377 45 3MB
English Pages [396] Year 2014
THE LAW AND ECONOMICS OF CLASS ACTIONS
RESEARCH IN LAW AND ECONOMICS Series Editor: James Langenfeld Recent Volumes: Volume 20:
An Introduction to the Law and Economics of Environmental Policy: Issues in Environmental Design Edited by T. Swanson
Volume 21:
Antitrust Law and Policy Edited by J. B. Kirkwood
Volume 22:
General Issues Edited by R. O. Zerbe and J. B. Kirkwood
Volume 23:
Introduction to Benefit-Cost Edited by R. O. Zerbe
Volume 24:
Law & Economics: Toward Social Justice Edited by Dana L. Gold
Volume 25:
Research in Law and Economics Edited by R. O. Zerbe and J. B. Kirkwood
RESEARCH IN LAW AND ECONOMICS VOLUME 26
THE LAW AND ECONOMICS OF CLASS ACTIONS EDITED BY
JAMES LANGENFELD Navigant Economics and School of Law, Loyola University Chicago, Chicago, IL, USA
United Kingdom North America Japan India Malaysia China
Emerald Group Publishing Limited Howard House, Wagon Lane, Bingley BD16 1WA, UK First edition 2014 Copyright r 2014 Emerald Group Publishing Limited Reprints and permission service Contact: [email protected] No part of this book may be reproduced, stored in a retrieval system, transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without either the prior written permission of the publisher or a licence permitting restricted copying issued in the UK by The Copyright Licensing Agency and in the USA by The Copyright Clearance Center. Any opinions expressed in the chapters are those of the authors. Whilst Emerald makes every effort to ensure the quality and accuracy of its content, Emerald makes no representation implied or otherwise, as to the chapters’ suitability and application and disclaims any warranties, express or implied, to their use. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN: 978-1-78350-951-5 ISSN: 0193-5895 (Series)
ISOQAR certified Management System, awarded to Emerald for adherence to Environmental standard ISO 14001:2004. Certificate Number 1985 ISO 14001
CONTENTS LIST OF CONTRIBUTORS
vii
THE LAW AND ECONOMICS OF CLASS ACTIONS: YESTERDAY, TODAY, AND TOMORROW James Langenfeld and Raleigh Richards
1
FROM HYDROGEN PEROXIDE TO COMCAST: THE NEW RIGOR IN ANTITRUST CLASS ACTIONS James Keyte, Paul Eckles and Karen Lent
11
THE CLASS CERT GAMES: COACH, COMMENTATOR, OR CRITIC? Joshua P. Davis
65
ANTITRUST CLASS PROCEEDINGS THEN AND NOW Michael D. Hausfeld, Gordon C. Rausser, and Gareth J. Macartney, with Michael P. Lehmann and Sathya S. Gosselin ECONOMETRIC TESTS FOR ANALYZING COMMON IMPACT Kevin W. Caves and Hal J. Singer ASSESSING MARKET EFFICIENCY FOR RELIANCE ON THE FRAUD-ON-THE-MARKET DOCTRINE AFTER WAL-MART AND AMGEN Mukesh Bajaj, Sumon C. Mazumdar and Daniel A. McLaughlin
v
77
135
161
vi
CONTENTS
EUROPEAN COLLECTIVE REDRESS: LESSONS LEARNED FROM THE U.S. EXPERIENCE Ethan E. Litwin and Morgan J. Feder
209
CARTEL OVERCHARGES John M. Connor
249
LIST OF CONTRIBUTORS Mukesh Bajaj
Navigant Economics, Oakland, CA, USA
Kevin W. Caves
Economists Incorporated, Washington, DC, USA
John M. Connor
Purdue University, West Lafayette, IN, USA
Joshua P. Davis
University of San Francisco School of Law, San Francisco, CA, USA
Paul Eckles
Skadden, Arps, Slate, Meagher & Flom LLP, New York, NY, USA
Morgan J. Feder
Hughes Hubbard & Reed LLP, New York, NY, USA
Sathya S. Gosselin
Hausfeld LLP, Washington, DC, USA
Michael D. Hausfeld
Hausfeld LLP, Washington, DC, USA
James Keyte
Skadden, Arps, Slate, Meagher & Flom LLP, New York, NY, USA
James Langenfeld
Navigant Economics and School of Law, Loyola University Chicago, Chicago, IL, USA
Michael P. Lehmann
Hausfeld LLP, San Francisco, CA, USA
Karen Lent
Skadden, Arps, Slate, Meagher & Flom LLP, New York, NY, USA
Ethan E. Litwin
Hughes Hubbard & Reed LLP, New York, NY, USA
Gareth J. Macartney
OnPoint Analytics, Inc., Emeryville, CA, USA
vii
viii
LIST OF CONTRIBUTORS
Sumon C. Mazumdar
Navigant Economics, Oakland, CA, USA; Walter A. Haas School of Business, University of California-Berkeley, Berkeley, CA, USA
Daniel A. McLaughlin
Sidley Austin LLP, New York, NY, USA
Gordon C. Rausser
University of California at Berkeley, CA, and OnPoint Analytics, Inc., Emeryville, CA, USA
Raleigh Richards
Navigant Economics, Chicago, IL, USA
Hal J. Singer
Economists Incorporated, Washington, DC, USA
THE LAW AND ECONOMICS OF CLASS ACTIONS: YESTERDAY, TODAY, AND TOMORROW James Langenfeld and Raleigh Richards Legal and economic analyses overlap and interact in many areas. Recent U.S. Supreme Court and lower court decisions on class action lawsuits clearly focus on the critical role that economic analysis plays in determining the outcome class actions. Class action decisions, such as The Wal-Mart Stores, Inc. v. Dukes1 and Comcast Corp. v. Behrend,2 have made national headlines, and are affecting how class action cases are evaluated and managed. For example, Comcast, Dukes, and other recent cases have arguably raised the bar in class certification for showing common impact and predominance through expert testimony. In these and other class actions, one critical question is whether there exists a plausible systematic way to identify the class members who have been economically impacted by a wrongful act and to formulate a reliable generally accepted economic methodology to measure their damages. If there is no reliable systematic approach to identifying which class members have been affected, then class members must pursue the case individually, which often means the case does not go forward because of the cost of the litigation is too high for a single class member. These class action decisions have turned on the adequacy of the analyses put forth by expert economists, finding the basic economic and
The Law and Economics of Class Actions Research in Law and Economics, Volume 26, 19 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0193-5895/doi:10.1108/S0193-589520140000026001
1
2
JAMES LANGENFELD AND RALEIGH RICHARDS
statistical analyses of the plaintiffs’ economists to be insufficient. The decisions are likely to have significant implications for use of expert testifiers in class certification and in estimation of monetary damages, and will present challenges to both attorneys and economists in antitrust and other class actions going forward. On October 25, 2013, Research in Law and Economics and Navigant Consulting sponsored the conference Class Action Landscape: Yesterday, Today, and Tomorrow in Boston. The conference was composed of four sessions. Each session was focused on one or more articles related to class actions that had been refereed and accepted for publication in this issue of Research in Law and Economics. In addition to the authors presenting their research, discussants involved in class actions and related research offered comments and critiques. These articles and critiques by lawyers and economists concern the changing landscape of class action law and its interaction with the economic analysis of key issues in class actions. These articles examine the elements of class action law from diverse viewpoints, featuring articles with defendant and plaintiff perspectives, concerning domestic and international law, and written by lawyers and economists. The balance of this introductory article briefly summarizes the articles and the comments that were offered at the conference. Well-known defense class action attorneys James Keyte, Paul Eckles, and Karen Lent of Skadden, Arps, Meagher and Flom LLP in “From Hydrogen Peroxide to Comcast: the New Rigor in Antitrust Class Actions” collect, assess, and categorize U.S. Rule 23(b)(3) direct purchaser antitrust cases since Hydrogen Peroxide3 and in the wake of the Supreme Court’s decision in Comcast. They find that economic testimony has become a critical part of class certification, and the U.S. Supreme Court has raised the standard for these analyses. The authors explore the meaning of “rigor” as it is used today in antitrust cases, suggest future developments based on current principles, and envision an increasingly important role for Daubert motions that are used to attempt to exclude experts and expert testimony. The new rigor the authors see suggests a more stepwise analytical framework for assessing class certification beginning with Daubert issues, proceeding to questions surrounding issues of common proof of impact, and finally, if necessary, concluding with a rigorous analysis of proffered classwide proof concerning damages. The article identifies the types of challenges to class certification that defendants should consider making and suggests where the future analytical framework for assessing class certification may eventually settle.
The Law and Economics of Class Actions: Yesterday, Today, and Tomorrow
3
Professor Joshua Paul Davis, of the University of San Francisco School of Law, is the author of several class action articles. He submitted a formal comment on the Keyte, Eckles, and Lent article, titled “The Class Cert Games: Coach, Commentator, or Critic?” In his comment, Davis finds that Keyte, Eckles, and Lent have done an excellent job presenting a “playbook” for defense counsel to challenge plaintiffs’ attempts to get class certification and to exclude plaintiffs’ experts. Davis does not see Keyte, Eckles, and Lent as completely objective in their reading of the cases, and he raises questions about some of the policy implications in their article. He points out some specific areas where he disagrees with Keyte, Eckles, and Lent’s interpretation of existing case law, particularly what is necessary to estimate classwide damages. In the “Antitrust Class Proceedings Then and Now,” leading plaintiffs’ attorney Michael Hausfeld of Hausfeld LLP and economists Gordon Rausser and Gareth Macartney of OnPoint Analytics Inc. track the evolution of Rule 23 on class certification since its inception, and analyze the impact of the Hydrogen Peroxide and Wal-Mart decisions on the Rule’s interpretation. The authors then examine the current and future role of economics in postComcast antitrust class certification, specifically in the areas of liability, common impact, and damages. They explain how liability evidence can be used by economists to support a finding of common impact for certification purposes, and how statistical techniques such as averaging, price-dispersion analysis, and multiple regressions have and should be employed to establish common proof of damages. They make several policy recommendations, and conclude that the emerging class-certification standards require greater diligence and impose greater obligations on both the parties and the courts. Robert Kneuper of Navigant Economics, who has worked on many class actions, provided general comments about the economic analyses needed to certify classes after Comcast and related decisions. First, he argued that showing common impact and damages need to be based on sound economic analyses that apply reliable economic and statistical techniques. Second, any economic analysis of common impact and damages must be done in the context of specific liability and causation theories of the case and the appropriate scope of the proposed class. Given these constraints and what the courts have found unacceptable, he believes it will often be better to narrowly focus the liability theory of the case so the classcertification analysis does not have to stretch to take into account too many different effects for different class members. Kneuper concluded that
4
JAMES LANGENFELD AND RALEIGH RICHARDS
defining a smaller class based on a clear liability theory is better than a larger class with a less precise liability theory with more chances for members who may not be impacted. Statistical and econometric analyses are often used in class certification to estimate the impact of an alleged wrongful act independent of other potential market influences, and now may be more important if there is increased rigor required. In “Econometric Tests for Analyzing Common Impact,” Kevin Caves and Hal Singer propose econometric methods and statistical tests for detecting the existence of common impact across class members and determining what proportion of a proposed class suffered injury. This work addresses the tension between today’s standard econometric methods that measure the average effects of challenged conduct on prices, and the legal standard of common impact that is concerned with determining whether individual class members were harmed. They propose a specific econometric approach for class-certification analyses that allow for the systematic estimation of differing effect of an action on the prices paid by individual class members and controls for the average impact of other market forces on prices on class members. This approach depends on the availability of sufficiently detailed customer by customer data before and during (or during and after) an allegedly wrongful act, such as pricefixing. Singer had used this approach in a litigated case under appeal at the time, and had found that 90 percent of customers were significantly affected based on standard statistical tests. Laila Haider of Edgeworth Economics is the coauthor of an article commenting on the Caves and Singer approach, where she evaluates the feasibility of systematically estimating the impact and effect of an allegedly wrongful act. Haider provided comments on the Caves and Singer article, and argued their approach inappropriately assumes influences other than the allegedly wrongful act should be estimated as the average across class members. Haider’s proposed econometric testing differs from Caves and Singer’s in that it not only allows the impact of the allegedly wrongful act to vary across class members, but also allows for the possibility that other influences on prices vary from customer to customer. The added complexity of her model can place additional demands on the amount of data needed to perform reliable tests, and this approach can find individual inquiry is needed when the Caves and Singer approach does not. Jeff Leon of Complex Litigation Group LLP has worked with both plaintiffs and defendants in his career. He commented on the work of both Caves and Singer and Haider, attempting to put the disagreements as to which economist model is more appropriate in the context of litigation and
The Law and Economics of Class Actions: Yesterday, Today, and Tomorrow
5
what is most effective. Given the complexity of the econometrics methods, Leon explained the need for economists to be simple and clear, or run the risk of being ignored. He raised some questions about whether Haider’s approach would almost invariably find a need to engage in individual inquiry, and Haider admitted that to date they had only run their model on complex markets where they did not find a certifiable class. “Assessing Market Efficiency for Reliance on the Fraud-On-The-Market Doctrine after Wal-Mart and Amgen” is authored by Mukesh Bajaj and Sumon Mazumdar of Navigant Economics and the Haas School of Business at the University of California-Berkley, and Daniel McLaughlin of Sidley Austin LLP. Although not presented at the conference, the article concerns class actions, was accepted for publication by the journal, and so has been included in this issue. In the article, the authors address specific aspects of the certification of securities class actions that have been discussed in Supreme Court cases as recently as 2013 in Amgen.4 In particular, the authors explain the economic analyses they believe are required for class plaintiffs to invoke the “rebuttable presumption of reliance on public, material misrepresentations regarding securities traded in an efficient market” (the “fraud-on-the market” doctrine) to prove classwide reliance. They find that lower courts have frequently granted class certification based on a mechanical review of some factors that are considered intuitive “proxies” of market efficiency, but do not actually show market efficiency according to recent studies and analyses that the authors present in the article. The authors argue that plaintiffs must first establish the security traded in an efficient market using well-accepted economic tests before invoking fraudon-the market. Only then do event study results, which are commonly used to demonstrate “cause and effect” across a class, have any merit. To show full classwide reliance, plaintiffs must additionally prove such cause and effect relationship throughout the class period, not simply on selected disclosure dates identified in the complaint. The authors point out that these issues have major policy implications because defendants frequently settle once a class is certified to avoid the magnified costs and risks associated with a trial. Ethan Litwin and Morgan Feder of Hughes Hubbard and Reed LLP have a great deal of experience with class action in the U.S. and the European Union (E.U.). In “European Collective Redress: Lessons Learned from the US Experience,” they raise a number of questions about the effectiveness of class action lawsuits in the U.S. and propose alternative approaches to reimburse class members using other approaches in the U.S. and the European Commission. Litwin and Feder describe the steps
6
JAMES LANGENFELD AND RALEIGH RICHARDS
already taken by the European Commission to research and develop European standards of compensatory collective redress, and then also propose, as well as criticize, alternatives to class actions based on both U.S. and E.U. experiences. They argue that all class action models essentially force the majority of class members to sacrifice their individual due process rights for the chance to aggregate claims. The opt-out mechanism in the United States presumes that all members of a proposed class are being represented by plaintiff lawyers unless they explicitly “opt-out” to pursue the case on their own. Litwin and Feder argue that this system where litigants cannot meaningfully participate in the adjudication of their rights (or even to decide whether a case should be brought at all) promotes just the kind of wasteful, unmeritorious over-litigation that E.U. lawmakers expressly wish to avoid. However, the authors also see the E.U. proposal as not being workable because of the limitations on third-party funding, preventing contingency fees, the opt-in rule, and the “loser pays” principle. Litwin and Feder instead describe how alternatives to class action litigations such as SEC Fair Funds and the September 11th Victim Compensation Fund could eliminate many of the disadvantages of the U.S. system and the problems with the E.U. proposals for collective redress. Barbara Hart, of Lowey Dannenberg’s securities litigation practice and Vice-Chair of the Executive Committee of the New York State Antitrust Committee, raised questions about how well the Litwin and Feder proposals could be implemented, and whether the U.S. mass tort system experience is as flawed as many Europeans and Litwin and Feder imply. Hart sees this article as a conscientious effort to propose a middle ground in light of the foreign aversion to adopting the contingent class action model. However, she believes that aversion to the U.S. system is born of false belief that the state of class actions has not evolved beyond its inception and the incidental abuses for which it is pilloried. Other countries should not stay focused on past shortcomings and dated caricatures of what the Rule 23 practice was. The old practice of simple oral argument and bare bones economic analysis for class certification are being supplanted by fullor multiday evidentiary hearings and detailed economic analyses based on the facts of the case. Industry and economic experts are put on the stand, fully examined and crossed. The standards for class certification have been ratcheted up as the defense bar often deluges the court with contentionsspecific transactions involve bargaining and variations in terms that make a proposed class unmanageable. Litigating a class case is complex from start to finish for this and other reasons, but Hart sees class actions as the best and ever improving model to address large-scale legal matters. Class
The Law and Economics of Class Actions: Yesterday, Today, and Tomorrow
7
certification depends on Rule 23, which is a case management invention, and she believes the problem with some class actions is not with the rule but with the execution. Courts and practitioners can and do engage in large-scale problem solving, taking care to derive the best possible solution. As such, she sees the state of class action practice in the U.S. as worthy, sophisticated, and remedial. Hart believes the U.S. class action provides the best model because it is constantly criticized, and thereby sharpened and improved. She argues that opt-in and government-sponsored victim redress funds cannot, have not, and will not succeed to the same measure as U.S. class actions. As other jurisdictions embark on the alternatives, she believes we will all learn from their experiments in mass problem solving and should increasingly appreciate the U.S. system. Heidi Dalenberg, partner with Schiff Hardin, LLP, has extensive experience representing clients in asserted class actions involving breach of contract, consumer fraud, and common law fraud. Unlike Hart, Dalenberg sees many of the same problems with the mass tort redress system in the U.S. that Litwin and Feder discuss. All agreed that the E.U. proposals with an opt-out system, etc., would not achieve the goal of providing redress to class members. Dalenberg also agreed with Hart that the Litwin and Feder’s proposals were not likely to be able to be workable in most instances. Finally, Professor John Connor of Purdue University contributed “Cartel Overcharges,” which is a comprehensive study of overcharges in antitrust cartel cases that surveys economic studies and judicial decisions that contain more than 2,000 estimates of overcharges from 532 cartels from the 19th century through present day. These sources met minimal quality standards, and Connor examined the estimates for systematic differences in reliability across sources and methods of calculating overcharges. In general, Connor did not apply any sophisticated quality filters for excluding possibly questionable sources for his overcharge estimates. Many of these matters involve class actions, and the author finds the median overcharge due to the cartel activity in these cases to be approximately 23 percent. The analysis indicates that average overcharges are much higher than the average level presumed by antitrust authorities. This sizable overcharge suggests that fines and damages may not adequately compensate class members or deter illegal behaviors such as pricefixing and bid-rigging. Among other interesting aspects of the data, Connor finds that international cartels on average impose higher overcharges, and that overcharges on average have fallen slowly since the 19th century.
8
JAMES LANGENFELD AND RALEIGH RICHARDS
Terry Calvani and John Kwoka discussed Connor’sarticle, and offered their views on the reliability and usefulness of Connors work. Professor John Kwoka of Northeastern University has a distinguished career as a scholar, former. Federal Trade Commission economist, and a widely published author on antitrust issues, including price-fixing class actions. Kwoka pointed out that before Connor’s work the Chicago School of Economics concluded, without an empirical basis, that cartels are hard to form and impossible to sustain. Kwoka noted that the U.S. Sentencing Commission in the 1980s concluded that average gain from a cartel was a 10-percent overcharge. However, Connor’s work shows that cartels have existed with many members, have been long-lasting, and have had a price effect well above 10 percent in many instances. Kwoka raised some questions about (1) the reliability of a number of the studies that Connor includes in his data set, and (2) how Connor counts different episodes of price-fixing in the same industry during different periods and by different authors. The latter could result in biasing upward some of Connor’s results if the multiple episode and multiple studies involve industries where the cartels were particularly effective. Nevertheless, Kwoka doubted that addressing these potential shortcomings would substantially affect one of the main implications of Connor’s work: fines and damage awards under deter cartel behavior, so “crime pays,” and more aggressive enforcement is merited. Terry Calvani practices law in antitrust and class actions at Freshfields, and has served as Acting Chairman of the Federal Trade Commission, and as a Member of the Irish Competition Authority and Director of the Criminal Cartels Division. Calvani did not comment on the quality of the studies that form the basis of Connor’s work and did not agree with all of Connor’s conclusions. However, Calvani was familiar with Connor’s work, and found it to be extremely useful when at the Irish Competition Authority because it provides an important empirical basis for cartel and class actions that are absent from other economic research. However, Calvani pointed out that fining cartel firms does not always benefit class members, especially in instances where the money may largely go to attorneys or antitrust think tanks through Cy Pres funds. Calvani also disagreed that there is insufficient deterrence of cartel activity if corporate fines are lower than the estimated overcharges. Calvani explained that individuals, and not corporations, form cartels. Companies are often fined for actions by individuals who no longer work for the company when the cartel actions are discovered. Instead, Calvani reasoned that sanctions should primarily be placed on the individuals who engaged in the illegal acts, and not the companies.
The Law and Economics of Class Actions: Yesterday, Today, and Tomorrow
9
In general, there does not appear to be a consensus on how well the current U.S. approach to class actions is working or on how to address any shortcomings. However, the articles and the comments agree that recent Supreme Court decisions appear to require more rigorous economic analysis to certify a class, and that this trend started several years ago with lower court decisions. As the lower courts now begin to apply these more rigorous standards, one can anticipate the development of more sophisticated economic approaches and more extensive discovery for class certification. The full impact of these developments on the number of successful class actions will likely depend on how effective the economic approaches prove to be, and on whether alternative approaches to consumer redress can be successfully implemented. These changes may have a substantial impact on the parties involved, and on the overall economy.
NOTES 1. 2. 3. 4.
Wal-Mart Stores, Inc. v. Dukes, 113 S. Ct. 2541 (2011). Comcast Corp. v. Behrend, 133 S. Ct. 1426 (2013). In re Hydrogen Peroxide Antitrust Litig., 552 F.3d 305 (3d Cir. 2008). Amgen v. Connecticut Ret. Plan & Trust Funds, 133 S. Ct. 1184 (2013).
FROM HYDROGEN PEROXIDE TO COMCAST: THE NEW RIGOR IN ANTITRUST CLASS ACTIONS$ James Keyte, Paul Eckles and Karen Lent ABSTRACT In 2009, the Third Circuit decided Hydrogen Peroxide, which announced a more rigorous standard under Federal Rule of Civil Procedure 23(b)(3) for assessing whether a putative class could establish antitrust injury. Earlier this year, the Supreme Court decided Comcast v. Behrend, a case that carries potentially broad implications for both antitrust cases and Rule 23(b)(3) class actions generally. A review of the case law starting with Hydrogen Peroxide and continuing through Comcast and its progeny reveals the new rigor in antitrust class action decisions and suggests what the future may hold, including the type of arguments that may provide defendants the most likely chance of defeating class certification. After Comcast, rigor under 23(b)(3) can no longer be avoided in assessing all class actions questions, and courts should now apply Daubert fully in the class setting concerning both impact and damages.
$
James Keyte, Paul Eckles and Karen Lent are partners in the antitrust group of Skadden, Arps, Slate, Meagher & Flom LLP.
The Law and Economics of Class Actions Research in Law and Economics, Volume 26, 1163 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0193-5895/doi:10.1108/S0193-589520140000026002
11
12
JAMES KEYTE ET AL.
Courts should also closely evaluate plaintiffs’ proposed methodologies for proving impact to determine if they apply to each class member. Finally, courts will inevitably have to determine how rigorously to scrutinize experts’ damages methodologies and whether Comcast requires or suggests more scrutiny in assessing common evidence for measuring damages. Keywords: Class certification; Daubert; common impact; Comcast JEL classifications: K41; K21
TABLE OF CONTENTS Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rigor Under 23(b)(3) Can No Longer be Avoided. . . . . . . . . . . . . . . . Step 1: Apply Daubert Fully in the Class Setting on the Questions of Both Impact and Damages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Where the Class Action Law on Daubert Has Been and Is Headed . Types of Rule 23(b)(3) Expert Analysis That Should Not Make It Past a Daubert Screen. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Where the Theory of Impact Does Not Match the Theory of Antitrust Injury . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Where the Methodology Cannot Pass the Most Basic Daubert Academic Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Where the Methodology “Presumes” or “Assumes” Impact or a Basis for Calculating Damages . . . . . . . . . . . . . . . . . . . . . . . . Other Analyses that “Mask” Potential Individual Questions of Proof . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . “Pricing Structure” or “Correlation Analysis” . . . . . . . . . . . . Methodologies Based on “Average” Impact . . . . . . . . . . . . . Methodologies that Ignore Critical Variables and Issues with the But-For World . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Step 2: Evaluate if the Proposed Methodology for Proving Impact Could be Used by Every Class Member if Applied in Individual Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Courts Must Perform a “Rigorous Analysis” of an Expert’s Opinion and Resolve the “Battle of the Experts,” Even if an Expert’s Report Survives Daubert . . . . . . . . . . . . . . . . . . . . . . . .
13 15 16 16 18 18 20 21 21 22 23 25
27
27
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 13
Defendants Should Attempt to Develop a Factual Record Demonstrating that Plaintiffs’ Expert’s Assumptions about the Actual World Are Incorrect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Price Lists and Price Increase Announcements. . . . . . . . . . . . . . Market Characteristics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Defendants Challenging an Expert’s Opinion Under Rule 23 Should Also Attack Any Unreasonable Assumptions Made about the “But-For” World . . . . . . . . . . . . . . . . . . . . . . . . . . . . . If Common Questions Predominate for Only Some of the Plaintiffs, the Court Can Sometimes Narrow the Scope of the Proposed Class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Step 3: Can Damages be Proven on a Class Wide Basis? . . . . . . . . . . . Preliminaries A Whole New World after Comcast? . . . . . . . . . . . Comcast Majority: The Need for Rigor Regarding Damages . . . Comcast Dissent: Attempting to Fence in the Majority . . . . . . . How Comcast Is Being Applied to Class Damages Analyses . . . . . . Greater Rigor Required . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cases Vacated and Remanded (GVR) by the Supreme Court Cases Applying Comcast’s Rigor Requirement to Proof of Damages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cases Purporting to Follow the Comcast Dissent. . . . . . . . . . Ways to Attack Damages Evidence under the New Rigor . . . . . . . . Does the Plaintiff Provide a Formula to Calculate Damages? . . Are Damages Susceptible to “Mathematical” or “Formulaic” Calculation? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Does the Product at Issue Involve Individualized Negotiations? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Do Individual Plaintiffs Differ in Ways that Make Proof of Damages Highly Individualized? . . . . . . . . . . . . . . . . . . . . Do Plaintiffs Propose a Damages Formula Clearly Inadequate to Calculate Individual Damages? . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
28 29 31
32
34 34 35 35 36 37 37 37 41 43 45 45 46 46 48 49 50 62 62
INTRODUCTION In the wake of the Third Circuit’s 2009 decision in Hydrogen Peroxide,1 numerous circuit and lower courts adopted a more rigorous standard under
14
JAMES KEYTE ET AL.
Federal Rule of Civil Procedure 23(b)(3) for assessing whether a putative class could establish antitrust injury or “impact” through common proof. And, no doubt, that newly found rigor resulted in a number of decisions rejecting class certification in cases where, on the proffered expert evidence, individual issues of proof of impact would likely predominate.2 Certainly, however, the cases denying class certification on this basis vary quite a bit in identifying the particular failure of evidence under Rule 23(b)(3), ranging from cases that exclude expert testimony at the class stage under Daubert3 to those that, in effect, make findings (for class purposes only) that plaintiffs’ proposed common proof of impact did not overcome the individualized issues identified by defendants’ expert. By contrast, on the subject of common proof of the amount of damages, relatively fewer courts both pre- and post-Hydrogen Peroxide rejected class certification on this basis alone, although some did where the proffered methodology clearly could not identify and measure a class member’s damage.4 The pervasive theme of many 23(b)(3) damages opinions is that, in contrast to the essential element of antitrust impact, plaintiffs and their experts had more flexibility in showing how damages would be assessed through common proof, just as in an individual case where proof of the “amount” of damages historically is given more leeway.5 Then, this past year, the Supreme Court decided Comcast Corp. v. Behrend, a Rule 23(b)(3) damages case.6 Read narrowly, Comcast merely stands for the proposition that any proposed common proof of damages under Rule 23(b)(3) must flow solely from the basis of antitrust liability accepted by the court considering class certification. But the Court’s majority opinion has much broader implications both in antitrust cases and in Rule 23(b)(3) actions generally. The opinion itself highlights that the same “rigor” the Supreme Court had earlier found applied to Rule 23 (a) also applied to Rule 23(b)(3), including presumably for assessing theories and evidence of damages. And while the dissent in Comcast did what it could to fence in the implications of the majority opinion attempting to limit it to its holding on the causal link between the basis for liability and damages the Supreme Court itself has remanded cases addressing broader impact and damages issues in light of Comcast.7 Similarly, several cases and commentators have read Comcast to require the same rigor now being applied on impact to be applied in the damages setting, whether on the subject of disaggregation, damages methodologies or simply whether plaintiffs met their burden of proving that common issues of proof concerning damages would predominate over any individual issues of proof.8
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 15
In contrast to most disclaimers for articles such as these, the objective here is to be reasonably comprehensive in collecting, assessing and categorizing successful challenges to putative Rule 23(b)(3) direct purchaser antitrust classes since Hydrogen Peroxide and in the wake of Comcast. Moreover, the goal is not only to understand what “rigor” has come to mean in antitrust cases both for experts and plaintiffs generally but also to suggest what the future may hold given the principles and trajectory we see. For example, Comcast and the cases in its wake (including those caught midstream) demonstrate that as a practical matter, Daubert could play an increasingly important role at the class certification stage e.g., practitioners are going to have to make more Daubert (or later, in limine) motions to avoid waiver issues. In turn, and given the rigor now clearly mandated by the Supreme Court under Rule 23(b)(3), a body of Daubert class action law will continue to expand as it has already since Hydrogen Peroxide. And, of course, courts will continue to hone analytical frameworks and principles for assessing Rule 23(b)(3) issues including those related to damages to the benefit of both experts and practitioners who must continually adapt to the new rigor. In this evolving legal environment, the body of Rule 23(b)(3) antitrust decisions themselves suggests some clear demarcations of where decisions have come out in the past and where they are likely to come out in the future. The rigor we see from Hydrogen Peroxide through Comcast and beyond suggests a more stepwise analytical framework for assessing class certification under Rule 23(b)(3) beginning with Daubert issues, proceeding to questions surrounding issues of common proof of impact, and finally, if necessary, concluding with a rigorous analysis of proffered classwide proof concerning damages. In the pages that follow, we seek both to identify the types of challenges to class certification that defendants should consider making going forward and to suggest where the future analytical framework for assessing class certification may eventually settle.
RIGOR UNDER 23(B)(3) CAN NO LONGER BE AVOIDED The underlying theme of Comcast is that the same (if not more) rigor applied to Rule 23(a) must now apply to Rule 23(b)(3) analysis. Hence, the Court confirmed that at the class certification stage, lower courts must “probe behind the pleadings” even if arguments “would be pertinent to the merits determination,” including with respect to expert testimony.9 Going
16
JAMES KEYTE ET AL.
forward, this definitively precludes any courts from essentially “kicking the can down the road” to avoid addressing “merits” issues at the class stage. Moreover, it in effect confirms that some of the “shortcuts” or presumptions utilized in past Rule 23(b)(3) cases e.g., based on Bogosian10 will no longer have traction with courts to the extent they were not already limited by Hydrogen Peroxide and other cases.11 These or similar truncated analyses cannot meet the Court’s directive that lower courts have a “duty to take a ‘close look’ at whether common questions predominate over individual ones.”12 Certainly, this is now the state of the law in assessing “predominance” with respect to the subject of “impact.” But, as we describe below, there remains some ambiguity on the scope of rigor that now is to be applied to proof of damages, which courts already are addressing in the post-Comcast environment. While the authority is split, we believe the better view is that the rigor required under Rule 23(b)(3) on the essential element of impact does not simply dissolve when assessing plaintiffs’ proffered expert evidence for calculating damages. Presumably a settled view will emerge, but one cannot rule out another Rule 23(b)(3) case making its way to the Court and perhaps another opportunity for a divided court to limit the scope of class actions damages litigation.
STEP 1: APPLY DAUBERT FULLY IN THE CLASS SETTING ON THE QUESTIONS OF BOTH IMPACT AND DAMAGES Where the Class Action Law on Daubert Has Been and Is Headed In the wake of Comcast, the subject of Daubert’s application at the class certification stage will be a primary focus, for two different reasons. First, as a purely practical matter, the waiver risks of not making a Daubert motion at the class stage are now sufficiently high that we are likely to see a steady flow of such motions in all but the most mundane price-fixing cases. Second, now that both Comcast (implicitly) and Dukes (more directly) have endorsed the use of Daubert scrutiny at the class certification stage,13 it is reasonable to conclude that those decisions requiring or endorsing a “full” Daubert screen prior to ruling on class certification are likely to emerge as the prevailing standard. When testimony challenged under Daubert relates to Rule 23(b)(3) requirements, to defer a Daubert screen
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 17
until after class certification would effectively remove, or at least severely diminish, the impact of the rigor requirement. This would ensure exactly the kind of provisional, wait-and-see approach that the rigorous approach to class certification rejects. Indeed, it makes little sense to claim rigor is required at the class certification stage but then push off tough (or even easy) Daubert calls for some later procedure in the case. While Dukes suggests that Daubert scrutiny applies at the class certification stage, other courts have held so more explicitly. This was precisely the reasoning employed by the Seventh Circuit in American Honda Motor Co. v. Allen, a case decided before Dukes and Comcast.14 There, purchasers of Honda motorcycles brought suit, alleging a design defect that prevented the front steering assembly from wobbling while riding.15 In support of its class certification motion, plaintiff’s expert presented a report that relied on a standard created by the expert himself.16 Although the District Court acknowledged it had reservations about the reliability of the expert’s standard, it refused to exclude the report and instead certified two classes of motorcycle purchasers.17 The Seventh Circuit reversed, holding that the district court should have performed a full Daubert analysis before certifying the two classes.18 The Seventh Circuit reasoned that if an expert’s information is relevant to proving any of the Rule 23 requirements for class certification, then the district court must resolve any challenge to the reliability of that information.19 A year later, but also before Dukes and Comcast, the Eleventh Circuit followed suit in Sher v. Raytheon Co.20 Relying on American Honda, the court overturned a district court’s certification of a class, holding that the district court’s failure to apply Daubert during its Rule 23 analysis was clear error.21 And, of course, the Supreme Court originally granted certiorari in Comcast for purposes of addressing the role of Daubert at the class certification stage before concluding that defendants had waived their rights by failing to file a motion objecting to the plaintiffs’ expert report.22 Accordingly, we are likely to see the maturing of a clear body of class certification law that is centered around traditional Daubert factors, but plays out in the landscape and language of class certification. Moreover, even past decisions that chose not to apply Daubert fully (or at all), yet clearly highlighted standard Daubert-like flaws in the expert’s qualifications or analyses, are a good resource for assessing the type of class-related Daubert exclusions we are likely to see going forward. These and other candidates are described below.
18
JAMES KEYTE ET AL.
Types of Rule 23(b)(3) Expert Analysis That Should Not Make It Past a Daubert Screen With rare exception, class action experts in the antitrust field are “qualified” in terms of their academic backgrounds and professional experience. In fact, there is a fairly robust cadre of antitrust economists that tend to work on the plaintiffs’ side of the bar and, in turn, have their work scrutinized in a number of court opinions. Instead, the problem for many of these economists is that their professional experience in the class action setting often is based largely on the pre-rigor legal framework. In the earlier framework, the validity of allegations was assumed, and courts did not address issues that went to the “merits” until after class certification. Under this framework, at times, the mere assurance that a methodology based on common proof existed or could be developed at some later point in the litigation was enough to support class certification.23 Those days are now officially behind us. Indeed, class experts must now be prepared to defend their work from any number of Daubert perspectives, all of which should be pressed by defendants.24 Where the Theory of Impact Does Not Match the Theory of Antitrust Injury One predictable Daubert motion in light of Comcast is one arguing that the expert has proffered a theory or analysis of impact or damages that does not clearly limit the harm to the exact basis of antitrust liability. In Comcast, cable television customers brought an antitrust action against Comcast and its subsidiaries, alleging that Comcast’s practice of “clustering” resulted in Comcast monopolizing the Philadelphia market. The District Court and Court of Appeals both approved certification of the class under Rule 23(b)(3). The Supreme Court reversed, holding that class certification was improper because the damages model developed by plaintiffs’ expert did not prove that damages could be measured on a classwide basis. Consequently, because individual questions would overwhelm the Court’s damages analysis, it held that plaintiffs did not satisfy Rule 23(b)(3)’s predominance requirement. The case arose out of a Comcast practice called “clustering,” a business strategy intended to concentrate operations within a particular region. In order to implement its strategy, Comcast purchased competing cable providers in the Philadelphia cluster25 and then replaced those providers with Comcast services. As a result of the practice, Comcast’s share of subscribers in the region allegedly jumped from 23.9% in 1998 to 69.5% in 2007.
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 19
Plaintiffs presented four theories of antitrust injury that allegedly affected the entire class.26 The district court rejected all but the “overbuilder deterrence theory,” which posited that Comcast’s strategy reduced competition by deterring “overbuilders” companies that chose to build competing networks in areas where an incumbent cable company already operates. But plaintiffs’ damages model was prepared before the court’s ruling regarding which liability theories were viable and hence relied on all four theories.27 The Supreme Court thus had to consider whether, when calculating damages, the plaintiff could use a model that measured harm based on the original four theories of injury or instead had to use a model based only on the remaining theory of overbuilder deterrence. The Court held that plaintiffs’ model could not meet the predominance requirement of Rule 23(b)(3). Plaintiffs were required to put forth a damages model that was specific to the overbuilder theory in order to sustain the predominance standard. Instead, plaintiffs’ model calculated damages based on the harm caused by “the alleged anticompetitive conduct as a whole” rather than the harm caused solely by overbuilder deterrence. This model could not differentiate between higher prices in general and higher prices attributable to overbuilder deterrence. Based on this flawed methodology, customers who were harmed by anything other than overbuilder deterrence still would be awarded damages. As a result, the trial court would have to investigate whether overbuilder deterrence or entirely different effects were the true cause of individual class members’ harm. Recognizing this, the Court found that individual questions necessarily would predominate over common questions during the damages inquiry, requiring reversal. As anticipated, since Comcast, a number of courts have applied this fundamental principle in either rejecting class certification or having to reassess class certification in light of Comcast.28 The practical implications of Comcast just on the subject of disaggregation are far reaching and potentially very complicated in cases where there is a variety of alleged misconduct. Experts are likely to find themselves offering several (or a myriad of) alternative impact and damages scenarios depending on the nature of the allegations. Moreover, the fact that, under Rule 23, class certification is supposed to be addressed early in a case29 puts plaintiffs (and courts having to manage the case) in somewhat of an awkward position: class experts have to account for alternative or completely different bases for antitrust liability often before discovery is over and plaintiffs themselves have settled on their preferred theory of liability. And, on top of this, all of these positions must be taken under a much more rigorous certification standard in light of Comcast.
20
JAMES KEYTE ET AL.
Finally, there also are disaggregation and causation issues that likely will come into play that go beyond the alternative bases for liability. Courts have long struggled with how to address the subjects of causation, impact and damages where plaintiffs’ harm flows in part from lawful or procompetitive conduct.30 Accordingly, not only will class experts have to deal with alternative analysis of alleged misconduct, but courts (and defendants) too are likely to consider whether economic experts can offer common proof of impact and damages that properly disaggregate other causes of plaintiffs’ harm, irrespective of whether it is other alleged misconduct. Moreover, the dissent in Comcast was quite upset that the majority opinion strongly suggests that plaintiffs and their experts must prove precisely “how” the alleged antitrust misconduct actually resulted in higher prices. The dissent argued that it should be sufficient simply to show “that” prices were in fact higher in the geographic market where Comcast had accumulated a high share.31 While these issues are beyond the scope of this article, they surely will be explored further in the lower courts.
Where the Methodology Cannot Pass the Most Basic Daubert Academic Requirements Perhaps the easiest application of Daubert in the class setting is where plaintiff’s expert essentially makes up his or her own analysis for the case i.e., where the expert is not relying on an academically established methodology. It is not uncommon for experts to rely on theories uniquely developed within a case and not based on an established methodology, and these situations often warrant a Daubert challenge. For example, in American Honda, plaintiffs’ expert created a report attempting to analyze the oscillations of Honda motorcycles’ steering columns.32 The report’s analysis rested on the application of a “wobble decay” standard, which was created by plaintiffs’ expert himself.33 The Seventh Circuit excluded the report under Daubert for three reasons. First, there was no indication that the standard used in the report had been generally accepted by anyone other than plaintiffs’ expert.34 Second, plaintiffs’ expert had no baseline to which he could compare his findings.35 Third, the expert’s methodology was flawed because it relied on an impermissibly small sample size.36 All of these reasons, the court found, indicated that the report did not pass the most basic Daubert testing requirements.37
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 21
Where the Methodology “Presumes” or “Assumes” Impact or a Basis for Calculating Damages Another common flaw though experts now are acutely aware of the pitfall is to offer a methodology that simply ignores the key question: is there common proof that could be used at trial to establish that the alleged misconduct in fact caused injury to each class member. Proposed expert testimony that avoids this central issue will not survive. For example, in Blades v. Monsanto Co.,38 purchasers of Monsanto’s genetically modified seeds alleged that Monsanto conspired with competitors to inflate the price of their seeds. Plaintiffs moved to certify two classes under Rule 23, but the district court rejected certification of both classes. The Eighth Circuit affirmed the district court’s decision and held that any attempt to analyze the harm caused by the alleged conspiracy would require consideration of individual rather than common questions.39 After describing the nationwide prices of genetically modified seeds, the court concluded that plaintiffs’ expert mistakenly assumed a classwide injury. Rather than rely on this assumption at the class certification stage, the court required plaintiffs’ expert to establish defendants’ alleged antitrust violations on a classwide basis through common proof. However, because the supply-and-demand conditions for the seeds varied so greatly, seed prices fluctuated within individual geographic markets across the country. Moreover, sellers throughout the country utilized heavy and variable discount approaches, which resulted in sellers charging a variety of prices for the seeds. In light of such differences in seed pricing, the court found that plaintiffs could not, through common proof, prove that each class member suffered injury by paying an inflated price for the seeds. In other words, when assessing whether each class member was harmed, the court would have to answer too many individual questions about which farmers were injured due to inflated prices and how significantly they were injured. As such, both groups of plaintiff seed purchasers failed to satisfy Rule 23’s predominance requirement and could not be certified as a class.40
Other Analyses that “Mask” Potential Individual Questions of Proof Separate and apart from these rather obvious defects, a rigorous approach to Rule 23(b)(3) will increasingly compel courts to reject expert methodologies and analyses that tend to avoid or mask potential individualized issues of impact or damages. Below are a few of the primary examples.
22
JAMES KEYTE ET AL.
“Pricing Structure” or “Correlation Analysis.” In vogue for some time was the so-called “pricing structure” analysis i.e., the notion that if prices of differentiated products move together in response to marketplace factors, then a price-fixing agreement also would affect all class members in a similar fashion.41 Such a structural approach, however, is ripe for masking individual issues of proof.42 For example, in In re Plastics Additives Antitrust Litigation,43 all direct purchasers of organotin heat stabilizers (“tins”) and epoxidized soybean oil (“ESBO”) alleged that defendant producers conspired to fix prices. Plaintiffs moved to certify two classes: one comprised of tin purchasers and the other comprised of ESBO purchasers. The plaintiffs’ expert opined that pricing for the relevant products displayed a “structure” in which prices “moved similarly” over time, such that if a conspiracy existed, it would have impacted every purchaser.44 After plotting transaction data on a graph, the expert relied on “visual observation” as the sole proof that prices were moving similarly.45 The court rejected the pricing structure methodology for three reasons. First, as a preliminary matter, some of the graphs presented did not, on their face, show that prices moved similarly throughout the class period.46 Second, the graphs relied on did not include transactional data for all customers and all products. Instead, the graphs considered a very limited sample size: 18 of the 256 tin products at issue, less than two dozen of the 508 tin class members, and less than two dozen of the 503 ESBO class members.47 Third, evidence at trial showed that, at certain times during the class period, prices moved in opposite directions.48 Thus, the court found that the pricing structure model was inadequate to measure antitrust injury on a classwide basis. Similarly, in In re Florida Cement & Concrete Antitrust Litigation, the court rejected the plaintiffs’ expert’s correlation analysis as an insufficient basis for showing common proof of impact.49 The court focused on the lack of connection between general pricing structure in the market and the impact of a price-fixing conspiracy, noting that “[p]laintiffs do not explain how the mere fact that average prices move similarly over time in response to general changes in market conditions necessarily implies those prices will also respond similarly to a price-fixing scheme such as the one alleged here.”50 As the court noted, “[t]he implementation of a price-fixing scheme may involve different considerations besides changes in market conditions.”51 Economists similarly have rejected a correlation analysis as an adequate way to determine whether antitrust impact exists.52 As two commentators note, “[t]he conclusion that the ‘prices move together’ is entirely in the eye
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 23
of the beholder, and this subjectivity is exactly what makes this ‘analysis’ non-scientific.”53 Methodologies Based on “Average” Impact. While regression analysis can be a valuable tool in determining whether common proof of impact exists, a common fatal flaw in plaintiffs’ experts’ regression models is a reliance on averages or average pricing. As the Northern District of Illinois recently explained, quoting an ABA publication: Sometimes the prices used by economists are averages of a number of different prices charged to different customers or for somewhat different products. Using such averages can lead to serious analytical problems. For example, averages can hide substantial variation across individual cases, which may be key to determining whether there is a common impact. In addition, average prices may combine the prices of different package sizes of the same product or of somewhat different products. When this happens, the average price paid by a customer can change when the mix of products that the customer buys changes even if the price of [no] single product changed.54
Hence, the problem is that an “average” impact, by definition, fails to demonstrate that all class members have been impacted whether in a Section 1 or a Section 2 Sherman Act case. In Bell Atlantic Corp. v. AT&T Corp., for example, plaintiffs alleged that AT&T attempted to monopolize the market for caller-ID services by blocking the free transmission of caller-ID signals over its long-distance network.55 Plaintiffs moved to certify two classes, one comprised of businesses and organizations that purchased AT&T’s long-distance service and a second comprised of businesses and organizations that were actual or potential purchasers of caller-ID services for long-distance calls.56 Because of AT&T’s actions, plaintiffs alleged they were unable to enjoy the substantial efficiency gains and cost savings that come with caller ID.57 Plaintiffs’ proposed damages formula utilized two national averages: the average cost of labor and the average amount of time that class members would have saved per call had caller ID been available.58 Using these averages as a baseline, plaintiffs argued, the court could assess the class members’ harm by taking the difference between plaintiffs’ averages and the national averages.59 The Fifth Circuit denied class certification, finding that plaintiffs did not satisfy Rule 23(b)(3)’s predominance requirement because their method to prove harm could not reasonably estimate the harm suffered by every class member.60 The court found that in order to adequately gauge harm, it had to consider important individualized questions like the varied nature of the businesses that made up the classes and, depending on those businesses, the
24
JAMES KEYTE ET AL.
range of uses for which caller ID could be employed.61 Because plaintiffs’ averaging formula failed to account for such critical and individualized factors, plaintiffs’ measure of harm and eventual damages projections could not be applied universally to all the businesses in the proposed classes.62 Consequently, the court held that individual questions rather than common questions dominated both the antitrust injury and damages inquiry.63 Similarly, in Reed v. Advocate Health Care, plaintiffs alleged that entities controlling several hospitals in the Chicago area conspired to suppress the wages of their registered nurse (“RN”) employees.64 To prove the conspiracy impacted all class members, plaintiff’s expert proffered econometric models referred to as “wedge” analyses, which compare the actual wages paid by defendants to the wages defendants would have been willing to pay given the actual supply of nursing hours.65 Calling plaintiffs’ expert’s statements “vague and inscrutable,” the court rejected the wedge analysis.66 The court found many problems with the wedge analysis, one of which was averaging.67 While the average wage may have been reduced from the alleged conspiracy, it does not follow that each class member’s wage was actually reduced as a result of the conspiracy.68 Conversely, because the evidence showed a substantial variation in the compensation of individual RNs during the class period, the court found that some members of the class were treated differently than others and thus not affected by the alleged conspiracy.69 In addition, the wedge analysis made no attempt to distinguish between registry nurses and nonregistry nurses despite the fact that registry nurses made up 20% of the class and that registry nurses’ base wages are calculated differently.70 The failure to account for such a disparity, the court found, rendered the model utterly unable to show common impact on a classwide basis. While the court did not exclude the evidence on the grounds that it failed under Daubert, it held that because the method failed to provide a reliable basis for plaintiffs to show common impact, it need not decide whether the wedge analysis passed muster under Daubert.71 In light of Comcast, future courts are much more likely to grant the Daubert motion in these circumstances and reserve on Daubert only where it is a much closer call. These principles apply with equal force outside of the price-fixing context. For example, in In re Wholesale Grocery Products Antitrust Litigation, the plaintiffs alleged that two large grocery wholesalers conspired to allocate territories and customers using an Asset Exchange Agreement (“AEA”).72 The plaintiffs presented several theories of common proof that
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 25
could be used to show impact, including price lists and two methodologies used by their expert: contrary hypothesis theory and a variance test. The court rejected these methodologies as insufficient to show damages on a classwide basis because they improperly relied on averages and oversimplified the individual plaintiff’s price negotiations. Specifically, the court rejected the contrary hypothesis theory because it could not confirm that each class member was impacted: Here, the contrary hypothesis test does not address the price levels after the AEA and whether each member of the New England Class was in fact charged a supracompetitive price. If accepted as valid, the contrary hypothesis test proves that SuperValu’s presence in New England influenced C & S’s prices throughout that region. However, even accepting that premise, the contrary hypothesis shows nothing about prices for C & S customers after the AEA (and in fact does not analyze upcharges after the AEA at all).73
The court similarly rejected the variance test, which “is a statistical measure of the spread of data calculated by comparing each data point’s value with the average value of the data set.”74 The court found that this methodology “only compares averages.”75 However, “[t]hat profits may have increased on average, does not mean that monopolist profits were extracted from each class member.”76 Rather than accept these theories that assumed all class members were impacted in the same way, the court focused on the fact that “prices are all the result of individual negotiations “influenced by, among other factors,” the size of orders, frequency of orders, and transportation costs.”77 Accordingly, the court denied class certification. Again, the court did not conduct a Daubert analysis, but after Comcast, courts will be much more likely to exclude expert testimony under Daubert where it erroneously relies on averages and oversimplified information that is not consistent with the anticompetitive theory of the case. Methodologies that Ignore Critical Variables and Issues with the But-For World. Finally, courts will continue to be presented with methodologies that, in theory, could pass Daubert, but remain flawed in the face of facts highlighted by the defendants that affect the price paid by individual putative class members. For example, regression models that ignore significant potentially individualized variables have been held to be an inadequate form of common proof. While these factors often exist in price-fixing cases, they are just as likely to appear in vertical nonprice restraint cases or monopolization cases where prices are negotiated on an individualized basis.
26
JAMES KEYTE ET AL.
For example, in In re Live Concert Antitrust Litigation, plaintiffs’ expert presented four types of statistical analyses, one of which was called the “yardstick approach.”78 At the first step of the “yardstick approach,” plaintiffs’ expert calculated the average ticket price for rock concerts in each geographic market. He then calculated the average ticket price for concerts promoted by the defendants in those markets and compared the two average prices. Finding that the concerts promoted by defendant had a higher average ticket price, the expert concluded that: (1) the average higher price was a direct result of the alleged conspiracy and (2) all class members were injured because they paid the higher price.79 The court flatly rejected this method, finding that the “yardstick” comparison ignored other critical factors that determine average ticket prices. Because the yardstick method did not account for other possible explanations for the difference in average ticket prices, including artist quality, popularity and venue size, it could not properly estimate impact in the butfor world, and the court excluded the evidence generated from the “yardstick” method.80 Similarly, in In re Cox Enterprises, Inc. Set-Top Cable Television Box Antitrust Litigation, plaintiffs alleged that defendant illegally forced its customers to rent a Cox set-top box in order to gain full access to defendant’s premium cable services.81 As a result of defendant’s alleged tying scheme, plaintiffs claimed they paid supracompetitive prices for the tied product, the set-top box. In order to prove that all class members actually were injured by paying the supracompetitive price, plaintiffs’ expert proposed the “GRS test.”82 The GRS test estimated the overcharged price by comparing the actual price paid by Cox customers to the “but-for” price the price Cox customers would have paid absent the alleged conduct. The proposed damages would be calculated by multiplying the overcharge price by the number of settop boxes rented. Although the court refused to decide whether the GRS test satisfied Daubert,83 it considered the Daubert requirements when evaluating whether the GRS test could be applied to the entire plaintiff class.84 The court particularly found flawed the test’s factor that weighed demand elasticities, faulting the expert for relying on academic estimates that were not Cox-specific to calculate demand elasticities. The court found that in order to estimate demand elasticities, an expert must consider the extent of market competition. And because market competition varies in different parts of the country, the court held that it was improper for plaintiffs’ expert to rely on common market data contained in the academic estimates. Thus, because the GRS test could not estimate demand elasticities unless it considered individualized market data, which it did not do in this case, the court denied certification.85
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 27
Going forward, courts are likely to apply the rigor required under Rule 23(b)(3) to regression models (or other quantitative analyses) that ignore obvious market facts or variables affecting price that, if addressed, may show a lack of impact to some number of class members. And where such potential is evident on the record, courts are very likely to exclude the expert’s opinion based on a threshold Daubert review without ever reaching the full review of the record that is required where an expert’s testimony is not facially defective.
STEP 2: EVALUATE IF THE PROPOSED METHODOLOGY FOR PROVING IMPACT COULD BE USED BY EVERY CLASS MEMBER IF APPLIED IN INDIVIDUAL CASES Where a plaintiff’s class action expert survives a Daubert challenge, this by no means suggests that class certification is inevitable or easy. Instead, survival of a Daubert motion is a necessary, but not in and of itself sufficient, condition for class certification.86 In these circumstances, plaintiffs must still meet their burden of showing that the record evidence, including the expert’s analysis and testimony, supports a finding that impact for each class member can be proven through evidence that is common to the class. As described below, the general framework for this inquiry is to (i) assess whether there are potential individualized issues of proof with respect to how the market operates, such as how prices are determined, (ii) understand how plaintiffs purport to show that the “but-for” price i.e., that which would exist without the alleged misconduct can be demonstrated with common proof, and (iii) analyze whether plaintiffs met their ultimate burden of demonstrating that, at trial, the essential element of impact can be established through common evidence. The most common pitfalls that plaintiffs face in making this showing are outlined below. Courts Must Perform a “Rigorous Analysis” of an Expert’s Opinion and Resolve the “Battle of the Experts,” Even if an Expert’s Report Survives Daubert As a threshold issue, it is the district court’s responsibility to determine the persuasiveness of admissible expert testimony when necessary to resolve class certification issues even if any flaws in an expert’s opinion do not rise to the level of excludability under Daubert. Plaintiffs who prevail under
28
JAMES KEYTE ET AL.
Daubert may try to sidestep this burden by emphasizing that courts historically have considered price-fixing conspiracies well-suited for class treatment.87 However, “it does not follow [from Amchem] that a court should relax its certification analysis, or presume a requirement for certification is met, merely because a plaintiff’s claims fall within one of those substantive categories.”88 The court still must determine if the expert’s opinion provides sufficient analysis to warrant certification. In Dukes, the Supreme Court clarified once and for all that a court considering class certification under Rule 23 must look beyond the pleadings and conduct a “rigorous analysis” of the evidence to assess whether plaintiffs’ proposed class meets the requirements of Rule 23.89 As the Dukes Court observed, this rigorous analysis often requires a close look at the merits of the evidence put forth by the plaintiffs: Frequently that “rigorous analysis” will entail some overlap with the merits of the plaintiff’s underlying claim. That cannot be helped. The class determination generally involves considerations that are enmeshed in the factual and legal issues comprising the plaintiff’s cause of action.90
Accordingly, it is the court’s job to resolve any inconsistencies between the opinions of the experts. Significantly, most of the problems addressed above that may provide the basis for a Daubert motion, also provide defendants a basis to argue that, regardless of whether the plaintiffs’ expert’s testimony is formally excluded, it is still insufficient to satisfy plaintiffs’ burden in proving that common issues predominate. Thus, for example, while reliance on a “price structure” analysis or “averages” may provide fruitful grounds for a Daubert motion, they also provide a basis for defendants to argue that plaintiffs have failed to meet their burden in demonstrating that they could prove each class member’s claim through common proof. Of course, there certainly was a trend in this direction in the case law well before the Supreme Court offered its own guidance in Dukes and Comcast.91 Dukes and Comcast make clear that this trend is now a requirement for lower courts. Thus, courts must perform a rigorous analysis and weigh the persuasiveness of the experts for both sides even where the expert’s testimony survives a Daubert analysis. Defendants Should Attempt to Develop a Factual Record Demonstrating that Plaintiffs’ Expert’s Assumptions about the Actual World Are Incorrect Even if a plaintiff’s expert uses an accepted methodology that is capable of demonstrating common impact under certain circumstances, that does not
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 29
mean those circumstances exist in every case. Expert’s opinions often mask or ignore inconvenient evidence regarding how a market actually operates, which may include individualized issues as to how prices are determined. The new rigor at the class certification stage opens up opportunities for defendants to challenge these experts’ opinions. In cases where the plaintiffs’ expert’s basic methodology can survive a Daubert challenge, the foundation for defeating class certification will often lie in the quality of the factual record that defendants can muster. Critically, even though plaintiffs bear the burden of proof, defendants are wise not to rely on a critique of the plaintiffs’ evidence, but rather to develop their own evidence demonstrating why the plaintiffs’ proffered methodology does not work in that particular case. Specifically, attacking the plaintiffs’ factual assumptions regarding how the market actually operates depends on the defendants developing a robust factual record through declarations, deposition testimony, analysis of the relevant pricing documents and analysis of the available data.
Price Lists and Price Increase Announcements To meet their burden of proof, plaintiffs commonly attempt to rely upon “price lists” or price increase announcements that at least ostensibly relate to all products purchased by putative class members. Prior to the more recent pronouncements regarding the rigor that courts should apply when determining class certification, courts were inconsistent in how to handle such evidence. Even in cases in which defendants argued that prices were set through complex, individualized negotiations, courts often found that a uniform price increase raises the base price at which those negotiations begin and hence lends itself to common proof sufficient for the class certification stage.92 Conversely, other courts declined to certify classes despite the use of price lists or price announcements when defendants were able to proffer evidence that individual variables impacted the setting of prices.93 More recently, courts have been more willing to take a critical look at the relationship between price lists or price increase announcements and actual prices, recognizing that such evidence is only susceptible to common proof if the price lists or announcements are actually used and followed. For example, in Plastics,94 plaintiffs proffered a collection of defendants’ documents that set forth price lists and price increase announcements for tins and ESBO. Because the plans detailed in the documents applied to all tin and ESBO products, plaintiffs claimed, all class members were impacted when they purchased those products.
30
JAMES KEYTE ET AL.
In response, the defendants offered evidence demonstrating that there were wide variations in the prices actually paid by members of the putative class, including that the prices paid by some purchasers remained constant after the price increase announcements were issued and that prices for at least some putative class members declined during the class period.95 Rather than blindly accepting the plaintiffs’ assertion that all putative class members were impacted by the conspiracy in a manner that could be demonstrated through common proof, the Plastics court considered the record evidence in the case and concluded: “[T]he evidence of record shows that the prices paid by customers did not correspond with Defendants’ price increases. Accordingly, the price lists and price increase announcements cannot serve as common evidence of impact.”96 Similarly, in In re Florida Cement and Concrete Antitrust Litigation, the court considered allegations that defendants conspired to raise the price of concrete by $25. Because the defendants all announced a $25 increase during the summer of 2008, plaintiffs asserted that the defendants’ price increase announcements could serve as common proof of impact. The court disagreed, finding evidence “[t]hat some customers avoided paying price increases or artificially-stabilized prices, is inconsistent with Plaintiffs’ theory of common impact.”97 The court highlighted the flaws in the expert’s uncritical reliance on the price lists: In fact, Plaintiffs present no empirical analysis of the actual effect of the price increase announcements on individual customers to counter the evidence presented by Defendants, and unlike Dr. Ordover, Dr. Mangum did not conduct any significant analysis at the individual customer level to determine whether any price changes were consistent across the putative Class. Thus, Plaintiffs have not shown how the price increase announcements even if they were intended to affect all customers across-theboardconstitute common evidence through which impact on the individual class members can be proven.98
Finally, in In re Graphics Processing Units Antitrust Litigation, the court considered allegations of a price-fixing conspiracy for the sale of graphics processors, which are “designed based on the specific application for which they will be used.”99 The court recognized that “[w]ithin these markets, defendants sold chips and cards of varying performance levels” based on the particular needs of the purchaser.100 The court also highlighted that the graphics processor “products at issue were sold to a variety of customers through a number of distribution channels.”101 The court focused on these complex characteristics that distinguished purchasers (all of whom were putative members of the same class) and
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 31
concluded that price lists were not an accurate way of determining how prices were set in the GPU market: [O]ver 99.5% of defendants’ revenue during the limitations period came from sales with large wholesale purchasers like Microsoft. … [T]he vast majority of sales were primarily executed after customized negotiations between wholesalers and either defendant. These sales were made without any regard to a price list. As such, defendants’ sales contracts varied significantly depending on the wholesale purchaser. … There is no doubt that a myriad of factors played a role in each large transaction. These factors each influenced the final sales price of each transaction.102
Thus, the court held that these price lists could not serve as common proof of impact for the class. As these cases demonstrate, in situations where the plaintiffs focus on price increase letters or price lists to establish common impact, defendants should make every attempt to determine whether individual customers actually paid prices that corresponded with those published prices. Developing evidence demonstrating that the prices in the real-world market do not correspond to those price lists, or that individuals were able to avoid published price increases, should force the plaintiffs’ expert to come up with something more to show common proof of impact. Market Characteristics Additionally, plaintiffs’ experts sometimes rely upon “market characteristics” to support their conclusion that impact can be shown through common proof. The critical inquiry is generally whether the record evidence demonstrates that the products are homogeneous.103 Even before the trend toward a more rigorous analysis, courts have long found that when “the distinctions among the [products] offered are substantial[,] the process of assessing impact would be anything but mechanical.”104 But this is an area that has seen increased scrutiny since Hydrogen Peroxide and that trend should continue in the post-Comcast environment. For example, in Plastics, the court recognized that, “in theory,” characteristics such as the homogeneity of the products at issue and defendants’ domination of the relevant market could make a market “vulnerable” to a price-fixing conspiracy.105 However, the court then discussed at length the evidence proffered by defendants demonstrating the heterogeneity of the products at issue: the different quality and strength properties in the defendants’ products affected whether such products were suitable for different types of buyers’ end uses, and customers clearly indicated a preference for the products of one defendant over another.106 These qualities made it
32
JAMES KEYTE ET AL.
impossible for the court to find that market characteristics could serve as common proof of antitrust impact. Similarly, in Florida Concrete,107 the court weighed the evidence offered by both sides and found that “Plaintiffs fail to demonstrate that Concrete is indeed a homogenous product such that it can be used interchangeably.”108 Importantly, the court rejected the notion that a defendant’s expert must “perform enough quantitative analyses to prove that Concrete is not interchangeable and that the market characteristics are not such that the impact of the purported conspiracy would have been unavoidable.”109 Rather, the court highlighted that “it is not Defendants’ burden to prove that the impact of the alleged conspiracy is not susceptible to common proof; to the contrary, it is Plaintiffs’ burden to prove the impact is susceptible to proof by common evidence.”110 Additional factual issues for defendants to consider include whether prices are set by a single individual or by a large group of individuals operating in the field. The same types of individual issues that can undermine an employment class action in which plaintiffs are challenging employment decisions made by scores of individual managers across different offices (i.e., Dukes) can exist in an antitrust case in which purchases or sales are handled by scores of individual employees. Even where prices are ostensibly dictated by a single individual, defendants should scrutinize whether the individuals on the front lines actually followed whatever policies or directives were issued. A related issue is whether the relevant industry is a “relationship” driven business in which sales and prices are determined more by the relationship of the individuals involved than by any directives received from corporate headquarters.111 Consequently, where there is record evidence of nonprice factors, such as perceived differences in service, quality, timeliness of deliveries, and willingness to accept returns, that may impact plaintiffs’ purchasing decisions, defendants should challenge the notion that a market for homogeneous products is “susceptible” to common proof based on market characteristics alone.
Defendants Challenging an Expert’s Opinion Under Rule 23 Should Also Attack Any Unreasonable Assumptions Made about the “But-For” World Defendants should attack any unreasonable assumptions built into an expert’s models proffered to show common impact. It is the task of an antitrust plaintiff to “establish a ‘but for’ baseline a figure that would show
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 33
what competitive prices would have been if there had been no antitrust violations.”112 Construction of the “but-for” world has never been a perfunctory task that plaintiffs (or courts) can take for granted; rather, damages can only be calculated “by comparing to that baseline what the actual prices were during the challenged period.”113 As the trend toward greater rigor in class certification continues, it will be essential that courts evaluate whether the plaintiffs’ “baseline” matches up with what really happened in the relevant market or what would likely happen if the alleged misconduct did not exist. In straightforward instances of an alleged price-fixing conspiracy in a market for a homogeneous product where prices are set by price lists, an expert may be able to construct an economic model that does not contain many or any suspect assumptions about “but-for” pricing and output in the relevant market. In these scenarios, if an expert uses an acceptable methodology and applies it properly to the facts of the case, that expert likely will have offered acceptable opinions that could prove antitrust impact. However, in more complicated cases where pricing is negotiated on an individualized basis or where the alleged misconduct involves nonpricerelated conduct experts are likely to make certain assumptions about the way the market operates that can skew their analyses. This is especially true in those cases where individual plaintiffs may have complicated specifications, preferences, or other characteristics that make it difficult to extrapolate broad principles for the behavior of the entire class. In such scenarios, experts cannot merely “assume” that all plaintiffs will be impacted in the same way.114 Indeed, such an assumption is grounds for the denial of class certification.115 Similarly, as made clear in Comcast, experts can no longer make assumptions that all theories of antitrust liability impact the plaintiffs in the same way.116 Therefore, an expert fails to establish an acceptable “but-for” world if that expert cannot distinguish between supracompetitive prices resulting from antitrust misconduct and “prices whose level above what an expert deems ‘competitive’ [is] caused by factors unrelated to an accepted theory of antitrust harm.”117 Defendants should highlight instances where it is likely that the plaintiffs’ expert is lumping together several theories of harm for example, when plaintiffs include in their complaints market allocation allegations or allegations related to other nonantitrust violations (such as misrepresentations or deceptive practices, or even competitive behavior). In such scenarios, if the court finds that one or more of these theories is not a viable antitrust theory of liability, then the expert’s
34
JAMES KEYTE ET AL.
model must be able to disaggregate those theories of antitrust impact to support certification of the class.118
If Common Questions Predominate for Only Some of the Plaintiffs, the Court Can Sometimes Narrow the Scope of the Proposed Class Even if the court does not deny class certification on the basis of the aforementioned individualized questions of fact, under the rigorous standard, a court may consider narrowing the proposed class based on the predominance requirement. For instance, the In re GPU court denied certification of the plaintiffs’ proposed class but granted certification for a more limited class of direct purchasers.119 The court focused on the fact that some of the proposed class members faced a nonnegotiable price, while other proposed plaintiffs negotiated prices individually with the defendants.120 Accordingly, the court held that “[p]roof that defendants conspired to fix [list] prices would hardly prove that defendants also conspired to fix the non-list prices for the transactions entered into with absent wholesale purchasers.”121 Defendants can ask the court to exclude customers who typically negotiate prices, if defendants find that some plaintiffs did make purchases solely based on list prices. While a court is generally under no obligation to redefine the class, it may be tempted to adopt this approach if it is particularly difficult to resolve the predominance issue. Moreover, in situations where defendants cannot make the aforementioned arguments on the basis of all members of the class, narrowing the proposed class to list-price customers can sometimes provide a way reducing the size of the potential damages in a price-fixing case.
STEP 3: CAN DAMAGES BE PROVEN ON A CLASS WIDE BASIS? The most controversial issue in the wake of Comcast is whether the “rigor” under Rule 23(b)(3) referenced by the majority opinion applies to class action damages methodologies and proof generally or whether, instead, the case can be limited to its facts i.e., the particular flaw in not disaggregating potential causes of harm that are not the basis of antitrust liability (discussed above).
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 35
We discuss below both the debate in the lower courts (and academic circles) on what Comcast means going forward, including how class action damages analysis is or will be affected in those cases where courts apply the full rigor of Rule 23(b)(3) in assessing whether damages for each class member can be proven at trial through common proof. Preliminaries A Whole New World After Comcast? Comcast contains potentially wide-ranging implications for the subject of common proof of the amount of damages. At a minimum, Comcast stands for the proposition that any proposed common proof of damages under Rule 23(b)(3) must flow from the basis of antitrust liability proposed by the plaintiffs and accepted by the court. Thus, an expert’s model calculating damages that failed to disaggregate potential causes of harm from those harms that the lower court accepted as a possible basis for antitrust liability was not consistent with plaintiffs’ liability case.122 In other words, the model did not link antitrust harm to damages, dooming the plaintiffs’ class certification motion. But the majority’s opinion suggests a new rigor for scrutinizing class action damages evidence well beyond the narrow proposition that a damages model must link damage to harm. The opinion suggests that the rigor Dukes required courts to apply under Rule 23(a) applies now with equal, if not greater, force under Rule 23(b)(3). This would include, presumably, rigorous analysis in assessing common evidence regarding the measurement of damages. The Comcast dissent claims the majority’s opinion could not possibly be read to require damages be measureable for each class member. But some courts, and commentators, do not see it the dissent’s way at all; suffice it to say that what the future holds is far from certain. Comcast Majority: The Need for Rigor Regarding Damages In addition to the uncontroversial holding that a damages model must link the theory of harm to impact, the Comcast majority applied additional rigor in evaluating common proof of the amount of damages. The Court started from the observation that the predominance criterion under Rule 23(b)(3) demands even stronger proof than Rule 23(a),123 and noted that Rule 23(b)(3) itself is “an adventuresome innovation designed for situations in which ‘class-action treatment is not as clearly called for’.”124 It held that a close look, including at arguments pertinent to the merits, is required to
36
JAMES KEYTE ET AL.
satisfy predominance.125 Under this rigorous standard, a proposed class cannot show Rule 23(b)(3) predominance in the damages context where “[q]uestions of individual damage calculations will inevitably overwhelm questions common to the class.”126 The Court added, even if the model had identified subscribers who paid more solely because of the deterrence of overbuilding, it still would not have established the requisite commonality of damages unless it plausibly showed that the extent of overbuilding (absent deterrence) would have been the same in all counties, or that the extent is irrelevant to effect upon ability to charge supra-competitive prices.127
Applying this rigorous standard to the damages model proffered by plaintiffs’ expert, the Court held that where the model measures “damages” that are not attributable to an accepted theory of liability, “the model cannot possibly establish that damages are susceptible of measurement across the entire class for purposes of Rule 23(b)(3).”128 The Court also identified three other situations where individualized damages might overwhelm common questions: (1) where plaintiffs’ methodology cannot prove harm for a large portion of the class;129 (2) where the “permutations” of damages are “nearly endless”;130 and (3) where the proposed damages methodology is speculative.131 Comcast Dissent: Attempting to Fence in the Majority The dissent, written by Justice Ginsburg and Justice Breyer, with Justices Sotomayor and Kagan joining, argued that the Court’s holding with respect to damages issues should be limited to the particular facts of the case. These four justices claimed that “[t]he decision should not be read to require, as a prerequisite to certification, that damages attributable to a classwide injury be measureable ‘on a class-wide basis.’ ”132 According to the dissent, plaintiffs must demonstrate and courts must find that questions of law or fact common to class members predominate over questions affecting only individual members, but predominance does not demand commonality as to all questions.133 The dissent claimed that, [r]ecognition that individual damages calculations do not preclude class certification under Rule 23(b)(3) is well nigh universal … In the mine rune of cases, it remains the ‘black letter rule’ that a class may obtain certification under Rule 23(b)(3) when liability questions common to the class predominate over damages questions unique to class members.134
Indeed, the dissent argued that if a class action is an efficient mechanism for resolving common liability questions, “the predominance standard is generally satisfied even if damages are not provable in the aggregate.”135
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 37
The dissent’s analysis with respect to damages is inconsistent with the rigor reflected in the majority opinion. Indeed, in the months since the case was decided, the Court has remanded several Circuit level antitrust cases that lacked common proof of damages “in light of Comcast.” Lower courts also have applied the Comcast court’s rigorous analysis to proof of damages in the class setting, although the responses to Comcast are far from uniform.
How Comcast Is Being Applied to Class Damages Analyses Greater Rigor Required Cases Vacated and Remanded (GVR) by the Supreme Court. Following Comcast, the Supreme Court has granted certiorari on, and vacated and remanded three cases “in light of Comcast.” In two of these cases, In re Whirlpool Corp. Front-Loading Washer Products Liability Litigation136 and Butler v. Sears, Roebuck and Co.,137 the likely cause for remand appeared to be a lack of rigor in analyzing proof of damages.138 In Whirlpool, consumers brought a putative class action against Whirlpool in connection with sales of front-load washing machines. Plaintiffs alleged breach of warranty, negligent design, and negligent failure to warn. According to the plaintiffs, Whirlpool’s front-load washing machines suffered from a design defect that made them prone to develop mold and mildew.139 Plaintiffs alleged a class comprised of current Ohio residents who purchased one of the specified washing machines for personal, family or household purposes.140 Whirlpool opposed class certification on the following grounds: • The vast majority of washing machine owners had not experienced a mold problem. • Whirlpool made dozens of changes between 2002 and 2009 to increase customer satisfaction and reduce service costs. • Washers owned by class members were built on two different (although substantially identical) platforms, involved 21 different engineering models and spanned 9 model years. • Consumer laundry habits are so diverse that plaintiffs presented individual liability questions.141 The district court held that common questions of liability predominated and certified the class under Rule 23(b)(3). It noted that “ ‘[n]o matter how individualized the issue of damages may be, these issues may be reserved
38
JAMES KEYTE ET AL.
for individual treatment with the question of liability tried as a class action.’ ”142 In a limited discussion, the Court noted that division of the class into subclasses would be sufficient to overcome any individualized issues with respect to damages.143 In Butler, consumers brought a class action against Sears regarding its sale of allegedly defective Whirlpool washers based on warranty laws in six states.144 The plaintiffs made claims regarding two alleged design defects: mold claims, similar to those made in Whirlpool, and control unit claims.145 The control unit claims involved allegations that a soldering defect caused control units to mistakenly identify a serious error with the washing machine and automatically shut down the machine mid-cycle.146 Plaintiffs sought to certify two classes: a class with “mold” claims and a class with “control unit” claims. For both classes, the court stated that there were common questions on liability. For the mold class, the question was, “were the machines defective in permitting mold to accumulate and generate noxious odors”; for the control unit class, the question was “whether the control unit was indeed defective.”147 The court acknowledged that individual questions existed on the issue of the amount of damages and, at least for the mold class, that most members did not experience a problem. The district court certified the control unit class and denied certification for the mold class. The Seventh Circuit affirmed certification of the control unit class and reversed denial of certification for the mold class,148 with Judge Posner favoring an “efficiency” approach to the predominance inquiry. According to the opinion, efficiency would be served through a single proceeding to resolve a central liability issue whether the washing machines were defective. This liability proceeding “could be followed by individual hearings to determine the damages sustained by each class member.”149 In both Whirlpool and Butler, the lower courts failed to apply rigor to their analysis of proof of damages, instead suggesting that it could be resolved later. At the same time, the lower courts acknowledged that numerous plaintiffs within the certified classes did not suffer any direct harm, and even those plaintiffs claiming actual harm suffered “nearly endless permutations” of alleged harm. This is precisely the type of harm the Court found inadequate to establish classwide proof of damages in Comcast. Although the Court did not explain why it was remanding these cases in light of Comcast, most commentators reasoned that it must be based on the greater rigor courts must use to assess theories of damages. For example, a Bloomberg Class Action Litigation Report suggested that the
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 39
Supreme Court’s remand of Whirlpool “could prompt additional scrutiny of the lack of injury for the majority of class members” as well as the 21 different types of washing machines purchased by class members. In both cases, any damages evidence must account for intraclass differences under Comcast.150 In another article commenting on the remand of both Whirlpool and Butler, the author noted: In both Comcast and Wal-Mart[, the] linchpin to certification is assuring that determination of whether defendant’s conduct caused injury to each class member can be made classwide and without resort to individualized assessments of each member’s circumstances … The major infirmity in Butler and Whirlpool is that each overlooked myriad permutations among hundreds of thousands of purchasers of different product models concerning the presence of mold and mildew, their causes, amounts of any resulting damages, customers’ care of washers, whether requests for warranty service were made and timely, and defendants’ responses to warranty claims.151
He suggested, drawing on the language from Comcast, that when “permutations” of plaintiffs’ alleged damages “are nearly endless,” damages may not be capable of classwide measurement and instead may devolve into “labyrinthine individual calculations.” The article also took specific issue with Judge Posner’s “efficiency standard” in Butler. The Butler court admitted that both whether each class member suffered any damages and, if so, the amount of those damages were individualized questions. The author argues the single common issue identified in Butler, “were the machines defective,” cannot satisfy the Rule 23 test.152 Another article concurred: “[T]he Butler court’s treatment of the need for individual damages trials seems flatly inconsistent with the Comcast court’s statements on the need for proof on a class-wide basis.”153 Surprisingly, the Supreme Court’s directive to reconsider the cases “in light of Comcast,” and the views of most commentators, largely fell on deaf ears. In Whirlpool, the Sixth Circuit subsequently issued an opinion holding that the front-load washing machine class was properly certified notwithstanding Comcast. Relying on the Comcast dissent, the Sixth Circuit viewed the Court’s decision of limited relevance to the case before it because the district court “certified only a liability class and reserved all issues concerning damages for individual determination.”154 In addition to essentially ignoring the GVR order on damages issues, the Sixth Circuit’s ruling overlooks the implications of Comcast for variations of injury within a putative class e.g., the odor problem that manifested in only a small percentage of the washing machines.155 Yet this only highlights that putative class members did not “suffer the same injury.”156
40
JAMES KEYTE ET AL.
