaintiffs are concerned about the illegality of price fixing and the mechanisms used to undertake it across firms. The incentive mechanisms within firms do not establish proof of price fixing and hence are not important for cases. Research into the incentives used in performance contracts for such individuals and across industries would be useful. The time horizon of individuals impacts their potential involvement in cartels. Sometimes the particular individuals involved in the cartel have moved on to other

Antitrust Corporate Governance and Compliance   589

companies and are not around to deal with the negative consequences of their illegal activity (Connor and Lande 2012). There is evidence in the airline industry that financially weak firms are more likely to cause price wars and thus more likely to defect from a collusive arrangement. The conjectured reason for this is that they effectively have a short time horizon since there is a nonnegligible probability they will exit the industry through bankruptcy or acquisition (Stephan 2010b). In other cases, the individuals remain within the company for a long time because they enjoy the rewards of their illegal activity, think that they will not be caught, or know that even if they move to a different company, they will still be criminally liable but no longer have the ability to cover up their misdeeds.

23.3.  Firm Indicia of Cartel Activity and Their Impact on Compliance Since the mid-1990s, most criminal antitrust violations targeted by US Department of Justice Antitrust Division involve international cartels. This focus on large-firm international price fixing is important given that the size of the firm affects its propensity for criminality. One study using event history analysis finds that larger firms are more prone to criminal behavior overall (Baucus and Near 2001). Size may be a factor because, as organizations get larger, agency costs increase and monitoring becomes more difficult. Organizational design issues may contribute to illegality. As organizations increase in complexity, firms develop various organizational structures in response (Tushman and O’Reilly 1997). The larger and more complex an organizational structure, the more difficult it is to coordinate various organizational subunits. Complexity of organizations also may increase agency costs (Chen and Hambrick 1995). Because of organizational size and complexity, it is possible to hide wrongdoing from government officials, and inside and outside gatekeepers such as in-house counsel or outside auditors. The more complex the organization, the higher the proclivity within the firm for illegal activity (McKendall and Wagner 1997). Given the link between organizational structure and wrongdoing, it seems to be the case that the internal governance structure within a corporation affects the likelihood of successfully monitoring illegal behavior and enforcing compliance. For example, independent outside directors seem to be more effective than inside directors (members of the firm’s management team) at policing against corporate fraud (Beasley 1996, 2000; Uzun, Szewczyk, and Varma 2004) and opportunistic grants of stock options (Bebchuk, Grinstein, and Peyer 2010). Gonzalez and coauthors (2013) note that cartel member firms are more likely to have busy directors (directors on many boards) than a typical board. Violations of antitrust law may occur because of organizational failure rather than merely profit seeking on the part of management (Beckenstein and Gabel 1986). Some

590   Rosa M. Abrantes-Metz and D. Daniel Sokol firms have a strong compliance culture because incentives have been put into place to reward strong antitrust compliance. These incentives may take the form of pay incentives, organizational structures that allow for effective monitoring by legal and compliance staff, and the overall creation of certain corporate cultures. For other firms, the social norms may work towards noncompliance for many of the same reasons. When individuals are rewarded for unlawful behavior, when monitoring by compliance staff is not strong, or where country-level norms push toward cartel behavior, these norms reinforce cartel behavior (Sokol 2012).

23.3.1.  Incentive Pay Firms may change the incentives for illegality for their employees via a focus in incentive pay on long-term rather than short-term gain (Bebchuk and Fried 2004). There is a principal-agent problem in firms in which the agents (employees) veer from what is in the shareholders’ best interest in order to maximize the individual employee’s best interest. One way in which firms reduce the agency cost problem is through incentive-based pay. It is more likely that firms that promote short-term gains for pay have individuals who may undertake criminal behavior to “meet the numbers,” as pay may be linked to performance. This is not to suggest that all pay for performance is problematic. If officers and directors have an equity stake in the firm, they have incentives to monitor the firm for illegal activity when the illegal behavior threatens firm returns (Alexander and Cohen 1999). In some cases pay for performance better aligns managers’ incentives with those of the firm (Baber, Janakiraman, and Kang 1996; Morgan and Poulsen 2001). Only when incentive-based pay is too large may this incentive problem lead to illegal behavior. Nonlinearities in payoffs, theory would suggest, encourages fraud. It may be that fraud happens in industries with highly specialized knowledge and high variance, and maybe incentive contracts work in such an environment. Work in the area of corporate wrongdoing in other regulatory fields offers some theoretical and empirical basis on what seems to be correlated with wrongdoing. There is some theoretical support for the incentives behind this behavior (Fudenberg and Tirole 1995; Hermalin and Weisbach 2012). A number of empirical findings show that CEOs whose pay is incentive-based are more likely to commit fraud by misreporting material information or manipulating earnings (Berhstresser and Phillippon 2006; Peng and Röell 2008; John, Ryan, and Tian 2009; Burns and Kedia 2006; Schnatterly 2003). CFO bonuses and stock options have also been empirically linked to manipulation of earnings (Oberholzer-Gee and Wulf 2012), and in the backdating of stock options (Lie 2005). A similar relationship of too much pay for performance is linked to corporate tax avoidance (Desai and Dharmapala 2006, 2009). The role of incentives in shaping behavior has been extended to cartels. If managers receive bonuses passed on certain profitability metrics, this may encourage members to meet their performance-based metric by any means necessary—including becoming

Antitrust Corporate Governance and Compliance   591

involved in a cartel (Buccirossi and Spagnolo 2008). Managers might be willing to risk joining a cartel if too much of pay is linked to performance, as the short-term incentive of a significant payout will increase (Gonzalez et al. 2013), especially if the risk of detection is low both inside the firm from compliance officers and outside the firm from antitrust enforcers.

23.3.2.  Enforcement and the Impact on Firm Behavior In public speeches and in the unwillingness to provide credit for penalty mitigation to companies for strong compliance programs, both the US DOJ and the EU Directorate-General for Competition in effect utilize a strict liability framework for cartel enforcement. Put differently, the fact that a cartel is detected is an indication of a failed compliance program. This strict liability theory involving cartels and compliance contradicts the main academic literature in the area of entity liability and punishment. Arlen and Kraakman (1997) advocate a mixed regime of negligence and strict liability. A strict liability regime will not ensure compliance in circumstances where the increased likelihood of uncovering harm by the compliance program outweighs the firm’s ability to prevent such harm. In such an environment, the firm has a disincentive for any sort of compliance program that would monitor firm behavior. Therefore, some sort of negligence regime needs to be incorporated into a composite approach to also punish behavior where an entity has demonstrated a “failure to discharge its policing duties.” This mixed regime is preferable to either a pure strict liability or negligence regime. In practice, most antitrust regimes are based on strict liability. The strict liability regime for antitrust shapes incentives within the firm to continue with criminality because there is very limited benefit to proactively spending on serious compliance when the firm (and individuals therein) benefit from nondetection. A  system that enforces individual liability only raises the individual’s expected cost of liability for wrongdoing but not necessarily that of the firm. By contrast, company-level liability imposes costs not on the individual but on the shareholders of the company (Arlen 2012). Thus, even if firms want to comply with the antitrust laws, individuals within firms may have a different set of calculations that may trigger illegal behavior. These “bad apples” may act illegally no matter how strong the corporate monitoring and compliance program implemented. The amount of compliance that a firm is willing to perform for antitrust or other areas of potential wrongdoing results from the nature of legal duties involving corporate governance of a firm. One reason for the sometimes anemic cartel compliance efforts by firms is that corporate law does not provide sufficient incentives to create the sort of internal compliance process that may actually create effective compliance for antitrust. These incentives suggest why antitrust compliance seems to encourage only weak compliance. Corporate boards under Delaware law have very weak legal duties to monitor the firms’ actions. The scope for violating such duties is narrow. Given the high threshold for

592   Rosa M. Abrantes-Metz and D. Daniel Sokol liability under Caremark,1 there seems to be little incentive for a serious proactive compliance program beyond the minimum required under corporate law. Empirical work on board liability shows that in practice, there are limited financial penalties for weak monitoring by the board (Black 2006). The one exception to this set of incentives is that the scope of liability, should a court find the board of directors to be liable, would make the violation of corporate law for noncompliance nonexculpable (Bainbridge 2009). The lack of strong corporate compliance mechanisms overall shapes the nature of firm compliance in antitrust. Many compliance programs are merely weak or cosmetic, and companies have incentives to undertake only cosmetic compliance since self-reporting may hurt the firm (Arlen 2011). The same seems to hold true in the antitrust context (Sokol 2012).

23.3.3.  Senior Management within an Organization A crucial dimension of better incentives (and to the power of various policies) for compliance takes into account the distinction between managerial incentives and shareholder incentives and between the incentives of a middle manager and those of a senior manager. This next section examines the interrelationships between different individuals within the firm and across firms to better understand what might constitute effective compliance in the antitrust setting. Senior management is an important component of firm governance and compliance (Hambrick 2007; Hambrick and Mason 1984). Different management styles affect corporate decision-making in a number of areas, such as investment and financial policy, tax compliance, and organizational strategy (Bertrand and Schoar 2003; Dyreng, Hanlon, and Maydew 2010). The focus on senior management in antitrust is particularly important. The majority of individual defendants in cartel cases have been at the level of a company’s corporate officers (Gallo et al. 2000; Stephan 2010b). The distinctiveness of criminality within top management of an organization may be due to the large amount of power that top management possesses and how it impacts firm culture (Schein 2010). Therefore, the preferences of top management will affect strategic outcomes of a corporation (Chatterjee and Hambrick 2007; Camerer and Lovallo 1999). Some work suggests that longer CEO tenure (Miller 2001) and top management team tenure (Finkelstein and Hambrick 1990) negatively affect the strategic dynamics of a corporation (Henderson, Miller, and Hambrick 2006). As stability usually favors cooperative outcomes (Stigler 1964), more stable firm management in an industry should facilitate collusion. Moreover, younger managers may be “trained” by the older generation to participate in cartels (Geiss 1967). Top management’s crimes differ from others within an organization, because the board of directors more closely monitors senior management than other parts of the

1 

In re Caremark Int’l Inc. Derivative Litig., 698 A.2d 959 (Del. Ch. 1996).

Antitrust Corporate Governance and Compliance   593

firm. There are various internal control devices to better align the incentives of shareholders and management, so as to improve the quality of oversight and reduce incentives for cartel activity. For example, companies might issue debt to constrain management from overinvestment (Harvey, Lins, and Roper 2004). Separation of the CEO and chairman position improves the board’s ability to monitor the CEO (Efendi, Srivastava, and Swanson 2007; Core 2003), as do independent directors who have financial expertise (Agrawal and Chadha 2005). Similarly, providing equity for directors might give rise to improved monitoring of management by directors by better aligning director interests with shareholder interests (Jensen 1993; Ertugrul and Hegde 2008). Moreover, board diversity serves to better monitor CEOs, based on CEO turnover (Adams and Ferreira 2009; Weisbach 1988). One cartel-related article (Simpson and Koper 1997)  that analyzes 43 antitrust offenders over 22  years finds that certain top management variables correlate more with antitrust illegality (CEOs with finance and administrative backgrounds, less top level turnover, and firm reliance on a single product market) than others. More recent finance-based work on cartels (Gonzalez et al. 2013) suggests that cartel members firms (1) tend to file an abnormally large amount of financial restatements, (2) have less effective monitoring due to foreign or busy (too many board positions) directors, (3) are less likely to replace directors who resign, and (4) switch auditing firms less often than the norm. In the antitrust context, the tone of senior management matters to compliance within the organization. From a procompliance standpoint, if the CEO mandates antitrust training, middle managers are more likely to take such compliance seriously. The CEO must project a sincere desire to comply. This will set the tone for the entire organization in terms of its antitrust compliance. The CEO must be fully committed to the antitrust compliance program and be consistent in such commitment (ABA 2010a). The more powerful the messenger, the more likely that others within the organization will conform to the message because of the CEO’s ability to offer compliant managers greater resources, legitimacy, and power (Oliver 1991). Therefore, the involvement by top management in criminal activities may merit tougher penalties, since senior management involvement signals compliance weakness and a corrupt culture overall, including at the level of board of directors.

23.3.4.  Middle Management and Other Employees Within the firm, middle management may not have the same incentives for compliance as senior management does. For example, in a divisional organizational model, each divisional unit may try to maximize the short-term profitability of that particular division instead of the entity as a whole (Shin and Stulz 1998). This suggests that organizational structure may be a contributing cause for misalignment of incentives. Culture also implicates the behavior of middle managers. To become successful leaders in companies, middle management and lower-level employees may mimic the

594   Rosa M. Abrantes-Metz and D. Daniel Sokol behavior of senior management (Haunschild and Miner 1997). This may include behavior such as cartel involvement if such cartel participation allows these middle managers to move up the ranks within the firm. Middle managers may be under significant pressure to meet various performance targets (Alexander and Cohen 1996). The financial rewards or possibilities for prestige or promotion for managers for divisional results may be different than for the firm as a whole (Berger and Ofek 1995). A cartelist may rationally risk criminality because he wants to save jobs in his group or division. The cartel participant believes that as long as other firms do the same during a time of economic downturn, a cartel will naturally break up when the economy improves. For other potential cartelists, the fear of being caught and punished is a deterrent. However, this deterrent does not apply to others who are genuine sociopaths, or are desperate, angry, and/or scared of losing a job if they do not participate in a cartel. In yet other cases, the cartelist may be convinced that she is smarter than everyone else.

23.4. Culture 23.4.1.  Norm Creation Antitrust scholarship has overlooked the importance of firm culture on compliance. Beyond the legal regime, there are other incentives within the firm that influence compliance regarding cartel policy. For a cartel to avoid detection by a participating firm’s employees, there needs to be some level of management that actively participates in the cartel and other employees who either are unaware of or who turn a blind eye to such behavior. As noted above, incentives within the firm shape firm behavior and the behavior of its agents. Thus, firm culture creates direct incentives for criminality. A compliance program, if not made part of a corporation’s culture, will be viewed antagonistically by midlevel management. Such managers may view it as a system to beat in pursuit of sales and commission. Illegal activity may become embedded in an organization over time and become a part of organization culture (Aguilera et al. 2007; Bettenhausen and Murnighan 1985). Unethical changes within an organization may be subtle and gradual, such that individuals do not realize that they are engaging in illegal behavior (Ashforth and Anand 2003). Over time, organizations reach a tipping point in their culture in which illegality becomes a defining element of the organization itself (Maclean and Behnam 2010). Additionally, rapid growth and unrealistic company performance forecasts are factors that indicate an increased likelihood of accounting fraud (Summers and Sweeney 1998). Culture plays out within the cartel and compliance culture of various firms. There are different types of cartelists among individuals and firms. Each of the types of cartelists

Antitrust Corporate Governance and Compliance   595

possesses its own unique traits based on nationality and size of the company. According to Sokol’s survey (2012), non-US firms, including Global Fortune 500 firms, seem not to have a strong cartel compliance culture. Respondents suggested that cartel crimes are accepted as part of doing business in many foreign firms, although in recent years respondents suggested that European firms have improved their awareness that cartel activity could lead to significant penalties and have begun a process of changing their internal tolerance of such activity. This is far less the case for Asian firms. The perception by society that illegal acts are immoral may also create increased deterrence within the firm based on a procompliance culture. There are social costs for individuals for wrongdoing, such as stigma (Rasmusen 1996). These costs amount to shaming penalties. These issues implicate corporate culture, which addresses issues of norms and customs (Hermalin 2012). Through more effective use of moral shaming, norms can be changed within companies and society at large. This can be done through changing incentives, such as highlighting the ethical value of compliance. This both decreases the cost of detection, because others will be on the lookout, and raises the potential cost to participation in illegal activity. This happens because those who might try to engage in illegal activity will see that it will hurt them given that such behavior will not be tolerated within the company. The more that people within the company who view cartel behavior and other white-collar crimes as on par with capital crimes, the greater the moral outrage that others will feel toward the perpetrators of such crime and the more stigma that will attach to the perpetrator. The mere threat of such stigma should be sufficient to deter some individuals from participating in cartel activity.

23.4.2.  Culture and Stock-Based Cartel Event Studies Stigma may be felt at the company level in terms of negative stock market returns due to the loss of “branding.” Reputation effects (via a drop in stock price) may have a significant effect on the decision whether or not to undertake crime. The deterrent effect may be even greater than that of jail or of civil penalties. This work has its origins in Klein and Leffler (1981). More recent empirical work in other areas of regulation suggests two possible reputational effects. Finance empirical work shows that the reputational loss for financial misreporting and fraud is greater than the penalties via the legal system (Karpoff and Lott 1993; Karpoff, Lee, and Martin 2008a; Engalin 2012). In contrast, in the environmental realm, the legal penalties imposed matter far more for optimal deterrence than do reputational penalties (Karpoff, Lott, and Wehrly 2005). Antitrust studies of reputational effects on stock performance have been limited and the results mixed. Motchenkova and van der Laan (2005) provide a theoretical model that generates reputational loss for large firms across jurisdictions for cartel activities. Bosch and Eckard (1991) examined the stock returns for firms involved in price fixing during the period 1962–1980. Their analysis suggested that the majority of the loss in stock value was due to the readjustment effect of what had been monopoly

596   Rosa M. Abrantes-Metz and D. Daniel Sokol profits moving to the competitive price. More recent studies have not been undertaken to examines the reputation effect of US cartel enforcement since that time, even though there have been significant changes both to enforcement (the introduction of the leniency program) and to the establishment of the importance of cartel enforcement. A study on Dutch listed firms that were the target of competition actions of both cartel and dominance cases from 1998 to 2008 as well as EC-level cases concerning Dutch cartels from the same time period finds that the reputational loss, taking into account the legal costs and readjustment effect of the profit stream, is a more significant deterrent than the legal effect (Van den Broek et al. 2012). Though information about a massive Dutch construction cartel was already publicly available (and financial penalties already built into the stock price), one study found that after a television show about the Dutch construction cartel appeared, the stock price of firms mentioned in the television show fell by 10%, which suggests a reputational penalty (Graafland 2004). At the EU level, the most recent work that examines the impact of cartel compliance (most of the observations) is by Langus, Motta, and Aguzzoni (2013) and Günster and van Dijk (2010). Langus, Motta, and Aguzzoni found that fines account for between a quarter and a third of the stock price drop, with the remainder due to a readjustment effect and almost none to potential reputation effects. Gunster and van Dijk find a 5% stock return around the time of the dawn raid and a 2% drop at the time of a final decision. They do not distinguish the loss among fines, the readjustment effect of the profit stream, and reputational effects. Other antitrust-related work on the impact of cartel enforcement on stock returns focuses on the long-term effect of stock prices. This line of work suggests that from a temporal perspective, companies might be willing to take a reputation hit because of a more limited effect of cartel enforcement. Thompson and Kaserman 2001, using the Bosch and Eckard data set, found that 85% of firms in their sample (on a market-adjusted basis) returned to the preindictment stock price within a year of the antitrust action. A more recent paper by Detre and coauthors (2005) uses a newer data set (1981–2001) and finds that the stock price returned to the preindictment level in a time period similar to that of Thompson and Kaiserman. To our knowledge, no study has examined the long-term effect of stock prices focusing only on the period since the introduction of the modern US leniency program.

23.4.3.  Compliance Culture Some of the lack of stigma for cartel cases is more directly tied to a poor compliance culture by cartel member firms. To date, there has not been a comprehensive study of the consequence for cartel participants akin to the work regarding managers involved in financial misrepresentations. Karpoff, Lee, and Martin (2008a) tracked the outcomes of 2,206 individuals for all 788 SEC and DOJ financial representations from 1978 to 2006. The one antitrust work on this topic, Connor and Lande 2012, reveals alarming statistics about the acceptance of illegal cartel behavior, although only with a small percentage

Antitrust Corporate Governance and Compliance   597

(34%) of a total of 103 managers who went to jail for cartel crimes between 1995 and 2010. Of that group of 25, nine (26%) remain employed by the same company for which they went to jail for their cartel activity, and another nine (26%) work for a different firm but within the same industry. Where there are no explicit restrictions (such as through the terms of a company’s plea agreement) to rehire convicted felons (as cartel convictions are felonies) who participated in cartel activities, social shaming could increase on some level the cost of participating in such activity. US antitrust is different from other areas of related law (such as securities law), where convicted (or civilly sanctioned) offenders with some frequency are barred from the industry or where companies can be debarred from doing business with the government. In the United Kingdom, there is director debarment for cartel offenses. However, the effect of director debarment on firms has not been studied empirically in detail, largely because there has only been one debarment. Blowing the whistle on cartel crimes requires employees to feel empowered within the organization. Employee incentives may not be aligned with the firm in terms of compliance because an employee or midlevel manager risks losing her job if she comes forward with information of illegal activity (Dyck, Morse, and Zingales 2010). Thus, in many cases, the cost of informing outweighs the benefit of remaining silent. The misalignment of incentives between employees and firm replicates itself in the cartel context. Where the cartel compliance culture at a firm is weak, there is little incentive for employees to come forward to report on others within the organization. Scholarship on the effectiveness of cartel compliance in the United States suggests that antitrust compliance programs are not effectively integrated within firm culture. Instead, only a select group of managers (such as at the senior management and general counsel level) understand the importance of antitrust compliance, whereas much of midlevel management and employees do not seem sensitive to the importance of such compliance generally, seem to forget their training; or seem insensitive to the particular nuances of what type of collaboration across competitors is illegal (Sokol 2012). Subsequent generations of employees and managers get trained by their more senior colleagues in industry practices and thereby pass on company and industry norms. In many cases entire industries become recidivists when a generation retires and the next generation relearns how to coordinate with competitors (Levenstein and Suslow 2006).

23.5.  Survey Evidence on Compliance The question of whether or not compliance is effective in embedding antitrust knowledge and to what extent this knowledge deters activity among individuals within a given firm remains unknown except through survey evidence. Some of this is quantitative and some is qualitative. A number of articles and reports have surveyed practitioners, whether lawyers both in-house and at law firms or businesspeople more generally.

598   Rosa M. Abrantes-Metz and D. Daniel Sokol Parker 2011a provides a review of such studies up to 2009. Studies in the UK have surveyed companies about competition compliance and awareness. These include both academic studies (Rodger 2000, 2005, 2009) and government studies (OFT 2007, 2011). Rodger’s studies involved surveys that tracked compliance with the UK Competition Act of 1998, including detailed analysis of some particular companies’ compliance efforts. The 2007 Office of Fair Trading study found that both businesses and their lawyers were most concerned with criminal sanctions for cartel offenses. The study also listed the most common compliance methods and noted that larger businesses tended to be the most compliant. The 2011 OFT study involved 2,009 telephone interviews across seven different industries in the United Kingdom. Compliance seemed to be uneven across all sectors. For a significant minority of firms, there is knowledge of competition law, but still no compliance. Larger companies are more likely to be aware of competition law than smaller firms. In the United States, there have been two studies that survey antitrust lawyers on compliance issues (Beckenstein and Gabel 1983; Sokol 2012). The Beckenstein and Gabel study surveys members of the ABA Section of Antitrust Law to understand compliance by its members. The Sokol study interviews antitrust practitioners (both members of the ABA Section of Antitrust Law as well as Chambers-ranked US cartel practitioners) in both quantitative and qualitative surveys and finds certain structural limitations in organizational behavior within firms that have prevented antitrust compliance programs from becoming embedded in a way that would reduce cartel activity. Twenty Swiss antitrust lawyers were the subject of an article (Hüschelrath, Leheyda, and Beschorner 2011) that illustrates the deterrent effects of enhancements to Swiss cartel policy—increased fines and increased evidence gathering. The article attempts to get at what sort of effect these enhancements have had on foreign companies as well as smaller and medium-sized Swiss companies, some of which have limited or no international exposure. The study noted the difficulty of getting top management to be involved in the successful creation and implementation of a compliance program. In Australia, Christine Parker has been the author or coauthor of a number of studies on competition compliance. Overall she and her coauthors find that cartel enforcement has changed social norms in Australia but that some compliance is only halfhearted or has certain implementation problems (Parker and Nielson 2006; Parker, Ainsworth, and Stepanenko 2004; Parker and Platania-Phung 2012). Some cartel studies that implicate cartel culture and effectiveness of compliance involve interviews with cartel participants. The most famous such work is the study by Geis (1967) of executives involved in the heavy electrical equipment cartel. He quotes industry leaders who said that they committed no wrongdoing because the activity was viewed as legitimate in the industry. That is, executives in their view were conforming with, not breaking, the rules. Herling 1962 provides a robust overview of the same conspiracy and its members. Similarly, Baker and Faulkner (1993) reviewed the court records from the heavy electrical equipment cartel to understand the motivations of individual cartel members. According to their analysis, the cartel was built around nondetection rather than efficiency of the cartel. John Fuller (1962) provides a journalistic

Antitrust Corporate Governance and Compliance   599

account of these court proceedings to understand the internal mechanisms of the cartel and its members. Although no cartel members were interviewed by Genesove and Mullin (2001), they reviewed weekly notes from the 1927–1936 sugar refining cartel. They explored the mechanisms of the cartel creation process across firms, including how individuals at the firms coordinated with each other in full knowledge, and the reasoning that underpinned the actions of particular firms and the individuals who represented them at the cartel meetings. The folding carton cartel of the 1970s is the subject of work by Sonnenfeld (1981) and Lawrence and Sonnenfeld (1978). These works provide detailed interviews with executives of companies that were involved in the cartel to understand the motivation for the involvement. A more recent study by O’Kane (2011) interviews a single cartel member in the marine hose conspiracy. Australian qualitative interview research is rich. A number of studies examine cartels, their individual participants, and their outside lawyers (Parker, Ainsworth, and Stepanenko 2004; Parker 2006, 2012a, 2012b; Parker et al. 2011).

23.6.  The Use of Bounties as Incentives to Induce Detection by Individuals The preceding discussion on compliance and culture suggests that there must be some level of trust based on the norms and culture of a firm. Leniency can destabilize this trust. However, most of the success of the leniency program around the world has been with leniency applicants at the firm level. Individuals rarely have been leniency applicants. Creating mistrust within firms could lead to the breakdown of cartels. Yet blowing the whistle on illegal behavior, such as cartels, is not easy. To do so presents significant risks. As Dyck, Morse, and Zingales (2010) note, employees who blow the whistle face significant costs in their companies for doing so. Bowen, Call, and Rajgopal (2010) correlate whistle-blowing to situations where (1) firms tend to be large companies with significant goodwill but with poor governance; (2) firms tend to have blurry lines of communication within the firm, in part because of personnel changes; and (3) these tend to be concentrated in industries in which there are significant government purchases. Similarly, there is little incentive for employees to come forward for cartel information without a reward since cartel compliance cultures at many firms are weak (Sokol 2012). Detection may be increased through individual financial incentives within the firm of blowing the whistle. An antitrust whistle-blowing bounty has been discussed in a few articles (Aubert, Rey, and Kovacic 2006; Kovacic 2001; Spagnolo 2008). Whistle-blowing is not foreign to the US enforcement scheme outside of antitrust. The qui tam statute enables whistle-blowers to receive between 15% and 30% of the money recovered by the government (Kovacic 2001).

600   Rosa M. Abrantes-Metz and D. Daniel Sokol A bounty may work in conjunction with leniency to improve detection, as the South Korean experience suggests (Spagnolo 2008; Sokol 2011). What has not been established is the impact of whistle-blowing on the internal dynamics of companies that have been the subject of antitrust whistle-blowing, or of multinationals that might do business in the United Kingdom or Korea, such as the breakdown of trust within the firm. Moreover, cartel bounties may create high administrative costs because employees may overreport information. Business may become paralyzed if managers fear that every possible decision might subject them to discipline internally or from antitrust authorities. These drawbacks have not yet been quantified or thoroughly modeled within the antitrust scholarship.

23.7.  The Use of Empirical Screens in Antitrust Compliance and Their Relevance for Corporate Governance Despite their success in many respects, antitrust compliance programs seem to play a minor role in detecting and possibly deterring antitrust violations relative to other detection methods. Why may that be the case? There are a variety of reasons. At the top is that key jurisdictions (e.g., the United States and the European Union) do not offer credit to corporations for their compliance programs in case antitrust violations are found. Were such credit offered, it would increase the incentive for corporations to enhance their compliance programs. This section explores the potential use of screening methodologies to improve antitrust compliance. It begins with an overview of empirical screens and detection of illegal behavior externally to the corporation. It then discusses the use of screens for internal monitoring and compliance. The ability to flag unlawful behavior through economic and statistical analyses is commonly known as screening. A screen is a statistical test based on an econometric model and a theory of the alleged illegal behavior, designed to identify whether manipulation, collusion, fraud, or any type of cheating for that matter, may exist in a particular market, who may be involved, and how long it may have lasted. Screens use commonly available data such as prices, bids, quotes, spreads, market shares, volumes, and other data to identify patterns that are anomalous or highly improbable. Over the last few years, economic analyses in general, and empirical screens in particular, have become increasingly important in cases of conspiracies and manipulations, a trend detailed in Abrantes-Metz and Bajari (2009, 2010), Harrington (2006, 2008), Hüschelrath (2010), and Laitenberger and Hüschelrath (2011). Competition authorities and other agencies worldwide have begun using screens to detect possible market conspiracies and manipulations, and so have defendants and plaintiffs at initial stages of litigation.

Antitrust Corporate Governance and Compliance   601

Broadly speaking, screens used in the literature employ two strategies. The first is to search for improbable events. This type of screen is similar to looking for a cheat in a casino. For example, the probability that a gambler at a Las Vegas casino will place a winning bet in roulette is roughly 0.5%. During her shift, a roulette dealer may see a handful of players win five, or even seven, times in a row. However, the probability of winning 20 times in a row is almost zero (though not impossible). If a pit boss sees this occur, he may not be able to prove that cheating has occurred, but he would be well advised to watch closely or risk losing a lot of money. One set of collusive screens generalizes this idea by looking for events that are improbable unless firms in the industry have coordinated their actions. The second type of screen uses the idea of a control group. A  somewhat extreme example illustrates the idea. In the 1980s, organized crime in New York City operated a concrete club that rigged bids on contracts over $2 million. During the 1980s, the price of concrete was 70% higher in New York City than other US cities. While it is true that the price of many goods and services is higher in New York City, relatively few of those prices are 70% higher than in other large cities. Prices that are anomalous compared to other markets suggest a competition problem. In this simple example, we form a control group for New York by using prices in other cities as a basis for comparison. Most collusion is not this blatant. When designing and implementing screens, there are two golden rules to keep in mind, which should be obvious when stated, but which may sometimes be forgotten. First, “One size does not fit all,” and second, “If you put garbage in, you get garbage out.” Screens can be very powerful, but these are econometric tools with all the usual caveats, and they may potentially be misused. Screens, just like any empirical technique, can be effective only when properly applied; otherwise they risk producing nonsense. See the discussion by Doane and coauthors in this volume. The first screening rule states that a screen needs to be designed or at least adjusted to the situation at hand. Just because a given set of variables and model specifications proves highly effective when estimating the demand for bread does not mean that those same variables or specifications work when estimating the demand for cars. In turn, the second screening rule states that, as is always the case in empirical work, a screen is only as good as the choices of what is put into it (Abrantes-Metz (2012) discusses in further detail issues related to the development of screens). In general, six requirements are key to developing and implementing a good screen: (1) an understanding of the market at hand, including the nature of competition and the potential incentives to cheat—both internal and external to a firm; (2) a view of the likely nature of cheating; (3) a view of how cheating will affect market outcomes; (4) a set of statistics that can capture both the implications of cheating as well as ordinary, natural relationships between key market variables; (5) empirical or theoretical support for the screen; and (6) the identification of an appropriate nontainted benchmark against which the evidence of cheating can be compared. Even a screen based on a solid theory of cheating and properly designed and implemented can still produce erroneous conclusions, just as is the case with any other

602   Rosa M. Abrantes-Metz and D. Daniel Sokol statistical test: it may indicate that cheating may have existed where one did not (type I error), or it may fail to flag cheating which did exist (type II error). Again, just as we would not argue that statistical tests are useless because they have margins of error, we should apply the same standards to screens. The hope is that type I and type II errors will not occur with high likelihood, though there is certainly a trade-off between the two.

23.7.1.  Types of Screens, Their Users, and Successes in the Detection of Illegal Behavior There are different types of screens used in a variety of markets, and this section reviews some of the most commonly applied. It starts by focusing on screens developed to detect bid rigging, a popular application of these methods since the strict rules of competitive bidding help to identify colluders. Next, it discusses screens when only price data are available, or some measure of cost is also available, including variance screens, which search for pockets of high or low variances in prices. Screens based on market shares are discussed next, followed by purely mathematical screens such as Benford’s law, which describes the rates at which certain digits occur in many data sets. Though these screens have been applied at the industry level, there is reason to believe, based on the more regular use of screens in areas such as accounting fraud and FCPA, that firms can undertake cartel-related screening as a form of monitoring its own employees.

23.7.1.1.  Bid Rigging Bid rigging in competitive tenders is a productive setting to apply screens for three reasons. First, competitive tenders are widely used not only for public sector procurement but also in financial markets, privatization of public assets, real estate, and many other transactions. Second, bid rigging is a common antitrust offense, representing a significant portion of all international cartels uncovered and associated with long-lasting cartels (Abrantes-Metz, Connor, and Metz 2013). Third, markets that use competitive bidding are frequently rich in data. In many countries, statutes require the public disclosure of bids. There is a large body of empirical literature on collusion in auctions that implements various types of screens. While these papers span a wide variety of industries, researchers have identified common patterns when collusion is known or suspected. In sealed auctions, firms submit their bids simultaneously, and these are often read at a previously set date. In the public sector, the lowest bidder is usually awarded the contract. When firms do not know about each other’s bids, they cannot take them into consideration when determining their own bid. Hence, each individual bid will likely primarily reflect market conditions such as costs and local market power. As a result, competitive bidding implies that bids will be independent after controlling for information that is observed by all of the bidders. When firms collude, by contrast, they need to coordinate their actions. This coordination is another common factor across bids of members that is observed or known to all

Antitrust Corporate Governance and Compliance   603

of them, and extends beyond the common factors given by costs and market power. This additional common factor is expected to increase the correlation across bids, as implied by an effective coordination. Therefore, collusive bids are highly correlated, even after controlling for costs and market power variables. The question then becomes “How much positive correlation across bids is high enough to raise suspicion of collusion?” The answer is “It depends.” Sometimes the correlation is so high that the likelihood that it could have been attained without explicit coordination is almost zero. A famous example is the bids received by the Tennessee Valley Authority to install conductor cables in the 1950s, in which seven firms submitted (presumably sealed) identical bids of $198,438,24. It is highly unlikely that independent bidders could have arrived at bid values identical to the eighth significant digit if acting independently. Other times, the correlation across bids is significantly higher across bidders in one particular market, say market A, than across bidders in another particular comparable market, say market B. If such difference in correlations between the two markets cannot be explained by a relevant legitimate market condition observable to firms in A but not observable to firms in B, then it is more likely that it can be explained by explicitly coordinated behavior across bidders in market A. But if the decision to submit bids is independent, then such a high degree of correlation is too large and too persistent to be likely only due to randomness or idiosyncrasies. This type of pattern was illustrated in an examination of a set of bids submitted to supply school milk in Ohio between 1980 and 1990, set forth by expert reports for the prosecution on behalf of State of Ohio v. Louis Trauth Dairies, Inc., in which several patterns in the data were found too improbable to have been reached by a competitive bidding system (Porter and Zona 1999).2 Economic theory predicts that bids should closely reflect costs in competitive markets. When firms collude, they break the relationship between bids and costs, with the objective of attaining higher than competitive profits. Therefore, another set of bid-rigging screens looks for disconnects between bids and costs, which can be evaluated, for example, by comparing the relationship between bids and costs in the allegedly tainted market, against that of the untainted markets. In contrast with the example of the price of concrete in New York discussed earlier, an examination of bids by highway contractors in the upper Midwest during the 1990s tended to disprove the existence of conspiracy (Bajari and Ye 2003). When computing the ratio of the winning bid to the cost estimate, the authors found it was almost equal to one, suggesting that bids were comparable to properly deflated bids from other markets, and representing evidence consistent with competitive markets for most of the bids. Nevertheless, the authors also found evidence consistent with collusion for two out of the 11 firms studied, the same two firms that had been previously sanctioned for bid rigging. 2  Other studies of the same nature have also been performed in markets suspected of collusion. Those include paving contracts (Porter and Zona 1993), Canadian timber (List et al. 2004), and Russian oil and gas leases (Marshall and Marx 2009).

604   Rosa M. Abrantes-Metz and D. Daniel Sokol In addition to the work of academics and consultants as discussed above, the US Department of Justice Antitrust Division has proposed a number of bidding patterns as being suggestive of collusion. All of these either look for bids that appear too improbable to have been independently submitted or for bids that do not reflect costs and other legitimate market factors in a way compatible with competition. Red flags for bid rigging suggested by the US Department of Justice are the following:3

1. The same company always wins a particular procurement. This pattern may be more suspicious if one or more companies continually submit unsuccessful bids. 2. The same suppliers submit bids and each company seems to take a turn being the successful bidder. 3. Some bids are much higher than published price lists, previous bids by the same firms, or engineering cost estimates. 4. Fewer than the normal number of competitors submit bids. 5. A  company appears to be bidding substantially higher on some bids than on other bids, with no apparent cost differences to account for the disparity. 6. Bid prices drop whenever a new or infrequent bidder submits a bid. 7. A successful bidder subcontracts work to competitors that submitted unsuccessful bids on the same project. 8. A company withdraws its successful bid and subsequently is subcontracted work by the new winning contractor.



When studying whether bidding patterns are the product of collusive agreements, investigators should realize that the failure to control for relevant components of costs or competitive factors will likely provide erroneous evidence that collusion existed and/ or that it materially impacted competition.