Shortly thereafter, in Butler, the Seventh Circuit followed suit and reinstated the ruling that the classes could be certified, even in light of Comcast. The Butler opinion continued to rely on a standard centered around judicial efficiency, rather than predominance, as the requirement for class certification under Rule 23(b)(3). It also fails to grasp the implications of Comcast; indeed, at one point the court asks, puzzled, “[b]ut if we are right that this is a very different case from Comcast, why did the Supreme Court remand the case to us for reconsideration in light of that decision?”157 Further, when the Seventh Circuit does address individualized damages issues, it sets up what appears to be a faulty dichotomy: If the issues of liability are genuinely common issues, and the damages of individual class members can be readily determined in individual hearings, in settlement negotiations, or by creation of subclasses, the fact that damages are not identical across all class members should not preclude class certification. Otherwise defendants would be able to escape liability for tortious harms of enormous aggregate magnitude but so widely distributed as not to be remediable in individual suits.158
This result, Judge Posner warns, “would drive a stake through the heart of the class action device.”159 But this conclusion only follows if one asserts that, after Comcast, satisfying predominance requires that plaintiffs suffer identical damages a premise neither the Comcast majority nor the Butler defendants ever suggest. Instead, the Comcast majority implies that where endless permutations of plaintiffs’ damages would result in “labyrinthine individual calculations,” individual damage issues may predominate over common questions. The Sixth and Seventh Circuits may not believe such labyrinthine damages calculations are necessary in these cases, but their failure to properly address the issue is difficult to reconcile with the Supreme Court’s directive that the cases be reconsidered “in light of Comcast.” More recently, however, the D.C. Circuit adopted the approach many anticipated the Sixth and Seventh Circuits would take. In In re Rail Freight Fuel Surcharge Antitrust Litigation, the D.C. Circuit exercised its jurisdiction over an interlocutory appeal from a class certification decision and vacated and remanded on certification grounds, in part due to the Comcast ruling.160 In that case, shippers brought a class action against the four major freight railroads alleging a price-fixing conspiracy regarding the railroads’ imposition of rate-based fuel surcharges.161 Plaintiffs sought to certify a class of shippers who paid the purportedly inflated fuel surcharges.162 Plaintiffs’ expert prepared two regression models that would work in conjunction to demonstrate predominance under Rule 23(b)(3).163 The
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 41
“common factor model[ ] attempted to isolate the common determinants of prices shippers paid,” and the “damages model” attempted to quantify the overage due to the alleged price fixing.164 The district court accepted the models as “plausible” and “workable.”165 The D.C. Circuit found plaintiffs’ model defective because it proved too much. While the model appeared to show injury to all class members, it also “detect[ed] injury where none could exist.”166 This was because one group of shippers had entered in “legacy contracts” with defendants.167 These contracts guaranteed fuel surcharges would be subject to formulae that predated the allegedly anticompetitive surcharge changes.168 The expert’s damages model showed similar damages for legacy contract shippers and class members.169 This critique of the expert’s model, the court explained, was “sharpen[ed]” by Comcast. After Comcast, “[i]f the damages model cannot withstand this scrutiny then, it is not just a merits issue. [Plaintiffs’ expert’s] models are essential to plaintiffs’ claim they can offer common evidence of classwide injury. No damages model, no predominance, no class certification.”170 Thus, when a damages model includes damages for parties that could not have been harmed by defendants’ actions, the model must fail. As the court stated, “we have no way of knowing the overages the damages model calculates for class members is any more accurate than the obviously false estimates it produces for legacy shippers.”171 The court’s analysis in Rail Freight further exposes the flaws in Whirlpool and Butler upon remand. As the D.C. Circuit pointed out, “[i]t is not enough to submit a questionable model whose unsubstantiated claims cannot be refuted through a priori analysis.”172 Were that the case, “at the class-certification stage any method of measurement is acceptable so long as it can be applied classwide, no matter how arbitrary the measurements may be.”173 This would contradict the import, if not the holding, of Comcast. A proper damages model must not only prove the possibly of damages, it must be limited to prove damages for only those plaintiffs who did or could possibly suffer harm. As courts continue to weigh in on the standard for damages models under Rule 23(b)(3), we expect courts to embrace the logic of the D.C. Circuit in Rail Freight and further examination of the Sixth and Seventh Circuits’ decisions, which seem wholly inconsistent with Comcast. Cases Applying Comcast’s Rigor Requirement to Proof of Damages. Applying the Comcast rigor requirement to proof of damages in putative class actions, several courts have held that class treatment under
42
JAMES KEYTE ET AL.
Rule 23(b)(3) was not warranted because individualized issues on damages would predominate over any common issues.174 For example, in Roach v. T.L. Cannon Corp.,175 employees sued Applebee’s franchises in New York and Connecticut alleging violations of the New York Labor Law (“NYLL”) and the Fair Labor Standards Act (“FLSA”). Plaintiffs sought to certify several classes, including a class of employees who were allegedly denied overtime wages when they worked more than 10 hours in any particular day (“spread of hours”). A magistrate judge recommended that a class be certified with respect to plaintiffs’ spread of hours claim, but the district court rejected the recommendation due to the subsequently issued Comcast decision. The court described Comcast as holding that the failure of a class proponent “to offer a damages model that is ‘susceptible of measurement across the entire class for purposes of 23(b)(3)’ was fatal to the certification question.”176 Specifically, the court found that the plaintiffs’ failure to offer a damages model susceptible of measurement across the entire class “[was] in contravention of the holding in [Comcast].” Evidence that “some employees, on various occasions, were denied their 10-hour spread payments” indicated to the court that damages in the putative class would require highly individualized proof.177 Thus, the court concluded that questions of individual damage calculations would overwhelm common questions, and as such, Rule 23 certification had to be denied. Similarly, in Cowden v. Parker & Associates, Inc.,178 the district court refused to certify a class because individual questions on damages would predominate over any questions common to the entire class. There, plaintiff insurance agents sued Parker & Associates, an insurance agency, for fraud, negligent misrepresentation, breach of contract, unjust enrichment, conversion and promissory estoppel, alleging that defendant failed to pay plaintiffs commissions for their sales of Medicare Advantage (“MA”) Plans. Plaintiffs sought to certify a class of insurance agents who had worked for defendant as insurance agents selling MA Plans.179 Plaintiffs recognized that their claims would require individual analyses for each agent’s sales and expenses. Plaintiffs’ compensation was commission-based, but could not be reduced to a formula or simple calculation. Instead, commissions were shared within a hierarchy of agents, through which an agent’s superiors shared in the commissions of any agent lower on the hierarchy. It appeared that compensation was decided based on oral representations to each agent. The defendant was also permitted to deduct from each agent’s commission check expenses and fees including mailing expenses, fees for leads, commissions advanced to the agent but
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 43
not earned because the carrier ultimately rejected the policy and policy premiums owed to the defendant. Citing In re Whirlpool, plaintiffs argued that “ ‘no matter how individualized the issue of damages may be, these issues may be reserved for individual treatment with the question of liability tried as a class action.’ ”180 The court noted that Whirlpool had been vacated for further consideration in light of Comcast, and found that, as in Comcast and Roach, plaintiffs “offered no manageable way to calculate damages across the entire class, and the individual damages calculations that would be required will inevitably overwhelm any questions common to the entire class.”181 In re Montano v. First Light Federal Credit Union,182 another postComcast decision, involved an adversary proceeding in bankruptcy court where the debtors sought class injunctive, declaratory, and monetary relief based on a credit union’s alleged violation of a discharge injunction through continued reporting of discharged debt as “past due” rather than “discharged.” The district court denied plaintiffs’ motion for reconsideration of the bankruptcy court’s decision to decertify the class. It cited several bases for this decision, including Comcast, which the Supreme Court handed down one week after the bankruptcy court’s decertification ruling. “Comcast,” the court stated, “bolsters the Court’s decision to decertify [the damages class], because there is no evidence [that] the damages they suffered would be susceptible of class-wide measurement.”183 Any damages would be individualized, “especially given the evidence that payment of the discharged debt likely would not be an element of damages in more than a few cases.” And even if a model existed that was capable of proving classwide damages, the court held that after Comcast, plaintiffs are not allowed “the luxury of waiting until trial” to come forward with proof that damages could be measured on a classwide basis.184 Cases Purporting to Follow the Comcast Dissent A few courts have found the dissent persuasive and followed its approach on damages.185 The Ninth Circuit has strongly resisted the Comcast opinion, instead siding with the dissent. The leading Ninth Circuit case on the issue is Leyva v. Medline Industries Inc., another employment class action seeking damages for violation of state labor laws.186 The case concerned alleged employer rounding down of employee time and improper application of nondiscretionary bonuses to overtime pay. The court rejected a broad reading of Comcast and found that in the Ninth Circuit, “the presence of individualized damages cannot, by itself, defeat class certification.”187 The court found that Comcast requires only that plaintiffs show
44
JAMES KEYTE ET AL.
their damages “stemmed from the defendant’s actions that created the legal liability.”188 One key fact in this case, which arguably distinguishes the case from most other Rule 23(b)(3) class actions, was that defendant’s removal notice relied on its own electronic databases to calculate the amount in controversy for each claim and totaled the exposure on all claims. The defendant’s apparent admission confirmed to the court that damages calculations would only entail a straightforward, mechanical application of readily available data to individual plaintiffs. The ability to calculate individualized damages through such a process likely would prevent individual questions from overwhelming questions common to the class. That interpretation of Leyva was relied on and adopted in Parra v. Bashas’, Inc., where the court found that plaintiffs’ methodology for calculating back pay showed that damages were capable of measurement on a classwide basis. The court acknowledged that Comcast requires proof that damages could be determined on a classwide basis, but found that the plaintiffs’ methodology for calculating damages correlated the legal theory of harm with the economic impact of that event. Further, similar to Leyva, the court found that once liability is established, calculating damages would be “a purely mechanical process” and thus there was no concern that individual questions would overwhelm common questions. Martins v. 3PD, Inc.,189 a post-Comcast case outside the Ninth Circuit, interpreted Comcast in a similar fashion. The court noted that even before Comcast, courts of appeal have found that predominance could be defeated where “questions of damage calculation are so complex or implicate so many potential class members, that they present a ‘Herculean task’ and overwhelm liability issues.”190 The court interpreted Comcast to leave open the possibility of class certification where some individual issues on the calculation of damages remain as long as those determinations “will neither be particularly complicated nor overwhelmingly numerous.”191 The Ninth Circuit has been criticized by commentators for its narrow interpretation of Comcast in the Leyva case. For example, on the issue of rounding, a plaintiff would need to establish that he clocked in and started working before his official start time. He would also need to establish the defendant rounded up to the official start time, illegally denying him compensation. As one author notes, “there could be significant individualized issue[s] as to when any particular employee started working on a given day.”192 Another article suggested that Leyva, and not Comcast, is limited to its facts. Leyva had “straight-forward damages at issue” and “did not involve
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 45
any argument that individual damages calculations would be expensive, extensive, time-consuming or complex.”193 The authors note that Leyva relies on a pre-Comcast Ninth Circuit decision to establish that damage calculations alone cannot defeat certification.194 According to the authors, that premise may still be valid where damages involve “mere mathematical computation[s] of damages based on known data.”195 But, if the Leyva court meant that individualized damage inquiries can never predominate over common liability issues, that opinion “cannot be squared with Comcast.”196
Ways to Attack Damages Evidence Under the New Rigor Defendants opposing class certification should make use of the Supreme Court’s language in Comcast and seize on situations where individualized questions might overwhelm common questions. The following section lays out possible ways to attack motions for class certification based on failure to offer common proof of damages. This advice is drawn from Supreme Court guidance, post-Comcast case law and commentary as well as case law predating Comcast that has been thrust back into the mainstream following the decision. In the sections that follow, the Article suggests defendants should first evaluate whether plaintiffs have provided a developed damages model. Then, assuming the plaintiffs have set forth a model, defendants should analyze whether damages modeled are susceptible to mathematical or formulaic calculation and attempt to identify individualized issues among the plaintiffs or the negotiation for the product. Does the Plaintiff Provide a Formula to Calculate Damages? At the most basic level, plaintiffs cannot establish that damages are susceptible of measurement across the entire class for purposes of Rule 23(b)(3) without providing a formula that endeavors to accomplish the task. After Comcast, plaintiffs cannot promise that they will develop a damages model; they must have a model in place at the class certification stage. One commentator noted: “[After Comcast,] if a case is too big to devise a damages process that matches actual damages to actual litigants, a majority of the current court is likely to treat it as ‘too big’ for litigation.”197 Roach v. T.L. Cannon Corp., a post-Comcast case discussed above, adopted this approach. Plaintiffs argued that they could address the liability question first without supplying a damages model because the damages issue “is separate from the question of liability.”198 Even if such damages
46
JAMES KEYTE ET AL.
might be highly individualized, Plaintiffs “contend[ed] that damages need not be considered for Rule 23 certification.”199 The court flatly rejected this argument and found plaintiffs’ failure to offer a damages model was fatal to class certification, following guidance from Comcast.200 Similarly, Rodney v. Northwest Airlines,201 a Sixth Circuit opinion that predates Comcast, denied certification partially due to plaintiffs’ failure to offer a formula to calculate damages. In Rodney, plaintiffs’ experts admitted they had not yet identified a methodology for calculating damages. Instead, they argued it was sufficient that they had a methodology that “could be used to calculate damages.”202 The court did not agree. Without a formula in place to compute damages, plaintiffs could not prove that class issues predominated.203 One commentator suggested that post-Comcast, a methodology capable of providing common proof of damages would be necessary even if plaintiffs wanted to separate the liability and damages portions of a class action. The author noted: [T]he opinion in Comcast is silent as to the availability of Rule 23(c)(4) to bifurcate the problematic damages portion of a class action. … Although Rule 23(c)(4) should remain available as a management tool in cases where common liability questions are found to predominate over individualized damages questions, an attempt to circumvent the holding of Comcast by carving out the individualized damages questions in order to satisfy Rule 23(b)(3) predominance would be at odds with the Supreme Court’s holding in Comcast.204
Are Damages Susceptible to “Mathematical” or “Formulaic” Calculation? If plaintiffs have at least attempted to provide a formula to establish damages across the entire class, defendants should next consider whether damages can be properly calculated under the proposed methodology in a straightforward fashion. Courts have identified three major areas where plaintiffs methodology fails to prove damages on a classwide basis: (1) where the product involves individualized negotiations; (2) where individual plaintiffs differ in ways that make proof of damages highly individualized; and (3) where plaintiffs offer a methodology to prove damages, but that methodology fails to account for differences between plaintiffs. Does the Product at Issue Involve Individualized Negotiations? The postComcast decisions are nearly universal that where the product at issue involves individualized negotiations on price, damages issues are likely to overwhelm any common issues. This approach was taken by the Fifth
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 47
Circuit nearly a decade ago in Piggly Wiggly and is a key consideration when attempting to defeat class certification.205 In Piggly Wiggly, plaintiffs brought a civil antitrust case alleging defendants conspired to fix the prices of bread and cake products in Texas and Louisiana.206 The proposed class included both wholesale purchasers and bid purchasers. Wholesale purchasers, typically small grocery and convenience stores, made purchases based on a price list provided by defendants, while bid purchasers, typically large grocery stores, school districts, and local government entities, received bids from defendants.207 For bid purchasers, the court accepted plaintiffs’ allegation that bids were based on the wholesale price lists given to wholesale purchasers.208 However, the court found that plaintiffs could not establish that class members suffered injury to their business or property as a result of the conspiracy through common proof and therefore denied certification under Rule 23(b)(3).209 Even starting from the wholesale price list, as plaintiffs alleged, the final price ultimately was arrived at after negotiation. The court noted that the final price resulted from many factors, including the amount of product purchased, geographic market, the particular services included, delivery costs, the discount negotiated, and the negotiating skill of the parties involved. Because no two negotiations were exactly the same, “it [would] be impossible to present evidence in a common manner as to the price each Plaintiff would have paid but for the alleged conspiracy.”210 The court further noted that predominance and manageability may be destroyed “solely by the complexity of determining damages when that determination does not lend itself to a mathematical calculation that can be applied to all class members.”211 Plaintiffs’ “mere assertions” that there are a number of methods or methodologies that could be utilized to prove damages were not sufficient.212 The court concluded, “Where the plaintiffs’ damage claims ‘focus almost entirely on facts and issues specific to individuals rather than the class as a whole’, the potential exists that the class action may ‘degenerate in practice into multiple lawsuits separately tried’.”213 “In such cases, class certification is inappropriate.”214 The Fifth Circuit affirmed the district court’s ruling,215 holding that “[t]he necessity of calculating damages on an individual basis, by itself, can be grounds for not certifying a class.”216 Although it acknowledged that in antitrust cases there is a relaxed burden for proving the amount of damages once the fact of damage is proven, the court agreed with the district court that damages may not be merely speculative. The Fifth Circuit agreed with the district court that individualized issues such as negotiating ability and geographic market could not be included in a general formula.
48
JAMES KEYTE ET AL.
Similarly, in Mekani v. Miller Brewing Co.,217 beer retailers brought a civil antitrust case alleging a conspiracy among brewers and beer distributors to fix prices and territories in violation of the Sherman Act as well as price discrimination in violation of the RobinsonPatman Act.218 Plaintiffs alleged that six major brewers and over 180 distributors engaged in horizontal and vertical conspiracies to fix the price of beer and establish and maintain noncompetitive geographic territories.219 The proposed class consisted of beer retailers in Michigan. The court denied certification of both the Sherman Act and RobinsonPatman Act claims. On the Sherman Act claims, the court found that even if common questions would predominate on liability, “plaintiffs are still required to prove fact of injury.”220 The damages allegedly suffered by the class were not a function of a single event or decision made by the defendants, but rather resulted from a combination of decisions and factors specific to a given retailerdistributor transaction: here, prices and costs could vary depending on the local markets, the different brands purchased, and the different sales packages chosen.221 These individualized facts made it impossible for plaintiffs to prove amount of injury on a classwide basis and rendered the case “unmanageable as a class action.”222 Dry Cleaning & Laundry Institute of Detroit, Inc. v. Flom’s Corp.,223 another pre-Comcast decision, also found individualized negotiations created an impediment to plaintiffs’ proof of classwide damages. In Dry Cleaning, Michigan dry cleaners and a trade association alleged Sherman Act price-fixing violations against dry cleaning and laundry supply companies. The court cited the same problems that doomed class certification in Mekani and Piggly Wiggly: In the instant action, thousands of transactions were involved over many years; each transaction was different; various plaintiffs may have purchased the dry cleaning supplies, which varies by brand and type, at different prices and varying quantities, in different ways under different credit terms. No formula proposed would be even-handed among class members or fair to defendants.224
Do Individual Plaintiffs Differ in Ways That Make Proof of Damages Highly Individualized? Even if plaintiffs do not engage in individualized negotiations, there may be important differences between the plaintiffs themselves that prevent mathematical or formulaic calculation of damages. Bell Atlantic Corp. v. AT&T Corp. provides one example of these types of differences.225 In Bell Atlantic, discussed above, plaintiffs alleged AT&T attempted to monopolize the market for caller-ID service by blocking the
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 49
free passage of caller-ID data over its long-distance network.226 Plaintiffs sought to certify two classes: a “reverse charge” class comprised of businesses that purchased AT&T reversed billed long-distance service (e.g., “800” numbers) and caller-ID service and a “call recipient” class of businesses and organizations that were actual or potential purchasers of callerID service for long-distance calls.227 The basic allegations were that but for AT&T’s monopolization of caller ID, plaintiffs could have avoided answering certain calls and spent less time on answered calls through the use of caller ID. Plaintiffs claimed damages due to wasted employee time on phone calls and greater long-distance charges assessed on calls that would otherwise not have been answered.228 The court found class treatment is not appropriate in cases “where the calculation of damages is not susceptible to a mathematical or formulaic calculation.”229 Class members ranged in size from primarily local sole proprietorships to large interstate corporations with call centers, used different types of equipment, and had different volumes of repeat business, which would, in turn, change how effective caller ID would be for call screening.230 These individualized differences demonstrate the damages are “not susceptible to a mathematical or formulaic calculation,” and therefore individualized proof overwhelms the common questions.231 The Supreme Court appears to agree that differences between individual plaintiffs can defeat common proof of damages. In Comcast, the court cautioned against certification where “permutations” of harm amongst plaintiffs “are nearly endless.”232 Further, both GVR’d cases arguably contain such endless permutations of harm amongst plaintiffs. Key considerations on remand in those cases should include how plaintiffs differed in their use of washing machines and the differences among the 21 types and 9 model years of allegedly defective machines.233 Although the defendants in those cases do not argue plaintiffs negotiated differently, they can allege that the plaintiffs differ in ways which overwhelm common proof of damages. Do Plaintiffs Propose a Damages Formula Clearly Inadequate to Calculate Individual Damages? Alternatively, defendants can attack plaintiffs’ methodology for failing to adequately calculate individual damages. As discussed above,234 the use of averages can be a fatal flaw in a plaintiff’s attempt to prove a common impact. While Defendants should challenge Plaintiffs’ methodologies that rely on averaging under Daubert, they can also make damages-specific challenges outside of the Daubert context. Averages can provide even greater problems in trying to prove damages across an entire class. For example in Bell Atlantic, the court stated, “[c]lass treatment is
50
JAMES KEYTE ET AL.
not appropriate in cases … where the formula by which the parties propose to calculate individual damages is clearly inadequate.”235 The plaintiffs’ method relied primarily on averaging. Damages were calculated based on “average number of seconds saved per call [both long-distance and local] through the use of caller ID, an average wage rate for the typical employee answering and processing telephone calls, and the total number of AT&T calls to class members made during the class period.” The formula was adjusted for the reverse charge class “using AT&T’s billing records, to include recovery of any long-distance charges assessed against class members that might have otherwise been avoided through the use of caller ID.”236 The court noted that plaintiffs’ averaging formula “ma[de] no effort to adjust for the variegated nature of businesses included in the classes” and thus could not reasonably approximate actual damages suffered by class members.237 Dry Cleaning, also discussed above, addressed similar methodological problems. In that case, plaintiffs attempted to supply a method for calculating damages that would overcome the problems of common proof caused by individualized negotiations. The method relied on “benchmarking,” which the court rejected, correctly noting that this method failed to account for the variation in markets, products, negotiating power and buyer size.238 As a result, plaintiffs could not establish that common questions predominated over individualized inquiry on damages.239 Finally, Reed v. Advocate Health Care faulted the plaintiffs’ expert for relying on averages.240 In Reed, registered nurses (“RNs”) alleged health care organizations conspired to depress wages and exchange compensation information. Although plaintiffs proposed a formula to calculate damages, the court found the formula “unacceptably masks the significant variation in RN base wages during the Class Period.” The court sided with defendants’ expert, who noted: [T]he relative movements of mere averages (means) do not prove common impact to individual RNs. For example, mean wages for Defendants’ RNs could move together even though particular Defendants gave larger increases to certain, hard to find nurses, and smaller increases to others. The issue is the feasibility of common proof regarding individual nurses, not a hypothetical ‘average’ nurse.241
CONCLUSION Comcast confirmed that there is a new rigor in Rule 23(b)(3) class actions, and it is not limited to issues of impact. And, in the wake of Comcast, both
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 51
practitioners and economists must adjust their tactics and proof to this new environment, which itself will evolve fairly rapidly as lower courts continue to grapple with the requirements that the Supreme Court has imposed.
NOTES 1. In re Hydrogen Peroxide Antitrust Litig., 552 F.3d 305, 318 (3d Cir. 2008). 2. See, for example, Reed v. Advocate Health Care, 268 F.R.D. 573 (N.D. Ill. 2009); In re Plastics Additives, 03-CV-2038, 2010 WL 3431837 (E.D. Pa. August 31, 2010); In re Intel Corp. Microprocessor Antitrust Litig., MDL 05-1717 JJF, 2010 WL 8591815 (D. Del. July 28, 2010). See also Thompson, Bloch, and GuerinCalvert (2010) (“In our view, Hydrogen Peroxide has done for class certification what Twombly and Iqbal did for the pleading standards plaintiffs must satisfy when faced with a motion to dismiss, and what Daubert did with respect to the standards for evaluating expert evidence.”). 3. Daubert v. Merrell Dow Pharm., 509 U.S. 579 (1993). 4. See, for example, Piggly Wiggly Clarksville, Inc. v. Interstate Brands Corp., 215 F.R.D. 523 (E.D. Tex. 2003), aff’d, 100 F. App’x 296 (5th Cir. 2004). 5. In Story Parchment, for example, the Court recognized that “although damages may not be determined by mere speculation or guess,” the damage calculations need not be exact. Story Parchment Co. v. Paterson Parchment Paper Co., 282 U.S. 555, 563 (1931). Instead, evidence that supports a “just and reasonable inference” will be accepted even if the result is “only approximate.” Ibid. See also Olden v. LaFarge Corp., 383 F.3d 495, 508 (6th Cir. 2004) (affirming class certification where liability could be determined as to the entire class but individual issues surfaced during the determination of damages). 6. Comcast Corp. v. Behrend, 133 S. Ct. 1426 (2013). 7. See infra section IV.B.1(a). 8. Ibid. 9. See Comcast, 133 S. Ct. at 1432. 10. See Bogosian v. Gulf Oil Corp., 561 F.2d 434, 454 (3d Cir. 1977). 11. See ibid. at 1433. 12. Ibid. at 1432 (quoting Amchem Prods., Inc. v. Windsor, 521 U.S. 591, 615, 617 (1997)). 13. In Comcast, the Court held that respondents could not bring a Daubert challenge to expert testimony after the class certification hearing occurred. By refusing to allow respondents to do so, the Court hinted that parties must raise, and courts must consider, Daubert issues at the class certification stage. Although Dukes also did not hold that trial courts must apply Daubert scrutiny at the class certification stage, the Court suggested it much more explicitly than it did in Comcast: “The parties dispute whether Bielby’s testimony even met the standards for admission of expert testimony … The District Court concluded that Daubert did not apply to expert testimony at the certification stage of class-action proceedings. We doubt that this is so, but even if properly considered, Bielby’s testimony does nothing to
52
JAMES KEYTE ET AL.
advance respondent’s case.” Wal-Mart Stores, Inc. v. Dukes, 113 S. Ct. 2541, 25532554 (2011). 14. 600 F.3d 813 (7th Cir. 2010). 15. Ibid. at 814. 16. Ibid. at 816. 17. Ibid. at 816817. 18. Ibid. 19. Ibid. 20. 419 F. App’x 887 (11th Cir. 2011). 21. Ibid. at 890. 22. Comcast, 133 S. Ct. 1426, 1431 n.4. 23. This was precisely the result in In re Lineboard Antitrust Litigation, a 2002 case decided by the Third Circuit before Hydrogen Peroxide. In re Lineboard Antitrust Litig. 305 F.3d 145, 154155 (3d Cir. 2002). There, the plaintiffs’ expert assured the court that she could rely on several accepted statistical or mathematical approaches to prove harm, including benchmarking, the yardstick approach, and simply comparing prices during the conspiracy period to prices during the nonconspiracy period. Ibid. Accepting the expert’s methods at face value methods that have been rejected by courts after Hydrogen Peroxide the court held that plaintiffs eventually could establish injury on a classwide basis. Ibid. at 155. 24. There are indications that, at least in the past, an unsuccessful Daubert challenge at the class certification stage can reduce the likelihood of successfully challenging that expert at a later phase. Accordingly, there might be some instances where a Daubert challenge at the class certification stage could be counterproductive for defendants. See James Langenfeld & Christopher Alexander, Daubert and other Gatekeeping Challenges of Antitrust Experts, Antitrust, Vol. 25, pp. 21, 24, 25 (2011). 25. The “Philadelphia cluster,” the relevant cluster in Comcast, was made up of 16 counties located in Pennsylvania, Delaware and New Jersey. Comcast, 133 S. Ct. at 1430. 26. The four theories were that (1) clustering acquisitions prevented or made it difficult for customers to compare prices; (2) direct broadcast satellite companies, one set of potential competitors, found it more difficult to gain rights to local sports content and decided not to enter the Philadelphia market as a result; (3) Comcast’s ability to obtain programming material at lower prices allowed it to raise prices; and (4) the “overbuilder deterrence theory” discussed above. Comcast, 133 S. Ct. at 1439. 27. In short, the model presented by plaintiff’s expert created a “but-for” baseline price a price the market would have settled on absent the effects of all four alleged theories of liability. 28. See infra section IV.B.1. 29. Federal Rule of Civil Procedure 23(c)(1)(A) (“Time to Issue. At an early practicable time after a person sues or is sued as a class representative, the court must determine by order whether to certify the action as a class action.”). See also Hydrogen Peroxide, 552 F.3d 305, 318 (3d Cir. 2008) (discussing 2003 Amendments to Rule 23(c)(1)(A) and their implications for the timing of class certification). 30. Compare M. Sean Royall, Disaggregation of Antitrust Damages, 65 Antitrust L.J. 311 (1997) with Conwood Co., L.P. v. U.S. Tobacco Co., 209 F.3d 768 (6th Cir. 2002). 31. See Comcast, 133 S. Ct. at 14401441.
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 53 32. American Honda Motor Co. v. Allen, 600 F.3d 813, 818 (7th Cir. 2010). 33. Ibid. 34. Despite the expert’s report being published in a journal for forensic engineers who testify as motorcycle experts, there is no indication that the standard itself was accepted by anyone other than the expert. In addition, the published article acknowledged there was no standard articulated by the government, the industry, or the Society of Automotive Engineers that determined acceptable response characteristics for motorcycles once they go into “wobble mode.” 35. Plaintiff’s expert had never conducted any “rider confidence studies” to determine when or how motorcycle riders perceive wobble nor did he perform a test to determine the minimum wobble amplitude at which a rider would detect oscillation. 36. The expert’s study relied on a test of a single motorcycle ridden by a single test rider and then extrapolated those findings to an entire fleet of motorcycles produced from 2001 to 2008. 37. Ibid. at 818819. 38. 400 F.3d 562 (8th Cir. 2005). 39. Ibid. at 569570. 40. See ibid. at 569571. 41. See In re Scrap Metal Antitrust Litig., 527 F.3d 517, 535 (6th Cir. 2008) (“Thus, even when there are individual variations in damages, the requirements of Rule 23(b)(3) are satisfied if the plaintiffs can establish that defendants conspired to interfere with the free-market pricing structure.”). 42. In re Hydrogen Peroxide Antitrust Litig., 552 F.3d 305, 313 (3d Cir. 2008) (rejecting this type of simplistic analysis of market pricing). 43. No. 03-CV-2038, 2010 WL 3431837 (E.D. Pa. August 31, 2010). 44. Ibid. at *13. 45. Ibid. 46. Ibid. at *14. 47. Ibid. 48. Ibid. 49. No. 09-23187-CIV, 2012 WL 27668 (S.D. Fla. January 3, 2012). 50. Ibid. at *10. 51. Ibid. 52. See, for example, Burtis and Neher (2011). 53. Johnson and Leonard (2008); see also Fisher (1986) (describing correlation analysis as “pseudoscience” and “poor statistical and econometric testimony masquerading as serious science”). 54. Reed v. Advocate Health Care, 268 F.R.D. 573, 591 (N.D. Ill. 2009) (quoting ABA Section of Antitrust Law, Econometrics: Legal, Practical, and Technical Issues 220 (2005)); see also In re Graphics Processing Units Antitrust Litig., 253 F.R. D. 478, 493 (N.D. Cal. 2008) (hereinafter “In re GPU”) (rejecting “correlation analyses” based on averages rather than specific “prices paid by individual consumers” and finding that plaintiffs’ expert “evaded the very burden that he was supposed to shoulder”); Forister and Hussain (2010) (“Because not all prices exactly follow the trend of the ‘average’ price, this aggregation or pooling of data precludes one from determining ways in which individual price series are related and how the alleged price-fixing agreement would affect different prices.”). 55. 339 F.3d 294 (5th Cir. 2003).
54
JAMES KEYTE ET AL.
56. Ibid. at 299300. 57. Ibid. at 300. According to plaintiffs, the business benefits of caller ID are screening out unwanted calls (and thereby reducing long-distance expenses), returning calls hours later (even in cases where the caller does not leave a message), tracking call volume (because some units can record information for later recall), and faster and more efficient customer service. 58. Ibid. at 300. 59. Ibid. 60. Ibid. at 307. 61. Ibid. at 306, 307. 62. Ibid. 63. Ibid. 64. Reed v. Advocate Health Care, 268 F.R.D. 573, 577578 (N.D. Ill. 2009). 65. See ibid. at 582584. 66. Ibid. at 589. 67. Ibid. at 591. 68. Ibid. at 592. 69. Ibid. 70. Ibid. at 593. 71. Ibid. at 594. 72. In re Wholesale Grocery Prod. Antitrust Litig., No. 09MD 2090 ADM/AJB, 2012 WL 3031085 (D. Minn. July 25, 2012). 73. Ibid. at *10. 74. Ibid. 75. Ibid. at *13. 76. Ibid. 77. Ibid. at *10. 78. 863 F. Supp. 2d 966, 972 (C.D. Cal. 2012). 79. See ibid. at 974. 80. Ibid. at 975. 81. No. 09-ML-2048-C, 2011 WL 6826813 at *1 (W.D. Okla. December 28, 2011). 82. Ibid. at *13. 83. Ibid. at *15. The court acknowledged that the test was “promulgated by serious economists who are experts in their field, peer reviewed, published and discussed in economic literature, and previously used in other litigation.” Thus, the court determined that whether the GRS test passes Daubert muster “is not easily resolved.” Ibid. 84. Ibid. 85. See ibid. at *1516. 86. See, for example, Blades v. Monsanto Co., 400 F.3d 562, 569 (8th Cir. 2005) (deferring ruling on Daubert motion in order to consider Plaintiffs’ arguments on class certification); Bacon v. Honda of Am. Mfg., Inc., 205 F.R.D. 466, 471 (S.D. Ohio 2001), aff’d, 370 F.3d 565 (6th Cir. 2004) (denying Daubert motion in order to consider class certification motion); In re Visa Check/MasterMoney Antitrust Litig., 192 F.R.D. 68, 77 (E.D.N.Y. 2000) (discussing preliminary role of Daubert inquiry in class certification analysis) aff’d, 280 F.3d 124 (2d Cir. 2001).
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 55 87. See, for example, Amchem Prods., Inc. v. Windsor, 521 U.S. 591, 625 (1997) (predominance is “a test readily met in certain cases alleging … violations of the antitrust laws.”). 88. In re Hydrogen Peroxide Antitrust Litig., 552 F.3d 305, 313 (3d Cir. 2008) (“Private damage claims by numerous individuals arising out of concerted antitrust violations may or may not involve predominating common questions.” (citing Fed. R. Civ. P. 23(b)(3) advisory committee’s note, 1966 Amendment)); see also Robinson v. Tex. Auto. Dealers Ass’n, 387 F.3d 416, 420421 (5th Cir. 2004) (“There are no hard and fast rules … regarding the suitability of a particular type of antitrust case for class action treatment.”). 89. See Wal-Mart Stores, Inc. v. Dukes, 113 S. Ct. 2541, 2551 (2011); see also Gen. Tel. Co. of Sw. v. Falcon, 457 U.S. 147, 156 (1982). 90. Dukes, 131 S. Ct. at 255; see also Rutstein v. Avis Rent-A-Car Sys., Inc., 211 F.3d 1228, 1234 (11th Cir. 2000) (collecting cases that considered merits when deciding a motion for class certification); Coopers & Lybrand v. Livesay, 437 U.S. 463, 467 & n.12 (1978) (“[T]he class determination generally involves considerations that are ‘enmeshed in the factual and legal issues comprising the plaintiff’s cause of action’ … ‘The more complex determinations required in Rule 23(b)(3) class actions entail even greater entanglement with the merits’.” (citation omitted)). Notably, for class certification, the burden of proof rests with plaintiffs to put forth common proof of impact by a preponderance of the evidence. See Hydrogen Peroxide, 552 F.3d at 320; see also In re Checking Account Overdraft Litig., Nos. 1:09-MD-02036-JLK et al., 2011 WL 3158998, at *2 (S.D. Fla. July 25, 2011) (to be published in F.R.D.) (“A district court may certify a class only if, after ‘rigorous analysis’, it determines that the party seeking certification has met its burden of a preponderance of the evidence.”). 91. See, for example, Blades v. Monsanto Co., 400 F.3d 562, 569570 (8th Cir. 2005) (court denied class certification where it “considered all expert testimony offered by both sides” and “afforded that testimony such weight as [the Court] deemed appropriate”); In re New Motor Vehicles Can. Export Antitrust Litig., 522 F.3d 6, 17 (1st Cir. 2008) (“It would be contrary to the ‘rigorous analysis of the prerequisites established by Rule 23 before certifying a class’ to put blinders on as to an issue simply because it implicates the merits of the case.”); West v. Prudential Sec., Inc., 282 F.3d 935, 938 (7th Cir. 2002) (same); Unger v. Amedisys Inc., 401 F.3d 316 (5th Cir. 2005) (same); In re Evanston N.W. Healthcare Corp. Antitrust Litig., 268 F.R.D. 56 (N.D. Ill. 2010) (citing Hydrogen Peroxide, 552 F.3d at 324) (“[r]esolving expert disputes to determine whether a class certification requirement has been met is always a task for the courtno matter whether a dispute might appear to implicate the ‘credibility’ of one or more experts, a matter resembling those usually reserved for a trier of fact”). Note, however, that before Comcast, a minority of circuits held that it is inappropriate to resolve merits issues related to expert testimony and would certify classes on the basis of plaintiffs’ expert opinions so long as the experts’ methodologies were not “worthless” or “inherently faulty.” See In re Polypropylene Carpet Antitrust Litig., 996 F. Supp. 18, 25, 29 (N.D. Ga. 1997); Drayton v. W. Auto Supply Co., No. 0110415, 2002 WL 32508918, at *6 n.13 (11th Cir. Mar. 11, 2002) (affirming the district court’s refusal to conduct a Daubert analysis or to resolve a
56
JAMES KEYTE ET AL.
“battle of the experts” and finding that expert testimony proffered was sufficient at class certification stage to demonstrate that plaintiffs could show “common” proof of impact); In re Commercial Tissue Prods., 183 F.R.D. 589, 596 (N.D. Fla. 1998) (“This case presents the familiar ‘battle of the experts’. The certification stage of this litigation is not, however, the proper forum in which to resolve this battle.”). 92. See, for example, In re Domestic Air Transp. Antitrust Litig., 137 F.R.D. 677, 689 (N.D. Ga. 1991) (inflated fares resulted in artificial “base” price which became benchmark for discounted or negotiated fares); J.B.D.L. Corp. v. Wyeth-Ayerst Labs., Inc., 225 F.R.D. 208, 217 (S.D. Ohio 2003) (finding that price negotiations with each customer do not undermine plaintiffs’ arguments related to common impact because conspiracy allegedly raised the starting price for negotiations); In re Flat Glass Antitrust Litig., 191 F.R.D. 472, 486 (W.D. Pa. 1999) (finding that “even though some plaintiffs negotiated prices, if plaintiffs can establish that the base price from which these negotiations occurred was inflated, this would establish at least the fact of damage, even if the extent of the damage by each plaintiff varied”); see also In re Indus. Diamonds Antitrust Litig., 167 F.R.D. 374, 383 (S.D.N.Y. 1996) (collecting cases); In re Catfish Antitrust Litig., 826 F. Supp. 1019 (N.D. Miss. 1993); Hedges Enters., Inc. v. Cont’l Grp. Inc., 81 F.R.D. 461, 475 (E.D. Pa. 1979). 93. See, for example, Kenett Corp. v. Mass. Furniture & Piano Movers Ass’n, Inc., 101 F.R.D. 313, 314, 316 (D. Mass. 1984) (finding that although moving service established standard fee schedules, individual questions predominated because the bundle of moving services provided differed from mover to mover and customer to customer “in light of the costs facing the company at the time, its competitive market, its reputation, and the particular services required for the move”); Am. Custom Homes v. Detroit Lumberman’s Ass’n, 91 F.R.D. 548, 550 (E.D. Mich. 1981) (finding that although defendant provided prices lists, individual questions predominated because actual purchases of defendant’s product involved individualized negotiations that varied based on produce design, credit terms, negotiating power, and specific purchase arrangements); In re Beef Indus. Antitrust Litig., 19862 Trade Cas. (CCH), 1986 WL 8890 at *2 (S.D. Tex. 1986) (finding that although defendant alleging part of cattle price was tied to “Yellow Sheet,” individual questions predominated because the price for cattle depended on a “multitude of varying factors differing from one area to another” including weather, minimum cattle requirements, price competition, and the beef by-product market). 94. In re Plastics Additives Antitrust Litig., No. 03-CV-2038, 2010 WL 3431837, at *13 (August 31, 2010). 95. Ibid. 96. Ibid. 97. No. 09-23187-CIV, 2012 WL 27668, at *9 (S.D. Fla. January 3, 2012). 98. Ibid. 99. 253 F.R.D. 478, 480 (N.D. Cal. 2008). 100. Ibid. 101. Ibid. (“First, cards and chips were sold to original equipment manufacturers (OEMs), such as Dell or Hewlett-Packard. The OEMs then installed the cards or chips in their own computers for later resale to individual consumers. Second, chips were sold to add-in-board manufacturers (AIBs) who in turn incorporated them into their own cards. These cards could then be sold as standalone products for a
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 57 retail price or to other computer manufacturers (e.g., OEMs) for incorporation into whole computers. Third, distributors could purchase chips or cards, which they in turn sold to other entities along the chain of distribution (i.e., AIBs, OEMs, or other distributors). Fourth, retailers, such as Best Buy, could purchase graphics cards for later sale to individual consumers. Fifth, original design manufacturers (ODMs) purchased graphics cards or chips that would be incorporated into parts that would later be branded by another firm along the chain of distribution for sale. Sixth, ATI sold graphics cards online through its website ATI.com directly to individual consumers.”). 102. Ibid. at 490491. 103. Nichols v. Mobile Cnty. Bd. of Realtors, No. 76-619-P, 1980 WL 1975, at *7 (S.D. Ala. May 16, 1980) (collecting and comparing antitrust cases in which products were homogeneous or distinct); In re Infant Formula Antitrust Litig., No. MDL 878, 1992 WL 503465, at *5 (N.D. Fla. January 13, 1992) (“Contentions of infinite diversity of product, marketing practices, and pricing have been made in numerous cases and rejected. Courts have consistently found the conspiracy issue the overriding, predominant question”). 104. Nichols, 1980 WL 1975 at *7. 105. In re Plastics Additives Antitrust Litig., No. 03-CV-2038, 2010 WL 3431837, at *8 (E.D. Pa. August 31, 2010). 106. Ibid. 107. In re Fla. Cement & Concrete Antitrust Litig., 09-23187-CIV, 2012 WL 27668 (S.D. Fla. January 3, 2012). 108. Ibid. at 13. 109. Ibid. 110. Ibid. 111. Another argument defendants can make is that there are variations across the geographic regions included in the plaintiffs’ proposed class. Although plaintiffs generally do not need to define a relevant geographic market in a per se price-fixing case, issues related to geographic market may be relevant where defendants show that localized markets and local pricing exist because these markets may raise individual questions about impact and damages. See Behrend v. Comcast, No. 10-2865, 2011 WL 3678805, at *78 (E.D. Pa. August 24, 2011) (citing Areeda & Hovenkamp, 19982007 & supp. 2008, ¶398b); see also Blades v. Monsanto Co., 400 F.3d 562 (8th Cir. 2005) (denying class certification because of the localized nature of the markets for genetically modified corn and soybean seeds that exhibited substantial price variation); Behrend, 2011 WL 3678805, at *78 (Jordan, J., concurring) (finding that geographic market is a relevant inquiry in a per se case where “the class definition includes a geographic component”). 112. Comcast Corp. v. Behrend, 133 S. Ct. 1426, 1434 (2013). 113. Ibid. 114. See Blades, 400 F.3d at 562 (“Simply put, plaintiffs presume class-wide impact without any consideration of whether the markets or the alleged conspiracy at issue here actually operated in such a manner as to justify that presumption. Dr. Leitzinger assumes the answer to this critical question …”). 115. Ibid.
58
JAMES KEYTE ET AL.
116. See Comcast, 133 S. Ct. at 1435. 117. Ibid. 118. See ibid. 119. In re Graphics Processing Units Antitrust Litig, 253 F.R.D. 478, 508 (N.D. Cal. 2008). 120. Ibid. at 489491. 121. Ibid. at 490; see also In re Indus. Diamonds Antitrust Litig., 167 F.R.D. 374, 383385 (S.D.N.Y. 1996) (“[W]e are persuaded that despite the wide range of products and prices involved, common proof of impact is possible on behalf of purchasers who bought list-price products. Common proof of impact is not possible, however, on behalf of those purchasers who bought non-list price products.”). 122. See generally Comcast Corp. v. Behrend, 133 S. Ct. 1426 (2013). 123. Ibid. at 1432. 124. Ibid. (citation omitted). 125. Ibid. at 14321433. 126. Ibid. at 1433. 127. Ibid. at 1435 n.6. This Article does not contend that plaintiffs must demonstrate that they suffered the same amount of harm to establish that common questions predominate after Comcast. The Court’s indication that plaintiffs must offer proof of damages that establishes the “requisite level of commonality” does, however, underscore the Court’s rigorous analysis of common proof of the amount of damages. 128. Ibid. at 1433. 129. See ibid. at 1434 (discussing how plaintiffs’ “damages” may not have been tied to actual harm in Philadelphia and thus it was possible that many individuals in the Philadelphia DMA did not suffer actionable harm). 130. See ibid. at 14341435. 131. See ibid. at 1433. 132. Ibid. at 1436 (citation omitted). 133. Ibid. 134. Ibid. 135. Ibid. at 1437. The dissenters noted that, instead of refusing to certify a class action, courts can divide classes into subclasses, certify classes for liability purposes only, or alter or amend cases as they develop. Ibid. at 1437 n.*. 136. 678 F.3d 409 (6th Cir. 2012) (GVR’d April 1, 2013). 137. 702 F.3d 359 (7th Cir. 2012) (GVR’d June 3, 2013). 138. In the third, Ross v. RBS Citizens, N.A., 667 F.3d 900 (6th Cir. 2013) (GVR’d April 1, 2013), Defendant RBS Citizens appealed solely on the ground that Plaintiffs’ proposed class did not comply with Rule 23(c)(1)(B). As such, the opinion does not discuss damages or Rule 23(b)(3). 139. Whirlpool, 678 F.3d at 415. 140. Ibid. at 412. 141. Ibid. at 415. 142. Ibid. at 419 (citation omitted). 143. Ibid. at 421. 144. Butler, 702 F.3d at 360361. The defects all concerned Whirlpoolmanufactured machines, but Sears was the defendant because it could still be
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 59 strictly liable for breach of warranty, with a possible contribution or indemnification claim against Whirlpool. See ibid. at 363. 145. Ibid. at 362. 146. Ibid. at 363. 147. Ibid. at 362, 363. 148. Ibid. at 363. 149. Ibid. at 362. 150. Beisner, Miller, and Wyatt (2013). 151. McLaughlin (2013) at 2. 152. Ibid. 153. Mass Tort Defense (2013). 154. In re Whirlpool Corp. Front-Loading Washer Products Liab. Litig., 722 F.3d 838, 860 (6th Cir. 2013). 155. See ibid. at 855856 (relying on an unsupported state law analysis that would allow class members to recover damages even when the defect had not manifested for plaintiffs). 156. Wal-Mart Stores, Inc. v. Dukes, 131 S. Ct. 2541, 2550 (2011). 157. Butler v. Sears, Roebuck & Co., 11-8029, 2013 WL 4478200 at *4 (7th Cir. August 22, 2013). 158. Ibid. at *5. 159. Ibid. 160. In re Rail Freight Fuel Surcharge Antitrust Litig.-MDL No. 1869, 12-7085, 2013 WL 4038561 at *8 (D.C. Cir. August 9, 2013). 161. Ibid. at *1. Fuel surcharges are added on top of the base rate railroads charge customers for freight. Ibid. Historically, fuel surcharges have taken two forms: “mileage-based fuel surcharges,” which raise total freight rates in proportion to shipping distances, and “rate-based fuel surcharges,” which impose a surcharge as a function of the base rate once fuel prices exceed a “strike” or “trigger” price. Ibid. In 20032004, all four major railroads had fully switched to rate-based surcharges, and all four lowered the trigger prices. Ibid. The Surface Transportation Board ended the practice, expressing concerns that, “the rationale for the fuel surcharges fuel cost and recovery” was disconnected from the base rates on which the surcharges were based. Ibid. at *2. Following decision, plaintiffs here alleged the defendants engaged in price fixing under §1 of the Sherman Act. Ibid. 162. Ibid. at *2. 163. Ibid. at *3. 164. Ibid. 165. Ibid. (quoting In re Rail Freight Fuel Surcharge Antitrust Litig. (Fuel Surcharge II), 287 F.R.D. 1, 67 (D.D.C. 2012). 166. Ibid. at *5. 167. Ibid. at *1. 168. Ibid. 169. Ibid. at *5. 170. Ibid. at *6 (internal citations omitted). 171. Ibid. at *7. 172. Ibid.
60
JAMES KEYTE ET AL.
173. Ibid. (citing Comcast, 133 S. Ct. at 1433). 174. To date, Rail Freight is the most defendant-friendly antitrust case applying post-Comcast rigor to proof of damages in class certification cases. There are also a number of precedents in other areas of the law discussed below, which involve rigorous damages analysis and which we expect will inform antitrust class certification analysis. 175. No. 3:10-CV-0591, 2013 WL 1316452 (N.D.N.Y. March 29, 2013). 176. Ibid. at *3. 177. Ibid. 178. No. 5:09-323-KKC, 2013 WL 2285163 (E.D. Ky. May 22, 2013). 179. Ibid. at *1. 180. Ibid. at *6 (citation omitted). 181. Ibid. 182. Bankruptcy No. 7-04-17866-TL, 2013 WL 2244216 (D.N.M. May 21, 2013). 183. Ibid. at *7. 184. Ibid. See also Smith v. Family Video Movie Club, Inc., No. 1 CV 1773, 2013 WL 1628176, at *5 (N.D. Ill. April 15, 2013) (“[After Comcast] damages must be susceptible to measurement across the entire class, and individual damage calculations cannot overwhelm questions common to the class.”); Phillips v. Asset Acceptance, LLC, No. 09 C 7993, 2013 WL 1568092, at *3 (N.D. Ill. April 12, 2013) (“[Comcast] may portend a tightening of class certification standards more generally, particularly as to the circumstances under which the task of measuring damages sustained by absent members destroys predominance under Rule 23(b)(3).”). 185. See Munoz v. PHH Corp., 2013 WL 2146925, at *24 (E.D. Ca. 2013) (“The Comcast decision does not infringe on the long-standing principle that individual class member damage calculations are permissible in a certified class under Rule 23(b)(3).”); In re High-Tech Employee Antitrust Litig., 2013 WL 1352016 (N.D. Cal. 2013) (quoting the Comcast dissent for the principle that “individual damage calculations do not preclude class certification under Rule 23(b)(3)” and finding that Plaintiffs satisfied their burden on damage calculations by providing a method of calculating damages consistent with their theory of liability); In re Diamond Foods, Inc., Sec. Litig., 2013 WL 1891382 (N.D. Ca. 2013) (withholding judgment on whether Comcast requires class certification to be denied absent affirmative evidence that damages are susceptible of measurement across the entire class); Harris v. comScore, Inc., 2013 WL 1339262 (N.D. Ill. 2013) (finding that “factual damages issues do not provide a reason to deny class certification when the harm to each plaintiff is too small to justify resolving the suits individually.”). In Harris, however, the Court cited Butler v. Sears, Roebuck & Co., 702 F.3d 359, 362 (7th Cir. 2012), for the principle that efficiency concerns trump factual damages questions. Butler, as discussed in section IV.B.1.a, supra, was vacated and remanded by the Supreme Court on June 3, 2013, two months after the Harris opinion. It is unclear if the Court would have resolved the issue the same way if it had known the Supreme Court would remand Butler. Compare Wang v. Hearst Corp., 2012 WL 1903787, at *9 (S.D.N.Y. 2013) (“Indeed, although one could certainly quibble with the significance of a grant, vacate, and remand (‘GVR’) order from the Supreme Court, one must pause at least for a moment when one sees that the Supreme Court,
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 61 ‘in light of Comcast’, has issued an order vacating and remanding a Seventh Circuit’s decision affirming the district court’s certification in an overtime misclassification case.” (quoting RBS Citizens, N.A. v. Ross, ——— U.S. ———, 133 S. Ct. 1722, ——— L.Ed.2d ——— (2013)). 186. 2013 WL 2306567 (9th Cir. 2013). 187. Ibid. at *514. 188. Ibid. 189. CIV.A. 11-11313-DPW, 2013 WL 1320454 (D. Mass. March 28, 2013). 190. Ibid. at *8 n.3 (citing Newton v. Merrill Lynch, Pierce, Fenner & Smith, Inc., 259 F.3d 154, 187 (3d Cir. 2001)). See also Broussard v. Meineke Discount Muffler Shops, Inc., 155 F.3d 331, 342343 (4th Cir. 1998); Windham v. Am. Brands, Inc., 565 F.2d 59, 68 (4th Cir. 1977). 191. Ibid. 192. Kaufman (2013). 193. Paley and Ryan (2013). 194. Ibid. 195. Ibid. 196. Ibid. 197. Mason, Little, and Yeroshalmi (2013). 198. No. 3:10-CV-0591, 2013 WL 1316452 at *3 (N.D.N.Y. March 29, 2013). 199. Ibid. 200. Ibid. 201. 146 F. App’x 783 (6th Cir. 2005). 202. Ibid. at 791. 203. Ibid. 204. Kazan and Vasquez (2013). 205. Piggly Wiggly Clarksville, Inc. v. Interstate Brands Corp., 100 F. App’x 296 (5th Cir. 2004). 206. Piggly Wiggly Clarksville, Inc. v. Interstate Brands Corp., 215 F.R.D. 523 (E.D. Tex. 2003). 207. Ibid. at 527. 208. Ibid. at 528. 209. Ibid. at 530. 210. See ibid. at 531. 211. Ibid. (citing Windham v. Am. Brands, Inc., 565 F.2d 59, 6768 (4th Cir. 1977); In re Beef Indus. Antitrust Litig., 1986 WL 8890, *1 (N.D. Tex. 1986); Dry Cleaning & Laundry Inst. of Detroit, Inc. v. Flom’s Corp., 1993 WL 527928, *3 (E.D. Mich. 1993)). 212. On this point, see also section IV.C.1, supra. 213. Ibid. (quoting Castano, 84 F.3d at 745 n.19) (citation omitted). 214. Ibid. 215. Piggly Wiggly Clarksville, Inc. v. Interstate Brands Corp., 100 F. App’x 296 (5th Cir. 2004). 216. Ibid. at 297 (citations omitted). 217. 93 F.R.D. 506 (E.D. Mich. 1982). 218. Ibid. at 507. 219. Ibid. at 509.
62
JAMES KEYTE ET AL.
220. Ibid. at 511. 221. Ibid. 222. Ibid. 223. 91-CV-76072-DT, 1993 WL 527928 (E.D. Mich. October 19, 1993) (hereinafter “Dry Cleaning”). 224. Ibid. at *4. 225. 339 F.3d 294 (5th Cir. 2003). For discussion of Bell Atlantic in the Daubert context, see section II.B.4(b), supra. 226. Ibid. at 297. 227. Ibid. at 299300. 228. Ibid. at 300. 229. Ibid. at 307. 230. Ibid. at 306, 307. 231. Ibid. at 307. 232. Comcast, 133 S. Ct. 1426, 14341435 (2013). 233. See generally In re Whirlpool Corp. Front-Loading Washer Products Liability Litigation, 678 F.3d 409 (6th Cir. 2012); Butler v. Sears Roebuck & Co., 702 F.3d 359 (7th Cir. 2012). 234. See section II.B.4(b), supra. 235. Bell Atl. Corp. v. AT&T Corp., 339 F.3d 294, 307 (5th Cir. 2003). 236. Ibid. at 300. 237. Ibid. at 307. 238. Dry Cleaning & Laundry Inst. of Detroit, Inc. v. Flom’s Corp., 91-CV-76072DT, 1993 WL 527928, at *5 (E.D. Mich. October 19, 1993). 239. Ibid. at *6. 240. 268 F.R.D. 573, 590 (N.D. Ill. 2009). For a discussion of Advocate Health Care in the Daubert context, see section II.B.4(B), supra. 241. Ibid. at 592 (citation omitted).
ACKNOWLEDGMENTS The authors would like to thank Marissa Troiano and Mike Folger, associates in the group, for their invaluable assistance in preparing this article.
REFERENCES Areeda, P. E., & Hovenkamp, H. (19982007). Antitrust law: An analysis of antitrust principles and their application (3rd ed., supp. 2008 ¶398b). Beisner, J. H., Miller, J. D., & Wyatt, G. M. (2013). From cable TV to washing machines: The Supreme Court cracks down on class actions. Class action litigation Report, 14 Class 616. Burtis, M. M., & Neher, D. V. (2011). Correlation and regression analysis in antitrust class certification. Antitrust Law Journal, 77, 495, 511512.
From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions 63 Fisher, F. M. (1986). Statisticians, econometricians, and adversary proceedings. Journal of the American Statistical Association, 81, 277, 282285. Forister, E. F., & Hussain S. (2010). Empirical approaches in assessing class certification in direct purchaser price-fixing cases, The Antitrust Review of the Americas, 22. Johnson, J. H., & Leonard, G. K. (2008). In the eye of the beholder: Price structure as junk science in antitrust class certification proceedings. Antitrust, 22, 108, 111. Kaufman, T. (2013, May 29). Ninth circuit rules that Comcast does not kill wage and hour class actions, Labor & Employment Blog. Retrieved from http://www.laboremploymentlawblog.com/wage-and-hour-ninth-circuit-rules-thatcomcast-does-not-kill-wage-andhour-class-actions.html Kazan, B. M., & Vasquez, G. Y. (2013). Viability of Rule 23(b)(3) Cases after “Dukes,” “Amgen” and “Comcast”: Court carves narrow path to class certification. New York Law Journal, June 10. Mason, C. M., Little, D. H., & Yeroshalmi, S. (2013). Supreme Court addresses problems of size: “Too big” and “too small” cases pose a struggle, New York Law Journal, June 10. Mass Tort Defense (2013, June 13). Supreme Court remand two class actions in light of Comcast. McLaughlin, J. M. (2013). Litigating the effect of “Comcast v. Behrend”, New York Law Journal, June 13. Retrieved from http://www.newyorklawjournal.com/PubArticleNY. jsp?id=1202603956620&Litigating_the_Effect_of_Comcast_v_Behrend&slreturn=20130 631142303 Paley, A., & Ryan, P. (2013). Ninth circuit reverses denial of class certification where determining damages is a purely mechanical exercise. Wage & Hour Litigation Blog, June 4. Retrieved from http://www.wagehourlitigation.com/rule-23-class-certification/ninthcircuit-reverses-denial-of-class-certification-wheredetermining-damages-is-a-purelymechanica-1/ Thompson, M. J. (Moderator), Bloch, R. E., & Guerin-Calvert, M. E. (2010). Anatomy of an information exchange healthcare antitrust case. AHLA Seminar Papers (Vol. 2010, Issue 1), Antitrust in Healthcare Conference, May 24.
THE CLASS CERT GAMES: COACH, COMMENTATOR, OR CRITIC? Comment on “From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions” Joshua P. Davis ABSTRACT This article responds to James Keyte, Paul Eckles, and Karen Lent’s article “From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions” (“The New Rigor”). It argues that The New Rigor offers valuable strategic advice to defense counsel and insight into defense counsel’s strategic thinking but is much less effective as an objective statement of the law or a normative argument for legal reform. In the parlance that I adopt, The New Rigor succeeds in the role of coach but much less so in the roles of commentator and critic. Keywords: Rigor; class actions; class certification; Daubert; common impact JEL classifications: K41; K21
The Law and Economics of Class Actions Research in Law and Economics, Volume 26, 6576 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0193-5895/doi:10.1108/S0193-589520140000026003
65
66
JOSHUA P. DAVIS
TABLE OF CONTENTS Coach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Commentator. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Critic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
66 68 71 73 76
How one gauges the success of James Keyte, Paul Eckles, and Karen Lent’s article, “From Hydrogen Peroxide to Comcast: The New Rigor in Antitrust Class Actions,” (hereinafter “The New Rigor”) depends in part on what it intends to be. Those intensions appear to be several. In the ongoing competition between those who seek to persuade courts to certify classes and those who oppose those efforts in the “class cert games,” if you will The New Rigor can be read as aspiring to play three roles: coach, commentator, or critic. For the purpose of this comment, I use these terms as follows: a coach offers advice in the spirit of enabling one side to succeed, not in a neutral spirit, although a coach may want to appear neutral when pleading to the referee; a commentator attempts to remain nonpartisan, not to criticize or to suggest reform; and a critic takes normative positions, departing from mere description or prediction by making suggestions for change in the law. The New Rigor at times attempts to adopt each of these roles. But there are tensions between them. The advocacy of the coach undermines the objectivity of the commentator, which in turn constrains the normativity of the critic. The New Rigor is more successful in the role of coach than in the roles of commentator or critic.
COACH The first role is coach. Coaches provide guidance about how to play games effectively. They also sometimes try to influence the referees, making public statements that display their teams in a favorable light and cast doubt on their opponents. The New Rigor often reads as if it were written as a form of coaching. Effective coaching. Informed coaching. Insightful coaching. But coaching nonetheless. And The New Rigor leaves little room for doubt as to which side the authors support. At numerous instances, it gives explicit advice about how to defend against proposed class actions. For
The Class Cert Games: Coach, Commentator, or Critic?
67
example, one of their headings says what “Defendants Should Attempt … ” to do to prevail on a motion under Daubert.1 Similarly, another explains what “Defendants … Should … Attack” to win under Daubert,2 and another suggests to defendants “Ways to Attack Damages” offered by plaintiffs to defeat class certification.3 Indeed, The New Rigor could be understood as a playbook for defense counsel, suggesting possible approaches to discovery, motion practice, and the like. In this regard, The New Rigor is quite successful. It displays an impressive knowledge of case law and practice and offers valuable research and advice to practitioners. Further, although it is written often quite explicitly for defense counsel, its utility is not so limited. After all, what team would not want to see the other side’s playbook? Plaintiffs’ attorneys would do well to take heed of The New Rigor, anticipating the kinds of arguments it suggests and contemplating possible responses. Interpreting The New Rigor as a guide to defense counsel, I am reluctant to offer any criticisms. Deference to the authors’ expertise seems appropriate. One point, however, may warrant consideration. As discussed in more detail below, the authors’ views are far from neutral. They offer an account of class certification and related law in the light most favorable to defendants. If the readers were to take at face value their interpretation of the law and their resulting recommendations that could give rise to an excessively aggressive stance in litigation. At class certification, that aggressiveness might mean making arguments that are too numerous and too attenuated. At some point, an advocate must prioritize. Otherwise, strong arguments can get lost among weak ones and a lawyer’s or an economist’s credibility can suffer. A related danger lurks before class certification, when defense counsel may be tempted to move to exclude under Daubert4 some or all of the expert testimony plaintiffs put forward. At first blush, Daubert motions may seem to offer defense counsel an additional opportunity to prevail with little cost. As the authors rightly recognize, a court can evaluate an expert under the relatively forgiving standard under Daubert where the issue is whether the testimony is merely admissible and separately at class certification where the issue becomes whether the expert testimony is persuasive.5 Two bites at the apple are often better than one, but not always. When courts rule on motions, they may seek to justify their decisions, both as a matter of psychology and explanation. A failed attack under Daubert may cause the court to become sympathetic to the plaintiffs’ experts and to make broad statements that later haunt defense counsel. A judge may not give defendants as large a second bite if they have already enjoyed a first one.6
68
JOSHUA P. DAVIS
But these criticisms are tentative and minor. At most, they apply in some contexts. No doubt The New Rigor provides many promising ideas for how defendants should argue their case. But it is much less likely to be successful at persuading judges about how best to interpret the law. Few readers could mistake its content for neutral description. As a result, to the extent The New Rigor aspires to influence the referees, it is not likely to be all that effective. For the same reason, a good coach is rarely an effective sports commentator.
COMMENTATOR Coaches do not generally make good commentators, at least not when their own teams are involved. They are rooting too hard to maintain objectivity. The New Rigor, alas, is no exception. This point should not be taken to suggest that none of the descriptive claims in The New Rigor is accurate. Some are quite plausible. It is surely true, for example, that the merits are no longer completely off limits at class certification as some courts had suggested in the past. As The New Rigor indicates, this issue was resolved by Dukes.7 The New Rigor is also right that courts are likely to apply Daubert at class certification. The New Rigor may overstate a bit the extent to which this proposition can be derived from the Supreme Court’s decisions in Comcast and Dukes.8 To be sure, it acknowledges that the Supreme Court has never held that Daubert applies at class certification, suggesting only that “it is reasonable to conclude that those decisions requiring or endorsing a ‘full’ Daubert screen prior to ruling on class certification are likely to emerge as the prevailing standard.”9 But Comcast ruled based on Federal Rule of Civil Procedure 23(b)(3), not Daubert,10 and, as the authors acknowledge, the Court in Dukes merely indicated in dicta its “doubt” that Daubert does not apply at class certification.11 An expression of leaning in one direction in dicta in an aside provides a pretty weak foundation for resolving a contested doctrinal issue, even if the aside appears in a Supreme Court opinion. Finally, it is also surely true that courts recently have been more demanding of plaintiffs at class certification than in years past, although there are some counterexamples (see Amgen). What is less clear is what we can extrapolate about the future from past decisions. The New Rigor appears to anticipate a continuing trend. The authors suggest that the “new rigor” of the title of their piece is indicative of “where [decisions] are likely
The Class Cert Games: Coach, Commentator, or Critic?
69
to come out in the future.”12 But it is also possible that we are witnessing a pendulum near the apex of its arc, one that will soon swing in the other direction. Only time will tell. Beyond these points, The New Rigor reads more like an aggressive advocacy piece than an objective assessment of the state of class certification doctrine or a prediction about its future shape. This is clear as a matter of both form and substance. Let’s begin with form. The New Rigor claims it will provide an “analytical framework” for assessing class certification under Rule 23(b)(3).13 But it does not. It proffers types of Rule 23(b)(3) expert analysis that should not make it past a Daubert screen,14 a list of reasons why courts should deny class certification,15 categories of evidence defendants should attempt to develop to prevail under Rule 23(b)(3),16 and a catalog of ways to attack damages evidence at class certification.17 In other words, instead of an analytical framework for assessing class certification, the authors offer what may be best understood as sets of strategies for defeating class actions. They never specify when a court should certify a class, only when it should not. That hardly suffices for a framework. It is a game plan. Similarly, The New Rigor does not provide an objective survey of various recent class certification decisions. Instead, it focuses on those that have been most sympathetic to defendants. It spends pages, for example, discussing the significance of the Supreme Court’s decision to vacate the Sixth Circuit’s decision in In re Whirlpool Corp. Front-Loading Water Products Liability Litigation,18 and the Seventh Circuit’s decision in Butler v. Sears, Roebuck and Co.19 approving class certification. It suggests those cases should have come out the other way under Comcast20 based on a detailed analysis of the facts and reasoning of cases reading Comcast broadly.21 In contrast, it does not explore the implications of the opinions of the Sixth and Seventh Circuits reaffirming class certification in Whirlpool and Butler22 on remand after Comcast and denigrates and dismisses opinions reading Comcast narrowly.23 Striking in this regard is its failure to address the Supreme Court’s recent decision in Amgen. A possible explanation is that Amgen did not involve an antitrust claim. But neither did Dukes, and Dukes receives ample attention.24 More generally, The New Rigor seems to discuss in depth only those cases ruling against class certification and praises only those cases.25 And it treats pre-Comcast cases as if they remain good only if they took a restrictive attitude toward certification.26 It is as if Comcast changed everything.
70
JOSHUA P. DAVIS
Which brings us to Comcast itself. The authors’ approach to the opinion is speculative at best, strained at worst, both as a matter of form and substance. As to form, the authors rely on unusual authorities for their broad reading of Comcast. They cite to various publications and blogs that share their perspective without citing to any that hold a contrary view.27 And they read a great deal into the Supreme Court’s decision to vacate and remand for reconsideration in light of Comcast.28 These are so-called grant, vacate, and remand (GVRs) orders from the Supreme Court granting a petition for certiorari, vacating a lower court opinion, and remanding it for further proceedings. I am not sure I have ever seen an analysis that infers so much from the Court’s decision to vacate a decision. How should we determine the significance of these GVRs? An obvious source of guidance might be the rulings of the federal appellate courts on remand. As noted above, in both of the relevant cases the Sixth and Seventh Circuits determined the classes they certified should remain certified after Comcast.29 These results may leave the authors baffled and frustrated. But they warrant significant weight in any objective description of the current state of the law. Similarly, an objective assessment would likely suggest that the case law interpreting Comcast is mixed and that the decision will have only a marginal effect on existing doctrine. But the authors are dismissive of the cases that read Comcast narrowly indicating that only “a few cases” have “purport[ed] to follow” the Comcast dissent.30 In reality, however, The New Rigor reports a D.C. Circuit opinion and various district court opinions that apparently see Comcast as changing class certification law significantly.31 Contrast that with three federal appellate court opinions Whirlpool in the Sixth Circuit, Butler in the Seventh Circuit, and Leyva v. Medline Industries, Inc.32 in the Ninth Circuit and various trial court opinions that read Comcast as having a much more modest effect.33 If there is a trend, it seems to run in a direction opposite to the one the authors imply. Clearly, the authors are critical of Whirlpool, Leyva, and Butler. Whether or not they are right, as a matter of pure description, the authors stray from objectivity by reading the lower court opinions broadly that read Comcast broadly, and the lower court opinions narrowly that read Comcast narrowly. It is hard to resist the conclusion that this analysis is doubly skewed. The authors’ substantive view of Comcast suffers from similar partisanship. The opinion held merely that a plaintiff’s theory of liability must match its measure of harm.34 The term “holding” should be used advisedly. A holding is the rule pronounced by a court that is necessary to its
The Class Cert Games: Coach, Commentator, or Critic?
71
decision.35 The Court’s reasoning in Comcast makes clear that the above holding was all that was necessary to its decision. Indeed, it suggested that its decision “turns on the straightforward application of class certification principles.”36 That would be a strange way to announce a radical change in class action doctrine. To be sure, some of the language in Comcast could be read more broadly, suggesting situations in which class certification might not be appropriate. None of them, however, was before the Court. That is not to say that its dicta may not or should not influence lower courts. And the authors are creative perhaps even ingenious about wringing every possible argument for defendants out of loose language that the Court included in its opinion. Some of those arguments may prevail. Some of them have. Others have not. Some have prevailed in some cases but not others. The key point is that it would be a mistake to confuse as an objective reading of Comcast what is in fact a suggestion about the various arguments it can potentially support, some of them plausible and others much less so. Indeed, as noted above, the authors have acknowledged that the Sixth Circuit on remand in Whirlpool37 and the Seventh Circuit on remand in Butler38 did not agree with their interpretation of Comcast. The New Rigor includes other substantive claims that should not be taken as settled law, although they are presented that way. The New Rigor, for example, suggests in passing that plaintiffs must show harm to “each class member” to support class certification.39 That is not the law. Or, to be more precise, whether that is the law is highly questionable. Various courts have held that a lesser showing can suffice under Rule 23(b)(3).40 A possible response to these concerns is that the authors are not merely playing the role of commentator, offering objective interpretations of what the law is or objective predictions of what it will become, but also playing the role of critic, contending how the law should be interpreted. To the extent the authors intend to adopt the latter role, however, they say far too little.
CRITIC The New Rigor’s efforts at playing commentator suffer, then, because of insufficient objectivity. At the same time, however, the effort to appear at least somewhat objective may undermine the strength of The New Rigor as critic. The authors make some suggestions about what the law should be,
72
JOSHUA P. DAVIS
but they tend to be brief and insufficiently supported. Two issues illustrate this point. The New Rigor suggests that the same rigor should apply to damages as to fact of damage at class certification.41 But courts have long treated damages differently than the elements of a claim on the merits.42 Plaintiffs must show fact of damage also called impact or antitrust injury to prevail on a private antitrust claim. Once they have done so, however, courts have long allowed plaintiffs greater leeway in proving the amount of damages they are owed. As the Seventh Circuit explained in Mishawaka v. Am. Elec. Power Co., “We recognize that the burden of proof on [plaintiffs] as to the amount of damages is less than that required for proof of antitrust injury. It need not be precise.”43 That distinction has informed class certification doctrine, as well it should. The more lenient standard for proving the amount of damages as opposed to fact of damage translates naturally to a more forgiving standard regarding the amount of damages at class certification. Indeed, to impose a stricter standard for proving damages in a class setting than in an individual action could well violate the Rules Enabling Act. After all, Rule 23 as a Federal Rule of Civil Procedure cannot abridge, enlarge, or modify substantive rights.44 To be sure, The New Rigor briefly acknowledges the distinct showings plaintiffs must make regarding proof of impact (or fact of damage) and proof of the amount of damages, citing some of the seminal cases.45 So did the Supreme Court in Comcast, recognizing that damage “[c]alculations need not be exact.”46 But The New Rigor never grapples adequately with the implications of this distinction for the class context. The relaxed standards for proving the amount of damages as compared to impact can help to explain the extensive case law holding individualized damage issues do not preclude class certification47 and allowing damages in an antitrust class action to be proven on an aggregate rather than an individual basis.48 This last point is crucial. Much of the authors’ analysis seems to assume but does not argue that courts should abandon the long line of authority allowing the award of aggregate damages in antitrust class actions. A similar problem besets The New Rigor’s assertion that Daubert should apply to expert reports relevant to class proceedings. It offers no meaningful support for this position. The authors simply suggest “it makes little sense to require rigor at the class certification stage but then push off tough (or even easy) Daubert calls for some later procedure in the case.”49 However, deferring any Daubert analysis may make good sense. After all, the judge is always the finder of fact in regard to class certification. There is
The Class Cert Games: Coach, Commentator, or Critic?
73
therefore much less of a need for judges to serve as gatekeepers, so that “junk science” does not improperly affect juries.50 Moreover, it is hard to imagine that a Daubert ruling would ever make a difference in a judge’s decision whether to certify a class. Consider what would have to happen for it to do so: A judge would have to conclude, on the one hand, that an expert’s opinion is so flawed that it is inadmissible but, on the other hand, that it is also so persuasive that it changes her mind about class certification. What are the odds a judge would hold both views? If they are very low, Daubert motions add to the time, cost, and paperwork involved in proposed class actions without contributing significantly to the quality of decision-making or the result of the class certification decision. Perhaps there are reasons nonetheless to apply Daubert to expert reports pertaining to class certification. If so, the authors of The New Rigor have not presented them.
CONCLUSION The success of The New Rigor depends in large measure on what its primary goal is. As the advice of a talented coach, it warrants high marks, even if the coach is engaging in a relatively transparent effort to influence the referees. As the objective assessment of a commentator, it does not fare nearly as well. And as a normative argument regarding controversial issues relevant to class actions, the analysis is incomplete. But that is only to be expected. We should not demand that a single article succeed in all of these endeavors at once. We would never hire someone to serve simultaneously as coach, commentator, and critic in the same game. That is too much to ask of anyone.