23.7.1.2.  Price Fixing Successful collusion will cause prices to be higher than the noncooperative oligopoly prices that would prevail absent collusion. With successful collusion, prices are also likely to be less volatile. Finally, prices may behave in otherwise unusual ways when collusion is initiated, when there are episodes of cheating and punishment (e.g., price wars), and when collusion ends. But there may be alternative explanations for unusual patterns that may warrant a thorough investigation. In a 2006 paper, Abrantes-Metz and coauthors propose a price-fixing screen based on low price variance. The authors used prices and costs of frozen perch fillets purchased by the Defense Personal Support Center between 1987 and 1989. Data extended from a collusive agreement, through its break, and the move into competition. During the first 3 

The US Department of Justice has identified a set of pricing patterns that are intended to help identify collusive behavior. See US Depart of Justice, Antitrust Division, An Antitrust Primer: Price Fixing, Bid Rigging, and Market Allocation Schemes: What They Are and What to Look For, at http://www.​ usdoj.gov/atr/public/guidelines/primer-ncu.htm (“DOJ Antitrust Primer”).

Antitrust Corporate Governance and Compliance   605

period there was collusion in the form of bid rigging among sellers of frozen perch fillets that led to the fixing of prices. This behavior was uncovered by the Department of Justice and at the time such investigation became known, the cartel broke apart. The authors illustrated four expected features of collusion:

1. The average price was higher with coordination than with competition, after taking account of differences in the cost of fresh perch.4 2. Prices dropped suddenly when the conspiracy collapsed. 3. Prices were more stable, and less responsive to demand and supply shocks with conspiracy than under competition. 4. Prices followed costs more closely when competition prevailed than when the conspiracy was in effect.5



These above features and others are used by competition authorities around the world to assist in the detection of price-fixing conspiracies, and in particular, feature 3 above relating to low price variance under collusion. Red flags for price rigging suggested by the Department of Justice are the following:

1. Prices remain identical for long periods of time. 2. Prices previously were different (and start becoming identical). 3. Price increases do not appear to be explainable by increased costs. 4. Discounts are eliminated, especially in a market where discounts historically were given. 5. Vendors are charging higher prices to local customers than to distant customers.



23.7.1.3.  Market Shares Another type of screen uses data on market shares. The literature and evidence from uncovered cartels demonstrate that these may attempt to collude by fixing market shares (Harrington 2006, 2007). Two screens are suggested by the literature: (1) market shares appear to be too stable over time; and (2) market shares for all firms in a particular market are negatively correlated over time. The first screen will detect an agreement by the cartels to divide the market. Examples of cartels with stable market share agreements include cartels in copper plumbing tubes, organic peroxides, and several vitamins (A, E, and folic acid, in particular). The second screen is suggested by dynamic models of collusion (Athey and Bagwell 2001 and Athey, Bagwell, and Sanchirico 2004). In these models, if a cartel member deviates from the collusive agreement, it will need to compensate other cartel members in subsequent time periods. As a result, abnormally high shares for a particular firm in one period should be followed by a reduction in shares in the following period. 4 

According to some estimates, the average cartel raises prices by 25%. See Connor (2007). Kim and Cotterill (2008) found that under collusion the pass-through rate for processed cheese was 21% to 31% while under competition the range was 73% to 103%. 5 

606   Rosa M. Abrantes-Metz and D. Daniel Sokol

23.7.1.4.  Screens Based on Mathematical Laws Benford’s law is a mathematical formula that describes the regularly occurring distribution of digits in many data sets. Studies have shown that the law applies to many diverse data sets, including populations of cities, electricity usage, word frequency, and the daily returns to the Dow Jones. Because Benford’s law is a naturally occurring pattern in those data, violations of the law can be used to detect irregularities. In the past, violations have been used to detect data tampering, manipulation of financial ratios, and tax evasion. For example, Varian 1972 uses Benford’s law to check the validity of purported scientific data in social sciences, Nigrini 1996 applies to taxpayer evasion, while Judge and Schechter 2009 usesthis law to detect problems in survey data. Most recently, Abrantes-Metz, Villas-Boas, and Judge (2011) used Benford’s law to test for unusual patterns in USD LIBOR, a matter now commonly known as the LIBOR conspiracy and manipulation. Other collusive markers used to screen for conspiracies are discussed in Harrington 2006.

23.7.1.5.  Screens’ Successes Economic analysis and empirical screening do trigger antitrust cases, such as an Italian cartel in baby milk and a Dutch cartel in the shrimp industry. Screens are also being successfully used to identify potential anticompetitive behavior in gasoline markets by the US Federal Trade Commission, and to prioritize complaints in the Brazilian gasoline retail market, leading to raids and the finding of direct evidence of collusion (Ragazzo 2012). In Mexico, the competition authority flagged a conspiracy in pharmaceutical markets through the use of bid-rigging screens (Mena Labarthe 2012). Market monitoring and screening programs have been adopted by several other competition authorities, such as the European Commission and the South African Competition Authority. Most recently, per a press article dated September 15, 2012, a Canadian reporter raised the flag for possible bid rigging in the pavement of roads in Quebec, Canada, by pointing to a variety of patterns that seemed unexpected under a competitive setting. Since then, investigations have been launched and senior officials have resigned. In 2013 the Quebec government took direct control over the city of Laval, the target of the investigation, and Quebec’s third largest municipality. The ongoing investigations have intensified in this escalating procurement scandal involving also Laval’s former Mayor. Screens have successfully detected cartels in US financial markets. Academics applied screens to NASDAQ inside spreads (the lowest selling offer minus the highest buying offer among all dealers) in 1994 and found these to be abnormally high when specific stocks were trading only at even and not at odd eighths, thus effectively raising dealers’ profit margins. The abnormal pattern of odd-eighths avoidance was published in May 27, 1994, in a paper by Christie and Shultz. Immediately after its publication, NASDAQ spreads significantly shrank, suggesting (if not revealing) a conspiracy among dealers and triggering regulatory investigations. Most recently, the US Department of Justice, the Securities and Exchange Commission, the Commodities Futures Trading Commission and other regulatory

Antitrust Corporate Governance and Compliance   607

agencies worldwide are investigating the possibility of a conspiracy to manipulate the US dollar LIBOR (“LIBOR”) by several major banks, as well as LIBOR denominated in other currencies, and similar benchmarks including Euribor in Europe and TIBOR in Japan. These investigations followed the application of empirical screens by the Wall Street Journal in April and May 2008 and by Abrantes-Metz and coauthors in August 2008. We will focus on this example in the last section of this chapter.

23.7.2.  Screens in Internal Monitoring As useful as screens have proven to be in the detection of illegal behavior by entities external to the corporation, the same tools used to provide circumstantial evidence of possible wrongdoing can be just as well internally applied by corporations (Abrantes-Metz and Bajari 2009). For example, there is in principle no reason why screens could not have detected the NASDAQ odd-eighths avoidance of inside spreads by NASDAQ itself, had these been internally used to monitor spreads by the exchange. Similarly, screens could have been used by LIBOR submitters to monitor possible wrongdoing in these submissions and potentially have avoided the exposure and liability currently faced by these institutions. Screens can have a variety of other specific applications internally to corporations. They could be used by managers to monitor for fraud in accounting and reimbursement statements, collusion on employee compensation surveys, or other forms of data manipulation. Furthermore, screens could be used to detect price fixing in purchasing or procurement. Next we will describe some potential applications below. However, note that the methods we have discussed in the section below can be very powerful and applied to detect a much wider range of attempts by employees or suppliers to manipulate data. A first example is the manipulation of financial statements or other forms of accounting fraud. In many industries, managers are under intense pressure to produce revenues and earnings that meet or exceed analysts’ forecasts. This is particularly true for managers with compensation that is heavily dependent on stock options. If the companies fail to meet these targets, even by a small amount, share prices may fall considerably. The screens discussed above could be useful in identifying this sort of fraud. For example, if a company is falsifying accounting statements not to miss analysts’ forecasts, it may smooth these out over time, and in this way produce revenues and earnings that are less volatile than those of comparable companies. This suggests using a variance screen on revenues, earnings, and stock prices. Additionally, Benford’s law could be applied to the accounting statements and share prices used to detect falsified accounting statements. In fact, this law has been used by accountants for several decades to flag potential data manipulations. As Gonzalez and coauthors (2013) have noted, cartel behavior might lead to income smoothing; hence screens that examine accounting irregularities could detect such manipulation based on cartel activity. Screens can be used to detect bid rigging or price fixing in purchasing or procurement. Many institutional bodies, like the federal government, may rely on a large network of

608   Rosa M. Abrantes-Metz and D. Daniel Sokol suppliers. Manufacturing companies, big box retailers, and HMOs are all examples of firms that may rely on a network of thousands of suppliers. Many companies use a competitive process to keep supplier margins competitive. However, suppliers often operate in concentrated industries and may be tempted to collude in order to earn profits above the competitive level. The methods we have described above could be applied to screen for outright supplier collusion or excessive supplier costs. A recent study by Hüschelrath and Veith (2011) illustrates the power of these techniques if applied internally to corporations. In particular, this study provides evidence on how screens, had these been used internally by buyer companies, could have detected the explicit collusive agreement in the upstream German cement markets before such a cartel was identified by the German Competition Authority. The authors use a data set with 340,000 market transactions from 36 smaller and larger customers of German cement producers, and apply a screen that searches for structural changes in prices paid by buyers, which could not be explained by legitimate market conditions. This application correctly identified the ongoing cartel (ex post), establishing the power of screens applied by buyers to internally detect this illegal behavior, and well ahead of its uncovering by the German Competition Authority. After having described how screens can be used internally to corporations to flag the possibility of illegal behavior, in the next subsection we turn to why, when, and how they should be used in this setting to enhance antitrust compliance.

23.7.3.  Why, When, and How to Use Screens in Antitrust Compliance The history of major international cartels, and to some extent that of smaller local conspiracies, suggests that compliance training, while a necessary tool, is not sufficient. Screens in the compliance setting may assist in identifying the high-risk areas of a business, allow better targeting of audits to those areas, and assist in the monitoring of these risks in a more efficient way. They employ techniques designed to highlight which parts of the company merit closer scrutiny, where there should be intensive reviews, and which units may call for intensive monitoring of internal communications and other direct actions. Screens can fulfill this role by looking at certain quantifiable red flags and applying statistical analysis to determine the priority areas for further focus, and in this way allow a more efficient allocation of resources. To date, there is a lack of theoretical work that links the use of antitrust screens with compliance to solve agency cost problems. A firm would use screens to be the first to detect any possible wrongdoing so it could fully benefit from, for example, leniency applications in the case of cartel behavior relative to other cartel members. The first consideration on whether to use screens for internal detection of wrongdoing is data availability. What types of data are available and for how long? Additionally, can other data be collected so that screens can be applied moving forward? Other considerations include the following: Has the company been involved in a conspiracy before? Is

Antitrust Corporate Governance and Compliance   609

this an industry where competition problems tend to exist, (i.e., history of violations or an industry with other characteristics such as overcapacity)? Are there several opportunities to rig bids and reach agreements with competitors (i.e., are trade association meetings and other industry gatherings common)? When several of these conditions are satisfied (and potentially others as well), then the company should engage in internal screening. This process does not have to apply to all areas within the company, simply to those where cheating is more likely, but expertise will be required in the development and implementation of these tools. As explained earlier and in more detail in the last section with respect to LIBOR, there have been multiple examples of entities external to the company detecting cartels through the use of screens. In terms of cartel detection through internal compliance programs and the use of screens, though, we are not aware of any example. There are three points that should be stressed: first, the successful external screens just mentioned could have, in every case, been developed internally first, meaning there is no a priori reason an aggressive compliance screening program would not be successful; second, corporations are not yet employing screens in their compliance programs, so there are not many chances for successful detections; and third, a strong compliance screen may deter inappropriate behavior, and such successes by their nature could never be identified.

23.7.4.  LIBOR Detection through Screening and Its Relevance for Corporate Governance There is a growing literature on the use of screens and detection regarding LIBOR. Such behavior was first flagged through screening by the Wall Street Journal (WSJ) in April and May 2008 and by Abrantes-Metz and coauthors (2008). The WSJ noted that LIBOR submissions seemed too low with respect to submitters credit default swaps after the beginning of the financial crisis. Abrantes-Metz and coauthors (2008) noted that LIBOR was constant day in, day out for many months prior to the financial crisis, while other comparable rates such as the Federal Funds Effective Rate were not; most banks’ quotes were identical to each other, while other market indicators, such as their pricing in the capital markets, indicated differences (even if slight) across banks; LIBOR did not respond to increasing risk at least since late spring and early summer of 2007; the authors saw signs of possible collusion dating back to prior to financial crisis.6 Other research on LIBOR has been conducted by Snider and Youle (2009, 2010), by Abrantes-Metz, Villas-Boas, and Judge (2011), and by Abrantes-Metz and Metz (2012). Snider and Youle present findings supporting their claim that the banks’ LIBOR quotes are difficult to rationalize by observable cost measures for the period under study, including a given bank’s quotes in other currency panels. They also introduce a model in which banks’ possession of LIBOR indexed contracts induces them to produce LIBOR quotes that are clustered around discontinuities, and using this model the authors show 6 

This study was later published in 2011.

610   Rosa M. Abrantes-Metz and D. Daniel Sokol that there was a severe clustering in the US dollar LIBOR for the three-month maturity throughout 2009. Snider and Youle further try to quantify gains from such behavior and present results showing large exposures to LIBOR by several banks through their interest rate derivative portfolios, allegedly enabling to profit from the rapid decline of LIBOR starting in late SUMMER 2007. The authors argue that these exposures may be the incentive behind a deliberate misreporting of LIBOR quotes by the banks. Abrantes-Metz, Villas-Boas, and Judge (2011) show that Benford’s law was violated for the US dollar LIBOR most noticeably from early 2006 through the summer of 2007, but with continuing anomalous patterns throughout 2009. Abrantes-Metz and Metz (2012) followed with an analyses similar to bid rigging to explain that, given that LIBOR quotes are submitted sealed, that they are supposed to be idiosyncratic to each of the banks as forecasts of their own interbank borrowing costs, and that, though all very highly rated, these banks are not completely identical to each other in terms of portfolios, exposures to different markets, and other differentiating factors, simultaneous moves by a large number of submitters from one day to the following to the exact same submission number are more consistent with explicit rather than with tacit collusion. The literature on LIBOR offers some lessons for antitrust compliance and the use of screens for internal firm detection. There seems to have been unawareness of the clearly anomalous data patterns by benchmark administrators, regulators, and banks management themselves. The auditing procedures at the banks applied to LIBOR submissions were either inexistent or clearly inadequate. Had such internal mechanisms based on screens been in place, banks would have been able to be the first to detect illegal behavior in such submissions, and hopefully to deter future illegal behavior (Abrantes-Metz and Sokol 2012). Research has yet to focus on the role of risk assessment and program evaluation. LIBOR illustrates that significant international cartel activity may occur even in situations where other types of sophisticated internal compliance mechanisms, including screens, such as for insider trading and accounting fraud, may be used. The specific relationship between antitrust compliance and other types of compliance remains understudied in the corporate governance literature. Relatedly, the role of internal detection in the merger-and-acquisition context of acquired (target) firms based on insufficient compliance remains an area for future study.

23.8. Conclusion This chapter provides an analysis of internal firm compliance and detection of antitrust cartel activity. Whereas there has been less antitrust literature devoted to internal firm incentives and compliance, a number of more recent works have incorporated insights from finance, management, and accounting to enrich understandings of antitrust issues. This trend of cross-pollination of ideas and empirical approaches will continue. Regarding policy, insights into internal detection may

Antitrust Corporate Governance and Compliance   611

better inform sentencing policy and the refinement of incentives for cartel detection by agencies around the world. We note that a number of agencies are increasing their reliance on econometric screens. As more agencies do so, firms that value high levels of antitrust compliance may do so as well as a way to better mitigate their antitrust risk profile.

References ABA. 2010a. Antitrust Compliance: Perspectives and Resources for Corporate Counselors. 2nd ed. ABA. 2010b. Proof of Conspiracy under Antitrust Federal Laws. Abrantes-Metz, Rosa. 2011. Design and Implementation of Screens and Their Use by Defendants. Competition Policy International Antitrust Chronicle, September (2). Abrantes-Metz, Rosa, and Patrick Bajari. 2009. Screens for Conspiracies and Their Multiple Applications. Antitrust Magazine 24(1): 66–70. Abrantes-Metz, Rosa, and Patrick Bajari. 2010. Screens for Conspiracies and Their Multiple Applications. Competition Policy International 6(2): 129–44. Abrantes-Metz, Rosa, Patrick Bajari, and Joseph Murphy. 2010. Enhancing Compliance Programs through Antitrust Screening. Antitrust Counselor 4(5). Abrantes-Metz, Rosa, John Connor, and Albert Metz. 2013. The Determinants of Cartel Duration. Working paper. Abrantes-Metz, Rosa, Luke Froeb, John Geweke, and Christopher Taylor. 2006. A Variance Screen for Collusion. International Journal of Industrial Organization 24: 467–86. Abrantes-Metz, Rosa, Michael Kraten, Albert Metz, and Gim Seow. 2012. LIBOR Manipulation? Journal of Banking and Finance 36: 136–50. First draft dated August 4, 2008, and available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1201389. Abrantes-Metz, Rosa, and Albert Metz. 2012. How Far Can Screens Go in Detecting Explicit Collusion? New Evidence from the Libor Setting. Competition Policy International Antitrust Chronicle, March. Abrantes-Metz, Rosa, and D. Daniel Sokol. 2012. Lessons from LIBOR for Detection and Deterrence of Cartel Wrongdoing. Harvard Business Law Review Online 3: 10–16. Abrantes-Metz, Rosa, Sofia Villas-Boas, and George Judge. 2011. Tracking the Libor Rate. Applied Economic Letters: 1–7. Adams, Renee B., and Daniel Ferreira. 2009. Women in the Boardroom and Their Impact on Governance and Performance. Journal of financial Economics 94(2): 291–309. Agrawal, Anup, and Sahiba Chadha. 2005. Corporate Governance and Accounting Scandals. Journal of Law and Economics 48(2): 371–406. Aguilera, R.V., D. Rupp, C.A. Williams, and J. Ganapathi. 2007. Putting the S Back in CSR:  A  Multi-level Theory of Social Change in Organizations. Academy of Management Review 32(3): 836–63. Alexander, Cindy R., and Mark A. Cohen. 1996. New Evidence on the Origins of Corporate Crime. Managerial and Decision Economics 17: 421–35. Alexander, Cindy R., and Mark A.  Cohen. 1999. Why Do Corporations Become Criminals? Ownership, Hidden Actions, and Crime as an Agency Cost. Journal of Corporate Finance 5: 1–34.

612   Rosa M. Abrantes-Metz and D. Daniel Sokol Angelucci, Charles, and Martijn A.  Han. 2011. Monitoring Managers through Corporate Compliance Programs. Amsterdam Center for Law and Economics, Working Paper No. 2010-14. Arlen, Jennifer. 2011. Removing Prosecutors from the Boardroom:  Limiting Prosecutorial Discretion to Impose Structural Reforms. In Anthony Barkow and Rachel Barkow, Prosecutors in the Boardroom: Using Criminal Law to Regulate Corporate Conduct, 62–86. New York: NYU Press. Arlen, Jennifer. 2012. The Failure of the Organizational Sentencing Guidelines. University of Miami Law Review 66: 321–62. Arlen, Jennifer, and Reiner Kraakman. 1997. Controlling Corporate Misconduct: An Analysis of Corporate Liability Regimes. New York University Law Review 72: 687–779. Ashforth, Blake E., and Vikas Anand. 2003. The Normalization of Corruption in Organizations. Research in Organizational Behavior 25: 1–52. Athey, Susan, and Kyle Bagwell. 2001. Optimal Collusion with Private Information. Rand Journal of Economics 32(3): 428–65. Athey, Susan, Kyle Bagwell, and Christopher Sanchirico. 2004. Collusion and Price Rigidity. Review of Economic Studies 71(2): 317–49. Aubert, Cecile. 2009. Managerial Effort Incentives and Market Collusion. TSE Working Paper 19-127. Aubert, Cecile, Patrick Rey, and William Kovacic. 2006. The Impact of Leniency Programs on Cartels. International Journal of Industrial Organization 24(6): 1241–66. Baber, William R., Surya N. Janakiraman, and Sok-Hyon Kang. 1996. Investment Opportunities and the Structure of Executive Compensation. Journal of Accounting and Economics 21(3): 297–318. Bajari, Patrick, and Lixin Ye. 2003. Deciding between Competition and Collusion. Review of Economics and Statistics 85(4): 971–89. Baker, Wayne E., and Robert R. Faulkner. 1993. The Social Organization of Conspiracy: Illegal Networks in the Heavy Electrical Equipment Industry. American Sociological Review 58(6): 837–60. Bainbridge, Stephen M. 2009. Caremark and Enterprise Risk Management. Journal of Corporation Law 34(4): 967–90. Baucus, Melissa S., and Janet P. Near. 1991. Can Illegal Corporate Behavior Be Predicted? An Event History Analysis. Academy of Management Journal 34: 1–28. Beasley, Mark S. 1996. An Empirical Analysis of the Relation between the Board of Director Composition and Financial Statement Fraud. Accounting Review 71: 443–65. Beasley, Mark S. 2000. Fraudulent Financial Reporting: Consideration of Industry Traits and Corporate Governance Mechanisms. Accounting Horizons 14: 441–52. Bebchuk, Lucian, and Jesse Fried. 2004. Pay without Performance: The Unfulfilled Promise of Executive Compensation. Cambridge: Harvard University Press. Bebchuk, Lucian, Yaniv Grinstein, and Urs Peyer. 2010. Lucky CEOs and Lucky Directors. Journal of Finance 65(6): 2363–401. Beckenstein, Alan R., and H. Landis Gabel. 1986. The Economics of Antitrust Compliance. Southern Economics Journal 52: 673–92. Berger, Philip G., and Eli Ofek. 1995. Diversification’s Effect on Firm Value. Journal of Financial Economics 37: 39–65. Bertrand, Marianne, and Antoinette Schoar. 2003. Managing with Style: The Effect of Managers on Firm Policies. Quarterly Journal of Economics 118(4): 1169–208.

Antitrust Corporate Governance and Compliance   613

Bettenhausen, Kenneth, and J. Keith Murnighan. 1985. The Emergence of Norms in Competitive Decision-Making Groups. Administrative Science Quarterly 30(3): 350–72. Black, Bernard S., Brian R. Cheffins, and Michael D. Klausner. 2006. Outside Director Liability. Stanford Law Review 58: 1055–159. Bosch, Jean-Claude, and E. Woodrow Eckard, Jr. 1991. The Probability of Price Fixing: Evidence from Stock Market Reaction to Federal Indictments. Review of Economics and Statistics 73(2): 309–17. Bowen, Robert M., Andrew C. Call, and Shivaram Rajgopal. 2010. Whistle-Blowing: Target Firm Characteristics and Economic Consequences. Accounting Review 84: 1239–74. Buccirossi, Paolo, and Giancarlo Spagnolo. 2008. Corporate Governance and Collusive Behavior. In Wayne D.  Collins, Issues in Competition Law and Policy, 2008, 3:  1219–40. Antitrust Section, American Bar Association. Burhop, Carsten, and Thomas Lübbers. 2009. Cartels, Managerial Incentives, and Productive Efficiency in German Coal Mining, 1881–1913. Journal of Economic History 69(2): 500–527. Burns, Natasha, and Simi Kedia. 2006. The Impact of Performance-Based Compensation on Misreporting. Journal of Financial Economics 79: 35–67. Camerer, Colin F., and Dan Lovallo. 1999. Overconfidence and Excess Entry: An Experimental Approach. American Economic Review 89(1): 306–18. Chatterjee, Arijit, and Donald C. Hambrick. 2007. It’s All About Me: Narcissistic Chief Executive Officers and Their Effects on Company Strategy and Performance. Administrative Sciences Quarterly 52: 351–86. Chen, Ming-Jer, and Donald C. Hambrick. 1995. Speed, Stealth, and Selective Attack: How Small Firms Differ from Large Firms in Competitive Behavior. Academy of Management Journal 38: 453–82. Christie, William, and Paul Shultz. 1994. Why Do NASDAQ Market Makers Avoid Odd-Eighth Quotes? Journal of Finance 49(50): 1813–40. Combe, Emmanuel, and Constance Monnier. 2011. Fines against Hard-Core Cartels in Europe: The Myth of Over-enforcement. Antitrust Bulletin 56(2): 235–76. Connor, John. 2007. Price-Fixing Overcharges: Legal and Economic Evidence. Research in Law and Economics 22: 59–153. Connor, John M., and Robert H. Lande. 2012. Cartels as Rational Business Strategy: Crime Pays. Cardozo Law Review 34: 427–90. Core, John E., Robert W. Holthausen, and David F. Larcker. 1999. Corporate Governance, Chief Executive Officer Compensation, and Firm Performance. Journal of Financial Economics 51(3): 371–406. Daboub, Anthony J., Abdul M.  A. Rasheed, Richard L.  Priem, and David Gray. 1995. Top Management Team Characteristics and Corporate Illegal Activity. Academy of Management Review 20: 138–70. Desai, Mihir A., and Dharmapala, Dhammika. 2006. Corporate Tax Avoidance and High Powered Incentives. Journal of Financial Economics 79(1): 145–79. Desai, Mihir A., and Dhammika Dharmapala. 2009. Corporate Tax Avoidance and Firm Value. Review of Economics and Statistics 91(3): 537–46. Detre, Joshua D., Alla Golub, and John M. Connor. 2005. The Profitability of Price Fixing: Have Stronger Antitrust Sanctions Deterred? Presented at the Third International Organization Conference, April 8–9. Dyck, Alexander, Adair Morse, and Luigi Zingales. 2010. Who Blows the Whistle on Corporate Fraud? Journal of Finance 65(6): 2213–53.

614   Rosa M. Abrantes-Metz and D. Daniel Sokol Dyreng, Scott, Michelle Hanlon, and Edward L. Maydew. 2010. The Effects of Executives on Corporate Tax Avoidance. Accounting Review 85: 1163–89. Efendi, Jap, Anup Srivastava, and Edward Swanson. 2007. Why Do Corporate Managers Misstate Financial Statements? The Role of Option Compensation and Other Factors. Journal of Finance 85: 667–708. Ertugrul, Mine, and Shantaram Hegde. 2008. Board Compensation Practices and Agency: Costs of Debt. Journal of Corporate Finance 14: 512–31. Finkelstein, Sydney, and Donald C.  Hambrick. 1990. Top-Management-Team Tenure and Organizational Outcomes: The Moderating Role of Managerial Discretion. Administrative Science Quarterly 35: 484–503. Fudenberg, D., and J. Tirole. 1995. A Theory of Income and Dividend Smoothing Based on Incumbency Rents. Journal of Political Economy 103: 75–93. Fuller, John G. 1962. The Gentlemen Conspirators. New York: Grove Press. Gallo, J.C., K. Dau-Schmidt, J. L. Craycraft, and C. J. Parker. 2000. Department of Justice Antitrust Enforcement, 1955–1997: An Empirical Study. Review of Industrial Organization 17: 75–133. Geis, Gilbert. 1967. White Collar Crime: The Heavy Electrical Equipment Antitrust Cases of 1961. In Marshall B.  Clinard and Richard Quinney, Criminal Behaviour Systems:  A  Typology. New York: Holt, Rinehart and Winston, 139–51. Genesove, David, and Wallace P. Mullin. 2001. Rules, Communication, and Collusion: Narrative Evidence from the Sugar Institute Case. American Economic Review 91(3): 379–98. Graafland, Johan J. 2004. Collusion, Reputation Damage and Interest in Code of Conduct: The Case of a Dutch Construction Company. Business Ethics: A European Review 2–3(13): 127–42. González, Tanja Artiga , Markus M. Schmid, and David Yermack, Smokescreen: How Managers Behave When They Have Something to Hide (November 25, 2013). NYU Working Paper No. 2451/31732 Günster, Andrea, and Mathijs A. van Dijk. 2010. The Impact of European Antitrust Policy: Evidence from the Stock Market. Working paper, ETH Zurich / Erasmus University Rotterdam. Hambrick, Donald C. 2007. Upper Echelons Theory:  An Update. Academy of Management Review 32: 334–43. Harrington, Joseph. 2007. Behavioral Screening and the Detection of Cartels. In Claus-Dieter Ehlermann and Isabela Atanasiu, European Competition Law Annual 2006: Enforcement of Prohibition of Cartels, 51–68. Oxford: Hart Publishing. Harrington, Joseph. 2008. Detecting Cartels. In Paolo Buccirossi, Handbook of Antitrust Economics, 259–304. Boston: MIT Press. Harvey, Campbell R., Karl V. Lins, and Andrew H. Roper. 2004. The Effect of Capital Structure When Expected Agency Costs Are Extreme. Journal of Financial Economics 74: 3–30. Haunschild, Pamela R., and Anne S. Miner. 1997. Modes of Interorganizational Imitation: The Effects of Outcome Salience and Uncertainty. Administrative Science Quarterly 42: 472–500. Helland, Eric. 1998. The Enforcement of Pollution Control Laws: Inspections, Violations, and. Self-Reporting. Review of Economics and Statistics 80: 141–53. Henderson, Andrew D., Danny Miller, and Donald C. Hambrick. 2006. How Quickly Do CEOs Become Obsolete? Industry Dynamism, CEO Tenure, and Company Performance. Strategic Management Journal 27(5): 447–60. Herling, John. 1962. The Great Price Conspiracy: The Story of the Antitrust Violations in the Electrical Industry. Washington, DC: R. B. Luce. Hermalin, Benjamin E. 2012. Leadership and Corporate Culture. In Robert Gibbons and John Roberts, Handbook of Organization, 432–78. Princeton, NJ: Princeton University Press.

Antitrust Corporate Governance and Compliance   615

Hermalin, Benjamin E., and Michael S. Weisbach. 2012. Information Disclosure and Corporate Governance. Journal of Finance 67: 195–233. Hüschelrath, Kai. 2010. Economist’s Note:  How are Cartels Detected? The Increasing Use of Proactive Methods to Establish Antitrust Infringements. Journal of European Competition Law and Practice 1: 1–7. Hüschelrath, Kai, Nina Leheyda, and Patrick Beschorner. 2011. The Deterrent Effect of Antitrust Sanctions: Evidence from Switzerland. Antitrust Bulletin 56(2): 427–60. Hüschelrath, Kai, and Tobias Veith. 2011. The Impact of Cartelization on Pricing Dynamics. ZEW Centre for European Economic Research Discussion Paper No. 11-067. Jensen, Michael C. 1993. The Modern Industrial Revolution, Exit, and the Failure of Internal Control Systems. Journal of Finance 48: 831–79. Jensen, Michael C., and William H. Meckling. 1976. Theory of the Firm: Managerial Behaviour, Agency Costs and Ownership Structure. Journal of Financial Economics 3(4): 305–60. Joh, Sung Wook. 1999), Strategic Managerial Incentive Compensation in Japan:  Relative Performance Evaluation and Product Market Collusion. Review of Economics and Statistics 81(2): 303–13. Johnson, Shane A., Harley E.  Ryan, and Yisong S.  Tian. 2009. Managerial Incentives and Corporate Fraud: The Sources of Incentives Matter. Review of Finance 13(1): 115–45. Judge, George, and Laura Schechter. 2009. Detecting Problems in Survey Data Using Benford’s Law. Journal of Human Resources 44: 1–24. Kaplow, Louis. 2011. An Economic Approach to Price Fixing. Antitrust Law Journal 77(2): 343–449. Karpoff, Jonathan M., D. Scott Lee, and Gerald S. Martin. 2008a. The Consequences to Managers for Financial Misrepresentation. Journal of Financial Economics 88: 193–215. Karpoff, Jonathan M., D. Scott Lee, and Gerald S. Martin. 2008b. The Cost to Firms of Cooking the Books. Journal of Financial and Quantitative Analysis 43: 581–611. Karpoff, J. M., and John R. Lott Jr. 1993. The Reputational Penalty Firms Bear from Committing Criminal Fraud. Journal of Law and Economics 26: 757–802. Karpoff, J. M., J. R. Lott, and E. W. Wehrly. 2005. The Reputational Penalties for Environmental Violations: Empirical Evidence. Journal of Law and Economics 48: 653–75. Kim, Donghun, and Ronald Cotterill. 2008. Cost Pass-Through in Differentiated Product Markets: The Case of U.S. Processed Cheese. Journal of Industrial Economics 56: 32–48. Kim, Donghun, and Ronald Cotterill. 2008. Cost Pass-Through in Differentiated Product Markets: The Case of U.S. Processed Cheese. Journal of Industrial Economics 56: 32–48. Klein, B., and K. Leffler. 1981. The Role of Market Forces in Assuring Contractual Performance. Journal of Political Economy 89: 615–41. Kovacic, William E. 2001. Private Monitoring and Antitrust Enforcement: Paying Informants to Reveal Cartels. George Washington Law Review 69: 766–97. Laitenberger, Ulrich, and Kai Hüschelrath. 2011. The Adoption of Screening Tools by Competition Authorities. Competition Policy International Antitrust Chronicle 2 (Sep.): 1–9. Levenstein, Margaret C., and Valerie Y. Suslow. 2006. Cartel Bargaining and Monitoring: The Role of Information Sharing. In The Pros and Cons of Information Sharing. Stockholm: Konkurrensverket, 59–64. Marshall, Robert, and Leslie Marx. 2009. The Vulnerability of Auctions to Bidder Collusion. Quarterly Journal of Economics 124: 883–910. Mason, Phyllis A., and Donald C.  Hambrick. 1984. Upper Echelons:  The Organization as a Reflection of Its Top Managers. Academy of Management Review 9(2): 193–206.

616   Rosa M. Abrantes-Metz and D. Daniel Sokol Mena-Labarthe, Carlos. 2012. Mexican Experience in Screens for Bid-Rigging. Competition Policy International Antitrust Chronicle 1 (Mar.): 1–8. Langus, Gregor, Massimo Motta, and Luca Aguzzoni. 2013. The Effect of EU Antitrust Investigations and Fines on a Firm’s Valuation. Journal of Industrial Economics 61(2): 290–338. Maclean, Tammy L., and Michael Behnam. 2010. The Dangers of Decoupling: The Relationship between Compliance Programs, Legitimacy Perceptions, and Institutionalized Misconduct. Academy of Management Journal 53(6): 1499–520. McKendall, Marie A., and John A. Wagner III. 1997. Motive, Opportunity, Choice, and Corporate Illegality. Organizational Science 8: 624–47. Miller, Danny. 2001. Stale in the Saddle: CEO Tenure and the Match between Organization and Environment. Management Science 37(1): 34–52. Montreal Gazette. 2012. Montreal Construction Bids Are Paved with Questions. September 15. Available at http://www.montrealgazette.com/news/Montreal+construction+bids+paved+​ with+questions/7248408/story.html. Morgan, Angela G., and Annette B. Poulsen. 2001. Linking Pay to Performance: Compensation Proposals in the S&P 500. Journal of Financial Economics 62(3): 489–523. Motchenkova, Evgenia, and Rob van der Laan. 2005. Strictness of Leniency Programs and Cartels of Asymmetric Firms. Center Discussion Paper 2005-74. Nielsen, Vibeke L., and Christine Parker. 2005. The ACCC Enforcement and Compliance Survey:  Report of Preliminary Findings. a report for the Centre for Competition and Consumer Policy, Australian National University, Canberra. Nigrini, Mark. 1996. A Taxpayer Compliance Application of Benford’s Law. Journal of the American Taxation Association 18: 72–91. Oberholzer-Gee, Felix and Julie M. Wulf. 2012. Earnings Management from the Bottom Up: An Analysis of Managerial Incentives below the CEO. Harvard Business School Strategy Unit Working Paper No. 12-056. Office of Fair Trading (OFT). 2011. Competition Law Compliance Survey. Office of Fair Trading (OFT). 2007. The Deterrent Effect of Competition Enforcement by the OFT. O’Kane, Michael. 2011. Does Prison Work for Cartelists? The View from behind Bars: An Interview of Bryan Allison. Antitrust Bulletin 56(2): 483. Oliver, Christine. 1991. Strategic Responses to Institutional Processes. Academy of Management Review 16: 145–79. Ormosi, Peter. 2011. How Big Is a Tip of the Iceberg? A Parsimonious Way to Estimate Cartel Detection Rate. Working paper. Parker, Christine. 2006. The “compliance” Trap: The Moral Message in Regulatory Enforcement Law and Society Review 40: 591–622. Parker, Christine. 2011. Criminal Cartel Sanctions and Compliance in Criminalising Cartels. In Caron Beaton-Wells and Ariel Ezrachi, Criminalising Cartels:  Critical Studies of an International Regulatory Movement, 239–62. Oxford: Hart Publishing. Parker, Christine. 2012a. Economic Rationalities of Governance and Ambiguity in the Criminalisation of Cartels. British Journal of Criminology 52(5): 974–96. Parker, Christine. 2012b. The War on Cartels and the Social Meaning of Deterrence. Regulation and Governance 7(2): 174–94. Parker, Christine, Paul Ainsworth, and Natalie Stepanenko. 2004. ACCC Enforcement and Compliance Project:  How ACCC Enforcement in Cartel Cases Works in the Face of

Antitrust Corporate Governance and Compliance   617

Deterrence Failure. Canberra: Australian National University, Centre for Competition and Consumer Policy. Parker, C., F. Haines, J. Kotey, J. Nankivell, and D. Round. 2011. Report on Interviews with Civil Respondents in Cartel Cases. Melbourne Law School Cartel Project. Parker, Christine, and Chris Platania-Phung. 2012. The Deterrent Impact of Cartel Criminalisation: Report of a Survey of Australian Business People. Working paper. Parker, Christine, and Vibeke Nielsen. 2006. Do Businesses Take Compliance Seriously? Melbourne University Law Review 30: 441–94. Peng, L., and A. Röell. 2008. Executive Pay and Shareholder Litigation. Review of Finance 12(1): 141–84. Porter, Robert, and Douglas Zona. 1993. Detection on Bid Rigging in Procurement Auctions. Political Economy 101(3): 518–38. Porter, Robert, and Douglas Zona. 1999. Ohio School Milk Markets: An Analysis of Bidding. Rand Journal of Economics 30(2): 263–88. Ragazzo, Carlos. 2012. Screens in the Gas Retail Market: The Brazilian Experience. Competition Policy International Antitrust Chronicle 1 (Mar.): 1–8. Rasmusen, Eric Bennett. 1996. Stigma and Self-Fulfilling Expectations of Criminality. Journal of Law and Economics 39: 519–44. Rodger, Barry J. 2000. Compliance with Competition Law: A View from Industry. Commercial Liability Law Review 2000: 249–81. Rodger, Barry J. 2005. Competition Law Compliance Programmes: A Study of Motivations and Practice. World Competition 28(3): 349–76. Rodger, Barry J. 2009. A Study of Compliance Post-OFT Infringement Action. European Competition Journal 5(1): 65–96. Scharfstein, David. 1988. Product-Market Competition and Managerial Slack. Rand Journal of Economics 19(1): 147–55. Schein, Edgar H. 2010. Organizational Culture and Leadership. 4th ed. San Francisco, CA: Jossey-Bass. Schmidt, Klaus M. 1997. Managerial Incentives and Product Market Competition. Review of Economic Studies 64(2): 191–213. Schnatterly, Karen. 2003. Increasing Firm Value through Detection and Prevention of White-Collar Crime. Strategic Management Journal 24: 587–614. Siltaoja, Marjo, and Meri Vehkaperä. 2010. Constructing Illegitimacy? Cartels and Cartel Agreements in Finnish Business Media from Critical Discursive Perspective. Journal of Business Ethics 92(4): 493–511. Shin, Hyun-Han, and René M. Stulz. 1998. Are Internal Capital Markets Efficient? Quarterly Journal of Economics 113(2): 531–52. Simpson, Sally S., and Christopher S. Koper. 1997. The Changing of the Guard: Top Management Team Characteristics, Organizational Strain, and Antitrust Offending, 1960–1986. Journal of Quantitative Criminology 13: 373–404. Snider, Connan, and Thomas Youle. 2009. Diagnosing the LIBOR:  Strategic Manipulation Member Portfolio Positions. Working paper. Snider, Connan, and Thomas Youle. 2010. Does the LIBOR Reflect Banks’ Borrowing Costs? Working paper. Sokol, D. Daniel. 2011. Detection and Compliance in Cartel Policy. CPI Antitrust Chronicle 9(2): 1–5.