NOTES 1. See Keyte, Eckles, and Lent (2014, p. 28) (discussing the application of Daubert v. Merrell Dow Pharm., 509 U.S. 579 (1993)). 2. Ibid, p. 32. 3. Ibid, p. 45. 4. Daubert v. Merrell Dow Pharm., 509 U.S. 579 (1993). 5. See Keyte et al. (2014, p. 27) (“Where a plaintiff’s class action expert survives a Daubert challenge, this by no means suggests that class certification is inevitable or easy. Instead, survival of a Daubert motion is a necessary, but not in
74
JOSHUA P. DAVIS
and of itself sufficient, condition for class certification.”) (citations omitted); ibid., p. 28 (“Significantly, most of the problems addressed above that may provide the basis for a Daubert motion, also provide defendants a basis to argue that, regardless of whether the plaintiff’s expert’s testimony is formally excluded, it is still insufficient to satisfy plaintiffs’ burden in proving that common issues predominate.”). 6. For an interesting empirical analysis suggestive of this possibility see Langenfeld and Alexander (2011) (“It is possible that the court’s review of expert methodologies at the class certification stage, even if such review is made under a relaxed Daubert standard, makes Daubert challenges less likely to succeed at later stages.”). This danger of trying to take two bites at the apple also may apply to overly aggressive motions to dismiss. 7. See Keyte et al. (2014, p. 28) (quoting Dukes, 131 S. Ct. at 255). But note The New Rigor is not always as clear as it might be that courts should address the merits only insofar as they are relevant to the class certification standard, as Amgen recently held. See Amgen, Inc. v. Conn. Ret. Plans & Trust Funds, 133 S. Ct. 1184, 1195 (2013) (“Merits questions may be considered to the extent but only to the extent that they are relevant to determining whether Rule 23 prerequisites for class certification are satisfied.”). It seems like an overstatement to say Comcast “held that a close look, including at arguments pertinent to the merits, is required to satisfy predominance” (Keyte et al., 2014, pp. 3536). It would be more precise to say it is permissible and appropriate to evaluate merits issues to the extent they are relevant to the Rule 23 standard. 8. Ibid., p. 16. 9. Ibid. 10. See Comcast, 133 S. Ct. 1426, 1432 (2013) (“Respondents’ class action was improperly certified under Rule 23(b)(3).”). The New Rigor treats the discussion in Comcast of petitioners’ waiver of a Daubert challenge, ibid. at 1431, n. 4, as if it implicitly ruled that petitioners could have raised the issue if they had not waived it (Keyte et al., 2014, p. 16). But that issue was not before the Court. 11. Ibid., p. 16, n. 13 (quoting Wal-Mart Stores, Inc. v. Dukes, 113 S. Ct. 2541, 25532554 (2011)). 12. Ibid., p. 15. 13. Ibid. 14. See ibid., p. 18 (“Types of Rule 23(b)(3) Expert Analysis That Should Not Make It Past a Daubert Screen”). 15. See ibid., p. 27 (“Step 2: Evaluate if the Proposed Methodology for Proving Impact Be Used by Every Class Member if Applied in Individual Cases”). 16. See ibid., p. 28 (“Defendants Should Attempt to Develop a Factual Record Demonstrating that Plaintiffs’ Expert’s Assumptions about the Actual World Are Incorrect”). 17. See ibid., p. 45 (“Ways to Attack Damages Evidence under the New Rigor”). 18. 678 F.3d 409 (6th Cir. 2012) (GVR’d 4/1/2013). 19. 702 F.3d 359 (7th Cir. 2012) (GVR’d 6/3/2013). 20. See Keyte et al. (2014, pp. 3743) 21. Ibid. 22. See, ibid. (criticizing the Whirlpool and Butler decisions after Comcast). 23. See ibid., pp. 4345.
The Class Cert Games: Coach, Commentator, or Critic?
75
24. See ibid., pp. 17, 28, 32, 35. 25. See, for example, ibid., pp. 1827 (discussing with approval cases that rejected plaintiffs’ expert testimony under Daubert); ibid., pp. 2934 (discussing cases denying class certification). 26. See ibid., pp. 4649 (including various cases denying class certification based on issues related to damages). 27. Ibid., pp. 3839, 4445. 28. See ibid., pp. 3738. 29. Glazer v. Whirlpool Corp., 722 F.3d 838 (6th Cir. 2013); Butler v. Sears, 2013 U.S. App. LEXIS 17748 (7th Cir. 2013). 30. Keyte et al. (2014, p. 43). 31. See ibid., pp. 4043. 32. Leyva v. Medline Industries, Inc., 716 F.3d 510 (2013). 33. See Keyte et al. (2014, p. 43, nn. 179180). 34. Comcast, 133 S. Ct. at 1433 (reasoning that plaintiffs’ “model purporting to serve as evidence of damages in [a] class action must measure only those damages attributable to” the liability theory available to them). 35. See, for example, Dorf (1994) and Greenawalt (1989). 36. Comcast, 133 S. Ct. at 1433. 37. Glazer v. Whirlpool Corp., 722 F.3d 838 (6th Cir. 2013). 38. Butler v. Sears, 2013 U.S. App. LEXIS 17748 (7th Cir. 2013). 39. See, for example, Keyte et al. (2014, pp. 11, 12, 21, 24, 25, 27, 28, 35, 38, 39). 40. Perhaps the clearest statement comes from Judge Posner in Kohen v. Pacific Investment Management Company LLC, 571 F.3d 672, 677 (7th Cir. 2009): [A] class will often include persons who have not been injured by the defendant’s conduct; indeed this is almost inevitable because at the outset of the case many of the members of the class may be unknown, or if they are known still the facts bearing on their claims may be unknown. Such a possibility or indeed inevitability does not preclude class certification, despite statements in some cases that it must be reasonably clear at the outset that all class members were injured by the defendant’s conduct.
See also DG v. Devaughn, 594 F.3d 1188, 1198 (10th Cir. 2010); Pella Corp. v. Saltzman, 606 F.3d 391, 394 (7th Cir. 2010); Cordes & Co. Fin. Servs., Inc. v. A.G., Edwards & Sons, Inc. 502 F.3d 91, 106107 (2d Cir. 2007); In re Urethan Antitrust Litig., 251 F.R.D. 629, 538 (D. Kan. 2008); Meijer, Inc. Warner Chilcott Holdings, Co. 246 F.R.D. 293, 310 (D.D.C. 2007); In re Rubber Chems. Antitrust Litig., 232 F. R.D. 346, 352 (N.D. Cal. 2005); J.B.D.L. Corp. v. Wyeth-Ayerst Labs, Inc., 225 F. R.D. 208, 219 (S.D. Ohio 2003); In re Cardizem CD Antitrust Litig., 200 F.R.D. 297, 321 (E.D. Mich. 2001). 41. Keyte et al. (2014, p. 14). 42. See also Bigelow v. RKO Pictures, Inc. 327 U.S. 251, 264 (1946); Story Parchment Co. v. Paterson Parchment Paper Co., 282 U.S. 555, 563 (1931). 43. 616 F.2d 976, 987 (7th Cir. 1980). 44. Shady Grove v. Allstate Ins. Co., 559 U.S. 393, 406407 (2010) (holding that, under the Rules Enabling Act, the Federal Rules of Civil Procedure cannot “abridge, enlarge, or modify” substantive rights and that Rule 23 does not do so).
76
JOSHUA P. DAVIS
45. See Keyte et al. (2014, p. 14, n. 5) (citing Story Parchment Co. v. Paterson Parchment Paper Co., 282 U.S. 555, 563 (1931); Olden v. LaFarge Corp., 383 F.3d 495, 508 (6th Cir. 2004)). 46. Comcast, 133 S. Ct. at 1433 (citing Story Parchment, 282 U.S. at 563). 47. See, for example, Leyva, 716 F.3d at 514 (holding Comcast does not change the rule that “the presence of individualized damages cannot, by itself, defeat class certification”); Whirlpool, 722 F.3d at 861 (noting the “well nigh universal” rule that “individual damages calculations do not preclude class certification under Rule 23(b)(3)”) (quoting Comcast, 133 S. Ct. at 1437 (Ginsburg, J. dissenting)). 48. See, for example, In re Flonase Antitrust Litig., 284 F.R.D. 207, 233 (E.D. Pa. 2012) (accepting aggregate class damages model); In re Neurontin Antitrust Litig., 2011 U.S. Dist. LEXIS 7453, at *4143 (D.N.J. January 25, 2011) (rejecting argument that “[plaintiffs’] damages model is inappropriate because it aggregates damages”); In re K-Dur Antitrust Litig., 2008 U.S. Dist. LEXIS 118396, at *74 (D.N.J. April 14, 2008), aff’d, 686 F.3d 197, 220 (3d Cir. 2012) (approving “aggregate damage model”). As a corollary, various courts have allowed experts to use averages in calculating aggregate damages. See, for example, Wellbutrin XL Antitrust Litig., 2011 WL 3563385, *1415 (E.D. Pa. August, 11, 2011) (expert “calculate[d] damages as the difference between the weighted-average price that class members paid … and the weighted average that class members would have paid but for the alleged conduct”); Meijer, Inc. v. Abbott Laboratories, 2008 WL 4065839, *10 (N.D. Cal. 2008) (“Plaintiffs have proffered methods for calculating aggregate damages for overcharges paid by class members, based on average market prices.”); In re: Titanium Dioxide Antitrust Litig., 2013 U.S. Dist. LEXIS 62394, *55 (D. Md. May 1, 2013) (“Plaintiffs … are not precluded from using Dr. Lamb’s multiple regression analysis to determine an average overcharge to the class as a whole.”). 49. Keyte et al. (2014, p. 17). 50. That can explain why some courts have declined to apply Daubert at class certification, In re Evanston Northwestern Healthcare Corp. Antitrust Litigation, 261 F.R.D. 154 (S.D. Ind. 2009), and others have imposed a lighter burden to survive a Daubert motion at class certification. In re Ready-Mixed Concrete Antitrust Litigation, 268 F.R.D. 56 (N.D. Ill. 2010).
REFERENCES Dorf, M. C. (1994). Dicta and Article III. University of Pennsylvania Law Review, 142, 1997, 2003. Greenawalt, K. (1989). Reflections on holding and dictum. Journal of Legal Education, 39, 431, 435. Keyte, J., Eckles, P., & Lent, K. (2014). From hydrogen peroxide to Comcast: The new rigor in antitrust class actions. In J. Langenfeld (Ed.), The law and economics of class actions (Vol. 26, pp. 11–63). Research in Law and Economics. Bingley, UK: Emerald Group Publishing Limited. Langenfeld, J., & Alexander, C. (2011). Daubert and other gatekeeping challenges of antitrust experts. Antitrust, 25, 21, 25.
ANTITRUST CLASS PROCEEDINGS THEN AND NOW Michael D. Hausfeld, Gordon C. Rausser, and Gareth J. Macartney, with Michael P. Lehmann and Sathya S. Gosselin ABSTRACT In class action antitrust litigation, the standards for acceptable economic analysis at class certification have continued to evolve. The most recent event in this evolution is the United States Supreme Court’s decision in Comcast Corp. v. Behrend, 133 S. Ct. 1435 (2013). The evolution of pre-Comcast law on this topic is presented, the Comcast decision is thoroughly assessed, as are the standards for developing reliable economic analysis. This article explains how economic evidence of both antitrust liability and damages ought to be developed in light of the teachings of Comcast, and how liability evidence can be used by economists to support a finding of common impact for certification purposes. In addition, the article addresses how statistical techniques such as averaging, price-dispersion analysis, and multiple regressions have and should be employed to establish common proof of damages.
The Law and Economics of Class Actions Research in Law and Economics, Volume 26, 77133 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0193-5895/doi:10.1108/S0193-589520140000026004
77
78
MICHAEL D. HAUSFELD ET AL.
Keywords: Class actions; damages; common impact; class certification; multiple regression; antitrust JEL classifications: B40; K00; K41; K21; L00
TABLE OF CONTENTS Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Early History of Rule 23 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Eisen and Bogosian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Rise and Fall of Bifurcated Discovery. . . . . . . . . . . . . . . . . . . . . . 2003 Amendments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hydrogen Peroxide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Wal-Mart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Comcast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Third Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Preliminary Inquiry into the Merits . . . . . . . . . . . . . . . . . . . . Impact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Daubert . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Damages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Supreme Court. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Post-Comcast Economic Testimony and Class Certification. . . . . . . . . The Role of Economic Analysis in Class Certification . . . . . . . . . . . Scientific Reliability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Causality and Common Impact . . . . . . . . . . . . . . . . . . . . . . . . . . . . Relationship Between Liability, Common Impact, and Damage Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Investigating Liability through Common Evidence . . . . . . . . . . . . . Investigating Impact through Common Evidence . . . . . . . . . . . . . . Proper and Improper Uses for Averages . . . . . . . . . . . . . . . . . . Price Dispersion and Its Implications for Common Impact . . . . Demonstrating Damages through Common Evidence . . . . . . . . . . . Determining whether a Proposed Damage Methodology Is “Workable” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Unreliable Decomposition of Regression Model Data Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion and Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
79 80 82 84 85 85 87 89 89 89 89 90 90 91 92 93 95 97 100 102 104 105 106 110 111 113 115 131
Antitrust Class Proceedings Then and Now
79
INTRODUCTION In the wake of the Supreme Court’s recent decision in Wal-Mart Stores, Inc. v. Dukes1 and Comcast, Inc. v. Behrend,2 scholars, judges, and attorneys are revisiting the requirements of Rule 23 and seeking to identify the precise effects of Wal-Mart and Comcast on future class-action litigation. The legal community can, however, agree on one central proposition: class certification has, in recent decades, become increasingly complex whether as an intellectual project, an expenditure of human and financial resources, or an evidence-gathering and demonstrating task and this trend is all the more evident in the antitrust field. This was not always the case, of course. Federal rule of Civil Procedure 23 was once a relatively simple procedural mechanism, and the class-certification determination typically occurred at the outset of any litigation, well in advance of significant discovery. With the advent of Wal-Mart and Comcast, however, and their expositions of the “rigorous analysis” required of courts, litigants can expect that the recent trend of deferred class-certification decisions (in favor of substantial precertification discovery) will continue. It is now clear more than ever before that the economic experts must carefully address each step in class-certification analysis liability, common impact, and measurement of damages based on reliable record evidence and data. Otherwise, the expert runs the risk of being discounted or ruled inadmissible. There are several practices that economics experts can adopt in order to ensure that their analysis both addresses the increased complexity of class certification and maintains professionally accepted standards. These practices are discussed in the Section “PostComcast Economic Testimony and Class Certification” of this article and include: (i) the rigorous adherence to scientific hypothesis formulation and empirical testing, rather than unacceptable data mining practices; (ii) the identification of causal relations, rather than mere correlations; (iii) a separation of liability, common impact, and damages into three distinct analyses, the results of each one feeding in as foundation for the subsequent analyses; (iv) the proper use of averages and the correct consideration of the importance or otherwise of price dispersion; and (v) the correct assessment of regression model reliability and interpretation of regression results, including when damage regressions are applied to individual class members. Carefully addressing these and related issues make class certification possible in many cases, even with decisions such as Wal-Mart and Comcast.
80
MICHAEL D. HAUSFELD ET AL.
As the legal community wrestles with the commands of Rule 23 and the attendant availability of and need for early nonbifurcated discovery, a brief retrospective of the timing of the class determination places this shift in evolutionary context and is useful for keeping in mind the goals of class actions. This retrospective is presented in the Sections “The Early History of Rule 23” through Eisen and Bogosian “The Rise and Fall of Bifurcated Discovery.”
THE EARLY HISTORY OF RULE 23 Class actions originated some 800 years ago, in what Prof. Stephen Yeazell has dubbed “medieval group litigation.”3 The earliest published sources record one instance in which a parish rector sued four of his parishioners (as representatives of the entire community) to recover certain parochial fees.4 Later came the English bill of peace, a procedural device that enabled representative parties to petition the English Courts of Chancery to aggregate multiple claims in a single equity proceeding.5 In the mid-19th century, the United States Supreme Court, borrowing from principles embodied in the bill of peace, created Federal Equity Rule 48 and with it “group representative litigation.”6 Equity Rule 48 provided, in relevant part: Where the parties on either side are very numerous, and cannot, without manifest inconvenience and oppressive delays in the suit, be all brought before it, the court in its discretion may dispense with making all of them parties, and may proceed in the suit, having sufficient parties before it to represent all the adverse interests of the plaintiffs and the defendants in the suit properly before it. But in such cases the decree shall be without prejudice to the rights and claims of all the absent parties.7
That rule proved too complicated to be of much utility, however, and it was seldom used.8 In 1912, the Supreme Court scrapped its earlier effort and rewrote the rule (renumbered Equity Rule 38), striving for simplicity with a single sentence: “when the question is one of common or general interest to many persons constituting a class so numerous as to make it impracticable to bring them all before the court, one or more may sue or defend for the whole.”9 In 1937, the Supreme Court promulgated the Federal Rules of Civil Procedure.10 Included was Rule 23, which was the first procedural device to permit class action suits for monetary damages in the United States.11 Rule 23, the advisory committee noted, was a “substantial restatement of [former] Equity Rule 38 (Representatives of Class),” although the new
Antitrust Class Proceedings Then and Now
81
rule “applie[d] to all actions, whether formerly denominated legal or equitable.”12 And the criteria remained the same: class treatment13 was appropriate if the litigation posed a question of “common or general interest to many persons constituting a class so numerous as to make it impracticable to bring them all before the court.”14 Notably, however, the first Rule 23 did not “pa[y] any attention to the details of the procedural management of class actions.”15 Nor was there a “routine certification procedure for the representative suit.”16 As Prof. Richard Marcus notes, “it was expected that judges would scrutinize the propriety of class treatment [at some unspecified time] before entering judgment. Indeed, it seems that the resolution of the question whether a case was a proper class action was enmeshed in the evaluation of the merits of the case and emerged from the resolution of the merits.”17 Other features of Rule 23 caused confusion as well. The three categories of class actions “true” class actions, which concerned “joint, common, or secondary rights” in which all plaintiffs had substantially identical interests; “hybrid” class actions, which described plaintiffs with individual causes of action tied to a common fund or property and united by at least one common question; and “spurious” class actions, which involved distinct rights but at least one common question concerning common relief “proved obscure and uncertain.”18 Further compounding the problem, Rule 23 did not require notice (at least for “hybrid” and “spurious” classes) or provide an opportunity for class members to opt out of the litigation.19 What is more, there was considerable uncertainty about the binding effect of a judgment on absent class members.20 Absent class members were bound by any judgment in “true” and “hybrid” class actions, but not with respect to “spurious” class actions.21 It was an open question, for example, whether absent “spurious” class members who were exempt from an unfavorable decision could nevertheless gain the benefit of a favorable decision after the fact.22 Recognizing these flaws, the rules drafters substantially revised Rule 23 in 1966. The 1966 amendments included the addition of the four foundational requirements required of all class actions today: numerosity, commonality, typicality, and adequacy.23 Additionally, the rule now clarified that a class action judgment binds all absent class members.24 Another noteworthy feature concerned timing: “In order to give clear definition to the action,” Rule 23(c)(1) required a court to determine “as soon as practicable” whether a class action was the appropriate procedural classification.25 That determination could be conditional, however, enabling a court to revisit its decision at some later point.26
82
MICHAEL D. HAUSFELD ET AL.
EISEN AND BOGOSIAN The “as soon as practicable” admonition prompted further questions about the procedural management of class actions. What was the precise decision sequence called for by the rule? What, if any, discovery should be permitted before a class-certification determination? And did Rule 23 authorize any consideration of the merits at this early stage? In essence, what could courts do with the facts and the merits at certification? What is more, any economic analysis, though potentially relevant to some class-certification questions, also seemed premature in light of the timing language embedded in the rule. Many judges found at least a partial answer in Eisen v. Carlisle & Jacquelin.27 Eisen, a price-fixing case against two “odd-lot” dealers for stock-exchange trades, chiefly concerned manageability and notice. The trial court judge, Judge Tyler, initially denied class status, reasoning that common questions of law or fact did not predominate over individual questions and plaintiffs’ publication notice plan was sorely inadequate for a class of nearly 4 million people.28 The Second Circuit overturned that decision, however, emphasizing the need for “a liberal rather than a restrictive interpretation” of the new rule.29 On remand, Judge Tyler allowed extensive discovery in order to probe various questions relevant to the class determination. After six months of evidence gathering, Judge Tyler concluded that class status was appropriate under Rule 23(a) and 23(b), leaving only the question of which side would bear the substantial costs of notice. Following a hearing and additional submissions, the court held that the defendants should bear 90% of the notice costs given the likelihood that plaintiffs would ultimately prevail on the merits.30 That prompted another appeal, culminating in the Supreme Court’s rejection of Judge Tyler’s novel notice-cost solution: We find nothing in either the language or history of Rule 23 that gives a court any authority to conduct a preliminary inquiry into the merits of a suit in order to determine whether it may be maintained as a class action. Indeed, such a procedure contravenes the Rule by allowing a representative plaintiff to secure the benefits of a class action without first satisfying the requirements for it. He is thereby allowed to obtain a determination on the merits of the claims advanced on behalf of the class without any assurance that a class action may be maintained. This procedure is directly contrary to the command of subdivision (c)(1) that the court determine whether a suit denominated a class action may be maintained as such “(a)s soon as practicable after the commencement of (the) action ….”31
Antitrust Class Proceedings Then and Now
83
Because of its seeming breadth, the Supreme Court’s pronouncement convinced many that class certification and merits analysis were mutually exclusive.32 That ambiguity lingered until Wal-Mart, where Justice Scalia explained that in Eisen, “the judge had conducted a preliminary inquiry into the merits of a suit, not in order to determine the propriety of certification under Rules 23(a) and (b) (he had already done that …), but in order to shift the cost of notice required by Rule 23(c)(2) from the plaintiff to the defendants. To the extent the … statement goes beyond the permissibility of a merits inquiry for any other pretrial purpose, it is the purest dictum and is contradicted by our other cases.”33 Without the benefit of such explicit guidance, though, courts maintained a fairly strict division between class-certification analysis and merits considerations for nearly forty years, from Eisen until the first signs of erosion in the circuit courts.34 That division became known as the “Eisen Rule” and was, for a long time, a pillar of class-action practice.35 The strongest form of the Eisen Rule was simply to assume the validity of plaintiffs’ allegations, with judges typically invoking this strong form as a shortcut to class certification.36 Another historical standard relevant for class certification is known as the “Bogosian Shortcut,” which follows the Third Circuit’s 1977 decision in Bogosian v. Gulf Oil Corp.37 The pertinent passage of Bogosian essentially states that if plaintiffs can prove that the prices in an industry exhibit a “price structure” such that the conspiratorially affected prices, even though they differ across regions or class members, were all higher than those that would have existed under competitive conditions, it is possible to infer that all class members suffered some economic harm.38 In the past, a number of district courts have seemingly relied on the Eisen Rule in conjunction with the Bogosian Shortcut to simplify their class-certification evaluation.39 Plaintiffs’ experts would argue that the products purchased by class members are homogenous and part of the same product market and the same geographic market. Under such circumstances, economics predicts that the Law of One Price holds, with the prices of different products differing only by fixed transaction characteristics (determined by well-specified physical characteristics and other observable transaction characteristics) and transportation costs. Under the Law of One Price, the prices of the products contained in the proposed class then exhibit “price structure” in that they rise and fall together. Plaintiffs would then invoke the Eisen Rule and assume that the defendants did agree to collectively raise prices (in the case of a monopolistic price-fixing conspiracy) as alleged. Under this assumption, and armed with the price structure
84
MICHAEL D. HAUSFELD ET AL.
opinion, plaintiffs contended that the defendants’ actions necessarily increased the prices paid by all class members, and therefore all class members suffered economic harm.
THE RISE AND FALL OF BIFURCATED DISCOVERY Mindful of the twin commands to determine class status “as soon as practicable” and to avoid any examination of the merits (while nevertheless conducting a “rigorous analysis”40), courts questioned the boundaries and permissibility of precertification discovery. As of 1977, the Manual for Complex Litigation “recommended that no discovery on the merits be permitted during the discovery of the class action issue, except as is relevant to the class determination. Only in exceptional circumstances, such as obvious lack of merit in the claim for relief, should a decision on the merits be made before scheduling discovery on the class action issue.”41 By 1985, however, the Manual reflected the challenges courts faced in making this artificial division: “Often … bifurcating [between class and merits] discovery … will be counterproductive. Discovery relating to ‘class issues’ is not always indistinguishable from other discovery. Nor will discovery into matters affecting other members of the putative class necessarily be wasted if a class is not certified, for in many cases this information will be valuable as circumstantial evidence.”42 That position did not hold, however, with the recognition in 1995 that “[b]ifurcating class and merits discovery can at times be more efficient and economical (particularly when the merits discovery would not be used if certification were denied), but can result in duplication and unnecessary disputes among counsel over the scope of discovery. To avoid this, the court should call for a specific discovery plan from the parties, identifying the depositions and other discovery contemplated and the subject matter to be covered.” Other limitations on precertification discovery were local rules in various districts demanding that any class-certification motion be filed within 90 days of commencement of the action.43 Not surprisingly, bifurcated discovery was, for a time, typical.44 Today, however, bifurcated discovery has virtually disappeared as courts acknowledge the difficulty, if not impossibility, of segregating “merits” and “class” evidence under a “rigorous analysis” of the susceptibility and plausibility of claimed facts and economic theories of violation and damage to common classwide proofs at trial.45
Antitrust Class Proceedings Then and Now
85
2003 Amendments In 2003, Rule 23 was amended in two significant respects. First, Rule 23(c)(1)(A) was altered to require the class determination “at an early practicable time” instead of “as soon as practicable after the commencement of the action.”46 Though subtle, the change was prompted by a recognition that the previous language “neither reflects prevailing practice nor captures the many valid reasons that may justify deferring the initial certification decision.”47 Time may be needed to gather information necessary to make the certification decision. Although an evaluation of the probable outcome on the merits is not properly part of the certification decision, discovery in aid of the certification decision often includes information required to identify the nature of the issues that actually will be presented at trial. In this sense it is appropriate to conduct controlled discovery into the “merits,” limited to those aspects relevant to making the certification decision on an informed basis. Active judicial supervision may be required to achieve the most effective balance that expedites an informed certification determination without forcing an artificial and ultimately wasteful division between “certification discovery” and “merits discovery.”48
Emphasizing the singular importance of the class-certification determination, the 2003 amendments also eliminated the provision (Rule 23(c)(1)(C)) permitting “conditional” class certification.49 The new timing provision had an immediate effect, further swaying courts already inclined to permit combined class/merits precertification discovery.50 Likewise, the 2003 amendments encouraged plaintiffs to “use this more liberalized standard to argue for adequate time and sufficient breadth of discovery. The notes wisely counsel a pragmatic approach to the neverending struggle over certification versus merits discovery.”51 Still, confusion lingered. “As soon as practicable” is not very specific. Defendants, for their part, complained that plaintiffs were obtaining class certification with minimal showings that seemed inconsistent with Falcon’s command of “rigorous analysis.” Plaintiffs, meanwhile, insisted that any more probing court analysis be accompanied (and preceded) by commensurate full-blown discovery. District courts were left to navigate this tension with little appellate guidance.
Hydrogen Peroxide The 2003 amendments roughly coincided with the first of many circuit court opinions insisting on examination of the merits at class certification
86
MICHAEL D. HAUSFELD ET AL.
insofar as they bear on the requirements of Rule 23, a position that bucked the longstanding conventional wisdom concerning Eisen.52 The most detailed and influential of these decisions is undoubtedly In re Hydrogen Peroxide Antitrust Litigation.53 In Hydrogen Peroxide, a case involving certification under Rule 23(b)(3) of a class of chemical purchasers alleging price fixing, the Third Circuit clarified the standards a district court must use when determining compliance with Rule 23.54 Noting that the trial court enjoys “broad discretion to control proceedings and frame issues for consideration,” the Third Circuit nevertheless reasoned that “proper discretion does not soften the rule: a class may not be certified without a finding that each Rule 23 requirement is met.”55 At issue on appeal was whether plaintiffs had satisfied Rule 23(b)(3)’s predominance requirement, which ensures that “questions of law or fact common to class members predominate over any questions affecting only individual members.”56 Defendants contended that the district court had erred in applying a lax standard of proof for class certification; dismissing defendants’ expert testimony without thorough examination; and incorrectly presuming antitrust injury.57 Taking up the first aspect, the Third Circuit explained that Rule 23 is dissimilar to a pleading standard; courts are free to probe beyond the pleadings in determining whether class certification is appropriate.58 “An overlap between a class certification requirement and the merits of a claim,” the court noted, “is no reason to decline to resolve relevant disputes when necessary to determine whether a class certification requirement is met.”59 The court added, “Eisen is best understood to preclude only a merits inquiry that is not necessary to determine a Rule 23 requirement,” a conclusion with which numerous circuit courts now agreed.60 A truly rigorous analysis, the court continued, cannot rely on “[a] party’s reassurance to the court that it intends or plans to meet the requirements.”61 In addition to the support of sister circuits, the court also cited the 2003 amendments to Rule 23(c)(1)(A), which “reflect[] the need for a thorough evaluation of the Rule 23 factors.”62 This alteration, when coupled with the elimination of conditional certification, “guide[s] the trial court in its proper task to consider carefully all relevant evidence and make a definite determination that the requirements of Rule 23 have been met before certifying a class.”63 The Third Circuit summarized its conclusions: First, the decision to certify a class calls for findings by the court, not merely a “threshold showing” by a party, that each requirement of Rule 23 is met. Factual determinations supporting Rule 23 findings must be made by a preponderance of the evidence. Second, the court must resolve all factual or legal disputes relevant to class certification, even if they overlap with the merits-including disputes touching on elements of the
Antitrust Class Proceedings Then and Now
87
cause of action. Third, the court’s obligation to consider all relevant evidence and arguments extends to expert testimony, whether offered by a party seeking class certification or by a party opposing it.64
Some suggested that Hydrogen Peroxide would radically reshape classcertification analysis, at least in antitrust class actions. But that prediction was not entirely accurate, as subsequent cases have demonstrated. For example, in In re Ethylene Propylene Diene Monomer (EPDM) Antitrust Litigation,65 the court noted the boundaries of its inquiry observing that its task is not to choose between competing economic views of pricing that would have existed absent a conspiracy. Rather, the court “need[s] only determine whether the plaintiffs have demonstrated that the issue of antitrust impact is susceptible to proof applicable to the whole class.”66 Cases like Hydrogen Peroxide and In re Initial Public Offerings Securities Litigation67 do not, the court continued, “require plaintiffs to prove the merits of their case-in-chief at the class certification stage.” They need not demonstrate that their multiple regression analysis captures all the proper variables and thus reaches the “right” answer, as the defendants would require them to do. The defendants have failed to convince me that it is methodologically impossible to use a single formula to estimate class-wide damages.68
Similarly, in In re TFT-LCD (Flat Panel) Antitrust Litigation,69 the court probed whether plaintiffs had demonstrated a “plausible methodology” to prove classwide impact and damages, insisting that it need not declare a victor in the battle among competing experts.70
Wal-Mart Three years after Hydrogen Peroxide, in Wal-Mart, the Supreme Court effectively endorsed the Third Circuit’s interpretation of Rule 23. WalMart took up the certification of a nationwide class of 1.5 million current and former female Wal-Mart employees “one of the most expansive class actions ever.”71 The Wal-Mart plaintiffs had alleged that Wal-Mart managers routinely abused their discretion over pay and promotions in favor of male employees, thereby violating Title VII of the Civil Rights Act of 1964. And according to the plaintiffs, a class action under Rule 23(b)(2) was the appropriate vehicle for their claims because Wal-Mart’s discrimination affected all of its female employees. As the Court explained, “[t]he basic theory of their case is that a strong and uniform ‘corporate culture’ permits bias against women to infect, perhaps subconsciously the discretionary
88
MICHAEL D. HAUSFELD ET AL.
decision making of each one of Wal-Mart’s thousands of managers thereby making every woman at the company the victim of one common discriminatory practice.”72 The Supreme Court considered two questions whether the Wal-Mart plaintiffs satisfied Rule 23(a)(2)’s commonality requirement and whether Rule 23(b)(2) permits monetary relief.73 Beginning with commonality, the Court observed that Rule 23(a)(2) requires more than a shared violation of the same provision of law; plaintiffs’ claims “must depend upon a common contention … capable of class-wide resolution which means that determination of its truth or falsity will resolve an issue that is central to the validity of each one of the claims in one stroke.”74 Significant dissimilarities within a proposed class, meanwhile, merit attention because they reduce the likelihood of common answers.75 Nevertheless, “even a single common question” will suffice to establish commonality.76 The Court also clarified the burden at class certification, explaining that “a party seeking class certification must affirmatively demonstrate his compliance with the Rule i.e., he must be prepared to prove that there are in fact sufficiently numerous parties, common questions of law or fact, etc.”77 Echoing language from General Telephone, the Court affirmed that the “rigorous analysis” called for by the rule may necessitate some examination of the merits of the underlying claim, but only to the extent necessary to conclude that Rule 23 has been satisfied.78 The Court emphasized as well that Eisen does not bar a preliminary inquiry into the merits to determine compliance with Rule 23, as some had mistakenly believed.79 Elsewhere, the Court hinted, in dicta, that Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993), may have some application at the class-certification stage: “The District Court concluded that Daubert did not apply to expert testimony at the certification stage of class-action proceedings … We doubt that is so ….”80 Finally, the dissent, led by Justice Ginsburg, expressed its concern that “[t]he Court blends Rule 23(a)(2)’s threshold criterion with the more demanding criteria of Rule 23(b)(3), and thereby elevates the (a)(2) inquiry so that it is no longer ‘easily satisfied’.”81 The “emphasis on differences between class members,” Justice Ginsburg continued, “mimics the Rule 23(b)(3) inquiry into whether common questions ‘predominate’ over individual issues. And by asking whether the individual differences ‘impede’ common adjudication … the Court duplicates 23(b)(3)’s question whether ‘a class action is superior’ to other modes of adjudication.”82 Ultimately, this “‘dissimilarities’ approach leads the Court to train its attention on what
Antitrust Class Proceedings Then and Now
89
distinguishes individual class members, rather than on what unites them.”83 Justice Scalia explained, however, that “for purposes of Rule 23(a)(2) … [w]e consider dissimilarities not in order to determine (as Rule 23(b)(3) requires) whether common questions predominate, but in order to determine (as Rule 23(a)(2) requires) whether there is ‘[e]ven a single [common] question’.”84 Wal-Mart was not an antitrust case. Most antitrust class actions raise damage claims, thus involving both the commonality requirement of Rule 23(a)(2) and the predominance requirement of Rule 23(b)(3). The inquiry in such cases is thus more extensive than in cases involving solely injunctive relief. Since Wal-Mart was decided classes involving direct purchasers of a price-fixed product or service continue to be certified.85
Comcast Third Circuit Less than two months after Wal-Mart issued, the Third Circuit had an opportunity, in Behrend v. Comcast Corp., 655 F.3d 182 (3d Cir., 2011) to synthesize the teachings of Wal-Mart and Hydrogen Peroxide. Comcast provides a helpful analytical roadmap for antitrust class-certification determinations under Rule 23(b)(3), even after its subsequent reversal.86 Preliminary Inquiry into the Merits. The first guideline of Comcast extrapolated from Wal-Mart and Hydrogen Peroxide is that, in performing its rigorous analysis, a court may consider the underlying merits of plaintiffs’ claim to the extent necessary to determine whether class certification is appropriate.87 According to the Third Circuit, however, these limited forays cannot supplant the ultimate fact finder; any factual determination is nonbinding at trial, and courts must be vigilant against the possibility that “class certification hearings … become actual trials in which factual disputes are to be resolved.”88 “Nothing in Hydrogen Peroxide requires plaintiffs to prove their case at the class certification stage,” the Third Circuit reasoned.89 “To require more contravenes Eisen and runs dangerously close to stepping on the toes of the Seventh Amendment by preempting the jury’s factual findings with our own.”90 Impact. The Third Circuit also articulated the test for evaluating antitrust impact under Rule 23(b)(3), explaining that the court’s task is to determine
90
MICHAEL D. HAUSFELD ET AL.
whether impact is “plausible in theory” and “capable of” or “susceptible to” common proof at trial.91 Determining whether an expert’s opinion of impact is capable of common proof at trial by means of a regression analysis or other common evidence is not necessarily capable of resolution by the application of a preponderance of the evidence. Although “plausible” falls below a requirement of perfection, it is above a “threshold” or a “not fatally flawed” standard. The application of plausibility should evaluate whether the plaintiff’s economist had persuasively established “through mathematical models or further data or other means the key logical steps behind [the impact] theory” and its rational fit to the facts and theory of the alleged conspiracy. In re New Motor Vehicles Canadian Exp. Antitrust Litig., 522 F.3d 6, 2526 (1st Cir. 2008). Daubert. Additionally, Comcast clarified the trial court’s duties and plaintiffs’ obligations at class certification with respect to Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993).92 The Third Circuit reasoned that, after Wal-Mart, the Supreme Court “require[s] a district court to evaluate whether an expert is presenting a model which could evolve93 to become admissible evidence, and [does] not require[e] a district court to determine if a model is perfect at the certification stage. This is consistent with our jurisprudence which requires that at the classcertification stage, expert models are evaluated to determine whether the theory of proof is plausible.”94 Daubert considerations should be satisfied upon presentation of expert models that show common evidence demonstrating impact of the conspiracy; based on data and record evidence. Corroborative government agency or law enforcement findings95 should, in fact, strengthen the reliability of such models. This type of presentation should end the district court’s Daubert inquiry: “When plaintiffs present multiple models created by expert witnesses that can show common evidence and those models are based on the appropriate transactional record data, a district court should not have to determine which model should be used in support of certification.”96 Damages. The Third Circuit also reaffirmed the burden on antitrust plaintiffs to show, under Rule 23(b)(3), that “the alleged damages are capable of measurement on a class-wide basis using common proof.”97 Even at trial, plaintiffs’ burden is to establish that the harm suffered from the antitrust violation is measurable, not absolutely certain.98 For purposes of class certification, antitrust plaintiffs should show, by a preponderance of evidence, that they will be able to ascertain damages across the class using common
Antitrust Class Proceedings Then and Now
91
proof.99 That some class members’ damages will exceed others’ does not prevent certification.100 Supreme Court The Supreme Court accepted certiorari in Comcast and rendered its opinion, and on March 27, 2013, reversed the Third Circuit’s decision. The majority opinion noted that class plaintiffs presented four theories of impact, one of which was an “overbuilder” theory: “Comcast’s activities reduced the level of competition from “overbuilders,” companies that build competing cable networks in areas where an incumbent cable company already operates.”101 Three of plaintiff’s theories were not deemed suitable for certification. Only the “overbuilder” theory survived motion practice. Plaintiffs’ economist (“PE”) “designed a regression model comparing actual cable prices in the Philadelphia DMA with hypothetical prices that would have prevailed but for petitioners’ allegedly anticompetitive activities. The model calculated damages of $875,576,662 for the entire class. … As the PE acknowledged, however, the model did not isolate damages resulting from any one theory of antitrust impact.”102 The majority noted that “We start with an unremarkable premise. If respondents prevail on their claims, they would be entitled only to damages resulting from reduced overbuilder competition, since that is the only theory of antitrust impact accepted for class action treatment by the District Court. It follows that a model purporting to serve as evidence of damages in this class action must measure only those damages attributable to that theory. If the model does not even attempt to do that, it cannot possibly establish that damages are susceptible of measurement across the entire class for purposes of Rule 23(b)(3).”103 This purported certainty of the opinion in Comcast, however, was critically questioned by the dissent. Justices Ginsburg, Breyer, Sotomayor, and Kagan said: “[t]he oddity of this case, in which the need to prove damages on a class-wide basis through a common methodology was never challenged by respondents … is a further reason to dismiss the writ as improvidently granted. The Court’s ruling is good for this day and case only. In the mine run of cases, it remains the ‘black letter rule’ that a class may obtain certification under Rule 23(b)(3) when liability questions common to the class predominate over damages questions unique to class members.”104 As the dissent explained, any language in the majority opinion that might suggest that plaintiffs must show that all putative class members suffered injury and damages is dicta at best.
92
MICHAEL D. HAUSFELD ET AL.
As one district court recently stated, The Supreme Court recently reversed a grant of class certification where “[q]uestions of individual damage calculations will in-evitably overwhelm questions common to the class. Comcast Corp. v. Behrend, No. 11-864, 2013 U.S. LEXIS 2544, 2013 WL 1222646 (U.S. Mar. 27, 2013). The Supreme Court’s holding came from its assumption, uncontested by the parties, that Rule 23(b)(3) requires that damages must be measurable based on a common methodology applicable to the entire class in antitrust cases. That assumption, even assuming it is applicable to privacy class actions in some way, is merely dicta and does not bind this court. See 2013 U.S. LEXIS 2544, [WL] at *9 (Ginsburg and Breyer, JJ., dissenting) ([T]the decision should not be read to require, as a prerequisite to certification, that damages attributable to a class-wide injury be measurable on a class-wide basis. (citation and quotation marks omitted)).”105
Other post-Comcast decisions similarly followed pre-Comcast precedent and have held that even if not all class members were impacted by defendants’ practices, that was not a disabling factor so long as all or virtually all of the putative class members can be shown to have been impacted by common evidence.106 Although raised in Comcast, the majority did not resolve the issue of the extent to which any Daubert analysis applies at the class-certification stage or the format of any such inquiry. The oral argument, however, reflected the opinions of several of the Justices with regard to their understanding of the proper interpretation of such a Daubert analysis or inquiry. It seems as though the Court accepted the petitioners’ characterization that a Daubert analysis encompassed “three distinct prongs”: (1) expert qualifications, (2) reliable methodology, and (3) fit of the facts and economic theories to the facts of the market in the litigation, i.e., is there a qualified expert that utilized a methodology that sufficiently fits the facts and is reliably based on a scientific method enabling proof of a plausible classwide theory of impact and damage at trial.107 The Justices seemed to have expressed a no “… magic words approach” to Daubert. As stated by Justice Sotomayor, district courts had to simply “come to the conclusion that the expert’s testimony is persuasive … reliable and probative[.]”108
POST-COMCAST ECONOMIC TESTIMONY AND CLASS CERTIFICATION After Comcast, plaintiffs in antitrust litigation moving for class certification will likely need to present expert testimony by a PE in most instances. The PE should still be able to use standard methodologies yardstick
Antitrust Class Proceedings Then and Now
93
comparisons, “before and after” analysis, regression models, and pricing analysis to analyze classwide impact. Damage calculations still need not be exact.109 However, in determining whether impact and damages are amenable to proof on a classwide basis, it is less likely that the PE can merely assert that he or she can come up with an “admissible” analysis after the certification decision. Courts are more likely to want to see a sample analysis at the time of certification, ideally based on at least a representative portrayal of the defendants’ transaction data. This likelihood is intensified when merits and class-related discovery are not bifurcated and where the class-certification determination is made at a juncture where significant discovery has already occurred and a substantial, if not completed, evidentiary record has been created. In such circumstances, the plaintiff can reasonably expect that defendants will insist on, and courts may apply, a more exacting scrutiny of the expert evidence on certification and, in doing so, will consider evidence concerning the merits to the extent it impinges upon the Rule 23 determination. Accordingly, the following sections present a view of the post-Comcast role of economics in antitrust class certification.
The Role of Economic Analysis in Class Certification In determining whether the standards for class certification have been met, courts handling antitrust cases must typically depend upon the testimony of expert witnesses including economists and statisticians. In an antitrust case, the assignment for the PE is typically to show that impact and the resulting damages can be proven on a classwide basis. Specifically, the PE must conduct common analysis that supports the proposition that the defendants acted in a manner consistent with the stated or demonstrated allegations (“liability” or “merits”), that these actions adversely affected all, or virtually all,110 class members (“common impact”), and that there is a method common to the proposed class that can be used to quantify the economic harm to the class in the aggregate (“damages”).111 The PE’s proposed analysis may draw on evidence from discovery documents and transaction data, as well as the academic literature and information in the public domain. Both discovery documents and transaction data provide economic evidence, with contemporaneous documents evidencing actual recorded market behavior supported by the data on transaction pricing. Further, the use of discovery documents and industry studies to corroborate and support transaction data analysis enables the PE to ensure that his analysis
94
MICHAEL D. HAUSFELD ET AL.
and results are consistent with the facts of the industry and the alleged actions by the defendants. This corroborating evidence serves an important function of providing a “reality check” for the PE’s proposed model and analysis. Similarly, the defendants’ economist (“DE”) should ensure that any data analysis he offers in opposition to the PE is consistent with the facts reflected in discovery documents and elsewhere. Armed with the available discovery, the PE’s task will be to investigate whether there is a predominant commonality in the economic circumstances faced by proposed class members a commonality that will facilitate the proof of liability, impact, and damages through common evidence at trial. In contrast, the DE will investigate whether the conclusions of the PE are reliable, credible, or flawed, often by attempting to show that proof of impact or damages is plagued by noncommon issues and/or inconsistent with industry practice and procedure, or contradicted by contemporaneous evidence. The standards for class certification emerging from Comcast, Hydrogen Peroxide, and Wal-Mart have seemingly created a new paradigm for both plaintiff and defendant experts. Whereas before, the merits of the case could be assumed to hold true, the PE will now in many instances be expected or challenged to investigate the merits to the extent that they are relevant or challenged for issues of commonality and predominance. For example, if the plaintiffs allege that the defendants adopted a universal policy that would necessarily impact all class members, it should now be within the scope of the PE’s assignment to examine whether such a policy was in fact implemented commonly across the class. In addition, the “rigorous analysis” standard requires that the PE not only propose common models and analysis to investigate the elements of the case, but also respond to criticisms of his factual premises, economic theories, and models, and if he has proposed them, his econometric and statistical methodologies. Expansion of precertification discovery now means that the class-certification experts are likely to have shared access to a large body of data, including historical transaction records. Inevitably, experts for the two sides are likely to draw conflicting inferences about commonality from their analysis of this shared evidence. Resolving disputes as to such differing opinions, which will require weighing the credibility of the inferences drawn and the quality of their underlying empirical support, may well be an additional obligation of the court. With this background, we proceed to analyze some of the factors including scientific reliability; reliable hypothesis formulation and the identification of causality; appropriate data treatments, such as the use of
Antitrust Class Proceedings Then and Now
95
averages; and the assessment of whether a damage model is workable that may inform the evaluation of the sufficiency of an expert analysis on classwide proof of impact and damages in antitrust class actions. These are not immutable guidelines. The factors may vary from case to case, but in general, they provide a framework on some of the main areas of expert testimony that would be encountered in antitrust class certifications after the Wal-Mart, Hydrogen Peroxide, and Comcast decisions.
Scientific Reliability When applying scientific principles to class-certification analysis, it is important to keep in mind “the nature of economic data.” As emphasized by econometrician Jeffrey Wooldridge, “rarely can we run a controlled experiment” to uncover a causal relationship between one economic variable and another.112 Instead, economic data that is generated and recorded as part of real-world interactions, rather than laboratory experiments, consists of two basic components: a systematic signal component that represents a causal relationship between measureable economic variables (for instance, the effect of increased income levels on price, through increased demand) and a random noise component. The random noise component can be due to measurement error or it can be due to idiosyncratic variations that are not readily explained by any economic model, yet are not relevant to the expert analysis. It is accepted in the profession that “[n]ot all possible variables that might influence the dependent variable [say, price] can be included … some cannot be measured, and others may make little difference.”113 As a result, “[n]o model could hope to encompass the myriad essentially random aspects of economic life.”114 The fact that there is random variation present in the detailed transaction data available to experts in a class action means that the PE should exercise caution in the identification of causal relationships (when his inferences would be potentially affected by random variation) and the development of hypotheses based on such relationships. Loose inference based on casual, anecdotal observations of random data patterns and correlations can lead to material errors, so caution must be exercised. As we describe in this section, the typical ordering in class-action litigation of liability, common impact, and damages lends itself well to the application of these principles. Scientific reliability may, for example, be demonstrated by formulating hypotheses and applying them to the record evidence.
96
MICHAEL D. HAUSFELD ET AL.
The process of formulating hypotheses has been formalized in science and economics over the centuries. The 18th century philosopher David Hume gave structure to the scientific process in An Enquiry Concerning Human Understanding, which argued the necessity of reasoning, rather than merely observation.115 Karl Popper over 100 years later wrote that “[a] scientist, whether theorist or experimenter, puts forward statements, or systems of statements, and tests them step by step. In the field of the empirical sciences, more particularly, he constructs hypotheses, or systems of theories, and tests them against experience by observation and experiment.”116 It may be tempting for experts engaged in certification analyses to comb through the available data, looking for information that appears to be supportive of their position. This approach, however, should be viewed with skepticism, since it is not necessarily consistent with sound scientific practice, which requires the investigator first to specify a theory with testable predictions and then to test those predictions with empirical observation. Popper writes that “a hypothesis can only be empirically tested and only after it has been advanced.”117 As the economist Earl Swanson writes, one starts with a theory and goes from it to “observations instead of vice versa.”118 This sentiment has been echoed elsewhere in the literature, “We are never justified in inferring theories from empirical observations.”119 Patterns are readily observed in the real world, and paying casual attention to them, rather than following a formal process of logical theory, model specification, and estimation, may lead to spurious inferences.120 The process of searching data (which can include running multiple diverse regression models) to find a hypothesis (rather than forming one first) is often referred to as “data mining.” In one study on the dangers of data mining, the authors investigate the supposedly substantial evidence that stock market returns are higher on certain days of the week, weeks of the month, months of the year, and so on.121 Although these patterns are found to be statistically significant using standard statistical tests, the authors demonstrate through 100 years of daily data that this statistical significance disappears once the distortions due to data mining are accounted for. The authors conclude that the traditional statistical tests do not take into account the relentless searching that has gone into finding patterns in stock market returns that are merely the product of chance. Tests of statistical significance involve a margin of error, often reported as a “confidence level.” This confidence level records the probability that the test result is found purely because of random variation in the data, and is not evidence of a real relationship between two variables. With relentless testing on the
Antitrust Class Proceedings Then and Now
97
same dataset, it is inevitable that results will be found that fall into this error margin and, although reported as statistically significant, such results are, in fact, due merely to random variation in the data.122 Therefore, caution should be exercised by the courts in the certification process to avoid attributing any unwarranted significance to economic conclusions based on data mining tactics. Scientifically reliable analysis in class certification should avoid data mining and instead apply hypothesis formulation to each of the three elements of liability, impact, and damages. An expert’s analysis at each element may test a relevant testable hypothesis. The result of that test may then serve as the foundation of the subsequent element as what is known in economics and statistics as a maintained hypothesis.123 For instance, PE may start by testing the liability hypothesis i.e., did the defendants engage in the alleged actions? If the liability hypothesis is supported by the evidence, it becomes a maintained hypothesis in the impact analysis because it has been supported by sufficient evidence. The PE can then test the common impact hypothesis that the defendants’ behavior caused injury to the class, taking the maintained hypothesis as a fact. If such causation is also demonstrated, the PE will then have, consistent with the rigorous application of scientific principles, constructed maintained hypotheses of both liability and common impact. With these as foundation, the expert can proceed to assess the damages hypothesis i.e., the quantum of damages suffered by the class. At the class-certification stage, the PE should not need to do more than plausibly demonstrate that this investigation can viably be performed using admissible evidence common to the proposed class. One complication in practice is that there is not a legal consensus about to what degree, or even whether, the class-certification expert needs to test the hypothesis of liability.124 So as an alternative, the PE may be asked to assume liability, in the form of accepting as true the conspiracy as alleged and pled by the plaintiffs, and then proceed to test causation in the context of the assumed conspiracy.
Causality and Common Impact Attempting to capture causal relationships is at the heart of economics. In antitrust class actions, economic and econometric evidence can be used by experts to investigate if events in the real world are consistent or inconsistent with having been caused by the alleged wrongful act. However, if such
98
MICHAEL D. HAUSFELD ET AL.
statistical evidence is to be presented in support or opposition of class certification, care must be taken. It is universally recognized by economists and statisticians that correlation does not imply causality; yet in classcertification analyses, this critical distinction is often overlooked. The most obvious example is the extensive use of univariate analysis (where pricing is investigated along only one dimension), rather than multivariate analysis (which attempts to control for each of the observable characteristics that would be expected materially to influence price). Univariate analysis such as a simple graph that presents price movements over time as evidence that proposed class members did (or did not) suffer injury establishes no causal relationship without further analysis. In order to identify the reasons for those observed price movements, the economic expert should, at a minimum, use multivariate analysis. Instead of investigating only how prices change as one price determining variable changes (univariate analysis), multivariate analysis seeks to explain how price changes as a multitude of determining variables change. So, instead of looking solely at price points graphed over time, a multivariate analysis can look at how various economic and transaction characteristics (product strength, distance shipped, fuel costs, and so on) cause prices to change both over time and across different transactions. The preferred method to implement multivariate analysis is to use a multiple regression model, which can “back out” the influence of a multitude of economic and transaction characteristics on price.125 However, an expert could at the very least perform data analysis and use summary statistics that measure variation (such as the standard deviation and the coefficient of variation) to investigate if price variation decreases as the transaction dataset is compartmentalized into isolated groups of transactions with the same economic and transaction characteristics. Such multivariate approaches can be used to demonstrate that factors that cause prices to vary can be controlled for, as they might otherwise confound isolation of any wrongful price manipulation. Similarly, most industries exhibit some price bargaining and discounting, but this alone does not necessarily preclude collusive and effective price manipulation from taking place across all members of a proposed class.126 Indeed, in industries with extensive price negotiation, the publication of collusively set price schedules may merely adjust the point at which those negotiations begin, artificially moving the entire schedule up or down while preserving the individual price differences that would otherwise exist.127 It is also possible to manipulate a particular component of price without manipulating, or triggering offsetting adjustments in, the other components. For example, in an industry where there has always been
Antitrust Class Proceedings Then and Now
99
negotiation over the base prices paid by customers, the imposition of a uniform tax or supplemental fee on top of those base prices may be an effective means for horizontal competitors to collusively implement acrossthe-board price increases. The mere fact that some class members experienced declining base prices during the class period does not justify an inference that they succeeded in countering the tax or supplemental fee that was pasted on top; indeed, their base prices might well have declined identically had the tax never been imposed.128 In order to establish this connection, it is necessary, empirically, to establish a causal link between the imposition of the tax and the downward adjustment in these particular customers’ base prices. This approach is consistent with the academic literature. As Wooldridge emphasizes: “[t]he notion of ceteris paribus i.e., holding all other (relevant) factors fixed is at the crux of establishing a causal relationship. Simply finding that two variables are correlated is rarely enough to conclude that a change in one variable causes a change in another.”129 Often economic variables are correlated because they are simultaneously determined by some other factor. As a result, we should “use econometric methods to effectively hold other factors fixed.”130 This is a basic focus of econometrics: “[a] first course in econometrics teaches students how to apply … [multivariate] regression analysis to estimate ceteris paribus effects.”131 However, even the utility of multivariate regression analysis can be limited because in some instances not all of the factors that influence price can be observed in the available data. For example, an individual customer might manage to achieve greater base price discounts for any number of reasons not reported in the data, such as his acquisition of an interest in a related business that would add to volume purchases, his election to leadership in an influential trade association, or his credible threats to leave for another supplier. These facts could, however, have been the same in the absence of the collusively imposed tax and do not mean that the customer had the ability to escape the effects of that tax. Therefore, reliance on anecdotal examples of unexplained price changes would constitute a professionally unsound and unreliable method. The approach adopted by econometricians to “identify” a causal relationship in the presence of unexplained price variation is referred to in the profession as an “identification strategy.” Ideally, an identification strategy involves a “natural experiment”132 in which a “treated” group is compared to an otherwise identical or near identical “untreated” group, in order to identify the effect of the treatment. In the context of our earlier example, suppose the conspiratorial tax was sequentially implemented and
100
MICHAEL D. HAUSFELD ET AL.
administered to different segments of the class at different times. If so, the expert could seek to compare in each period the base prices paid by certain proposed class members subjected to the tax, with the base prices for proposed class members to whom the tax had not yet been applied. With sufficient controls for any differences between the two groups, a decrease in base prices for certain class members who suffered the tax, relative to those who at the same time did not (supported by sound economic theory as to why certain proposed class members may have negotiated discounts), could provide evidence that some discounting was being implemented to offset the tax. In this fashion, a PE should investigate if common evidence can be used to establish a dual maintained hypothesis of liability and common impact, which will serve as foundation for a quantification of damages. Both the PE’s investigation and the DE’s challenges should adhere to reliable hypothesis formulation and the identification of causal relationships. The next section describes more formally the relationship between the three branches of the analysis liability, common impact, and damages and suggests certain nonexclusive factors that can be helpful in evaluating expert testimony concerning them.
Relationship between Liability, Common Impact, and Damage Analyses Although the three branches of analysis liability, common impact, and damages are factually and logically connected, they are also distinct and subject to different forms of proof. If they were entirely separable, the results for any one of the three branches would have no implications for the other two. However, this may not be the case. The common impact and damage analyses can depend upon either a presumption or some investigation of liability, while the damage analysis must measure the type of injury posited to have occurred given the common impact analysis. The ongoing evolution in legal standards (and particularly some courts’ clarification of the Eisen Rule) appears to recognize the implications that one may have for the others. From an economic perspective, there is a natural and common sense ordering. Liability must be evaluated first because identifying the wrongful conduct is a necessary precursor to analyzing how that conduct may, or may not, have affected putative class members. Impact analysis follows to test whether, as a result of the identified wrongful conduct, all or most of the class would actually have suffered economic harm, regardless of its quantum. The classwide damages that result from
Antitrust Class Proceedings Then and Now
101
that common impact can then be quantified using a professionally accepted regression model or other methodologies.133 At each of the three stages, the economic expert operates within the boundaries of accepted paradigms, such as the assumption that agents on average act rationally, and that the laws of supply and demand apply. Thomas Kuhn had originally proposed the notion of a paradigm for the natural sciences, but it has now been fashionable in economics for several decades.134 A paradigm in economics, as in the natural sciences, “defines the type of relationships to be investigated and the methods and abstractions which are regarded as legitimate within a particular problem area.”135 As discussed above, the paradigm helps to define a set of possible maintained hypotheses from which those most consistent with economic logic will be selected.136 Not all hypotheses are, in fact, testable, given the limitations of real-world data; from the selected maintained hypotheses, it is therefore necessary to identify one or more testable hypotheses.137 Economic experts engaged in a class-certification analysis generally apply the accepted paradigm of neoclassical economics. Under this paradigm, most agents are rational and act in their self-interest in order to maximize their utility.138 In this setting, the PE may either test or assume the liability hypothesis that the defendants acted as alleged and in a fashion that would be expected to cause injury to the putative class. This then establishes a maintained hypothesis (a hypothesis that is plausibly determined or assumed to be true) of liability, which can serve as the foundation for the expert to assess the hypothesis that such behavior had a common impact on the putative class. If the results of the inquiry support that second hypothesis, then the expert can approach the damage analysis and quantify classwide damages armed with the demonstrated maintained hypothesis of common impact. This type of analytical progression is generally widely accepted in economics and most other fields of scientific inquiry. For example, performing a damage analysis without the maintained hypothesis of common impact may mean the damage regression model is applied to a group of products or regions different than those for which prices could plausibly have been manipulated. As a further example, the most commonly used methods for computing classwide antitrust damages rely on the quantification of a price “delta,” computed as the difference between prices actually paid for the class products and those paid for competitive benchmarks understood to be unaffected by the alleged misconduct.139 Typically, these benchmarks are based upon: (1) the same product and time period but in geographic regions unaffected by the alleged manipulation, or (2) the same product and region
102
MICHAEL D. HAUSFELD ET AL.
but in prior or subsequent time periods unaffected by the alleged manipulation, or (3) a comparable product in regions and time periods as similar to the class products as possible. Differences in economic conditions between the class transactions and the benchmark transactions are accounted for by control variables, and the resulting “delta” by which class transactions are priced differently than the benchmark transactions may be attributed to the defendants’ alleged misconduct and used to quantify the resulting classwide harm.140 To make this critical attribution of the price “delta” to the alleged actions, the PE should either first assume or provide evidence to support the maintained hypotheses of liability and common impact. Attempting to work backward up the logical chain and test the maintained hypothesis of common impact by using the damage model may also lead to spurious inference. Most damage models, such as pricing regressions, rely on economic data. As discussed in subsection Demonstrating Damages through Common Evidence, it is important to keep in mind the “nature of economic data.”141 Such data is generated in the real world, and not in a laboratory. For this reason, it can exhibit substantial random variation, not all of which can be explained by any economic model.142 This unexplained variation may be unimportant for accurately quantifying classwide damages. As discussed below, it may be simply random variation, unrelated to the effect of the alleged acts and will not, therefore, bias an estimate of damages either up or down.
Investigating Liability through Common Evidence When a PE addresses the issue of liability, she should do so based on the existence of common economic and noneconomic factors that describe the defendants’ alleged behavior. The PE may choose to determine an economic model that establishes hypotheses regarding the alleged behavior of the defendants. These hypotheses can then be evaluated against the evidence, to demonstrate if the evidence is common to the class and how that evidence will be used to prove that the defendants acted in a manner considered by economists to be consistent with the allegations against them. The expert should be able to draw from both documentary evidence and actual transaction data. Both are economic evidence, with contemporaneous documents evidencing actual recorded market behavior, supported by the data on transaction pricing. Both PE and DE should use documentary evidence to perform “reality checks” on their statistical analysis and results.143 Are the models used consistent with the reality of the industry
Antitrust Class Proceedings Then and Now
103
and the alleged behavior of the defendants? Do the results of those models and statistical analyses fit the facts of the industry and the case? Any inconsistency with undisputed facts, or facts as found by the court, serves to undermine the reliability of an expert’s analyses. The DE may choose to criticize the PE’s liability model as being inconsistent with accepted economics. There may, for example, be implications of the model that would require some agents to act in an economically irrational manner, which would serve to undermine its credibility. The discovery record may contain evidence of differential behavior by the defendants across the class or evidence of behavior before or after the class period about the contended commonality of behavior during that period. Such evidence would need to be evaluated by the court in assessing whether, based on a preponderance of the evidence, plaintiffs’ model is either incomplete or belies common evidence of liability. The exact analysis required to investigate whether liability can be demonstrated through common evidence will depend on the facts of the industry and the nature of the allegations or record evidence against the defendants. Among a nonexhaustive list of factors a court could consider in assessing common evidence of liability are the following: 1. Can common evidence on industry characteristics be used to support the factual foundations of the PE’s model and hypothesis? For instance, industry characteristics such as high barriers to entry and a lack of competing products would serve to support that the defendants would have expected to succeed in an alleged price manipulation. 2. Is it economically rational that the defendants acted in a fashion that was common across the proposed class, so that plaintiffs’ model can be appropriately applied on a common basis? Rationality could be demonstrated through an analysis of the costs and benefits of a common classwide manipulation, relative to one aimed at only certain proposed class members. 3. Can a common economic model or other acceptable methodology be constructed or used that explains how the alleged behavior by the defendants was rational? For example, do the potential benefits to the defendants of their alleged actions in terms of profitability outweigh the cost of losing customers to nondefendant suppliers of similar products? 4. Are there any implications of such a model that are inconsistent with the accepted economic principles of rational agents acting to serve their own best interest? Such implications would undermine the validity of the model.
104
MICHAEL D. HAUSFELD ET AL.
5. Does an analysis of the documentary record and the transaction data support the allegation that the defendants acted in a common fashion across the putative class that is consistent with the PE’s model and hypotheses? For instance, if the plaintiffs allege a classwide imposition of a conspiratorial tax, was that tax assessed on all (or virtually all) proposed class members, where possible. 6. Does the economic model account for all (or at least the predominant) common influences on price or output? If academic studies or testimony by market participants suggest that, say, strength, color, and distance shipped predominantly determine prices, are these factors included in the model? 7. Does the economic model account for material factors or peculiarities of the market and transactions?
Investigating Impact through Common Evidence In per se cases under Section 1 of the Sherman Act (e.g., horizontal price fixing), direct evidence of anticompetitive conduct and injury is more relevant than indirect evidence concerning the defendants’ intentions or their market power. In contrast, in rule of reason cases, whether or not there is indirect evidence that the defendants had market power typically becomes one of the more contentious issues and may prove dispositive at the classcertification stage. For both types of violations, the current standards for class certification are leaning toward proving common impact through a fuller assessment of the appropriate facts of the case necessary to respond to the requirements of class certification. The expert’s assessment may encompass relevant documents, data and possibly testimony produced in the litigation, as well as publicly available information such as academic, governmental, and industry studies or findings. Where available, data specifically enumerating the transactions alleged to have been affected should be subjected by the experts to rigorous hypothesis formulation and empirical analysis separately from the quantification of classwide damages the objectives and necessary form of the two exercises differ. The resulting body of evidence may support or disprove common impact through a sequence of steps, from identification of the product and geographic scope of the enquiry, through the characteristics of the industry that may or may not make it conducive to collusion, to an evaluation of any evidence from the discovery record that defendants implemented a scheme across the entire class. If available, both sides could
Antitrust Class Proceedings Then and Now
105
also examine the transaction data to determine whether pricing was reasonably uniform across the proposed class period, but only after accounting for the set of observable pricing determinants common to the class transactions (“common factors”), and only after considering that all transaction datasets exhibit some amount of random variation that cannot be readily explained. In common impact analysis in antitrust cases, there are two issues that almost always arise in an analysis of transaction data, which we now examine: (1) the appropriate use of averages and (2) the importance of price dispersion in the industry. Proper and Improper Uses for Averages Often, as with many statistical analyses using large transaction datasets, class-certification analysis will require the use of averages. Occasionally average price movements have been used as evidence that different products are substitutes and/or that their prices are subject to the same forces of supply and demand. Averages can be revealing, but if misused can distort data and mask important differences. The case that brought the use of averages under the spotlight in recent years was Hydrogen Peroxide. The court in Hydrogen Peroxide criticized the use of averages by the PE, on the basis that they masked differences between the distinct types of hydrogen peroxide, which had fundamentally different uses, customer bases, demand drivers, and pricing. Post-Hydrogen Peroxide, “average” became something of a dirty word in class actions. However, the criticism of the use of averages in Hydrogen Peroxide should be understood in the context of that case’s unique facts and not read as erecting a taboo against all uses of averages. Averages are an essential tool for understanding large volumes of complex transaction data. Such tools are at the heart of econometrics. The founding of the Econometric Society in 1930 was in response “to an unprecedented accumulation of statistical information … [and] a need to establish a body of principles that could organize what would otherwise become a bewildering mass of data.”144 The purpose of an average in econometric analysis is, specifically, to “average out” individual differences that are not relevant to the matter being investigated. In determining whether a particular average runs the risk of masking important differences or instead provides a useful tool, it is important to keep in mind the purpose of the analysis. Where the analytical objective is to evaluate common impact, individual variation takes on a special importance and averages should be cautiously employed. However, where common impact has already been
106
MICHAEL D. HAUSFELD ET AL.
satisfactorily established, averages may be an appropriate even essential tool in quantifying classwide damages.145 Because this third branch of inquiry is specifically intended to estimate aggregate classwide damages (which normally requires applying an average overcharge to classwide unit sales), averages are often helpful in generating a concrete and reliable result. Accurate inferences generally require assessment of summary statistics from individual transaction data such as the probability distribution, mean, variance, higher moments, and coefficient of variation, each of which depends upon an average. These standard statistical tools provide a sound basis for investigating differences in the data and it is neither necessary nor possible to explain all variation on an individual basis; attempting to do so can produce misleading results. For instance, it is not possible to determine whether the differences in prices paid by one type of customer and those paid by another type of customer are statistically significant without using the full sample of pricing data to capture the magnitude of total variation in pricing, both systematic and random. Similarly, it is not possible to determine whether such price differences are economically significant without comparing them to the average prices paid in the industry and the variation in those prices. Without recourse to averages, the analysis can become anecdotal and unreliable. Given that there is always variation in prices in real-world data and that simply averaging over that variation can be undesirable, a determination of the importance (or otherwise) of such price dispersion for class certification also requires careful analysis. Price Dispersion and Its Implications for Common Impact Prices are never entirely standard, and all products even the most highly commoditized exhibit unexplained price dispersion.146 Whether the variation creates an impediment to price manipulation or refutes the hypotheses of liability and common impact depends upon its magnitude and the likelihood that it is systematically related to the allegedly unlawful conduct. In answering these questions, an economist’s first step would be to quantify the observed price dispersion. The second step would be an attempt to determine how much of that dispersion can be explained using a set of common transaction characteristics, as well as exogenous factors which economic logic suggests would affect supply or demand for the product. A DE may choose to present price dispersion in the form of a scatter plot and then argue that prices are so substantially different for different transactions that an individual analysis is necessary. The DE’s inference from such
Antitrust Class Proceedings Then and Now
107
a scatter plot would be that the scatter is in and of itself evidence that some proposed class members may have managed to negotiate their way out of the effects of the conspiracy, and this possibility would have to be investigated on a customer-by-customer basis. However, a presentation of price dispersion in the form of a scatter plot can be misleading, if not supplemented with some form of objective quantification. Such quantification can be performed using objective, summary statistics, rather than relying solely on the subjective visual inspection of graphics. One professionally accepted summary statistic is the coefficient of variation (the standard deviation divided by the average). This statistic, like the standard deviation, is a representation of the probability distribution of individual data points arrayed around the average, but it offers the advantage that it is normalized by the average. The resulting statistic appears as a percentage which can be compared to price dispersion in other industries. These comparisons can shed light on whether the degree of price dispersion present in the industry under study is evidence of intra-class heterogeneity that may be problematic for class treatment, because some proposed class members may be successfully negotiating out of the effect of the conspiracy, or whether the dispersion is merely the highly typical and random price variation found in all markets, even those with homogenous and interchangeable goods. Even when price variation may be large, this would not necessarily foreclose class treatment if it can be systematically controlled using causal factors unrelated to the alleged misconduct, such as transaction characteristics, economic conditions, and seasonality.147 A multivariate regression analysis would enable the expert to adjust for these systematic price differences in order to investigate the impact on class members of the alleged wrongful acts. This type of regression utilizes a hedonic approach proposed by some authors for use in class-certification analysis, so named because it measures how attributes or “hedonic” characteristics of products are commonly valued.148 These control variables may more simply be referred to as “common factors” because, even though they may take on different values, their application is common to all transactions.149 Examples include delivery distance, product strength, size of delivery, and transaction date. The analysis of price which controls for these shared characteristics may be referred to as the “common factors regression.” The need to control for multiple transaction characteristics does not turn a common one-model method into an individual analysis, as long as the characteristics are shared across transactions and there is sufficient data so that the parameters associated with each can be estimated reliably.
108
MICHAEL D. HAUSFELD ET AL.
Often the success of a common factors regression is focused on its explanatory power, measured by the R-squared. Whereas economists are comfortable with a wide range of R-squared values as long as other elements of the regression model perform well, courts are sometimes uncomfortable with the inability to perfectly explain all price variation. As noted by Franklin Fisher, statisticians are used to the idea that regression models do not perfectly fit the available data; relatively minor and random influences on prices are left to the error term. He states that “[l]awyers may understand this in principle, but they do not like it.”150 As the court ruled in Reed,151 discussing a regression model with an R-squared that ranged between 51% and 97% (depending on the specification) of individual nurse wage variation, this rate “may be sufficient for the economics literature but it falls far short of satisfying plaintiffs’ legal burden to establish a means of demonstrating by common proof that the members of the putative class were injured and, if so, by how much.”152 The ruling that a regression model with an R-squared as high as 97% does not satisfy the court is difficult for an economist with sound statistical training to accept.153 However, there are no bright lines to determine what constitutes a sufficiently high R-squared. What matters is the significance statistical and economic of the included explanatory variables and the properties of the residual unexplained variation. If all of the systematic variation can be explained, with only unexplained random variation left over, the level of the R-squared is not determinative. There are many reasons why markets for even undeniably homogenous products exhibit price dispersion that cannot be accounted for by an economic model. One such reason may be that not every market participant has the same information at the same time. At any given moment, not all buyers may know each of the prices on offer from all sellers and vice versa. As a result, even given perfect rationality, different parties will agree to different prices for the same transaction at the same point in time. This is reflected in the economics literature, which includes a wide range of studies on price dispersion in homogenous goods. These studies investigate transactions involving highly similar or identical products, occurring at around the same time (in some cases on the very same day), in the same geographic region.154 Even purchases of identically branded products in the same size and packaging involve a certain amount of price dispersion,155 and even purchases of products reasonably well accepted to be homogenous, such as gasoline.156 The challenge for economic experts involved in class certification is to determine which consequences of unexplained variation are important and
Antitrust Class Proceedings Then and Now
109
which are not. A factor which is missing from a model because it cannot be observed and thus contributes to the error term (the unexplained residual) is often referred to as an omitted variable. In econometrics, it is well understood that the important consideration is whether any omitted explanatory variable is correlated with the variable of interest. “[I]n a discrimination case, the variable of interest may be the race or sex of the individual. In an antitrust case, it may be a variable that takes on the value 1 to reflect the presence of the alleged anticompetitive behavior and the value 0 otherwise.”157 No regression model applied to a real-world industry can explain 100% of the variation in pricing; some portion may always be left unexplained.158 It is accepted in the profession that “[n]ot all possible [explanatory] variables that might influence the dependent variable can be included if the analysis is to be successful; some cannot be measured, and others may make little difference.”159 What matters is whether any omitted explanatory variables are correlated with the variable of interest:160 Failure to include a major explanatory variable that is correlated with the variable of interest in a regression model may cause an included variable to be credited with an effect that actually is caused by the excluded variable … Other things being equal, the greater the correlation between the omitted variable and the variable of interest, the greater the bias caused by the omission.
As a result, what is important for class certification is not whether unexplained price variation may exist (which may often be the case), but whether it is related to the alleged illegal conduct. If so, it may confound the ability to determine impact and quantify classwide damages.161 In order to assist in an evaluation of whether or not antitrust impact may be proven through common evidence, the court may consider another set of nonexhaustive factors, as listed here. Again, the exact analysis that the PE will perform will depend on the facts of the industry, the alleged actions by the defendants and the available discovery. 1. Are class members generally in a similar situation regarding economically viable alternatives to escape the defendants’ alleged harmful actions? 2. Are there pretextual or other reasons that defendants proffered to explain their action in a uniform manner across the class? 3. Are there business practices in the industry, such as the linking of prices of different products or the tying of prices to an underlying index, that would have ensured that the defendants’ alleged actions would have had a common impact across class members?
110
MICHAEL D. HAUSFELD ET AL.
4. Can a common regression model or other generally accepted methodology be used to explain pricing variation using a set of common factors? 5. Is there a convincing economic theory that the remaining unexplained pricing variation is relevant to proving impact and quantifying damages? That is, is the unexplained variation related to the conspiracy and the alleged actions by the defendants, or is it merely random and/or anomalous? 6. Has the PE appropriately used averages in his analysis of common impact? In a pricing comparison, it may be inappropriate to average prices over different product strengths thereby masking important variation, but it can be appropriate to present average prices for each of those product strengths (averaging over small random variations across customers) to investigate if the prices for different products generally move together. 7. Has DE offered more than scatter plots that simply record that there exists variation in prices, which he merely speculates will confound a common impact analysis? Has the DE demonstrated that the lack of common impact professed is not the result of mere speculation and is consistent with accepted principles of ceteris paribus and identification?
Demonstrating Damages through Common Evidence The final step in the analysis offered by the PE at the class-certification stage, and inevitably destined to be challenged by the DE, is generally to demonstrate that damages can be quantified on a classwide basis. Comcast makes certain that a PE must quantify classwide damages in a manner that is consistent with the theory of liability. Merely promising later to come up with a workable damage methodology can be insufficient to support certification after Hydrogen Peroxide. However, it is equally clear that the actual or probable amount of such damages need not be established at the certification stage, nor need the expert commit to the precise model specification that will be presented at trial. The emergent standards focus heavily (and appropriately) on the soundness of the PE’s proposed damage model. Critical questions include the professional standards applied to the model, availability of necessary data, the acceptable margin of error, and the reliable interpretation of regression results. All of these questions bear on the “workability” of the proposed damage model.
Antitrust Class Proceedings Then and Now
111
Determining whether a Proposed Damage Methodology Is “Workable” The “reduced form” pricing model is a workhorse of antitrust litigation and represents the most commonly used, but not the only, damage methodology in such cases.162 The model investigates the effect a set of transaction characteristics and economic factors have on market equilibrium prices. The included transaction characteristics are usually the same set of “common factors” identified as important price determinants in the common impact analysis. The reduced form model investigates the market equilibrium price, which reflects both demand and supply forces.163 The damage model quantifies the common classwide impact by estimating the “delta” between prices for class transactions and prices for transactions in a benchmark (whether a different time period, geographic region, or product), which is unaffected by the allegations against defendants. The model’s variables control for differences between the class transactions and the benchmark transactions. After these nonconspiratorial factors have been adjusted for, the price delta between the class transactions and the benchmark transactions is usually captured by a dummy variable that distinguishes the two groups of transactions.164 To be reliable, this comparison requires that there are a sufficient number of data points in both the class period and the benchmark period.165 Three common misapprehensions about such models involve the role of economic intuition, the acceptability of unexplained price variation, and the treatment of measurement error in the underlying data. Economic theory and intuition are indispensable tools in evaluating model specification and regression results.166 In this context, “theory” and “intuition” are not synonyms for speculation, but instead reflect the application of a body of economic knowledge to evaluate whether those variables likely to be important are present in the model, and whether the resulting coefficients for each variable are consistent with economic logic, the facts of the industry, and the alleged actions by the defendants. If important variables are missing or the coefficients behave contrary to economic logic, the facts of the industry, or the alleged behavior of the defendants, these may be signs that the model has been mis-specified or is unreliable. As previously noted, regressions are never “exact,” and cannot be expected to explain all of the variation in the dependent variable (e.g., price); there may be some random variation left unexplained. Similarly, there is typically measurement error in the underlying data. Such errors most commonly arise because of mistakes in initially recording or inputting the data, such as the failure to include the correct number of digits which
112
MICHAEL D. HAUSFELD ET AL.
can easily convert an entry of 1,000 into an entry of 10.167 Statisticians and economists routinely deal with this problem through establishing a protocol to identify and exclude “outliers” in the data, which are random, unsystematic values outside the normal range otherwise represented in the data, or which violate economic logic.168 This is a generally accepted practice; the question is not whether it should be undertaken by each side, but instead whether the outlier exclusions have been too strictly or too leniently set and what is the effect on the regression estimates from outlier removal.169 With these observations in mind and considering that the exact form of the PE’s analysis will depend on the case and the available discovery, the court may assess whether a proposed damage model is theoretically plausible and practically workable by considering the following nonexclusive factors: 1. Is there an appropriate benchmark of transactions that could reasonably be expected to have been unaffected by the defendants’ alleged actions?170 2. Does the model have sufficient explanatory power given the purposes for which it is specified, such that the independent “right-hand side” variables are jointly significant and explain a reasonable amount of variation in the dependent variable (e.g., price)? 3. Is the model applied to data of sufficient quality and quantity in order to both estimate and account for the effect of nonconspiratorial economic factors on prices and estimate the effect of the defendants’ actions on prices? 4. Are the coefficients on each of the important transaction characteristics and economic factors both statistically and economically significant, and do they have signs (positive or negative) that are consistent with economic theory, the facts of the industry and the alleged actions by the defendants? 5. Does the model include as controls all variables that are: (i) expected to have significant explanatory power over the variable used to measure effects of the alleged antitrust violation (e.g., price) and (ii) expected to be different between the benchmark transactions and the class transactions? Failure to include such factors may lead to omitted variable bias in the estimated damages. Variables claimed by a DE to be omitted should be corroborated by transactional data and industry facts, and not merely reflect hypothetical speculation.171 6. Is the model free from problems of “endogeneity”? This problem arises when explanatory variables are themselves affected by the dependent
Antitrust Class Proceedings Then and Now
113
variable, which can result in biased and inaccurate estimates of the effect of each independent variable on the dependent variable.172 Endogenous variables should not be included unless instrumentation is used to filter out the endogenous relationship and identify the causal effect of the explanatory variable on the dependent variable. 7. Is any residual unexplained variation (i.e., the error term) random or anomalous? If nonrandom unexplained variations exists, then estimates of the impact of an illegal act can be biased and unreliable. Have appropriate steps been taken to control for any nonrandomness such as heteroscedasticity or serial correlation, which can often make tests of statistical significance unreliable?173 Unreliable Decomposition of Regression Model Data Samples Plaintiffs’ damage models are sometimes rerun by DEs on an individual customer basis, estimating all the model parameters and the delta for each one, as a means of attempting to disprove common impact. Such an approach may be extremely misleading and may often be unreliable, due in large part to the small amount of data that is available for each individual customer. For example, the DE may find widely varying and erratic coefficients on supply and demand variables in his individual customer regression runs, with rising demand increasing prices for some customers (as one would expect), but decreasing it for others. He may claim that each run finds differing coefficients, because each customer’s demand is different and the results are simply reflecting this. In Plastic Additives, the DE reran the PE’s damage regression one class member at a time. The court noted that of the “115 [class members that he did this for] 81 show[ed] no evidence of a statistically significant increase in prices as a result of the alleged conduct.”174 The PE replied that the regressions were not appropriately specified for this individual use, as evidenced by the fact that the coefficients on supply and demand variables changed with each regression run, and were at times statistically insignificant or had a sign inconsistent with economic theory (with prices going down as demand increased, for example). The PE further noted that there was insufficient data to allow an application of the model for individual customers, leading to the erroneous regression coefficients that changed sign or statistical significance between each customer-specific regression. DE disagreed, arguing that the coefficients on supply and demand variables differed across the customer-specific regressions because each customer may react to supply and demand factors differently.175 The court ruled in favor of the defendants.176
114
MICHAEL D. HAUSFELD ET AL.
Such claims that a regression model applied to individual customers finds widely different coefficients on supply and demand variables, because the model is merely reflecting customer-specific supply and demand dynamics can be highly misleading. The reduced form regression models routinely used in class actions are predicated on the underlying structural market supply and demand equations.177 As is well known, the reduced form models the market equilibrium price, which is determined by the intersection of these supply and demand equations. Thus, the coefficients on the supply and demand factors utilized by PE’s reduced form regression model are correctly interpreted as measuring the relationship between market supply and demand factors and the market equilibrium price. They do not measure the supply and demand relationships for individual buyers/sellers in the market.178 This is because the market equilibrium price which determines the market equilibrium that prices will tend toward and therefore determines the prices that will be observed in the data is not determined solely by any individual consumer’s demand function or any individual firm’s supply function. Rather, it is determined by a horizontal aggregation of those individual demand and supply functions.179 That the reduced form model cannot (and need not) estimate individual customer demand dynamics is relatively easy to understand for the simple reason that in a market customers do not necessarily pay the maximum amount they are willing to pay for a good; instead, they pay the market price. A customer may value a good at $1,000 (determined by his individual demand for that good), but he will pay the market price of $500. Market demand may increase (say, people become richer due to good fortune) and the market price increases to $600. Our customer in question will then pay $600. At no point need he pay his true value of the good, which is $1,000. Therefore, there is no information in the available pricing data on that individual consumer’s demand. The reduced form model will reliably estimate how the shift in demand caused prices to move from $500 to $600, but even if applied to an individual customer’s prices in isolation, will not be able to estimate anything about that customer’s individual demand. Indeed, the model does not need to, because it is marketwide demand and supply that determine that the individual customer paid $500 and then $600 for the product. Simply, the underlying structural demand and supply equations do not identify a consumer’s individual demand function and therefore the reduced form regression does not either, even if it is run for that individual consumer. Such small subset data analyses for individual customers are often used by DEs in an attempt to show that impact varied widely across the
Antitrust Class Proceedings Then and Now
115
proposed class members, perhaps with many supposedly suffering no impact or even experiencing a benefit of an alleged price manipulation. Often in such decomposition exercises, the coefficients on supply and demand variables change substantially with each individual customer regression run. Any claim by an expert that this is because his individual customer regressions are detecting individual customer supply and demand dynamics is likely unreliable, given our preceding discussion. It is much more likely that any unintuitive coefficients found are due to the small data samples that are available for each customer, making the analysis and the inferences regarding the lack of common impact unreliable. Simply, an expert’s interpretation of the coefficients on individual customer regression runs must be considered, as with empirical results in general, in light of generally accepted economic theory and econometric practices.
CONCLUSION AND RECOMMENDATIONS What do these cases and assessments mean for class certification? From the inception of Rule 23 until the present, the process for certification has undergone a number of transformations. The timing to move for certification is no longer mandated at a fixed point early in the litigation; discovery of class issues and merits are integrated; and the court must conduct a rigorous analysis of the facts, merits, and economics to the extent they bear on class-certification requirements. The court is also expected to resolve, by a preponderance of the evidence, all factual and economic disputes and to be persuaded as to the reliability, plausibility, and workability of all economic opinions and methodologies in so far as they respond to the elements of Rule 23. All in all, the process has evolved from one which encompassed briefs based on complaint allegations alone and a possible truncated oral argument, to an extensive development of record evidence supported by corroborating and complementary expert evaluation and analysis, culminating in increasingly detailed hearings, potentially including presentation of live witness testimony. The shift has many implications for both the nature of the proceedings and the responsibilities of the parties and the court. The application of a rigorous analysis to the three essential elements of antitrust liability (conspiracy, impact, and damage) involves a sequential resolution of disputed issues as to each. The findings and conclusions in one segment cannot be disregarded in determinations needed to be made in
116
MICHAEL D. HAUSFELD ET AL.
subsequent segments. Theories of liability, impact, and damages combined with supporting economics must reflect the facts of the conspiracy. Reliability and credibility of the parties’ representation and expert opinions must be judged in the context of the strength or weakness of the factual record confirming or undermining contradictory positions.180 As a whole, class certification has taken on characteristics of a summary judgment framework.181 Rule 23 is designed to concentrate on the ability of the plaintiff class to demonstrate that there is evidence susceptible to common proof at trial that defendants engaged in a conspiracy that impacted and caused damage to all or virtually all members of the class. It is neither a function of Rule 23 nor an obligation of the court to allow the process to be used or to use the process as a forum to decide the ultimate merits of the parties’ differences circumventing the right to a jury trial. Rule 23 must not become a surrogate for summary judgment or replacement for a jury trial.182 The facts, as bounded by the record, are now the touchstone or benchmark of certification. They must be tested equally by the court on all positions relied upon by either party. Both plaintiff’s claims and defendant’s contentions must be judged as to how well or reliably they explain or “fit” the findings of the court. Although certification findings or conclusions are not binding in future pretrial proceedings, they should materially influence decisions within the class proceeding. Findings by the court concerning the existence, nature, and scope of a conspiracy, for example, should be the significant factors in any analysis of the conspiracy’s purported class impact and damage. Similar to the Supreme Court holding in Comcast that the economic measurement of conspiracy damage must match the claimed theory of conspiracy, findings related to class impact and damage should be consistent with a court’s findings of fact defining the conspiracy. In the sequential approach to class antitrust liability elements, findings as to the challenged conspiracy may be established with or without corroborating economic opinion. These findings may be achieved through common factual evidence of the agreement as disclosed in contemporaneous writings, communications, electronic exchanges, declarations, depositions, industry customs, practice, usage, transaction documents, regulatory proceedings, if any, and other available public information. Proof or refutation of this issue will not vary among class members since the essence of the challenged violation is the joint behavior of the defendants to the market as a whole as opposed to any singular market participant. Although economic opinion concerning the predominance of the issue of conspiracy to
Antitrust Class Proceedings Then and Now
117
the class in some instances may not be necessary, it may assist the court by either corroborating or challenging the economic significance of the underlying factual evidence, consistent with economic theory and principles. Economic analysis, for example, of industry structure and concentration, product or service homogeneity, and behavior inconsistent with independent self-interest are factors that may aid the court in determining the existence, nature, scope, or duration of the alleged agreement. However, agreements to manipulate market prices in the absence of countervailing market factors will most likely have some impact on all market participants. In such situations, courts may consider imposing a presumption based upon record evidence of class impact.183 In the approach being taken by European authorities in follow-on actions, a finding of regulatory accountability in price-fixing market infringements seems to be advocated as rebuttably presumptive evidence of market damage causation. The presumption appears to apply whether the collective action mechanism is opt-out184 or opt-in185. The function of such follow-on civil proceedings is to focus the claimant’s proof on demonstrating the amount of the claimant’s damage and not to require the claimant to comment on the fact or absence of damage to others that may have been collectively harmed. In class proceedings in the United States, a court’s findings of fact and resolution of disputes regarding conspiracy in many ways resemble the European approach in that the analysis of common impact and damage can “follow on” any findings regarding the conspiracy. The element of injury in fact requires the establishment of a causal link186 between the violation and an injury to business or property. A court need not find that a conspiracy was the sole cause of injury, only that it was “a material cause” and evidence of that causation is capable of common proof. When, for example, a conspiracy is found to be based on an agreement among competitors to uniformly and universally manipulate prices or price levels in a market; defendants behaved accordingly; economics confirms that such pricing changes occurred and there are no material nonconspiratorial explanations for the pricing movements, then it would appear reasonable to shift the burden of proof to the defendant to disprove common proof causation.187 As to the final element of antitrust class certification, the court’s obligation is to analyze whether there are economically reliable188 and plausible formulaic methods for estimating damage tied to the conspiracy. This inquiry is largely in the province of economic theory and methodology. After Comcast, courts have reached different conclusions as to its import on the analysis of antirust class economics, both as to form and
118
MICHAEL D. HAUSFELD ET AL.
substance.189 Defendants seem to be of the almost unanimous opinion that Comcast alone has fundamentally altered the landscape of antitrust certification by not only requiring a full-blown Daubert analysis and mini-trial, but by also obligating the court to resolve all differences in economic opinion, whether or not they reflect on scientific reliability. In engaging in the appropriate level of class damage inquiry, courts should be guided by whether the dispute among experts involves the application of appropriate methodology, the construction of an appropriate formulaic model, or whether the proffered analysis is consistent with or contradicted by the record evidence, as found by the court in its evaluation of the issues of conspiracy and impact. Daubert inquiries should not devolve into resolving disputes concerning the results of an expert’s methodology as opposed to its feasibility, plausibility, or workability. Resolution of the former is the province of the jury. Similarly, economic opinions cannot be viewed in a vacuum. An economic opinion, no matter how seemingly sound in the abstract, that is contradicted by the record is of diminished credibility. With respect to Daubert challenges, there is no consensus among courts as to the precise type of such a hearing or argument on class certification. Some courts have held extensive proceedings over several days with the admission of documentary and testimonial evidence. Others have held mini-trials, which in addition to argument and the exchange of evidence, included testimony by live witness. Some courts have required a formal proceeding as an integral part of the class process, while others have not. There is no consensus even as to the extent of Daubert inquiries.190 Regardless of the precise form of the “look” used to address Daubert challenges, courts must not become referees of academic economic debate.191 In the final analysis, the emerging class-certification standards require greater diligence and impose greater obligations on both the parties and the court. Unbundled discovery provides the opportunity and ability to tether class findings and resolution of disputes, noneconomic and economic, to a fully developed fact-based record. The reliability and credibility of all positions of the parties and resolution of all challenges to inferences and opinions should be weighed against their “fit” to the record findings. As well as being scientifically reliable, economic opinion must not be divorced from any class fact findings of the court. Class damage opinions must demonstrate, as opposed to speculate, that there are workable accepted methodologies for estimating aggregate class damage. Parties and courts will have to work harder. Strong cases of liability will likely succeed. Weak cases should and will falter. Rule 23, however,
Antitrust Class Proceedings Then and Now
119
should survive as an effective procedure affording meaningful access to justice for matters of collective wrongs.
NOTES 1. 131 S. Ct. 2541 (2011). 2. 133 S. Ct. 1426 (2013). 3. Yeazell (1997) 687. 4. Ibid. 5. Yeazell (1987) and Pastor (2000) 774. 6. Yeazell, supra n. 2, at 221. 7. Fed. R. Eq. 48 (cited in 42 U.S. (1 How.) xIii, Ivi (1843)). 8. Ibid.; Pastor, supra n. 2, at 785. 9. Equity R. 38, 226 U.S. 659 (1912). 10. See Order of December 20, 1937, 302 U.S. 783 (1937). 11. Alba Conte and Newberg (2007). 12. Fed. R. Civ. P. 23 advisory committee’s note, Note to Subdivision (a) (1938). 13. “Class certification” is a relatively new term. It first appeared in the text of Rule 23 in 1998, although the Supreme Court had by that time employed the term for some two decades (Marcus, 2011) 324, 330. 14. Fed. R. Civ. P. 23 advisory committee’s note (1938). 15. Kaplan (1967) 356, 379. 16. Bone (1990) 213, 284. 17. Richard Marcus, supra n. 4, at 330. 18. Fed. R. Civ. P. 23 advisory committee’s note (1966); see Yeazell (1987) 229 (“Unable to discern the ‘jural relationships’ asserted to lie behind the rules classifications and reluctant to venture into constitutional thickets that the Court had identified, lawyers and judges trod warily around this procedural device.”); see, for example, Shipley v. Pittsburgh & L. E. R. Co., 70 F. Supp. 870, 874876 (W.D. Pa. 1947). 19. See Spence (2002) and Rabiej (2003) 323. 20. See Rabiej, supra n. 10, at 330333. 21. Bassett (2006) 1415. 22. See Rabiej, supra n. 10, at 330333. 23. Fed. R. Civ. P. 23 advisory committee’s note (1966). 24. Ibid. 25. Ibid. 26. Ibid. 27. 417 U.S. 156 (1974). 28. Eisen v. Carlisle & Jacquelin, 41 F.R.D. 147, 151152 (S.D.N.Y. 1966). 29. Eisen v. Carlisle & Jacquelin, 391 F.2d 555, 563 (2d Cir. 1968). 30. Eisen v. Carlisle & Jacquelin, 50 F.R.D. 471, 473474 (S.D.N.Y. 1970). 31. 417 U.S., at 177178. 32. For example, Blackie v. Barrack, 524 F.2d 891, 901 n. 17 (9th Cir. 1975); In re Tableware Antitrust Litig., 241 F.R.D. 644, 648 (N.D. Cal. 2007); In re Dynamic
120
MICHAEL D. HAUSFELD ET AL.
Random Access Memory Antitrust Litig., No. M 02-1486 PJH, 2006 WL 1530166, at *9 (N.D. Cal. June 5, 2006). 33. Wal-Mart, 131 S. Ct., at 2552 & n. 6 (emphasis added). 34. See infra [ ]; [string cite]; Szabo v. Bridgeport Machines, Inc., 249 F.3d 672, 676677 (7th Cir. 2001); In re Initial Pub. Offerings Sec. Litig., 471 F.3d 24 (2d Cir. 2006). 35. Miller (2004), at 2. 36. Miller, at 16. 37. 561 F.2d 434 (3d Cir. 1977). 38. The passage reads: “[i]f, in this case, a nationwide conspiracy is proven, the result of which was to increase prices to a class of plaintiffs beyond the prices which would obtain in a competitive regime, an individual plaintiff could prove fact of damage simply by proving that the free market prices would be lower than the prices paid and that he made some purchases at the higher price. If the price structure in the industry is such that nationwide the conspiratorially affected prices at the wholesale level fluctuated within a range which, though different in different regions, was higher in all regions than the range which would have existed in all regions under competitive conditions, it would be clear that all members of the class suffered some damage, notwithstanding that there would be variations among all dealers as to the extent of their damage.” Ibid., at 455. 39. For a discussion, see Ripley and Glueck (2009). 40. See General Tel. Co. of the S.W. v. Falcon, 457 U.S. 147, 160 (1982). 41. Manual for Complex Litigation § 1.40 (1977). 42. Manual for Complex Litigation § 30.12 (1985). 43. See Local Rules § 23-3 (C.D. Cal.); Local Rules § 23.1(b) (D. D.C.). 44. See Manual for Complex Litigation § 21.14 (4th ed. 2004) (“Courts often bifurcate discovery between certification issues and those related to the merits ….”). See also Larson v. Burlington Northern & Santa Fe Ry. Co., 210 F.R.D. 663, 665666 (D. Minn. 2002) (“the mandate of Rule 1 … [warrants] completing discovery as to the claims of the four named-Plaintiffs, prior to extensive discovery on the merits”); Rodriguez v. Banco Central, 102 F.R.D. 897, 902 (D.P.R. 1984) (“[t]he class determination is preferable before substantial discovery on the merits has been conducted.”); Am. Nurses Ass’n v. Illinois, No. 84-4451, 1986 WL 10382, at *3 (N.D. Ill. September 12, 1986) (“[i]f class certification is denied, the scope of permissible discovery may be significantly narrowed; if a class is certified, defining that class should help determine the limits of discovery on the merits.”). 45. See, for example, In re Rail Freight Fuel Surcharge Antitrust Litig., 258 F.R.D. 167, 172174 (D.D.C. 2009) (noting that permitting defendants to pare their evidence unilaterally would “in effect amend[ ] the Federal Rules of Civil Procedure to create a unique form of discovery for class actions” and observing that class evidence and merits evidence can, at least in price-fixing instances, be “so closely intertwined” that segregating the two would be arbitrary); See also In re Plastics Additives Antitrust Litig., 2004 WL 2743591, at *4 (E.D. Pa. November 29, 2004) (“[t]here will be a substantial overlap between what is needed to prove plaintiffs price-fixing claims, as well as the information needed to establish class-wide defenses, and what is needed to determine whether the elements of class certification are met.”); Bodner v. Banque Paribas, 202 F.R.D. 370, 373 (E.D.N.Y. 2000) (denying bifurcation request and
Antitrust Class Proceedings Then and Now
121
noting that “discovery on the merits, reasonably structured, is essential to determining whether class certification is appropriate”); Manual for Complex Litigation (Fourth) § 21.14. 46. Fed. R. Civ. P. 23, advisory committee’s note (2003). 47. Ibid. 48. Ibid. (emphasis added). 49. Ibid. 50. See, for example, Weiss v. Regal Collections, 385 F.3d 337, 347 & n. 17 (3d Cir. 2004) (“[o]f course, the federal rules do not require certification motions to be filed with the class complaint, nor do they require or encourage premature certification determinations. … Allowing time for limited discovery supporting certification motions may also be necessary for sound judicial administration. See … 7B Wright and Miller, Fed. Practice and Procedure § 1785, at 107 (“[t]he [certification] determination usually should be predicated on more information than the complaint itself affords.”).”); Gariety v. Grant Thornton, LLP, 368 F.3d 356, 365368 (4th Cir. 2004) (overturning class-certification determination that simply accepted plaintiffs’ allegations as true for purposes of Rule 23 and urging the district court to consider the fruits of discovery on remand). 51. Larkin and Coukos (2004), at 4. 52. See, for example, Szabo v. Bridgeport Machines, Inc., 249 F.3d 672, 675-78 (7th Cir. 2001); In re Initial Pub. Offering Sec. Litig., 471 F.3d 24, 26 (2d Cir. 2006). 53. 552 F.3d 305 (3d Cir. 2008). 54. Hydrogen Peroxide, 552 F.3d, at 305. 55. Ibid., at 310. 56. Ibid., at 310311; Fed. R. Civ. P. 23(b)(3). 57. Hydrogen Peroxide, 552 F.3d, at 312. 58. Ibid., at 316. 59. Ibid., at 316317. 60. Ibid., at 317 & n. 17. 61. Ibid., at 318. 62. Ibid., at 319. 63. Ibid., at 320. 64. Ibid., at 307. 65. 256 F.R.D. 82 (D. Conn. 2009). 66. Ibid., at 95. 67. 471 F.3d 24 (2d Cir. 2006). 68. Ibid., at 101. 69. 267 F.R.D. 291 (N.D. Cal. 2010). 70. Ibid., at 313; see also In re Southeastern Milk Antitrust Litig., No. 2:08-md1000, 2010 WL 3521747, at *10 (E.D. Tenn. September 7, 2010) (“It appears to the Court that once again the defendants are asking the Court to resolve certain of the issues in this litigation on the merits. A class certification motion is not the appropriate vehicle for accomplishing that. While the Court must rigorously examine the allegations of the plaintiffs in evaluating a class certification motion, the plaintiffs are not required, at this stage of the litigation, to establish that they can succeed on the merits.”). 71. 131 S. Ct., at 2547.
122
MICHAEL D. HAUSFELD ET AL.
72. Ibid., at 2548. 73. As for the second question, the Court held that claims for monetary relief may not be certified under Rule 23(b)(2), at least where, as with the Dukes plaintiffs, each class member would be entitled to an individualized award. 131 S. Ct., at 25572558 (prohibiting claims for monetary relief under Rule 23(b)(2) if the monetary relief is not incidental to the injunctive or declaratory relief). 74. Ibid., at 2545. 75. Ibid. 76. Ibid., at 2556 (emphasis added and quotation marks omitted). A literal application of this expression would virtually render class certification in antitrust conspiracy cases mandatory. Although the issue and evidence of conspiracy will always be predominantly common to a class, it is doubtful that its presence alone would be a sufficient foundation up on which to grant certification. Issues of impact, damage, nexus and methodology would still need to be established. 77. Ibid., at 2551 (emphasis in original). 78. Ibid. 79. Ibid., at 25512552 & n. 6. 80. Ibid., at 2554. The Court provided no guide as to what type of format or any Daubert gatekeeper inquiry would be appropriate or necessary. 81. Ibid., at 2565 (citation omitted). 82. Ibid., at 2566. 83. Ibid., at 2567. 84. Ibid., at 25562557 (quotation marks and citation omitted). 85. See, for example, In re Rail Freight Fuel Surcharge Antitrust Litig., 287 F.R.D. 1 (D.D.C. 2012); In re Titanium Dioxide Antitrust Litig., 284 F.R.D. 328 (D. Md. 2012); In re Vitamin C Antitrust Litig., 279 F.R.D. 90 (E.D.N.Y. 2012); In re Aftermarket Auto. Lighting Prods. Antitrust Litig., 276 F.R.D. 364 (C.D. Cal. 2011); In re Chocolate Confectionary Antitrust Litig., No. 1:08-MDL-1935, 2012 U.S. Dist. LEXIS 174681 (M.D. Pa. December 7, 2012); Allen v Dairy Farmers of Am. Inc. Case No. 5:09-cv-230, 2012 U.S. Dist. LEXIS 164718 (D. Vt. November 19, 2012). 86. See ibid., at *5 n. 6 (recognizing that “[t]he Supreme Court confirmed our interpretation of the Rule 23 inquiry in Wal-Mart”). 87. Ibid., at *5. 88. Ibid., at *14. 89. Ibid. (emphasis added). 90. Ibid. 91. Ibid., at *6, *7, *12,*19 n. 13, *20. 92. Ibid., at *19 n. 13. 93. Although the Supreme Court in Comcast, in accepting certiorari, undertook to hear the issue of what is an “admissible” economic model, it decided the matter on other grounds. There is therefore, no definitive precedent mandating that plaintiffs’ expert must, at the certification stage, present a workable damage model as opposed to presenting merely a theoretically acceptable economic model. Despite this uncertainty, it would seem that the preferable practice would be to present a workable model at certification, with the express understanding that it could be modified or revised for trial.
Antitrust Class Proceedings Then and Now
123
94. Ibid. (emphasis added). 95. Ibid., at *10, *13, *19 n. 13. 96. Ibid., at *19 n. 13 (emphasis added); see ibid., at *13 (reasoning that a district court’s choice to credit one expert’s theory of common impact over another’s cannot be clearly erroneous if both experts’ views are “permissible” interpretations of the evidence); See also In re Currency Conversion Fee Antitrust Litig., 264 F.R.D. 100, 115 (S.D.N.Y. 2010) (expert dispute over which of two “but for” worlds would have existed absent conspiracy does not preclude certification); In re Ethylene Propylene Diene Monomer (EPDM) Antitrust Litig., 256 F.R.D. 82, 100-02 (D. Conn. 2009) (experts’ competing views of market structure and the “but for” world present merits questions). 97. Ibid., at *15 (emphasis added); see ibid., at *18 (characterizing the inquiry as whether plaintiffs’ damages presentation is “amenable” to the class action mechanism). 98. Ibid., at *18 (“Given the inherent difficulty of identifying a ‘but-for world,’ we do not require that damages be measured with certainty, but rather that they be demonstrated as ‘a matter of just and reasonable inference’.” (citation omitted)). 99. Ibid. Not surprisingly, Comcast attempted to compare the damages model proffered by the plaintiffs’ expert, Dr. James McClave, with the expert sociologist’s model in Wal-Mart asserting that Dr. McClave’s work was similarly deficient. The Third Circuit squarely rejected that analogy, though, noting the unique features of Wal-Mart and explaining its limited application: “The factual and legal underpinnings of Wal-Mart which involved a massive discrimination class action and different sections of Rule 23 are clearly distinct from those of this case. Wal-Mart therefore neither guides nor governs the dispute before us.” Ibid., at *17 n. 12. 100. Ibid. 101. Ibid., at 14301431. 102. Ibid., at 1431. 103. Ibid., at 1433. 104. Ibid., at 1437. 105. Harris v. comScore, Inc., No. 11 C 5807, 2013 U.S. Dist. LEXIS 47399, at *27 n. 6 (N.D. Ill. April 2, 2013). 106. See In re Urethanes Antitrust Litig., MDL No. 1616, 2013 U.S. Dist. LEXIS 69784, at *39*40 (D. Kan. May 15, 2013) (denying decertification despite Comcast, saying “Dr. McClave gave his opinion that the conspiracy alleged by plaintiffs a horizontal price-fixing conspiracy impacted nearly every class member because prices during the alleged conspiracy period exceeded those that would have prevailed absent that conspiracy, which competitive prices were determined from an analysis of prices during a post-conspiracy benchmark period”); In re High-Tech Employees Antitrust Litig., No.: 11-CV-02509-LHK, 2013 U.S. Dist. LEXIS 49784, at *92 (N.D. Cal. April 4, 2013) (court declined to certify classes because of the concern that the “proposed classes may be defined so broadly as to include large numbers of people who were not necessarily harmed by Defendants’ allegedly unlawful conduct”). For pre-Comcast cases making a similar point (both in and out of the antitrust context), see Butler v. Sears, Roebuck & Co., 702 F.3d 359, 361 (7th Cir. 2012); Messner v. Northshore Univ. HealthSystem, 669 F.3d 802, 823824 (7th Cir. 2012) Kohen v. Pac. Inv. Mgmt. Co., LLC, 571 F.3d 672, 677 (7th
124
MICHAEL D. HAUSFELD ET AL.
Cir. 2009), cert. denied, 130 S. Ct. 1504 (2010); DG ex rel. Stricklin v. Devaughn, 594 F.3d 1188, 1198 (10th Cir. 2010); Mims v. Stewart Title Guar. Co., 590 F.3d 298, 308 (5th Cir. 2009); In re Rail Freight Fuel Surcharge Antitrust Litig., 287 F.R.D. 1, 40 (D.D.C. 2012); Yarger v. ING Bank, 285 F.R.D. 308, 312 (D. Del. 2012); Brooks v. GAF Materials Corp., 284 F.R.D. 352, 363 (D.S.C. 2012); Connor B. ex rel. Vigurs v. Patrick, 272 F.R.D. 288, 295 (D. Mass. 2011); In re Light Cigarettes Mktg. Sales Practices Litig., 271 F.R.D. 402, 419 (D. Me. 2010); Hamilton v. First American Title Ins. Co., 266 F.R.D. 153, 163 (N.D. Tex. 2010); Sheet Metal Workers Local 441 Health & Welfare Plan v. GlaxoSmithKline, PLC, No. 04-5898, 2010 WL 3855552, at *93 (E.D. Pa. September 30, 2010). 107. Comcast Corp. v. Behrend, No. 11-864, Transcript of Oral Argument, 22 (November 5, 2012) (“Comcast Transcript”). 108. As expressed by Justice Alito, the question of probative value in Comcast apparently subsumed the Daubert question. Ibid., at 25, 2627. See, for example, Justice Scalia’s footnote noting that a plaintiff will not have established “the requisite commonality of damages” unless it “plausibly” showed that the theory of damage had either been “the same in all counties or was irrelevant to the effect upon the ability to charge supra-competitive prices.” Comcast Corp. v. Behrend, 133 S. Ct. 1426, 1435 n. 6 (2013). 109. Comcast, 133 S. Ct. at 1433. 110. There is not yet consensus concerning the quantity contemplated by “virtually all.” We do not answer that question here. 111. Supra Section “The Rise and Fall of Bifurcated Discovery” 112. Wooldridge (2002), at 3. 113. Rubinfeld (2011), at 314. 114. Greene (2012), at 6. 115. Hume (1748). 116. Popper (2005), at 3. 117. Popper, at p. 7 (emphasis in the original). 118. Swanson (1960), at 1484. 119. Christ and Rausser (1973), at 273. 120. Shapiro (1973), at 252. 121. Sullivan, Timmerman, and White (2001). 122. Traditional statistical tests, such as a t-test, report a confidence level. An estimated statistical effect of an explanatory variable on a dependent variable is said to be statistically significant at a 95% level of confidence if there is only a 5% probability that the chosen sample exhibits the effect (due to random variation in the data), even though in reality there is no true effect. With many repeated statistical testing of the same dataset (in this case the dataset of stock returns), researchers will eventually find statistical variations that fall into this 5% margin of error, but will be reported by the t-test and the researcher as statistically significant (this is known as a Type I error). 123. Hayashi (2000), at 33: “Sometimes the maintained hypothesis is somewhat loosely referred to as ‘the model’. We say that the model is correctly specified if the maintained hypothesis is true” (emphasis added). 124. It should be noted that in practice, the PEs are often asked to assume the existence of liability and proceed directly to issues of impact and damages. Even in
Antitrust Class Proceedings Then and Now
125
such cases, however, the analysis of impact often takes into account evidence of defendants’ conduct and statements. See, for example, In re Aftermarket Auto. Lighting Prods. Antitrust Litig., 276 F.R.D. 364, 371 (C.D. Cal. 2011) (PE considered “the existence of a pricing structure adopted by Defendants” as part of the impact analysis). 125. Greene, at pp. 2930. 126. Regarding the pricing system of the citric acid price-fixing cartel: “In some industries, like retail stores, the price listed on the item or shelf is the actual price a buyer will pay at check-out. However, in many other industries, like automobiles, most consumers regard the list price as simply the highest price a seller hopes to get for the product, whereas after searching for alternative offer prices most buyers will purchase an automobile at a negotiated discount. List prices for most industrial commodities like citric acid follow the automobile model of pricing.” Connor (2007), at 156. 127. A conspiracy that manipulates those list prices by adjusting the starting point for negotiations was recognized in In re Ethylene Propylene Diene Monomer (EPDM) Antitrust Litigation, 256 F.R.D. 82, 90 (U.S.D.C. Conn. 2009). 128. In In re Plastic Additives Antitrust Litigation, 2010 U.S. Dist. LEXIS 90133, at *24*26 (U.S.D.C. E.D. Pa. 2010) (at 12), the court ruled against class certification because prices for some individual transactions were observed not to increase after the implementation of supposedly conspiratorial list price announcements. 129. Wooldridge, at p. 3 (emphasis in the original). 130. Ibid. 131. Ibid., at p. 4. 132. The term “natural experiment” is used because, of course, econometricians cannot conduct these experiments in the laboratory; rather we exploit instances where they have occurred naturally in the real world. See, Coleman and Langenfeld (2008). 133. Not all classwide damage methodologies will necessarily require a regression model. For instance, a case that alleges the imposition of a conspiratorial tax, where the entirety of that tax is alleged to represent the damage to each class member, may require simply the summation of all such taxes in order to calculate classwide damages. 134. Kuhn (1970). 135. Christ and Rausser, at p. 274. 136. Hayashi at 33: “Sometimes the maintained hypothesis is somewhat loosely referred to as ‘the model.’ We say that the model is correctly specified if the maintained hypothesis is true” (emphasis added). 137. Christ and Rausser, at p. 275: “the selected maintained hypotheses … isolate a still smaller set which represents the testable hypotheses.” 138. “Neoclassical economics assumes that man, homo economicus if you will, has consistent preferences if he prefers apples to oranges and oranges to nuts, then he prefers apples to nuts and makes choices that maximize his utility at all times.” An alternative paradigm is that of “behavioral economics.” Under this paradigm, agents “have only limited or ‘bounded’ rationality and therefore sometimes make choices that satisfy their preferences but do not maximize their utility. This
126
MICHAEL D. HAUSFELD ET AL.
idea flows from the observation that humans do not possess the cognitive capacity required to process all the information necessary to maximize utility at all times; instead, they use heuristics or ‘shortcuts’ to make decisions that sometimes fail to jibe with the predictions of neoclassical economics.” However, although its influence in legal scholarship is growing, “behavioral economics has not yet affected judicial decisions in the United States in any substantive area of law,” Ginsburg and Moore (2010), at 89. 139. Baker and Rubinfeld (1999), at 391. See also Brander and Ross (2006), at 351: “… estimation of reduced-form price equations is the preferred and most commonly applied method for damage estimation by economists in price-fixing cases.” 140. See ABA Section of Antitrust Law (2010). 141. Supra, citing Wooldridge. 142. Supra, citing Rubinfeld, at p. 314 and Greene, at p. 6. 143. Kennedy (2002), at 577: “… check that the results make sense. Are the signs of coefficients as expected? Are important variables statistically significant? Are coefficient magnitudes reasonable? Are the implications of the results consistent with theory?” 144. Greene, at p. 1. 145. See, for example, Greenhaw v. Lubbock County Beverage Assocs., 721 F.2d 1019, 1028 (5th Cir. 1983) (“it was more reasonable to apply the market average overcharge figure in calculating damages rather than separate figures for each store in which purchases were made”); LCD II, 267 F.R.D., at 605 (noting that averaged and aggregated data are used by courts in granting class certification); In re Static Random Access Memory (SRAM) Antitrust Litig., 264 F.R.D. 603, 614 (N.D. Cal. 2009) (same); In re Nifedipine Antitrust Litig., 246 F.R.D. 365, 371 (D.D.C. 2007) (“the Court can find no fault in [plaintiffs’ expert’s] consideration of aggregate changes in price”) (emphasis in original); Cardizem, 200 F.R.D., at 345 (use of average prices did not undermine defendants’ due process rights); NASDAQ, 169 F.R.D., at 523 (proper to show that the price range was affected “generally” by defendants’ prices); In re Pressure Sensitive Labelstock Antitrust Litig., 03-MDL1556, 2007 WL 4150666, at *20 (M.D. Pa. November 19, 2007) (“[d]efendants argue that [plaintiffs’ expert’s] proposed multiple regression analysis is flawed because it yields an average overcharge. … This will not defeat Plaintiffs’ class certification motion because Plaintiffs do not have to present a precise damages formula at this stage”); In re Scrap Metal Antitrust Litig., 02 CV 0844, 2006 WL 2850453, at *16 (N.D. Ohio, September 30, 2006), aff’d, 527 F.3d 517 (6th Cir. 2008), cert. denied, 129 S. Ct. 673 (2009) (in situation where plaintiffs’ expert used average price spreads to compute average undercharges, court upheld jury verdict for plaintiffs; “[r]ecognizing that antitrust cases present damages questions that often are not amendable to concrete mathematical calculations, courts have regularly recognized averaging of overcharges as a valid method”); Presidio Golf Club of S.F., Inc. v. Nat’l Linen Supply Corp., No. C 71-431 SW, 1976 WL 1359, at *5 (N.D. Cal. December 30, 1976) (only an “illustration of generalized injury” is needed for class certification); Gordon v. Microsoft Corp., No. MC 00-5994, 2003 WL 23105550, at *3 (Minn. Dist. Ct. December 15, 2003) (use of average overcharge analysis was permissible in determining what classwide damages would have been in the absence of defendants’ conduct).
Antitrust Class Proceedings Then and Now
127
146. In a seminal article on the economics of information George Stigler notes that “it is important to emphasize immediately the fact that dispersion is ubiquitous even for homogenous goods.” Stigler (1961), at 213. 147. Nieberding and Cantor (2007), at 7980. 148. Hartman and Doane (1987), at 352: “It is so named because it measures how the attributes, or ‘hedonic’ characteristics, of goods and services are commonly valued in consumption or in production. Its usefulness for class certification procedures lies in its ability to identify and measure the commonality in a group of apparently heterogeneous products, services, or individuals. When such commonalities are identified and measured, the courts can support class certification and can calculate the common effect of illegal actions, correcting for any apparent heterogeneity.” 149. In addition, see ABA Section of Antitrust Law (2010). 150. Fisher (1986), at 278. 151. Reed v. Advocate Health Care, 268 F.R.D. 573 (N.D. Ill. 2009). 152. Ibid., at 593. 153. For example, R-squared statistics of 26% or lower in regressions testing hypotheses on market concentration are cited by Carter (1978); another litigation research article on examining employment discrimination reports a series of regressions with R-squareds lower than 80%, Whiteside and Narayanan (1989), at 405, Table 4; Rubinfeld (at p. 345), presents a regression with an R-squared of 56% as an example, when explaining the utilization of multiple regressions in antitrust analysis. 154. Sorensen (2000), at 833850; Adams (1997) and Lach (2002). 155. Adams, at p. 801. 156. Barron, Taylor and Umbeck (2004). 157. Rubinfeld, at p. 313. 158. Greene, at p. 6: “No model could hope to encompass the myriad essentially random aspects of economic life.” 159. Rubinfeld, at p. 314. 160. Rubinfeld, at p. 314. See also, Baker and Rubinfeld, at p. 396. 161. Although, it should be noted that the omission of variables that determine price from the model, even if unrelated to the included variables, will inflate the standard errors on the estimated coefficients on the included variables and therefore decrease the precision of the model’s estimates. 162. Baker and Rubinfeld, at p. 391. See also Nieberding (2006) and ABA Section of Antitrust Law (2010). 163. This is in contrast to a “structural model” which estimates the supply and demand curves separately. Part of the PE’s assignment will be to decide between the appropriateness of a reduced form model versus a structural model both are used in antitrust. Reduced form models are more commonly used and this section will focus on such models, although the general principles described herein are applicable to structural models, with some technical modifications. 164. A similar method is known as the predictive approach. The predictive approach estimates the model for the benchmark transactions first then uses the estimated coefficients from that model to predict but-for prices for the class
128
MICHAEL D. HAUSFELD ET AL.
transactions. The difference between actual prices and the estimated but-for prices facilitates the quantification of damages. ABA Section of Antitrust Law (2010). 165. Rubinfeld describes, using the example of a patent infringement case, how it is important to have sufficient data to accurately perform a “but-for” analysis: “For example, if the question at issue in a patent infringement case is what price the plaintiff’s product would have been but for the sale of the defendant’s infringing product, sufficient data must be available to allow the expert to account statistically for the important factors that determine the price of the product,” (emphasis added) Rubinfeld, at p. 311. See also Brander and Ross, at pp. 353354: “The investigator will need data from both inside and outside the cartel period …. In addition to needing data of high enough quality, we need it to be of sufficient quantity. Precise estimates require a large number of observations.” 166. In Plastic Additives 2010 U.S. Dist. LEXIS 90135 at *67, the court rejected criticisms of a model’s specification because those criticisms were “based solely on economic theory.” 167. Wooldridge (2009), at 325: “From a practical perspective, outlying observations can occur for two reasons. The easiest case to deal with is when a mistake has been made in entering the data. Adding extra zeros to a number or misplacing a decimal point can throw off the [regression] estimates, especially in small sample sizes.” 168. Wooldridge (2009), at p. 325: “Outliers can also arise when sampling from a small population if one or several members of the population are very different in some relevant aspect from the rest of the population. The decision to keep or drop such observations in a regression analysis can be a difficult one … [regression] results should probably be reported with and without outlying observations in cases where one or several data points substantially change the results.” 169. Wooldridge (2009), at p. 677: “Good papers in the empirical social sciences contain sensitivity analysis … [discussing outliers] If some observations are much different from the bulk of the sample … do your results change much when those observations are excluded from the estimation?” 170. Brander and Roos, at pp. 342343 and at p. 346. 171. Economist and Nobel laureate, Wassily Leontief, criticizes economic modeling based on “preoccupation with imaginary, hypothetical, rather than with observable reality.” (Leontief, 1971), at 3. 172. For example, consider a regression model designed to measure the effect of an alleged conspiracy on prices. In such a model, it is important to control for the effect of demand on prices. However, if one includes the quantity of the product consumed as an attempt to control for demand, this may lead to problems of endogeneity derived from reverse causality. Simply, the quantity of a product consumed is itself affected by price (consumers buy more when the product is cheaper), as well as being an indication of how much of the product is inherently demanded by consumers. If not accounted for, this could lead to a biased estimate of the effect of demand on prices (because the coefficient on the quantity variable reflects both the effect of demand on price and the effect of price on quantity) and unreliability in the regression model in general. More technically, an explanatory variable is endogenous if it is correlated with the error term in the regression. This can be due to reverse causality from the dependent variable to a specified (endogenous)
Antitrust Class Proceedings Then and Now
129
explanatory variable, as just discussed, but other sources of endogeneity include omitted variable bias and nonclassical measurement error; Wooldridge, at pp. 5051. Econometricians use “instrumental variables” as an identification strategy to filter out the endogenous relationship and identify the causal effect of an explanatory variable on the dependent variable. 173. The classical assumptions for an ordinary least squares regression model to produce reliable tests of statistical significance include that the residuals (the variation in price for each data point, unexplained by the model) are random, in that they are not correlated with each other and their variance is not correlated with explanatory variables included in the regression. Heteroscedasticity is a potential problem in regression models where the variance of the residuals is not constant across observations, but instead is correlated with the explanatory variables. If not accounted for, heteroscedasticity can lead to inaccurate standard errors and, therefore, an erroneous inference concerning the statistical significance of coefficient estimates, see Greene, at pp. 268269. Serial correlation (or autocorrelation) is a situation common in time-series data where the residuals in a regression are correlated with each other over time i.e., if the residual is high in one time period, it is more likely to be high in the next period. This can occur if there is an explanatory variable omitted from the regression that varies over time. As with heteroscedasticity, if not corrected for, serial correlation may lead to inaccurate standard errors and unreliable tests of statistical significance, see Greene, at pp. 903906. 174. Plastic Additives 2010 U.S. Dist. LEXIS 90135 at *61. 175. Ibid., at *69: “For example, one customer may be a ‘very aggressive negotiator,’ and so even as costs go up, that customer is able to ‘extract lower prices’.” 176. Ibid.: “in a hypothetical textbook example … [an] increase in cost will increase prices … in the real world … there are many reasons why … the relationship between the two variables is not positive.” Defendants’ expert argued that the regression could be used to estimate and control for each individual customer’s demand responses and the possibility that some aggressive negotiators may be able to extract lower prices as costs increased. 177. Varian (1992), at 202203, describes how the reduced form model is derived from the “demand and supply system” that defines the “demand and supply for some good … its price … [in terms of] … variables that affect supply and demand …. The reduced-form parameters can be used to predict how the equilibrium price will change as the [supply and demand] variables change,” (emphasis added). Economists define the equilibrium price as the “price at which the quantity of a good supplied [by the market] is equal to the quantity demanded [by the market]” (Black, 2012, at 150). 178. Varian, at p. 219, describes how the equilibrium price, which he previously established is the focus of a reduced form model, is determined by industry supply and demand: “The industry supply function measures the total output supplied at any price. The industry demand function measures the total output demanded at any price. An equilibrium price is a price where the amount demanded equals the amount supplied” (emphasis in the original). 179. Varian, at p. 152 describes how the “aggregate demand for [a] good” is derived from a summation of the individual consumers’ demand functions;
130
MICHAEL D. HAUSFELD ET AL.
Mas-Colell, Whinston and Green (1995), at 105109, describes how aggregate demand is determined by a sum of consumers’ individual demand functions. 180. See, for example, In re High-Tech Employee Antitrust Litig., 11-CV-02509LHK, 2013 WL 5770992, at *40 (N.D. Cal. October 24, 2013) where the court noted: “Before the Court turns to an analysis of the competing methodologies … the Court notes that the importance of these statistical models is diminished in light of the extensive documentary evidence that supports Plaintiffs’ theory of impact. In other contexts, courts have long noted that statistical and anecdotal evidence must be considered in tandem. … After all, class certification requires a holistic, qualitative assessment … the class certification analysis is not ‘bean counting’.” 181. See, for example, Judge Weinstein’s Opinion in Tobacco which combined class certification, summary judgment and Daubert determinations in a single proceeding. Schwab v. Philip Morris USA, Inc., 449 F. Supp. 2d 992 (E.D.N.Y. 2006) (later reversed on other grounds). There is also a further potential consequence of merits-type findings in the course of class procedures with respect to separate subsequent summary judgment motions. If a court makes such findings in certifying a class, would it not be inherently inconsistent for defendants to prevail on a motion for summary judgment since the court would have already decided that there are material facts in dispute? 182. For a divergent view as to the elements of class certification, see the opinions of the Supreme Court of Canada in Pro-Sys Consultants Ltd. v. Microsoft Corp. [2013] S.C.R. 57 (Can.), Infineon Tech. AG v. Option Consommateurs [2013] S.C.R. 59 (Can.), and Sun-Rype Prods. Ltd. v. Archer Daniels Midland Co. [2013] S.C.R. 58 (Can.). 183. Although some courts have referenced such a presumption, few have actually applied it. See, for example, In re Chocolate Confectionary Antitrust Litig., 289 F.R.D. 200 (M.D. Pa. 2012). In the chronology of class certification, those decisions did not make findings of fact a court is now expected to make in the certification process. This new dynamic may require at least a revisiting of this practice. See also In re Hydrogen Peroxide Antitrust Litig., 552 F.3d 305, 325 (3d Cir. 2008), where the Court held that “the question at [the] class certification stage is whether, if such impact is plausible in theory, it is also susceptible to proof at trial through available [economic] evidence common to the class.” Such a “plausible” formulation of class impact, when bonded with the facts specific to the claim and contemporaneous industry information would be consistent with the application of a presumption. 184. Binds all class members unless they specifically exclude themselves from the litigation in an appropriate court approved form. 185. Binds only those entities that specifically agree, in accordance with a court approved procedure, to be included as litigant claimants. 186. See In re Nexium (Esomeprazole) Antitrust Litig., CIV.A. 12-MD-02409, 2013 WL 6019287, at *14 (D. Mass. November 14, 2013) where the court states that “Comcast has not changed the rule on what is required for damages models in establishing Rule 23(b)(3) predominance. Comcast simply requires the moving party to present a damages model that directly reflects and is linked to an accepted theory of liability under Rule 23(b)(3)” (internal citations omitted). 187. Under such a presumption, the burden of proof would shift to defendants to demonstrate, for example, in situations where it is asserted that pricing was
Antitrust Class Proceedings Then and Now
131
individually negotiated, the precise identity of the class members who engaged in such negotiations, the transaction price actually negotiated and the total number of class members so affected. 188. In re High-Tech Employee Antitrust Litig., 11-CV-02509-LHK, 2013 WL 5770992, at *19 (N.D. Cal. October 24, 2013) notes, “Ultimately, the Court is not tasked at this phase with determining whether Plaintiffs will prevail on these theories. Rather, the question is narrower: whether Plaintiffs have presented a sufficiently reliable theory to demonstrate that common evidence can be used to demonstrate impact. The Court finds that, based on the extensive documentary evidence, economic theory, data, and expert statistical modeling, Plaintiffs’ methodology demonstrates that common issues are likely to predominate over individual issues.” In re U.S. Foodservice Inc. Pricing Litig., 729 F.3d 108, 123, n. 8 (2d Cir. 2013) states that “In Comcast, the Supreme Court held that courts should examine the proposed damages methodology at the certification stage to ensure that it is consistent with the classwide theory of liability and capable of measurement on a classwide basis. [Comcast v. Behrend, 133 S. Ct. 1426, 14331435 (2013)] (finding that plaintiffs’ damages ‘model failed to measure damages from the particular antitrust injury on which petitioners’ liability in this action is premised’). As discussed in Part B, infra, the district court carefully examined plaintiffs’ damages model, finding it appropriate and feasible to redress the common harms alleged, and therefore did not abuse its discretion in determining that common issues predominate.” 189. Compare In re Rail Freight Fuel Surcharge Antitrust Litig., __ F.3d __, 2013 WL 4038561 (D.C. Cir. 2013) (remanding class certification decision in light of Comcast, and explaining that Comcast requires a “hard look at the soundness of statistical models that purport to show predominance”), with In re Cathode Ray Tube (CRT) Antitrust Litig., MDL No. 1917, at *14 (N.D. Cal. September 24, 2013) (finding that “Defendants’ arguments misread Comcast and relevant precedent to require proof of the merits of their damages claim as opposed its methodology at the class certification stage”). 190. See Transcript of Supreme Court Justices’ oral comments in Comcast cited at subsection “Supreme Court” supra. 191. The function of the law in a civil proceeding is to provide a forum in which a fact finder may determine which party’s explanation of events is more likely descriptive of what happened. It is not the role of a court to declare the truth on scientific disagreement. See also, for example, Appelbaum (2013).
REFERENCES ABA Section of Antitrust Law. (2010). Econometrics and regression analysis. In Proving antitrust damages: Legal and economic issues (2nd ed.), ABA. Adams, A. (1997). Search costs and price dispersion in a localized, homogeneous product market: Some empirical evidence. Review of Industrial Organization, 12, 801–808. Alba Conte, A., & Newberg, H. B. (2007). Newberg on class actions § 1:9 (4th ed.). Appelbaum, B. (2013). Economist clash on theory, but will still share the Nobel. New York Times, October 14.
132
MICHAEL D. HAUSFELD ET AL.
Baker, J., & Rubinfeld, D. (1999). Empirical methods in antitrust litigation: Review and critique. American Law and Economics Association, 391. Barron, J., Taylor, B., & Umbeck, J. (2004). Number of sellers, average prices, and price dispersion. International Journal of Industrial Organization, 22, 1041–1066. Bassett, D. L. (2006). Constructing class action reality. Brigham Young University Law Review, 2006, 1434–1436. Black, J. (2012) Oxford dictionary of economics. Oxford: Oxford University Press. Bone, R. G. (1990). Personal and impersonal litigative forms: Reconceiving the history of adjudicative representation. Boston University Law Review, 70, 213, 284. Brander, J., & Ross, T. (2006). Estimating damages from price-fixing. In S. Pitel (Ed.), Litigating conspiracy: An analysis of competition class actions (p. 351). Toronto: Irwin Law. Carter, J. (1978). Collusion, efficiency and antitrust. Journal of Law and Economics, 21(2) (October), 435–444. Christ, C., & Rausser, G. (1973). Discussion on Shapiro and Fromm. American Journal of Agricultural Economics, 55(2) (May), 271–279. Coleman, M., & Langenfeld, J. (2008). Natural experiments, issues in competition law and policy. ABA Section of Antitrust Law, 1, 743–772. Connor, J. (2007). Global price fixing (2nd ed.). New York, NY: Springer. Fisher, F. (1986). Statisticians, econometricians, and adversary proceedings. Journal of the American Statistical Association, 81(394) (June), 277–286. Ginsburg, D., & Moore, D. (2010). The future of behavioral economics in antitrust jurisprudence. Competition Policy International, 6(1) (Spring), 92–93. Greene, W. (2012). Econometric analysis (7th ed.). Prentice Hall. Hartman, R., & Doane, M. (1987). The use of hedonic analysis for certification and damage calculations in class action complaints. Journal of Law, Economics and Organization, 3(2) (Autumn), 351–372. Hayashi, F. (2000). Econometrics, Princeton University Press. Hume, D. (1748). An enquiry concerning human understanding. Kaplan, B. (1967). Continuing work of the civil committee: 1966 amendments of the Federal Rules of Civil Procedure (I). Harvard Law Review, 81, 356, 379. Kennedy, P. (2002). Sinning in the basement: What are the rules? The ten commandments of applied econometrics. Journal of Economic Surveys, 16(4), 570–589. Kuhn, T. (1970). The structure of scientific revolutions (2nd ed.). Chicago, IL: The University of Chicago Press. Lach, S. (2002). Existence and persistence of price dispersion: An empirical analysis. The Review of Economics and Statistics, 84(3), 433–444. Larkin, J. D., & Coukos, P. (2004, February). Rule 23 Revisions: Ten Tips for Plaintiffs’ Lawyers, Retrieved from http://www.impactfund.org/downloads/Resources/Rule23 RevisionTips.pdf Leontief, W. (1971). Theoretical assumptions and nonobserved facts. The American Economic Review, 61(1) (March), 1–7. Marcus, R. (2011). Reviving judicial gatekeeping of aggregation: Scrutinizing the merits on class certification. George Washington Law Review, 79, 324, 330. Mas-Colell, A., Whinston, M., & Green, J. (1995). Microeconomic theory (Oxford Student Edition., pp. 105–109). Oxford: Oxford University Press. Miller, G. P. (2004). Review of the merits in class action certification. The Berkeley Electronic Legal Series.
Antitrust Class Proceedings Then and Now
133
Nieberding, J. (2006). Estimating overcharges in antitrust cases using a reduced-form approach: Methods and issues. Journal of Applied Economics, IX(2) (November). Nieberding, J., & Cantor, R. (2007). Price dispersion and class certification in antitrust cases: An economic analysis. Journal of Legal Economics, 14(2), 62–84. Pastor, N. M. (2000). Equity and settlement class actions: Can there be justice for all in Ortiz v. Fibreboard. American University Law Review, 49, 783–784. Popper, K. (2005/1935). The logic of scientific discovery. In Routledge classics. Taylor & Francis e-Library. Rabiej, J. K. (2003). The making of class action Rule 23 – What were we thinking? Mississippi College Law Review, 24, 330–333. Ripley, A. R., & Glueck, M. J. (2009). In re Hydrogen Peroxide antitrust litigation bleaches clean the class certification standard. The Antitrust Source, 8(3), (February). Rubinfeld, D. (2011). Reference guide on multiple regression. In Reference manual on scientific evidence (3rd ed.). Federal Judicial Center. Shapiro, H. (1973). Is verification possible? The evaluation of large econometric models. American Journal of Agricultural Economics, 55(2) (May), 250–258. Sorensen, A. (2000). Equilibrium price dispersion in retail markets for prescription drugs. The Journal of Political Economy, 108(4), 833–850. Spence, S. T. (2002). Looking back … in a collective way: A short history of class action law. Business Law Today, 11(6), (July/August). Retrieved from http://apps.americanbar.org/ buslaw/blt/ 2002-07-08/spence.html Stigler, G. (1961). The economics of information. The Journal of Political Economy, LXIX(3), 213–225. Sullivan, R., Timmerman, A., & White, H. (2001). Dangers of data mining: The case of calendar effects in stock return, Journal of Econometrics, 105, 249–286. Swanson, E. (1960). Are agricultural economists becoming mechanics? Journal of Farm Economics, 42(5), Proceedings of the annual meeting of the American Farm Economic Association, December, 1483–1486. Varian, H. (1992). Microeconomic analysis (pp. 202–203). (3rd ed.). W. W. Norton. Whiteside, M., & Narayanan, A. (1989). Reverse regression, collinearity, and employment discrimination. Journal of Business & Economic Statistics, 7(3) (July), 403–406. Wooldridge, J. M. (2002). Econometric analysis of cross section and panel data. MIT Press. Wooldridge, J. M. (2009). Introductory econometrics: A modern approach (4th ed.). SouthWestern. Yeazell, S. C. (1987). From medieval group litigation to the modern class action. Yale University Press ix. Yeazell, S. C. (1997). The past and future of defendant and settlement classes in collective litigation. Arizona Law Review, 39, 687–688.
ECONOMETRIC TESTS FOR ANALYZING COMMON IMPACT Kevin W. Caves and Hal J. Singer ABSTRACT In antitrust class-action litigation, courts are increasingly unlikely to accept the presumption that all class members were harmed by pricefixing among a group of firms or by exclusionary behavior by a single firm. Econometric methods typically applied in antitrust and other settings estimate the average effect of the challenged conduct, but do not inform impact for individual class members. We present classwide econometric methods and statistical tests for detecting the existence (or lack thereof) of common impact and determining what proportion (if any) of the proposed class suffered injury in many class actions. We conclude that econometric tools can meaningfully inform the legal process, even when courts demand proof of common impact. Keywords: Common impact; econometric modeling; common proof paradox; class actions; class certification JEL classifications: C52; K21; C19
The Law and Economics of Class Actions Research in Law and Economics, Volume 26, 135160 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0193-5895/doi:10.1108/S0193-589520140000026005
135
136
KEVIN W. CAVES AND HAL J. SINGER
TABLE OF CONTENTS Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Allowing for Individualized Effects in Price-Formation Models . . . . . Model Specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Key Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Testing for Individualized Effects with a Common Econometric Model. Case I: Common and Identical Impact for All Class Members . . . . Case II: Individualized Impact for Some Class Members . . . . . . . . . Case III: Individualized Impact for All Class Members . . . . . . . . . . Illustrative Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The “Common Proof Paradox” . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
136 140 141 143 144 144 147 148 149 152 154 159
INTRODUCTION In antitrust class-action litigation, courts are increasingly unlikely to accept the presumption that all members of a proposed class were harmed by pricefixing or by other allegedly anticompetitive conduct such as exclusive dealing. Instead, courts have exhibited a clear tendency in recent years to require that plaintiffs offer empirical proof of common impact or at least demonstrate that a common and viable method of proof for showing impact exists before certifying the class.1 Common impact implies that a large proportion of buyers (or purchase volumes) were adversely affected by the challenged conduct. In contrast, the econometric methods commonly employed in antitrust settings often measure the average effect of the challenged conduct (typically on prices) across individual buyers, after controlling for other factors.2 This has created a tension between evolving legal standards and econometric methods as they have typically been applied in antitrust cases.3 For example, in Plastics Additives, which concerned alleged price-fixing involving organotin heat stabilizers and epoxidized soybean oil, the district court rejected plaintiffs’ use of regressions designed to measure the average, industry-wide price effects of the allegedly anticompetitive conduct.4 The court found that a “fundamental failure”5 of the plaintiffs’ regressions was that they “tell us nothing about individual class member experience.”6 Consistent with this finding, the defense expert estimated regressions for individual class members suggesting that substantial numbers of class
Econometric Tests for Analyzing Common Impact
137
members showed no evidence of elevated prices during the period of the alleged price-fixing conspiracy and that only a minority of customers showed evidence of a “significant increase in price.”7 In Blades et al. v Monsanto, plaintiffs alleged that the defendant had secured the agreement of licensees to inflate the prices of genetically modified (GM) corn and soybean seeds. In denying class certification, the district court stated that “[t]he market for seeds is highly individualized, depending upon geographic location, growing conditions, consumer preference and other factors,”8 and noted that “defendants and their distributors often lowered the ‘overall’ price of certain seeds, or gave discounts or rebates to certain farmers to offset any alleged premium, and that some farmers in fact paid no premium.”9 The court found that plaintiffs’ expert had failed to show that impact could be demonstrated on a classwide basis, stating that it was insufficient for plaintiffs to simply “presume class-wide impact without any consideration of whether the markets or the alleged conspiracy at issue here actually operated in such a manner so as to justify that presumption.”10 Similarly, in Medical Waste Services, the district court refused to certify a proposed class in a monopolization claim, noting that plaintiffs had simply instructed their expert to “assume impact for all their causes of action.”11 The court was evidently persuaded by defendants’ experts’ “indepth [sic] analyses of the evidence that will be required for each specific claim, underscoring the necessity of the individualized evidence that will be required in this case.”12 The Third Circuit’s decision in Hydrogen Peroxide has been particularly influential in reshaping the relevant legal standard. To analyze plaintiffs’ allegations that defendants had jointly coordinated on price, plaintiffs’ expert proposed in general terms (but did not implement)13 a regression analysis to estimate the effect of the alleged anticompetitive conduct on prices, by modeling “the relationship between price of hydrogen peroxide, sodium perborate, and sodium percarbonate and the various market forces that influence prices, including demand and supply variables.”14 The district court certified the class, finding it sufficient that plaintiffs’ expert had proposed a reliable method for proving impact,15 and noted that plaintiffs could likely demonstrate common impact on all purchasers “merely by showing that defendants kept list prices that were artificially high because of their conspiracy.”16 However, the Third Circuit vacated the district court’s decision, stating that, despite the general plausibility of plaintiffs’ theory of harm, “the question at [the] class certification stage is whether, if such impact is plausible in theory, it is also susceptible to proof at trial through available evidence common to the class.”17 In its decision, the Third
138
KEVIN W. CAVES AND HAL J. SINGER
Circuit observed that defendants’ expert had presented evidence that prices charged to individual customers did not move together, suggesting that the alleged conspiracy was not amenable to proof of common impact and that individualized inquiries would be necessary.18 The Third Circuit held that the district court could not “decline to resolve a genuine legal or factual dispute because of concern for an overlap with the merits,”19 including disputes over plaintiffs’ ability to prove impact through common evidence. The Supreme Court’s decision in Comcast raised the evidentiary burden for plaintiffs, but it did not alter the standards regarding common proof of impact articulated in Hydrogen Peroxide.20 Instead, the major thrust of Comcast was that plaintiffs’ expert evidence must link damages to plaintiffs’ theory of harm and must be able to “bridge the differences”21 between supra-competitive prices in general versus those specifically attributable to a given theory of harm.22 If multiple theories of harm are asserted, but only some of those theories are susceptible to common proof (or deemed anticompetitive), then the damages model must be able to accommodate the specific alternative but-for world for each theory.23 Comcast appears to have been interpreted by the lower courts as requiring increased analytic rigor. For example, in In re Rail Freight Fuel Surcharge, plaintiffs presented a model that predicted antitrust injury across the board, even to legacy shippers “bound by rates negotiated before any conspiratorial behavior was alleged to have occurred.”24 Noting that the district court had not considered the damages model’s flaw, and had also lacked guidance from the Comcast decision, the D.C. circuit vacated and remanded the district court’s class certification decision.25 In light of developments such as these, it is not difficult to see why antitrust practitioners have cautioned against using an econometric model at the class-certification stage that “assumes that a conspiracy has the same effect on every purchaser and focuses on an average effect ….”26 Such a model “may hide variation across class members. If one is attempting to test whether there is an impact on all members of a proposed class … that assumption is not valid, as it assumes the very proposition that is being tested.”27 The tendency for economic experts to focus on average effects in the past appears to have been motivated by courts’ historical tendency to allow plaintiffs to prove impact simply by proving that, by virtue of the allegedly anticompetitive conduct, market prices were inflated above their but-for levels.28 But this “Bogosian Short-Cut”29 appears to have been eliminated by cases such as Hydrogen Peroxide, which emphasize the need for the expert to provide, at the class-certification stage, an empirical
Econometric Tests for Analyzing Common Impact
139
assessment of the relative importance of individual and common elements within the proposed class. It has sometimes been asserted that this evolving legal standard undermines (and perhaps even rules out) the applicability and utility of common econometric tools in class-certification proceedings. For instance, Cremieux, Simmons, and Snyder (2010) operating under the assumption of a regression that measures averages across individuals (as opposed to averages over time for a given individual) argue that regression analysis is, on its own, inherently insufficient to prove common impact.30 Similarly, Burtis and Neher (2011) note the potential for “price dispersion across customers,”31 among a set of factors that “complicates the analysis further and increases the likelihood that individualized analysis of each proposed class member over time is necessary in order to determine impact.”32 Johnson and Leonard (2007, 2011) advance a more extreme version of this line of reasoning in the form of a so-called “Common Proof Paradox” (CPP).33 According to the CPP, one cannot determine whether individualized evidence is necessary to establish common impact without first performing individualized analyses, which defeats the purpose of the class-certification mechanism. In essence, the CPP posits that a common, classwide regression model cannot be implemented without first allowing for the possibility that each individual class member may be subject to its own unique priceformation process. By necessity, econometric models are an approximation of the “true” price-formation process; if the process were sufficiently simple, no model would be necessary to understand it. Accordingly, we argue in this article that the economic expert’s choices in specifying a regression model should strike a balance between parsimony and complexity guided by both economic theory and facts specific to the case. In our view, a CPP-based approach, which assumes any and all uniformity across buyers, is skewed too far in the direction of complexity to be meaningful in most antitrust class actions. Such cases often involve large defendants selling to large numbers of buyers; in such contexts, it is difficult to justify a model that effectively assumes that the defendant employs a “customer-specific” pricesetting formula individually tailored to each of its customers. As an alternative, we present a broad class of price-formation models that allow for variation across class members with respect to the effect of the conduct, which can still be estimated empirically using regression frameworks common to all class members. Standard statistical tests can then be applied to detect the existence (or lack thereof) of individualized price effects that may deviate from the overall effect of the challenged
140
KEVIN W. CAVES AND HAL J. SINGER
conduct (if any). To the extent that some (or all) class members were “immunized” from the allegedly anticompetitive effects of conduct at issue, the class can be said to lack “cohesion” with respect to the challenged conduct. The higher the share of class members immunized from the conduct, the less likely it is that the proposed class satisfies the “all or almost all” standard for classwide impact. Determining the appropriate threshold for this standard in any given case would obviously be a matter of law. We conclude that there exist common econometric tools capable of meaningfully informing the legal process, even when courts demand proof of common impact. Of course, this is not to say that all types of allegedly anticompetitive conduct are amenable to common proof. Instead, the salient point is that there exist reliable econometric tests of common impact that should “rule in” classes that are cohesive with respect to the challenged conduct and “rule out” classes that are not.
ALLOWING FOR INDIVIDUALIZED EFFECTS IN PRICE-FORMATION MODELS The purpose of an econometric model is to generate a reasonable approximation of the “true” process that ultimately determines the prices paid by potential class members.34 As explained later, econometric models are most likely to yield meaningful results when guided by economic theory, industry knowledge, and facts specific to the case at hand. Although it is generally not possible to model all aspects of the price-formation process, a wellconstructed model can provide meaningful insight into the factors that influence the prices paid by potential class members, and it can distinguish between price changes attributable to the allegedly anticompetitive conspiracy and those attributable to unrelated factors (e.g., changes in the defendant’s costs). While many econometric models traditionally applied in antitrust settings have relied upon averages across potential class members to identify the effect of the challenged conduct, the models discussed later allow for buyer-specific effects to be measured by making use of buyerspecific averages. In effect, they compare the average price paid by a given class member before (or after) the alleged conspiracy (or restraint of trade) with the average price paid by that same buyer while the alleged conspiracy was in place. The proposed framework is adaptable to a range of antitrust cases, including those with large numbers of potential class members. It is also potentially relevant to class-certification matters outside antitrust, such
Econometric Tests for Analyzing Common Impact
141
as a class of buyers suing for damages due to fraud. Given a database of purchases by class member, and a method of quantifying the conduct (e.g., a start date and/or end date of the allegedly fraudulent conduct), the same basic framework could be applied.
Model Specifications In this section, we use a reduced form econometric model to illustrate the proposed framework. Reduced form methods capture the price-formation process by modeling various factors that may influence the prices paid by potential class members, including supply-side factors (e.g., costs) and demand-side factors (e.g., buyer income).35 Specifically, consider a customer price-formation equation of the following form: pit = Xit β þ Zt γ þ δDt þ μi þ ɛ it
ð1Þ
Eq. (1) is a representative of a class of reduced form models commonly employed in antitrust litigation. The model can be estimated using a panel of individual-level data, such as a transactions database obtained in antitrust litigation, that reflect multiple purchases by potential class members over time. Above, the left-hand side variable is the price (or the natural logarithm of price) paid by customer i at time t. In a reduced form framework, price is a function of cost and demand factors, which appear on the right-hand side of the equation. The purpose of including cost and demand variables is to control for factors other than the challenged conduct itself that may cause prices to rise or fall. For instance, if the alleged conspiracy took place at a time when the defendant’s input costs were rising, it will be necessary to control for this because any observed increase in prices may be wholly explained by rising costs. Conversely, if the defendants’ costs were falling substantially during the alleged conspiracy, failure to include input costs as a control variable may cause the expert to significantly underestimate the inflationary effect of the conduct, relative to the but-for world. Potential control variables may include individual-specific factors, Xit , such as the customer’s size or scale; it may also include common factors that vary over time, such as input costs and macroeconomic factors, denoted by Zt . Provided the impact of these factors does not vary over time, all individual-specific factors influencing the prices paid by customer i
142
KEVIN W. CAVES AND HAL J. SINGER
are captured by the customer fixed effects, μi .36 Such factors need not be included in Xit or in Zt ; it would be redundant to do so. The effect of the challenged conduct is measured using the variable Dt , an indicator set equal to one during the period of the allegedly anticompetitive conduct, and zero otherwise. Accordingly, the key parameter of interest is δ, which measures the extent to which prices were (or were not) elevated at during the time frame when the allegedly anticompetitive acts occurred, after controlling for other factors. Importantly, δ measures the average effect of the conduct across all individual buyers, controlling for all other variables included in the price-formation model. Finally, note that Eq. (1) is common to the class in the sense that only a single regression model needs to be estimated for all potential class members. The relevant inquiry during the class-certification stage of an antitrust litigation is whether there is common impact from the challenged conduct that is, whether the class is sufficiently cohesive with respect to the conduct. Common impact “concerns whether the conduct’s effect on individual prices can be characterized by relying exclusively on common factors.”37 The obvious concern, as articulated in Plastics Additives and In re Rail Freight, is that models such as Eq. (1) might “demonstrate[] impact where there in fact was none”38 owing to individual-specific deviations from the average effect of the conduct. Thus, the focus of the inquiry should be on whether and to what extent the effect of the conduct on prices varies across class members. Eq. (1) can be adapted to allow for this possibility by introducing individualspecific interactions between the individual fixed effects and the conduct variable. The result is a modified version of the initial specification: pit = Xit β þ Zt γ þ δDt þ δi Dt þ μi þ ɛit
ð2Þ
Eq. (2) is nearly identical to Eq. (1), with the exception of the individualspecific parameters δi . Note that Eq. (2) requires that a “baseline” customer be designated with δi constrained to zero for this customer. The effect of the conduct (or lack thereof) on the baseline customer is therefore equal to δ. The effect of the conduct on all other class members is measured against the baseline, where an individual’s impact differs from the baseline if and only if δi ≠ 0. In other words, the effect of the conduct on individual j’s prices would be ðδ þ δj Þ, while the effect for individual k would be ðδ þ δk Þ.39 Finally, although it allows for individualized effects, Eq. (2) is still common to the class. As before, the same regression model is applied to all potential class members.40
Econometric Tests for Analyzing Common Impact
143
Key Definitions We are now in a position to define key concepts that inform the analysis of common impact. Definition 1: Assume that some or all of the proposed class experienced identical impact (or lack thereof) as a result of the challenged conduct. The Common Effect is defined as the extent to which prices were (or were not) altered for these class members, holding other factors constant. In Eq. (2), the Common Effect is captured by the coefficient δ, which captures the effect of the conduct on both the baseline customer and all other customers for whom δi = 0. In the event that δi = 0 for all i, then impact is common and identical, and is given by δ for every class member. Importantly, note that δ may be less than or equal to zero; this means that the conduct had no common inflationary effect on prices.41 Definition 2: Assume that some or all of the proposed class experienced differential impact as a result of the challenged conduct. The Individual Effect is defined as the extent to which a given individual’s prices deviated from the Common Effect. In Eq. (2), the Individual Effects are captured by the δi coefficients. If δi ≠ 0; then the individual experienced differential impact relative to the Common Effect. For example, if δ > 0; but δ þ δi = 0; then individual i was fully immunized from the Common Effect.42 Definition 3: The Average Effect is defined as the average level of impact (or lack thereof) experienced across all potential class members, and is given by a weighted average of the Common Effect and the Individual Effects. Let n1 be the number of putative class members who experienced the Common Effect only, and let n2 be the number of putative class members who experienced Individual Effects. Define the total number of potential class members as N ≡ n1 þ n2 . In the context of Eq. (2), the Average Effect (AE) can be written as: " # n2 n2 X 1 1X AE = n1 δ þ ðδ þ δi Þ = δ þ δi ð3Þ n1 þ n2 N i=1 i=1
144
KEVIN W. CAVES AND HAL J. SINGER
Based on Eq. (3), if all Individual Effects are equal to zero, the Average Effect is identical to the Common Effect. More generally, the Average Effect converges to the Common Effect to the extent that the sum across the Individual Effects (δi ) converges to zero (such that negative and positive deviations cancel each other), and also to the extent that n2 is small relative to n1 . Therefore, the absence of Individual Effects is sufficient, but not necessary, for the Common and Average Effects to be equal.43 Eq. (3) illustrates why it is that model such as Eq. (1), which measures only the Average Effect, may provide only a crude assessment of the presence (or absence) of common antitrust injury. Even if the Average Effect is estimated to be large and economically significant, such a finding does not preclude the possibility that some individuals experienced large, negative deviations from the Common Effect, as these could have been erased from the Average Effect by offsetting positive deviations. Conversely, even if the Average Effect is small, it remains possible that a nontrivial subclass suffered significant injury.
TESTING FOR INDIVIDUALIZED EFFECTS WITH A COMMON ECONOMETRIC MODEL In this section, we demonstrate how the framework developed earlier can be used to analyze common impact by testing for and quantifying common and individualized effects. As explained later, standard statistical tests can be applied to detect the existence (or lack thereof) of individualized price effects and to estimate the extent to which some proposed class members experienced Individual Effects on prices that deviated from the Common Effect.