618   Rosa M. Abrantes-Metz and D. Daniel Sokol Sokol, D.  Daniel. 2012. Cartels, Corporate Compliance and What Practitioners Really Think about Enforcement. Antitrust Law Journal 78(1): 201–40. Sonnenfeld, Jeffrey. 1981. Executive Apologies for Price Fixing: Role Based Perceptions of Causality. Academy of Management Journal 24: 192–98. Sonnenfeld, Jeffrey, and Paul R. Lawrence. 1978. Why Do Companies Succumb to Price Fixing? Harvard Business Review 56: 145–57. Spagnolo, Giancarlo. 2000. Stock-Related Compensation and Product-Market Competition. Rand Journal of Economics 31(1): 22–42. Spagnolo, Giancarlo. 2005. Managerial Incentives and Collusive Behavior. European Economic Review 49(6): 1501–23. Spagnolo, Giancarlo. 2008. Leniency and Whistleblowers in Antitrust. In Paolo Buccirossi, Handbook of Antitrust Economics, 259–304. Cambridge, MA: MIT Press. Stephan, Andreas. 2010a. Price Fixing during a Recession:  Implications of an Economic Downturn for Cartels and Enforcement. Working paper. Stephan, Andreas. 2010b. See No Evil: Cartels and the Limits of Antitrust Compliance Programs. Company Lawyer 31: 231–39. Stigler, George J. 1964. A Theory of Oligopoly. Journal of Political Economy 72(1): 44–61. Summers, Scott L., and John T. Sweeney. 1998. Fraudulently Misstated Financial Statements and Insider Trading: An Empirical Analysis. Accounting Review 73: 131–46. Thompson, John S., and David L. Kaserman. 2001. After the Fall: Stock Price Movements and the Deterrent Effect of Antitrust Enforcement. Review of Industrial Organization 19(3): 329–34. Tushman, Michael, and Charles A.  O’Reilly. 1997. Winning through Innovation:  A  Practical Guide to Leading Organizational Change and Renewal. Boston, MA:  Harvard Business School Press. Uzun, Hatice, Samuel H. Szewczyk, and Raj Varma. 2004. Board Composition and Corporate Fraud. Financial Analysts Journal 60: 33–43. van den Broek, Stijn, Ron G. M. Kemp, Willem F. C. Verschoor, and Anne-Claire de Vries. 2012. Reputational Penalties to Firms in Antitrust Investigations. Journal of Competition Law and Economics 8(2): 231–58. Varian, Hal. 1972. Benford’s Law. American Statistician 26(3): 65–6. Weisbach, Michael S. 1988. Outside Directors and CEO Turnover. Journal of Financial Economics 20: 431–60.

Case Cited In re Caremark Int’l Inc. Derivative Litig., 698 A.2d 959 (Del. Ch. 1996)

Index

A&P. See Great Atlantic and Pacific Tea Company A&W, 389 Abrantes-Metz, Rosa M. on corporate governance and compliance,  586, 600, 604, 606–607, 609–610 on screening,  528, 535–540, 542–544 Abreu, Dilip,  478–479, 490 Abuse-of-dominance, 234–251 behavioral definition,  238 in China,  236, 241–242, 244–245 competition and,  237–239 corruption and,  238 in Costa Rica,  240 divestiture, 250 Efficient Component Pricing Rule (ECPR) and, 250 in El Salvador,  240 enforcement, 237–239 essential facilities doctrine and,  250 in EU,  236 examples of,  239–242 in Hungary,  237 in India,  236–237, 245–248 in Jamaica,  237, 240–241 in Latvia,  238, 241 legal standard,  235–236 in Lithuania,  237 in Mexico,  237 overview,  234–235, 250–251 in Peru,  241 in Poland,  237 remedies in,  249–250 in Romania,  237 in Russia,  236, 241–244 sanctions, 249–250

in Senegal,  241–242 in Slovak Republic,  237 in South Africa,  237–238, 241, 248–249 structural definition,  238 in Ukraine,  238 in United States,  236 in Uzbekistan,  237 Addanki, S.,  537, 542 Added value,  260–261 Adelman, M.A.,  43 Adobe,  556, 565 Advanced Micro Devices (AMD),  92, 95, 114 AEC (As-efficient-competitor) test,  168–169 Agency costs, corporate governance and compliance and,  587 Aghion, Philippe,  148, 171, 176, 204, 317–318, 369 Aguzzoni, Luca,  596 Alcoa, 58 Alexander, Barbara,  449 Allied, 562 Allocative distortion,  79–80 Amemiya, Kei,  205 American Bar Association, Antitrust Law Section, 598 American Medical Association,  559 American National Standards Institute (ANSI),  557, 561, 567, 569 American Recovery and Reinvestment Act of  2009, 453 American Society of Mechanical Engineers (ASME), 564–565 American Sugar Refining Company,  368 American Tobacco Company,  44, 58 Amsterdam second-price auction,  518n8 Angelucci, Charles,  588 ANSI. See American National Standards Institute

620   Index Anthem, 269 Anticompetitive nature of exclusionary conduct, 7–19 diminished ability to compete and,  26–29 enhanced market power and,  30 harm to consumers and,  30 negative contracting externalities and, 31–32 Antitrust Division. See Justice Department Antitrust Modernization Commission intellectual property rights and,  148 predatory buying and,  93 predatory pricing and,  41, 50, 52 quality commitment discounts and,  110, 116 Aoyagi, M.,  508–509 Apesteguia, Jose,  436–437 Apple,  129, 260, 565–566, 572 Areeda, Phillip on exclusive dealing,  320 on predatory pricing,  46–47, 49, 54 on quantity commitment discounts,  94 on squeeze claims,  125 on tying,  338, 347 Areeda-Turner rule,  47–49, 52–54 Arlen, Jennifer,  591 “Arrive by reasoning,”  482–486 Arrow, Kenneth,  148, 171, 560 Arrow's Information Paradox,  135 As-efficient-competitor (AEC) test,  168–169 Asker, John,  368, 394, 502 ASME (American Society of Mechanical Engineers), 564–565 Assessment errors, dominant firms and, 169–170 AstraZeneca, 167 Asymmetry, collusion and,  418n6 Athey, Susan,  422–423, 508–509, 513–514, 530, 533 Attribution test flaws with,  111–116 liability thresholds and,  110 overview, 105–109 uses of,  111–116 Aubert, Cecile,  429–430 Auctions Amsterdam second-price auction,  518n8

bid rigging in,  501–503. See also Bid rigging collusion in,  501–503 common value auction,  499–500 defined, 498 Dutch auction,  499–500 English auction,  499–500 first-price sealed-bid (FPSB) auction,  499–500, 503–505, 510, 516, 518 Japanese auction,  499 one-shot auctions, bid rigging in,  504–507 overview, 498–499 private value auction,  499–500 repeated auctions, bid rigging in,  507–510 second-price sealed-bid (SPSB) auction,  499–500, 504–505 standard models,  499–500 Audretsch, David B.,  444 Aumann, Robert J.,  481–482, 488–490, 492 Australia cartels in,  445 corporate governance and compliance in, 598 vertical restraints in,  384 Avraham, Raphael,  120 Azoulay, Pierre,  395–397 Bagwell, Kyle,  422–423, 478, 508–509, 530, 533 Baidu, 245 Bain, Joe S.,  478 Bajari, Patrick,  514, 546, 548, 600 Baker, Jonathan B.,  431, 494 Baker, Wayne E.,  598 Baldwin, L.,  511–512 Bank of Japan,  536 Bar-Isaac, Heski,  368 Barkoff, Rupert M.,  403, 408 Baskin-Robbins,  329–330, 343, 400 Baumol, William J.,  47–48, 55 Bayesian hypothesis testing, screening and, 525–528 Bayesian Nash equilibrium,  488 Beckenstein, Alan R.,  598 Beijing Netcom,  219–220 Beijing Qihoo Technology,  216, 228–230 Belgium, franchising in,  395, 407 Below-cost pricing in Japan,  202–203

Index   621

Beltone, 309 Benford's Law,  602, 606–607, 610 Benoit, Jean-Pierre,  479 Ben-Porath, Elchanan,  491 Bergen, Mark,  304, 394 Bergin, James,  492–493 Bernheim, B. Douglas,  3, 311–312, 365, 419, 434, 483 Bertrand markets bid rigging in,  498, 508 collusion in,  416, 435 screening in,  530, 541 Besanko, David,  254, 266, 431–432 Bessen, James,  138 Betamax, 558 Biddle, Brad,  555 Bid rigging,  498–519 in auctions,  501–503 in Bertrand markets,  498, 508 in China,  501 in Cournot markets,  498 deterrence of,  517–518 empirical evidence,  510–515 in EU,  501 experimental evidence,  515–517 incentive for collusion,  503–504 Nash equilibrium and,  500, 504 in one-shot auctions,  504–507 overview,  498–503, 518–519 in repeated auctions,  507–510 screening for,  602–604 theory of,  503–510 in United States,  501–502 Bigoni, Marta,  436–437 Blair, Roger D.,  343, 400, 405 Blanckenberg, K.,  547, 549 Block, Michael K.,  431, 453 Blue Cross and Blue Shield Association,  314 Blundering into tacit collusion,  492–493 Blu-ray, 558 Bohannan, Christina,  132, 148, 334, 343, 345–346 Böhm, Franz,  161–163, 182 Bolotova, Y.,  536, 543 Bolton, Patrick,  52, 204, 317–318, 369 Bork, Robert H.,  9, 311–312, 318, 337

Bosch, Jean-Claude,  595–596 Bounties as incentives to induce detection, 599–600 Bowen, Robert M.,  599 Bowman, Ward S.,  333 Bradburd, Ralph,  449 Brandeis, Louis,  333–334 Brandenburger, Adam,  481–482, 492 Brannon, I.,  540 Brazil cartels in,  453–454 screening in,  524, 606 Breakup of cartels, causes of,  455–458 Brenkers, Randy,  398 Bresnahan, T.,  540–541, 545 Breyer, Stephen,  58, 382 British Banking Association,  536 British Standards Institution,  557 Broadway-Hale,  123–125, 127 Brock, William A.,  479 Brodley, Joseph F.,  52 Bryant, Peter G.,  444, 448 Bundling as exclusionary conduct,  18, 22–24 explicit bundling,  104–105 implicit bundling,  104–105 intellectual property rights,  343–344 quality commitment discounts and,  102, 104–105 Burger King,  354, 389 “Burning the dollar” game,  491 Burns, Malcolm R.,  44 Bush, George H.W.,  194 Business-format franchising,  390 Business strategy,  253–272 added value,  260–261 competitive advantage,  261–266. See also Competitive advantage complementarities in,  258 consumer surplus and,  255, 258–260 economic framework,  253–254 enhancing value creation,  256–258 “Five Forces,”  255, 258–260, 263 overview,  253–255, 271–272 producer surplus and,  255, 258–260 rivalry and,  258–260

622   Index Business strategy (cont.) scope of firm,  266–271 sustainability of competitive advantage, 264–266 synergies, 267–268 trade-offs in,  257–258 value creation and capture framework, 255–261 “But for” prices,  180–182 Buyback options,  352 California Air Resources Board,  575 Call, Andrew C.,  599 Canada Competition Bureau Canada,  470, 538 franchising in,  388–389 screening in,  606 vertical restraints in,  353, 372, 384 Can System,  203 Caremark, 101 Carlsson, Hans,  488–489 Carlton, Dennis W.,  370, 469–471 Carrefour, 329 Cartels in Australia,  445 in Brazil,  453–454 breakup, causes of,  455–458 competition, effect on formation of,  447 competition and,  162 concerted squeezing and,  128–129 corporate governance and compliance, effect of firm indicia of cartel activity on, 589–594 detection, 430–433 effects on prices,  453–455 empirical evidence of collusion,  442–459 enforcement and,  424–433 in EU,  443–446 experimental evidence of collusion, 433–435 in Finland,  443 formation of,  446–449 in Germany,  444–445, 455 hub-and-spoke cartels,  128–129 in Japan,  191, 445 leniency programs and,  424–430, 435–437

in Mexico,  454n33 Nash equilibrium and,  426–428 in Netherlands,  596 overview,  415–416, 442–443, 458–459 prevalence of,  443–446 price setting by,  449–452 recession, effect on formation of,  447–449 screening, 430–433 in Sweden,  443–444 in United Kingdom,  443–444 in United States,  443–446 CDS (Credit default swaps),  536–537 Census Bureau,  390 Certification services,  280–281 Chang, Myong-Hun,  457 Charness, Gary,  489 Che, Y.-K.,  517–518 Chen, Chia-Wen,  394, 446 Chen, J.,  531–532 Chevron, 575 Chicago School decision theory and,  165 dynamic efficiencies and,  171 intellectual property rights and,  140 on tying,  337–338 on vertical restraints,  352 Chicken Delight,  409 China abuse-of-dominance in,  236, 241–242, 244–245 Anti-Monopoly Commission,  211–212, 215 Anti-Monopoly Law of 2007 (AML),  188, 210–221, 225–231 Anti-Unfair Competition Law,  213, 218 bid rigging in,  501 buying at unfairly low price in,  217–218 China Telcom, 219–220 China Unicom, 219–220 definition of market dominance in,  214 discriminatory practices in,  218–219 dominant market position in,  214–215 Draft Guidelines for Anti-Monopoly Enforcement in the Field of Intellectual Property Rights,  225–226 EU enforcement structure compared,  217– 218, 225, 227, 244

Index   623

exclusive dealing in,  222–224 Guidelines Concerning the Definition of Relevant Markets,  215–216, 228 Intellectual Property Division,  212 intellectual property rights in,  225–226 Japan compared,  230–232 joint dominance in,  215 Judicial Interpretation on the Application of Laws to Anti-Monopoly Private Actions, 213 legal provisions of AML,  211–212 liability in,  213 Ministry of Commerce,  212, 227 monopoly in,  210–232 National Development and Reform Commission (NDRC),  211–213, 215, 217–219, 221–227, 230–231, 244–245 non-price-related abuse of dominance in, 221–230 overview,  188, 210–211, 227–230 predatory pricing in,  218 presumption of market dominance in,  215 price discrimination in,  220–221 Price Law,  213–214, 218, 220–221, 231 price-related abuse of dominance in, 217–221 private actions in,  212–213 Qihoo v. Tencent, 228–230 refusals to deal in,  221–222 Regulation on Anti-Price Monopoly,  212, 217–218, 221–224, 227, 230 Regulation on Prohibiting the Abuse of a Market Dominant Position,  212, 214, 217–218, 221–222, 224, 227, 230 Regulation on the Prevention of Below Cost Dumping Conduct,  218 relevant laws,  213–214 relevant market in,  215–216 rule-of-reason analysis in,  216–217 selling at unfairly high price in,  217–218 State Administration for Industry and Commerce (SAIC),  211–212, 214–215, 217–218, 221–222, 224–225, 227, 230–231, 244–245 state-owned enterprises (SOEs),  226–227 Supreme People's Court,  212–213, 215 tying in,  224–225

US enforcement structure compared,  225 China Internet Network Information Center, 229 China Netcom,  219 China Telecom,  215, 219, 227 China Unicom,  215, 219, 227 Choi, Jay Pil,  415, 420, 430–431, 587 Christie, William G.,  523–525, 540, 545, 606 Church & Dwight,  326 Cisco, 565 Clark, John M.,  41 Clayton Act of  1914 exclusive dealing and,  305–306 Japanese antitrust law compared,  189 predatory pricing under,  45 quality commitment discounts and,  90n4 tying under,  331, 347 vertical restraints under,  372, 379 Coase, Ronald H.,  267 Coca-Cola,  35, 95, 283, 388–389 Cohen, Mark A.,  453 Collusion asymmetry and,  418n6 in auctions,  501–503 in Bertrand markets,  416, 435 bid rigging,  498–519. See also Bid rigging in cartels. See Cartels concentration and,  418 corporate governance and compliance and, 586–611. See also Corporate governance and compliance in Cournot markets,  421, 433–434 demand conditions and,  418–419 empirical evidence,  442–459 experimental evidence,  433–435 factors facilitating,  417–424 folk theorem and,  417n5 game theory and,  416–417 imperfect observability and monitoring and, 420–422 incomplete information and communication and,  422–423 multimarket contact and,  419–420 in NASDAQ,  523–525, 540 number of firms and,  418 screening for,  523–551. See also Screening self-enforcing nature of,  416