Case I: Common and Identical Impact for All Class Members The most straightforward case occurs when the effect of the conduct (if any) is both common and identical across all potential class members. In this scenario, all Individual Effects are zero by definition. In terms of Eq. (2), this can be formulated as the hypothesis that δi = 0 for all i, such that all individuals experienced impact identical to that of the baseline customer. Alternatively, one could also estimate Eq. (4) below, which is a
Econometric Tests for Analyzing Common Impact
145
reformulation of Eq. (2). Rather than specifying a baseline customer, Eq. (4) captures each individual’s price effect (or lack thereof) without reference to other individuals: pit = Xit β þ Zt γ þ δi Dt þ μi þ ɛit
ð4Þ
In Eq. (4), there is no baseline effect. Instead, there is a separate δi that summarizes the Individual Effect in prices for buyers (i = 1,2,…, N). In addition, there is no Common Effect (δ = 0). This means that δi measures the full extent to which prices for buyer i were elevated when exposed to the allegedly anticompetitive conduct. The advantage of using Eq. (4), rather than Eq. (2), is that doing so obviates the need to designate a baseline class member, effectively placing all class members on equal footing. Although the two approaches are asymptotically equivalent both will yield identical results in sufficiently large samples the latter is generally preferable in practice, because the results of the former may be sensitive to which baseline is designated. Standard statistical tests can be applied to test the null hypothesis of equality across all Individual Effects as shown in Eq. (5) below: δ1 = δ2 = ⋯ = δN
ð5Þ
This relationship can be tested using the Wald test, the Likelihood Ratio (LR) test, or the Lagrange Multiplier (LM) test.44 Each of these three methods provides a statistical assessment of the validity of the restrictions on the δi parameters summarized in Eq. (5). In each case, the test statistic measures the extent to which the imposition of the parameter restrictions reduces the ability of the model to fit the data. If the reduction is not statistically significant, one accepts the null hypothesis that the restrictions are valid. These three tests are asymptotically equivalent: with a sufficient number of data points, each will give the same answer as the others. However, for in any given application, some tests may reject the null hypothesis of equal effects, and others may accept it. Fortunately, there is a consistent basis for ranking each test statistic: In the case of linear models (such as Eqs. (2) and (4)), the Wald test statistic is always greater than or equal to the LR test statistic, which in turn is greater than or equal to the LM test statistic.45 The greater the test statistic, the easier it is to reject the null hypothesis of equal individual effects in Eq. (5). This implies that the Wald test can be viewed as the most conservative in the sense of being the least likely to detect a Common Effect when there is none (i.e., the most likely to
146
KEVIN W. CAVES AND HAL J. SINGER
correctly reject the null hypothesis when it is false). However, this does not imply that the Wald test has greater statistical power than the others, because the Wald test is also more likely to incorrectly conclude that a Common Effect is not present when it is (i.e., to incorrectly reject the null hypothesis when it is true). In fact, all three tests have approximately equal statistical power.46 Thus, the expert can present the fact finder with a less equivocal statistical inference when all three tests yield the same answer that is, when all three uniformly reject or accept the null hypothesis. If the null hypothesis Eq. (5) is accepted, this provides a basis for modeling the price-formation relationship by the relatively simple process in Eq. (1), in which the Common Effect is identical to the Average Effect. Given an estimate of δ from that model, the data can support a simple “all or nothing” inference regarding impact. If δ > 0; then all potential class members suffered antitrust injury. Otherwise, impact cannot be proven for any member of the proposed class, let alone “all or almost all” potential class members. As with any econometric model, one must observe a sufficient number of data points to accurately estimate the parameters and test hypotheses. The power of the statistical test given in Eq. (5) is the probability that the test will correctly reject the null hypothesis that δi is identical across individual buyers, assuming that, in reality, the coefficients are different. Thus, a high-powered test is less likely than a low-powered test to conclude incorrectly that individual-specific effects are absent. Databases with higher transactional volumes per individual will, all else equal, allow for tests with a higher degree of statistical power.47 In addition, when buyer-fixed effects, μi , are included, only buyers that are observed both before and during (and possibly after) the period of the allegedly anticompetitive conduct can inform the model. In a model with buyer-fixed effects, the effect of the conduct is in essence measured by the extent to which the prices paid by an individual buyer are higher after the conduct than before. If a buyer is not observed prior to the conduct, this comparison cannot be made (the regression cannot distinguish between μi and δi ). On the other hand, if buyer-fixed effects are excluded, then the model can also be applied to any buyer who made purchases during the alleged conduct that is, any buyer who might have suffered impact. Under this specification, the regression measures the effect of the conduct on a given buyer by comparing that individual’s prices after the conduct to the prices that an average individual would have paid but-for the conduct controlling for Zt (e.g., costs, demand), as well as the individual factors Xit .48
Econometric Tests for Analyzing Common Impact
147
Case II: Individualized Impact for Some Class Members If the null hypothesis of Case I is rejected, the analysis can proceed to Case II, which provides a framework for analyzing situations in which at least some class members but not all experienced individualized price effects. Under Case II, the data can support inferences regarding the percentage of the class (if any) that suffered impact, as well as the identity of class members that were and were not injured. Specifically, one can apply the same statistical methods noted earlier to Eq. (4) to test for the existence of a Common Effect among subsets of class members. For example, given 100 potential class members, the economic expert could test the hypothesis that a Common Effect applies to 99 of them by testing the null hypothesis that: δ1 = δ2 = ::: = δ99
ð6Þ
If a Common Effect is accepted for some subset of class members, this provides a basis for modeling a price-formation relationship akin to Eq. (2), with the caveat that δi = 0 for many class members. For example, if the hypothesis in Eq. (6) is accepted, then only δ100 is nonzero; all other class members share the Common Effect given by δ. Given estimates of δ and δi , the data allow for inferences regarding which class members suffered impact, and which class members did not. If δ > 0; then all potential class members who experienced only the Common Effect suffered antitrust injury. Those who also experienced individual effects may or may not have been injured, depending on whether δ þ δi > 0: On the other hand, if δ ≤ 0; impact cannot be proven for any member of the proposed class that experienced only the Common Effect, although it is possible that class members with Individual Effects were injured (i.e., δ þ δi > 0). By the same token, the economic expert can obtain an estimate of the percentage of class members who were injured. Continuing the prior example, if the Common Effect is found to be positive and statistically significant, but the Individual Effect δ þ δ100 is not, this provides evidence that 99 percent of potential class members were injured and that the remaining one percent was immunized from the effects of the conduct. The higher the share of immunized class members, the less likely it is that the proposed class would satisfy the “all or almost all” standard for classwide impact. Of course, determining whether or not the share of class members injured is sufficient to satisfy the “all or almost all” standard would be a matter of law.49
148
KEVIN W. CAVES AND HAL J. SINGER
Setting up the hypothesis test contemplated in Case II first requires the identification of a subset of potential class members that may be subject to a Common Effect. This may be informed by economic theory or facts specific to the industry, as well as the pricing data itself, including estimates of δi obtained from Eq. (4). Information from any or all of these sources may allow the expert to infer which individual class members may be “outliers” relative to the others and which might sensibly be grouped together. On the other hand, if there is insufficient information in the record to identify any plausible grouping, or if the data reject the hypothesis of a (within-group) Common Effect, the analysis proceeds to Case III.50
Case III: Individualized Impact for All Class Members In the event that it proves impracticable to identify a Common Effect, the next and final step in the analysis is Case III, which evaluates common impact while allowing individual-specific price effects to vary arbitrarily. This would involve estimating Eq. (4) without any restrictions on the parameters, yielding separate estimates of δi for each potential class member. As before, this allows for inferences regarding which class members (if any) suffered impact and which class members (if any) were immunized. Specifically, impact can be shown for individual i if and only if δi is positive and statistically significant. All else equal, the more transactions observed for a given individual, the more precise the individual’s estimate of δi will be. Conversely, for individuals engaging in smaller numbers of transactions, the likelihood of a Type II error accepting null hypothesis that δi = 0 when the individual actually suffered impact is higher. In this situation, one alternative would be simply to exclude such individuals from the class, particularly if there are relatively few of them. Another possibility would be to perform the hypothesis test at a nonconventional level of statistical significance (e.g., to reject the null hypothesis that δi = 0 for any p-value less than or equal to say 0.15 or even higher level of statistical significance depending on the circumstances). Determining the appropriate significance level ultimately depends on the appropriate balance between Type II errors and Type I errors (which occur when the null hypothesis is incorrectly rejected that is, when the expert concludes impact occurred when it did not).51 Depending on what the data indicate, Case III could result in a finding of impact anywhere along the continuum from zero percent to 100 percent of potential class members. At one extreme, if all of the δi have the same
Econometric Tests for Analyzing Common Impact
149
(positive) sign (and are statistically significant), the data would support the inference that 100 percent of the class suffered injury. At the other end of the continuum, if all δi are negative and/or statistically indistinguishable from zero, the proper inference would obviously be that no class member was injured. In many cases, the answer is likely to fall somewhere in between. As before, the greater the share of class members for which impact can be proven, the greater the likelihood that the “all or almost all” requirement is satisfied. Although no Common Effect is identified under Case III, this does not imply that the effect of the conduct is economically insignificant or that the class lacks cohesion. From Eq. (3), it is clear that the Average Effect may be substantial and all class members may suffer impact even when δ = 0: In fact, a class that lacks a Common Effect, but for which δi > 0 for all i, is arguably more cohesive than a class in which the majority suffered injury equal to δ > 0, but with a nontrivial minority that were immunized from the effects of the challenged conduct. In any case, it is clear that the degree of class cohesion decreases with the number of individuals with zero (or negative) values of δi . However, the mere existence of Individualized Effects should not preclude proof of common antitrust injury. This would be tantamount to a requirement that each class member suffer the same quantum of damages.
Illustrative Example In this section, we provide an illustrative analysis based on a masked data set from an alleged price-fixing conspiracy employed by Cremieux et al. (2010).52 We modify the data set slightly by introducing random noise to the pricing data.53 Both the original, underlying pricing series and the randomized series are displayed in Table 1.54 For ease of exposition, the data set used in this illustrative example contains only six buyers. As seen in Table 1, the data set is comprised of six distinct buyers, with prices observed before and after an allegedly collusive price increase announcement. Four of the six buyers exhibit an increase in the underlying price after the announcement, but two buyers (B-2 and B-4) exhibit no increase whatsoever. Furthermore, although buyers B-1 and B-5 exhibit a relatively large ($5) increase in the underlying price, buyers B-3 and B-6 exhibit smaller increases ($1 for B-3 and $2 for B-6). In Table 2, we report the results of various regression analyses using these data. All specifications include buyer-fixed effects. The first column
150
Table 1. Invoice Date 6/11/1998 7/30/1998 12/17/1998 3/18/1999 9/17/1999 9/29/1998 12/31/1998 2/26/1999 3/31/1999 9/30/1999 8/31/1998 12/31/1998 7/2/1999 10/6/1999 12/7/1999 7/17/1998 8/12/1998 11/11/1998 6/10/1999 6/25/1999 8/28/1999 12/7/1998 2/25/1999 4/14/1999 12/17/1999 4/22/1998 11/16/1998 10/27/1999 4/22/1999 6/21/1999
KEVIN W. CAVES AND HAL J. SINGER
Masked Transactional Data Set Based on Cremieux et al. (2010). Product ID Product 1A Product 1A Product 1A Product 1A Product 1A Product 2A Product 2A Product 2A Product 2A Product 2A Product 3 Product 3 Product 3 Product 3 Product 3 Product 1B Product 1B Product 1B Product 1B Product 1B Product 4 Product 4 Product 4 Product 4 Product 4 Product 2B Product 2B Product 2B Product 2B Product 2B
Buyer ID B-1 B-1 B-1 B-1 B-1 B-2 B-2 B-2 B-2 B-2 B-3 B-3 B-3 B-3 B-3 B-4 B-4 B-4 B-4 B-4 B-5 B-5 B-5 B-5 B-5 B-6 B-6 B-6 B-6 B-6
Underlying Price
Randomized Price
Announcement Indicator
$129.00 $129.00 $129.00 $134.00 $134.00 $90.00 $90.00 $90.00 $90.00 $90.00 $90.00 $90.00 $91.00 $91.00 $91.00 $129.00 $129.00 $129.00 $129.00 $129.00 $106.00 $106.00 $111.00 $111.00 $111.00 $119.00 $119.00 $121.00 $121.00 $121.00
$128.86 $129.03 $128.86 $134.74 $133.65 $89.29 $90.46 $90.15 $89.85 $90.06 $89.49 $89.73 $90.70 $91.04 $90.89 $128.95 $128.54 $128.85 $128.75 $128.63 $106.30 $105.93 $110.68 $111.36 $110.90 $117.98 $119.11 $120.67 $120.38 $121.01
0 0 0 1 1 0 0 0 1 1 0 0 1 1 1 0 0 0 1 1 0 0 1 1 1 0 0 1 1 1
Source: Cremieux et al. (2010), at Table 5. The randomized price series adds a normally distributed random variable to each price observation with a mean of zero and a standard deviation of $0.50.
shows the results of a regression that includes a single indicator variable for the allegedly anticompetitive conduct, equal to one after the announcement and zero otherwise. The coefficient on the conduct variable, Announce, is positive and statistically significant, reflecting the fact that prices increased, on average, by approximately $2.24 after the allegedly collusive announcement. Of course, in reality, prices increased for some buyers, but not all. Based on just such a result, Cremieux et al. (2010) found that regression
151
Econometric Tests for Analyzing Common Impact
Table 2. Specification Announce Announce_Customer B1 Announce_Customer B2 Announce_Customer B3 Announce_Customer B4 Announce_Customer B5 Announce_Customer B6 Announce_Customer B_1_5 Constant Buyer-fixed Effects? Obs. R-Squared
Regression Output.
Average Effect 2.24*** (0.004) n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a n/a 110.37*** (0.000) Yes 30 99%
Individual Effects n/a n/a 5.28*** (0.000) −0.078 (0.983) 1.27*** (0.002) −0.092 (0.798) 4.87*** (0.000) 2.14*** (0.000) n/a n/a 110.32*** (0.000) Yes 30 99%
n/a n/a n/a n/a −0.078 (0.982) 1.27*** (0.002) −0.092 (0.797) n/a n/a 2.14*** (0.000) 5.07*** (0.000) 110.31*** (0.000) Yes 30 99%
Notes: Stars (*) represent significance thresholds: *p < 0.05, **p < 0.01, and ***p < 0.001. Buyer-fixed effects not reported.
analysis could not be used to assess common impact with such data, because a regression would serve “only to mask the underlying mix of impact and no impact cases.”55 In the second column of Table 2, we implement our proposed approach by allowing the effect of the conduct to vary across class members; that is, we estimate a separate Announce coefficient for each of the six potential class members. As seen in Table 1, buyers B-1 and B-5 are both estimated to have experienced statistically significant price increases of $5.28 and $4.87, respectively. The Announce indicators for buyers B-2 and B-4 are not statistically significant, consistent with the fact that neither of these two potential class members appears to have been impacted. Finally, the point estimates for buyers B-3 and B-6 are $1.27 and $2.14, respectively. Both estimates are statistically significant, and both reflect the fact that each of these two potential class members appear to have suffered impact (albeit to a lesser extent than buyers B-1 and B-5).
152
KEVIN W. CAVES AND HAL J. SINGER
A Wald test correctly rejects the null hypothesis of a Common Effect on the four buyers that appear to have suffered impact (p < 0.000). That is, the data refute the hypothesis that the announcement had equal effects on the prices paid by buyers B-1, B-3, B-5, and B-6. In addition, the hypothesis that B-2 and B-4 both suffered impact equal to zero is correctly accepted (p = 0.9668). The null hypothesis of equal effects for B-3 and B-6 is rejected, albeit only at the 10 percent significance level (p = 0.099). Additional statistical power (i.e., more observations) would be necessary to distinguish between underlying impact of $1 versus $2 with a greater degree of statistical certainty. Based on these hypothesis tests, the final column of Table 2 reports the results of a regression in which a Common Effect is estimated for buyers B-1 and B-5 (i.e., the Announce coefficient is constrained to be equal for these two buyers). As seen in Table 2, the point estimate for the Common Effect is $5.07, which is closer to the true value ($5) than either of the individual estimates obtained in the second column. (The remaining coefficients are essentially unchanged in the third column.) Accordingly, when the effect of the conduct is permitted to vary across buyers, the regression reliably distinguishes between buyers that incurred impact and those that did not. In this example, the analysis reveals that two thirds of the proposed class paid higher prices after the allegedly collusive price announcement. Without loss of generality, the same logic applied here could be applied to estimate the share of potential class members that incurred impact using much larger data sets containing a much greater number of potential class members. For example, suppose there were 600 buyers, rather than six: The analysis would proceed simply by computing the proportion of potential class members with positive and significant price effects. If there were 400 such buyers, this would again imply that approximately two thirds of the proposed class suffered impact.56 As before, based on the estimated proportion of buyers that incurred impact, it would be up to the fact finder to determine whether that the proportion is sufficiently high to permit a finding of common impact. To reiterate, because the same regression model is applied to all potential class members and because the same procedure is applied to assess impact for all buyers, the fact finder’s decision would rely exclusively on evidence common to the proposed class.
The “Common Proof Paradox” In contrast to the common regression models presented earlier, the “Common Proof Paradox” approach discussed earlier posits that the
Econometric Tests for Analyzing Common Impact
153
application of econometric tools class to certification proceedings is an inherently individualized exercise. The CPP is predicated on priceformation models of the following form:57 pit = Xit βi þ Zt γ i þ δi Dt þ μi þ ɛ it
ð7Þ
In contrast to Eqs. (1), (2), and (4), both Xit and Zt now have differential effects on prices, which vary from one buyer to the next. Thus, the CPP represents a particular type of reduced form model in which both the supplyside factors and the demand-side factors affect prices differently from one buyer to the next. For example, suppose that a defendant passes on input cost increases to different retailers at different rates depending on bargaining power, such that a large retailer (such as Wal-Mart) might absorb a smaller portion of a cost increase than a smaller retailer. This implies that the parameter γ w (for Wal-Mart), would be smaller than γ s , the corresponding pass-through parameter for the small retailer. Accordingly, Eq. (7) assumes that each individual class member may be subject to its own unique priceformation process. Proponents of the CPP argue that attempts to utilize such a model to assess common impact are “inherently individualized.”58 Econometric models are generally most useful in antitrust settings when informed by economic theory, industry knowledge, and the question they are designed to answer.59 The relevant inquiry at the class-certification stage of an antitrust litigation is whether there is a common impact from the challenged conduct that is, whether the class is sufficiently cohesive with respect to the allegedly anticompetitive conduct.60 Thus, the nature of the inquiry lends itself naturally to models (such as Eqs. (2) or (4)) that allow for the effect of the conduct and only the conduct to vary from one class member to another. This is not to say that a common model cannot be adjusted to fit the industry in question (given a reasonable theoretical or empirical basis for doing so). For example, if theory, data, or record evidence suggests that the defendant passes through a higher share of its cost increases to large customers than to small customers, the expert could incorporate this into the price-formation model using buyer-fixed effects and/or control variables capturing transaction volumes or buyer size. Similarly, if some purchasers of an allegedly monopolized input use that input to make a final product for (say) the aerospace industry, while others use it to make a final product for (say) automobiles, one could include industry-specific demand or cost factors. Importantly, such adjustments to the model do not involve tailoring the analysis to individual class members.
154
KEVIN W. CAVES AND HAL J. SINGER
More fundamentally, it is both possible and statistically beneficial to incorporate the CPP framework into a single, common regression model estimated for all class members at once.61 Of course, one could attempt to estimate separate regression models of the form given in Eq. (7) for each class member individually, but doing so would merely discard potentially valuable information. In particular, performing separate regressions excludes, ex ante, any and all variation in prices from one buyer to the next, even though such variation is likely to inform the price-formation process.62 In contrast, estimating a common regression model preserves degrees of freedom by pooling data across a large number of buyers allowing the expert to evaluate the degree of uniformity in the price-setting process through statistical tests. For example, the classwide approach would allow the expert to test whether the cost pass-through parameter (γ) is equal across all customers (or across some subset of customers), which is possible even when there are insufficient observations to run separate regressions for each customer.63 Performing only individualized regressions, as suggested by the CPP, simply imposes a separate pass-through parameter γ i for each customer ex ante, while ignoring the cross-buyer variation that is available to test the validity of this assumption. In summary, all econometric models are, by necessity, approximations that are tools for sorting out meaningful patterns.64 Accordingly, the expert’s choices in specifying a regression model should strike a balance between parsimony and complexity, guided by both economic theory and facts specific to the case. If too simple a process is specified, the model will not accurately distinguish class members who suffered impact from those that did not. But if the specification is overly detailed or complex, the expert runs the risk of overfitting the model, attempting to discern meaningful relationships where there are none, and robbing the model of its ability to evaluate the existence of common impact by forcing estimation of an excessive number of parameters.65 In our view, the CPP approach is skewed too far in the direction of complexity to be meaningful in most antitrust class actions. While it is always possible to add detail to a model, it is not always desirable to do so.
CONCLUSION Courts are increasingly skeptical of the notion that all or nearly all members of a proposed class would have necessarily suffered antitrust injury as
Econometric Tests for Analyzing Common Impact
155
a result of allegedly anticompetitive conduct. The ability to analyze impact on a classwide basis turns on whether the effect of the allegedly anticompetitive conduct on prices paid by individual class members can be captured with a model that relies exclusively on factors common to the class. In this article, we have presented common econometric tools that satisfy this requirement. Applying standard regression frameworks, we demonstrate a broad class of price-formation models that allow for arbitrary variation across class members with respect to the effect of the conduct. A variety of standard statistical tests can then be applied to assess the extent to which the data support the existence (or lack thereof) of individualized price effects and to determine what proportion (if any) of the proposed class suffered antitrust injury. Accordingly, econometric tools can meaningfully distinguish between forms of allegedly anticompetitive conduct that are and are not amenable to common proof, by estimating the degree of cohesion (or lack thereof) among proposed class members.
NOTES 1. See, for example, In re Plastics Additives Antitrust Litig., No. 03-CV-2038, 2010 U.S. Dist. LEXIS 90135 (E.D. Pa. August 31, 2010) (hereinafter Plastics Additives); see also Blades et al. v. Monsanto Company et al., 400 F 3d 562 (8th Cir. 2005) (hereinafter Blades et al v. Monsanto); see also In Re Medical Waste Services Antitrust Litig. (Slip Copy, 2006 WL 538927 (D. Utah)) (hereinafter Medical Waste Services); see also In re Hydrogen Peroxide Antitrust Litig., 552 F.3d 305 (3d Cir. 2008) (hereinafter Hydrogen Peroxide). In Hydrogen Peroxide, the Third Circuit’s opinion stressed courts’ obligation to resolve factual and legal disputes regarding class certification, even those that overlap with the merits of the case. 2. See, for example, Schmalensee (2008) (“Most economic analysis in support of litigation is concerned with averages or totals. To show that a practice is anticompetitive, for instance, one typically wants to show that it tends to increase market price, which is generally an average, or to decrease total output. But the focus in class certification is on differences: the commonality and typicality requirements for certification have to do with the relative importance of differences versus common elements within the asserted class.”) (Emphasis in original). 3. As noted below, the importance of correctly applying the appropriate economic tests for class certification has recently been underscored by the Supreme Court in its decision in Comcast Corp. v. Behrend, No. 11-864 (U.S. March 27, 2013) (hereinafter Comcast v. Behrend). One of the authors (Singer) served as an expert for plaintiffs in this matter. 4. Plastics Additives, at 5557. 5. Ibid. 6. Ibid.
156
KEVIN W. CAVES AND HAL J. SINGER
7. Ibid., at 6061. 8. Blades et al. v. Monsanto, at 17. 9. Ibid. 10. Ibid., at 16 (Emphasis in original). 11. Medical Waste Services, at 21 (Emphasis in original). 12. Medical Waste Services, at 21. 13. Hydrogen Peroxide, at 29. 14. Ibid., at 2122. 15. Ibid., at 29. 16. Ibid. 17. Ibid., at 59. 18. Ibid., at 23. 19. Ibid., at 57. 20. Comcast v. Behrend, supra. 21. Ibid., at 2. 22. In Comcast, the court accepted only one of four proposed theories of harm, which alleged that Comcast’s actions had lessened competition from so-called “overbuilders” (rival cable companies that build competing networks in areas where incumbents already operate). Ibid., at 1. 23. It bears noting that if plaintiffs assert two mechanisms of harm (say, horizontal and vertical restraints), and if plaintiffs can demonstrate that both restraints were necessary for defendant to exercise market power, then there need not be an allocation of damages between the two restraints even if one restraint is deemed procompetitive and therefore persists in the but-for world. In that case, 100 percent of damages would be attributable to the other restraint. 24. See In re Rail Freight Fuel Surcharge Antitrust Litigation MDL No. 1869 (D. C. Cir. August 9, 2013) (hereinafter In re Rail Freight), at 13. 25. Ibid., at 18. 26. Nelson, McFarland, and Smith (2005). 27. Ibid. 28. See, for example, Cremieux, Simmons, and Snyder (2010), at 942943 (hereinafter Cremieux et al.). See also Bogosian v. Gulf Oil Corp 561 F.2d 434 (3d Cir. 1977). 29. Ibid. 30. Cremieux et al., at 955. 31. See Burtis and Neher (2011) at 520. 32. Ibid. 33. See Johnson and Leonard (2011). With Dr. Laila Haider, the same authors submitted an amicus brief to the Supreme Court in August 2012 that lays out their views of the appropriate standards for evaluating the viability of computing damages on a classwide basis. See Brief of Economists as Amici Curiae in Support of Neither Party, Comcast Corporation, et al., Petitioners, v. Caroline Behrend, et al., Respondents. On Writ of Certiorari to the United States Court of Appeals for The Third Circuit (“One reliable method is to have the proposed regression model run on both the class as a whole and separately on one or more sub-groups of the putative class. The results of those separate models are then compared to the results provided when the regression model was run class-wide. If this comparison
Econometric Tests for Analyzing Common Impact
157
shows a statistically significant difference between the sub-groups and the class, then the class-wide regression model does not provide a sound method for measuring damages on a class-wide basis.”). Such a test would appear to bar the formation of a class so long as two sets of buyers experienced differential impact, even if both sets experienced an adverse and statistically significant impact. In any case, the Supreme Court did not accept or reject such an approach in its Comcast decision. 34. Depending on the case, the advanced econometric techniques that we explore in this section may not be feasible (e.g., buyers may not purchase the product in sufficient quantities both before and during the class period) or necessary (e.g., in the presence of a rigid pricing structure). See, for example, Singer (2011) (showing how common impact can be proven in the presence of a pricing structure). The appropriate techniques for assessing common impact will vary according to the nature of the industry and the conduct being challenged. 35. See, for example, Langenfeld, Li, Leonard, and Morris (2010), at 125192. 36. For example, one individual might be more costly to serve than most, and incur a constant surcharge on all purchases; another might be contractually entitled to a 10 percent discount at all points in time. Of course, to the extent that these factors do change with time for example, if an individual’s discount decreases from 10 percent to 5 percent then individual-specific factors will not be fully captured by μi . 37. Johnson (2011), 544. 38. Plastics Additives at 64; see also In re Rail Freight at 4. 39. To illustrate, assume that the baseline customer’s prices were inflated by $5as a result of the conduct, that individual j’s prices were inflated by $9, and that individual k’s prices were not inflated at all. In this case, δ = $5, δj = $4, and δk = − $5. 40. To continue the example earlier, the model would indicate impact for the baseline customer, since ðδ = $5Þ and for individual j, since ðδ þ δj = 9Þ but not for individual k, since ðδ þ δk = 0Þ. If these were the only three potential class members, then the (common) model would indicate that two thirds of the class had incurred antitrust impact. 41. To illustrate, if ðδ = $5Þ and ðδi = $0Þ for all individuals, then all class members incurred common and identical impact in the amount of $5. On the other hand, if ðδ = − $5Þ and ðδi = $0Þ for all individuals, then no class members were impacted. 42. For example, if ðδ = $5Þ and ðδj = − $3Þ, then individual j experienced differential impact of $2. However, if ðδj = − $5Þ, then individual j was fully immunized. 43. To illustrate, assume, as above, that there are the only three potential class members, with δ = $5, δj = $4, and δk = − $5. Therefore, ðδ þ δj = $9Þ and ðδ þ δk = $0Þ. According to Eq. (3), the average effect is then $5 + ($4 − $5)/3 = $4.67. Note that this is equal to the average total impact across all three class members, equal to ($5 + $9 + $0)/3 = $4.67. 44. See, for example, Engle (1984). The Wald test is an asymptotic approximation of the familiar F-test, and the two are often referred to interchangeably. Ibid., at 780. 45. Ibid., at 792. 46. Ibid., at 793. 47. Although the power of a test increases with the number of observations, it also depends on other characteristics of the data. A given number of data points may lead to relatively high statistical power in one case but not in another. In
158
KEVIN W. CAVES AND HAL J. SINGER
general, there is no absolute rule governing the “acceptable” number of data points required to conduct regression analysis, nor is there an absolute rule for the “acceptable” level of power that a test should have. 48. To clarify, while it is preferable to observe buyers who make purchases both before and during the alleged conspiracy, the model can also be applied to buyers who only make purchases during the alleged conspiracy. In the former case, the model involves a “difference-in-difference” approach by, in effect, measuring whether the difference between a given individual’s prices and the average price changes during the alleged conspiracy. This will be the case if, for example, a given individual pays prices $5 less than comparable individuals before the conspiracy, but $5 more than comparable individuals during the alleged conspiracy. In the latter case, the model can identify whether the individual pays $5 more than comparable individuals during the alleged conspiracy; however, it is not possible to determine whether that individual would have also paid $5 more than comparable individuals before the alleged conspiracy. 49. Impact could also be assessed by weighting each class member by its purchase volume over the class period. This would provide the fact finder with an estimate of the share of class purchases whose prices were inflated as a result of the allegedly anticompetitive conduct. 50. In theory, one could apply a “brute force” method to search systematically for subsets of class members with a Common Effect. However, there are significant theoretical and computational drawbacks to such an approach. From a computational perspective, given n potential class members, there are a total of n!=k!ðn − kÞ! different ways to select k candidates for a Common Effect; this number can quickly run into the millions and beyond. From a theoretical perspective, implementing a large number of hypothesis tests increases the likelihood that one or more of these tests will generate a “false negative,” or “Type II Error,” causing the analyst to incorrectly accept the null hypothesis of a Common Effect when there is none. 51. The task of determining the appropriate significance level would presumably fall to the fact finder; accordingly, the expert should generally report the results of hypothesis tests under standard significance levels as well. To illustrate the relative error costs, consider an example where the individual effect of a small buyer (with say 10 purchases) is approximately equal to the average effect of impact across all buyers, yet the coefficient for that buyer’s impact is not statistically significant at the 10 percent level. The cost of incorrectly including that small buyer in a class of direct purchases comprised of buyers with a large number of purchases that clearly suffered impact (the relevant Type I error at the class-certification stage), however large, is arguably less than the cost of incorrectly finding that the defendant violated the antitrust laws (the relevant Type I error at the merits phase). With the former, defendant would have several opportunities, including at summary judgment and at trial, to disprove plaintiffs’ theory of harm (or even to reverse the class-certification decision at a court of appeals). 52. Cremieux et al., at Table 5. 53. The data are modified primarily because the original pricing data can be predicted perfectly when our proposed specification is employed (i.e., we obtain an R-squared of 100 percent). Accordingly, we added a normally distributed random variable to each price observation, with a mean of zero and a standard deviation of $0.50.
Econometric Tests for Analyzing Common Impact
159
54. In addition, we multiply each of the original pricing observations by 100. Although this does not affect the substance of our results, it allows us to interpret the regression output without the need to distinguish between fractions of a penny. 55. Cremieux et al., at 955. 56. Because this procedure considers a set of statistical inferences simultaneously, it may be desirable to adjust for “multiple comparisons,” particularly when the number of potential class members is large. (In essence, the issue arises because the probability of committing at least one Type I Error increases with the number of hypotheses that are tested. For example, if a significance level of one percent is chosen, then one would expect to conclude that one out of every 100 class members suffered impact, even if none were impacted in reality.) To account for this, one can employ procedures to control the “false discovery rate,” defined as the proportion of null hypotheses incorrectly rejected (i.e., the frequency of Type I Errors). See, for example, Benjamini and Hochberg (1995). However, as noted above, the benefits of limiting Type I Errors (concluding erroneously that a given class member incurred impact) should be weighed against the costs of committing Type II Errors (concluding erroneously that a given class member did not incur impact). 57. Johnson and Leonard (2007) and Johnson and Leonard (2011) discuss nearly identical versions of this model. See Johnson and Leonard (2007), at 349 and Johnson and Leonard (2011), at 576. 58. Johnson and Leonard (2011), at 578. 59. See, for example, Rubinfeld (2011). 60. See, for example, Singer (2011), at 34. 61. Johnson and Leonard (2007), at 349 (noting correctly that such a model “could, as a technical matter, be estimated using a ‘single’ regression”); see also Johnson and Leonard (2011), at 583584. 62. For example, if one buyer purchases an item when the firm’s costs are $5 and another buyer purchases the same item after the firm’s costs have risen to $6, this variation can inform the degree to which cost increases are passed on to buyers. 63. Johnson and Leonard (2007), at 349, n. 17, and at 351, n. 18. 64. See Box and Draper (1987) (“Essentially, all models are wrong, but some are useful.”). 65. The problem of overfitting in model selection is sometimes referred to as a tendency to see “signals in the noise.” See Silver (2012).
REFERENCES Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society: Series B (Methodological), 57, 289–300. Box, G. E. P., & Draper, N. R. (1987). Empirical model-building and response surfaces (p. 424). Wiley. Burtis, M. M., & Neher, D. V. (2011). Correlation and regression in antitrust class certification. Antitrust Law Journal, 77, 495–532.
160
KEVIN W. CAVES AND HAL J. SINGER
Cremieux, P., Simmons, I., & Snyder, E. A. (2010). Proof of common impact in antitrust litigation: The value of regression analysis. George Mason Law Review, 17, 939–967. Engle, R. F. (1984). Wald, likelihood ratio, and Lagrange multiplier tests in econometrics. In Z. Griliches, & M. Intriligator (Eds.), Handbook of Econometrics (Vol. 2, pp. 776–825), North Holland. Johnson, J. H., & Leonard, G. K. (2007). Economics and the rigorous analysis of class certification in antitrust cases. Journal of Competition Law and Economics, 3, 341–356. Johnson, J. H., & Leonard, G. K. (2011). Rigorous analysis of class certification comes of age. Antitrust Law Journal, 77, 569–586. Johnson, P. A. (2011). The economics of common impact in antitrust class certification. Antitrust Law Journal, 77, 533–567. Langenfeld, J., Li, W., Leonard, G., & Morris, J. (2010). Econometrics and regression analysis. In Proving antitrust damages: Legal and economic issues (2nd ed.). American Bar Association. Nelson, P., McFarland, H. B., & Smith, D. D. (2005). The use of econometrics in class certification. In Econometrics: Legal, practical, and technical issues (p. 222). American Bar Association. Rubinfeld, D. L. (2011). Reference guide on multiple regression. In Reference manual on scientific evidence (3rd ed., pp. 416, 423428). Federal Judicial Center. Schmalensee, R. (2008). Economic analysis of class certification, Global Competition Policy, June. Silver, N. (2012). The signal and the noise: Why so many predictions fail-but some don’t. Penguin Press. Singer, H. (2011). Economic evidence of common impact for class certification in antitrust cases: A two-step analysis. Antitrust (Summer), 34.
ASSESSING MARKET EFFICIENCY FOR RELIANCE ON THE FRAUDON-THE-MARKET DOCTRINE AFTER WAL-MART AND AMGEN$ Mukesh Bajaj, Sumon C. Mazumdar and Daniel A. McLaughlin ABSTRACT Following the Supreme Court’s 1988 decision in Basic, securities class plaintiffs can invoke the “rebuttable presumption of reliance on public, material misrepresentations regarding securities traded in an efficient market” [the “fraud-on-the-market” doctrine] to prove classwide reliance. Although this requires plaintiffs to prove that the security traded in an informationally efficient market throughout the class period,
$
Bajaj is the Global Head of the Finance and Securities Practice of Navigant Economics. Mazumdar is the Lead Director of the Finance and Securities Practice of Navigant Economics. Both are members of the Finance faculty at the Walter A. Haas School of Business, University of California-Berkeley. Daniel A. McLaughlin is Counsel with Sidley Austin LLP. The opinions expressed herein are opinions of the authors alone and not of their respective organizations or their clients.
The Law and Economics of Class Actions Research in Law and Economics, Volume 26, 161207 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0193-5895/doi:10.1108/S0193-589520140000026006
161
162
MUKESH BAJAJ ET AL.
Basic did not identify what constituted adequate proof of efficiency for reliance purposes. Market efficiency cannot be presumed without proof because even large publicly traded stocks do not always trade in efficient markets, as documented in the economic literature that has grown significantly since Basic. For instance, during the recent global financial crisis, lack of liquidity limited arbitrage (the mechanism that renders markets efficient) and led to significant price distortions in many asset markets. Yet, lower courts following Basic have frequently granted class certification based on a mechanical review of some factors that are considered intuitive “proxies” of market efficiency (albeit incorrectly, according to recent studies and our own analysis). Such factors have little probative value and their review does not constitute the rigorous analysis demanded by the Supreme Court. Instead, to invoke fraud-on-the-market, plaintiffs must first establish that the security traded in a weak-form efficient market (absent which a security cannot, as a logical matter, trade in a “semi-strong form” efficient market, the standard required for reliance purposes) using wellaccepted tests. Only then do event study results, which are commonly used to demonstrate “cause and effect” (i.e., prove that the security’s price reacted quickly to news a hallmark of a semi-strong form efficient market), have any merit. Even then, to claim classwide reliance, plaintiffs must prove such cause-and-effect relationship throughout the class period, not simply on selected disclosure dates identified in the complaint as plaintiffs often do. These issues have policy implications because, once a class is certified, defendants frequently settle to avoid the magnified costs and risks associated with a trial, and the merits of the case (including the proper application of legal presumptions) are rarely examined at a trial. Keywords: Securities class actions; fraud-on-the-market; arbitrage limits JEL classifications: G14; K22
TABLE OF CONTENTS Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 The Link Between the Fraud-on-the-Market Theory (a Judicial Doctrine) and Market Efficiency Hypothesis (an Economic Theory) 164
Assessing Market Efficiency for Reliance
163
Basic and Class Action Law. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Efficient Capital Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . From Basic to Cammer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tests of Efficiency Used in the Economics Literature. . . . . . . . . . . . . . Weak-Form Efficiency Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Semi-Strong Form Efficiency Tests . . . . . . . . . . . . . . . . . . . . . . . . . Law of One Price Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Cammer Factors are Not Conclusive Proof of Market Efficiency . The Global Liquidity Crisis of 20072008, Limits to Arbitrage and Market Inefficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
164 170 172 174 175 177 182 183 190 192 204 205
OVERVIEW In its seminal decision in Basic1 in 1988, the U.S. Supreme Court permitted plaintiffs to prove classwide reliance by endorsing the “fraud-on-themarket” doctrine, which allows plaintiffs to “invoke a rebuttable presumption of reliance on public, material misrepresentations regarding securities traded in an efficient market.”2 Basic did not clarify what constituted adequate proof of efficiency for reliance purposes.3 Instead, left to develop their own standards of proof in this regard, lower courts typically rely on a “jumbled”4 list of factors (the best-known coming from Cammer v. Bloom,5 a district court decision shortly after Basic) that are considered intuitive indicators or “proxies”6 of market efficiency. However, according to a large body of economic evidence documented since Basic, even large and actively followed publicly traded securities do not always trade in efficient markets. Tellingly, the 2013 Nobel Prize in Economics was shared by Eugene Fama, the father of market efficiency theory, and Robert Shiller, one of that theory’s leading critics. We believe that continued reliance on ad hoc Cammer factors is inadequate to meet the Supreme Court’s requirement that, to obtain class certification, plaintiffs must prove that the security at issue traded in an efficient market throughout the class period based on rigorous analysis. To see why, we begin in the section entitled “The Link Between the Fraud-on-the-Market Theory (a Judicial Doctrine) and Market Efficiency Hypothesis (an Economic Theory)” by discussing the link between fraudon-the-market (a judicial doctrine) and market efficiency theory (an economic theory). In the section entitled “Tests of efficiency used in the
164
MUKESH BAJAJ ET AL.
economics literature”, we review the economic underpinnings of market efficiency theory and the commonly accepted empirical tests of the theory. In that section, we also discuss how economic reports related to class certification often ignore the economic literature and approach the plaintiffs’ burden to establish market efficiency as if it was at best a perfunctory step with a foregone conclusion. In this connection, we discuss how most of the Cammer factors (and other similar indirect “indicators” of market efficiency) have almost no power to detect a violation of market efficiency. Even the last Cammer factor (the cause and effect factor), which is the only Cammer factor that is a direct test of market efficiency, is typically analyzed using event studies in a perfunctory manner and in ways that cannot detect inefficiency. In the section entitled “The Cammer factors are not conclusive proof of market efficiency”, we demonstrate that such analyses have little probative value by focusing on a sample of well-documented cases of stocks that unquestionably violated the Law of One Price and hence did not trade in efficient markets over identified periods. Yet we show that even over these periods, these stocks nevertheless satisfied the Cammer factors. In the section entitled “The global liquidity crisis of 20072008, limits to arbitrage and market inefficiency”, we discuss the limits to arbitrage that arose during the recent financial crisis, rendering the market for numerous securities inefficient. Such evidence, coupled with the large body of literature in finance developed over the past twenty-five years following Basic, indicates that a mechanical review of ad hoc factors is clearly inadequate to assess market efficiency for class certification purposes. In the last section, we offer our concluding remarks.
THE LINK BETWEEN THE FRAUD-ON-THE-MARKET THEORY (A JUDICIAL DOCTRINE) AND MARKET EFFICIENCY HYPOTHESIS (AN ECONOMIC THEORY) Basic and Class Action Law In securities class actions, for a plaintiff class to be certified under Rule 23(b)(3) of the Federal Rules of Civil Procedure, a court must find “that the questions of law or fact common to class members predominate over any questions affecting only individual members,” which as the Supreme Court noted in its 2011 Halliburton decision, “often turns on the element of reliance.”7 In fact, the Supreme Court has recognized that, in
Assessing Market Efficiency for Reliance
165
the absence of some method for resolving questions of reliance on a classwide basis, the inherently individual question of what investors relied on in making a purchase or sale decision would make class action treatment of securities fraud claims impractical and unmanageable.8 In its seminal decision in Basic9 in 1988, the Supreme Court adopted a judicially created presumption to permit plaintiffs to dispense with proving individualized reliance on alleged misrepresentations where they purchased in the open market, but stressed that reliance was nonetheless a crucial element of a securities fraud case, as it provided the causal link between a misrepresentation and harm to the investor: Reliance provides the requisite causal connection between a defendant’s misrepresentation and a plaintiff’s injury … There is, however, more than one way to demonstrate the causal connection. Indeed, we previously have dispensed with a requirement of positive proof of reliance, where a duty to disclose material information had been breached, concluding that the necessary nexus between the plaintiffs’ injury and the defendant’s wrongful conduct had been established. Requiring a plaintiff to show a speculative state of facts, i.e., how he would have acted if omitted material information had been disclosed … or if the misrepresentation had not been made …, would place an unnecessarily unrealistic evidentiary burden on the Rule 10b-5 plaintiff who has traded on an impersonal market.10
Driven by this practical concern, the Court permitted plaintiffs to prove classwide reliance by endorsing the “fraud-on-the-market” doctrine, which allows plaintiffs to “invoke a rebuttable presumption of reliance on public, material misrepresentations regarding securities traded in an efficient market.”11 As Justice Thomas noted in his dissenting opinion in the Court’s 2013 Amgen decision, fraud-on-the-market is “a judicially invented doctrine based on an economic theory adopted to ease the burden on plaintiffs bringing claims under an implied cause of action.”12 The Court adopted this economic theory as a legal presumption on the basis of (1) the legislative history of the Securities Exchange Act of 1934 showing a Congressional “premise” that “competing judgments of buyers and sellers as to the fair price of a security brings [sic] about a situation where the market price reflects as nearly as possible a just price,” and (2) its assessment of “common sense and probability” derived from “[r]ecent empirical studies [through the use of sophisticated statistical analysis and the application of economic theory that] have tended to confirm Congress’ premise that the market price of shares traded on well-developed markets reflects all publicly available information, and, hence, any material misrepresentations.”13 Thus, the Court concluded that, “[b]ecause most publicly available
166
MUKESH BAJAJ ET AL.
information is reflected in market price, an investor’s reliance on any public material misrepresentations, therefore, may be presumed for purposes of a Rule 10b-5 action.”14 After Basic, courts have viewed the link between the market efficiency hypothesis and fraud-on-the-market theory as a “syllogism: (a) an investor buys or sells stock in reliance on the integrity of the market price; (b) publicly available information, including material misrepresentations, is reflected in the market price; and therefore, (c) the investor buys or sells stock in reliance on material misrepresentations. This syllogism breaks down, of course, when a market lacks efficiency, and the market does not necessarily reflect the alleged material misrepresentation.”15 The Court more recently, in Halliburton, distilled as “Basic’s fundamental premise that an investor presumptively relies on a misrepresentation so long as it was reflected in the market price at the time of his transaction.”16 The Court in Amgen further anchored the presumption in the view of market reality that market efficiency theory portrays: This presumption springs from the very concept of market efficiency. If a market is generally efficient in incorporating publicly available information into a security’s market price, it is reasonable to presume that a particular public, material misrepresentation will be reflected in the security’s price. Furthermore, it is reasonable to presume that most investors knowing that they have little hope of outperforming the market in the long run based solely on their analysis of publicly available information will rely on the security’s market price as an unbiased assessment of the security’s value in light of all public information.17
The validity of this premise therefore depends on plaintiffs using economically sound evidence to show that the market in question was actually acting the way the theory posits. Because the presumption is not dictated by statute but is adopted as a method of proving actual causation through economics, that places an obligation on courts to consider the advances in economic understanding of markets in the quarter century since Basic. The efficient market hypothesis, which was widely accepted by financial economists at the time of Basic,18 hypothesizes that a security’s price quickly and correctly19 impounds material new public information.20 If a security is temporarily mispriced (given available information) then such mispricing is quickly corrected in an efficient market through “arbitrage.” Lo (2008) describes this process as trading by “an army of investors [who] pounce on even the smallest informational advantages at their disposal [to earn a profit], and in doing so they incorporate their information into market prices and quickly eliminate the profit opportunities that first motivated their trades. If this occurs instantaneously, which it must in an idealized
Assessing Market Efficiency for Reliance
167
world of ‘frictionless’ markets and costless trading, then prices must always fully reflect all available information.”21 At its most basic level, for instance, two securities with identical cash flows must trade at the same price in an efficient market (referred to as “Law of One Price”). In short, it is the looming presence of arbitrage trading that keeps a market efficient. Given such arbitrage activity, no obvious arbitrage opportunity should persist in an efficient market. If it does (i.e., if the “no-arbitrage condition” is violated) then the market is considered inefficient. The Supreme Court, in Amgen, explained: The fraud-on-the-market theory rests on the premise that certain well developed markets are efficient processors of public information … Few investors in such markets, if any, can consistently achieve above-market returns by trading based on publicly available information alone, for if such above-market returns were readily attainable, it would mean that market prices were not efficiently incorporating the full supply of public information. See R. Brealey, S. Myers, & F. Allen, Principles of Corporate Finance 330 (10th ed. 2011) (“[I]n an efficient market, there is no way for most investors to achieve consistently superior rates of return.”).22
Importantly, to invoke Basic’s rebuttable presumption of reliance in the context of a class action, plaintiffs must prove at the class certification stage that the security at issue traded in an “efficient market” during the alleged class period, as the Supreme Court has reiterated in two recent decisions (Halliburton and Amgen). For instance, in Halliburton, the Court noted that “it is undisputed that securities fraud plaintiffs must prove certain things in order to invoke Basic’s rebuttable presumption of reliance … . [F]or example, that plaintiffs must demonstrate that the alleged misrepresentations were publicly known (else how would the market take them into account?), that the stock traded in an efficient market.”23 In Amgen, the Court, citing Wal-Mart, reiterated: “‘[P]laintiffs seeking 23(b)(3) certification must prove that their shares were traded on an efficient market,’ an element of the fraud-on-the-market theory (emphasis added).”24 The Amgen opinion noted that “without that proof, there is no justification for certifying a class.”25 Moreover, the courts have increasingly emphasized that such proof cannot be perfunctory. As the Supreme Court held in Wal-Mart, class certification is proper only if the “trial court is satisfied, after a rigorous analysis, that the prerequisites of Rule 23(a) have been satisfied” and that “Rule 23 does not set forth a mere pleading standard. A party seeking class certification must affirmatively demonstrate his compliance with the Rule that is, he must be prepared to prove that there are in fact sufficiently numerous parties, common questions of law or fact, etc.”26 The reason for this is that
168
MUKESH BAJAJ ET AL.
the defendant’s potential liability to a class “must be of such a nature that it is capable of classwide resolution which means that determination of its truth or falsity will resolve an issue that is central to the validity of each one of the claims in one stroke.”27 In addition to liability, the Supreme Court in its 2013 Comcast decision held that damages must be capable of determination on a common basis according to a common damages model that flows from the determination of liability; otherwise, “[q]uestions of individual damage calculations will inevitably overwhelm questions common to the class.”28 Consistent with Basic’s focus on practical market realities, Wal-Mart and Comcast emphasized that a class action is designed for the situation where a jury can make a final, up-or-down decision as to the class that is actually true as to all class members, rather than use a sampling of cases to conduct what the Court in Wal-Mart described as “Trial by Formula” while ignoring defenses that the defendant would have been entitled to raise in particular cases.29 In Comcast, the Court noted that a damages model in an antitrust case against a cable television company would not show that the case was a proper class action unless it “plausibly showed that the extent of [the antitrust violation] would have been the same in all counties” where the company did business.30 The Court in Amgen stressed that the question of materiality, by contrast, could properly be determined on a classwide basis because “[t]he alleged misrepresentations and omissions, whether material or immaterial, would be so equally for all investors composing the class … a failure of proof on the issue of materiality would end the case, given that materiality is an essential element of the class members’ securities-fraud claims. As to materiality, therefore, the class is entirely cohesive: It will prevail or fail in unison.”31 Amgen contrasted this to the situation where issues individual to particular investors were raised by the investor buying at a time when the price was not affected by the misrepresentation: A security’s market price cannot be affected by a misrepresentation not yet made, and in an efficient market, a misrepresentation’s impact on market price is quickly nullified once the truth comes to light. Thus, a plaintiff whose relevant transactions were not executed between the time the misrepresentation was made and the time the truth was revealed cannot be said to have indirectly relied on the misrepresentation through its reliance on the integrity of the market price. Such a plaintiff’s claims, therefore, would not be “typical” of the claims of investors who did trade during the window between misrepresentation and truth revelation.32
The theory must assume that if the market is efficient, each purchaser during the class period will be affected in the same way and in the same
Assessing Market Efficiency for Reliance
169
amount, just as the Court required for antitrust damages in Comcast that “price inflation” enters the price of the stock as soon as the misrepresentation enters the market (and exits the price as soon as it is corrected).33 If that common price impact is not present, then the presumption is a false guide: not every purchaser of stock was affected in the same way, and the class will include some members who did not actually buy stock at a time when misrepresentations were “reflected in the security’s price.”34 Thus, the Basic presumption provides the cohesion and “unison” required by Wal-Mart, Comcast, and Amgen only if the underlying economic theory can reliably show a common price impact on all purchasers over every day of the class period, such that price inflation can reasonably be presumed simply from a purchase during the period (and damages can later be computed mathematically from a showing of when the price inflation was dissipated by a corrective disclosure). The longer the class period and the larger the number of misrepresentations alleged, the greater the importance of assuring that the market remained consistently efficient at all times to avoid inclusion in the class of investors who purchased at prices unaffected by misrepresentations.35 Finally, the test for market efficiency, to be useful to courts, must be capable of testing by evidence in the adversary process and not merely the subject of speculation or the ipse dixit of the expert. This requirement is more important given that the Supreme Court has held that neither materiality nor loss causation is a required element of proof at the class certification stage. The Supreme Court has long warned, especially in the context of the judicially created Rule 10b-5 cause of action, against creating liabilities that depend on proof of speculative facts that cannot be tested by objective, extrinsic evidence.36 Indeed, that was one of the justifications for the Basic presumption itself. It is imperative that courts, in basing class certification on a theory of market efficiency, actually put that theory’s applicability to a test that can be evaluated on some objective, scientific basis and not just presumed from the existence of a loss on the plaintiff’s investment. There is substantial basis in the economic literature for finding markets to be informationally efficient, but not for simply presuming them efficient without examination. As discussed later, market efficiency theory cannot simply be assumed to be always true given the large body of finance literature has emerged since Basic demonstrating that the theory does not always hold. Instead, objective and rigorous analysis of evidence supporting the plaintiffs’ claims that the security at issue traded in an efficient market throughout the alleged class period is imperative at the class certification stage, as we discuss in detail in this article.
170
MUKESH BAJAJ ET AL.
Efficient Capital Markets In a market without frictions, an arbitrageur can earn an immediate riskfree profit with no up-front capital by simultaneously buying and selling identical securities at different prices. Such arbitrage trading in mispriced securities leads prices to correct quickly and restores market efficiency. In theory, as Shleifer and Vishny (1997) note, such arbitrage “requires no capital and entails no risk” and thus can be readily implemented. But in practice, market frictions may limit arbitrage, the mechanism by which markets are rendered efficient and the Law of One Price is enforced. Indeed, since Basic was decided, a very large body of economic literature has established that while actively traded and widely followed securities that trade on welldeveloped public markets are generally informationally efficient, violations of market efficiency are far more numerous and systematic than was previously thought, given limits to arbitrage. As a result, security prices can deviate from what would be observed in efficient markets, sometimes over extended periods of time. What is more, such instances of market inefficiencies (or “anomalies”) are observed even for stocks issued by some of the largest publicly traded companies that are extensively followed by analysts and widely traded by investors. Tellingly, the 2013 Nobel Prize in Economics was shared by Eugene Fama, the father of market efficiency theory and Robert Shiller, one of that theory’s principal critics.37 The large body of research that has developed since Basic to examine this economic issue has identified several real-world phenomena that may limit arbitrage even when investors are assumed to act rationally.38 We discuss a few well-accepted limits to arbitrage here. First, if gathering information is costly, then an arbitrageur would be willing to invest in gathering information only if he could profit from it. But if the security’s price is always efficient then such arbitrage profits are impossible. Knowing this, no arbitrageur would then be willing to invest in gathering information, and then, paradoxically, the security’s price could not be informationally efficient.39 This paradox, first articulated by Nobel laureate Joseph Stiglitz and Sanford Grossman, implies that given information-gathering costs, security prices cannot fully reflect all available information at all times. Instead, a certain degree of inefficiency is necessary to incentivize arbitrageurs to invest in gathering more information about a security.40 Thus, as Fama (1991, p. 1575) noted, when information-gathering and trading costs exist, “[a] weaker and economically more sensible version of the efficiency hypothesis says that prices reflect information to the point where the
Assessing Market Efficiency for Reliance
171
marginal benefits of acting on information (the profits to be made) do not exceed the marginal costs.” Second, the risk that mispriced securities’ prices may diverge further from their informationally correct values before correcting may limit the supply of capital necessary for arbitrage. As Shleifer and Vishny (1997) noted, arbitrage is a specialized activity that is generally conducted by profession fund managers who use other people’s capital. Such investors may become nervous and demand their money back if the securities purchased by the fund manager on behalf of these outside investors decline further in value (which can occur if mispriced securities’ prices temporarily diverge further from their correct values). To honor such redemption requests by investors, the fund manager may be forced to sell the mispriced securities prematurely (before the mispricing is corrected) and incur a loss. Anticipating this problem, arbitrageurs (fund managers) may not undertake certain arbitrage trades, and security mispricing may persist. Studies have documented that even under “normal” market conditions, stocks of certain large publicly traded companies have violated the Law of One Price (the most basic condition for a market to be deemed efficient). Notably, these violations persisted for extended periods even though they were discussed extensively in the popular press at the time.41 These cases are considered glaring instances of inefficiency in the academic literature. The lack of capital to conduct arbitrage was especially evident during the unprecedented global financial crisis of 20072009. The “depletion of dealer capital was so severe that, among other effects, large distortions in arbitrage-based pricing relationships appeared,”42 as Stanford economist, Darrell Duffie noted in his presidential address to the American Finance Association in 2010 [See Duffie (2010)]. As we will discuss later, such a liquidity crunch severely limited arbitrage and resulted in Law of One Price violations in several markets. In short, given this large body of academic evidence, economists today do not consider market efficiency a foregone conclusion in every situation, as might have been the case decades ago.43 Instead, the academic debate about the validity of the efficient markets hypothesis in various contexts based on decades of rigorous, peer-reviewed research continues to this day.44 An economist approaching a particular market would examine the efficiency of the market for the particular security at issue at the time in question, rather than assuming it. In contrast, such rigorous analysis is often absent in the “proof” that courts have routinely been provided at class certification stage.
172
MUKESH BAJAJ ET AL.
From Basic to Cammer Basic did not clarify what constituted adequate proof of efficiency for reliance purposes.45 Instead, left to develop their own standards of proof in this regard, lower courts typically rely on a “jumbled”46 list of factors (the best-known five coming from Cammer v. Bloom,47 from a district court decision shortly after Basic) as such factors are considered intuitive indicators or “proxies”48 of market efficiency. In addition to the five factors originally listed in Cammer, three more factors have been generally considered. This augmented list of eight factors are as follows: (1) the average weekly trading volume expressed as a percentage of total outstanding shares; (2) the number of securities analysts following and reporting on the stock; (3) the extent to which market makers and arbitrageurs trade in the stock; (4) the company’s eligibility to file SEC registration Form S-3 (as opposed to Form S-1 or S-2); (5) the existence of empirical facts “showing a cause and effect relationship between unexpected corporate events or financial releases and an immediate response in the stock price;” (6) the company’s market capitalization; (7) the bid-ask spread for stock sales; and (8) float, the stock’s trading volume without counting insider-owned stock.49
Courts have varied in their application of these factors, and some courts have tended to gloss over them when a stock trades on a well-known market such as the New York Stock Exchange (NYSE) or NASDAQ.50 Others, such as the Second Circuit, have emphasized that “factor five,” the rapid share price reaction to unexpected corporate news, is the primary focus of the market efficiency inquiry: Evidence that unexpected corporate events or financial releases cause an immediate response in the price of a security has been considered the most important Cammer factor … and the essence of an efficient market and the foundation for the fraud-onthe-market theory. Without the demonstration of such a causal relationship, it is difficult to presume that the market will integrate the release of material information about a security into its price. An event study that correlates the disclosures of unanticipated, material information about a security with corresponding fluctuations in price has been considered prima facie evidence of the existence of such a causal relationship.51
However, as several economic studies have recognized, many of the Cammer factors do not prove that a security traded in an efficient market, as they fail to distinguish between securities that trade in efficient markets from those that do not.52 According to some studies, stocks which did not trade in informational efficient markets according to more rigorous and direct tests of efficiency nevertheless satisfied the Cammer factors.53 Below, we also demonstrate that the Cammer factors were “satisfied” by several stocks that violated the Law of One Price the most basic market
Assessing Market Efficiency for Reliance
173
efficiency condition for extended periods according to studies published in leading academic journals. Even though several of the Cammer factors are only secondarily related to market efficiency, and others may fail consistently to detect market inefficiency, a plaintiff’s claim of reliance based on such factors “is not disputed in the vast majority of shareholder class actions,” and, until recently, “has not been rebutted in any case involving actively traded securities.”54 As a result, classes have been frequently certified55 which generally forces defendants to settle, given the magnified costs and risks associated with a trial.56 Not surprisingly, billions of dollars have been paid out in settlements in 10b-5 lawsuits “as a result of Basic”57 and only “8 percent of all federal class action securities fraud claims ever result in a ruling on a motion for summary judgment.”58 The merits of the plaintiffs’ claims, which are assessed only at the trial stage, are thus made largely irrelevant in many cases.59 Given the rigor required by Wal-Mart, this is not an acceptable result in cases where a closer examination would have revealed that the market was not actually efficient. Several justices’ opinions in the Supreme Court’s February 2013 Amgen decision echo the view that the policy consequences demand a closer look at the theory under which such classes are certified. The Court was not asked to revisit Basic’s fraud-on-the-market presumption or market efficiency in Amgen, as the defendant had conceded that the market for its stock was efficient.60 Nevertheless, Justice Scalia opined that, by eliminating one predicate for applying the theory (materiality), the Amgen majority opinion “does not merely accept what some consider the regrettable consequences of the fourJustice opinion in Basic; it expands those consequences from the arguably regrettable to the unquestionably disastrous.”61 Justice Thomas called the continued reliance on the fraud-on-the-market doctrine “questionable.”62 He noted that the concerns about the fraud-on-the-market doctrine which Justice White had expressed in his dissenting opinion in Basic that “[c]onfusion and contradiction in court rulings are inevitable when traditional legal analysis is replaced with economic theorization by the federal courts” and that the Court is “not well-equipped to embrace novel constructions of a statute based on contemporary microeconomic theory” “remain valid today.” 63 Even the majority opinion in Amgen took note of arguments presented in the briefs that “modern economic research [tends] to show that market efficiency is not a binary, yes or no question,”64 and in his concurring opinion, Justice Alito noted that “more recent evidence suggests that the presumption may rest on a faulty economic premise. … In light of this development, reconsideration of the Basic presumption may be appropriate.”65
174
MUKESH BAJAJ ET AL.
TESTS OF EFFICIENCY USED IN THE ECONOMICS LITERATURE Fraud-on-the-market theory was based on some influential academic studies of market efficiency published in the 1960s and 1970s.66 Since then, many implications of the Efficient Market Hypothesis (“EMH”) have been empirically examined for a wide range of financial securities67 and even other assets, such as real estate.68 In this vast body of literature in financial economics, certain types of empirical tests have become well-accepted and standard. We begin by reviewing how degrees of market efficiency are defined in the financial economics literature and review the standard empirical tests of efficiency commonly used in that literature. In an efficient market, a security’s price fully reflects all available information. Hence, in his seminal article on the subject, Fama (1970) classified empirical tests of market efficiency into three “forms” (weak, semi-strong, and strong) based on the particular subset of available information considered. Weak-form efficiency tests examine if a security’s current price fully reflects its historical prices (as well as the prices of other assets). As one court has observed, “the weak form, which asserts simply that the current share price in an efficient market reflects all information about past share prices. If the weak form of the hypothesis accurately describes a market, it is impossible to predict future prices using only past prices.”69 “[T]he weak form … posits that all previous stock prices, which are necessarily a subset of all public information, are already incorporated into current stock prices.”70 Such tests include examining whether prices are “serially correlated” e.g., whether they follow a predictable trend based on one or more prior day’s movement or instead follow a “random walk.” A momentum-based strategy such as “buy-high/sell-low” which is based on stale information about historical prices should not outperform a simple “buy-and-hold” strategy in an efficiently functioning market after accounting for transaction costs.71 Semi-strong form efficiency tests if a security’s current price fully reflects a larger set of information, namely all public information (not only historical prices). It is generally accepted by the courts that the fraud-on-themarket theory is based on the semi-strong form of market efficiency.72 It follows that, if a stock’s price does not fully impound information related to historical prices (i.e., does not trade in a market that is even weak-form efficient), then it cannot be said to trade in a semi-strong form efficient
Assessing Market Efficiency for Reliance
175
market.73 Therefore, tests of semi-strong form efficiency such as event studies used to establish cause and effect the most important Cammer factor are relevant only if it is first confirmed that the security at issue trades in a weak-form efficient market. In theory, in a strong form efficient market, a security’s price fully reflects not just historical price data or all publicly available information, but all possible information, public and private, including information available to insiders. Financial economists believe that markets are not strong form efficient; the concept of strong form of market efficiency thus is simply a theoretical benchmark.74
Weak-Form Efficiency Tests The central feature of an informationally efficient market is succinctly summarized in the title of Nobel Laureate Paul Samuelson’s seminal article published in 1965: “Proof that Properly Anticipated Prices Fluctuate Randomly.” In other words, for a stock to trade in an informationally efficient market (i.e., a market in which the stock’s price fully incorporates available information), it must be impossible to forecast future price changes from observing prior price changes. Therefore, tests of the predictability of stock returns are commonly used to preliminarily assess market efficiency. A common measure of such returns predictability is “serial correlation,” which measures mathematically the “correlation between the current return on a security and the return on the same security over a later period.”75 As Ross et al. (2002), a well-known basic finance text notes:76 A positive coefficient of serial correlation for a particular stock indicates a tendency toward continuation. That is, a higher-than-average return today is likely to be followed by higher-than-average returns in the future. Similarly, a lower-than average return today is likely to be followed by lower-than-average returns in the future. A negative coefficient of serial correlation for a particular stock indicates a tendency toward reversal. A higher-than-average return today is likely to be followed by lower-than average returns in the future. Similarly, a lower-than-average return today is likely to be followed by higher-than-average returns in the future. Both significantly positive and significantly negative serial-correlation coefficients are indications of market inefficiencies; in either case, returns today can be used to predict future returns.
Thus, contrary to some recent commentary,77 statistically significant serial correlation in publicly traded stocks is rare, and according to
176
MUKESH BAJAJ ET AL.
research by Nobel Laureate Clive Granger, if serial correlation in a stock’s returns is found to be “large enough to cover the size of transaction costs,” the finding “invalidate[s]” the conclusion that the stock trades in an efficient market.78 Thus, serial correlation tests, which date back to studies in the 1960s by Cootner (1962) and Fama (1965, 1970), continue to be used in current research. For instance, a recent study by Erenburg et al. (2011) used serial correlation and other common tests of weak-form efficiency developed in the finance literature (runs tests and tests of profitable momentum strategies) to identify stocks at issue in federal class actions filed in 1996 and 1997 that did not trade in weak-form efficient markets according to such empirical tests. It next examined if five indicators of efficiency commonly considered by the courts could also detect such weak-form inefficiency and concluded that the factors courts commonly considered exhibited “little relation to weak-form market efficiency.”79 Specifically, it found the “blanket presumption of market efficiency for NYSE-listed firms or the Cammer court’s presumption of greater market efficiency for NYSE-listed firms relative to NASDAQ” unsupportable because both groups displayed “evidence of weak-form inefficiency” such as serial correlation to a degree that could not be reasonably attributed to chance alone.80 The study also found the “legal presumption” that stocks with coverage by more analysts will be more efficiently priced was “backward” because analyst following was related to one-day serial correlation and stocks followed by more analysts were candidates for profitable oneday momentum strategies.81 The study also concluded that trading volume (turnover) was not indicative of efficiency, contrary to the courts’ intuition. Instead, “inefficient pricing (or “slow price adjustment”) induces more trading.”82 In a similar vein, Erenburg et al. (2011) also found that highmarket-cap firms [with market capitalization of at least $1 billion] have more positive serial correlation relative to other firms in their sample, which again contradicts the court’s intuition.83 Finally, Erenburg et al. (2011) found that wide bid-ask spreads, which is considered an indicator of inefficiency according to the Krogman court, also impeded profitable momentum strategies (which exist if the market is weak-form inefficient). Hence, again counter to the court’s opinion, the evidence “indicates that profitable momentum trading is more likely for firms with narrower spreads.”84 Significant serial correlation is indicative of weak-form inefficiency, but it is not the only test. Another way of measuring weak-form inefficiency is through the use of “trading rules” tests to see if a profit could have been earned simply by buying and selling the stock based on its observed price
Assessing Market Efficiency for Reliance
177
movement in order to capitalize on momentum.85 Fama (1970) noted that is it “difficult to judge what degree of serial correlation would imply the existence of trading rules with substantial expected profits,” so he suggests that “for many reasons it is desirable to directly test the profitability of various trading rules.” While a variety of filter rules can be considered, a simple one is the “y-filter” based trading rule, which examines if a momentum strategy, devised without the benefit of hindsight, earns greater profits than a buy-and-hold strategy.86 In a weak-form efficient market, such excess profits should be impossible, net of transaction costs.87 If net of reasonable cost assumptions, y-filter rules indicate profits in excess of the buy-andhold benchmark then the evidence suggests that the security at issue did not trade in even a weak-form inefficient market over the relevant period. The scope of weak-form efficiency tests may be expanded to include “tests for return predictability,” e.g., “forecasting returns with variables like dividend yields and interest rates.”88 Hence, in the spirit of Fama (1991), one may examine (using regression analysis and daily or intra-day data if available) if the security’s returns are statistically significantly related to lagged returns of other securities (in addition to serial correlation tests) to assess weak-form efficiency. If a stock’s future price change can be forecast in a statistically significant manner based on past changes in the price of a related security (such as preferred stocks or bonds issued by the same issuer, or the stocks of other companies in the same industry) then such predictable price trends based on information that was already known to investors would indicate that the stock at issue did not trade in a weakform efficient market. Recent academic research has shown that not all publicly traded stocks always trade in a market that is even weak-form efficient. In fact, “momentum” (the tendency for rising asset prices to rise further and falling prices to keep falling)89 is a pervasive anomaly that has been repeatedly documented and is considered by many to be an indicator of market (weak-form) inefficiency.
Semi-Strong Form Efficiency Tests Courts have generally concluded that to invoke the fraud-on-the-market doctrine plaintiffs must prove that the security at issue traded in a semistrong form efficient market90 in which the stock price quickly impounds all publicly available information (i.e. is informationally efficient). Semi-strong form efficiency is commonly assessed by examining if a security’s price
178
MUKESH BAJAJ ET AL.
quickly and correctly impounds new information (the “cause and effect” Cammer factor) using a statistical technique known as “event study.”91 The mechanics of an event study are straightforward. The study has the following steps: (1) The event window (day or days) on which the market received unexpected news is identified. (2) The stock’s return or price change over the event window is calculated. (3) The stock’s “residual” return (also referred to at times as the “market-adjusted” or “abnormal” return) over the event window is calculated. The residual return is typically estimated by calculating the “predicted” return “using a [regression] model that takes into account market and industry effects on stock price returns … [and subtracting] the predicted return from the actual return.”92 (4) The residual return fluctuates daily even in the absence of any news. Hence, a statistical measure referred to as the “t-statistic” is then commonly used to determine if the residual return on a particular day was abnormally large (or significantly different from zero), given its past fluctuations.93 If a stock’s abnormal return is statistically insignificant (or not distinguishable from zero with a high degree of certitude), then the event study result does not support the conclusion that the stock price reacted to the identified news. The converse, however, is not necessarily true. A statistically significant price reaction on any day (even following some contemporaneous news release) does not necessarily prove that such news caused the observed price reaction, because correlation alone is not proof of causation. To draw an economic conclusion regarding “cause and effect,” further analysis is critical for several reasons. First, the mechanical steps of an event study can be conducted for any day and the results of such a mechanical calculation alone do not explain what caused the observed price reaction. Although an event study can be conducted for any security, the study’s results (which are used to assess semi-strong for efficiency) have little probative value unless it is first established that the security at issue traded in a weak-form efficient market. If a security does not trade in even a weak-form efficient market, then its price change could in part be attributable to a delayed reaction to stale information (including information about its past prices) or momentum rather than any new information released that day. The presence of such an alternative explanation for the price movement makes it unreasonable to reject the “null hypothesis” of no cause and effect solely on the basis of correlation between the event and the price movement; a study simply cannot conclude what caused the movement with confidence. Thus, the empirical finding that the security’s residual return following a news release is statistically significant does not prove that observed price reaction was a quick and
Assessing Market Efficiency for Reliance
179
correct reaction to such news. Without such proof, a statistically significant price movement does not demonstrate cause and effect, the “most important” Cammer factor related to efficiency.94 Second, it is important to recognize that the first step of an event study (identifying the event date when unexpected news first became public) is critical. Instead of arbitrarily focusing on some selected dates, or certain dates identified in the complaint, a careful review of prior mix of information available in the market is necessary. Otherwise, it is impossible to ascertain if the identified event was unexpected new information (“news”) or simply stale information. If a security trades in an efficient market, then its price should not react to stale information because such information would have been impounded into the stock price soon after it was first released.95 Therefore, a statistically significant abnormal price reaction attributed to such information would not prove cause and effect. To the contrary, such a finding would support the opposite conclusion, i.e., that the security did not trade in an efficient market. Third, a proper event study for purposes of determining market efficiency over a class period must look for consistent demonstration of market efficiency, not just occasional price movements on days with news. The results of an event study based on a small sample of dates when the security price was already known to have declined sharply are not probative. For instance, to prove the cause-and-effect relationship, some economic reports focus on only a handful of dates out of the alleged class period which are typically those identified in the complaint when the stock price was known to decrease significantly. Some have argued that such a limited enquiry (rather than proving that the security at issues traded in an efficient market throughout the class period) is sufficient to prove reliance for class certification purposes.96 We believe such a claim does not constitute the “rigorous analysis” demanded by Wal-Mart for certification purposes. One reason why is hindsight bias. Some of the news releases may be identified as material misrepresentation or curative disclosure dates in the complaint.97 When a security’s observed daily return is large, its residual return is typically also large and statistically significant.98 Thus, confirming that the stock’s price declined “significantly” on a large price-drop date amounts to merely repeating the complaint’s allegations, as such a date was likely identified as at least a partial curative disclosure in the complaint because the stock price was known to have dropped sharply that day. Statistical significance on a particular date, a technical finding, does not by itself prove that a consistent cause-and-effect relationship existed throughout the relevant class period.99
180
MUKESH BAJAJ ET AL.
A second reason is that episodic events are not proof of a consistent cause-and-effect relationship. A limited showing of a relationship between news and price movements on a handful of dates does not demonstrate that the stock traded in an informationally efficient market throughout the alleged class period. From an economic perspective, a rigorous analysis of the security’s price reaction over time is necessary to prove reliance. In PolyMedica II, for example, the plaintiff’s expert had found that the stock price reacted significantly to news on 5 (or 3.13%) of 160 trading days in the contested period in that case.100 However, the Court opined that a “mere listing of five days on which news was released and which exhibited large price fluctuations proves nothing.”101 Ferrillo et al. (2004) also reason that, to establish that the security at issue traded in an informationally efficient market, it is not adequate to demonstrate that its price reacted statistically significantly on a handful of news dates. Instead they argue that, to prove efficiency for reliance purposes, plaintiffs must demonstrate that the security’s price reacted statistically significantly more often on days with news (“news dates”) compared to days without news.102 However, the power of Ferrillo et al.’s test depends critically on how such news dates are identified. If “news dates” are indeed those dates when the market learnt material new information then the security’s price should reacted significantly to “most new, material news” as the court noted in Freddie Mac.103 If, instead, news dates are picked mechanically by simply identifying all dates when the company was “mentioned” in the press as Ferrillo et al. (2004) suggest,104 such purported “news” may in fact be stale information or immaterial. Even in an efficient market one would not expect to see a significant price reaction on such days. Notably, large companies are mentioned in the press almost daily. Hence almost all days in any sample period could be identified as news dates in such cases, which would limit the applicability of the Ferrillo et al. test. Given these limitations, in Freddie Mac, the Court found the plaintiff expert’s use of Ferrillo et al.’s test unreliable.105 Moreover, simply showing that the market reacted more often to news than to non-news does not, logically, support presuming that it always reacted to news of the type contained in the alleged misrepresentations the very presumption that is the entire purpose of using fraud-on-the-market as a basis for class certification. For purposes of litigation, a focus on price reactions following a few major negative news events affecting the company does not prove that the security traded in a market that was consistently efficient throughout the class period, as plaintiffs must in order to claim reliance on statements that
Assessing Market Efficiency for Reliance
181
may not have been as dramatically newsworthy. The corrective event at the end of a class period in litigation is often a drastic event such as a bankruptcy, receivership or major restatement of earnings. A security’s price will typically show some reaction to such a drastic event. However, such a price reaction on a single date (or even a handful of selected disclosure dates) does not prove that the security traded in market that was sufficiently efficient to incorporate the effect of all prior statements at issue throughout the relevant class period. It is not infrequent in securities class actions that there is no statistically significant price reaction on the dates of the alleged misrepresentations. The plaintiffs in such cases often argue the speculative proposition that the misstatement merely confirmed expectations and should not have been expected to move the market price and that the amount of price inflation should be inferred from the subsequent stock drop.106 For instance, Langevoort (2009) argues, “This is particularly apt when what plaintiffs allege is an omission rather than affirmative lie: the omission will not necessarily lead to an identifiable market move rather, plaintiffs’ claim is that the market would have adjusted had the truth been told.” The problem with such a theory is that it is not testable. It is doubly problematic in such cases, where price impact is inferred from a later price decline, to also use that decline as proof of market efficiency. The better course is to require a rigorous showing by other evidence that the price actually did react consistently to unexpected events before the stock drop. In other cases, plaintiffs have sought to show efficiency based on proof that the market reacted on some percentage of dates, but not others a conclusion that, applied to a class, would suggest that misstatements could be presumed to have caused harm to all class members when the actual empirical evidence suggests that only a much lower percentage were affected. This is an unacceptable basis for certifying a class and awarding damages, as the Supreme Court held in Comcast.107 After Comcast, a court would not certify an antitrust case where 40% of the class suffered no injury; neither should it certify a securities class where an event study shows that the market failed to react to new, material information 30% or 40% of the time, or where an event study simply failed to examine most of the days in the class period. For example, in Freddie Mac, the plaintiffs’ expert had set aside 136 of the 193 trading days in the alleged class period for the security at issue (Freddie Mac’s Series Z preferred stock) and reported that, over the remaining segment of the class period, the security’s abnormal return was statistically significant on 28% of the news days he had considered. The
182
MUKESH BAJAJ ET AL.
Court found this proof inadequate to establish efficiency for reliance purposes, noting that “[a] plaintiff must show that the market price responds to most new, material news.”108 Citing Bajaj (one of this article’s authors who had testified as an expert for the defendants109), the Court noted that “an economist may conclude that a market is efficient if it reacts to news 80 to 90% of the time, depending on the number of news dates at issue.”110 In China Automotive Systems, the plaintiff’s expert found that the stock price reacted significantly to news on 7 out of 16 identified dates but in the wrong direction on another day. The Court opined that “[e]ven assuming that the methodology was proper, showing that only seven out of sixteen days resulted in a market reaction is an insufficient foundation upon which to pronounce market efficiency.”111 To confirm from an economic perspective that investors could have relied on the integrity of the security’s price throughout the alleged class period and to provide an economic basis for a legal presumption that all investors were equally affected at all times it is necessary to conduct a comprehensive analysis of all news released during that entire period and confirm using well-accepted scientific tests that the security’s price consistently reacted quickly and correctly to all such news releases.
Law of One Price Tests The Law of One Price is a central tenet in financial economics and a fundamental criterion for an efficient market. Hence, tests for Law of One Price violations constitute a basic and direct test of efficiency.112 For example, put-call parity tests are commonly used to tests for violations of the Law of One Price. In this test, a portfolio of a call and a put option on the stock and a risk-free bond is constructed to have the same cash flow as the stock by design. The portfolio and the stock are therefore, in effect, identical securities that should always trade at identical prices in an efficient market. This fundamental relationship between the value of the portfolio and the stock is known as put-call parity and is described in most standard finance texts.113 As contemporaneous (and past) prices of the options, bond and stock are publicly known, violations of this parity relationship that are sufficiently large to yield arbitrage profits net of transaction costs such as “bid-ask spreads” and commissions suggest that the security at issue does not trade in an informationally efficient market. Such direct tests of efficiency are more probative in distinguishing between efficient and inefficient securities markets114 whereas indirect tests
Assessing Market Efficiency for Reliance
183
using ad hoc proxies for efficiency such as Cammer factors cannot, as we explain in the next section.