624   Index Collusion (cont.) standard setting and,  554–581. See also Standard setting strong explicit collusion,  468–477 symmetry and,  418 tacit collusion,  464–494. See also Tacit collusion theory of,  416–417 vertical mergers and,  423–424 vertical restraints and,  423–424 weak explicit collusion,  468–477 Comanor, William B.,  288 Commerce Clearing House Trade Regulation Reporter, 444 Commerce Department, franchising and, 389–390 Commodities Futures Trading Commission, screening and,  606–607 Common value auction,  499–500 Communication in tacit collusion,  466–468 Compatibility standards,  555 Competition cartels, effect on formation of,  447 cartels and,  162 in developing countries,  237–239 diminished ability to compete, exclusionary conduct and,  26–29 economic power versus,  162–163 monopsony compared,  74 standard setting and,  554–581. See also Standard setting Competitive advantage,  261–266 accumulated market experience and,  266 benefit-based advantage,  262–263 cost-based advantage,  262 defined, 261 niche-based advantage,  263 resource protection and,  264–265 sustainability of,  264–266 Competitive discounts. See Quantity commitment discounts Complementarities in business strategy,  258 Complements, tying of,  339–341 Compliance. See Corporate governance and compliance Compte, Olivier,  418 Concentration, collusion and,  418

Concerted squeezing cartels and,  128–129 overview, 128–129 railroad industry and,  128 unilateral squeezing compared,  122–125 Connor, Robert A.,  443, 454, 536, 543, 596 Consignment selling arrangements,  352 Consumer Goods Pricing Act of  1975, 373 Consumers, exclusionary conduct and harm to, 30 Consumer surplus,  255, 258–260 Contract adaptation, franchising and, 405–408 Contracts that reference rivals (CRRs),  89–90 Convergent-elimination view,  483–486 Conwood, 325–326 Cooper, James,  355, 371, 527 Corporate governance and compliance, 586–611 agency costs and,  587 in Australia,  598 bid rigging, screening for,  602–604 bounties as incentives to induce detection, 599–600 cartel activity, firm indicia of,  589–594 compliance, use of screens in,  608–609 compliance culture,  596–597 culture, 594–597 empirical evidence,  597–599 employees and,  593–594 enforcement and,  591–592 in EU,  596 in Germany,  588 incentive pay,  590–591 internal monitoring, screens in,  607–608 in Japan,  588 LIBOR, screening for,  609–610 market share, screening for,  605 mathematical laws, screens based on,  606 middle management and,  593–594 norm creation,  594–595 organizational environment,  587–589 overview,  586, 610–611 price fixing, screening for,  604–605 screens, use of,  600–610 senior management and,  592–593 in South Korea,  600

Index   625

stock-based cartel event studies,  595–596 success of screens,  606–607 in Switzerland,  598 in United Kingdom,  597–598, 600 in United States,  597–598 Corruption, abuse-of-dominance and,  238 Costa Rica, abuse-of-dominance in,  240 Cotropia, Christopher A.,  138 Cotter, Thomas F.,  132 Counterstrategies, RRC theory and,  64 Cournot, Augustin,  355 Cournot markets bid rigging in,  498 collusion in,  421, 433–434 screening in,  530 Credit default swaps (CDS),  536–537 Cross-licenses of patents,  571–572 CRRs (Contracts that reference rivals),  89–90 Dach, Christian,  279 Dairy Queen,  343 Dal Bo, Pedro,  433 Damages in Japan,  196–199 liquidated damages,  317n40 tying, 347–348 Dandong Yichuang Yaoye Co., Ltd.,  223 David, Paul A.,  554 Davidow, Joel,  444 Dealer property rights, exclusive dealing and, 324–326 Decision theory,  164–171 assessment errors and,  169–170 Chicago School and,  165 in EU,  165–166 information deficiencies and,  168–169 overview,  164–165, 170–171 self-enforcement and,  166–170 special responsibility of dominant firms and, 166–167 TFEU and,  166–171 Defense Department, cartels and,  453 Defense Personnel Support Center,  535, 604 Degussa, 449 Dekel, Eddie,  491 Dell,  92, 575 Demand conditions, collusion and,  418–419

Demand curves, quality commitment discounts and,  95–97 Demand effects, RRC theory and,  64 DeNA Co., Ltd,  209 Deneckere, Raymond,  364 Dentsply, 321–324 Denture industry, exclusive dealing in, 321–324 Detection, cartels and,  430–433 Deterrence of bid rigging,  517–518 Detre, Joshua D.,  596 Deutsches Institut für Normung,  557 Deutsche Telekom,  167, 240 Developing countries, monopoly in,  234–251. See also Abuse-of-dominance Dick, Andrew R.,  445 Dicke, Thomas S.,  388 Digicel, 240 Diminished ability to compete, exclusionary conduct and,  26–29 Direct interaction, tacit collusion and,  467n7 Discount Pricing Consumer Protection Act (proposed), 373n27 Discounts. See Quantity commitment discounts; specific discount Discriminatory practices in China,  218–219 in Japan,  207–209 Disparagement of alternative proposals, standard setting and,  564–565 Divestiture in abuse-of-dominance cases,  250 DLF, 247 Dnes, Antony W.,  389 Doane, Michael J.,  523, 587 DOJ. See Justice Department Dominant firms,  153–184 as-efficient-competitor (AEC) test,  168–169 assessment errors and,  169–170 attempted monopolization,  156 “but for” prices,  180–182 competition on merits,  158, 174–176 decision theory and,  164–171. See also Decision theory dynamic efficiencies and,  171–178 in EU,  153–184 excessive pricing and,  178–182. See also Excessive pricing

626   Index Dominant firms (cont.) exclusionary conduct and,  154–155 exploitative abuses,  155–156 information deficiencies and,  168–169 innovation and,  171–178 interaction between,  176–177 interpretation differences between Sherman Act and TFEU,  156–160 monopolization versus abuse,  154–155 ordoliberalism and,  160–164. See also Ordoliberalism overview,  153–154, 182–184 protection of competitive process,  158–159, 172–174 self-enforcement by,  166–170 special responsibility of,  157, 166–167 in United States,  153–184 wording differences between Sherman Act and TFEU,  154–156 Domino's Pizza,  343, 402–403 Donggang Hongda Pharmaceutical Co., Ltd., 223 Dorfman, Robert,  360 Dorfman-Steiner theorem,  360–361 Double marginalization, tying and,  339–341 Dufwenberg, Martin,  436–437 du Pont,  58 Dutch auction,  499–500 Dutta, Shantanu,  304, 394 DVDs, 558 Dyck, Alexander,  599 Dynamic efficiencies Chicago School and,  171 dominant firms and,  171–178 Easterbrook, Frank H.,  53, 67, 156–157, 159, 165, 527 Eaton, 92 Eckard, E. Woodrow, Jr.,  444, 448, 595–596 Economies of joint provision,  341–344 intellectual property bundling,  343–344 overview, 341 package licensing,  343–344 production cost efficiencies,  342 quality control efficiencies,  342–343 Economies of scale, quality commitment discounts and,  103n33

ECPR (Efficient Component Pricing Rule), abuse-of-dominance and,  250 Edlin, Aaron S.,  47, 51 Edwards, Corwin D.,  419 Efficiencies production cost efficiencies,  342 quality control efficiencies,  342–343 RRC theory and,  64 vertical restraints and,  359–364 Efficient Component Pricing Rule (ECPR), abuse-of-dominance and,  250 Elhauge, Einer,  52 Eli Lilly & Co.,  92 El Salvador, abuse-of-dominance in,  240 Elzinga, Kenneth G.,  40, 43 Employees, corporate governance and compliance and,  593–594 Empresa de Servicios Publicos de Heredia, 240 Enforcement regimes cartels and,  424–433 in developing countries,  237–239 in Japan,  195–199 predatory pricing and,  45–49 England. See United Kingdom English auction,  499–500 Enhanced market power, exclusionary conduct and,  30 Entrant-tax-by-contract theory of exclusive dealing, 316–318 Eruthku, C.,  538, 547 Essential facilities,  121, 250 ETSI (European Telecommunications Standards Institute),  557, 568 Eucken, Walter,  161–163, 178, 182 Euribor, 607 European Central Bank,  536 European Commission abuse and,  155 Block Exemption Regulation on Vertical Restraints, 405–406 cartels and,  445, 448–449, 454, 456, 591 competition on merits and,  175 decision theory and,  167–169 Directorate-General for Competition,  157, 591 excessive pricing and,  179–181

Index   627

exclusionary conduct and,  24n45 Guidance Paper,  157, 159 Guidelines on Vertical Restraints,  297–301, 358, 367, 375–378, 380–383, 405–406 nonhorizontal merger guidelines,  69–70 RRC theory and,  69–70 screening and,  606 tacit collusion and,  494 Technology Transfer and Research and Development Block Exemption Regulations, 147n37 European Community Treaty, intellectual property rights under,  147 European Court of First Instance intellectual property rights and,  141 vertical restraints and,  378 European Court of Justice and General Court decision theory and,  168–169 excessive pricing and,  179 predatory pricing and,  46 protection of competitive process and,  158–159, 173–174 European Telecommunications Standards Institute (ETSI),  557, 568 European Union. See also Treaty on the Functioning of the European Union abuse-of-dominance standard in,  236 bid rigging in,  501 cartels in,  443–446 Chinese enforcement structure compared,  217–218, 225, 227, 244 corporate governance and compliance in, 596 decision theory in,  165–166 dominant firms in,  153–184. See also Dominant firms essential facilities doctrine in,  250 exclusive dealing in,  306–307 franchising in,  387, 405–408 intellectual property rights in,  142, 146–147 Japanese enforcement structure compared, 201 mergers in,  269n9 predatory pricing in,  46, 51 resale price maintenance in,  297–301 vertical restraints in,  352–353, 366, 371–384 Evans, David S.,  342

Evans, Phil,  449 Ex ante bargaining, patents and,  579–580 Excel (software),  206 Excessive pricing,  178–182 “but for” prices,  180–182 ordoliberalism and,  178–179 overview, 178 practical objections against enforcement, 179–180 Exclusionary conduct,  3–36 anticompetitive nature of,  7–19, 26–32 bundling,  18, 22–24 defined, 3 determination of,  20–26 diminished ability to compete and,  26–29 duration of,  29n52 enhanced market power and,  30 evaluation of,  19–32 examples of procompetitive nature,  34–35 exclusionary conditions,  15–19 harm to consumers and,  30 “hold-up problem,” 7  n10 in Japan,  200–209 loyalty discounts,  5n5, 18n37, 32n57 mechanism of,  8–14 modes of conduct implicating mechanism, 15 negative contracting externalities and,  7–13, 31–32 100% exclusive deals,  15 overview,  3–6, 35–36 patents and,  4n3 practices involving exclusionary conditions, 15–17 practices not involving exclusionary conditions, 17–19 predatory pricing,  17–18, 21 procompetitive nature of,  32–35 quality commitment discounts and,  102 rule-of-reason analysis,  4–6 “signal jamming problem,”  27n48 single-product pricing,  18, 24 in soft drink industry,  35 standard setting as venue for,  559 tying, 5n6, 19, 25 volume discounts,  18n37, 24–25 Exclusive dealing,  304–326

628   Index assessment of harm,  320–324 in China,  222–224 dealer property rights and,  324–326 in denture industry,  321–324 entrant-tax-by-contract theory of,  316–318 in EU,  306–307 ex post exclusive dealing,  318–320 externalities, role of,  311–313 franchising and,  393–395 in health insurance,  314–315 in hearing aid industry,  307–309 in Japan,  203–205 linear versus nonlinear contracts,  311n29 naked exclusion theory of,  313–316 in optical industry,  310–311 overview,  304–307, 326 partial exclusive dealing,  304 as property right creation,  307–311 quantity commitment discounts and,  100 theories of,  311–320 vertical restraints and,  379–381 Exclusive distribution. See Franchising Exclusive territories franchising and,  393, 395–398 vertical restraints and,  374–376 Exclusivity restraints,  352, 365–367, 369–370 Explicit bundling,  104–105 Export Trading Company Act of  1982, 445 Ex post exclusive dealing,  318–320 Express agreements, tacit collusion and, 467n6 Externalities exclusive dealing, role in,  311–313 negative contracting externalities, exclusionary conduct and,  7–13, 31–32 Facebook, 176 Farrell, Joseph,  266, 554 Faulkner, Robert R.,  598 Federal Acquisitions Regulations,  519 Federal Reserve Bank,  536 Federal Trade Commission (FTC) bid rigging and,  501 exclusionary conduct and,  16n30 exclusive dealing and,  306, 309–311 Japanese enforcement structure compared, 195

Merger Guidelines. See Merger Guidelines quality commitment discounts and,  92 RRC theory and,  67 screening and,  524, 606 standard setting and,  559, 570, 574–575, 578, 580 tacit collusion and,  477 tying and,  331 vertical restraints and,  353 Federal Trade Commission Act of  1914, 189, 306, 331, 559, 574–575 Feinberg, Robert M.,  431, 433 Ferrari, Stijn,  407 Fidelity discounts as exclusionary conduct, 5n5, 18n37, 32n57 Finland, cartels in,  443 First-price sealed-bid (FPSB) auction,  499– 500, 503–505, 510, 516, 518 Fisher, Franklin M.,  48 “Five Forces,”  255, 258–260, 263 Focal points, tacit collusion and,  477–478, 483–486 Folk theorem collusion and,  417n5 tacit collusion and,  469, 479–480 Fölster, Stefan,  443 Fonseca, Miguel A.,  434 Foreclosure quality commitment discounts and,  102 tying and,  337–338 Forest Service,  511–513 Fox, Eleanor M.,  159 FPSB (First-price sealed-bid) auction. See First-price sealed-bid (FPSB) auction Franchising, 387–410 in Belgium,  395, 407 in Canada,  388–389 contract adaptation and,  405–408 in EU,  387, 405–408 exclusive dealing and,  393–395 exclusive territories and,  393, 395–398 in Germany,  388 historical background,  388–392 mandatory purchase requirements in,  401 overview,  387–388, 408–409 resale price maintenance and,  393, 403–405 role of antitrust law,  393–408

Index   629

statistics, 390–392 traditional versus business-format franchising, 390 tying and,  393, 399–403 in United States,  387–392 Frandata Corporation,  395–396 Freedom of Information Act of  1966, 519 Free-riding, 278–281 in brand image retail services,  280–281 certification services and,  280–281 in retail services,  278–279 “showrooming,” 279 Freiburg School. See Ordoliberalism French Association for Standardization,  557 Friedman, James W.,  478 Froeb, Luke M.,  453, 523, 535, 542 FTC. See Federal Trade Commission Fudenberg, Drew,  492 Fuji Iron and Steel Company,  193 Fuller, John,  598 Gabel, H. Landis,  598 Gal, Michal S.,  159 Gallo, Joseph C.,  444 Game theory “burning the dollar” game,  491 collusion and,  416–417 tacit collusion and,  478–479 Garces-Tolon, E.,  494 Gasoline industry, tacit collusion in,  473–474 Gates, Bill,  338 Geis, Gilbert,  598 Geist, A.,  547, 549 Geithman, Frederick E.,  395 General Agreement on Tariffs and Trade (GATT), 193 General Electric,  368, 444 Genesove, David,  452, 539, 542, 599 Geradin, Damien,  180 Gerlach, Heiko,  415, 420, 423, 430–431, 587 Germany antitrust law in,  163 cartels in,  444–445, 455 Competition Authority,  608 corporate governance and compliance in, 588 franchising in,  388

screening in,  608 vertical restraints in,  384 Gertner, Robert H.,  469–471 Gilbert, Richard,  554 Gillette, 268 Glaxo, 101 Goeree, Jacob K.,  500 Gonzalez, Tanja Artiga ???,  589, 607 Google,  176, 260, 272 Gould, J. R.,  287 Gould, Jay,  128 Governance. See Corporate governance and compliance Graham, D.,  502, 504–506 Granitz, Elizabeth,  67 Grantback provisions, intellectual property rights and,  140–141 Great Atlantic and Pacific Tea Company (A&P),  43, 58, 268 Great Britain. See United Kingdom GREE, 209 Green, Edward J. on collusion,  421 on corporate governance and compliance, 587 on screening,  530, 532 on tacit collusion,  464, 471, 476–477, 479, 490 Greenstein, Shane,  554 Grimes, Warren,  403 Günster, Andrea,  596 Haltiwanger, John,  419 Han, Martijn A.,  588 Hanazono, Makoto,  423 Hand, Learned,  126 Harm to consumers,  30 Harper, Martha Matilda,  389 Harper Beauty Shops,  389 Harrington, Joseph E., Jr. on cartels,  419, 422, 427–429, 432–433, 446, 452, 457 on corporate governance and compliance, 600 on screening,  531–532 on tacit collusion,  477 Harris, H. Stephan,  195 Harris, J. H.,  523, 525

630   Index Harsanyi, John C.,  487–488 Hart-Scott-Rodino Act of  1976, 268 Hatch-Waxman Act of  1984, 68, 143, 149 Health insurance, exclusive dealing in, 314–315 Health maintenance organizations (HMOs), 95 Hearing aid industry, exclusive dealing in, 307–309 Heeb, Randall,  3, 365 Heide, Jan B.,  304, 394 Heimler, Alberto,  234, 250 Hemphill, C. Scott,  56 Hendricks, Ken,  498, 506, 541, 587 Herling, John,  598 Hewlett-Packard,  92, 556 Heyer, Ken,  370 Higgins, Richard S.,  62 High-technology markets, tying in,  345–346 Hildebrand, V.,  538, 547 Hillman, Arye L.,  564 Hinloopen, Jeroen,  436 HMOs (Health maintenance organizations), 95 “Hold-up problem,”  7n10 Hopenhayn, H.,  507–509 Hörner, J.,  480 Housekeeper (software),  229–230 Hovenkamp, Erik,  329 Hovenkamp, Herbert on exclusive dealing,  320 on innovation,  132, 148 on predatory pricing,  50, 53 on quantity commitment discounts,  94 on tying,  329, 334, 338, 343, 345–347 on vertical restraints,  380 Howard Johnson,  389 Hu, A.,  518 Hub-and-spoke cartels,  128–129 Hubei Salt Industry Group,  224–225 Huck, Steffen,  434 Hungary, abuse-of-dominance in,  237 Hüschelrath, Kai,  600, 608 Husted, Thomas A.,  433 Iacobucci, Edward M.,  351 Ichitaro (software),  206

ICN. See International Competition Network Idiosyncratic per se rule, tying and,  330–331 IEC (International Electrochemical Commission), 556 IFA (International Franchise Association), 395–396 Image Technical Service,  401 IMF (International Monetary Fund),  193 Imhof, Lorens A.,  492 Imperfect observability and monitoring, collusion and,  420–422 Implicit bundling,  104–105 Incentive pay,  590–591 Incomplete information and communication, collusion and,  422–423 India abuse-of-dominance in,  236–237, 245–248 Competition Act of  2002, 245, 247 Competition Commission,  246–247 MCX Stock Exchange,  247 National Stock Exchange,  247 Indian Head,  562 Industry standard setting. See Standard setting Inference, screening as problem of,  525–530 Bayesian hypothesis testing and,  525–528 testing versus screening,  528–530 Information deficiencies, dominant firms and, 168–169 Initial capital investment as communication of tacit collusion,  490–491 Injunctions, patents and,  572–574 Innovation, 132–149 aggressiveness of antitrust law and,  145–148 dominant firms and,  171–178 intellectual property rights and,  134–141 leniency toward joint conduct, role of antitrust law in,  144–145 overview,  132–134, 148–149 patents and,  134–141 penalizing noncompetitive practices, role of antitrust law in,  143–144 role of antitrust law in promoting,  141–148 Intel Corporation,  92, 95, 101, 176–177, 204–205, 260 Intellectual property rights bundling, 343–344

Index   631

in China,  225–226 empirical evidence,  138–139 in EU,  142, 146–147 exclusive rights in,  135–136 grantback provisions and,  140–141 innovation and,  134–139 patents. See Patents prospect theory and,  137–138 relationship with antitrust law,  139–141 social benefits versus social costs,  136–137 standard setting and,  567–580 venture capitalists and,  138 Internal monitoring, screening in,  607–608 International Competition Network (ICN),  234, 237–240, 449 International Electrochemical Commission (IEC), 556 International Franchise Association (IFA), 395–396 International Monetary Fund (IMF),  193 International Organization for Standardization (ISO),  555–557 International Telecommunications Union (ITU), 556–557 International Trade Commission (ITC),  68, 573, 581 Internet distribution restrictions,  367 online retailing, resale price maintenance (RPM) of,  277–301. See also Resale price maintenance (RPM) Internet Explorer,  175, 259, 561, 565 Interoperability standards,  555 Isaac, R. Mark,  466, 515–517 ISO (International Organization for Standardization), 555–557 Italy, screening in,  606 ITC (International Trade Commission),  68, 573, 581 ITU (International Telecommunications Union), 556–557 Ivaldi, Marc,  469, 494 Ivoclar, 322–324 Jamaica, abuse-of-dominance in,  237, 240–241 Jamison, Julian,  480

Japan Act against Delay in Payment of Subcontract Proceeds, Etc. to Subcontractors, 192 Action of Prohibition of Privation Monopolization and Maintenance of Fair Trade (AMA),  188–197, 199, 201–203, 209–210, 230–231 below-cost pricing in,  202–203 cartels in,  191, 445 China compared,  230–232 civil procedure in,  199 competition policy versus industrial policy in, 193 control in,  201 corporate governance and compliance in, 588 discriminatory treatment in,  207–209 enforcement structure in,  195–199 entrepreneurs in,  200 EU enforcement structure compared,  201 exclusionary conduct in,  200–209 exclusive dealing in,  203–205 Fair Trade Commission,  189–190, 193–210, 231 future trends,  209–210 Guidelines for Exclusionary Private Monopolization under the Antimonopoly Act,  200–204, 206–207, 231–232 historical background of AMA,  190–194 Intel, 204–205 Intellectual Property Guidelines,  209 JASRAC, 205 Law of the Elimination of Excessive Concentration of Economic Power (LECP),  191, 193 Microsoft, 206 Ministry of Internal Affairs and Communications, 208 Ministry of International Trade and Industry, 191 monopoly in,  188–210 1977 Amendment to AMA,  193–194 original AMA,  190–193 overview,  188–190, 209–210 particular field of trade in,  201–202

632   Index private monopolization in,  199–202 procedures in monopoly cases,  196 refusal to supply in,  207–209 sanctions in,  196–199 screening in,  606 Shinpan hearings,  196 SSNIP test,  201–202 Structural Impediments Initiative,  194 substantial restraint of competition in,  202 Supreme Commander of Allied Forces and, 191 Supreme Court,  196, 202, 208 surcharges in,  196–199 Toshiba Elevator, 206–207 tying in,  205–207 US enforcement structure compared,  189, 201 Japanese auction,  499 Japanese Industrial Standards Committee,  557 Japanese Society for Rights of Authors, Composers and Publishers (JASRAC), 205 JEDEC (Joint Electron Device Engineering Council), 575 Jenny, Frédéric,  418 Jiménez, J. L.,  539, 545 John Hancock,  314 Johnson & Johnson,  95, 100–101, 105, 113 Joint Economic Committee,  540 Joint Electron Device Engineering Council (JEDEC), 575–576 Joint venture, standard setting as,  559–561 Joskow, Paul L.,  48, 50, 56 Juang, Wei-Torng,  492 Judge, George,  606, 609–610 Jullien, Bruno,  423–424 Justice Department (DOJ) bid rigging and,  514, 604 cartels and,  444–445, 448, 454, 456–457, 459, 589, 591 CRRs and,  89–90 exclusive dealing and,  314–315 intellectual property rights and,  139–140 Japanese enforcement structure compared, 195 Merger Guidelines. See Merger Guidelines monopolization and,  155

price fixing and,  605 quality commitment discounts and,  93–94, 110, 112–114, 116–117 RRC theory and,  67 screening and,  523, 534, 606–607 standard setting and,  565, 579–580 tacit collusion and,  477 vertical restraints and,  353 Just Systems,  206 Kandori, Michihiro,  492 Kaplow, Louis,  425 Karpoff, Jonathan M.,  596 Kaserman, David L.,  596 Katz, Michael L.,  554 Kell West Regional Hospital,  112–113 Kim, J.,  517–518 Kime, Posey T.,  190 Kitch, Edmund,  137 Klein, Benjamin on exclusive dealing,  304, 325 on franchising,  389 on quantity commitment discounts,  99 on reputation effects,  595 on resale price maintenance,  277 on RRC theory,  67 on tying,  343 on vertical restraints,  363 Klein-Murphy theorem,  368 Klemperer, Paul,  266 Klevorick, Alvin K.,  48, 50, 56 Klor's, 123–125 Kodak,  58, 377, 401–402, 408 Koller, Roland H.,  43 Koukou Guard (software),  229 Kovacic, William,  429–430, 465, 526, 528 Koyak, Robert A.,  453, 535, 542 Kraakman, Reiner,  591 Krattenmaker, Thomas G.,  65, 127 Kreps, David,  44 Krishna, Vijay,  479 Krishnan, Ranjani,  364 Kroc, Ray,  396 Kwasnica, A.,  516–517 Lafontaine, Francine,  343, 371, 387, 400, 405 Laitenberger, Ulrich,  600

Index   633

Lambson, Val E.,  479 Lande, Robert H.,  454, 596 Landeo, Claudia M.,  315–316 Langus, Gregor,  596 Lanham Act of  1946, 399 Larouche, Pierre,  153 Latvia, abuse-of-dominance in,  238, 241 Lawrence, Paul R.,  599 Lee, D. Scott,  596 Leegin Creative Products, Inc.,  403–404 Leffler, K.,  595 Lemley, Mark A.,  138, 555 Leniency programs cartels and,  424–430, 435–437 experimental evidence,  435–437 Lenovo, 101 LePage's,  91, 105 Lerner Index, monopsony and,  73 Levenstein, Margaret C.,  442, 443–445, 447–448, 451–452, 454–457, 587 Leverage theory, tying and,  333–334 Levin, J.,  513–514 LIBOR, screening and,  536–538, 606–607, 609–610 Liebowitz, Stan J.,  345 Liefmann, Robert,  444 Lin, Ping,  188 Lipman, Barton L.,  492–493 Li Qing,  219 Liquidated damages,  317n40 Lithuania, abuse-of-dominance in,  237 Localized price cutting,  43n7 Lock-in doctrine,  332 Long, Clarisa,  138 “Long purse” theory,  42, 45 Lopatka, John E.,  72 Lopomo, G.,  505–506 Lotus 1-2-3 (software),  206 Loyalty contracts,  351 Loyalty discounts economics of,  100–102 as exclusionary conduct,  5n5, 18n37, 32n57 LTE,  558, 569 Magrane-Houston, 379 Mailath, George J.,  492, 505–506 Malik, Arun,  425

Mandatory purchase requirements in franchising, 401 MAP (Minimum advertised price) policies, 291–292 Margolis, Stephen E.,  345 Market power enhanced market power, exclusionary conduct and,  30 tying and,  332–333 Market share, screening for,  605 Market share discounts CRRs and,  90n2 as exclusionary conduct,  5n5, 18n37, 32n57 Markey, Edward,  524 Marshall, Robert C. on bid rigging,  501–506, 511–512, 517 on corporate governance and compliance, 587 on screening,  531, 533, 548 on tacit collusion,  464, 472–473 Martin, Gerald S.,  596 Marvel, Howard P.,  287, 304, 308, 364, 365, 399 Marx, Leslie M. on bid rigging,  501–503, 505–506, 517 on corporate governance and compliance, 587 on screening,  531, 533, 548 on tacit collusion,  464, 472–473 Mason, Charles F.,  434–435 Masten, Scott E.,  317 Master Kang,  217–218 Mathematical laws, screening based on,  606 Matsushima, Hitoshi,  420 Matutes, Carmen,  555 Mayer Laboratories,  326 Mazzeo, Michael J.,  253 McAfee, R. Preston,  498, 505, 507–508, 587 McCafferty, Stephen,  287 McCormick Harvesting Company,  388 McDevitt, Ryan C.,  253 McDonald's,  343, 354, 389, 396, 405 McGee, John S.,  40–45, 48, 51, 56 McGuire Act of  1952, 373 McMillan, J.,  505, 507–508 Medco, 101 Medicaid, 314 Medicare, 314

634   Index Mehta, Kirtikumar,  234, 250 Merger Guidelines Japanese enforcement structure compared, 201 tacit collusion and,  477 Mergers. See also Vertical mergers in EU,  269n9 Mestmäcker, Ernst-Joachim,  156–157, 163 “Metering,” 334–335 Metz, Albert,  537, 543, 609–610 Meurer, Michael J.,  138 Mexico abuse-of-dominance in,  237 cartels in,  454n33 Competition Commission,  454n33 screening in,  524, 606 Michael, Steven C.,  400 Microsoft competition on merits and,  175 competitive advantage and,  272 as dominant firm,  176–177 exclusive dealing and,  319n44 intellectual property rights and,  141 in Japan,  206 RRC theory and,  64 standard setting and,  565–566 tying and,  338 unilateral squeezing and,  126 value creation and capture framework and, 259 vertical restraints and,  372, 378, 384 Microsoft Office (software),  556 Microsoft Word (software),  206 Middle management, corporate governance and compliance and,  593–594 Miksch, Leonhard,  178–179 Milgrom, Paul,  53, 499–500 Miller, D.,  536, 543 Miller, Nathan H.,  457 Miller-Tydings Act of  1937, 373 Mills, David E.,  40 Minimum advertised price (MAP) policies, 291–292 Misuse, tying and,  345–346 Miwa, Yoshiro,  195 Monopoly business strategy and,  253–272. See also

Business strategy in China,  210–232. See also China in developing countries,  234–251. See also Abuse-of-dominance dominant firms,  153–184. See also Dominant firms exclusionary conduct and,  3–36. See also Exclusionary conduct innovation and,  132–149. See also Innovation in Japan,  188–210. See also Japan predatory buying,  72–87. See also Predatory buying predatory pricing,  40–58. See also Predatory pricing quantity commitment discounts,  89–117. See also Quantity commitment discounts RRC theory and,  62–70. See also Raising rivals' costs (RRC) theory squeeze claims,  120–129. See also Squeeze claims Monopsony competition compared,  74 dominant buyer and,  75–76 economics of,  73–77 input markets and,  74–76 Lerner Index and,  73 output markets and,  76–77 predatory buying and,  72–73. See also Predatory buying Morse, Adair,  599 Morton Salt,  345–346 Mossel Jamaica Limited,  240 Motchenkova, Evgenia,  595 Motion Picture Patents Company,  399 Motorola, 572 Motta, Massimo,  428–430, 596 Mueller, Willard F.,  395 Mullin, Wallace P.,  452, 539, 542, 599 Multimarket contact, collusion and,  419–420 Muris, Timothy J.,  68 Murphy, Kevin M.,  89, 99, 304, 325 Myerson, R.,  500 Naked exclusion theory of exclusive dealing, 313–316 NASDAQ, screening and,  523–525, 540,

Index   635

606–607 Nash equilibrium Bayesian Nash equilibrium,  488 bid rigging and,  500, 504 cartels and,  426–428 tacit collusion and,  469–470, 480–482, 490–493 vertical restraints and,  370 National Electrical Code,  562 National Fire Prevention Association,  562 National Industrial Recovery Act of  1933, 446 National Macaroni Manufacturer's Association, 559 National Semiconductor,  574 NavTeq, 69 Negative contracting externalities, exclusionary conduct and,  7–13, 31–32 Negotiated Data Solutions,  574 Netherlands cartels in,  596 screening in,  606 Netscape (browser),  338, 566 Neven, D.,  494 New York Transportation Department,  510–511, 541 Nigrini, Mark,  606 Nihon Network Vision,  203 Nike, 299 “Nine No-Nos,”  139–140 Nippon Soda,  449 Nippon Steel,  193 Nippon Telegraph,  208 Nocke, Volker,  423 Nokia, 69 Nold, Frederick C.,  431, 453 Nonlinear pricing schedules,  351 Normann, Hans-Theo,  434 Norm creation,  594–595 Nowell, Clifford,  435 Number of firms, collusion and,  418 Nurski, Laura,  395, 407 O'Brien, Daniel P.,  355, 371 Observable posted prices, tacit collusion and, 473–474 OECD Global Forum on Competition Policy, 240–241

Oechssler, Jorg,  434 Offerman, Theo,  500, 518 Ohashi, Hiroshi,  188, 195 Oligopoly, tacit collusion in,  464–494. See also Tacit collusion Oligopoly and the Theory of Games (Friedman), 478 Omnibus Trade Act of  1988, 194 Onderstal, S.,  518 One-shot auctions, bid rigging in,  504–507 Online retailing, resale price maintenance (RPM) of,  277–301. See also Resale price maintenance (RPM) Optical industry, exclusive dealing in,  310–311 Orbach, Barak,  120 Ordoliberalism, 160–164 competition versus economic power, 162–163 excessive pricing and,  178–179 as general legal and economic theory, 160–161 overview,  160, 164, 183–184 in TFEU,  163–164 threat from economic power and,  161–162 Ordover, Janusz A.,  42, 44, 50, 65, 235 Organization for Economic Cooperation and Development (OECD) Global Forum on Competition Policy,  240–241 Over, Mead, Jr.,  449 Overpurchasing model, RRC theory compared, 64 Package licensing,  343–344 Pareto efficiency, vertical restraints and,  358 Parker, Christine,  598 Partial exclusive dealing,  304 Partial exclusivity discounts as exclusionary conduct, 5n5, 18n37, 32n57 Patent Act of  1952, 139n15, 143, 330–331, 345 Patent and Trademark Office, RRC theory and, 67 Patents cross-licenses of,  571–572 empirical evidence,  138–139 ex ante bargaining and,  579–580 exclusionary conduct and,  4n3 exclusive rights in,  135–136

636   Index failure to disclose standard-essential patents, 574–577 injunctions and,  572–574 innovation and,  134–141 overly broad declarations of standard-essential patents,  577–579 package licensing of standard-essential patents, 577–579 prospect theory and,  137–138 reasonable and non-discriminatory licensing commitments,  569–574 royalties, 570–571 sale and assignment of,  574 social benefits versus social costs,  136–137 standard setting and patent holdup, 568–569 venture capitalists and,  138 Pathmark, 268 Pearce, David G.,  479, 482, 490 Peck, James,  364 Peltzman, Sam,  443 PepsiCo,  35, 95, 270, 283 Perdiguero, J.,  539 Pereira, P.,  540 Peru, abuse-of-dominance in,  241 Pesendorfer, M.,  514–515, 517 Petroleum Marketing Practices Act of  1978, 398 Philips, 578 Phillips, Owen R.,  434–435 Phlips, Louis,  49 Pinto, Brijesh P.,  523 Pioneer Foods,  248 Pitofsky, Robert,  280, 287 Pittman, Russell,  235 Plastic bags, tacit collusion and,  474–477 Plott, Charles R.,  466, 515–516 Poland, abuse-of-dominance in,  237 Polo, Michele,  428–430 Porter, Michael,  255–260, 263 Porter, Robert H. on bid rigging,  506, 510–511, 513 on cartels,  421 on screening,  530, 532, 541, 545–546 on tacit collusion,  471, 476–477, 479, 490 Posner, Richard A.,  444, 493 Predatory buying,  72–87

allocative distortion,  80 caselaw, 77–78 completely different rivals and monopoly power, 82–84 completely different rivals and no monopoly power,  78–82 dominant seller and,  83 economics of,  73–77 monopsony and,  72–73 no allocative distortion,  79 overview,  72–73, 87 same competitors in input and output markets, 84–86 Predatory pricing,  40–58 Areeda-Turner rule,  47–49, 52–54 Brooke Group, 49–53 in China,  218 economics of,  41–45 enforcement and,  45–49 in EU,  46, 51 as exclusionary conduct,  17–18, 21 intent, 46–47 localized price cutting,  43n7 “long purse” theory,  42, 45 overview,  40–41, 57–58 price-cost comparisons in,  49–50, 54–56 recoupment in,  50–51, 56–57 Preston, L. E.,  287 Price ceilings,  352 Price-cost comparisons in predatory pricing, 54–56 Price discrimination in China,  220–221 tying and,  334–337 Price fixing screening for,  604–605 standard setting as venue for,  559 Price floors,  352 Price leadership, tacit collusion and,  472n17, 477–478 Price setting by cartels,  449–452 Price squeezes,  121 Pricing, predatory. See Predatory pricing Princo, 578–579 Private value auction,  499–500 Procompetitive nature of exclusionary conduct

Index   637

examples, 34–35 overview, 32–34 Procter & Gamble,  268 Producer surplus,  255, 258–260 Production cost efficiencies,  342 Property right creation, exclusive dealing as, 307–311 Prospect theory, patents and,  137–138 PSKS, Inc.,  403–404 Purchase share discounts as exclusionary conduct, 5n5, 18n37, 32n57 Qihoo 360 (software),  229 Qualcomm, 576 Quality and safety standards,  555 Quality control efficiencies,  342–343 Quantity commitment discounts,  89–117 attribution test for exclusion,  105–109 bundling and,  102, 104–105 contracts that reference rivals (CRRs), 89–90 demand curves and,  95–97 division of gains from,  98–100 economics of,  94–102 economies of scale and,  103n33 exclusionary conduct and,  102 flaws with attribution test,  111–116 foreclosure and,  102 HMOs and,  95 indicators of harm to competition,  116 legal context,  91–94 liability thresholds and attribution test,  110 loyalty discounts,  100–102 mutual gains from,  95–98 overview,  89–91, 117 policy context,  91–94 potential harm to competition,  102–105 tests for harm to competition,  105 uses of attribution test,  111–116 Queen City Pizza,  343, 402, 408 Raiff, M.,  531, 548 Railroad industry, concerted squeezing in, 128 Raising rivals' costs (RRC) theory,  62–70 assessment of,  65–66 counterstrategies, 64

demand effects and,  64 efficiencies and,  64 limitations of,  65–66 overpurchasing model compared,  64 overview, 70 policy implications of,  67–68 strengths of,  65 vertical mergers and,  68–70 Rajgopal, Shivaram,  599 Rambus, 575–576 Ramey, Garey,  516 Ramseyer, J. Mark,  195, 313–316 RAND (Reasonable and non-discriminatory) licensing commitments. See Patents Rapson, David,  541 Rasmusen, Eric B.,  313–316 RealNetworks, 319n44 Reasonable and non-discriminatory (RAND) licensing commitments. See Patents Recession, effect on formation of cartels, 447–449 Recoupment in predatory pricing,  56–57 Refusals to deal in China,  221–222 overview, 121 Refusal to supply in Japan,  207–209 Regibeau, Pierre,  555 Remedies in Japan,  196–199 liquidated damages,  317n40 tying, 347–348 Renren, 245 Repeated auctions, bid rigging in,  507–510 Resale price maintenance (RPM) anticompetitive motivations for,  292–297 antitrust standard of analysis,  297–298 brand image retail services, control of free-riding in,  280–281 certification services and,  280–281 compensation of retailers for providing interbrand retailing services,  283–286 discretion over retail services with interbrand but not interretailer demand effects, 281–283 effective retail distribution, preserving, 281–292 in EU,  297–301

638   Index Resale price maintenance (RPM) (cont.) franchising and,  393, 403–405 free-riding, control of,  278–281 impulse purchases and,  287 manufacturer control of online retailing and, 299–301 manufacturer-motivated, 292–294 manufacturer's distribution network damaged by retailer price discounting, 286–289 minimum advertised price (MAP) policies and, 291–292 of online retailing,  277–301 outlets hypothesis,  287 overview,  277–278, 301 procompetitive rationales for,  298–299 protection of retail distribution network, use for,  289–292 regulation of,  297–301 retailer-motivated, 294–297 retail services, control of free-riding in, 278–279 “showrooming,” 279 slotting allowances and,  285 in United States,  297–301 vertical restraints and,  368, 372–374 Restraint of trade in Japan,  202 Rey, Patrick,  369, 418, 423–424, 429–430 Rey-Stiglitz theorem,  369 Rhône-Poulenc,  449, 457 Richard, J.-F.,  502, 511–512 Riley, John G.,  500, 564 Riordan, Michael H.,  52 Risk dominance as obstacle to tacit collusion, 486–490 Risk-sharing, tying and,  344–345 Rob, Rafael,  492 Roberts, John,  53 Rodger, Barry J.,  598 Rojas, Christian,  434 Romania, abuse-of-dominance in,  237 Rosenfield, Andrew M.,  469–471 Rotemberg, Julio J.,  419, 530, 532 Royalties patents, 570–571 vertical restraints and,  351 RPM (Resale price maintenance),  277–301.