THE CAMMER FACTORS ARE NOT CONCLUSIVE PROOF OF MARKET EFFICIENCY Some scholars have identified defects in the use of Cammer factors for assessing market efficiency. Langevoort refers to the list as a “jumble,” some redundant, and “[a]s with most multi-factor lists, Cammer is unclear what is to be done except examine the factors in order. It invited an ad hoc approach informed by expert testimony, but in fact largely unconstrained.” Barber et al. (1994) also refer to the Cammer factors as “ad hoc,” and note that, “We know of no systematic body of evidence showing that these or any other criteria distinguish between efficient and inefficient stocks. Nor are we aware of evidence supporting specific cutoff values of these criteria.”115 Yet, limited by the reports experts have presented to them, courts have continued to rely on Cammer factors to evaluate efficiency, viewing them as intuitive proxies for efficiency. Such a view, however, is at odds with conclusions based on rigorous economic analysis. For instance, Erenburg et al. (2011) construct different measures of weak-form efficiency (including serial correlation) for securities at issue in 236 federal securities class actions cases filed during 1996 and 1997 and find that the “actual relation of the [Cammer] factors to weak-form efficiency is sometimes the converse of the courts’ intuition.” The study concludes that:116 (1) some cases certified for class action status do not satisfy the conditions for even weak-form efficiency; (2) numerous opportunities exist for cost-effective investors (those who can trade quickly and at low cost) to profit by using simple momentum-based strategies; (3) including such investors as class members effectively subsidizes their strategies and overstates damages from reliance on market efficiency; (4) when such investors can profit by rejecting market efficiency, standard measures of damage overstate the fraud-related damage of other investors; (5) because of endogeneity, the factors that commonly are relied on by the courts for determining market efficiency bear little or no relation to weak-form efficiency. [emphasis added]
Thus, Erenburg et al. (2011) noted, there is an “inconsistency between the efficient markets hypothesis and the way U.S. courts have applied the hypothesis in cases involving allegations of fraud-on-the-market.”117 Their “findings raise serious questions about the standards used by the courts in
184
MUKESH BAJAJ ET AL.
granting motions for class action status and about the economic appropriateness of routinely presuming universal reliance on market efficiency when certifying broad classes consisting of all investors who traded during the class period.”118 We also assess the probative value of Cammer factors next by examining seven well-known instances in which stock prices violated the Law of One Price and thus did not trade in informationally efficient markets according to prior studies. Yet, we demonstrate, they satisfied the Cammer factors indicating once again these factors have little probative value in assessing efficiency. The case of the “twin stocks,” Royal Dutch/Shell, identified by Lamont and Thaler (2003a) is one of the stocks in our sample. In this case, a single firm, the Royal Dutch/Shell Group, has two sets of traded shares (Royal Dutch shares traded in Amsterdam and Shell shares traded in London). Contractually, Royal Dutch shares receive 60% and Shell shares receive 40% of the firm’s cash flows. Therefore, the market value of the Royal Dutch shares should be 1.5 times the market value of Shell shares. However, as Lamont and Thaler (2003a) found, this ratio has varied considerably from its theoretical value, “from 30 percent too low in 1981 to more than 15 percent too high in 1996.” The authors noted that such a large and persistent violation of the Law of One was “surprising since both Royal Dutch and Shell trade in highly liquid and open markets in Europe and, additionally, have ADRs trading in the United States. Thus, to profit from the mispricing, a U.S. investor doesn’t even need to trade in international markets. All that is necessary is to short the overpriced shares, buy the underpriced shares and hold forever.”119 In another study, Lamont and Thaler (2003b) identified a sample of 18 cases from April 1996 to August 2000 of companies that conducted initial public offerings (IPOs) in a subsidiary (known as a “carve-out”)120 with the announced intention of spinning off the rest of the subsidiary to the parent company’s shareholders at a later date. In six cases, the aftermarket price of the subsidiary was so high that if the same value was attached to the remaining shares owned by the parent, the implied value of the rest of the parent’s assets would be negative, which is of course impossible in an efficient market because a stock investment represents limited liability (i.e., a shareholder’s loss is limited to the amount he paid for the shares). For example, in 1999 the Silicon Valley technology company, 3Com spun off a small portion of its handheld computer division, Palm, through an IPO, while retaining 95% of Palm which would be distributed in about six
Assessing Market Efficiency for Reliance
185
months to 3Com shareholders, when each 3Com shareholder would receive 1.5 shares of Palm. Hence, following Palm’s IPO, each 3Com share was expected to be valued at 1.5 times the price of a Palm share, plus the per share value of the remainder of the 3Com business (excluding Palm). However, after its IPO, Palm’s stock price rose significantly while that of 3Com actually fell,121 to the point that the market value of the 95% stake in Palm that 3Com still held was higher than the remaining business of 3Com, i.e. the rest of 3Com’s business had a negative value (−$22 billion). This irrational mispricing was widely discussed in the press the day after Palm’s IPO, “including in two articles in the Wall Street Journal, one in the New York Times, and … USA Today.”122 In another instance, Fedenia and Hirschey (2009) found Chipotle’s Class A shares traded at a premium to Class B shares even though the former had inferior voting rights.123 Table 1 lists the cases we have analyzed and their relevant analysis periods.124 Each of these represents a well-known case of securities that have not traded in informationally efficient markets for specific periods. As Table 2 shows, despite violating the Law of One Price and trading in an inefficient market over particular analysis periods, these securities satisfied the Cammer factors (such as sufficient average weekly trading volume, analyst coverage, market capitalization, and a low bid-ask spread)125 over their respective analysis periods and were publicly traded on NYSE/ NASDAQ.126 Among the criterion for eligibility to file form S-3, the Court in Cammer noted that it is “the number of shares traded and value of shares outstanding that involve the facts which imply efficiency.” As Table 2 shows, all the firms had significantly higher market capitalization compared to the required public float to be eligible to file form S-3.127 To assess the fifth Cammer factor, we analyzed the stock price reaction to earnings announcements in the cases of (i) the five parent companies which had negative stub values following their subsidiaries’ spin-offs listed in Table 1, panel A; and (ii) the Royal Dutch ADR, the Shell ADR, the Chipotle Class A, and the Chipotle Class B shares, listed in panel B of Table 1.128 We identified 38 instances in which these companies announced quarterly earnings that constituted a “surprise”129 relative to the consensus (mean) analyst forecast and were not coupled with any confounding news (See Table 3). In 35 of these 38 cases, these companies’ stock prices moved immediately in the right direction, i.e., increased following a positive earnings surprise and declined following a negative surprise. We also confirmed that in these 35 cases the price reactions were in the right direction even on a market-adjusted basis, using the typical market model we see in many
Parent
Subsidiary
a
Creative Computers stock HNC Software stock Daisytek stock 3Com stock Methode Electronics Class A stock
UBID stock Retek stock PFSWeb stock Palm stock Stratos Lightwave stock
Chipotle Class A Stock
7
Shell Trading & Transport Co. ADR Chipotle Class B Stock October 16, 2006
January 1, 1997
December 3, 1998 November 17, 1999 December 1, 1999 March 1, 2000 June 26, 2000
August 31, 2008
June 30, 2002
June 7, 1999 September 29, 2000 July 6, 2000 July 27, 2000 April 28, 2001
End Date
In the five “negative stub” cases, the implied market value of the parent company’s residual assets (excluding the value of its shares in a spun-off subsidiary) was negative following the spin-off (or IPO of shares) of a subsidiary [See Lamont and Thaler (2003b)]. For the negative stub cases identified by Lamont and Thaler (2003b) that we have analyzed, the analysis periods are as defined by Lamont and Thaler (2003b). The analysis period for the Chipotle case is as defined in Fedenia and Hirschey (2009), which documented Chipotle’s violation of the Law of One Price. See Lamont and Thaler (2003a), which documented the Royal Dutch and Shell pricing anomaly from 1990 to August 2002. For illustrative purposes, we have focused on the sub-period January 1, 1997June 30, 2002, which spans more than five years, in our analysis of the Royal Dutch/Shell example. b In the Royal Dutch/Shell example, the Royal Dutch ADR’s claim to the cash flows of the Royal Dutch/Shell Plc. was 1.5 times than that of the Shell ADR’s claim. Yet, the market value of the Royal Dutch ADR violated parity (i.e., 1.5 times the value of the Shell ADR) consistently for almost a decade [See Lamont and Thaler (2003a)]. In the Chipotle case, Chipotle had two classes of traded shares. Even though they had inferior voting rights, however, the Class A shares were found to have traded “at a persistent price premium of as much as 20% more than superior Class B shares.” See Fedenia, M. and Hirschey, M. (2009), The Chipotle paradox. Journal of Applied Finance, Issues 1 & 2, pp. 116.
a
Royal Dutch ADR
6
B: Other Pairs of Securities for Which the Law of One Price Was Violatedb
1 2 3 4 5
Start Date
Analysis Period
Analyzed Securities that Violated the Law of One Price Unger Factors Were Satisfied.
A: Instances in Which the Parent Company’s Assets Had “Negative Stub” Value
#
Table 1. 186 MUKESH BAJAJ ET AL.
Security
July 27, 2000 April 28, 2001
June 30, 2002 June 30, 2002 August 31, 2008 August 31, 2008
January 1, 1997 January 1 1997 October 16, 2006 October 16, 2006
June 7, 1999 September 29, 2000 July 6, 2000
End Period
December 3, 1998 November 17, 1999 December 1, 1999 March 1, 2000 June 26, 2000
Start Period
5.78
23.28
3.39
1.45
8.76
15.42
8.10
8.92
38.43
Average Weekly Turnover (%)
101 Analysts, 1455 Reports 101 Analysts, 1455 Reports 27 Analysts, 172 Reports 27 Analysts, 172 Reports
3 Analysts, 8 Reports 9 Analysts, 49 Reports 6 Analysts, 17 Reports 4 Analysts, 10 Reports 5 Analysts, 36 Reports
Number of Analysts
N/A
N/A
July 20, 2005
July 20, 2005
August 8, 1995
June 22, 1995
February 26, 1998
March 4, 1997
October 17, 2007
Date of Earliest S-3/F-3 Filingsa
$1,484,489,924
$1,297,857,287
$1,990,918,927
$45,399,477,394
$1,090,986,919
$19,037,647,538
$298,065,642
$1,786,106,220
$335,177,990
Average Market Capitalization
Cammer Unger Factor Related Evidence for Selected Securities.
0.28
0.17
0.33
0.11
0.49
0.13
0.82
0.37
0.81
Daily Average BidAsk Spread (%)
a Date of earliest filing of form S-3 is from EDGAR database maintained by SEC. Foreign companies file form F-3. Per SEC, the general eligibility conditions to file Form F-3 are the “same as Form S-3, except the company must be a foreign private issuer (that is, not a domestic (U.S.) company).” [“Eligibility of Smaller Companies to Use Form S-3 or F-3 for Primary Securities Offerings,” available at SEC website on http://www.sec.gov/info/smallbus/secg/s3f3-secg. htm]. Sources: Stock price, volume and shares outstanding data is from CRSP. Number of analysts covering a company was determined using Thomson Reuters database.
5 Methode Electronics (Class A shares) 6 Royal Dutch/Shell (Amsterdam) 7 Shell Transport & Trading 8 Chipotle Class A shares 9 Chipotle Class B shares
4 3Com
3 Daisytek
1 Creative Computers 2 HNC Software
#
Table 2.
Assessing Market Efficiency for Reliance 187
16. 17. 18. 19. 20. 21. 22. 23. 24.
12. 13. 14. 15.
7. 8. 9. 10. 11.
2. 3. 4. 5. 6.
1.
Creative Computers (PC Mall Inc) HNC Software HNC Software HNC Software Royal Dutch Petroleum
Chipotle Mexican Grill Inc (Class B)
Chipotle Mexican Grill Inc (Class A)
Security
August 5, 1999 February 10, 2000 February 8, 2001 May 2, 2002 August 7, 1997 February12, 1998 August 6, 1998 November 5, 1998 February 11, 1999
January 26, 2000 April 19, 2000 July19, 2000 May 6, 1999
May 1, 2007 July 31, 2007 February 14, 2008 July 23, 2008 February 8, 1999
May 1, 2007 July 31, 2007 February 14, 2008 July 23, 2008 February 15, 2007
February 15, 2007
Release Date
August 5, 1999 February 10, 2000 February 8, 2001 May 2, 2002 August 7, 1997 February 12, 1998 August 6, 1998 November 5, 1998 February 11, 1999
January 27, 2000 April 20, 2000 July 20, 2000 May 6, 1999
May 2, 2007 August 1, 2007 February 15, 2008 July 24, 2008 February 8, 1999
May 2, 2007 August 1, 2007 February 15, 2008 July 24, 2008 February 16, 2007
February16, 2007
Event Datea 6.0 16.9 11.2 −3.6 −16.6 7.2 16.5 10.3 −1.9 −17.8 −5.0 9.1 −4.6 −5.0 3.7 3.1 4.8 0.2 2.1 −3.4 −2.1 −8.0 −4.6 −4.1
17.9* 12.3* −3.1* −19.7* 6.9* 17.3* 11.5* −1.5* −21.2* −3.3* 9.3* −3.5 −3.6 2.8* 3.6* 5.0* 0.2* 2.0* −4.0* −1.6* −7.0* −3.3* −2.1*
Residual Return (%)
5.9*
Raw Return (%)
Table 3. Earnings Surprise and Price Reactions.
1.4 2.6** 0.1 1.5 −3.5** −1.6 −5.5** −2.6** −2.0**
1.5 −0.8 −0.9 1.8
10.5** 4.9** −0.7 −6.5** −0.5
8.1** 5.2** −1.4 −6.1** 4.5**
2.5**
t-Statistic
+ + + + − − − − −
+ + + +
+ + − − −
+ + − − +
+
Earnings Surpriseb
188 MUKESH BAJAJ ET AL.
Shell
November 4, 1999 November 2, 2000 May 6, 1999 August 5, 1999 February 10, 2000 February 8, 2001 May 2, 2002 August 7, 1997 February 12, 1998 August 6, 1998 November 5, 1998 February 11, 1999 November 4, 1999 November 2, 2000
Release Date November 4, 1999 November 2, 2000 May 6, 1999 August 5, 1999 February10, 2000 February 8, 2001 May 2, 2002 August 7, 1997 February 12, 1998 August 6, 1998 November 5, 1998 February 11, 1999 November 4, 1999 November 2, 2000
Event Datea Residual Return (%) −5.4 −7.0 4.6 4.4 7.9 −0.1 3.6 −4.2 −1.5 −5.1 −5.2 −4.2 −5.4 −6.7
Raw Return (%) −5.0* −6.8* 3.9* 4.9* 8.2* −0.2 3.4* −4.7* −1.1* −4.4* −4.1* −2.6* −5.1* −6.4*
−2.9** −3.6** 2.3** 2.1** 4.1** −0.1 2.3** −4.3** −1.2 −4.0** −3.2** −2.3** −2.9** −3.2**
t-Statistic − − + + + + + − − − − − − −
Earnings Surpriseb
**
Raw return is consistent with earnings surprise. Market-adjusted return is statistically significant and in the right direction. a Event date is the trading day on which impact of earnings announcement is analyzed. For earnings announced after trading hours, event day equals the first trading day following announcement of earnings. b Positive and negative earnings surprise are denoted by “+” and “−”.
*
25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38.
Security
Assessing Market Efficiency for Reliance 189
190
MUKESH BAJAJ ET AL.
economic reports for class certification purposes.130 Further, the marketadjusted price reaction was statistically significant and in the right direction in 25 of the 38 cases analyzed.131 In short, the securities analyzed above satisfied the Cammer factors, including appearing to satisfy the test for cause and effect. However, these securities traded in markets that were clearly inefficient over the relevant analysis periods. This finding confirms that even if a stock satisfies an ad hoc list of factors considered indications of efficiency, it does not constitute adequate proof that the stock actually traded in an efficient market.
THE GLOBAL LIQUIDITY CRISIS OF 20072008, LIMITS TO ARBITRAGE AND MARKET INEFFICIENCY A global liquidity and credit crisis of unprecedented magnitude erupted on August 9, 2007, when France’s largest bank, BNP Paribas, announced that it had frozen redemptions for three investment funds, because of the “complete evaporation of liquidity in certain market segments of the US securitization market [which had] … made it impossible to value certain assets fairly regardless of their quality or credit rating.”132 When liquidity, a key ingredient for arbitrage evaporates, it triggers several knock-on effects that amplify economic shocks to a “full-blown financial crisis.”133 As Brunnermeier (2009) noted, “When asset prices drop, financial institutions’ capital erodes and, at the same time, lending standards and margins tighten. Both effects [which Brunnermeier referred to as ‘liquidity spirals’] cause fire-sales, pushing down prices and tightening funding even further.” The mortgage crisis also amplified because (i) runs on financial institutions, like those on Bear Stearns, Lehman Brothers, and Washington Mutual, suddenly eroded bank capital and (ii) financial counterparties began to hold additional funds to protect themselves from counterparty risk that are not netted out (which can occur when multiple trading parties fail to cancel out offsetting positions because of concerns about counterparty credit risk). The liquidity crunch resulted in the LIBOR-OIS spread,134 which former Fed Chairman Greenspan described as a “barometer of fears of bank insolvency”135 tripling overnight from 13.4 basis points (bps) on August 8 to nearly 40 bps on August 9, 2007. Central banks around the world injected liquidity into their national banking systems to prevent asset sales at
Assessing Market Efficiency for Reliance
191
depressed “fire-sale” prices and further deleveraging by broker dealers.136 Noting that loss of confidence in markets due to false rumors could result in panic selling, and led stock prices to “artificially and unnecessarily decline well below the price level that would have resulted from the normal price discovery process,”137 the Securities and Exchange Commission (“SEC”) banned naked short selling of the stocks of 19 major financial institutions on July 15, 2008 and nine days later announced that it would expand the list of companies covered by the ban to the entire market.138 Despite these measures, by September 2008 the financial markets were on the brink of collapse and illiquidity had reached unprecedented proportions. By the end of that month, the financial landscape was forever changed as several large financial institutions were either taken over by the U.S. government, failed or merged with other firms. On Sunday, September 7, 2008, Freddie Mac and Fannie Mae (government-sponsored entities established to provide liquidity in the secondary mortgage market) were put into receivership by the U.S. government in one of the largest bailouts in U.S. history. On Monday, September 15, 2008, Lehman Brothers Holdings Inc. filed for bankruptcy protection “one of the biggest credit events in history”139 and Merrill Lynch announced it would be acquired by Bank of America. On September 16, 2008, AIG, a major insurer of credit risk, was taken over by the U.S. government.140 A wave of deposit withdrawals triggered the failure of the Seattle thrift, Washington Mutual, the largest bank failure in U.S. history141 and Goldman Sachs and Morgan Stanley, the only two large remaining U.S. investment banks announced plans to become bank holding companies to obtain liquidity from the U.S. Federal Reserve.142 As noted earlier, markets are rendered efficient through the costly efforts of arbitrageurs to gather relevant information and trade to profit from such information. Such arbitrage is essential for a market to be informationally efficient. Arbitrage is conducted mainly by institutional investors such as hedge funds that rely on external funding.143 Thus, when investors withdrew their capital from hedge funds during the crisis, the funds’ ability to arbitrage by purchasing additional undervalued assets became severely limited and funds were forced to sell assets at a huge loss. In 2008, the typical fund lost about one-fifth of its value and “convertible arbitrage” funds (that try to exploit price anomalies among corporate bonds) lost 46%, their worst losses since 1990, according to The Economist. Over the next few quarters, the number of funds, which had risen to over 7,000, was estimated to fall by half.144 The Economist article noted that hedge funds’ lack of liquidity was “simply another part of a vast, debt-dependent ecosystem that is now being starved of oxygen.”145
192
MUKESH BAJAJ ET AL.
Numerous recent academic studies too have documented that the liquidity crisis limited arbitrage and rendered markets inefficient. For instance, Pedersen (2009) focuses on “quants” (quantitative traders who use algorithmic trading strategies) and finds that in early August 2007 such traders “ran for the exits” as liquidity froze. As a result, the cumulative return to a longshort market-neutral value and momentum strategy for U.S. largecap stocks displayed “amazing short-term predictability and volatility,”146 which was clear evidence that even large-cap stocks’ prices during this period were not informationally efficient. Hu, Pan, and Wang (2010) demonstrate that even the U.S. Treasury market, the largest, safest, and most liquid asset market in the world,147 was rendered inefficient during the fall of 2008.148 Griffoli and Ranaldo (2010) from the Swiss National Bank found that the covered interest parity (a noarbitrage condition) was persisted violated during the crisis.149 Garleanu and Pedersen (2011) study the credit default swaps-bond “basis”, which is a measure of the price discrepancy between corporate bonds and CDS, which are securities with nearly identical economic exposures. Thus in a wellfunctioning market, the basis should be zero, which was not the case during the crisis (i.e., the Law of One Price was violated). In sum, security markets are not always efficient. For markets to be efficient, it is essential that arbitrageurs have the incentive and the ability to take advantage of temporary mispricing in securities (given available information). Indeed, it is through such arbitrage that markets become efficient. Thus, arbitrage is the lifeblood of an efficient market and limits to arbitrage can result in a security trading at prices that are obviously informationally inefficient for extended periods. The most recent and dramatic instance of such inefficiency occurred during the global financial crisis of 20072009 when several assets markets became acutely dislocated.
CONCLUSION To seek class certification under Rule 23, securities fraud plaintiffs must prove with rigorous analysis that the security at issue traded in an efficient market in which its price fully reflects all available information throughout the alleged class period. Instead of using well-accepted empirical tests of market efficiency that are grounded in the economic principles of market efficiency theory, courts are often presented with expert testimony that is limited to a mechanical review of ad hoc Cammer factors. Courts usually grant class certification without having the benefit of rigorous economic
Assessing Market Efficiency for Reliance
193
analysis. Our analysis confirms that the factors have little probative value in assessing efficiency. We show that even stocks that have violated the Law of One Price, a fundamental condition for efficiency according to several well-known prior studies, satisfy the Cammer factors. Thus, a review of Cammer factors does not constitute rigorous analysis that the Supreme Court in Wal-Mart ruled was necessary for class certification. Further, the market efficiency hypothesis on which the Supreme Court adopted its judicial doctrine of fraud-on-the-market in Basic in 1988 has been challenged on several fronts in the last 25 years. Today, market efficiency is not considered a foregone conclusion in every instance. In particular, during severe financial crises such as the one during 20072008, arbitrage (the mechanism through which markets become efficient) can be severely restricted. Then, as numerous studies have confirmed, various asset markets can be dislocated and security prices become informationally inefficient. In light of such modern economic research, the continued reliance on the fraud-on-the-market doctrine (and market efficiency) is questionable, as dissenting opinions in the recent Supreme Court decision in Amgen have noted. At the very least, courts should require plaintiffs invoking the fraudon-the-market doctrine to prove the market for the security at issue was efficient throughout the relevant class period using rigorous analysis based on scientific methods that have been well developed in the economics profession over the past 25 years. If security prices are obviously severely dislocated due to limits to arbitrage it is unreasonable to assume that investors would have relied on the “integrity” of such prices. Hence, a mechanical review of Cammer factors which are ad hoc and fail to distinguish between securities that trade in an efficient market from those that do not, does not constitute sufficiently rigorous economic analysis for proving reliance based on the fraud-on-the-market doctrine.
NOTES 1. Basic, Inc. v. Levinson, 485 U.S. 224 (1988). 2. Amgen v. Connecticut Ret. Plan & Trust Funds, 133 S. Ct. 1184, 1188 (2013) (citing Basic, 485 U.S., at 241249). 3. See, for example, Langevoort (2009). 4. Langevoort (2009), p. 167. 5. Cammer v. Bloom, 711 F. Supp. 1264 (D.N.J. 1989). 6. Erenburg et al. (2011), p. 263. 7. Erica P. John Fund, Inc. v. Halliburton Co., 563 U.S. ___, 131 S. Ct. 2179, 2181 (2011) (“Halliburton”).
194
MUKESH BAJAJ ET AL.
8. See Wal-Mart Stores, Inc. v. Dukes, 564 U.S. ___, 131 S. Ct. 2541, 2552 n. 6 (2011); Halliburton, 131 S. Ct., at 218485; Basic, Inc. v. Levinson, 485 U.S. 224, 242 (1988). 9. Basic, Inc. v. Levinson, 485 U.S. 224 (1988). 10. Basic, 485 U.S., at 243, 245 (citations omitted). See also Halliburton, 131 S. Ct., at 2185; Stoneridge Invest. Partners, LLC v. Scientific-Atlanta, Inc., 552 U.S. 148, 159 (2008) (“[U]nder the fraud-on-the-market doctrine, reliance is presumed when the statements at issue become public. The public information is reflected in the market price of the security. Then it can be assumed that an investor who buys or sells stock at the market price relies upon the statement.”). 11. Amgen v. Connecticut Ret. Plan & Trust Funds, 133 S. Ct. 1184, 1188 (2013) (citing Basic, 485 U.S., at 241249). 12. Amgen, 133 S. Ct., at 1213 (Thomas, J., dissenting). “The fraud-on-themarket rule says that purchase or sale of a security in a well-functioning market establishes reliance on a material misrepresentation known to the market. This rule is to be found nowhere in the United States Code or in the common law of fraud or deception; it was invented by the Court in Basic Inc. v. Levinson, 485 U.S. 224 (1988).” Ibid., at 1205 (Scalia, J., dissenting). 13. Basic, 485 U.S., at 246247 & n. 24 (quotations omitted; citing literature on the “efficient-capital-market theory”). 14. Ibid., at 247. 15. PolyMedica I, 432 F.3d, at 8. 16. Halliburton, 131 S. Ct., at 2186 (emphasis added). See also Amgen, 133 S. Ct., at 1195 (Basic presumption “is premised on the understanding that in an efficient market, all publicly available information is rapidly incorporated into, and thus transmitted to investors through, the market price.”). 17. Amgen, 133 S. Ct., at 1192. 18. As Grundfest (2013) noted, “Basic was decided at a time when confidence in the efficient market hypothesis was at its historic peak. Since then, a large literature challenging the efficient market hypothesis has emerged, but that literature has spawned an equally vigorous defense. The debate over market efficiency is nuanced and complex, and it implicates fine points of econometrics and finance theory. It splits leading scholars.” 19. “Correctly” in this context merely means that the market has placed its best aggregate estimate on the value of information; the efficient capital markets hypothesis does not assume that the valuation of a security at a particular point in time will be accurate as a predictive matter, only that it incorporates the collective judgment of the market as to the best estimate of that valuation based on all available public information as of that point in time. As a legal matter, the First Circuit has ruled that reliance requires only “informational efficiency” and not “fundamental value efficiency” i.e., the market price must rapidly reflect all public information but not necessarily be the best possible estimate of the stock’s actual worth. In re PolyMedica Corp. Sec. Litig., 432 F.3d 1, 1419 (1st Cir. 2005) (“PolyMedica I”) (holding that “the market price of the stock fully reflects all publicly available information. By ‘fully reflect,’ we mean that market price responds so quickly to new information that ordinary investors cannot make trading profits on the basis of such information”). This distinction may be
Assessing Market Efficiency for Reliance
195
meaningful in clarifying what the courts mean by efficiency, in the sense that they ordinarily do not need to receive evidence on the generally speculative question of whether the market price was “right” about the value of the company. But as an economic matter, it is not very important. A security’s price on any current date is simply the sum of its historical price at some earlier date and the cumulative price changes thereafter through the current date. If a security trades in an informationally efficient market in which it correctly impounds all new information, then, as a cumulative effect of such price reactions, the security’s price converges over time to its fundamentally efficient value. 20. If information-gathering and trading costs exist, “[a] A weaker and economically more sensible version of the efficiency hypothesis says that prices reflect information to the point where the marginal benefits of acting on information (the profits to be made) do not exceed the marginal costs.” [Fama (1991), p. 1575]. 21. Lo (2008), pp. 23. 22. Amgen, 133 S. Ct., at 1192. 23. Halliburton, 131 S. Ct., at 2182, 2185. 24. 133 S. Ct. at 1198 (citing Wal-Mart, 131 S. Ct., at 2542 n. 6). 25. Ibid., at 1210. 26. Wal-Mart, 131 S. Ct., at 2551 (emphasis in original). 27. Ibid. 28. Comcast Corp. v. Behrend, 133 S. Ct. 1426, 1433 (2013). 29. Wal-Mart, 131 S. Ct., at 2561. 30. Comcast, 133 S. Ct., at 1435 n. 6. 31. Amgen, 133 S. Ct., at 1191. 32. Amgen, 133 S. Ct., at 1198. 33. As the Court held in Halliburton, proof of how and when inflation exits the price is not required to invoke the presumption at the class certification stage, although the Court in that case did not address any situation in which there were individual variations in how inflation was dissipated from the price. 34. Amgen, 133 S. Ct., at 1192. 35. This is true not only with regard to misrepresentations but with regard to other statements considered in the mix of public information. For example, under Basic, courts will commonly presume that markets incorporate disclosures that rendered the alleged misrepresentation stale or immaterial. See, for example, Teachers’ Ret. Sys. of Louisiana v. Hunter, 477 F.3d 162, 187188 (4th Cir. 2007); In re Merck & Co., Inc. Secs. Litig., 432 F.3d 261, 270 (3d Cir. 2005) (“[a]n efficient market for good news is an efficient market for bad news”); Greenberg v. Crossroads Sys., Inc., 364 F.3d 657, 665666 (5th Cir. 2004) (“confirmatory information has already been digested by the market and will not cause a change in stock price.”); Oran v. Stafford, 226 F.3d 275, 282 (3d Cir. 2000) (Alito, J.). The entire project of determining the effect of information in the market over the course of the class period on a common basis assumes that each piece of information entering the market was so reflected. 36. See, for example, Virginia Bankshares, Inc. v. Sandberg, 501 U.S. 1083, 10921096 (1991) (declining to permit litigation confined solely to questions of motivation, unless testable by objective evidence); Blue Chip Stamps v. Manor Drug Stores, 421 U.S. 723, 743 (1975) (warning against litigation of “hazy issues of
196
MUKESH BAJAJ ET AL.
historical fact the proof of which depended almost entirely on oral testimony” regarding whether investors would have bought or sold stock). See also Anza v. Ideal Steel Supply Corp., 547 U.S. 451 (2006) (rejecting “speculative” theory of RICO damages in light of “the difficulty that can arise when a court attempts to ascertain the damages caused by some remote action” and the “intricate, uncertain inquiries” needed to ascertain damages); The Wharf (Holdings) Ltd. v. United Int’l Holdings, Inc., 532 U.S. 588, 594595 (2001) (permitting litigation over oral sales where, unlike Blue Chip Stamps, “both parties would be able to testify as to whether the relevant events had occurred”); Banca Cremi, S.A. v. Alex. Brown & Sons, Inc., 132 F.3d 1017, 1035 (4th Cir. 1997) (court “acutely uncomfortable” with theory of liability where “every element of fraud materiality, reliance, scienter, and proximate cause of damages is inferred or can be presumed.”). 37. “The two men, leading proponents of opposing views about the rationality of financial markets a dispute with important implications for investment strategy, financial regulation and economic policy were joined in unlikely union Monday as winners of the Nobel Memorial Prize in Economic Science. Mr. Fama’s seminal theory of rational, efficient markets inspired the rise of index funds and contributed to the decline of financial regulation. Mr. Shiller, perhaps his most influential critic, carefully assembled evidence of irrational, inefficient behavior and gained a measure of fame by predicting the fall of stock prices in 2000 as well as the housing crash that began in 2006.” (Binyamin, 2013). 38. According to the “behavioral finance” literature (which is now widely accepted and for which the Nobel Prize was recently awarded), investors do not necessarily act rationally, rationally, and their “irrational” behavior affects their attitudes toward risk and their assessment of probabilities [See Barberis and Thaler (2003). The effect of such irrational behavior, especially if it is systemic, may make a market inefficient. Standard graduate texts also now cover behavioral finance. See, for example, Brealey et al., pp. 326327.] 39. This result, known as the “GrossmanStiglitz Paradox,” was first articulated by Nobel laureate Joseph Stiglitz and Sanford Grossman [See Grossman and Stiglitz (1980)]. 40. As Fisher Black, a preeminent financial economist (of BlackScholes option pricing theory fame) noted in his presidential address to the American Finance Association, “All estimates of value are noisy, so we can never know how far away price is from value. However, we might define an efficient market as one in which price is within a factor of 2 of value, i.e., the price is more than half of value and less than twice value.1 The factor of 2 is arbitrary, of course. Intuitively, though, it seems reasonable to me, in the light of sources of uncertainty about value and the strength of the forces tending to cause price to return to value.” [See Black (1986), p. 533]. 41. See, for example, Lamont and Thaler (2003a, b) and Mitchell, Pulvino, and Stafford (2002). 42. Duffie (2010), p. 1245. 43. See Brealey et al., p. 321. 44. See, for example, Fama and French (2012) and Erenburg et al. (2011). 45. See, for example, Langevoort (2009). 46. Langevoort (2009), p. 167. 47. Cammer v. Bloom, 711 F. Supp. 1264 (D.N.J. 1989).
Assessing Market Efficiency for Reliance
197
48. Erenburg et al. (2011), p. 263. 49. Unger v. Amedisys, Inc., 401 F.3d 316, 323 (5th Cir. 2005) (collecting cases). Cammer identified the first five factors, see Cammer v. Bloom, 711 F. Supp. 1264, 12861287 (D.N.J. 1989); later courts added the other three. See Krogman v. Sterritt, 202 F.R.D. 467 (N.D. Tex. 2001). For ease of discussion, unless otherwise noted, we shall henceforth refer to all factors identified in Krogman, Unger, and Cammer (except, the cause-and-effect factor), collectively as the “Cammer factors.” 50. See, for example, In re Merck & Co., Inc. Secs., Deriv. & “ERISA” Litig., MDL No. 1658 (SRC), 2013 WL 396117, at *11 (D.N.J. January 30, 2013). But see In re Fed. Home Loan Mort. Corp. (Freddie Mac) Sec. Litig., 281 F.R.D 174, 178 (S.D.N.Y. 2012) (“[I]t would be illogical to apply a presumption of reliance merely because a security traded within a certain ‘whole market’ without considering the trading characteristics of the individual stock itself.”). 51. Teamsters Local 445 Freight Div. Pension Fund v. Bombardier Inc., 546 F.3d 196, 207208 (2d Cir. 2008) (emphasis added; quotations and citations omitted). 52. For example, Barber, Griffin, and Lev (1994) examine if Cammer factors can differentiate stocks that are (semi-strong form) efficiently priced (display a price response to an announcement of an extreme earnings surprise) from those that are not (i.e., lack such a price response to an announcement of an extreme earnings surprise). They find that except for trading volume and number of analysts following the stocks, other factors commonly used (firm size, percentage of bid-ask spread, return volatility, price, and institutional holdings) “either fail to significance test or yield results counter to our expectations” [Barber et al. (1994), p. 310] i.e., have no probative value in this context. Note however, that even the two factors that Barber et al. (1994) identified as potentially useful in distinguishing stocks as semi-strong form efficient were not useful in detecting weak-form efficiency (an essential first step in any efficiency analysis) according to Erenburg et al. (2011)’s study which we discuss in greater detail later. 53. See Erenburg et al. (2011) and Barber et al. (1994). The only Cammer factor that is recognized in the economics profession is the “cause-and-effect” factor, which plaintiffs typically seek to demonstrate by providing “event study evidence.” However, as we discuss later, such evidence constitutes proof of “semi-strong form efficiency,” which is moot if the stock does not even trade in a “weak-form efficient” market. 54. Dunbar and Heller (2006), Abstract, p. 455. Notable exceptions in this regard being the case of Freddie Mac’s NYSE-traded Series Z preferred stocks in Freddie Mac, and more recently, George v. China Auto. Sys., Inc., 11 CIV. 7533 KBF, 2013 WL 3357170 (S.D.N.Y. July 3, 2013), in which the plaintiffs’ proof of market efficiency related to the securities at issue was considered inadequate by the court. 55. Dunbar and Heller (2006). 56. See China Auto. Sys., 2013 WL 3357170, at *1. 57. Langevoort (2009), p. 152. 58. Brief for Former SEC Commissioners and Officials and Law and Finance Professors as Amici Curiae Supporting Petitioners, Amgen v. Connecticut Ret. Plans and Trust Funds, 133 S. Ct. 1184 (2013) (No. 111085), 2012 WL 3555291. “Of the 92% of cases resolved prior to issuance of a ruling on summary judgment, 41% are dismissed, and 51% are settled.” Ibid.
198
MUKESH BAJAJ ET AL.
59. Grundfest (1995). 60. Amgen, 133 S. Ct., at 1193 & 1197 n. 6. 61. Ibid., at 1206 (Scalia, J., dissenting). 62. Ibid., at 1208 n. 4 (Thomas, J., dissenting). 63. Ibid. 64. Ibid., at 1197 n. 6 [quoting Langevoort (2009)]. 65. Amgen, 133 S. Ct., at 1204 (Alito, J., concurring). 66. Erenburg et al. (2011). 67. In their survey of the efficient market literature, economists at the Reserve Bank of Australia noted that “Within a decade, the efficient market hypothesis was so well established that Jensen (1978) was prompted to write that he believed there to be ‘no other proposition in economics which has more solid empirical evidence supporting it’. Such confidence portends a reversal, and the subsequent twenty years of research and asset-market experience have rendered the efficient market hypothesis a much more controversial proposition.” [Beechey, Gruen, and Vickery (2000), p. 21]. 68. Maier and Herath (2009) provide a recent survey of tests of market efficiency in real estate markets. They conclude based on their comprehensive survey of the empirical literature on the topic that “the result found in the literature is inconclusive. Majority of studies provide evidence supporting inefficiency of the real estate market while several studies maintain the notion of real estate market efficiency.” [Maier and Herath (2009), p. 1]. 69. In re IPO Securities Litigation, 260 F.R.D. 81, 98 n. 148 (S.D.N.Y. 2009). 70. Ferrillo, Dunbar, and Tabak (2004), p. 103, citing Fama (1970). 71. Fama (1970) discusses a momentum strategy known as the y% filter strategy, He notes, “But when one takes account of even the minimum trading costs that would be generated by small filters, their advantage over buy-and-hold disappears.” [Fama (1970), p. 396]. Ferrillo et al. (2004), p. 103. (“The main implication of the EMH [Efficient Market Hypothesis], when it holds, is that an investor cannot earn an above-average return by using stale or previously known information.”). 72. See, for example, PolyMedica I, 432 F.3d, at 10. 73. See Erenburg et al. (2011). 74. As explained in the text notes of Bodie et al., “This version of the hypothesis is quite extreme. Few would argue that the proposition that corporate officers have access to pertinent information long enough before public release to enable them to profit from trading on that information. Indeed, much of the activity of the Securities and Exchange Commission is directed toward preventing insiders from profiting by exploiting their privileged situation.” [Bodie, Kane, and Marcus (2008), p. 361]. 75. Ross et al. (2002, p. 349). 76. Ross et al. (2002, p. 349). 77. Hu and Marcus (2012) incorrectly claim that “it is a generally accepted and articulated principle in the finance literature that serial correlation in daily stock returns and deviations from the random walk model are not necessarily proof of market inefficiency,” citing certain theoretical studies by LeRoy (1973) and Lucas (1978) to buttress their claim. These theoretical studies, which consider abstract theoretical paradigms that bear little relevance to the real world, do not demonstrate that that periods of serial correlation are “not incompatible with efficiency of the
Assessing Market Efficiency for Reliance
199
market,” as Hu and Marcus assert. For instance, Lucas (1978) analyzes the “stochastic behavior of equilibrium asset prices in a one-good, pure exchange economy,” i.e., an economy in which there is only a single good produced, all consumers are identical (in terms of their preferences and wealth) and the good is “perishable,” i.e., there is no means of storing the good for future consumption. In other words, financial securities that allow one to save for future do not exist in this hypothetical world. Given such perishability, consumers are willing to pay relatively more for the good in a drought year compared to a normal year (i.e., declining marginal utility of consumption which manifests itself in risk aversion) which creates predictability in prices. Such a theoretical model and its conclusions bear little resemblance to the real world. Similarly, Leroy (1973) assumes, among other things that, “all earnings on stock are paid out in dividends” (i.e., companies cannot retain any earnings for future investment needs), each investor holds “an equal share of the equity and an equal allocation of cash,” and “can always make an unbiased forecast of the price of stock.” Leroy himself notes that some of his model’s assumptions are “particularly unrealistic,” and consequently, his conclusion that, in theory, the “return on stock will not satisfy the martingale property [i.e., display serial correlation] was proved only in a very restricted context.” He concludes that his article “demonstrated, then, not that any particular systematic departure from the martingale property is to be expected, but. … if capital markets are efficient, rates of return will follow a martingale distribution as a fair approximation [i.e., data will not empirically show serial correlation].” Contrary to Hu and Marcus’ assertion that “If one were to accept the premise that deviations from the random walk model make a security market inefficient, then one would have to conclude that almost all publicly traded stocks … are inefficient as well,” the Lo and MacKinlay (1988) study that Hu and Marcus themselves noted that there was no evidence of significant serial correlation on average across their sample of 625 stocks. While minor “spurious” negative serial correlation due to market frictions, such as lack of trading as Hu and Marcus note, in practice the odds of detected serial correlation being spurious is extremely low as Lo and MacKinlay (1990) confirm. These authors demonstrate that for even minimal spurious serial correlation (of −0.37%) to be detected in daily returns data, the stock’s lack of trading would have to be extreme, i.e., the stock could trade no more than once every “35.4 days.” In short, as, a study by Professor Timmermann and Nobel Laureate Clive Granger (which Hu and Marcus also cite) noted, if serial correlation in a stock’s returns is found to be “large enough to cover the size of transaction costs,” the finding “invalidate[s]” the conclusion that the stock trades in an efficient market. 78. Timmermann and Granger (2004), a study that Hu and Marcus (2012) cite. 79. Erenburg et al. (2011), Abstract, p. 260. 80. Erenburg et al. (2011), p. 289. 81. Erenburg et al. (2011), p. 290. 82. Erenburg et al. (2011), p. 291. 83. Erenburg et al. (2011), p. 291. 84. Erenburg et al. (2011), p. 291. 85. As noted above, Erenburg et al. (2011) consider serial correlation as well as momentum trading rules to detect weak-form inefficiency.
200
MUKESH BAJAJ ET AL.
86. Fama (1970) describes the y-filter rule as: “If the price of a security moves up at least y%, buy and hold the security until its price moves down at least y% from a subsequent high, at which time simultaneously sell and go short. The short position is maintained until the price rises at least y% above a subsequent low, at which time one covers the short position and buys. Moves less than y% in either direction are ignored.” [Fama (1970), pp. 394395]. A buy-and-hold strategy is a passive strategy in which the stock is bought and held until the end of the analysis period. The weak-form market efficiency model that Fama (1970) considers in this connection is referred to as a sub-martingale in prices, according to which in an efficient market investors expect next-period returns on a stock to be greater than or equal to zero [See Fama (1970), p. 386]. 87. Fama (1970) cautions that y-filter trading rule profits may arise because transaction costs such as commissions are ignored. Thus, to have probative value from an economic perspective, y-filter tests of weak-form efficiency should incorporate the transaction costs (commissions and bid-ask spreads) that implementing such a strategy would actually entail. 88. Fama (1991), p. 1576. 89. For instance, Jegadeesh and Titman (1993) documented that stocks with strong past performance continue to outperform stocks with poor past performance in the next period with an average excess return of about 1% per month over the 312 month horizon. In a more recent study, Fama and French (2012) examined momentum in four regions (North America, Europe, Japan, and Asia Pacific) and found “momentum everywhere” except Japan [See Fama and French (2012)]. 90. “An efficient market is one in which the market price of the stock fully reflects all publicly available information.” PolyMedica I, 432 F. 3d, at 10. 91. Fama (1991) recognized that event studies test for semi-strong form efficiency. He noted, “Instead of semi-strong-form tests of the adjustment of prices to public announcements, I use the now common title, event studies.” [Fama (1991), p. 1577]. 92. Tabak and Dunbar (2001), section 19.2. 93. The t-statistic is a ratio of the residual return and its standard error (or the “standard deviation” of the residual return over the estimation period, where the term “standard deviation” refers to the average squared difference of a variable’s observed values compared to the variable’s mean value). A t-statistic with an absolute value of 1.96 or greater means that the residual return is statistically significant at the 95% confidence level, i.e., there is only a 5% chance that the residual return can be attributed to chance. In such instances, the residual returns are typically considered “statistically significant.” 94. Teamsters v. Bombardier, 546 F.3d, at 207208. 95. The courts have recognized that this is a corollary of the efficient markets hypothesis. See, for example, In re Omnicom Group, Inc. Secs. Litig., 597 F.3d 501, 512 (2d Cir. 2010). See also supra n. 32. 96. For instance, Macey, Miller, Mitchell, and Netter (1991) argue, “We suggest that the focus of the Supreme Court’s holding in Basic is misplaced: what determines whether investors were justified in relying on the integrity of the market price is not the efficiency of the relevant market but rather whether a misstatement distorted the price of the affected security … [which can be determined using a] simple empirical technique, called an event study … Whenever event study methodology
Assessing Market Efficiency for Reliance
201
shows that a fraudulent event has had a statistically significant effect on the price of a firm’s securities, courts are justified in presuming reliance under the fraud-on-themarket theory.” [Macey et al. (1991), p. 1018.] 97. Analysis of the security’s price change following on such dates may well be relevant to the plaintiffs’ loss causation and damage claims, which the Supreme Court has noted need not be proven for class certification. See Halliburton, 131 S. Ct., at 21862187. 98. This follows from the fact that the adjustment made for contemporaneous changes in market and industry benchmarks in standard event studies is typically small over daily horizons. 99. By the same logic, an event study of selected dates during the alleged class period (which may or may not be mentioned in the complaint) when the price change was known to be large and significant, and the expert found some positive (or negative) news ex post that were consistent with the direction of the observed price change does not prove cause-and-effect. Instead, such a claim is an instance of the well-known “post hoc” fallacy. (The Latin term for the fallacy is post hoc, ergo propter hoc [“After this, therefore because of this”] and refers to the incorrect conclusion that if an event of kind A is followed in time by an event of kind B, then A must have caused B.) 100. In re PolyMedica Corp. Sec. Litig., 453 F. Supp. 2d 260, 270 (D. Mass. 2006) (“PolyMedica II”). 101. Ibid. 102. That is Ferrillo, Dunbar, and Tabak’s test compares the percentage of days with news that have a statistically significant price movement to the percentage of days without news that have a statistically significant price movement [Ferrillo et al. (2004, p. 120)]. 103. Freddie Mac, 281 F.R.D., at 180 (emphasis added). 104. Ferrillo et al. (2004), p. 120. 105. “The Fisher’s Exact and Chi-Square tests [proposed by Ferrillo, Dunbar, and Tabak] do not show that the price of Series Z consistently responded to unexpected, material news.” (Freddie Mac, 281 F.R.D., at 180). 106. This is one reason why the Fifth Circuit does not permit claims to be asserted on such a theory of confirmatory misrepresentation. See Greenberg v. Crossroads Sys., Inc., 364 F.3d 657, 665666 (5th Cir. 2004). 107. Comcast, 133 S. Ct., at 1433. 108. Freddie Mac, 281 F.R.D., at 180 (emphasis added). 109. Another one of this article’s authors, Mr. McLaughlin, was one of the counsel representing one of the defendants in Freddie Mac. 110. Ibid. 111. China Auto. Sys., Inc., 2013 WL 3357170, at *10. 112. See Lamont and Thaler (2003a, b), Ofek, Richardson, and Whitelaw (2004), and Battalio and Schultz (2006). As Lamont and Thaler (2003b) noted, “Do arbitrage trades actually enforce the law of one price? This empirical question is easier to answer than the more general question of whether prices reflect fundamental value. Tests of this more general implication of market efficiency force the investigator to take a stance on defining fundamental value. Fama (1991, p. 1575) describes this difficulty as the “joint-hypothesis” problem: “market efficiency per se is not testable.
202
MUKESH BAJAJ ET AL.
It must be tested jointly with some model of equilibrium, an asset-pricing model.” “In contrast, one does not need an asset-pricing model to know that identical assets should have identical prices.” [Lamont and Thaler (2003b), p. 228]. 113. See Battalio and Schultz (2006) for a discussion of Law of One Price tests using put-call parity including the use of such tests for dividend paying stocks. 114. Barber et al. (1994). 115. Barber et al. (1994), p. 290. 116. Erenburg et al. (2011), Abstract, p. 260. 117. Erenburg et al. (2011), Abstract, p. 260. 118. Erenburg et al. (2011), p. 264. 119. All quotes in this paragraph are from Lamont and Thaler (2003a), p. 195. 120. Lamont and Thaler (2003b), p. 233. 121. Following its IPO, Palm closed at $95.06 on its first day of trading while 3Com closed at $81.81 [See Lamont and Thaler (2003b), p. 230]. 122. Lamont and Thaler (2003a), p. 198. 123. See Fedenia and Hirschey (2009). 124. For the “negative stub” cases identified by Lamont and Thaler (2003b) that we have analyzed, the analysis periods are as defined by Lamont and Thaler (2003b). The analysis period for the Chipotle case is as defined in Fedenia and Hirschey (2009), which documented Chipotle’s violation of the Law of One Price. Lamont and Thaler (2003a) documented the Royal Dutch and Shell pricing anomaly from 1990 to August, 2002. For illustrative purposes, we have focused on the January 1, 1997June 30, 2002 sub-period, which spans more than five years, in our analysis of the Royal Dutch/Shell example. 125. Historical data over the relevant analysis period regarding the number of market makers for the securities analyzed are unavailable on widely recognized databases for such data, e.g., Bloomberg and the Center for Research on Security Prices (CRSP). Public float data were only available for Chipotle Class A and Class B shares over the relevant analysis period. However, the average bid-ask spread during the analysis period for most of the securities listed in Table 2 was lower than 22.15% which represent the average traded stock’s percentage quoted spread on the NYSE and Nasdaq, respectively, for a sample spanning the two weeks immediately after the beginning of decimal trading in the Nasdaq market: April 920, 2001 [“Comparing Bid-Ask Spreads on the New York Stock Exchange and Nasdaq Immediately Following Decimalization,” prepared by NYSE Research, July 26, 2001]. Such low bid-ask spreads are indicative of sufficient activity by market makers who provide liquidity in a security by posting bid and ask quotes at which they are willing to buy or sell the security immediately, respectively. Over the analysis period, the public float for Chipotle Class A and Class B shares was 98% and 93%, respectively, as a percentage of total shares outstanding [Source: Bloomberg L.P.]. 126. Source: CRSP U.S. stock database. 127. A company was required to have an outstanding float over $150 million held by nonaffiliates, or $100 million of such float coupled with annual trading volume exceeding 3 million shares at the time of Cammer Opinion, and the threshold for float has since been reduced to U.S. $75 million. As we noted earlier, public float data are available only for Chipotle Classes A and B shares over the analysis period. For both Chipotle class A and class B shares, average public float during
Assessing Market Efficiency for Reliance
203
analysis period was $1,268,113,512 and $1,381,714,099, respectively, more than 89 times the threshold of U.S. $150 million. 128. We determined the earnings announcement date and time through Factiva, a Dow Jones product, which is a widely used database containing news from worldwide sources. 129. We compute earnings surprise for the firm as the excess of actual earnings over the mean estimate of analysts. We obtained data on mean earnings estimate and actual earnings from Thomson Reuters I/B/E/S data base, which is a well-known database that “provides detailed and consensus estimates featuring up to 26 forecast measures including GAAP and pro-forma EPS, revenue/sales, net income, pre-tax profit and operating profit, and price targets and recommendations for more than 60,000 companies in 67 countries worldwide.” (http://thomsonreuters.com/products_ services/financial/financial_products/a-z/ibes/, downloaded August 1, 2011). We verified the earnings surprise with information contained in news articles through Factiva, if such information was available. In case the earnings surprise based on news articles was different from the earnings surprise based on IBES data, I used the earnings surprise based on news articles because IBES updates its database once a month. 130. The event study first uses a regression model to estimate the historical relationship between the subject stock’s daily return to that of a market index (in this case the S&P 500 index), or the index “beta,” over an estimation period. We defined the estimation period to be the 252 days preceding each earnings surprise date analyzed, except in three instances, when price data were unavailable for 252 days prior to the earnings surprise. In these three cases, the estimation period comprised of the maximum number of days for which such price data were available. Then, given the observed change in the market index on the event date at issue (when the earnings surprise was announced), and the estimated beta, the stock’s expected return is calculated. 131. Even though the price reaction was not in the right direction on three occasions, the market-adjusted returns on these occasions were insignificant, confirming that the earnings news considered in these cases was immaterial, after adjusting for contemporaneous changes in the market index. 132. BNP Paribas Press Release, “BNP Paribas Investment Partners temporaly [sic] suspends the calculation of the Net Asset Value of the following funds: Parvest Dynamic ABS, BNP Paribas ABS EURIBOR and BNP Paribas ABS EONIA,” August 9, 2007. 133. Brunnermeier (2009), p. 78. 134. “The 3-month London Interbank Offered Rate (LIBOR) is the interest rate at which banks borrow unsecured funds from other banks in the London wholesale money market for a period of 3 months. Alternatively, if a bank enters into an overnight indexed swap (OIS), it is entitled to receive a fixed rate of interest on a notional amount called the OIS rate. In exchange, the bank agrees to pay a (compound) interest payment on the notional amount to be determined by a reference floating rate (in the United States, this is the effective federal funds rate) to the counterparty at maturity. … Entering into the OIS exposes the bank to future fluctuations in the reference rate. However, the bank can guarantee itself longer-term funding while still paying close to the overnight rate. Because the alternative would be rolling over the funds on a daily basis at changing overnight rates, banks are willing to pay a premium. This is reflected in the LIBOR-OIS spread (defined as
204
MUKESH BAJAJ ET AL.
the difference between the LIBOR rate and the OIS rate).” [Sengupta and Tam (2008)]. 135. Former U.S. Fed Chairman Alan Greenspan quoted in Thornton (2009). 136. Bernanke (2008); Chairman Bernanke presented identical remarks to the Risk Transfer Mechanisms and Financial Stability Workshop, Basel, Switzerland, on May 29, 2008. 137. Taking Temporary Action to Respond to Market Developments Emergency Order, Exchange Act Release No. 58166, 73 Fed. Reg. 42379-01 (July 21, 2008). 138. Wutkowski and Younglai (2008). The SEC’s ban on naked short selling meant that in order to short sell a company’s stock investors would have to first borrow the stock. 139. BIS (2008), p. 4. 140. For example, on September 16, 2008, the government extended a two-year emergency loan of $85 billion to AIG, a major insurer of credit risk, and in exchange was entitled to 79.9% equity ownership of the company through preferred stock (AIG Annual Report for 2008, p. 1). 141. “WSJ Update: JPMorgan To Buy Most Ops of Washington Mutual,” Dow Jones News Service, September 25, 2008. 142. Robert (2008). 143. See Shleifer and Vishny (1997). 144. See “Hedge funds in trouble: The incredible shrinking funds,” The Economist, October 23, 2008 (citing statistics from Hedge Fund Research), p. 1 of 6. 145. See “Hedge funds in trouble: The incredible shrinking funds,” The Economist, October 23, 2008 (citing statistics from Hedge Fund Research), p. 2 of 6. 146. Pedersen (2009), p. 179. 147. Hu et al. (2011) focus on the U.S. Treasury market for three reasons: (i) the market’s global importance and use of Treasury yields as pricing benchmarks for other securities; (ii) its safety (or absence of credit risk) which implies that price deviations in the U.S. Treasury market are likely to provide information about liquidity shortages per se and not be contaminated by other risk factors that typically affect other security prices; and (iii) its significant liquidity relative to other asset markets, which implies that a shortage of liquidity in this market would constitute a strong signal about liquidity in the overall market [See Hu et al. (2011), pp. 12]. 148. Hu et al. (2011), p. 1. 149. Covered interest parity (CIP) arbitrage entails borrowing in one currency and lending in another to take advantage of interest rate differentials while avoiding exchange rate risk. “Arbitrage normally ensures that covered interest parity (CIP) holds. Until recently, excess profits, if any, were documented to last merely seconds and reach a few pips. Instead, … following the Lehman bankruptcy, these were large, persisted for months and involved strategies short in dollars.” [Griffoli and Ranaldo (2010)].
ACKNOWLEDGMENTS The authors thank Carl Vogel and Debo Sarkar of Navigant Economics, and Howard Privette and D. Scott Carlton at Paul Hastings for their
Assessing Market Efficiency for Reliance
205
comments on earlier drafts, Christina Hang for her valuable research assistance, and the editor Jim Langenfeld and an anonymous referee of this volume for their helpful comments.
REFERENCES Barber, B. M., Griffin, P. A., & Lev, B. (1994). The fraud-on-the-market theory and the indicators of common stock efficiency. Journal of Corporation Law, 19, 285312. Barberis, N., & Thaler, R. (2003). A survey of behavioral finance. In G. M. Constantinides, R. Stulz, & M. Harris (Eds.), Handbook of the economics of finance (Ed. 1, Vol. 1, Chapter 18, pp. 10511121). Amsterdam: Elsevier Science B.V. Battalio, R., & Schultz, P. (2006). Options and the bubble. The Journal of Finance, 61, 20712102. Beechey, M., Gruen, D., & Vickery, J. (2000). The efficient market hypothesis: A survey. Research discussion paper 200001. Economic Research Department, Reserve Bank of Australia, January 2000. Bernanke, B. S. (Chairman, Federal Reserve). (2008). Remarks on liquidity provision by the Federal Reserve at the Federal Reserve Bank of Atlanta Financial Markets Conference, Sea Island, Georgia, May 13. Retrieved from http://www.federalreserve.gov/newsevents/speech/bernanke20080513.htm Binyamin, A. (2013). Economists clash on theory, but will still share the Nobel. The New York Times, October 14. Black, F. (1986). Noise. The Journal of Finance, 41(3), 529543. Bodie, Z., Kane, A., & Marcus, A. J. (2008). Investments (7th ed.). Irwin: McGraw-Hill. Brealey, R. A., Myers, S. C., & Allen, F. (2011). Principles of corporate finance (10th ed.). Irwin: McGraw-Hill. Brunnermeier, M. K. (2009). Deciphering the liquidity and credit crunch 20072008. Journal of Economic Perspectives, 23(1), 77100. Cootner, P. H. (1962). Stock prices: Random vs. systematic changes. Industrial Management Review, 3, 2445. Duffie, D. (2010). Asset price dynamics with slow-moving capital. The Journal of Finance, 65, 12371267. Dunbar, F. C., & Heller, D. (2006). Fraud on the market meets behavioral finance. Delaware Journal of Corporate Law, 31(2), 455532. Erenburg, G., Smith, J. K., & Smith, R. L. (2011). The paradox of “fraud-on-the-market theory”: Who relies on the efficiency of market prices? Journal of Empirical Legal Studies, 8(2), 260303. Fama, E. F. (1965). The behavior of stock market prices. Journal of Business, 38(January), 34105. Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical work. The Journal of Finance, 25(2), 383417. Fama, E. F. (1991). Efficient capital markets: II. The Journal of Finance, 46, 15751617. Fama, E. F., & French, K. R. (2012). Size, value, and momentum in international stock returns. Journal of Financial Economics, 105, 457472. Fedenia, M., & Hirschey, M. (2009). The Chipotle paradox. Journal of Applied Finance, 19, 144159.
206
MUKESH BAJAJ ET AL.
Ferrillo, P. A., Dunbar, F. C., & Tabak, D. (2004). The less than efficient capital markets hypothesis: Requiring more proof from plaintiffs in fraud-on-the-market cases. St. Johns Law Review, 78(Winter), 81129. Garleanu, N., & Pedersen, L. H. (2011). Margin-based asset pricing and deviations from the law of one price. The Review of Financial Studies, 24(6), 19802022. Gilson, R. J., & Kraakman, R. (2013, forthcoming). Market efficiency after the financial crisis: It’s still a matter of information costs. University of Virginia Law Review. Griffoli, T. M., & Ranaldo, A. (2010). Limits to arbitrage during the crisis: Funding liquidity constraints and covered interest parity. Swiss National Bank Working Paper 201014, June. Grossman, S. J., & Stiglitz, J. (1980). On the impossibility of informationally efficient markets. American Economic Review, 70(3), 393408. Grundfest, J. A. (1995). Why Disimply? Harvard Law Review, 108, 727747. Grundfest, J. A. (2013). Damages and reliance under section 10(b) of the exchange act, rock center for corporate governance. Working Paper Series No. 150. Stanford University. Hedge Fund Research. (2008). Hedge funds in trouble: The incredible shrinking funds. The Economist, October 23 (citing statistics from Hedge Fund Research), pp. 1–6. Hu, G., & Marcus, M. (2012). Emerging issues in evaluating market efficiency. Law 360, July 18, 2012, Securities and Class Action Expert Analysis Sections. Hu, X., Pan, J., & Wang, J. (2010). Noise as information for illiquidity. NBER Working Paper No. w16468. Jegadeesh, N., & Titman, S. (1993). Returns to buying winners and selling losers: Implications for stock market efficiency. The Journal of Finance, 48(1), 6591. Lamont, O., & Thaler, R. (2003a). The law of one price in financial markets. Journal of Economic Perspectives, 17, 191202. Lamont, O., & Thaler, R. (2003b). Can the market add and subtract? Mispricing in tech stock carve-outs. Journal of Political Economy, 111, 227268. Langevoort, D. C. (2009). Basic at twenty: Rethinking fraud on the market. Wisconsin Law Review, 2009(2), 151198. LeRoy, S. F. (1973). Risk aversion and the martingale property of stock prices. International Economic Review, 14, 436446. Lo, A., & MacKinlay, A. C. (1988). Stock market prices do not follow random walks: Evidence from a simple specification test. Review of Financial Studies, 1(1), 4166. Lo, A. W. (2008). Efficient markets hypothesis. In S. N. Durlauf & L. E. Blume, (Eds.), New Palgrave Dictionary of Economics (2nd ed.). New York, NY: Palgrave McMillan. Lo, A. W., & MacKinlay, A. C. (1990). An econometric analysis of nonsynchronous-trading. Journal of Econometrics, 45, 181–212. Lucas, R. E., Jr. (1978). Asset prices in an exchange economy. Econometrica, 46(6), 14291445. Macey, J. R., Miller, G. P., Mitchell, M. L., & Netter, J. M. (1991). Lessons from financial economics: Materiality, reliance, and extending the reach of Basic v. Levinson. Virginia Law Review, 77, 10171050. Maier, G., & Herath, S. (2009). Real estate market efficiency: A survey of literature. Research Institute for Spatial and Real Estate Economics, WU Wien, Vienna University of Economics and Business Working Paper SRE-Discussion 2009/07. Mitchell, M., Pulvino, T., & Stafford, E. (2002). Limited arbitrage in equity markets. The Journal of Finance, 57, 551584.
Assessing Market Efficiency for Reliance
207
Ofek, E., Richardson, M., & Whitelaw, R. F. (2004). Limited arbitrage and short sales restrictions: Evidence from the options markets. Journal of Financial Economics, 74, 305342. Pedersen, L. H. (2009). When everyone runs for the exit. The International Journal of Central Banking, 5, 177199. Robert, S. (2008). Goldman, Morgan to become holding companies. Companies get access to Fed lending in exchange for oversight. Market Watch, September 21. Ross, S. A., Westerfield, R. W., & Jaffe, J. (2002). Corporate finance (6th ed.). p. 349. Sengupta, R., & Tam, Y. M. (2008). The LIBOR-OIS spread as a summary indicator. Economic Synopses, Number 25. Federal Reserve Bank of St. Louis. Shleifer, A., & Vishny, R. W. (1997). The limits of arbitrage. The Journal of Finance, 52(1), 3555. Tabak, D. I., & Dunbar, F. C. (2001). Materiality and magnitude: Event studies in the courtroom. In R. L. Weil, M. J. Wagner, & P. B. Frank (Eds.), Litigation services handbook: The role of the financial expert (3rd ed., Chapter 19, pp. 19.119.22). New York, NY: Wiley. Thornton, D. L. (2009). What the LIBOR-OIS spread says. Economic Synopses, Number 24. Federal Reserve Bank of St. Louis. Timmermann, A., & Granger, C. W. J. (2004). Efficient market hypothesis and forecasting. International Journal of Forecasting, 20, 1527. Wutkowski, K. & Younglai, R. (2008). UPDATE 2-US SEC looks to expand short rule to entire market. Reuters News, July 24.
EUROPEAN COLLECTIVE REDRESS: LESSONS LEARNED FROM THE U.S. EXPERIENCE$ Ethan E. Litwin and Morgan J. Feder ABSTRACT In recent years, the European Commission and various Member States, citing increasingly integrated markets and higher levels of cross-border activity within the European Union (“E.U.”), have called for the adoption of effective collective redress mechanisms for victims of violations of E.U. law. Although many Member States have already adopted collective action procedures under national law, these procedures have been ineffective in stimulating private enforcement of E.U. law and are often divergent in their approach to consolidating claims. E.U. lawmakers, after a lengthy period of investigation and study, have identified a set of guiding principles for the Member States to use in enacting new collective redress procedures within their national systems. The studies and papers solicited from the public during the Commission’s deliberations
$
This article was originally presented at the Spring Meeting of the ABA Section of International Law, which was held in New York in April 2012. We have updated the original version to account for recent events.
The Law and Economics of Class Actions Research in Law and Economics, Volume 26, 209248 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0193-5895/doi:10.1108/S0193-589520140000026007
209
210
ETHAN E. LITWIN AND MORGAN J. FEDER
are explicit in their rejection of the U.S.-style opt-out class action mechanism. In their effort to avoid similarly calamitous results, European lawmakers propose that Member States adopt “opt-in” class actions, while rejecting many of the economic incentives that some believe lead to filing nonmeritorious claims, such as punitive damages and contingency fee arrangements. The European proposal is unlikely in the authors’ view to stimulate private enforcement of European law or increase victims’ access to compensation, given the flaws inherent in the opt-in class action device. Instead of looking to adopt a “U.S.-lite” approach to victim redress which is fundamentally incompatible with many judicial systems within the E.U., the authors propose that Europeans consider adopting a regulatory administered compensation system, modeled after such U.S. examples as the Securities and Exchange Commission Fair Funds and the September 11th Victim Compensation Fund. The authors also propose that regulatory administered funds can provide more effective and efficient restitution to victims than traditional litigation. Keywords: Class actions; opt-out; opt-in; collective redress alternatives; Fair Funds JEL classifications: K41; K21
TABLE OF CONTENTS Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Consultation on Collective Redress in the E.U. . . . . . . . . . . . . . . . . . . Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Priorities and Proposals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Opt-In . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Judicial Supervision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Representative Organizations. . . . . . . . . . . . . . . . . . . . . . . . . . . No Punitive Damages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Loser Pays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Class Actions in the United States . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Development of the Class Action . . . . . . . . . . . . . . . . . . . . . . . Criticisms of Class Actions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Opt-In Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
211 213 213 216 217 217 217 218 218 218 219 220 222 222
European Collective Redress
211
The Opt-Out Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Implications of the U.S. Experience . . . . . . . . . . . . . . . . . . . . . . Nonlitigation Mechanisms for Collective Redress in the United States Agency Distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . SEC Fair Funds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Federal Trade Commission . . . . . . . . . . . . . . . . . . . . . . . . . . . . Other Public and Private Funds . . . . . . . . . . . . . . . . . . . . . . . . . . . . September 11th Victim Compensation Fund . . . . . . . . . . . . . . . Gulf Coast Claims Facility. . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion: A Modest Proposal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
222 224 225 225 225 227 228 228 231 234 247
INTRODUCTION As the European Union (“E.U.”) has grown in size, European markets have become more integrated, and cross-border activity and trade have increased.1 In light of these changes, E.U. lawmakers have recognized a growing need to develop effective E.U.-level mechanisms for compensatory collective redress for victims of unlawful cross-border behavior that would complement the existing system of public enforcement.2 In the views of many European legislators, regulators, and commentators, the need for a collective approach to victim compensation is particularly apparent in the areas of competition, consumer and environmental law, and in the financial services, telecommunications, transport, and package travel/tourism sectors,3 where consumers in multiple countries are likely to be affected by the same legal infringement but individual damages may be small. The same legal justification is often invoked to justify the U.S. class action system: plaintiffs with only small damage claims often find it economically unfeasible to litigate on their own, and their injuries tend to go uncompensated in the absence of a procedure permitting them to share the costs and risks of litigation with others.4 Collective redress is not a completely alien concept to European legal systems; to the contrary, many Member States have debated and enacted national mass action legislation of one form or another.5 The structure and effectiveness of these collective redress mechanisms, however, vary significantly from country to country. For this reason, the European Commission (“E.C.”) expressed concern that the “lack of a consistent approach to collective redress at E.U. level may undermine the enjoyment of rights by citizens and businesses and [give] rise to uneven enforcement of those rights.”6
212
ETHAN E. LITWIN AND MORGAN J. FEDER
The E.C. therefore initially undertook to research and develop European standards of compensatory collective redress, considering models and data from within and outside the E.U. and inviting comments on proposals.7 In the numerous papers, reports, and other documents that have resulted from this process, European lawmakers have raised and rejected as a model the American-style “opt-out” class action, which has facilitated in their view the undesirable and wasteful proliferation of “abusive and unmeritorious” litigation in the United States.8 Recently, the E.C., finding the task of imposing a set of E.U.-mandated procedural rules on the national court systems of the various Member States perhaps too challenging in the near term, proposed that the Member States develop collective redress procedures themselves under their own national laws. To assist the Member States in this task, the E.C. set forth a set of guiding principles. The majority of the comments received in response to the public consultation on the Collective Redress Green Paper and the E.P. Report, whether from the business community, consumer advocates, or the various European bar associations, largely echoed these principles. For a number of reasons discussed below, collective actions embracing these principles are unlikely to lead to increased compensation for victims. In the United States, collective redress has not solely been achieved through class action lawsuits. Particularly in recent years, the United States has explored methods outside of the traditional litigation context for compensating plaintiffs who have been damaged on a mass scale. Although the degree of success enjoyed by these alternative methods of resolving mass claims is debatable, these approaches may serve as useful models for the E.C. and the Member States to draw from in designing their own collective action frameworks, and might even be considered more often in the United States to address cases currently addressed by class actions. The first example is compensation through agency funds, the most well-known example of which is the U.S. Securities and Exchange Commission’s (“SEC”) “Fair Funds” mechanism. Through Fair Funds, the SEC distributes to injured investors the disgorged profits of wrongdoers’ activities as well as penalties collected from companies found to have violated U.S. securities laws.9 In addition to agency funds, other public or private administered funds, such as, among others, the September 11th Victim Compensation Fund and the Gulf Coast Claims Facility created to compensate those injured in the BP oil spill, are also valuable examples from which the E.C. may draw when designing its own procedures for collective redress. While these two funds originated out of unique circumstances that are unlikely to be repeated, the examples of the methods employed by the administrators in designing
213
European Collective Redress
claims processes and in determining awards may help the E.C. and the Member States to devise procedures that are fair, efficient, and consistently applied. Part II of this article will more specifically describe the steps that the E.C. has already taken, the findings of its investigation, and the key considerations it identified for any European collective redress mechanism. Part III will describe U.S. experience with litigating class actions. Part IV will describe some nonlitigation approaches used in the United States, including Fair Funds and other private and public administered funds. We conclude by proposing that mass litigation, in whatever form, is unlikely to achieve the E.C.’s aspirations for effective and efficient collective redress. We propose, instead, that the E.C. follow the recent U.S. example of providing for administrative redress funds and utilizing other alternative means of compensating victims on a mass scale.
CONSULTATION ON COLLECTIVE REDRESS IN THE E.U. Background The E.C. began its exploration of possible E.U.-level methods of compensatory collective redress in the fields of consumer and competition law.10 In 2005, it published a Green Paper on Damages Actions for Breach of the E.C. Antitrust Rules (the “Antitrust Green Paper”), seeking to identify deficiencies in existing systems and calling for comments on proposed solutions to increase efficiency, to improve access to justice and compensation for injured parties, and to strengthen enforcement and deterrence.11 The E.C. followed this in April 2008 with a White Paper on Damages Actions for Breach of the E.C. Antitrust Rules (the “Antitrust White Paper”). In both of these papers, the E.C. acknowledged the inadequacy of existing mechanisms for victim compensation, noting that victims of antitrust violations rarely obtained reparation because they faced difficult legal and procedural obstacles in the Member States’ rules.12 Without a collective redress procedure, the E.C. came to the same conclusion that lawmakers in the United States came to over 75 years ago, and that consumers or purchasers with only small damage claims are almost always deterred from bringing actions on their own, because as a practical matter, it is all but impossible to make the pursuit of such claims financially worthwhile.13 The E.C. estimated that
214
ETHAN E. LITWIN AND MORGAN J. FEDER
victims of antitrust violations forgo compensation in the range of several billion euros per year.14 As it examined the need for collective redress in the antitrust field, the E.C. came to recognize that competition law was not the only area in which such a mechanism was necessary.15 It observed that in an increasingly consumer-oriented, global, digital economy where markets are rapidly expanding in size and crossing borders, ever-larger numbers of consumers in a widening geographic area could be affected by seemingly small infringements of laws in many areas.16 The effects of these infringements could be so widespread that they could distort markets, lower consumer confidence, and harm the economy.17 The E.C. cited several examples of how such small infringements could create widespread harm, including allegations that U.K. banks systematically imposed excessive charges on hundreds of thousands of customers whose accounts became overdrawn, as well as alleged abuses in the telecommunications sector, where supposedly “free” ringtones came with hidden charges or subscriptions either disguised in small print or simply not disclosed at all.18 The E.C. concluded that, like antitrust victims with only small claims, European consumers in general with small claims were unlikely to pursue those claims individually.19 The E.C. therefore broadened the scope of its exploration of collective redress procedures, looking at ways in which consumer rights beyond the antitrust sector could be recompensed collectively. As part of this initiative, the E.C. examined existing mass action procedures at the Member State level to determine how an E.U. mass action directive could improve access to justice.20 The E.C. found that many Member States had already introduced national collective redress mechanisms of one form or another, and many still were considering doing so in the future.21 Each of these mechanisms was, however, unique; the procedures and rules governing mass actions in the Member States could therefore result in widely divergent results in any given case.22 In its Evaluation of Collective Redress, the E.C. grouped the approaches taken in the various Member States into four broad categories: • Group Actions: These are the most common collective redress mechanisms available in many of the Member States such as Belgium. Individual countries have their own variations, but generally this category includes (a) actions in which individual claims are combined into one proceeding, (b) actions that are brought by groups of consumers, and (c) actions that are brought by one claimant (either an individual consumer, a consumer organization or a consumer ombudsman), who
European Collective Redress
215
seeks a decision on behalf of a group with equal or similar problems, giving group members the right to enforce their rights in accordance with that decision.23 • Representative Actions: In these actions, such as those in the United Kingdom, the representative obtains a decision that the representative can enforce. Typically, European laws authorize certain organizations to bring claims under these acts. Representative Actions are different from Group Actions, in which all members have the right to enforce their rights separately.24 • Test Case Procedures: A few countries have introduced procedures that give effect to a judgment in a test case beyond the parties to the test case itself. Test cases in general can be persuasive, though not binding, on lower courts, and can aid in negotiations. In Greece, for example, consumer associations may seek a declaratory judgment on the defendant’s liability for damages that then has a res judicata effect in favor of individual claimants who may initiate follow-on procedures to claim payment for their individual losses. An unfavorable decision would have no res judicata effect against consumers, however, meaning that even if the consumer association lost, claimants could still sue individually.25 • Procedures for Skimming off Profits: This approach, introduced in Germany, seeks not to compensate individual consumers, but to force wrongdoers to disgorge the profits from unfair competition. The action can be initiated by consumer associations, but the disgorged profits go to the treasury, not to the victims.26 These national mechanisms vary also in terms of the business sectors or fields of law in which they are available,27 the persons or entities endowed with standing to bring group claims,28 whether businesses or corporate entities may sue (or only individual consumers), the effect of a judgment,29 the point at which claimants must be identified (at the time the action is brought, or later), funding of collective actions, distribution of awards, and the use of alternative dispute resolution procedures.30 Although collective redress mechanisms are available in many Member States, few collective actions have actually been brought.31 Several factors may contribute to this perceived underuse. According to the E.C., many of the national mechanisms seem unable to fulfill the purposes for which they were enacted, financing remains a significant obstacle because of the “loser pays” principle common in the E.U., and litigation costs remain high for representatives.32 The E.C. further reported, however, that the introduction of collective procedures in these Member States has not unreasonably
216
ETHAN E. LITWIN AND MORGAN J. FEDER
raised costs for businesses or caused reputable businesses to fail, while at the same time it has increased consumers’ access to justice.33 Ultimately, the E.C. concluded that while almost all existing mechanisms enable some efficiencies or are more effective in certain ways than individual actions, improvement is needed to harmonize procedures across the E.U. and to ensure consistent treatment of both plaintiffs and defendants in collective redress actions, especially in cases with a cross-border dimension.34 In June 2013, the E.C. determined that justice would best be served through the development of collective redress procedures at the Member State level.35 Based on the various consultations and studies conducted over the past several years, the E.C. Recommendation sets forth a collection of nonbinding principles intended to be used by the Member States in developing collective redress mechanisms at a national level that are intended to facilitate access to justice for victims of violations under E.U. law. The common principles applied through this horizontal approach are likewise intended to ensure that the fundamental rights of the parties are preserved and enforced and that abuses are prevented, while providing sufficient flexibility to accommodate the disparate legal traditions of the various Member States. Key among the principles enumerated by the E.C. are, as described in more detail below, (1) the “loser pays” principle, (2) the “opt-in” principle, (3) no contingency fees, (4) no punitive damages, and (5) primacy of proceedings brought by public authorities.36 The Commission proposes to allow the Member States two years to enact procedures consistent with the principles enumerated in the E.C. Recommendation. Two years after that, the E.C. plans to conduct a community-wide survey to determine whether further enhancements or changes are warranted.37 As discussed in the next section, there are a number of reasons that the combination of these principles will not lead to increased compensation for victims. At the same time it issued the E.C. Recommendation, the Commission proposed a directive to facilitate damages claims by victims of antitrust violations, portraying this legislation as a kind of competition-specific complement to the E.C. Recommendation.38 The Proposed Directive would apply to actions for damages available in Member States, but would not oblige Member States to introduce collective actions if they are not available.39
Priorities and Proposals The E.C. Recommendation reflects the European Parliament’s (“E.P.”) earlier resolution (the “E.P. Resolution” and, collectively with the E.C.