See also Resale price maintenance (RPM) RRC theory. See Raising rivals' costs (RRC) theory Rubinstein, Ariel,  488 Rule-of-reason analysis in China,  216–217 exclusionary conduct,  4–6 Russia abuse-of-dominance in,  236, 241–244 Federal Antimonopoly Service,  242, 244 Sabourian, Hamid,  492 Saft, Lester F.,  343 Saks Fifth Avenue,  367 Salant, Steven W.,  431 Salinger, Michael A.,  342 Saloner, Garth,  44, 65, 419, 530, 532, 554 Salop, Steven C.,  52, 65–66, 127, 312 Samuelson, W.,  500 Sanchirico, Chris,  422, 508, 530, 533 Sanctions in abuse-of-dominance cases, 249–250 San Francisco Peace Treaty,  191 Sannikov, Yulij,  421 Sass, Tim R.,  394, 398 Saurman, David S.,  398 Scalia, Antonin,  172, 415n1 Scharfstein, David,  588 Schechter, Laura,  606 Scheffman, David T.,  62, 65–66, 312, 453 Scheinkman, Jose A.,  479 Schelling, Thomas,  478 Schenone, P.,  479 Scherer, Frederic M.,  47, 288, 478 Schinkel, Maarten Pieter,  153 Schmalensee, Richard,  334 Schmidt, Klaus M.,  588 Schott Glass,  246 Schultz, Paul H.,  523–525, 540, 545, 606 Schumpeter, Joseph,  137, 148, 171 Schwartz, W.,  65 Scott Morton, Fiona,  254 Screening, 523–551 Bayesian hypothesis testing and,  525–528 in Bertrand markets,  530, 541 for bid rigging,  602–604 in Brazil,  524, 606

Index   639

in Canada,  606 cartels and,  430–433 collusive markets other than price screens, based on,  548–549 compliance, use in,  608–609 in corporate governance and compliance, 600–610 in Cournot markets,  530 empirical screens,  534 first and second moments of price distribution, based on,  534–540 in Germany,  608 inference, as problem of,  525–530 in internal monitoring,  607–608 in Italy,  606 in Japan,  606 LIBOR and,  536–538, 606–607, 609–610 for market share,  605 mathematical laws, based on,  606 in Mexico,  524, 606 NASDAQ and,  523–525, 540, 606–607 in Netherlands,  606 overview,  523–525, 549–550 for price fixing,  604–605 in South Africa,  606 structural models, based on,  540–548 success of,  606–607 testing versus,  528–530 theoretical screens,  530–534 SCSI (Small Computer Systems Interface),  555 Seabright, Paul,  494 Sealy, 395 Second-price sealed-bid (SPSB) auction,  499–500, 504–505 Securities and Exchange Commission, screening and,  523, 606–607 Seira, E.,  513–514 Self-distribution, tying and,  344–345 Selten, Reinhard,  44, 436–437, 487 Senegal, abuse-of-dominance in,  241–242 Senior management, corporate governance and compliance and,  592–593 Separate products test, tying and,  341 Shanda Interactive Entertainment Ltd.,  244 Shandong Weifang Shuntong Pharmaceuticals Co., Ltd.,  223–224, 245 Shane, Scott,  395–397

Shanghai Xuanting Entertainment Co. Ltd., 244 Shapiro, Carl,  147–148, 554–555 Shavell, Steven,  425 Shenzhen Tencent Computer System Co. Ltd., 228 Sherman Act of  1890 attempted monopolization under,  156 collusion under,  415 competition on merits under,  158 decision theory and,  169 dominant firms under,  153–184. See also Dominant firms exclusive dealing and,  305–306 franchising under,  395 intellectual property rights and,  140–142 Japanese antitrust law compared,  189 monopolization under,  154–155 patents and,  139n15 predatory pricing under,  40, 45, 52, 58 quality commitment discounts and,  110 special responsibility of dominant firms under, 157 standard setting and,  562, 574 tacit collusion and,  464, 493 tying under,  330–331 vertical restraints under,  372, 379 Shire, 123 “Showrooming,” 279 Sibley, David S.,  523 Sidak, Joseph G.,  431, 453 “Signal jamming problem,”  27n48 Simcoe, Timothy,  555, 567 Sinclair, 394 Singer Sewing Machine Company,  388 Single-product pricing as exclusionary conduct,  18, 24 Sivadasan, Jagadeesh,  454 Skrzypacz, Andrzej,  421–422, 452, 507–509 Slade, Margaret E.,  371, 387 Slotting allowances resale price maintenance and,  285 unilateral squeezing and,  127n34 as vertical restraints,  352 Slovak Republic, abuse-of-dominance in,  237 “Small but significant and nontransitory increase in price” (SSNIP) test,  201–202, 216

640   Index Small Computer Systems Interface (SCSI),  555 Smith, Richard L., II,  398 Snider, Connan,  609 Snyder, Edward A.,  89, 317 Soetevent, Adrian,  436 Soft drink industry, exclusionary conduct in, 35 Sokol, D. Daniel,  586, 595, 598 Sonnenfeld, Jeffrey,  599 Sony,  204, 578 South Africa abuse-of-dominance in,  237–238, 241, 248–249 Competition Authority,  606 Competition Commission,  248 Competition Tribunal,  249 screening in,  606 South African Airways,  248 South Korea, corporate governance and compliance in,  600 Southwest Airlines,  265 Spagnolo, Giancarlo,  420, 429, 457, 588 Spar, Debora L.,  445 Special responsibility of dominant firms,  157, 166–167 Spengler, Joseph,  355 Spier, Kathryn E.,  315–316 SPSB (Second-price sealed-bid) auction,  499–500, 504–505 Spulber, Daniel F.,  431–432 Squeeze claims,  120–129 cartels and,  128–129 concerted squeezing,  122–125, 128–129 essential facilities,  121 in nonintegrated firms,  127 overview,  120–122, 129 price squeezes,  121 railroad industry and,  128 refusals to deal,  121 slotting allowances and,  127n34 unilateral squeezing,  122–127 in vertically integrated firms,  125–127 SSNIP (“Small but significant and nontransitory increase in price”) test,  201–202, 216 SSOs. See Standard-setting organizations Stacchetti, Ennio,  479, 490

Standard auction models,  499–500 Standard Fashion,  379 Standardization Administration of China,  557 Standard Oil Company,  41–42 Standard setting,  554–581 compatibility standards,  555 corruption of process,  561–565 development of standards,  555–557 disparagement of alternative proposals and, 564–565 ex ante bargaining and,  579–580 exclusionary conduct, as venue for,  559 failure to disclose standard-essential patents, 574–577 intellectual property rights and,  567–580 interoperability standards,  555 as joint venture,  559–561 overly broad declarations of standard-essential patents,  577–579 overly broad standard specifications, 577–579 overview,  554–555, 580–581 package licensing of standard-essential patents, 577–579 patent holdup and,  568–569 price fixing, as venue for,  559 quality and safety standards,  555 reasonable and non-discriminatory licensing commitments,  569–574 types of standards,  555 unilaterally determined standards,  565–567 vote stacking and,  561–564 Standard-setting organizations (SSOs) ANSI-approved SSOs,  557 competition principles,  557–565 innovation and,  144, 146 international SSOs,  556 national SSOs,  557 regional SSOs,  557 Steiner, Peter O.,  360 Stewart, Potter,  116 Stigler, George J.,  455, 465, 470, 478, 518–519, 530 Stiglitz, Joseph,  369, 423 Stock-based cartel event studies,  595–596 The Strategy of Conflict (Schelling),  478 Strong explicit collusion,  468–477

Index   641

Structural Impediments Initiative,  194 Sumitomo, 449 Summit Technology,  578 Sun, P.,  505 Super  301, 194 Suppiger, 345–346 Suslow, Valerie Y.,  442, 443–445, 447–448, 451–452, 454–457, 587 Sustainability of competitive advantage, 264–266 Sutton, John,  454 Sweden, cartels in,  443–444 Switzerland, corporate governance and compliance in,  598 Sylvania, 375 Symeonidis, George,  443, 455 Symmetry, collusion and,  418 Synergies, 267–268 Tacit collusion,  464–494 antitrust litigation and,  493–494 “arrive by reasoning,”  482–486 Bayesian Nash equilibrium,  488 blundering into,  492–493 “burning the dollar” game and,  491 communication in,  466–468 convergent-elimination view,  483–486 defined, 464 direct interaction and,  467n7 economics versus legal terminology,  467 evolution of theory,  477–480 express agreements and,  467n6 focal points and,  477–478, 483–486 folk theorem and,  469, 479–480 game theory and,  478–479 in gasoline industry,  473–474 illustrations, 473–477 initial capital investment as communication, 490–491 initial phase of collusion,  480–493 Nash equilibrium and,  469–470, 480–482, 490–493 observable posted prices and,  473–474 overview, 464–468 plastic bags and,  474–477 price leadership and,  472n17, 477–478 risk dominance as obstacle to,  486–490

self-enforcement of,  479n26, 480–482 strong explicit collusion versus,  468–477 unobservable bid prices and,  474–477 weak explicit collusion versus,  468–477 Taiwan, abuse-of-dominance in,  241 Tan, G.,  506 Tan, Tommy C.-C.,  483 Tarbell, Ida M.,  41 TeleAtlas, 69 Telephone East Corporation,  208 Telser, Lester,  368 Tencent Technology Co. Ltd.,  216, 228–230 Tennessee Valley Authority,  603 Terminal Railroad Association of St. Louis, 128 Territorial restraints,  352, 364–365, 369 Testing versus screening,  528–530 TFEU. See Treaty on the Functioning of the European Union Thompson, John S.,  596 3M,  91, 105 TIBOR, 607 Tiffany, 367 TOK, 242 Tom Tom,  69 Topel, Robert H.,  89 Toshiba, 204 Toshiba Elevator,  206–207 Toys “R” Us,  129 Trade-offs in business strategy,  257–258 Traditional franchising,  390 Transitions Optical,  310–311 Treaty of Rome (1957),  179 Treaty on the Functioning of the European Union (2009) abuse under,  154–155 collusion under,  415 competition on merits under,  158, 174 decision theory and,  166–171 dominant firms under,  153–184. See also Dominant firms excessive pricing and,  178–181 exclusive dealing and,  306–307 exploitative abuses under,  155–156 intellectual property rights under,  147n37 interaction between dominant firms and, 176–177

642   Index Treaty on the Functioning of the European Union (2009) (cont.) ordoliberalism in,  163–164 predatory pricing under,  46 protection of competitive process under,  158–159, 173–174 special responsibility of dominant firms under, 157 vertical restraints under,  372, 374 Turner, Donald F.,  46–47, 49, 54, 493 Twitter, 176 Tyco, 95 Tying, 329–348 in China,  224–225 of complements,  339–341 damages, 347–348 double marginalization and,  339–341 economies of joint provision and,  341–344. See also Economies of joint provision effects of,  333–338 as exclusionary conduct,  5n6, 19, 25 foreclosure and,  337–338 franchising and,  393, 399–403 in high-technology markets,  345–346 identification of,  329–330 idiosyncratic per se rule and,  330–331 in Japan,  205–207 leverage theory and,  333–334 lock-in doctrine,  332 market power and,  332–333 “metering,” 334–335 minimum conditions for competitive harm, 332–333 misuse and,  345–346 motives, 333–338 overview,  329–330, 348 price discrimination and,  334–337 remedies, 347–348 risk-sharing and,  344–345 self-distribution and,  344–345 separate products test and,  341 statutory coverage,  330–331 variable proportion ties,  336 vertical restraints and,  352, 365–367, 369–370, 376–378

Ukraine, abuse-of-dominance in,  238 UNCTAD (United Nations Conference on Trade and Development),  236 Unilateral squeezing,  122–127 concerted squeezing compared,  122–125 in nonintegrated firms,  127 slotting allowances and,  127n34 in vertically integrated firms,  125–127 Union Oil Company of California,  575 United Kingdom Board of Trade,  443 cartels in,  443–444 Competition Act of  1998, 598 Competition Commission,  407 corporate governance and compliance in,  597–598, 600 Office of Fair Trading,  407, 598 vertical restraints in,  353 United Nations Conference on Trade and Development (UNCTAD),  236 United Regional Health System,  112–113, 314–315 United States. See also specific Act or governmental entity abuse-of-dominance standard in,  236 bid rigging in,  501–502 cartels in,  443–446 Chinese enforcement structure compared, 225 corporate governance and compliance in, 597–598 dominant firms in,  153–184. See also Dominant firms franchising in,  387–392 Japanese enforcement structure compared,  189, 201 resale price maintenance in,  297–301 vertical restraints in,  352–353, 366, 371–384 United States Tobacco Company,  127, 325 Uno, Sosuke,  194 Unobservable bid prices, tacit collusion and, 474–477 U.S. Steel,  58, 193 Usen Broadband Networks,  203 Uzbekistan, abuse-of-dominance in,  237

Index   643

Value creation and capture framework, 255–261 added value,  260–261 consumer surplus and,  255, 258–260 enhancing value creation,  256–258 “Five Forces,”  255, 258–260 overview, 255–256 producer surplus and,  255, 258–260 rivalry and,  258–260 Van Baal, Sebastian,  279 van Damme, Eric,  488–489, 491 van der Laan, Rob,  595 van Dijk, Mathijs A.,  596 Variable proportion ties,  336 Varian, Hal,  606 Varney, Christine,  93 Veith, Tobias,  608 Venture capitalists, patents and,  138 Verboven, Frank,  395, 398, 407 Vertical integration exclusive dealing and,  304–326. See also Exclusive dealing franchising and,  387–410. See also Franchising resale price maintenance (RPM),  277–301. See also Resale price maintenance (RPM) tying and,  329–348. See also Tying unilateral squeezing and,  125–127 vertical restraints,  351–384. See also Vertical restraints Vertical mergers collusion and,  423–424 RRC theory and,  68–70 Vertical restraints,  351–384 anticompetitive uses,  368–371 in Australia,  384 buyback options,  352 in Canada,  353, 372, 384 collusion and,  423–424 consignment selling arrangements,  352 economics of,  353–367 efficiencies and,  359–364 empirical evidence,  371 in EU,  352–353, 366, 371–384 exclusive dealing and,  379–381

exclusive territories and,  374–376 exclusivity restraints,  352, 365–367, 369–370 in Germany,  384 Internet distribution restrictions,  367 law versus economics,  381–384 loyalty contracts,  351 Nash equilibrium and,  370 nonlinear pricing schedules,  351 overview,  351–353, 381–384 Pareto efficiency and,  358 price ceilings,  352 price floors,  352 resale price maintenance and,  368, 372–374 royalty contracts,  351 single firm incentives,  359–364 slotting allowances,  352 territorial restraints,  352, 364–365, 369 tying and,  352, 365–367, 369–370, 376–378 in United Kingdom,  353 in United States,  352–353, 366, 371–384 VHS videotapes,  558 Vickers, John,  51 Vickrey, W.,  500 Video Electronics Standards Association (VESA), 575 Villas-Boas, Sofia,  606, 609–610 VISX, 578 Vita Zahnfabrik,  322–324 Vladivostok Avia Open Co.,  242, 244 Volume discounts as exclusionary conduct, 18n37, 24–25 Vote stacking, standard setting and,  561–564 W Walker, J.,  516–517 Walmart,  268, 329 Weak explicit collusion,  468–477 Webb-Pomerene Act of  1918, 445 Weber, R.,  499–500 Weifang Huaxin Pharmaceuticals & Trading Co., Ltd.,  223–224, 245 WellPoint, 269 Wendy's, 354 Werden, Gregory L.,  453, 535, 542

644   Index Werlang, S.R. da Costa,  483 Westinghouse,  368, 444 Whinston, Michael D. on collusion,  416, 419, 434 on exclusive dealing,  311–312, 317 on Japan,  204 on tying,  334 on vertical restraints,  370 White, Andrew,  555 White, Lucy,  423 Wiley, John S., Jr.,  313–316 Williams, Michael A.,  498, 516, 587 Williamson, Oliver,  47–48, 360 Willig, Robert D.,  42, 50 Wilson, Robert,  44 WiMax, 558 Windows (operating system),  175, 372, 378, 561, 565–566 Windows Media Player,  372, 378 Winter, Ralph A.,  351, 363–364

Woods, Sean,  555 World Wide Web Consortium,  567, 569 Wuchang Salt,  224–225 Xu Kunlin,  221 Yamamoto, Y.,  479 Yamey, Basil,  43 Yang, Huanxing,  423 Yawata Iron and Steel Company,  193 Ye, L.,  514, 546, 548 Youle, Thomas,  609 Young, Peyton,  492 Zambia, abuse-of-dominance in,  237, 241 Zanarone, Susanne,  398, 407 Zemsky, P.,  505–506 Zhang, X. J.,  217 Zingales, Luigi,  599 Zona, Douglas J.,  510–511, 513, 541, 545–546