European Collective Redress
217
Recommendation, the “European Proposals”) that identified a broad set of considerations that reflect the consensus of European thinking on mass actions.40 The E.P. emphasized that any mechanism implemented must follow European principles and traditions, and avoid incorporating “alien” procedural elements that could create an abusive, litigious atmosphere.41 In not so many words, the E.P. summarily rejected some 75 years of U.S. experience with collective redress actions, implying that Europeans could do better than their American counterparts. The following is a summary of the most significant aspects of the European Proposals. Opt-In42 In any collective action, claimants must present common issues. While the current American system allows representative plaintiffs to litigate commonality at the outset of the lawsuit without first identifying all potential class members, Europeans reject this approach and uniformly call for procedures that would identify class members prior to commencement of any mass action.43 Any decision would therefore bind only those claimants who expressly indicated a wish to join the action at the outset. The European Proposals recommend that, in addition to being able to opt in to a class, individuals must retain the right to seek individual compensatory redress before a court.44 The E.P. and the E.C. intend that the opt-in approach will reduce abusive litigation, and at the same time will comport with the Constitutions of several Member States that prohibit actions on behalf of unknown victims.45 Judicial Supervision The European Proposals favor a system of judicial review of mass actions, in which a judge or similar body conducts an initial admissibility check of proposed collective actions to confirm that all criteria have been met and that the action may properly proceed.46 As discussed below, the E.C. and the E.P. either have not considered or have rejected without discussion vesting redress authority for mass injuries in regulatory bodies instead of courts. Representative Organizations In addition to opt-in class actions, the European Proposals would further vest certain organizations such as consumer groups and ombudsmen with the authority to bring representative actions on behalf of victims.47 To reduce the financial incentive for filing unmeritorious claims, representative organizations are prohibited under the European Proposals from sharing in any compensation paid to victims.48 Compensation for prosecuting the
218
ETHAN E. LITWIN AND MORGAN J. FEDER
litigation, therefore, would have to be derived from sources unrelated to any eventual damage award. No Punitive Damages Victims would be entitled to seek compensation only for actual damages suffered; the European Proposals strongly reject any incentivizing economic factors such as punitive damages or multiple damages (such as the treble damages available in U.S. antitrust litigation). Damages, under the system envisioned by the European Proposals, would be distributed to victims in proportion to the harm they sustained individually.49 Evidence The European Proposals reject the broad discovery rights that are a defining characteristic of U.S.-style litigation. Rather than requiring the defendant to provide evidence to support the claimants’ case, the European Proposals require each claimant to provide sufficient evidence (outside of the discovery process) to support his claim. To European eyes, U.S.-style discovery unduly raises the cost of litigation and encourages the prosecution of unmeritorious claims.50 Despite the strong European tradition of limited discovery, the U.S. experience shows that litigating mass actions uniquely requires substantial discovery of defendants, and Europeans may soon find themselves amending their discovery rules to permit greater access to defendant files in the mass action context. Moreover, as the U.S. experience suggests, concessions made in the context of mass actions tend to influence the prosecution of individual actions as well. One wonders, therefore, whether European distaste for expansive discovery can survive the introduction of collective redress. Loser Pays The loser pays rule, which is common in European jurisdictions already, requires the unsuccessful party to bear the litigation-related costs of the prevailing party. In principle, the loser pays rule discourages the prosecution of unmeritorious claims.51 For this reason, the European Proposals also reject contingency fees and third-party funding solutions, which Europeans believe contribute to excessive litigation in the United States and are contrary to European traditions.52 The loser pays rule, while certainly providing some benefit, is not a cure-all for the flaws of collective redress; for one, it tends to discourage settlement, as continued prosecution of winning claims will be compensated by the losing party at the end. Indeed, because litigation costs must factor into the analysis of any
European Collective Redress
219
eventual recovery, settlement costs are typically higher in jurisdictions where the loser must pay the winner’s costs. There are other flaws to the loser pays rule as well. To cover potential reimbursement claims, claimants in loser pays jurisdictions frequently are forced to obtain litigation insurance to protect themselves in the event of an adverse ruling. In effect, the loser pays rule therefore substitutes insurance actuaries for judges as the primary gatekeepers to justice in these jurisdictions. It is a risk, therefore, that conditioning collective redress on the adoption of loser pays rules will result in limiting mass actions to only the most meritorious cases, while continuing to deny claimants with more contestable, yet not frivolous, actions their day in court. The loser pays rule is also not a panacea for all the ills of collective actions. As discussed below, in an opt-in system as supported by the E.P., victims with small claims are unlikely to risk joining an early opt-in class when the outcome is uncertain; they will have an incentive to wait until early classes win and the risk of having to pay is reduced before joining in. The loser pays rule only exacerbates that problem. Accordingly, opt-in classes are likely to struggle to assemble a large enough class to begin with to make a collective action possible, and will not serve to improve access to justice for those with small claims.53 In the many responses to the E.C. Public Consultation and in the growing body of commentary in legal journals and law reviews, proposed solutions cover a range of possibilities.54 Based on the objectives and priorities outlined in the European Proposals, however, implementing any private litigation-based system of collective redress is unlikely to result in an efficient, effective system of victim compensation without the attendant disadvantages of the U.S. class action system that the E.U. wishes to avoid. Ultimately, mass litigation is not the answer to Europe’s identified need for collective redress. The E.U. would benefit from looking beyond mass litigation as modified to fit the Commission’s and the Parliament’s criteria, and consider other alternatives. We suggest that a regulatory administered compensation system can provide benefits equivalent to those of a judicially administered compensation system with a higher degree of efficiency.
CLASS ACTIONS IN THE UNITED STATES In the United States, the class action “is a procedural device that was adopted with the goals of economies of time, effort and expense, uniformity
220
ETHAN E. LITWIN AND MORGAN J. FEDER
of decisions, the promotion of efficiency, and fairness in handling large numbers of similar claims.”55 Although it was created with these aspirations in mind, the class action has always been viewed with a certain ambivalence by judges and legislators, because it creates numerous problems for plaintiffs and defendants and for the judicial system generally.56 As one federal court put it, “class actions are unique creatures with enormous potential for good and evil.”57 As discussed above, it is not surprising that the E.U. has expressed concerns about adopting the U.S. system. As explained below, there are alternative approaches to consumer redress that both the E.U. and the United States should consider.
The Development of the Class Action It has been a long-standing, fundamental principle in U.S. law that no individual is personally bound by a judgment in a litigation to which he is not a party.58 In keeping with this principle, U.S. law originally made no provision for class or representative suits of any kind, enforcing instead the “necessary parties rule,” in which all persons, no matter how numerous, with a material interest in the subject of the litigation were required to be made parties.59 Exceptions to this rule eventually emerged to avoid unfairness to litigants that occasionally resulted when mandatory joinder made certain suits simply impracticable, where the number of materially interested parties was so large that it was impossible for all of them to appear in court.60 In exceptional situations, courts sometimes permitted actions to proceed without all “necessary parties,” although these judgments still did not serve to bind absent parties.61 In 1937, the Federal Rules of Civil Procedure set forth the precursors to the modern class action, which included, among other things, the same sort of opt-in procedure that Europeans favor adopting now.62 Although it required parties affirmatively to join the class to be bound by a decision, the opt-in class action device also permitted absent class members to intervene in the case even after a judgment was handed down (known as “oneway intervention”), and thus take advantage of favorable decisions while avoiding being bound before the result was known.63 The original version of Rule 23 of the Federal Rules of Civil Procedure, as approved in 1937, recognized three different types of class actions: the “spurious” class action, where members of the class each brought individual claims based on common questions of law or fact; the “true” class action, which concluded the rights of all class members, whether named or
European Collective Redress
221
not; and the “hybrid” class action, in which class members made separate claims against a common fund or property. U.S. courts struggled to apply Rule 23 as originally drafted.64 In particular, the original Rule 23 led to inconsistent judgments, proved unable to adequately protect the rights of absent class members, did not materially increase judicial efficiency in disposing of similar claims, and failed to facilitate the spreading of litigation costs among multiple claimants. In short, Rule 23 as originally drafted and in a form similar to what Europe contemplates today was a complete failure. Underused and ineffective, Congress acted in 1966 to revise and revitalize the class action rule. But where common sense may have suggested abandoning the class action device altogether given its inherent flaws, there was no longer any going back for the United States. Empowered by class litigation, consumer advocates staunchly opposed any retrenchment of consumer litigation rights. Congress ultimately approved the modern “opt-out”-style class action rule, which sought to address some of the problems inherent in the earlier opt-in version through the “adventuresome innovation” of permitting a decision to bind absent class members unless they requested to be excluded.65 The opt-out rule was intended to clarify when absent class members would be bound, and to bind them whether or not the judgment was favorable to the class.66 By doing so, Congress intended to eliminate what was perceived to be the central flaw of the opt-in model, namely the “one-way intervention” option, which incentivized potential claimants to delay litigation and, in effect, ride the coattails of other successful litigants, with the result that no one wanted to be the first to litigate.67 Congress, and subsequently the courts, endorsed the opt-out rule as an exception to the long-standing principle of U.S. law that one is not bound to a judgment without being a party to the lawsuit, because, through the class action mechanism, “absent class members’ interests are deemed to be adequately represented by another similarly situated entity and unitary adjudication of similar claims or issues is efficient and fair to the parties.”68 The procedural requirements of Rule 23 (i.e., numerosity, commonality, typicality, and adequacy of representation) are intended to “assure, to the greatest extent possible, that the actions are prosecuted on behalf of the actual class members in a way that makes it fair to bind their interests.”69 Despite its numerous and serious flaws, the modern opt-out class action mechanism has produced some beneficial efficiencies in the administration of U.S. justice. First, class actions reduce the need to try numerous, duplicative suits on a single subject matter where many litigants need to establish the same set of facts.70 This enables the efficient use of time, money, and
222
ETHAN E. LITWIN AND MORGAN J. FEDER
other resources for all parties, and lowers the strain on the system, which could be overwhelmed by hundreds or thousands of individual repetitive lawsuits.71 Second, by spreading litigation costs over a larger group of claimants, class actions enhance access to justice for those who would be unable to make an individual suit financially worthwhile.72 Third, class actions help to reduce inconsistent results, which is fairer to both sides.73
Criticisms of Class Actions The Opt-In Model Opt-in class actions present certain obstacles at the outset of the case during the joinder process. Where harm is small or obscure, many potential class members may not even realize that they have been injured. Even where class members are aware of their injury, they may not be aware of, or may choose to proceed independently of, existing actions, which leads to the more serious problem of competing classes. Where several such actions proceed simultaneously, they compete for members, waste time and money on duplicative work, and risk inconsistent results. As noted above, this leaves defendants vulnerable to repeated attack by successive actions until one succeeds or the statute of limitations runs out. Permitting competing lawsuits also encourages litigants to “game” the system by seeking out the most favorable jurisdictions and by waiting until earlier lawsuits have already borne much of the effort and expense of discovery. Opt-in models, as the U.S. experience counsels, invariably lead to inefficiency and delays, overburdened courts, inconsistent results, and ultimately, less access to compensation for victims. By 1966, Congress and many U.S. legal commentators had concluded that the flaws inherent in the opt-in model could not be rectified. It is therefore surprising that Europe appears to have embraced the very procedural device for giving effect to collective redress that proved to be unworkable half a century ago in the United States. The Opt-Out Model The opt-out model promised to address many of the problems experienced under the opt-in model. First, the opt-out model resolved the difficulty of establishing numerosity by binding, by default, the majority of potential claimants; accordingly, opt-out classes do not typically struggle at the outset to identify enough claimants to satisfy Rule 23’s numerosity requirement. The opt-out model also promotes finality for defendants: once a class action is concluded, defendants do not remain vulnerable to a long and
European Collective Redress
223
unpredictable stream of lawsuits. Defendants face potential litigation only from an identifiable subset of class members who affirmatively opted out. These “benefits” of opt-out class actions, however, ultimately proved illusory as the new class action model introduced new and more serious problems into the U.S. legal system. While the opt-out model permitted actions to proceed more readily at the outset by removing the requirement that claimants identify themselves at the outset, the effect was merely to delay this hurdle, not to eliminate it. Eventually, class members must make themselves known to prove their individual entitlement to recovery and to collect their share of any award. The notice requirement creates an enormous burden at a later stage of the litigation, when all prospective class members must be informed of their rights to opt out or claim an award. This requirement represents perhaps an even greater burden than that of collecting claimants at the outset of an opt-in class: there, the class must simply become large enough to warrant collective action, and when enough claimants are signed on, representatives need not notify anyone else. In an opt-out model, by comparison, because a decision will bind absent class members, the representatives of the class must notify as many of these class members as possible, and in some cases, many or most of them may prove to be unknown and difficult, if not impossible, to identify.74 Even where potential members are readily identifiable, the larger the class, the more expensive and resource-intensive it is to mail notice or post information in bulletins, newspapers, and other locations where class members are likely to see it. In a large widespread class, even the best methods of notice inevitably leave potential class members out. These class members are therefore denied any opportunity to defend their rights; if they miss the opt-out deadline, they will be precluded from litigating independently later even if they never received notice of the class. Even if effective notice is given, it is no assurance of a high response rate, especially if the potential recovery on any individual claim is low. Indeed, the price of the stamp to mail in a claim form has, in some prior class action cases, cost more than the entire award for each claimant.75 Even larger amounts could seem like too much trouble to pursue if significant time or effort is required to claim the award (e.g., if the evidence required to prove injury is scattered or otherwise difficult to assemble).76 The most significant problem presented by the current opt-out mechanism, however, is the potential conflict of interest that exists between class counsel and the prospective class. Without identified class members at the start of the lawsuit, some class counsel may proceed on their own initiative, motivated by the prospect of large fees.77 Lacking a true client whose
224
ETHAN E. LITWIN AND MORGAN J. FEDER
interests and decisions guide the progress of the lawsuit, the lawyers alone make the critical choices related to the goals and priorities of the litigation, the strategy to be followed, and most importantly, settlement negotiations. Where lawyers’ fees are tethered to the size of a recovery, their incentive to conduct a lengthy litigation is greatly reduced in the absence of a motivated client; it is in their best interest to settle early, collect their fees, and move on to the next project. The opt-out model also diminishes the likelihood of a just result for defendants. With the leverage presented by the potentially hundreds or thousands of claimants, class counsel come to the bargaining table with enormous power to coerce defendants to settle rather than risk litigation and judgment for an unpredictably enormous sum. The imbalance in bargaining power created by the opt-out model reduces defendants’ incentive to challenge unmeritorious claims: facing a potentially catastrophic loss at trial, defendants may find it safer to cut their losses and settle to get rid of lawsuits, even if they believe the plaintiffs’ claims are weak or even frivolous. Knowing this, lawyers have even more motivation to abuse the system, and class members, defendants, and the system at large pay the price.78 Implications of the U.S. Experience All class action models essentially force the majority of class members to sacrifice their individual due process rights for the chance to aggregate claims. While the opt-out mechanism has doubtless enabled many meritorious lawsuits and effectively compensated many deserving victims, a system in which litigants must give up the chance meaningfully to participate in the adjudication of their rights (or even to decide whether a case should be brought at all) promotes just the kind of wasteful, unmeritorious overlitigation79 that E.U. lawmakers expressly wish to avoid. The right to “opt-in” or “opt-out” provides scant protection for individual due process rights. Certainly, the provision of timely notice affords a potential claimant with a choice, whether to proceed individually or as part of a collective action. But for claimants whose potential claims are dwarfed by the time, effort, and expense of individual litigation, that choice is illusory. Real due process is the right to be heard, and the right to tilt at windmills is no choice at all. Although Rule 23’s certification requirements (commonality, typicality, and adequacy of representation), which could be applied to opt-in or opt-out class actions, are designed to ensure that the representative plaintiffs will protect the due process rights of all class members, the reality is somewhat divorced from this aspiration. In a collective action of any kind, seemingly by definition, the individual voice is completely lost.
225
European Collective Redress
NONLITIGATION MECHANISMS FOR COLLECTIVE REDRESS IN THE UNITED STATES In addition to class actions, the United States has developed several alternative methods of compensating victims, the characteristics of which would fit better with the E.U.’s aims and legal culture than mass litigation would. These include agency funds, such as the SEC’s Fair Funds framework, that distribute disgorged profits and penalties to victims of violators of securities laws. Other kinds of public and private funds, which have also been used to compensate victims in circumstances where no agency enforcement exists, may also provide useful guidance in crafting an effective and efficient collective redress mechanism for Europe.
Agency Distribution SEC Fair Funds Reacting to the series of catastrophic financial scandals such as Enron and WorldCom that cost investors billions at the turn of the millennium, Congress passed the SarbanesOxley Act (“SOX”) of 200280 to tighten reporting and controls on American business to try to prevent such disasters from recurring. SOX included what has come to be known as the Fair Funds Provision, which allows SEC through an administered fund to distribute to the victims of securities law violations the sums it collects from violators through disgorgement, penalty, and settlement.81 Before SOX, it had not been one of the SEC’s primary functions to return funds to investors; rather, its objective was to detect wrongdoing through investigation and take steps to prevent or punish violations.82 Although it was permitted to distribute disgorged profits to investors, penalties imposed either in court or through administrative proceedings were required to go to the U.S. Treasury.83 Fair Funds broadened the SEC’s power to compensate victims by permitting it to distribute both disgorgement and penalties to investors, making significantly more money available for victim compensation.84 As a result of this legislation, the SEC has greatly increased the amounts of penalties it seeks from violators.85 While the SEC aims to distribute money to injured parties whenever possible, it may in its discretion choose to direct these sums to the Treasury, typically in cases where penalties are small compared to the population of victims, and economic considerations make it unfeasible to establish a distribution mechanism.86
226
ETHAN E. LITWIN AND MORGAN J. FEDER
When the SEC elects to establish a Fair Fund, it must devise a distribution plan. Sometimes, as in the WorldCom litigation, a distribution plan is established by court order.87 There, the court selected a Distribution Agent and an Equity Manager, and directed WorldCom to pay the specified amounts into the distribution fund while the SEC drafted a proposed distribution plan to submit to the court for approval.88 Following approval, the Distribution Agent became responsible for identifying eligible investor claimants and processing claims for payment from the fund.89 In other instances, the Fair Fund process is governed by the SEC Rules of Practice 11001106.90 Any party may submit a proposed plan, notice of which is then placed in the SEC Docket, on the SEC website, and in other publications as needed.91 There follows a period for comments,92 after which a hearing officer must review the plan for fairness before final approval and the commencement of distribution.93 Other than during the comment period provided by the SEC Rules of Practice, no challenges to this process are permitted, either to orders of disgorgement, creation of a fund, the distribution plan or eligibility to participate.94 Distribution typically takes place on a pro rata basis, with each victim receiving compensation in proportion to the degree of harm suffered.95 This is not the only method, however; in the WorldCom case, for example, the SEC distributed Fair Funds money to investors, but excluded large groups of unsecured creditors, whose remedies available in bankruptcy were limited.96 These groups challenged the distribution plan in court, arguing that they should share in the Fair Funds distribution, but the court held that the choice to exclude them was a fair and reasonable exercise of the SEC’s discretion, as it “equitably ensure[d] that the limited funds available to claimants will be directed to the individuals most severely affected by the fraud.”97 Because of the magnitude of the fraud perpetrated in some of these cases, and the correspondingly enormous losses suffered by investors, the harm may vastly exceed the profits disgorged and the penalties assessed that make up the funds available for distribution.98 Fair Funds should not be seen as a way of making investors whole; rather, the addition of penalties to the sums available for distribution represents a way of reducing losses to a greater degree, not eliminating them.99 This is not dissimilar from the hard choices that have to be made in private litigation as well, but in contrast to Fair Funds, recoveries in private actions are greatly reduced by lawyers’ compensation (often on the order of 30%) and other litigation costs, as well as the costs of notice, processing, and distribution, which can be significant. It has not been easy for the SEC to implement the Fair Funds changes and add such a comprehensive victim compensation dimension to its
European Collective Redress
227
primary function as an investigation and enforcement agency.100 Generally, thus far, distribution to victims under Fair Funds has been limited because of the challenges the SEC faced in developing “adequate systems and data to fulfill its oversight responsibilities.”101 This may be due, in part, to the fact that the SEC has had to divert resources from other areas to develop and run the distribution function.102 To handle its new duties, the SEC created a dedicated office, the Office of Collections and Distributions, to oversee Fair Fund distributions.103 Despite these challenges, the Fair Fund distribution rate has steadily increased over the years since its inception. During its first five years, the SEC collected approximately $8 billion in disgorgement and penalties, but as of June 2007 had distributed only $1.8 billion.104 By February 2010, a total of $9.5 billion had been ordered over the course of the program, $9.1 billion (96%) had been collected, and $6.9 billion (76%) had been distributed.105 Federal Trade Commission The U.S. Federal Trade Commission (“FTC”) is the administrative agency responsible for enforcement of the Federal Trade Commission Act (“FTCA”), which is aimed at regulating corporate trade practices and preventing unfair competition.106 The FTCA specifically empowers the FTC to provide restitution to injured consumers in the form of, inter alia, “rescission or reformation of contracts, the refund of money or return of property, [or] the payment of damages.”107 In addition to this express grant of authority under FTCA §57b, the FTC also may seek restitution, disgorgement, and other forms of consumer redress pursuant to its broad authority under FTCA § 13(b) to enjoin unfair trade practices and to seek “any ancillary equitable relief.”108 The FTC aggressively monitors and pursues violators on behalf of injured consumers, successfully returning hundreds of millions of dollars in consumer redress or disgorgement of ill-gotten gains.109 Despite these successes, weaknesses exist in the FTC’s approach that could be points for improvement in an E.U. model. For example, there should be clearer guidelines for the agency’s decision when and how to return collected funds to consumers. Additionally, more procedural safeguards would better ensure adequate representation of different victims’ groups’ interests and ensure that their due process rights are protected. The FTC also provides no opportunity to intervene in cases, nor to contest or object to distributions.110 This could be remedied by an expanded opportunity to comment on proposed distribution plans. Any successful collective redress mechanism must include reasonable provisions for claimant involvement to minimize their sacrifice of due process rights. Finally, consistency
228
ETHAN E. LITWIN AND MORGAN J. FEDER
and fairness to consumers and defendants would be better ensured by increased judicial checks on distribution plans; the FTC’s distribution processes involve little judicial oversight and an E.U. model should include a requirement for a judge to approve the final plans.
Other Public and Private Funds In addition to administrative agency programs distributing funds collected through enforcement processes, the United States has also applied the fund concept in other circumstances where litigation was either impractical or unpalatable. For example, Congress created the September 11th Victim Compensation Fund (the “Fund”) to compensate the victims of the September 11 terrorist attacks and their families, helping to rein in mass litigation that would have otherwise threatened to overwhelm the courts and potentially bankrupt the airline industry. Similarly, after the BP oil rig Deepwater Horizon exploded and caused billions of dollars’ worth of damage in one of the worst oil spills in U.S. history, the Gulf Coast Claims Facility (the “Facility”) was established to pay claims from the victims of that disaster. These were unique solutions to unique disasters, but certain principles or lessons from these experiences could be valuable in the E.U.’s creation of a collective redress framework. September 11th Victim Compensation Fund After the terrorist attacks of September 11, 2001, Congress passed the Air Transportation Safety and System Stabilization Act111 (“ATSSSA”) to try to avert the inevitable landslide of litigation that could have otherwise overwhelmed the courts and potentially bankrupted the airline industry. The purpose of the Fund was “to provide compensation to any individual (or relatives of a deceased individual) who was physically injured or killed as a result of the terrorist-related aircraft crashes of September 11, 2001.”112 Although it left open the option of private litigation, the ATSSSA sought to steer victims toward the Fund and lessen their incentive to file in court by imposing a cap on damages,113 as well as limitations on jurisdiction114 and available substantive law.115 These factors, combined with the Fund’s unlimited financial resources116 and speedy (compared to litigation) claims process, helped the Fund serve as a successful alternative to litigation, ultimately disbursing awards to approximately 97% of eligible mortality claimants.117 If they chose to participate in the Fund, however,
European Collective Redress
229
claimants had to waive or release all claims in litigation against the airlines, airports, and aircraft manufacturers.118 The ATSSSA, passed a mere 10 days after the terrorist attacks, did not set forth a comprehensive distribution plan. Rather, it called for the appointment of a Special Master to devise a compensation program and administer the Fund.119 The Special Master, appointee Kenneth Feinberg, had almost unlimited discretion to develop claim forms, conduct factual inquiries, determine eligibility of claimants, and determine awards.120 As part of the process of developing a distribution plan for the Fund, Feinberg published tables of presumed economic loss calculations with an explanation of his intended procedures and assumptions to aid claimants in the decision whether to participate in the Fund and waive their right to litigate.121 Awards were determined based on a claimant’s “economic loss,” defined as “any pecuniary loss resulting from harm,” including, inter alia, losses related to earnings, benefits related to employment, and loss of business or employment opportunities,122 from which was subtracted “collateral sources” such as life insurance, pension funds, and death benefit programs.123 Added to the remaining sum is an award for “noneconomic losses,” including physical and emotional pain and suffering.124 During this calculation, the Special Master had broad discretion to consider “individual circumstances” in raising or lowering the award.125 Once their claim form was submitted, claimants could elect to follow one of two tracks: under Track A, a claimant would receive a preliminary award determination and then decide whether to accept it outright or request a hearing before the Special Master; under Track B, the claimant would proceed straight to a hearing, after which the Special Master would deliver a final award determination.126 During these hearings, claimants or their attorneys could call witnesses and present evidence relating to the determination of their awards, including information about the victim, his or her spouse and dependents and their financial needs, facts related to noneconomic losses such as pain and suffering, and any other relevant factual or legal arguments they thought would be relevant to the Special Master’s decision.127 Claimants were notified in writing of the final award.128 These procedures enabled claimants to play an active part in the claims determination process and to advocate individually before the decisionmaker. Critics have questioned, however, whether claimants’ due process rights were adequately protected.129 The Special Master was not required to provide any explanation or justification for his claim determination, and no provision existed for appeal of the final determination, either before the Special Master or before a court.130 Without these procedures, the process
230
ETHAN E. LITWIN AND MORGAN J. FEDER
was arguably faster and cheaper, and resulted in greater certainty and finality. But without the checks and balances of written reasoning and judicial review, criticism has arisen that “errors, lapses in judgment, poor organization, and the appearance of arbitrariness and bias can arise when the review is limited. The lack of a written explanation or reason for the outcome allows an adjudicator to abandon responsibility and reject deference to the fairness of due process. In the absence of judicial review, the importance of reasoned and articulated explanations is heightened because they are the only forms of accountability remaining.”131 The Fund operated from 2001 to 2004, and paid over $7 billion to survivors of 2,880 people who died in the attacks, and to 2,680 who were injured in the attacks or the rescue efforts.132 Seven years later, in response to calls for compensation for a broader group of victims than those eligible under the Fund’s original mandate, President Obama signed the James Zadroga 9/11 Health and Compensation Act of 2010 (the “Zadroga Act”).133 Among other things, the Zadroga Act reactivated the Fund and broadened its scope to compensate additional individuals, including, for example, debris removal workers who suffer from cancer, respiratory diseases, and other conditions as a result of their work at the crash sites.134 The Zadroga Act established a five-year claims period, open from October 2011 to October 2016.135 In contrast to the earlier period of operation where there was no funding cap, the Zadroga Act provides for a limited $2.775 billion appropriation, with $875 million available in the first five years, for victim compensation and the Fund’s administrative costs.136 Because of the cap, successful claimants will initially receive only a portion of their allowed compensation, with the remainder to be paid once all claims are determined. This may mean that claims are pro-rated to ensure that all eligible claimants receive an award.137 Sheila Birnbaum, the appointed Special Master, issued proposed rules designed to implement the reactivated Fund. During the public comment period, 95 formal comments were received and considered, and hundreds of potential claimants participated in several town hall meetings across New York City. Additionally, Ms. Birnbaum met to discuss the reactivated Fund and the proposed rules and procedures with members of New York City agencies, unions, and communities affected by the events of September 11. Final rules took effect on October 3, 2011.138 The claims evaluation process is similar, but not identical to the Fund’s initial process. Once complete, each claim is presented to the Special Master’s Office for review. The Special Master then has 120 days to make a decision according to the new eligibility and economic loss models
European Collective Redress
231
developed for this process.139 This decision is final and not subject to judicial review; however, the claimant may appeal the decision to the Special Master and is entitled to a hearing before a hearing officer, where s/he may present evidence or witnesses.140 As of October 2012, over 14,000 potential claimants were registered on the online claims system, and 697 had submitted eligibility forms.141 Of these, a small number have been deemed eligible so far and are progressing toward completing their claims. The lessons learned from the Fund’s initial incarnation helped inform and streamline this new process. For example, the initial Fund was bankrolled by an unlimited “blank check” from Congress, but the limited appropriation for the current Fund is slated both for awards and for the administrative costs of running the Fund. Therefore, because every dollar spent on administration is a dollar less for claimants, the Special Master has had to create and maintain a more streamlined, efficient process to ensure the maximum recovery for each claimant.142 These processes have not yet been active long enough to permit an objective, data-based evaluation of their merit as compared to the processes used in the previous Fund. As that information becomes available, however, future creators and administrators of mass claim resolution mechanisms will benefit from being able to compare two approaches to similar issues. Gulf Coast Claims Facility The Gulf Coast Claims Facility was created following the explosion of BP’s Deepwater Horizon oil platform in the Gulf of Mexico, which killed 11 oil rig workers, spilled millions of gallons of oil into the water, polluted hundreds of miles of coastline, and caused billions of dollars’ worth of damage. In contrast to the Fund, which was created by an act of Congress, the Facility traces its origins to two sources: the Oil Pollution Act143 (“OPA”) and an agreement between BP executives and the Obama Administration.144 Following the spill, the U.S. Coast Guard designated BP as a “responsible party” under the OPA, which triggered a duty on the part of BP to pay for all costs related to the spill.145 In response, BP set up claims offices and hired claims adjusters to pay walk-in claims, and began paying emergency compensation to individuals and businesses on May 3, 2010, a mere 13 days after the explosion. As the OPA prescribes no particular procedures to which a “responsible party” must adhere to satisfy its duty to pay costs, however, the initial BP claims process was “chaotic, ad hoc, unsystematized, and largely unregulated.”146 Two months later, on June 16, 2010, BP and President Obama agreed that BP would replace this system with a fund of $20 billion to be paid into an escrow account to be administered by the September 11
232
ETHAN E. LITWIN AND MORGAN J. FEDER
Special Master, Kenneth Feinberg.147 Thus, the Facility is a privately created, privately administered entity; even though the Obama Administration played a role in its creation, there is no Executive Order, Congressional mandate, or other official order creating the Facility and its duties and powers are far less clear than the Fund’s.148 Although the Fund and the Facility share certain characteristics, the ways in which they differ provide a useful contrast of approaches and considerations that should help the E.U. in designing its own approach. Once the Facility was in place, Feinberg devised a system of standards and regulations to guide claimant compensation. Lacking even the basic guidelines provided by the ATSSSA, however, this process was less transparent and less formal than the process he undertook in designing the distribution plan for the Fund.149 Some critics have asserted that Feinberg made no public study of proposed Facility standards, did not share the sources on which he relied in drafting these standards, and kept the involvement of affected parties vague.150 While he did visit affected areas immediately following the spill and held several “town hall” meetings with residents and concerned parties, he did not provide any opportunity for claimants to comment on proposed standards.151 Critics argued that the procedures Feinberg ultimately developed were similarly unclear and seemingly arbitrary. Unlike the Fund, for example, which had clearly defined eligibility standards from the outset limiting the identity of claimants and compensable injuries, the group of potential claimants against the Facility was undefined and potentially enormous, and Feinberg, understandably, had to decide where and how to draw the line. He attempted to impose a “proximity” requirement, such that victims in the Gulf States who were directly affected by the spill would be eligible but indirect victims (such as a restaurant owner in Boston whose cost of shrimp increased) would not.152 This standard ultimately proved difficult to apply, Feinberg changed his mind repeatedly, and claimants accordingly expressed great frustration with the process.153 Responding to his critics, Feinberg modified his eligibility standards multiple times, gradually expanding the kinds of claims and the geographic regions that would be permitted to make claims.154 Ultimately, Feinberg settled on a three-tiered claims structure. Under the Quick Payment option, claimants who received emergency advance payments from the Facility would be eligible to receive a final payment of $5,000 for individuals or $25,000 for businesses without having to submit any additional documents or go through any more review.155 In exchange, they would sign a release waiving all claims against the Facility and the
European Collective Redress
233
defendants.156 Alternatively, claimants could apply for a full review final payment claim, receiving individualized evaluation of their claims (also releasing their rights in exchange for payment) or interim payments with proof of continuing losses (no release).157 Unlike the Fund, however, the Facility did not make available any compensation grid or guidance to help claimants assess their potential recovery for different kinds of damages, and the standards the Facility may be following in conducting their individualized evaluation and calculating awards are murky.158 Feinberg indicated that there might be a potential for review, however; whereas in the Fund, awards were final. Initially, he contemplated review by a three-judge panel; later, he indicated the panel would not be composed of judges, but of similarly credentialed professionals, such as law professors.159 The process for appeal was left undefined, and in the face of urging from the Department of Justice to implement an appeals process, Feinberg endorsed an appeals process through the U.S. Coast Guard.160 As of spring 2011, the Coast Guard had affirmed the Facility’s decision in every appeal it had processed.161 Perhaps deterred by the Facility’s opaque procedures and the difficulty in estimating a potential recovery, many individuals and entities with claims arising out of the Deepwater Horizon incident elected not to rely on the Facility for compensation. These claimants brought hundreds of civil lawsuits against BP and other defendants, which were consolidated in August 2010 in a Multi-District Litigation in the U.S. District Court for the Eastern District of Louisiana, styled In re: Spill by the Oil Rig “Deepwater Horizon” in the Gulf of Mexico on April 20, 2010.162 Trial was scheduled to begin on February 27, 2012. As they prepared for trial, the parties also engaged in continuous settlement negotiations, and ultimately reached an agreement as to economic and property damages. On December 21, 2012, the District Court in New Orleans entered an Order granting final approval to the Deepwater Horizon Economic and Property Damages Settlement.163 Recognizing that the Settlement would render the Facility superfluous, after the parties reached an agreement in principle, the court laid the groundwork for transitioning claims from the Facility process to the settlement program. On March 8, 2012, it ordered a Transition Process to handle claims until the new program was established. This Transition Process evaluated claims under the same protocols and methodologies as the Facility (with certain changes).164 The Transition Process ended when the settlement program formally opened on June 4, 2012, and began accepting claims and processing them pursuant to its terms, supervised by the court.165 The Facility is no longer functional.166
234
ETHAN E. LITWIN AND MORGAN J. FEDER
The Facility stands in marked contrast to the September 11th Victims Compensation Fund as an example of how the failure to involve the affected parties, lack of transparency and consistency in the decision-making process, and opaque eligibility and compensation guidelines can lead to the failure of even a well-funded, well-intentioned privately administered mass claims resolution mechanism, and drive claimants back to the courts. The Fund was particularly robust in these areas, and thereby efficiently compensated victims while avoiding all but a tiny number of court litigations. By comparison, the Facility was inconsistent, unclear, and ultimately failed in its effort to streamline and speed the compensation process, as ongoing court intervention and supervision continue to be necessary to provide claimants the compensation they seek. The lessons offered by these examples may be particularly useful in the creation and administration of future mass claims resolution mechanisms. Both the Gulf oil spill and the terrorist attacks of September 11 were, of course, one-time events that are unlikely to repeat, at least not in the same way. Problems such as that of notice to the class were not present in these examples because they were significant events drawing national attention. Applying these models in more typical contexts in which the need for class actions arises such as, for example, the purchase by millions of people of a price-fixed commodity product like CD jewel cases, where careful records of who purchased what quantities at what prices are not maintained centrally will still be a challenge when it comes to certain necessary steps like notice to the class. The benefits of a regulatory administered compensation system, however, lie outside of the ever-present problem of notice. By bypassing the judicial system and its attendant delays and inefficiencies, regulatory administered compensation systems have the potential for compensating victims more quickly than does traditional litigation.167 Especially where robust civil enforcement exists already, there is little need to duplicate the cumbersome process of gathering evidence and proving violations to another finder of fact. The investigating agencies are best placed to use the results of their investigation as a starting point for developing an administered fund to compensate victims.
CONCLUSION: A MODEST PROPOSAL In light of the foregoing, we propose that the E.U. generally reject the concept of the private class action mechanism or other mass or representative
European Collective Redress
235
litigation models, in favor of administered funds, modified to meet the needs of the particular circumstances presented by specific kinds of cases. Where a robust system of public civil enforcement exists (for example, in the E.U. competition law sector), the underlying facts and civil liability of the defendants have already been determined by a governmental body. Learning from the United States’s experience with the SEC Fair Funds legislation, the E.C. could introduce a process that would complement the enforcement function and segue smoothly into victim compensation. The agency-administered fund concept is a strong alternative to private litigation that would serve many of the purposes and meet many of the criteria enumerated by the E.C. and the E.P. for an E.U. collective redress framework. First and foremost, redress through an agency eliminates the disadvantages and inappropriate incentives that can accompany private attorney representation, obviating the issues of third-party funding, contingency fees, and the loser pays requirement that are, as discussed above, problematic in the mass litigation context. Public agencies, which do not rely on contingency fee or hourly rate funding, are better positioned to return a greater portion of the recovered funds and penalties to victims.168 Although additional funding would be necessary to support an agency’s distribution activities beyond what is required simply to maintain an enforcement office, efficiencies result when work product and individual knowledge can be carried over from the enforcement side to the distribution side. This contrasts with parallel or “follow-on” private litigation in court, which would require a private firm to conduct afresh all the legal and factual research the agency has already done, and then prove to a new fact finder what the agency has already established. Eliminating all this repetitive work and proceeding straight to a determination of each claimant’s eligibility for recovery will be faster and more efficient than private litigation. The Commission is well-placed to administer a victims’ restitution fund. Take, for example, the Commission’s work in the area of competition law, one of the primary targets of the collective redress proposal. The Commission is one of the leading agencies worldwide in investigating violations of competition law, regularly topping the charts both in the number of investigations and the amount of fines issued. In its work in the competition sector, the Commission develops a robust evidentiary file, assesses whether the targets of the investigation have breached European competition law, and drafts a reasoned ruling explaining its decision. Having already determined whether there is a civil violation of E.U. competition law, the Commission would not need to relitigate that issue in a victim compensation setting. Rather, the Commission could simply quantify the
236
ETHAN E. LITWIN AND MORGAN J. FEDER
gross amount of damage to European victims caused by any conduct deemed to have been anticompetitive and order the infringing companies to pay that amount into a victim restitution fund. The Commission could, as Congress has done, appoint a Special Master to administer the fund and order the infringing parties to pay the Special Master’s fees. In addition to diverting a higher percentage of damage awards to victims than traditional litigation, an agency-administered fund arguably provides for greater due process protections. Although class action devices preserve, in theory, the opportunity to litigate individually, that option is not realistic for the vast majority of potential claimants. Mass litigation fails to preserve essential due process rights, including the right to notice, the right to be heard, and the right to adequate representation. Class actions fail either to protect absent class members by failing to provide adequate notice or to adequately represent known class members because of complicated individual issues and the inability to identify a truly representative plaintiff. Finally, class actions fail in both absent and known class members by denying them a true opportunity to be heard by the court and to influence the prosecution of the litigation. While imperfect, agency-administered funds address most of these concerns.169 For example, the Fair Funds mechanism’s comment period does offer a certain opportunity for individual victims to be involved in the development of the distribution process that will determine their rights, although we would suggest that greater involvement would be beneficial and would reduce the loss of individual due process rights that must inevitably accompany any collective procedure. Agencies, given their extraordinary access to government records, can provide superior notice than private lawyers can. Finally, there are no concerns about adequacy of representation with respect to agency funds, as there is no need to litigate the merits of the case through a representative.170 In sum, a balance must be struck between efficiency and due process; perhaps a model that provided for a longer or more in-depth opportunity to comment, or even an opportunity for parties to vote on alternatives, could raise the level at which victims may participate in the determination of their due process rights. In areas of regulatory under-enforcement, it may be necessary to draw on the lessons from the Fund and the Facility in designing a framework from scratch. Because situations creating a need for victim compensation will not all relate to mass disasters, the framework adopted may not ultimately correspond to one-off funds like these two examples. Rather, in light of the information assembled during the public consultation on
237
European Collective Redress
collective redress, an E.U.-level committee could be established that would be tasked with identifying particular areas of business or law that are most in need of a collective compensation mechanism, where the risk of small injuries to large numbers of people in multiple Member States is highest. Resources could then be devoted either to strengthening or adapting existing enforcement systems that could put in place a Fair Funds-type model of administration, modified to meet the needs of that particular area. This would require a significant financial investment, but the alternative private litigation solution would similarly increase the burden on courts and judges and create additional costs there, too. Alternatively, funding could be increased by diverting a certain portion of fines to the maintenance of the new victim compensation system. The E.C. has, in the past, suggested that the adoption of collective redress procedures would further two important policy goals: compensating victims of unlawful conduct and enhancing enforcement of European law. The authors submit that mass litigation procedures, in whatever form they ultimately would take, would neither effectively nor efficiently accomplish either goal. All forms of mass litigation suffer from the same fundamental flaw: whether structured in the form of representative actions, opt-in or opt-out class actions, all mass litigation procedures require the sacrifice of individual due process rights. Simply put, all collective actions require individual claimants and defendants to sacrifice their rights so that rough justice may be done. European courts, much like their American brethren, are ill-equipped to adapt what is essentially a forum for resolving individual legal complaints to the largely political process of mass compensation schemes. The U.S. experience strongly suggests that entrusting important policy objectives, such as the compensation of victims, to private entities is an inefficient and ineffective means of accomplishing these goals. Litigation is best reserved for individual parties who have failed to resolve their disputes amicably. Victim compensation and enforcement of E.U. law is best left to government officials, who act with the best interests of the consumer foremost in mind.
NOTES 1. European Parliament Committee on Legal Affairs, Report on “Towards a Coherent European Approach to Collective Redress,” 4 (2011/2089(INI) January 12, 2012) (“E.P. Report”).
238
ETHAN E. LITWIN AND MORGAN J. FEDER
2. E.P. Report, at 11. See also European Commission Staff Working Document, Public Consultation: Towards a Coherent European Approach to Collective Redress, 2 (SEC(2011) 173, February 4, 2011) (“E.C. Public Consultation”). 3. See E.C. Public Consultation at 24; Commission of the European Communities, Green Paper on Consumer Collective Redress, 4 (COM(2008) 794, November 27, 2008) (“Collective Redress Green Paper”); European Commission DG SANCO, “Evaluation of the Effectiveness and Efficiency of Collective Redress Mechanisms in the European Union,” Final Report Part I: Main Report, 4 (SANCO/2007/B4/004, August 26, 2008) (“Evaluation of Collective Redress”). 4. See Collective Redress Green Paper at 4, 6; E.C. Public Consultation at 3; European Commission Flash Eurobarometer 299, “Consumer Attitudes Towards Cross-Border Trade and Consumer Protection,” 8 (March 2011). 5. The Netherlands, Sweden, Denmark, the United Kingdom, France, Italy, Norway, Germany, and Poland, among others, have proposed, adopted, or significantly revised procedures for aggregating litigation claims. 6. E.C. Public Consultation, at 4. 7. See, for example, ibid., at 45; Collective Redress Green Paper; Commission of the European Communities, Green Paper on Damages Actions for Breach of the E.C. Antitrust Rules (COM(2005) 672, December 19, 2005) (“Antitrust Green Paper”); Commission of the European Communities, White Paper on Damages Actions for Breach of the E.C. Antitrust Rules (COM(2008) 165, April 2, 2008) (“Antitrust White Paper”). 8. For example, E.P. Report, at 11 (“The recent decision of the U.S. Supreme Court in a class-action bias case shows once more that the U.S. legal system itself is fighting against abusive and unmeritorious class actions resulting from the excesses of the U.S. system that were certainly not envisaged when such actions were introduced decades ago. Europe must stand firm against any intention to changing E.U. legal traditions by incorporating alien procedural elements allowing for abusive collective action.” (referring to Wal-Mart Stores Inc. v. Dukes, 131 S. Ct. 2541 (2011))). 9. The SarbanesOxley Act of 2002 provides for the use of Fair Funds at 15 U.S.C. § 7246(a) (“Fair Funds Provision”). 10. E.C. Public Consultation at 5. 11. Antitrust Green Paper at 3. 12. Ibid., at 4 (noting that “this area of law in the 25 Member States presents a picture of ‘total underdevelopment’,” citing Ashurst (2004); Antitrust White Paper, at 2. 13. Antitrust Green Paper, at 8; Antitrust White Paper, at 2. 14. Antitrust White Paper, at 2. 15. See Collective Redress Green Paper, at 23; see also Commission of the European Communities, Communication from the Commission to the Council, the European Parliament and the European Economic and Social Committee, E.U. Consumer Policy Strategy 20072013, 11 (COM(2007) 99, March 13, 2007) (“E.U. Consumer Policy Strategy”). 16. See Collective Redress Green Paper, at 23. 17. Ibid., at 23. See generally E.U. Consumer Policy Strategy. 18. See Collective Redress Green Paper, at 3. 19. High litigation costs and lengthy, complex procedures mean that one in five European consumers say they will choose not to go to court on their own for claims
European Collective Redress
239
worth less than 1,000 euros. Half will not go to court for claims worth less than 200 euros. Collective Redress Green Paper, at 4. A study by the U.K. Office of Fair Trading found that an average of only 62% of harmed U.K. consumers complain; a figure that drops to 54% for purchases below 10 GBP. Ibid. 20. See, for example, Evaluation of Collective Redress; Collective Redress Green Paper, at 5 n.18 (the E.C. held several workshops from 2007 to 2008 in which it consulted with consumers, business stakeholders, and legal practitioners about effective and efficient collective redress approaches). 21. E.C. Public Consultation, at 4. 22. E.C. Public Consultation, at 4; Collective Redress Green Paper, at 45; Evaluation of Collective Redress, at 6 et seq. 23. Evaluation of Collective Redress, at 67, 27. 24. Ibid., at 7, 2930. 25. Ibid., at 7, 31. 26. Ibid., at 7, 31. 27. For example, specific mechanisms exist only for the recovery of capital investment losses in Germany or damages caused by anticompetitive practices in the U.K. E.C. Public Consultation, at 4. 28. In addition to permitting individuals to bring claims on behalf of a group, certain Member States vest standing in public authorities, such as Finland’s Ombudsman; others, like Bulgaria, permit private organizations such as consumer associations to bring claims. Many Member States have a combination of these and other standing rules and requirements. Ibid. 29. Most Member States follow the “opt-in” format, binding only those who have expressly joined a proceeding. Others have rules that bind all members of a group unless they expressly “opt-out,” such as Portugal, Denmark, and the Netherlands. E.C. Public Consultation, at 4. The United Kingdom is in the process of implementing what appears to be a limited opt-out collective action mechanism for claimants seeking redress for violations of competition laws. See Department for Business Innovation and Skills, Private Actions in Competition Law: A Consultation on Options for Reform Government Response (2013). The right to file an opt-out collective action would not be automatic under the pending draft legislation; the Competition Appeal Tribunal would decide on a case-by-case basis whether to permit any such action to proceed. To further efforts to reduce the potential for frivolous litigation, the draft legislation also provides for strict judicial certification of cases, a ban on treble or punitive damages, a prohibition on contingency fees for lawyers, and a “loser pays” rule. The draft legislation would permit collective actions to be brought by individual claimants or businesses, and also by representative bodies such as trade associations. The opt-out policy would be limited to U.K. class members, though class members domiciled elsewhere in the European Union would have the option of opting in. Ibid. Commentators have suggested that the proposal, if adopted into law, would make the United Kingdom an attractive forum for private actions alleging violations of competition law. See, for example, Vane (2013). 30. Ibid. For a comprehensive review of existing national mass action mechanisms in E.U. Member States, see generally Evaluation of Collective Redress. 31. Collective Redress Green Paper at 5; Evaluation of Collective Redress, at 4 (over the study period of approximately a decade, 326 consumer-relevant cases were initiated in the Member States that had introduced collective redress procedures.
240
ETHAN E. LITWIN AND MORGAN J. FEDER
The highest numbers of cases were concentrated in France, Spain, Germany, and Austria, and focused on wrongdoing in the financial services and telecommunications fields. Roughly 10% of these cases presented cross-border issues). 32. Evaluation of Collective Redress, at 45. 33. Ibid., at 5. 34. See Collective Redress Green Paper, at 5. 35. European Commission, “Commission Recommendation of Common Principles for Injunctive and Compensatory Collective Redress Mechanisms in the Member States Concerning Violations of Rights Granted under Union Law,” C(2013) 3539/3 (June 11, 2013) (“E.C. Recommendation”). 36. See generally E.C. Recommendation, at 510. 37. Ibid., at 10. 38. European Commission, “Proposal for a Directive of the European Parliament and of the Council on Certain Rules Governing Actions for Damages under National Law for Infringements of the Competition Law Provisions of the Member States and of the European Union,” COM(2013) 404 (June 11, 2013) (“Proposed Directive”). 39. FAQ on Proposed Directive, MEMO 13/531. Retrieved from http://europa. eu/rapid/press-release_MEMO-13-531_en.htm (last visited February 3, 2014). 40. European Parliament, Resolution of 2 February 2012 on “Towards a Coherent European Approach to Collective Redress” (2011/2089(INI)) (“E.P. Resolution”); see also E.P. Legislative Observatory. 41. E.P. Resolution, at ¶2; E.P. Report, at 11 (referring to “alien” procedures that no doubt refer to U.S. rules regarding class actions, discovery, contingency fee arrangements, and punitive damages). 42. As will be discussed in greater detail in Part III, “opt-in” class actions require that each claimant affirmatively express the desire to participate, and a decision binds only those who have chosen to join the class. “Opt-out” class actions, by contrast, bind anyone fitting the class definition, unless they affirmatively express their desire not to be bound. 43. E.P. Resolution, at ¶20; E.P. Report, at 13. 44. E.P. Resolution, at ¶20; E.P. Report, at 13. 45. The authors are unaware of any opt-in regimes that have operated successfully to achieve the aims of collective actions as outlined in this article. 46. E.P. Resolution, at ¶20. 47. Ibid.; E.P. Report, at 13. See also Antitrust White Paper, at 4. 48. E.P. Report, at 14. 49. E.P. Resolution, at ¶20; E.P. Report, at 14. 50. E.P. Resolution, at ¶20; E.P. Report, at 14. 51. E.P. Resolution, at ¶20; E.P. Report, at 14. 52. E.P. Resolution, at ¶20; E.P. Report, at 14. 53. The loser pays rule would not cure the flaws of the opt-out model either. Presenting plaintiffs with the prospect of having to incur their opponents’ expenses, the loser pays rule in the opt-out context would not only deter the filing of unmeritorious claims, but also discourage filing worthy lawsuits where the result was uncertain, thereby frustrating the class action’s central purpose to compensate those whose claims are so small they would otherwise not be brought.
European Collective Redress
241
54. See Replies to the E.C. Public Consultation. Retrieved from http://ec.europa. eu/competition/consultations/2011_collective_redress/index_en.html (last visited February 3, 2014). 55. McLaughlin (2011) (quoting In re West Virginia Rezulin Litigation, 214 W. Va. 52 (2003)). 56. See ibid., at 3, n.5. 57. Ibid. (quoting Besinga v. U.S., 923 F.2d 133, 135 (9th Cir. 1991)) (internal quotation marks omitted). 58. Ibid., at 2 (citing Hansberry v. Lee, 311 U.S. 32, 40 (1940); Taylor v. Sturgell, 553 U.S. 880 (2008); Pelt v. Utah, 539 F.3d 1271, 1281 (10th Cir. 2008); In re Telelectronics Pacing Systems, Inc., 221 F.3d 870, 872 (6th Cir. 2000); In re Vivendi Universal, S.A. Securities Litigation, 2009 WL 855799, at *7 (S.D.N.Y. 2009)). 59. Ibid., at 4 (citing West v. Randall, 29 F. Cas. 718, 721, No. 17424 (C.C.D. R. I. 1820) (citing Bonanno v. Quizno’s Franchise Co., LLC, 2009 WL 1068744, at *9 (D. Colo. 2009))). 60. Ibid. (citing West, 29 F. Cas. at 721; In re Ski Train Fire in Kaprun, Austria on November 11, 2000, 220 F.R.D. 195, 210 n.31 (S.D.N.Y. 2003), rev’d on other grounds, 393 F.3d 120 (2d Cir. 2004)). 61. Ibid. (citing West, 29 F. Cas. at 721; In re Ski Train Fire, 220 F.R.D. at 210 n.31). See also ibid., at 5, n.11, 12 (describing Equity Rules that provided for a form of “class action” where parties were very numerous, but also did not bind absent parties). 62. Ibid., at 5 (citing Amchem Products, Inc. v. Windsor, 521 U.S. 591, 615 (1997); Pelt, 539 F.3d, at 1284). 63. Ibid., at 7 (citing Fireside Bank v. Superior Court, 40 Cal. 4th 1069, 1078 (2007) (“One-way intervention left a defendant open to ‘being pecked to death by ducks. One plaintiff could sue and lose; another could sue and lose; and another and another until one finally prevailed; then everyone else would ride on that single success. This sort of sequence, too, would waste resources; it also could make the minority (and therefore presumptively inaccurate) result the binding one.’”) (citation omitted). 64. Ibid., at 56, n.17. 65. Ibid., at 6 (citing Amchem Prods., Inc., 521 U.S., at 614615). 66. Ibid., at 78. 67. See ibid. (citing Fed. R. Civ. P. 23 Advisory Committee Note; White v. Deltona Corp., 66 F.R.D. 560, 563 (S.D. Fla. 1975)). 68. Ibid., at 2, n.3. 69. Ibid., at 8 n.26 (quoting In re General Motors Corp. Pick-Up Truck Fuel Tank Prods. Liab. Litig., 55 F.3d 768, 785 (3d Cir. 1995) (internal quotation marks omitted)). 70. Ibid., at 9. 71. Ibid., at 910 (citing one court’s rationale for approving aggregated proof of damages in an asbestos class action: “[u]nless we can use the class action and devices built on the class action, our judicial system is not going to be able to cope with the challenges of the mass repetitive wrong.” Cimino v. Raymark Industries, Inc., 751 F. Supp. 649, 652 (E.D. Tex. 1990), aff’d in part, vacated in part on other grounds, 151 F.3d 297 (5th Cir. 1998)).
242
ETHAN E. LITWIN AND MORGAN J. FEDER
72. Ibid., at 10. 73. Ibid., at 11, n.35 “This protection from inconsistent adjudications derives from the fact that the class action is binding on all class members …. By contrast, proceeding with individual claims makes the defendant vulnerable to the asymmetry of collateral estoppel: If [defendant] lost on a claim to an individual plaintiff, subsequent plaintiffs could use offensive collateral estoppel to prevent [defendant] from litigating the issue.” (quoting Gunnells v. Healthplan Services, Inc., 348 F.3d 417, 427 (4th Cir. 2003) (internal quotation marks omitted)). 74. See, for example, Willging, Hooper, and Niemic (1996) (discussing difficulties inherent in and failures of notice to be given in class actions). See also generally Leslie (2007). 75. In Maffei v. Alert TV of North Carolina, 342 S.E.2d 867, 872 (N.C. Sup. 1986), the court rejected a proposed class settlement which would have given each class member a grand total of $0.29 less than the postage to file the claim. 76. Claim rates in many opt-out classes are low for this and similar reasons. See, for example, Perish v. Intel Corp., No. CV-75-51-01 (Cal. Super. Ct. Santa Clara Cty 1998) (150 class members out of an estimated 500,000 claimed their $50 coupon award); Strong v. BellSouth Telecommunications, Inc., 173 F.R.D. 167, 172 (W.D. La. 1997) (out of a $64 million fund, only $1.8 million was claimed). 77. Rule 23 requires the presence of a “lead plaintiff” whose interests are aligned with the class to serve as their representative, guide the lawyers during the litigation and settlement processes, and “fairly and adequately protect the interests of the class.” Fed. R. Civ. P. 23(a)(4). Often it is not a lead plaintiff who is the motivating force behind these lawsuits, however, it is opportunistic lawyers who selectively initiate suits designed to yield large settlements, of which they receive a percentage. Just as in the Milberg Weiss scandal (see, e.g., McDonald, 2008), some lawyers in effect “hire” lead plaintiffs who fit the definition of a class action they wish to bring, sometimes inducing them to sue with promises to share in the legal fees or other compensation. Some lawyers thus create figureheads whose interests are more closely aligned with the lawyers’ than with those of the rest of the class whom they are supposed to protect. See also Leslie, supra n.74, at 8081 (citing Brunet, 2003). 78. It remains an open question whether current regulatory enforcement successfully deters illegal behavior, or whether it suffers from either Type 1 or Type 2 error (i.e., whether it over-deters, limiting legal behavior along with illegal behavior, or it falls short and under-deters illegal behavior). Depending on the answer, private litigation may not be needed to further deterrence-related objectives. In any event, an examination of this issue is beyond the modest scope of this article. This article assumes that some private actions will have some deterrent benefit and instead focuses on victim compensation and the efficiency of the systems designed to achieve it. For a robust examination of the deterrent effect of regulatory enforcement of competition laws, see, for example, Centre for European Policy Studies, 2007. 79. Some of the settlements to come out of the opt-out system have been in the interest of the class counsel but do not appear to have been in the interest of the class members. In Kamilewicz v. Bank of Boston Corp., 100 F.3d 1348 (7th Cir. 1996), each class member received $8.76, but class counsel received at least $8.5 million. In a class action against cable company Charter Communications, lawyers received $5.5 million in fees, while class members received only a new late-payment
European Collective Redress
243
policy, for which the company compensated itself by increasing prices. See Brueggeman (2002), at A40. One class member complained, “please don’t sue anyone else on my behalf. I can’t afford any more of these brilliant legal victories.” Ibid. 80. SarbanesOxley Act of 2002, Pub. L. No. 107-204, 116 Stat. 745 (codified in scattered sections of 11, 15, 18, 28, and 29 U.S.C.). 81. Winship (2008) 1103, 1121 (citing the Fair Funds Provision, 15 U.S.C. § 7246 (a); Report Pursuant to Section 308(c) of the SarbanesOxley Act of 2002, 2. Retrieved from http://www.sec.gov/news/studies/sox308creport.pdf (last visited February 3, 2014)). See also Bromberg and Lowenfels (2011). 82. Winship, supra n.81, at 11101111; Zimmerman (2011). 83. Bromberg & Lowenfels, supra n.81. 84. Aitkins and Bondi (2008). Fair Funds empower the SEC to distribute penalties to victims, but only where the SEC also gets disgorgement from the wrongdoer of ill-gotten gains: disgorgement is a prerequisite for the SEC to be able to add penalty sums to a Fair Fund. 85. Herr, M. E. (2012); see also Atkins & Bondi, supra n.84, at 399400 (“in 2002, the SEC achieved its first $10 million penalty against a public corporation in its settlement with Xerox Corporation. Since then, the Commission has levied many civil penalties in that amount or larger. In 2003, the Commission obtained twenty penalties in that range or greater, while in 2004, it obtained forty such penalties …. From 2003 to 2007, approximately $13.8 billion in disgorgement and penalties were ordered to be paid to the SEC, courts or other appointed trustees.” (citations omitted)). 86. Zimmerman, supra n.82, at 532, n.146; Winship, supra n.81, at 11191120. 87. Bromberg & Lowenfels, supra n.81. 88. Ibid. (citing SEC v. WorldCom Inc., SEC Litig. Rel. No. 34-18451 (S.D.N.Y. November 10, 2003, 81 SEC Docket 1936 (2003))). 89. Ibid. (citing WorldCom Inc., SEC Litig. Rel. No. 34-18451, 81 SEC Docket 1936). The WorldCom Distribution Plan, SEC Litig. Rel. No. 34-18789 (S.D. July 23, 2004), 83 SEC Docket 1407 (2004). Retrieved from www.sec.gov/litigation/ litreleases/worldcom163a.pdf (last visited February 3, 2014). 90. 17 C.F.R. §§ 201.1100201.1106, 6 Fed. Sec. L. Rep. ¶¶66, 310366, 310T. 91. Bromberg & Lowenfels, supra n.81. 92. Zimmerman, supra n.82, at 532 n.147 (citing 17 C.F.R. § 201.1103). 93. Ibid., at 532 n.149 (citing 17 C.F.R. § 201.1104); Bromberg & Lowenfels, supra n.81. 94. Bromberg & Lowenfels, supra n.81 (citing SEC Rule of Practice 1106 (17 C.F.R. § 201.1106)). This lack of claimant involvement has drawn criticism from some commentators (see, e.g., Zimmerman, supra n.82, at 532). If the E.U. were to use Fair Funds as a model, individual due process would be better preserved if a more robust victim involvement procedure were added. 95. Zimmerman, supra n.82, at 532 (citing SEC v. Bear, Stearns & Co., 626 F. Supp. 2d 402, 407 (S.D.N.Y. 2009)). 96. SEC v. WorldCom, Inc., No. 02-CV-4963, 2004 WL 1621185, at *12 (S.D.N.Y. July 20, 2004). 97. Ibid., at *2. The court noted that “when funds are limited, hard choices must be made.” Ibid.
244
ETHAN E. LITWIN AND MORGAN J. FEDER
98. Winship, supra n.81, at 11241126. In the WorldCom case, the court estimated that the loss to shareholders was approximately $200 billion. In the SEC action, the money penalty was set at $2.25 billion, later reduced to $750 million in bankruptcy. 99. Ibid., at 11241125. 100. Ibid., at 1122. 101. Ibid., (quoting U.S. Government Accountability Office (2007) (“GAO Report”). 102. Ibid., at 1136. 103. Ibid., at 11361137 (citing 73 SEC Ann. Rep. 30 (2007); GAO Report, at 30. 104. Ibid., at 11221123 (citations omitted). 105. U.S. Government Accountability Office (2010). 106. Federal Trade Commission Act, 15 U.S.C. § 41 et seq., § 45. 107. 15 U.S.C. § 57b(b). The FTC is not, however, authorized to impose exemplary or punitive damages in this context. Ibid. 108. Zimmerman, supra n.82, at 534 (citing FTC v. Amy Travel Servs., Inc., 875 F.2d 564, 572 (7th Cir. 1989) (“We hold that in a proceeding under section 13(b), the statutory grant of authority to the district court to issue permanent injunctions includes the power to order any ancillary equitable relief necessary to effectuate the exercise of the granted powers.”); FTC v. Febre, 128 F.3d 530, 534 (7th Cir. 1997) (relying on Amy Travel Servs. to hold same, stating “the power to grant ancillary relief includes the power to order repayment of money for consumer redress as restitution or rescission.”)). The FTC may also seek disgorgement and restitution for violations of the Clayton Antitrust Act of 1914. Ibid. 109. Ibid. (citing, e.g., the FTC’s 2009 Annual Report, FTC, the FTC in 2009, at 54 (2009) (Retrieved from http://www.ftc.gov/os/2009/03/2009ftcrptpv.pdf (last visited February 3, 2014)) (reporting that “from March 2008 to February 2009, the FTC filed 64 actions in federal district court and obtained 83 judgments and orders requiring defendants to pay more than $371.2 million in consumer redress or disgorgement of ill-gotten gains.”)). The FTC has also achieved multimillion-dollar recoveries for victims of cold-remedy marketing scams, subprime home mortgage lenders, and telemarketing schemes. The FTC in 2009 at 45. In FY 2012, the FTC reported that it returned over $36 million in redress funds to consumers and nearly $123 million to the U.S. Treasury from fees, redress disgorgements, and fines, and obtained the largest litigated judgment ever by the agency, $478 million, against get-richquick system marketers that prey on financially distressed consumers. FTC, 2012 Performance Snapshot, at 1 (2012) (Retrieved from http://www.ftc.gov/reports/ 2012-one-page-ftc-performance-snapshot, last visited February 3, 2014). 110. Zimmerman, supra n.82, at 535. 111. Pub. L. No. 107-42, 115 Stat. 237 (2001) (codified at 49 U.S.C. § 40101 Note). 112. ATSSSA § 403. 113. The ATSSSA caps litigation damages at the limits of the airlines’ insurance coverage, approximately $6 billion. Ibid. § 408(a). 114. The ATSSSA vests exclusive jurisdiction in the U.S. District Court for the Southern District of New York. Ibid. § 408(b)(3). 115. The ATSSSA limits available substantive law to that of New York, Pennsylvania, and Washington, DC, the sites of the plane crashes. Ibid. § 408(b)(2).
European Collective Redress
245
116. The Fund is paid for by taxpayers. Congress set no limit in the ATSSSA on the amount of money available to be disbursed; as appointed Special Master Kenneth Feinberg put it, Congress gave him a blank check. 117. Chen (2004), at A1. After the application deadline, some 70 private lawsuits were pending, and 13 families elected neither to sue nor to apply for compensation from the Fund. Ibid. 118. ATSSSA § 405(c)(3)(B)(i). 119. Ibid. § 404(a)(1)(3). 120. Ibid. §§ 405(a)(2)(A), 405(b). 121. September 11th Victim Compensation Fund (2002). 122. ATSSSA § 402(5). 123. Ibid. § 402(4). 124. Ibid. § 402(7). 125. Ibid. § 405(b)(1)(B)(ii). 126. September 11th Victim Compensation Fund (2002). 127. Ibid. See also Berkowitz (2006). 128. September 11th Victim Compensation Fund (2002) (“Fund FAQ”). 129. See, e.g., Berkowitz, supra n.127, at 15. 130. ATSSSA § 405(b)(3); Fund FAQ. 131. Berkowitz, supra n.127, at 15. 132. Birnbaum (“2012 Status Report”) (2012). 133. Pub. L. No. 111-347, 124 Stat. 3623 (2011). 134. 2012 Status Report, at 1. 135. Ibid. 136. Ibid. 137. September 11th Victim Compensation Fund, Frequently Asked Questions (“New Fund FAQ”) 1.5. 138. Ibid., at 12. 139. Ibid., at 1, 4. 140. Ibid. 141. Ibid., at 5. 142. New Fund FAQ 1.5. 143. 33 U.S.C. § 2701 et seq. 144. See BDO Consulting (2012) (the “DOJ Report”). 145. See ibid. 146. Mullinex (2011). 147. See DOJ Report, at 1213. 148. Mullinex, supra n.146, at 836, 842 (“Simply, the GCCF is a largely privatized enterprise not subject to public legitimacy constraints.”). 149. Ibid., at 842. 150. Ibid., at 843844 (citing King (2010) (contending that Feinberg was laying down rules for the claims fund with little input from the states most affected)); Editorial, The Gulf Claims Racket (2010), at A14. 151. Mullinex, supra n.146; King, supra n.150 (dozens of town hall meetings had not shielded Feinberg from accusations of secrecy and coziness with BP; criticism from Alabama Attorney General Troy King accused Feinberg of acting as a “corporate shill”). 152. Mullinex, supra n.146, at 847848.
246
ETHAN E. LITWIN AND MORGAN J. FEDER
153. Ibid. 154. Ibid., at 850. At one point, seeking to minimize the burden of proof on fishermen, shrimp haulers, and similarly situated claimants, Feinberg announced that he would accept as proof of their claims letters from their local pastors. He later retracted this, however. He also ceased processing claims on an individual basis in some circumstances, instead “grouping” claimants by business or profession to try to increase consistency for similarly situated claimants, but as before left the procedures to be followed in administering these groups unclear. Ibid. at 851853. 155. See DOJ Report at 34. 156. Ibid. 157. Ibid., at 35. 158. Mullinex, supra n.146, at 858. 159. Ibid., at 879. 160. Ibid., at 880. 161. Editorial, supra n.150. 162. In re: Spill by the Oil Rig “Deepwater Horizon” in the Gulf of Mexico on April 20, 2010, MDL No. 2179 (E.D. La.). See also Order and Reasons Granting Final Approval of the Economic and Property Damages Settlement Agreement, No. 10-md-2179, at *1 (E.D. La. December 21, 2012). 163. Order and Judgment Granting Final Approval of Economic and Property Damages Settlement and Confirming Certification of the Economic and Property Damages Settlement Class, No. 10-md-2179 (E.D. La. December 21, 2012). 164. First Amended Order Creating Transition Process, No. 10-md-2179 (E.D. La. March 8, 2012). 165. Deepwater Horizon Claims Center, Frequently Asked Questions No. 47. 166. Ibid. Because the Settlement does not resolve all claims against BP arising out of the Deepwater Horizon incident, however, BP has also created an OPA Claims Process to handle claims that do not fall within the terms of the Settlement, or claimants who are not class members or who wish to opt out of the Settlement. Information about the OPA Claims Process can be found at http://www.bp.com/ claims (last visited February 3, 2014). 167. Critics of regulatory compensation systems rightly note that outside of mass disasters (e.g., September 11) that garner significant political support and attention, extra-judicial methods of compensating victims have struggled at times to provide meaningful levels of compensation. The initial difficulties faced by regulatory compensation programs such as Fair Funds should not, however, be seen as conclusive evidence that such systems are unworkable. Given a clear legislative mandate and sufficient support and staffing, regulators such as the SEC or the European Commission should be able to efficiently and effectively administer a civil restitution program for the victims of illegal behavior those regulators have already investigated. 168. See Winship, supra n.81, at 11341135. 169. The most serious challenge to an effective and efficient regulatory compensation system remains the motivation and resources of the relevant agencies. Administering victim compensation funds, admittedly not a top regulatory priority at present, would overwhelm limited agency resources which are already overtaxed by current enforcement efforts. Effective victim redress, therefore, requires more than simply dumping responsibility for fund administration on responsible agencies. Legislation providing for regulatory compensation systems must provide the
247
European Collective Redress
resources needed to administer those funds efficiently and effectively, including the authority to delegate, where necessary, to private firm. 170. The primary drawback of our proposal would appear to be in that small number of cases where the private bar has initiated the investigation and prosecution of the alleged violation. However, as the U.S. experience counsels, the vast majority of class actions “follow-on” governmental investigations, somewhat mitigating this concern.
REFERENCES Aitkins, P. S., & Bondi, B. J. (2008). Evaluating the mission: A critical review of the history and evolution of the SEC Enforcement Program. Fordham Journal of Corporate & Financial Law, 13, 367, 395396. Ashurst. (2004, August 31). Study on the conditions of claims for damages in case of infringement of E.C. competition rules. Retrieved from http://ec.europa.eu/competition/anti trust/actionsdamages/comparative_report_clean_en.pdf. Accessed on August 14, 2013. BDO Consulting. (2012, June 5). Independent Evaluation of the Gulf Coast Claims Facility Report of Findings & Observations to the U.S. Department of Justice 1114. Retrieved from http://www.justice.gov/opa/documents/gccfrpt-find-obs.pdf. Accessed on August 15, 2013. Berkowitz, E. (2006). The problematic role of the special master: Undermining the legitimacy of the September 11th Victim Compensation Fund. Yale Law & Policy Review, 24(Winter), 1, 14. Birnbaum, S. L. (Special Master). (2012, October). September 11th Victim Compensation Fund First Annual Status Report (“2012 Status Report”), Retrieved from http:// www.vcf.gov/pdf/VCFStatusReportOct2012.pdf. Accessed on August 15, 2013. Bromberg, A. R., & Lowenfels, L. D. (2011). Investor recovery as an enforcement objective. Bromberg & Lowenfels on securities fraud § 12:246 (2d ed. December), 5. Brueggeman, B. (2002). Hall of shame for announced class-action lawsuits in the country. Belleville News-Democrat, September 6. Brunet, E. (2003). Class action objectors: Extortionist free riders or fairness guarantors. University of Chicago Legal Forum, 2003, 403, 405. Centre for European Policy Studies. (2007, December 21). Making antitrust damages actions more effective in the EU: Welfare impact and potential scenarios, 51 et seq. Retrieved from http://ec.europa.eu/competition/antitrust/actionsdamages/files_white_paper/impact_ study.pdf. Accessed on October 13, 2013. Chen, D. W. (2004). Striking details in final report of 9/11 fund. New York Times, November 18. Deepwater Horizon Claims Center. (no date). Frequently asked questions no. 47. Retrieved from https://cert.gardencitygroup.com/dwh/fs/faq?.delloginType=faqs. Accessed on August 15, 2013. Department for Business Innovation and Skills. (2013, January). Private actions in competition law: A consultation on options for reform Government response. Retrieved from https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/ 70185/13-501-private-actions-in-competition-law-a-consultation-onoptions-for-reformgovernment-response1.pdf. Accessed on October 13, 2013.
248
ETHAN E. LITWIN AND MORGAN J. FEDER
Editorial (2010, August 26). The Gulf claims racket. Wall Street Journal. E.P. Legislative Observatory. (no date). Summary of E.P. resolution. Retrieved from http:// www.europarl.europa.eu/oeil/popups/summary.do?id=1189229&t=e&l=en. Accessed on August 15, 2013. Herr, M. E. (2012). PLI Corporate Compliance Answer Book, Q. 30.26.1. King, N., Jr. (2010, August 24). Feinberg criticized for spill-compensation terms. Wall Street Journal. Leslie, C. R. (2007). The significance of silence: Collective action problems and class action settlements. Florida Law Review, 59, 71, 9192. McDonald, J. P. (2008). Note, Milberg’s monopoly: Restoring honesty and competition to the plaintiffs’ bar. Duke Law Journal, 58 (December), 507. McLaughlin, J. M. (2011). McLaughlin on class actions: Law and practice (8th ed. pp. 23). West Publications. Mullinex, L. S. (2011). Prometheus unbound: The Gulf Coast claims facility as a means for resolving mass tort claims a fund too far. Louisiana Law Review, 71(Spring), 819, 833. September 11th Victim Compensation Fund. (no date). Frequently asked questions (“New Fund FAQ”) 1.5. Retrieved from http://www.vcf.gov/faq.html#gen2. Accessed on August 15, 2013. September 11th Victim Compensation Fund. (2002). Instructions Compensation form for deceased victims, 4. Retrieved from http://www.usdoj.gov/archive/victimcompensation/ deceasedvictims.pdf. Accessed on August 15, 2013. September 11th Victim Compensation Fund. (2002). Presumed economic loss calculation tables. Retrieved from http://www.usdoj.gov/archive/victimcompensation/vc_matrices. pdf. Accessed on August 15, 2013. September 11th Victim Compensation Fund. (2002). Victim compensation fund frequently asked questions § 9.1. Retrieved from http://www.usdoj.gov/archive/victimcompen sation/faq.html. Accessed on August 15, 2013. U.S. Government Accountability Office. (2007). Rep. No. GAO-07-830, SEC: Additional actions needed to ensure planned improvements address limitations in enforcement division operations 2. Retrieved from http://www.gao.gov/products/GAO-07-830. Accessed on August 15, 2013. U.S. Government Accountability Office. (2010). Rep. No. GAO-10-448R, SEC: Fair fund collections and distributions 3. Retrieved from http://www.gao.gov/assets/100/96667.pdf. Accessed on August 15, 2013. Vane, H. (2013, October 17). Brealey: UK to become “extremely friendly” for litigation, global competition review. Retrieved from http://globalcompetitionreview.com/news/article/ 34384/brealey-uk-become-extremely-friendly-litigation/?utm_medium=email&utm_ source=Law+Business+Research&utm_campaign=3217797_GCR+Headlines&dm_ i=1KSF,1WYV9,9GQ5QE,6VIK3,1. Accessed on October 18, 2013. Willging, T. E., Hooper, L. L. & Niemic, R. J. (1996). Empirical study of class actions in four federal district courts: Final report to the Advisory Committee on civil rules. Federal Judicial Center, 46, 63. Winship, V. (2008). Fair funds and the SEC’s compensation of injured investors. Florida Law Review, 60(December). Zimmerman, A. S. (2011). Distributing justice. New York University Law Review, 86(May), 500, 527.
CARTEL OVERCHARGES$ John M. Connor ABSTRACT Many jurisdictions fine illegal cartels using penalty guidelines that presume an arbitrary 10% overcharge. This article surveys more than 700 published economic studies and judicial decisions that contain 2,041 quantitative estimates of overcharges of hard-core cartels. The primary findings are: (1) the median average long-run overcharge for all types of cartels over all time periods is 23.0%; (2) the mean average is at least 49%; (3) overcharges reached their zenith in 18911945 and have trended downward ever since; (4) 6% of the cartel episodes are zero; (5) median overcharges of international-membership cartels are 38% higher than those of domestic cartels; (6) convicted cartels are on average 19% more effective at raising prices as unpunished cartels; (7) bid-rigging conduct displays 25% lower markups than price-fixing cartels; (8) contemporary cartels targeted by class actions have higher overcharges; and (9) when cartels operate at peak effectiveness, price
$
The author is Professor Emeritus at Purdue University, West Lafayette, IN. He is indebted to Professor Robert H. Lande, who worked with the author on earlier law review articles on cartel overcharges; he also was responsible for locating several overcharges from antitrust verdicts in U.S. courts and provided meticulous comments on this version.
The Law and Economics of Class Actions Research in Law and Economics, Volume 26, 249387 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0193-5895/doi:10.1108/S0193-589520140000026008
249
250
JOHN M. CONNOR
changes are 6080% higher than the whole episode. Historical penalty guidelines aimed at optimally deterring cartels are likely to be too low. Keywords: Cartel; collusion; price fixing; overcharge; antitrust; optimal deterrence JEL classifications: L12; L42; K22; B14; F29
TABLE OF CONTENTS Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Overcharge Defined . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Overcharge Rates Computed . . . . . . . . . . . . . . . . . . . . . . . . . . Overcharges are Important in Economics and the Law . . . . . . . Overcharges and Cartel Fines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The United States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Other Jurisdictions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Overcharges and Cartel Deterrence . . . . . . . . . . . . . . . . . . . . . . . . . Literature Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Early Cartel Studies in Brief . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . PostWorld War II Cartel Studies . . . . . . . . . . . . . . . . . . . . . . . . . Quantitative Estimates of Cartel Overcharges . . . . . . . . . . . . . . . . . Surveys of Cartel Price Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . General Description of the Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . Number of Cartelized Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . Number of Episodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Number of Episodic Overcharges . . . . . . . . . . . . . . . . . . . . . . . . . . Defining Time Periods of Analysis . . . . . . . . . . . . . . . . . . . . . . . . . Numbers of Episodic Overcharge Estimates over Time . . . . . . . . . . Data Reliability Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Analysis of the Overcharge Results . . . . . . . . . . . . . . . . . . . . . . . . . . . Number of Overcharge Observations . . . . . . . . . . . . . . . . . . . . . . . Height of Episodic Overcharges over Time and by Type . . . . . . . . Overcharges over Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . International-Membership Cartels . . . . . . . . . . . . . . . . . . . . . .
251 252 252 253 255 257 257 258 259 261 262 263 266 267 268 269 272 273 275 277 285 287 287 288 290 293
Cartel Overcharges
251
Bid-Rigging Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sanctioned versus Unsanctioned Cartels . . . . . . . . . . . . . . . . . . Buyers’ Cartels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Market Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Unsuccessful Cartel Episodes . . . . . . . . . . . . . . . . . . . . . . . . . . Size Distribution of Overcharges . . . . . . . . . . . . . . . . . . . . . . . . . . Looking in Detail at Extreme Observations . . . . . . . . . . . . . . . . . . Peak Overcharges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Overcharges by Location of Cartel . . . . . . . . . . . . . . . . . . . . . . . . . Geographic Spread and Price Effects . . . . . . . . . . . . . . . . . . . . . . Overcharges and Duration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Overcharges and Market Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Decisions of Antitrust Authorities . . . . . . . . . . . . . . . . . . . . . . . . . . . Economists versus Courts or Commissions . . . . . . . . . . . . . . . . . . . Cartels Targeted by Class Actions . . . . . . . . . . . . . . . . . . . . . . . . . Summary and Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Implications for Economics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Implications for Public Policy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix: Data Sources and Collection Methods . . . . . . . . . . . . . . . . Selection Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Social Science Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Episodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Decisions of Antitrust Authorities . . . . . . . . . . . . . . . . . . . . . . . . . Laboratory Market Experiments . . . . . . . . . . . . . . . . . . . . . . . . . .
295 295 296 297 298 298 299 304 306 309 310 312 312 312 316 316 316 317 319 340 340 375 375 377 381 383 384
INTRODUCTION For at least 125 years, hundreds of economists, historians, commissioners, and jurists have labored mightily to assess the effectiveness of cartels.1 Several criteria have been devised to assess effectiveness, including longevity, stability, efficiency, and profitability, but chief among them is market price effects.2 The particular price effect of interest in cartel studies is the increase in selling prices3 caused by the collusive conduct of suppliers in a market.
252
JOHN M. CONNOR
Objective The principal purpose of this article is to assemble and analyze the most comprehensive collection of quantitative estimates of monopoly overcharges generated by private, hard-core cartels.4 Candidates are cartels that operated in all geographic locations of the world and in all historical eras. The estimates are assembled from serious published social science studies by disinterested authors and from the decisions of competent judicial bodies (see Appendix). Although the sources met minimal quality standards, no effort was made to apply additional subjective quality filters during the collection phase. Later, however, the estimates were examined for systematic differences in reliability across types of sources or methods of calculating overcharges. Analysis in this article is limited to descriptive tabulations using categories that have been shown to be significantly different by more formal analyses. However, this article attempts to convey its findings in a style approachable by practitioners and policy makers who may not be professional economists. Compilation should serve two subsidiary concerns. First, the results of the survey can be used as benchmarks to assess the ability of current antitrust penalties to deter illegal cartels. Second, these data may demonstrate empirical regularities that may suggest hypotheses for formal economic model-building.
Overcharge Defined The increase in purchase costs to buyers due to an effective sellers’ cartel is commonly called an overcharge by economists and legal writers.5 When multiplied by the quantity sold be a cartel, it becomes the major portion of cartel damages.6 The overcharge rate is calculated by comparing actual cartel-enhanced prices to an appropriate noncollusive benchmark price7 (Connor, 2008). To be precise, if a sellers’ cartel is effective in raising the market price Pm for a period of time because of collusion, then the unit monetary overcharge is Pm − Pc, where Pc is the competitive or benchmark price that would have been observed in the market absent overt collusion. Given the quantity sold during the conspiracy (Qm), the total overcharge is: Dollar overcharge = ðPm − Pc Þ × Qm
253
Cartel Overcharges
The price difference Pm − Pc is conventionally converted to a rate (a ratio or percentage) by dividing the price wedge by the benchmark price. That is, Overcharge rate = ðPm − Pc Þ=Pc The overcharge can in theory range from zero to infinity, though the latter is highly unlikely. If Pc is properly measured, an overcharge of 0% would imply that the cartel was ineffective in controlling market price and that buyers from the cartel had suffered no antitrust injuries. Overcharge Rates Computed There are a couple of reasons why overcharge ratios may be systematically under-reported. First, commentators can err, even when the dollar overcharge and the affected sales are known precisely. Calculating an overcharge rate is straightforward when working with prices, but converting a monetary overcharge into a percentage overcharge can easily lead to an underestimate of the overcharge rate.8 Let us examine a specific overcharge calculation. In 19921995, the world’s five producers of synthesized lysine (an amino acid that accelerates the growth of muscle tissue in animals) conspired to raise its global price. In the U.S. market, the cartel obtained a dollar overcharge of $80 million on sales of $460 million (Connor, 2007b, pp. 200, 220235). Most observers would readily infer that the overcharge rate is (80/460) × 100 = 17.4%. This is the method commonly followed by counsel when reporting how well they have represented their clients. However, the appropriate calculation is more complicated. It involves dividing the overcharge by the competitive or but-for sales, not the actual (affected) sales.9 The correct formula is: Overcharge rate = ðPm − Pc Þ × Qm =ðPm × Qm ð1 − ððPm − Pc Þ × Qm =Pm × Qm ÞÞÞ So, in the lysine example, the divisor ought to be competitive rather than overcharge-inflated affected sales. That is, the proper divisor is $460 − $80 = $380 million, and the true overcharge rate is (80/380) × 100 = 21.1%. Note that when working with prices, under-reporting overcharge rates should not be an issue. The average monthly prices were about $0.945 and the but-for price about $0.78, which also yields an overcharge of 21%. A recent example of using the wrong denominator to calculate the overcharge rate can be seen in a widely read report commissioned by the
254
JOHN M. CONNOR
European Commission from a respected consultancy: prices of an Austrian cartel fell from h1140 during collusion to h900 after a raid; the report computes the overcharge to be 22%, whereas the correct overcharge is 26.7% (Komninos et al., 2009, p. 53). A second cause of low-reported overcharge rates is under-reporting of affected sales (see Box).
The Iowa Ready-Mix Concrete Antitrust Case An order handed down by U.S. District Court Judge Mark W. Bennett in this case contains the following information: The combined settlement fund of $18.5 million is sufficient to repay completely each class member’s actual overcharge damages … even after fees and costs … [which is] “very unusual” in an antitrust class action… . The $18.5 million sum is especially remarkable, given that the United States Department of Justice estimated that the total volume of commerce affected by the price fixing conspiracies was only $5,666,348.61. (Bennett, 2011, p. 4)
Later in this decision we learn that the settlement fund includes $7,638,113 in fees and costs, which implies that the overcharges were $10,861,887. To compute the rate, the first impulse of counsel is to take the overcharges ($10,861,887) and divide them by the sales during the collusive period ($5,670,000). The result is 191.7%. Hold on: This is an impossible number! The overcharges to direct buyers cannot exceed 100% of the value of their purchases. Overcharges can exceed competitive sales but not affected sales. The solution to this conundrum is that the affected sales mentioned in the judicial order are far too small. The Judge was quoting from affected sales calculated by the U.S. Department of Justice for its criminal prosecution in which it counted only projects within 15 miles of each concrete plant. Private litigants used a more expansive approach to identifying geographic market boundaries, which resulted in more logical affected sales above $18.5 million.
Under-reporting of cartel sales is a common practice by antitrust authorities. One reason for this tendency is that authorities in many jurisdictions must defend their decisions when the alleged cartelists appeal their fines to a higher court. Because fines are directly, positively related to affected sales,
Cartel Overcharges
255
the authorities customarily (1) cut down the list of products that probably were cartelized to list only products that were incontrovertibly subject to price fixing, (2) exclude regions within the jurisdictions that arguably were subject to price fixing, and (3) foreshorten the collusive time period either because early-period written documents are incomplete or because there is possibly contradictory testimony by cartel managers concerning start or ending dates. Lengthy appeals over imposed fines are common in the EU, Brazil, and many other legal systems. Appeals are also possible when defendants go to trial in criminal antitrust regimes. In criminal jurisdictions like the United States, cartel fines are also linked to the size of affected sales, but the size of fines is usually the result of guilty-plea negotiations; the resulting agreements cannot be appealed. However, prosecutors have incentives to carve down affected sales during plea negotiations in order to avoid the risk of trials, where the standard of proof is “beyond a reasonable doubt.” Concessions may be and are offered to defendants about which products, geographic regions, and time periods to include in affected sales (or the degree of harm caused). For example, a plea agreement may state that price fixing began “… as early as May 1, 2000,” when in fact collusion is later proven to have begun much earlier.10 Overcharges are Important in Economics and the Law A price-fixing overcharge is a transfer of income or wealth from buyers to the members of the cartel that occurs as a result of an overt collusive agreement.11 Ceteris paribus when a cartel achieves high levels of effectiveness (i.e., longevity, stability, and high overcharge rates), it tends to generate large customer losses, i.e., measurable reductions in consumer surplus.12 Although there are other economic effects of price fixing, legal-economic scholarship on antitrust injuries tends to focus on the overcharge.13 Effective cartels are also viewed as destructive of the competitive process in the sense that they weaken the natural effects of demand and supply in price formation and cause deadweight social losses.14 That is, effective cartels cause economic, allocative inefficiency. The deadweight losses result from the costs incurred by customers when they are forced to substitute inferior substitutes, if any, the costs incurred by the members of the cartel in managing the collusive enterprise, and rent-seeking behavior by the cartel such as efforts directed at forestalling entry. “Umbrella pricing” or “free riding,” the tendency of suppliers outside the cartel to sell at the cartel’s elevated price, creates further harm for customers of fringe suppliers. In this article, I focus on cartel overcharge rates as the root indicators of the many harms or damages created by price fixing.
256
JOHN M. CONNOR
Direct purchasers from an effective sellers’ cartel are the immediate losers. However, if the cartel is comprised of manufacturers (the most common story), then other buyers farther down the distribution channel are also harmed. These indirect purchasers typically will be other manufacturers, wholesale distributors, retail distributors, and the final consumers of the cartelized product.15 Indirect buyers pass on part of or all of the overcharge contained in the direct purchase. Under simplifying assumptions, indirect purchasers in perfectly competitive industries pass on 100% of the initial overcharge, but if the indirect buyer is a monopolist then only 50% will be passed on at any one stage.16 If all the distributors use percentage markup rules, a fairly common situation, then the consumer pass-through rate is 100%. If the cartelized product is highly differentiated, then the pass-on rate will exceed 100%. Until about 1990, scholarly literature surveys of the economics of cartels seldom addressed overcharges, but interest in this subject has blossomed in the past decade. For example, Levenstein and Suslow (2006, §6.1), while focusing their article on duration, examine 8 cross-industry and 54 “selected” case studies of cartels in 19 industries for evidence about price or profit effects.17 They conclude that (1) almost half of the industry case studies do not address the issue, (2) when addressed, nearly all find at least short-run price changes due to cartelization, but (3) few of the latter are explicit about the counterfactual (i.e., the but-for price) (ibid., pp. 8182). Today textbooks of economics conventionally devote considerable space to the market price effects of cartels.18 While empirical studies of cartels routinely survey selected antecedents as a prelude to the study being presented, to my knowledge no one has published a work aimed principally at comprehensively surveying and analyzing cartel overcharges.19 This article is aimed principally at filling this gap in the legal-economic literature. The actual size of cartel overcharges is an issue at the heart of a number of legal and economic controversies. First, knowing the size and distribution of cartel overcharges is necessary to justify the underpinnings of antitrust authorities’ guidelines for sanctioning illegal cartel conduct. Many commentators on government fining practices have noted the absence of appropriate empirical data for the rational design of such policies. Second, because the typical harm from cartel operations was mainly anecdotal, there are widely varying opinions among experts on the critical issue of the size of sanctions needed for optimal deterrence of cartel formation.20 The following sections discuss these issues.
257
Cartel Overcharges
Overcharges and Cartel Fines The United States The Sentencing Reform Act of 1984 created the U.S. Sentencing Commission (USSC), a judicial-branch unit charged by the U.S. Congress with devising guidelines for criminal sentencing for the federal judiciary (USSG Advisory Group, 2003). The first set of guidelines was promulgated in 1987, and after public comment was made law in 1989. The guidelines included sanctions for organizations guilty of horizontal price fixing and bid rigging (Cohen & Scheffman, 1989, p. 332). Although the Sherman Act of 1890 is a criminal statute that encompasses other types of restrictive business practices, by long tradition only horizontal price-fixing and market-sharing agreements have triggered criminal indictments by the Department of Justice (DOJ).21 The issue of how high cartels typically raise prices was crucial when the USSC established the fine levels for cartel violations. The USSC’s formulas for calculating cartel fines follow from an embedded assumption: “It is estimated that the average gain from price-fixing is 10 percent of the selling price.”22 The Commission added: “The purpose for specifying a percent of the volume of commerce is to avoid the time and expense that would be required for the court to determine actual gain or loss.”23 As the Sixth Circuit noted, the Sentencing Commission “opted for greater administrative convenience instead of undertaking a specific inquiry into the actual loss in each case.”24 The USSC appears to have adopted the 10% presumption because its use was advocated by the then-head of the Antitrust Division, Douglas Ginsburg.25 The origin of Ginsburg’s 10% figure is not publicly known. However, a prominent analysis of the issue by Cohen and Scheffman (1989), published shortly after the antitrust sentencing Guidelines were promulgated, asserts that the economic evaluation of only three price-fixing conspiracies was particularly important in shaping Ginsburg’s views. It says further that “… there is little credible statistical evidence that would justify the Commission’s assumptions which underlie the Antitrust Guidelines” (p. 333). If this analysis is correct, a critical assumption in setting cartel penalties in the United States is supported by a surprisingly small amount of evidence. In the history of antitrust before 1990, the sum of all cartel penalties amounted to less than $100 million (Gallo, Dau-Schmidt, Craycraft, & Parker, 1994).26 From 1990, a series of record corporate fines and other penalties were imposed for criminal price fixing by U.S. courts, most of
258
JOHN M. CONNOR
which were prosecutions of international cartels (Connor, 2011c). A similar upswing may be noted for fines imposed by the European Commission, the EU’s Member States, and a few antitrust authorities in Asia, Africa, and Latin America. By 2010, U.S. and EU governments and private monetary penalties amounted to at least $84 billion (p. 31). In early 2012, worldwide cartel penalties surpassed $100 billion. This figure does not include legal fees, corporate reputational effects, or penal sanctions. The consensus of scholars is that current antitrust regimes are underdeterring price fixing (Connor & Lande, 2012; Ginsberg & Wright, 2010; Harrington, 2012). However, some attorneys engaged in defending alleged international price-fixing conspiracies have argued that the Guidelines have resulted in excessive penalties. For example, just as the DOJ’s campaign against international cartels was gathering steam, Adler and Laing (1997) assert that “the fines being imposed against corporate members of international cartels are staggering” (p. 1), placing the blame on the “uniquely punitive” requirements of the U.S. Sentencing Guidelines.27 Denger (2003) too decries the prevalence of excessive price-fixing fines and private settlements. He places the blame for excessive fines on the Corporate Guidelines base fine calculation (p. 3). This approach, he notes, unlike all other white-collar federal crimes, means that the actual degree of direct harm caused does not have to be proven by prosecutors.28 Denger blames this state of affairs on a gap in the economic-legal literature: “… we have little information on what level of criminal or civil exposure is needed to deter most cartels” (p. 4). Concern about the lack of empirical evidence on the size of overcharges caused by price fixing is not confined solely to those sympathetic to the increased exposure of corporate defendants. DOJ official Graubert (2003) notes that the controversy over whether antitrust payments are excessive is largely attributable to the “… difficulty of gathering useful data.” In a law review article noting the sharp increase in U.S. criminal fines on international cartels in the late 1990s, Klawiter (2001) believes that these fines and other related antitrust penalties “… have substantially increased the level of deterrence in antitrust criminal cases” (ibid., p. 756).29 Yet, he laments the paucity of information needed to make a more sweeping conclusion. “There are no known applicable empirical studies on the adequacy of the present mix of criminal and civil antitrust sanctions from the standpoint of deterrence” (ibid., note 79). Other Jurisdictions The U.S. antitrust enforcement has been a model for many other countries that have more recently adopted such laws (Wells, 2002). Germany and
259
Cartel Overcharges
Japan had antitrust laws imposed on them by the U.S. occupation authorities in the late 1940s.30 After a vigorous debate, Germany revised its competition law in 1958; it, in turn, became one of the principal influences on the adoption of a similar statute by the original six members of the European Economic Community (Goyder, 1998, pp. 1833). After four years of confidential political discussions31 within the EEC’s Commission, Regulation 17 was passed in 1962; it lays out the powers of the Directorate General for Competition (DG-COMP) to fine companies for competitionlaw infringements (ibid., p. 45). That rule sets a maximum corporate fine of 10% of the company’s total sales in the year prior to the Commission’s decision and specifies that the specific fine will depend on the duration and seriousness of the offense.32 Harding and Joshua (2003) state that EC fines are supposed to incorporate both compensatory and punitive components, the latter meant to serve deterrence (p. 240). Methods of calculating EC cartel fines are explained in 1998 and 2006 Notices (Connor, 2010). Under the earlier guidelines, EC cartel fines were loosely related to overcharges because cartels with large damages that are geographically widespread and relatively large companies were given larger fines. Since late 2006, EC fines have been tied to affected sales in the EU, and they have become more severe. After considering a number of culpability factors, the Commission ensures that the fine does not exceed 10% of a defendant’s global sales in the year prior to the date of the decision. Rarely does the EC need to worry about reaching the 10% cap. Canada is another jurisdiction with relatively tough sentencing for cartels. The Canadian Competition Bureau (CCB) uses a fairly simple standard for setting fines. Although not spelled out in any administrative guidelines, decisions of Canadian courts have, in the absence of aggravating and mitigating circumstances, imposed fines hew closely to 20% of Canadian affected sales (Connor, 2003; Low, 2004).33 A former Canadian prosecutor comments that “there has not been any economic or judicial analysis of the assumptions behind this proxy for harm that this represents …” (Low, 2004, p. 19). The Canadian 20% rule seems to mimic the base fine of the USSGs.
Overcharges and Cartel Deterrence Concerns about the inadequacy or excessiveness of antitrust sanctions are part of the larger issue of the effectiveness of antitrust interventions. Most legal scholars accept that the fundamental objective of price-fixing laws is
260
JOHN M. CONNOR
deterrence: that is, to minimize the future formation of new cartels or recidivism by previous cartel violators. To make any headway in assessing empirically the adequacy of anticartel enforcement, analysts must have reliable information about the degree of harm generated by private cartels. Antitrust sanctions should be calibrated to cartels’ overcharges. Total cartel injuries to purchasers are positively related to three economic factors: the size of the cartel’s market, the duration of the conspiracy, and the percentage overcharge. Cartel deterrence can also be affected by other enforcement rules. Amnesty programs and general investigatory procedures can increase the probability of cartel detection or reduce the duration of cartels. The sentencing guidelines developed in the United States, the EU, and elsewhere for fining hard-core cartels are consistent with the optimal deterrence standard first suggested in a seminal article by Becker (1968) and elaborated by William Landes (1983). Landes showed that to achieve optimal deterrence, the damages from an antitrust violation should be equal to the violation’s expected “net harm to others,” divided by the probability of detection and proof34 (Landes, 1983, pp. 666668). Critics of the U.S. Sentencing Guidelines suggest that their assumed average overcharges are too high. For example, Cohen and Scheffman (1989) argue that fines based on the USSGs, when coupled with civil and marketplace sanctions will cause “a serious overdeterrence problem” (p. 334). That is, they and other critics of the Guidelines believe that there is a disparity between the size of the corporate fines mandated35 for antitrust violations and the amount of the economic injuries caused by overt price fixing. Specifically, Cohen and Scheffman argue that actual overcharges are well below the 10% level assumed in the Guidelines (pp. 343347).36 During recent years, their criticism has been repeated with perhaps even more intensity. For example, in a provocative essay that quickly drew rebuttals,37 Crandall and Winston (2003) argue that extant empirical evidence demonstrates that U.S. antitrust policy has been ineffective in deterring anticompetitive conduct. To support their view that the prosecution of overt price fixing is misdirected, they cite five empirical studies of overt collusion that find no upward effects on prices of conspiracies convicted in U.S. courts. In his comment on Crandall and Winston, Kwoka (2003) faults them for their “startlingly selective” body of evidence. He suggests that they should have included “… studies from any source with appropriate evaluation of their credibility” (p. 4). There are few empirical studies of cartel deterrence. Even the most ambitious have focused on strictly national data (Connor & Lande, 2012).
261
Cartel Overcharges
Yet since about 1995, a large majority of the overcharges generated by cartels have been international in membership and global in their geographic impact (Connor, 2001a, 2003, 2008). To assess the likelihood of deterrence in the context of international schemes, worldwide monetary sanctions must be considered. A previous analysis by Connor (2012, Figs. 8 and 9) summarizes a large data set on the severity of penalties on global cartels during 19902010.38 He finds that total monetary penalties worldwide average about 11% of affected sales (higher in North America and the EU, lower elsewhere). Penalties disgorge at most 40% of the worldwide overcharges generated.39 Given that the odds of being caught are less than 100%, optimal deterrence requires cartel sanctions to be somewhat punitive. That is, disgorgement must exceed 100% of overcharges. Because it does not, punitive sanctions are the exception not the rule for illegal international price fixing. Clearly, information on both damages and penalties are needed on a worldwide basis. In sum, there does indeed seem to be a broad consensus among legal and economic writers that the question of the optimality of price-fixing penalties turns mightily on the actual degree of harm caused by cartel conduct, and that not enough is known about this issue. Moreover, even if the creators of the USSC Guidelines were correct that in the 1980s cartels generally raised prices by 10%, the harsher cartel sanctions imposed more recently could mean that this presumption is no longer justified. The contents of this article could provide a factual foundation for dialogs on optimal deterrence and rational anticartel policies.
LITERATURE OVERVIEW40 This article was prepared by examining approximately 1500 social science publications and legal documents.41 Of these, 524 contained usable quantitative overcharge estimates.42 The major portion of the overcharge estimates included in this article is taken from books, book chapters, conference proceedings, or papers published in economic, historical, and legal journals whose readers and contributors are mainly academics. The great majority of these publications are peer reviewed. A minority of the estimates are taken indirectly from newspapers, magazines, and similar journalistic outlets; from reports issued by governments; from academic working papers; and from decisions rendered by courts or antitrust commissions. This section focuses on the evolution of social science concepts about cartels and their price effects.
262
JOHN M. CONNOR
Early Cartel Studies in Brief Adam Smith (1776) has a claim as a founder of industrial-organization economics. He explicitly examined business collusion, which he called “a conspiracy against the public.” From 1880 to 1920, there was vigorous debate in the economics profession over public policies to address market power, market regulations, and the “trust problem” (Martin, 2005, p. 5). However, these discussions were hampered by the nearly exclusive reliance of the economics profession on the models of pure competition and monopoly.43 What changed in the 1930s was the development, slow at first, of conceptual models of oligopoly (ibid., pp. 611).44 At that point, the subfield of industrial economics was born and flourished. Cartel studies spent 70 years being practiced before it had a name. The empirical economics literature on cartels up to the 1940s is characterized by a groping toward a conceptual understanding of the nature of private cartels and the first tentative steps toward quantitative evaluation of the market effects of overt collusion. Formal economic studies of cartels began in Germany in the 1870s; books and articles written in German continued to dominate the literature up to the 1920s. Among German scholars, the ideas of Smith, Ricardo, and the other classical economists spread only slowly during the early 19th century (Gerber, 1998, pp. 8188). While the core concepts of classical economics continued to be accepted, during the late 19th century, the “historical school” came to dominate the scholarship of German academic economists. The historical school emphasized the importance of unique temporal and institutional factors in explaining empirical phenomena; it consciously rejected abstract theories as a guide to empirical studies. Cartels were usually seen as an inevitable response to historical overproduction. Despite their understanding of the monopolistic tendencies of cartels, evaluation of cartels was almost solely from the producers’ perspective rather than consumers’ interests. Especially influential was the German economist Liefmann (1897, 1932). His concept of a cartel as a voluntary, contractual association of independent firms intent on profit maximization45 and monopolistic control of a market became the accepted definition. An unfortunate legacy of the German historical school of cartel studies was its view that gauging price effects was either fruitless or impossible, a presumption that discouraged Continental European economists from attempting to estimate overcharges until the late 20th century.46 However,
263
Cartel Overcharges
U.S. social scientists inherited a more pragmatic tradition driven by an awareness of the country’s new antitrust law, which was passed in 1890 after a long debate that highlighted the negative effects of cartels on small businesses. Court decisions interpreting the Sherman Act in the early 1900s stimulated further scholarship on cartels. As a result, most quantitative estimates of overcharges made prior to 1945 were produced largely by American social scientists.47 Some highlights include Jenks’ (1888) path-breaking analysis of the Midwest salt cartel; Jones’ (1914) book on the anthracite coal industry; Edgerton’s (1897) superb analysis of price effects of a short-lived but highly effective international cartel, the U.S. Wire Nail Association; Andrews’ (1889) sketch of what is quite possibly the world’s first global cartel, the Secre´tan copper syndicate of 18871889; and Stevens’ (1912) study of the convicted Gunpowder Trust, notable for focusing on what was believed to be the longest-running discovered cartel in the Nation’s history (it lasted 35 years, of which 17 were illegal).48 In the decade after World War I, hundreds of cartels were established (or reestablished) in a wide range of commodities and industrial products, gaining control if nearly half of world trade in the 1930s. Nearly all of them operated in the open. Contemporary scholars now regard the interwar era as something of a Golden Age of Cartels. Yet, exceedingly little published work by professional economists dates from this era.
PostWorld War II Cartel Studies During and immediately after World War II, a surge in publications examined the roles of cartels in international trade and in war production. Hexner (1946) produced the most comprehensive economic study of international cartels yet published. Hexner had an insider’s knowledge of cartels (Barjot, 1994, p. 65). Marlio (1947), a French economist, who wrote a detailed account of the international aluminum cartel, was also a cartel manager (ibid., p. 66). Both of these authors found much to admire in international cartels, whereas postwar works by American authors tended to be distinctly more skeptical, if not hostile regarding the economic and political effects of the interwar cartels (e.g., Berge, 1944; Edwards, 1946). Perhaps the first publications to attempt to quantify systematically the price effects of cartels were a pair of books produced by a team of economists that had access to information handed over to investigators of
264
JOHN M. CONNOR
Congressional committees and to criminal court proceedings (Stocking & Watkins, 1946, 1948).49 These books were the culmination of eight years of study by a team of economists.50 They set a new intellectual standard for the economics literature on cartels, because they were the first to apply rigorous modern concepts of the emerging field of industrial economics; because of access to ample quantitative information spawned by numerous Congressional investigations, the Federal Trade Commission, and law suits; and because they were among the first to focus on the market effects of international cartels.51 Numerous and continuing citations to their books by leading contemporary scholars attest to their status as seminal works and classics in the field (Mueller, 2007, p. 188). The increasing evidence of negative impacts of cartels during 19201945 began to bring about a reappraisal of cartels among Europeans just after World War II. In Germany, there was a healthy parliamentary debate over its cartel laws in 19511957 (Gerber, 1998, pp. 270277; Wells, 2002, pp. 165174). Through the early 1950s, a majority of the UK’s manufacturing output was affected by cartels (Swann et al., 1974; Symeonidis, 2002). A long series of empirical studies by the Monopolies Commission investigated the structure and performance of British industries and made recommendations to the government about restrictive practices, dominant firms, and mergers.52 By the late 1950s, anticartel legislation was adopted in the United Kingdom that placed the burden of proof on cartels to prove the economic benefits of their price fixing and related conduct. Germany was the prime mover behind the adoption of tough anticartel provisions in the Treaty of Rome, which solidified the antitrust tradition in the EU and its Member States. In the second half of the 20th century, relatively few books were written about the empirical economics of cartels, but there have been three brief periods of interest. First, there was intense but short-lived U.S. attention to domestic cartels when the “Great Electrical Equipment Conspiracy” burst onto the Nation’s consciousness in 19601961.53 The great electrical equipment conspiracy resulted in the release of more publications in a few years than any other single historical event since the beginning of cartel literature. The scope of the conspiracy, the fame of the leading companies involved, and the U.S. Government’s aggressive prosecution of the violators all these factors led to a degree of public fascination and publicity about an antitrust action not seen since the Supreme Court decisions against the Standard Oil and American Tobacco trusts in 1911.54 Several trials provided unusually detailed pictures of the cartel’s organization.55
Cartel Overcharges
265
This cartel has become a standard example in textbooks in industrial organization (e.g., Carlton & Perloff, 1990, 2005). Second, there was a brief revival of interest in international cartels after 1973 when the Organization of Petroleum Exporting Countries (OPEC) first used its power to raise crude petroleum prices.56 Some economic studies tried to predict OPEC’s staying power by studying previous international cartels.57 Griffin (1989), who has several cartel studies to his credit, specifies a formal cartel model which allows for a fringe of competitive, noncooperating producers outside the cartel. From this theoretical model, Griffin derives a simple empirical model that explains variation in the Lerner Index58 of market power for a large sample of cartels. Third, scholarship seems to have been stimulated by the large number of well-publicized U.S. and EU prosecutions of global cartels that commenced in the mid-1990s. Many of these cartels were organized by some of the world’s most recognizable multinational companies. The first global case in decades in both jurisdictions was Lysine, which was capped in the United States by a notorious 1998 criminal trial of three executives of the Archer Danieal Midlands Co. The trial record provided a degree of testimonial evidence which is unique for international cartels discovered after World War II (Connor, 2007b; Eichenwald, 2000; Lieber, 2000). EC decisions have become major sources of information about contemporary cartel conduct (Harrington, 2007). After about 1973, many empirical analyses of cartel effects began to appear in professional academic journals. The shift away from monographs to journal papers is remarkable. Of the 125 journal papers in this survey with useful overcharge information, 88% were published after 1973.59 While a few are historical narratives, the later articles tend to focus on statistical tests of theoretical hypotheses or demonstrations of the superiority of a novel estimation technique. In general, these journal papers supplied only about one-fifth of the estimates in the vast literature in economics that measures the price effects of cartels. It is small because external information is needed to identify markets in which sellers overtly colluded from the much larger number of markets characterized by presumptively tacit collusion. These papers for the most part depend heavily on statistical methods of analysis. Around the early 1970s, statistical methods started to become standard for proving cartel damages (Finkelstein & Levanbach, 1983). Other important sources of scores of overcharge estimates are the decisions of courts and competition-law commissions, most published since 1990.
266
JOHN M. CONNOR
Quantitative Estimates of Cartel Overcharges Most cartel studies published in academic journals since 1974 use econometric methods to estimate overcharges. The first published work that uses econometrics to estimate a cartel overcharge is Sultan’s (1974) analysis of the U.S. electrical equipment conspiracy of the 1950s.60 Fisher (1980) and Finkelstein and Levenbach (1983) show that econometric evidence of price fixing was being presented by experts in U.S. civil trials as early as 1970. Econometric evidence on monopoly overcharges was also published to critique government-enforced compulsory cartels; Kwoka (1977) is the first of many analyses of the price effects of agricultural marketing orders. However, quantitative analyses of the size of buyers’ cartels’ undercharges are rare; Daggett and Freedman (1985) seem to have been the first to publish such a study. Sophisticated econometric modeling has spread into historical studies of cartels: a notable pair of studies by Hausman (1980, 1984) examines two UK coal markets from 1699 to 1845, and Levenstein (1997) analyzes the century-old bromine cartel. Genesove and Mullin (2001) is a rare example of a widely cited cartel study in a leading journal that does not employ statistics. A new development in the cartel literature was the statistical analysis of auctions and bid rigging, much of it inspired by the urge to test gametheoretic notions (Porter, 2001 surveys this literature). Howard and Kaserman (1989) study collusion in public tenders for sewer construction; Froeb, Koyak, and Werden (1993) federal-government procurement of frozen fish; Brannman and Klein (1992) state road-building contracts; and Lee (1999), Porter and Zona (1999), and Pesendorfer (2000) school milk procurement. These studies were made possible by U.S. “freedom-ofinformation” laws that mandate public access to bids for public project tenders. Although such laws exist outside the United States, few have been used there to obtain data on bid rigging of public tenders. Novel methods continue to be applied to estimating cartel markups. There is substantial work focused on understanding cartel stability from which price effects can be derived. Grossman (1996) looked at the 18511913 railroad express delivery market, and several have studied the 19th-century Joint Economic Committee railroad cartel (Briggs, 1992; Ellison, 1994; Porter, 1983). Bajari and Ye (2003) applied the Bayesian statistical method to a U.S. seal-coating conspiracy. Clarke and Evenett (2003) apply a trade model to importing countries to estimate price increases during the 1990’s bulk vitamins cartel. Dynamic estimation methods have begun to yield insights into cartel conduct (e.g., de Roos, 2006).
267
Cartel Overcharges
Surveys of Cartel Price Effects Given the importance of the topic for legal-economic discourse, there have been surprisingly few compilations of the empirical findings of cartel overcharges. Economics textbooks devote limited space to the subject.61 I have been unable to find any research publication that has as its principal aim collecting or analyzing information on the price effects of overt collusion. However, I have found seven works that mention a significant number of studies of markups due to overt collusion. The overcharges are assembled as a prelude to scholarly research or policy analysis, not as an end in itself; none claims to be a comprehensive survey. The seven brief surveys are summarized in Table 1. They report on overcharges from 127 to 129 cartels, most of which operated in the interwar era. The median average markup is 27.1% and the mean average is 32.4%.62
Table 1.
Summary of Seven Economic Surveys of Cartel Overcharges.
Reference
1. Cohen and Scheffman (1989) 2. Werden (2003) 3. Posner (2001) 4. Levenstein and Suslow (2002) 5. Griffin (1989) 6. OECD (2003), excluding peaksa 7. Davies and Majumdar (2002)b Total, simple average of the seven entries Total, weighted average of the seven entries
Number of Cartels
Episodic Overcharge Mean (%)
Median (%)
57 13 12 22 39 13 23
7.710.8 21 49 43 28.0c 21.6 24.933.9
14.0 18 38 44.5 28.7c 14.0 2025
127129 127129
32.4 38.1
27.1 31.0
a One overcharge in the OECD survey with missing affected sales (U.S. lysine) was converted to percentages using affected sales data in a published U.S. court decision. One overcharge reported to be “more than 13%” was recorded as 14%. If a range, the midpoint is used for averaging. Three percentages cited to be “as high as” were omitted because they are not likely to be representative of the overcharge rate for the whole episode. The OECD report states that its sample median is “between 15% and 20%.” b The present author did not discover this estimable survey until 2011, perhaps because of its title. c Because one does not know what the benchmark prices are for these observations, I show average Lerner Indexes. If the benchmark is perfect competition, the mean and median overcharges would be higher, 53.2% and 38.9%, respectively.
268
JOHN M. CONNOR
GENERAL DESCRIPTION OF THE SAMPLE Technically, the observed cartel overcharges collected for this article are a sample of a larger population of cartel overcharges, both seen and unseen. The unobserved overcharges are the vast majority of the total for two reasons. First, since about the middle of the 20th century (and earlier in the United States) most cartels are clandestine. The great body of expert opinion is that in the past few decades fewer than one-third of all cartels are discovered by antitrust authorities (Connor & Lande, 2012, Table 1). Second, among those cartels that never hid themselves or that were discovered by antitrust authorities, sufficient price data were unavailable (or of no interest to the writer) for roughly half or more.63 Thus, the sample of overcharges in this article, while quite large, is no more than one-fourth of the total of all cartel overcharges. Because the sample of observed cartels may be different in some respects from the total population of all cartels, the features of the sample about to be described may be subject to “selection bias.” Only samples that are selected randomly from a list of the whole population are fully representative of that population, but that process is not possible in the case of cartels. Fortunately, a recent study from Germany suggests that selection bias may be minor. Haucap, Heimeshoff, and Schultz (2010) compared all illegal cartels with state- or federal-authorized German cartels during 19582004, hundreds of the latter being permitted for a wide range of reasons.64 In terms of industry distribution, the legal cartels had a greater share in mining, textiles, machinery, and metals manufacturing than did illegal cartels. Surprisingly, there was virtually no difference in the average number of firms per cartel between the two types. The major difference was that the median duration of legal cartels, having state support, was 2.75 times the illegal cartels, and legal cartel with few members or in the food industry tended to be the most durable (ibid., p. 18). What Haucap et al. (2010) suggest is that the cartels sampled for this study may well be representative of all cartels, except for their endurance. The data are organized according to three levels of analysis: markets, episodes, and specific overcharge estimates. By “market” is meant the industry or product market that was subject to price fixing. Markets are precisely self-identified by the participants in the conspiracy, though occasionally there are alternative names for the same market.65 The name of the market is eponymous for the cartel. The range of cartelized markets is impressive.66 Episodes, discussed more fully below, are distinct periods of
269
Cartel Overcharges
collusion separated by price wars, temporary lapses in agreements, changes in cartel membership, or modifications of internal organization. Episodes may be adjacent in time or may be separated by significant gaps of time.67 Markets marked by adjacent multiple episodes will typically be regarded by antitrust law as one infraction, but as economic phenomena they could be classified as multiple cartels. Because there are sometimes multiple publications about the same episode and because a single analyst will sometimes apply alternative methods of estimation, this article often records several overcharged estimates for a single episode. Most of the analyses in this article will use overcharges as the units of observation. Each episode will in principle have one true “episodic” overcharge and one “peak” overcharge.68
Number of Cartelized Markets My search yielded useful overcharge or undercharge information on cartels that operated in 532 markets (Table 2). If one group of sellers decided to fix prices of a product in one geographical region and a different group colluded on the same product in a separate geographical region, these may be counted as two markets. Of the 532 markets, 55% were cartelized by international agreements, where “international” describes the membership composition of the cartel and not necessarily the geographic spread of the cartel’s effects. Some international cartels affected directly the commerce of only one nation, though the vast majority was international in a geographic sense as well. National-membership cartels account for the remaining 45% of the cartelized markets.69 In this category, I count some purely national price-fixing cartels that were formed for the sole purpose of controlling a nation’s export sales of a particular product; in the United States, these export cartels70 are called Webb-Pomerene Associations. In addition, some domestic cartels had side agreements with international cartels that protected their domestic market from exports from the international cartel’s members. One-third (34%) of the sample consists of markets affected by bidrigging cartels (Table 2).71 Although many cartels have some sales to government entities or industrial customers that purchase by tenders, these cartels are explicitly described to have been principally or exclusively engaged in bid rigging. The proportion of bid-rigging schemes in the sample is probably underestimated because some sources did not always provide enough detail on the cartels to be certain of the degree of bid rigging. Recall
270
JOHN M. CONNOR
Table 2. Number of Cartelized Markets and Episodes, by Characteristics. Characteristic of Cartel Membership: International membership Members from one nation Conduct: Bid-rigging schemes Classic price-fixing cartels Buyers’ cartels Legal status: Cartel found guilty or liable for damages Known to have been operating legally No record of sanctions (presumed “legal”) Membership location: North America EU-wide Nations of Europe Asia and Oceania Africa, Latin America, and Eastern Europe Global (members from two or more continents) Market location: North America EU-wide Nations of Europe Asia and Oceania Africa, Latin America, and Eastern Europe Global (operations in two or more continents) Geographic reach: Single nation, of which: Local/subnational Cross-national, of which: Global Total sample
Number of Cartels
Percent of Sample
Number of Episodes
Percent of Sample
294 238
55 45
515 359
59 41
179 353 34
34 66 6.4
212 612 36
24 76 4.1
399
75
629
72
97
18
36
6.8
245
38
166 57 131 88 20
31 11 25 17 3.8
246 77 179 103 30
28 8.8 20 12 3.4
13
240
27
34 8.3 27 16 3.8
288 67 205 106 46
33 7.7 23 12 5.3
63
12
168
19
425 114 109 64
80 21 20 12
608 186 267 162
70 21 31 19
532
100
874
100
70
178 44 143 84 20
Sources: Connor (2014: Appendix Tables 1 and 2), which summarize Price Fixing Overcharges Master Data Set, spreadsheet dated October 2013.
Cartel Overcharges
271
that the U.S. sentencing guidelines assume that bid rigging leads to higher overcharges than otherwise identical conspiracies. The remaining 66% of the cartelized markets may be called “classic” price-fixing cartels, those that set market prices and/or market quotas for each of its members.72 Cartels may profit by attempting to either raise selling prices of their outputs or suppress the prices of their purchased inputs. Buyers’ cartels are often overlooked in the literature. I find that 6.4% of the cartels are buyers’ cartels; that is, 1 out of 10 of the price-fixing cartels fixed the prices of their inputs, not their outputs. This ratio is likely to be higher than many experts would have expected. Three-fourths of the cartels (75%) were found to be in violation of antitrust laws by at least one legal body.73 Sometimes these are called “discovered” or detected cartels. The determination of guilt or liability may take the form of guilty pleas (or nolo contendere in U.S. courts up until the early 1960s); of a decision at trial by judge or jury; of a commission decision to impose fines, consent decrees, or other sanctions; of the payments of civil penalties; or of negotiated settlements by defendants in a suit. Eighteen percent of the remaining cartelized markets are known or believed to be “legal,” because they operated prior to the enactment of antitrust laws in the jurisdictions in which they functioned or because they were organized and registered under antitrust exemptions, such as export cartels or ocean shipping conferences. About 7% of the cartels may be described as “extralegal” because there was nothing in the case material indicating that they were punished by an antitrust authority. Who ran these cartels and where did they function? Regarding membership composition, the largest number (187 or 35%) hail from Western or Central Europe, of which about 40% were comprised of companies from a single European nation. The next highest number is North American cartels (165 or 31%), followed by Asian (16.5%), and rest of the world (ROW = 3.8%). The final category is one that will loom large in the discussion below global cartels. These are the 70 cartels (13%) with at least two members from different continents, though typically North America, Western Europe, and East Asia are represented. The loci of operations are somewhat different (Table 2). The large majority of price fixing by cartels (80%) is directed within the boundaries of a single national jurisdiction (and one-fourth of that is more localized). The rest involves cross-national operations (and more than half of that is global). The largest single geographic category (34%) is North American cartels those operating in the United States, Canada, or both markets.
272
JOHN M. CONNOR
The second largest geographic group (27%) is cartels that functioned in only one nation in Western Europe; if these are combined with trans-EU cartels, then Western Europe is the largest continent with 35.1% of the sample. Global cartels (trans-continental cartels) comprise merely 12% of the sample; these tended to fix prices in North America, Western Europe, and Asia. Asian and ROW cartels (20%) tend to be domestic schemes populated by local companies. The apparently heavy location of cartels in only two continents is somewhat misleading. It is an artifact of the relatively early enforcement of anticartel laws in North America and Western Europe, giving rise to numerous well-documented cartel cases that could be studied by academics in those regions. The numbers likely understate cartel activity in Asia and the ROW. Going forward, cartel numbers are more likely to reflect the geographical distribution of antitrust convictions and the local capacities to analyze the cases.
Number of Episodes A more precise way of accounting for the distribution of cartel activity is by counting cartel episodes rather than whole cartels. This term episode is commonly used in modern cartel studies. If a cartel had more than one episode, then each episode is marked by a change in membership composition, the terms of the collusive agreement, method of management, geographic focus, or other major organizational innovation.74 In other words, when a cartel is re-formed, it adopts a new organizational configuration. The end of an episode is often instigated by expansion of fringe sales, by an intolerable level of cheating by cartel members, or by the appearance of a new process or product technology that redefines the market boundary. Between episodes, pricing discipline often breaks down; for some of the cartels, the interregnum is a period of contract renegotiation. The interwar global aluminum cartel, for example, went through six distinct phases from 1901 to 1939 that sometimes were adjacent in time and sometimes were several years apart. This heavily researched cartel has 28 overcharge observations (Connor 2014: Appendix Table 2). The total number of episodic overcharges is undercounted. Some single episodes reported are in fact averages of groups of episodes. For example, one episode summarized the results of 109 bid-rigging convictions in numerous distinct fluid milk markets of the Southeastern United States that occurred within a few years of each other (Lanzillotti, 1996). Each of
273
Cartel Overcharges
the 109 convictions should be counted as a separate episode because each conviction represented a distinct buyer. For 49% of the cartels found, only one episode was reported. The Bulk Vitamins cartels had 78 episodes, or about five for each vitamin product. The most impressive single-product cartel was the Newcastle Coal cartel, for which 22 distinct episodes were recorded during its impressively long life from 1699 to 1845. An additional 17 cartels have had five or more episodes, most of them global commodity cartels. Table 2 lists several key characteristics of cartel episodes. They are generally distributed in a similar fashion to the cartels themselves. International cartels tend to have more episodes than noninternational cartels, and this is especially true of geographically global cartels. So, while global cartels comprise only 13% of the sample, their episodes are 27% of the sample. On the other hand, bid-rigging cartels (34% of the sample) tend to have single episodes written up (24%).
Number of Episodic Overcharges The present study’s sample consists of 1530 cartel episodic overcharges (Table 3). In the simplest and most common situation, a cartel has only one episode. However, about half of the markets experienced multiple phases or episodes; they had an average of about six episodes. Researchers usually report the average price increases over a whole episode or a representative portion of it. Episodic averages are the measure most relevant for forensic purposes and are the measures that will be the focus of most analyses in this article. Many, probably most, episodic overcharges are conservative numbers.75 In some cases, the episodic prices are carefully weighted by the sales in each year or month of the episode, but in most cases the authors give equal weights to the price changes in each subperiod during the total affected period. Sometimes it is not clear from the source whether the averages are weighted or unweighted; if the conspiracy period is marked by steady, slow market growth, it matters little which is reported. Less commonly, some authors report minimum overcharge estimates. To be conservative, all minimum estimates are counted as episodic averages.76 If analysts give minimum and maximum estimates, I employ the center of the range for calculation purposes. The distribution of episodic overcharges across types of cartels is given in Table 3. In general, that distribution is similar to the distribution of cartelized markets across cartel characteristics (cf., Table 2). International
274
JOHN M. CONNOR
Table 3.
Number of Episodic Overcharge Observations, by Type of Cartel.
Characteristic of Cartel
Number
Percent of Sample
Episodes per Cartel
International membership National members only Bid-rigging schemes Classic price-fixing cartels Buyers’ cartels Cartels found guilty or liable No record of sanctions (“legal”) Membership composition: North America EU-wide Nations of Europe Asia and Oceania Africa, Latin America, and Eastern Europe Global (two or more continents) Market/pricing location: North America EU-wide Nations of Europe Asia and Oceania Africa, Latin America, and Eastern Europe Global (two or more continents)
1042 548 341 1249 72 1137 453
65.5 34.5 21 79 4.5 71.5 28.5
4.3 1.0 1.9a 3.3 2.7 3.2
414 195 289 142 51 500
26.0 12.3 18.2 8.9 3.2 31.4
2.3 4.4 2.0 1.7 2.6 7.9
512 141 292 146 61 383
32.2 8.9 18.4 9.2 3.9 24.1
2.3 4.0 2.2 1.6 3.0 7.7
Total (episodes with either episodic or peak estimates)
1590
100.0
2.9
a
An episode is very likely to encompass a large number of bids, perhaps hundreds. Sources: Connor (2014: Appendix Tables 1 and 2), which summarize Price Fixing Overcharges Master Data Set, spreadsheet dated October 2013.
cartels tended to have above-average number of multiple overcharges than did domestic ones and bid-rigging cartels lower. However, global international cartels really stand out with six to eight overcharges per cartel on average. The number of overcharges per market does not vary significantly across other type categories. Therefore, international cartels seem to be uniquely able to fall apart and re-form, often with better internal organization than before. This ability to renew and generate new episodes is a major factor that accounts for their longevity. Two kinds of cartel markup data are available: episodic and peak. Peak overcharges are interesting because they indicate the effectiveness of cartels
275
Cartel Overcharges
when internal and external conditions are briefly optimal. Comparisons of the two measures will be made below.
Defining Time Periods of Analysis One of the advantages of this sample is the broad time span of the data collected more than three centuries.77 To simplify exposition, tabulations are organized into seven time periods. The seven periods were selected to represent different antitrust regimes in the United States and abroad.78 In addition, the periods correspond roughly to the major changes in the relationship of antitrust jurisprudence to economics (Kovacic & Shapiro, 2000). The era up to 1890 is an obvious first period because of the enactment of the Sherman Act in the United States. Prior to 1890, no effective antitrust statute had been passed, mainly because of weak sanctions.79 During the early decades of the 20th century, numerous U.S. court decisions made the scope and power of the U.S. anticartel law apparent to lawyers, enforcement officials, and businesspersons (Wells, 2002).80 The year 1919 is chosen as a break point because it represents the end of a period of U.S. antitrust activism and because of World War I. During 19141919 nearly all international cartels, some with U.S. corporate members, had ceased operating. Many of the prewar cartels were reestablished after 1919, but in the majority of instances without the active participation of U.S. firms. The years 19451946 are another logical break point. During 19391945 nearly all of the interwar international cartels were disbanded; moreover, wartime price controls and cost-plus government contracts made cartels superfluous. Scores of U.S. criminal prosecutions of international cartels during 19401945 clarified the illegality for U.S. firms of many more subtle forms of cartel participation, such as patent pools, cross-licensing of technologies, and the creation of overseas subsidiaries as loci for cartel participation. The postWorld War II era is characterized by the emergence of industrial organization as a separate discipline within economics, of rapid advances in empirical methods of analysis, and of the adoption of effective anticartel laws outside of North America. Kovacic and Shapiro (2000) note that in the United States by the 1940s “… there was considerable consistency between judicial decisions and economic thinking …” (pp. 5152). Moreover, the vast expansion of higher education in North America and Europe brought about a parallel expansion of the economics profession as a whole and, consequently, an acceleration in the total resources devoted to
276
JOHN M. CONNOR
theoretical modeling (particularly after 1980) and related empirical testing on collusion.81 Beginning in the 1960s, economists in North America began to work more closely with prosecutors and the private bar in antitrust cases, and many of them began to analyze and write about those activities. This is a major factor responsible for the fact that nearly 80% of the estimates of “national” cartels (most of them prosecuted in North America) are drawn from the post-1945 time period. The postwar era is divided into four subperiods. The transition years 19451973 correspond with three relevant changes in anticartel enforcement. First, the antitrust idea became firmly implanted in the laws of countries outside North America for the first time: Germany and Japan in 1947, the United Kingdom in 1956, and the European Economic Communities (EEC) in 1958. Second, the European Commission (EC), the administrative arm of the EEC, after a decade of registering cartels, successfully fined its first cartel in 1969. Third, U.S. price-fixing enforcement penalties became significantly more severe in 1974. A change in U.S. anticartel legislation was the 1974 law that made price fixing a felony, thereby lengthening maximum individual prison sentences and strengthening the bargaining power of the DOJ. Class action suits became far more common by the mid-1970s because of changes in federal court rules, a change that permitted plaintiffs to attract better lawyers and economic expertise (White, 1988, Table 1.1). Finally, 1974 was the year the first econometric analysis of an overcharge was published. Although the prosecution of price fixing of relatively inconsequential domestic conspiracies was at a high level in 19741990, the DOJ did not give a high priority to investigating international cartels, nor did it have any success in the courtroom in the few international cases it did pursue (Connor, 2001a). Kovacic and Shapiro (2000) identify 19731991 as the years during which the Chicago school of economics had its greatest influence on antitrust law and enforcement. Developments in auction theory began to draw economists to empirical studies of bid rigging. By 1990, all the present criminal sanctions available to the U.S. government were in place. In 1990, penalties for corporations rose from $1 million to $10 million.82 Moreover, in the early 1990s, the DOJ had in place three devices that improved detection and prosecution of cartels: the U.S. Sentencing Guidelines for corporations (1989), the automatic amnesty policy for corporate whistle-blowers meeting certain criteria (1993), and a demonstrated ability since 1994 to impose fines above the $10-million statutory cap by means of an alternative sentencing provision. These devices were in some cases adopted by the EU and other antitrust authorities,
Cartel Overcharges
277
which significantly improved the investigation and prosecution of international cartels from the late 1980s. Both U.S. and EU prosecutions of international cartels increased markedly. The decade of the 1990s was when leniency programs were new and experimental; in the 2000s, leniency programs were standard features of antitrust enforcement. Because of these and other shifts in antitrust enforcement, this article distinguishes data of the last decade of the 1990s and from the first 14 years of the 2000s. To summarize, there are seven time periods distinguished in the present analysis: the years up to 1890, 18901919, 19201945, 19461973, 19741989, 19901989, and 20002013. Connor and Bolotova (2006) demonstrated in formal econometric testing that these time periods were significantly different with respect to the level of overcharges.
Numbers of Episodic Overcharge Estimates over Time One of the first cartels subject to historical scholarship is a London coalbuyers’ cartel that began as early as 1595 and persisted on and off for about 200 years.83 The buyers were lightermen, wholesale coal merchants who were able to manipulate the prices paid to the owners of coal-laden ships in London’s harbor. The government took many actions that proved ineffectual. Acts of parliament against bid rigging were passed in 1642 and 1665. Later, in 1729, a parliamentary investigation found that 10 lightermen controlled 67% of purchases in London, and the investigation report specifically blamed them for 17221729 price increases. Moreover, price controls for London coal were legalized in 1744, to be administered by three judges. In 1788, a law made any agreements among or partnerships of more than five coal buyers illegal “combinations in restraint of trade.” A 17001702 bid-rigging episode in London is the first overcharge estimate in the present study. London consumers of coal were later affected by a seller cartel of coalmine owners. The first mining cartel for which price effects can be found is the Coal Gild of northeastern England (later known as the “Newcastle Vend”), which made its first collusive agreement on London coal prices in 1699.84 In the early 19th century when the Vend was well organized, Tan (2003, p. 22) estimates that various episodes resulted in coal overcharges of from 12% to 16%. Although highly unstable, the Vend did not finally collapse until 146 years later in 1845. It the most durable cartel in the data set.85 The total number of episodic estimates is summarized in Fig. 1. Because of the long period covered by the sample, the mix of overcharge numbers
278
JOHN M. CONNOR 500 400 300 200 100
-2 0 00 20
19
19
13
99 90
-1 9 74
46 19
-1 9
89
73 -1 9
45 19
20
-1 9
19 -1 9 91 18
17
70
-1 8
90
0
Year Episode Ended
Fig. 1.
Table 4. Cartel Episode End Date
Number of Episodic Overcharge Estimates Collected over Time.
Number of Episodic Overcharge Estimates, by Year and Type. Membership
Legal Status
National International Found Legal guilty
Bid Rigging
Classic Price Fixing
Buyers’ All Cartels Types
Number Before 1890 18911919 19201945 19461973 19741989 19901999 20002013 All years
78 109 30 84 79 74 76 530
5 51 169 32 36 386 326 1005
47 42 60 73 96 436 360 1103
36 118 139 43 19 24 42 432
8 43 1 42 53 74 144 330
75 117 198 74 62 386 258 1205
1 1 5 5 9 15 34 70
83 160 199 116 115 460 402 1535
Sources: Connor (2014: Appendix Tables 1 and 2), which summarize Price Fixing Overcharges Master Data Set, spreadsheet dated October 2013.
changes quite a bit. Except for a bump in the two decades following the 1890 Sherman Act, the number of domestic-cartel estimates does not vary much across the seven time periods (Table 4). However, the number of international-membership cartel estimates tend to increase and peak twice:
Cartel Overcharges
279
first in the interwar years (19201945) and then in the last 24 years (19902013). Although not shown, dual peaks for global cartels are even more pronounced in those years. From 1920 to 1945, most data are drawn from studies of international cartels. Five to seven overcharge estimates are available per year during these periods. The proportion of international schemes is especially high during the interwar period and especially low during 19461990. It is likely that there were more domestic cartels operating legally in Europe in the early 20th century than there were international cartels, but the latter were given more publicity because they appeared to be novel forms of business organization.86 The increasing awareness of the illegality of price fixing in the United States may also account for the absence of internal records of domestic cartels in the United States after 1890. Moreover, because the penalties were so low (a maximum of $5,000 per count), relatively few court decisions bothered to give details about sales or prices during the conspiracy. Another trend is that the proportion of estimates from cartels that were judged guilty by a government or competent antitrust authority rose very slowly until 1989, but reversed positions thereafter. The majority (71%) of the guilty-cartel overcharges occur in episodes ending after 1989. Before 1990, 40% of the episodic estimates were from guilty cartels, whereas after 1990, 83% were. It is likely that these estimates’ patterns reflect objective market conditions, that is, the globalization of many markets in the early 20th century, recessions in the interwar period, and the surge in anticartel detection after 1990. One other change in the mix of cartel pricing conduct may reflect the availability of data and the changing preferences of economists rather than objective market conditions. In particular, the number and proportion of episodes involving bid-rigging increased markedly. Before 1945, bid-rigging episodes accounted for only 11.8% of all sample overcharges; during 19461989 it was 19.7%; and after 1989 it rose to 28.4% (Table 4 and Fig. 2). Rather than a trebling of bid rigging in natural markets, more likely explanations may lie in the direction of newly available data sets and keen interest by economists in testing new theories in auction theory.87 Prior to the 1950s, overcharges could be located for only six cartels that primarily engaged in bid-rigging conduct.88 Remarkably, in the 19451989 periods, almost half of all the overcharge observations in the sample were primarily bid-rigging conspiracies. Awareness of the importance of bid rigging among economists may have been triggered by the well-publicized U.S. electrical equipment conspiracies discovered around 1957. In addition,
JOHN M. CONNOR
13 10 20
-2 0 90 19
-2 0
09
89 19
74
-1 9
73 19
46
-1 9
45 19
20
-1 9
-1 9 91 18
-1 8 70 17
19
50 45 40 35 30 25 20 15 10 5 0 90
Percent
280
Year Episode Ended
Fig. 2.
Proportion of Overcharges from Bid-Ridding Episodic.
there were advances in economic theories of auctions that spurred interest in empirical testing of the theories. Postwar studies of bid-rigging cartels focused on national cartels in the United States, most of them local milk or construction conspiracies. The immediate victims of most of these bidrigging conspiracies were governments. Relatively few international cartels rely primarily on rigging auctions or tenders for public projects. What may seem like a refocus in effort may in fact be a consequence of changes in data availability. Most of the articles on bid rigging have drawn on public records of state or federal agencies that have been the objects of these conspiracies. It is possible that the increase in bid-rigging cases seen in the data is simply due to the advent of open-records laws in the 1960s and 1970s at the state and municipal levels similar to the federal Freedom of Information Act. Except for dips in 19461973 and 20002013, the number of observations per year has grown over time (Fig. 3). The growth in observations in episodes ending in 19901999 was extraordinary. The primary factor that explains the upward trend in the number of overcharges is the growth in the number of international cartels with usable data (Fig. 4).89 Up until 1890 when price fixing was legal everywhere in the world, only one estimate is available about every six months on average. During this early period, the vast majority of price effects are reported for domestic cartels operating in the United States, the United Kingdom, and Germany. Although there were large numbers of domestic cartels extant in the late 19th century, the
281
Cartel Overcharges 45.00 40.00 35.00 30.00 25.00 20.00 15.00 10.00 5.00 13
99
-2 0 00
90
-1 9
89
20
19
19
19
74
-1 9 46
20 19
-1 9
73
45 -1 9
19 -1 9 91 18
17
70
-1 8
90
0.00
Year Episode Ended
Number of Episodic Overcharge Estimates per Year.
Fig. 3.
90 80 70 Percent
60 50 40 30 20 10 13 20 20 00 -
99 19 19 90 -
89 19 19 74 -
73 19 19 46 -
45 19 19 20 -
19 19 18 91 -
17 70 -
18
90
0
Year Episode Ended
Fig. 4.
Proportion of Episodic Overcharges International.
small size of the fledgling economics profession, a literary approach to writing in economics, and inevitable destruction of most business records over time contributed to the fewness of quantitative overcharge observations for 19th-century cartels.
282
JOHN M. CONNOR
During 18911919, there were 5.4 price observations per year; the rate rises to 7.7 per year in the interwar period. More data are available for international cartels during 19201945 than for cartels composed of companies from a single nation. One reason is that international cartels mostly were based in Europe, where they operated with legal impunity. That is, other than Weimar Germany for a few years after 1923, cartels had freedom to set prices. In a few European countries, cartels were required to register with the government. In others, private cartel contracts were enforceable in the courts. Many of the interwar international cartels were organized as federations of national cartels and were aimed primarily at creating national monopolies and assigning shares for export sales.90 As nearly all of them were legal under the national laws of the time, their activities often were openly reported by the business press.91 Members of these cartels did not attempt to hide their activities; indeed they often publicized their operations, particularly if they achieved putatively efficiency-enhancing industry rationalization, protected national markets, increased national employment during stressful economic times, or promoted price stability. During this period, many countries passed legislation specifically authorizing cartels that controlled national exports, even if that meant agreements on prices in various overseas markets. In a few cases, including the United States, these cartels were used as cover organizations for domestic price fixing. In the early and mid-1940s, many of the interwar cartels were investigated by the U.S. Congress, indicted by the DOJ, and sued by private parties. Combined with the expanding size of the economics profession and the growing interest among economists in imperfect competition, the transparency of non-U.S. cartels led to a large number of empirical cartel studies. For 50 years after the end of World War II, the number of known international cartels declined markedly. Perhaps because of the aggressive prosecution of cartels by the DOJ in the early 1940s, it appears that international cartels were by and large driven underground for decades after 1945. From 1946 to 1989, an annual average of five or six overcharge estimates could be found, nearly all of them domestic conspiracies. Few international cartels were discovered or prosecuted until the early 1990s one or two international cartel episodes every year. Several explanations have been offered for the hiatus in international cartel formation in the two decades following 1945. The destructiveness of World War II left the United States with as much as 65% of world industrial capacity in the late 1940s. As a result, manufacturers in Europe and Japan were oriented mainly toward rebuilding their domestic markets; not
Cartel Overcharges
283
only were few industrial partners available for international agreements, it seems that U.S. firms were less prone to form cartels than firms from countries with no or weaker antitrust cultures. In the 1950s and accelerating in subsequent decades, U.S. firms embarked on a period of rapid foreign direct investment as the preferred means of improving profits; leading European and Asian firms adopted this strategy increasingly after the late 1960s. Until the early 1980s, most U.S. markets were subjected to little import competition, but by the 1990s, imports were exerting a powerful influence on price competition across a wide spectrum of commodity markets. Most international cartels have arisen only in industries with internationally traded merchandise and populated by multinational corporations with strong leading positions. For all these reasons and probably several others as yet unknown, international cartel formation was seemingly at an historically low level until the 1980s. Since 1989, the number of overcharges available has exceeded 35 per annum more than double the previous period. In part this may be ascribed to the launching of an historically high number of international cartels in the early and mid-1980s. Many of these cartels could not have been contemplated without the direct participation or passive cooperation of leading U.S. companies in the cartelized markets. Other factors that may be responsible for the surge in overcharge estimates may include greater interest in collusive phenomena by economists, shifts in antitrust enforcement priorities, expansion in the sheer number of antitrust authorities worldwide, and improved cartel detection programs. A second important reason for the surge in overcharges is that most cartel data now arise from prosecuted cartels (Fig. 5). Prior to 1946, about one-third of the observations refer to cartels known to have been sanctioned.92 Prior to the late 1940s, U.S. anticartel sanctions were weak by today’s standards, but increasingly after 1911 or so businesspersons became aware of the legal dangers of overt collusion in the domestic market. However, until the early 1970s, national and international cartels comprised of European companies could form cartels subject only to registration requirements in most European countries (and the EEC after 1960).93 The European Commission began imposing fines on unregistered cartels that affected EEC trade beginning in 1969 (Harding & Joshua, 2003, p. 121). During 19741990, U.S. corporate sanctions on cartels became significantly more severe, and the European Union’s prosecutions moved in the same direction (Connor, 2003). Both jurisdictions imposed historically unprecedented penalties on international cartels beginning in the late 1990s. After 1990, virtually all the observed cartels in the sample were
JOHN M. CONNOR
13
99
-2 0 00 20
19
90
-1 9
89 19
74
-1 9
73 -1 9 46 19
19
20
-1 9
19 -1 9 91 18
-1 8 70 17
45
100 90 80 70 60 50 40 30 20 10 0 90
Percent
284
Year Episode Ended
Fig. 5.
Proportion of Overcharges from Sanctioned Episodic.
prosecuted or fined by one or more antitrust authority. This pattern suggests a marginal improvement in cartel deterrence (albeit still suboptimal), but it does not necessarily mean that the probability of discovery by prosecuting bodies has gone up. However, it probably does represent a heightened aggressiveness in anticartel enforcement by a much larger number of authorities as well as more productive research methods by social scientists.94 Finally, it is important to note the special role of global international cartels in this study. There are 383 episodes of episodic global-cartel overcharges, which account for 25.0% of the total number of overcharge estimates in the sample. There are few global-cartel overcharges available prior to 1900. The first recorded global cartel was the French-Belgian zinc national cartel that began in 1847 and later merged into a global export cartel; the first global-export cartel episode that began in 1862 was reportedly ineffective (Devos, 1994). The next global cartel episode, Secre´tan copper syndicate of 18871889, was highly effective. Although organized by four French and British firms, the syndicate cornered the supply of copper by signing long-term supply contracts with mine owners all over the world. The syndicate made profits when the contraction of supply forced up the price of copper contracts traded on European commodity exchanges (Andrews, 1889). It was unanticipated supplies of Asian recycled copper that brought this pioneering global cartel to its end. It is noteworthy that nearly all the global-cartels episodes in the sample that ended before 1890 were minerals, metals or metallic ores all industries with very high fixed, sunk costs.
285
Cartel Overcharges
The interwar period witnessed an explosion in global cartels, many of them Europe-based export cartels. From 1920 to 1945, more than 163 episodic overcharges of global price fixing were recorded; these global cartels account for an impressive 77% of all episodes in the interwar period, a record-high proportion that still stands. Discovered global cartels remained relatively scarce until the 1990s when 77 episodes of price fixing ended, but these episodes comprise only 17% of all of the episodes in that period. One can only speculate as to why global collusion apparently first became feasible in the 1920s and later during 19801999 revived in popularity. The availability of improved international and trans-oceanic communication and transportation very likely played a role. International trade and foreign investment surged in the 1920s and 1980s, factors that may account for intensified price competition or the formation of strategic conjectures about all the world’s major suppliers in an industry.
Data Reliability Issues Many readers will have prior beliefs about the most appropriate data and methods that ought to be employed to derive estimates of the price effects of cartels. Some might regard a lengthy historical investigation with access to the internal communications of a cartel’s managers as the surest path to the truth. Others might give greater credence to such communications only where the cartelists had reason to believe that their activities were legal or where the managers are writing about an illegal cartel years after the statute of limitations had passed. Some might assume that disinterested social scientists are likely to be closer to the mark than prosecutors, plaintiffs’ counsel, defendants’ counsel, or other interested parties. Indeed, the crosschecks of a more global retrospective analysis might contradict delusions, if they are delusions, of cartel managers about their power over markets. Among economists, ever cognizant of the march of progress in quantitative research methods, there may be a tendency to regard peer-reviewed studies applying methods of the most recent vintage to highly disaggregated, detailed data the most reliable.95 Among legal scholars, many will view criminal trials or other procedures with criminal protections as the gold standard of fact-finding, whereas civil-law administrative hearings likely to contain more errors. By design, this research project did not filter out some groups of cartel studies because they are purportedly surpassed by other groups in quality.96
286
JOHN M. CONNOR
Rather, three approaches are taken to learn whether the various overcharge estimates are sensitive to the methods utilized, data sources, time period, or disciplines of the authors. In my view, statistical meta-analysis applied after data collection is a more appropriate tool to handle such issues, and that is the approach taken in Connor and Bolotova (2006). Their model uses a large subset of the data shown in this article and controls for a large number of cartel-estimation factors and predicts quite satisfactorily. I summarize their findings in this section. Confidence in the estimates may be judged in part by the high quality of the publication sources from which the overcharge estimates were derived (see Bibliography). The large majority of the estimates are drawn from the traditional end-product outlets of academic research: academic books, book chapters, and peer-reviewed journals account for 65% of the total (Connor, 2005a, Table 11). In addition, 15% of the estimates were taken from economist’s working papers, most of which examine modern international cartels and are intermediate versions of subsequently published book chapters and journal papers. The majority of the government reports (4% of the estimates) were authored by civil servants with specialized training in economics, and some were written by academics commissioned by the agency; typically, these reports would be vetted by a panel of experts. For example, the legal decisions of the UK Monopolies Commission were reviewed and approved by panels that contained a couple of leading professors of industrial economics working alongside senior civil servants attached to the Commission. Much the same process was used for United Nations, OECD, and Congressional Committee reports on cartels. Court and competition-law commissions accounted for 12% of the estimates. In sum, four-fifths of the estimates are drawn from the formal or informal writings of academic social scientists, and most of the remainder was the product of professionally trained individuals subject to the checks and balances of internal reviews. The types of publication outlets have changed over time. Before 1974, books and chapters in edited collections accounted for 58% of the publications that contained usable overcharge data. Most of these earlier works show evidence of meticulous scholarship, but the share of them subject to blind reviews is small. After about 1973, books became a very minor component of this survey’s source materials. Instead, the three-fourths of overcharge estimates shifted to the published decisions of courts and commissions and academic working papers and journal papers. That is, in recent decades, most estimates are drawn from papers that have been peer-reviewed, from an adversarial forum, or from decisions likely to be
287
Cartel Overcharges
reviewed by courts of appeal. Some may regard the latter review processes as likely to induce more reliable calculations. Controlling for other factors, Connor and Bolotova (2006, Table 6) find that government reports tended to have systematically lower overcharges than the reference group, books, and monographs. Estimates published in all other publication forms were not statistically different from books. A singular characteristic of science is its tendency to improve on the past. I examined whether there are systematic differences between the episodic overcharges across time, using the date of publication of the study as a proxy for analytical advances. The intuition here is that the authors of more recent empirical studies of cartels have learned to avoid the methodological pitfalls of their predecessors.97 Among the economic studies that dominate the sample, there is an undeniable trend away from mere narrative historical case studies sometimes embellished with simple graphical illustrations toward more formal statistical modeling. Correspondingly, in industrial economics, generally, there is a trend away from evaluating cartels from the point of view of the theory of pure monopoly toward a more sophisticated and nuanced view informed by game theory and other conceptual advances. Controlling for other factors, Connor and Bolotova (2006, Table 6) find that overcharge estimates decline slightly over time, but the effect is not completely monotonic. Of course, other things are changing over time as well, including generally tougher anticartel enforcement with respect to cartel discovery and severity of penalties. A more direct test involves qualitative variables for the author’s estimation method. In this case, the yardstick method tends to result in significantly higher estimates than the reference group, which is the “after” method. The rest of the analytical methods are not significantly different from each other or the reference group. Thus, with one minor exception, methods used do not cause bias in estimating overcharges. In sum, apart from minor exceptions, neither sources nor methods suggest unreliability.
ANALYSIS OF THE OVERCHARGE RESULTS Number of Overcharge Observations There is a total of 2044 quantitative estimates of overcharges and undercharges drawn from about 350 publications.98 The sample consists of 1535
288
JOHN M. CONNOR
episodic (long-term) and 470 peak estimates (highest price achieved for one year or less). Every estimate is assigned to one episode. Of the 1589 price-fixing episodes in the sample, 1536 (96.7%) have only an episodic estimate99 and 53 (3.3%) have only a peak estimate, but 455 episodes (28.6%) have both types of overcharge estimates.100 A large majority (65%) of the episodic overcharge estimates are drawn from international-membership cartel episodes (Table 4). More than twothirds of the estimates (71%) come from episodes that were legally sanctioned and almost four-fifths (78%) from “classic” price-fixing schemes. The smallest cartel type by far is buyers’ cartels (4.6%). The episodes made must be classified according to their geographic extent or geographic region of operation. Of the episodic overcharges, 17% are local/subnational, 47% cover entire nations, 36% involved multiple nations within one continent, and 25% are global. In regional terms, the great majority of episodic overcharge estimates are drawn from Western European (30%) or North American (25%) conspiracies (Table 12). However, the share of episodic estimates drawn from episodes of global price fixing is also quite large (31%). Information on African, Asian, or Latin American cartels is relatively sparse. International-membership and global international collusion tends to be more durable and to spawn far higher numbers of episodes per cartel than any other types of collusion. Twenty-three percent of the 2005 overcharge figures that were assembled are peak price effects. In some cases, the peak price was reached for only one day during a cartel episode; in other cases, the peak may be the highest one of several years; most often it was an intermediate length. Peak price changes indicate the potential for maximum harm when a cartel is at its most disciplined or when market conditions were most congenial. Classifying a particular estimate as an average or peak figure in a minority of cases required judgment. If the original source is unclear about which type of estimate is being presented, in order to be conservative I have assumed it is a peak estimate. Peak estimates are separately analyzed below.
Height of Episodic Overcharges over Time and by Type Table 5 and Fig. 6 display the medians of all episodic overcharges reported, distinguished by membership type, legal type, mode of pricing conduct, and time period. Median averages may be preferred by some readers because nearly all the cells contain negatively skewed figures. That is, a few very high overcharges in any particular category tend to overwhelm the larger number of low-to-medium percentages when calculating the more
289
Cartel Overcharges
Median Average Episodic Overcharges, by Year and Type.
Table 5. Cartel Episode End Date
Membership
Legal Status
Bid Classic Rigging Price National International Found Legal Fixing guilty
Buyers’ All Cartels Types
Median percenta Before 1890 1890–1919 1920–1945 1946–1973 1974–1989 1990–1999 2000–2013 All years
18.2 20.5 5.0 15.4 16.7 17.7 20.7 18.2
34.0 54.0 30.5 24.1 29.1 25.0 20.0 25.1
16.1b 20.5 36.2 15.0 20.0 24.0 20.1 22.3
22.0 39.0 27.0 25.0 6.8 20.8 17.5 26.0
16.2 39.0 34.0 14.2 20.0 18.9 18.2 18.7
19.8 31.6 29.0 23.0 16.8 24.8 20.9 24.5
36.5 430.0 6.4 47.0 24.0 16.6 17.6 19.8
19.3 34.5 29.0 19.5 18.1 24.0 20.0 23.0
a
Medians of the point estimates or, where appropriate, of the midpoint of range estimates. Includes many zero estimates. See Table 4 for the numbers of observations in each cell. b Only three cartels (but with 47 episodes) were deemed guilty prior to 1890: Wholesale Grain Merchants in Greece (guilty by public trial), Anthracite Coal (by U.S. court), and Newcastle Coal (by the UK Parliament). Sources: Connor (2014: Appendix Tables 1 and 2), which summarize Price Fixing Overcharges Master Data Set, spreadsheet dated October 2013.
30
Percent
25 20 15 10 5
Fig. 6.
l ar te ’C
Bu
ye
rs
e ic Pr
R Bi d
Fi xi ng
ig gi ng
l ga Le
ui lty G
at io er n In t
N
at io
na
l
na l
0
Median Episodic Overcharges by Cartel Type.
common type of average, the mean. Moreover, while there is no upper limit on overcharge estimates, they cannot fall below zero. In such situations, the means are larger than the medians, and the median may be a better representation of central tendency. The median cartel overcharge for all
290
Table 6. Cartel Episode End Date
JOHN M. CONNOR
Median Average of Positive Episodic Overcharges, by Year and Type. Membership
Legal Status
National International Found Legal guilty
Bid Rigging
Classic Price Fixing
Buyers’ All Cartels Types
Median percenta Before 1890 18911919 19201945 19461973 19741989 19901999 20002013 All years
20.6 24.8 20.0 18.8 16.9 18.9 23.3 20.0
74.0 59.8 39.5 42.0 43.5 25.1 20.2 27.0
24.4b 24.4 44.9 28.2 20.0 24.6 20.3 23.3
17.6 44.0 36.7 17.9 9.7 21.5 18.4 30.0
16.2 24.5 34.0 14.2 20.0 18.9 18.0 19.8
22.6 44.0 37.0 23.9 21.5 25.0 21.0 26.6
36.5c 430c 6.4c 47c 23.0 16.9 17.6 21.6
22.0 36.8 36.9 22.5 20.0 24.0 20.0 25.0
a
Medians of the point estimates or, where appropriate, of the midpoint of range estimates. This table excludes zero estimates. On average, 94% of all episodic overcharges are above zero, and that percentage increases over time. Very few peak overcharges are zero. b Only three cartels (but with 47 episodes) were deemed guilty prior to 1890: Wholesale Grain Merchants in Greece (guilty by public trial), Anthracite Coal (by U.S. court), and Newcastle Coal (by the UK Parliament). c Only five peak observations before 1974, so comparisons with totals are problematic. Sources: Connor (2014: Appendix Tables 1 and 2), which summarize Price Fixing Overcharges Master Data Set, spreadsheet dated October 2013.
types and time periods is 23.3% and for effective (“successful”) cartels 26.0%.101 I will demonstrate later below that the highest overcharge estimates are in no sense aberrations. They are generally taken from well-conducted studies of cartel episodes that arose from monopolistically structured markets and, therefore, do not deserve to be rejected. Hence, the mean average also has a strong claim to represent the central tendency of the sample. The mean episodic overcharge is 49%; for effective episodes it is 59% (Table 9 and Fig. 7). Overcharges over Time Cartel markups vary according to time period, but it is hard to tell from the raw data whether the 300-year trend is rising or falling (Table 5 and Fig. 8). They are above average for two periods (18911945), below average during two periods (19461989), and closer to the all-periods average
291
Cartel Overcharges 60
Percent
50 40 30 20 10
l ar te C
Bu
ye
Pr ic
d Bi
In
rs ’
e
R
ig
Fi
xi
gi
ng
ng
l ga Le
ui G
te
N
rn a
at
io
tio
na
na
l
l
lty
0
Fig. 7.
Mean Episodic Overcharges by Cartel Type.
35 30
Percent
25 20 15 10 5
01 3 20
10
-2 90 19
-2
00 9
9 98 19 74 -1
73 19 46 -
19
5 94 19 20 -1
19 19 18 91 -
17 70 -
18
90
0
Year Episode Ended
Fig. 8.
Median Episodic Overcharges over Time.
for the other three time periods (before 1890 and 19892013). Variation over time appears to be related primarily to changes in the mix of cartels types. For example, overcharges are relatively high when the time-period mix is rich in unpunished and/or international cartels but poor in bidrigging cartels (cf., Table 4).
292
JOHN M. CONNOR
Looking more deeply into the micro data (1531 episodes), a strong upward linear time trend in international-membership cartel episodes is apparent.102 Episodic overcharges are slightly positively correlated with international membership (r = 0.06), but there is no evidence of a simple correlation of overcharges with any of the other cartel characteristics. A finding emphasized in this study is the superior price effectiveness of international cartels relative to domestic ones (and correspondingly the higher markups of geographically expansive collusion over small-scale schemes). However, this disparity is disappearing over time. A steep secular decline in episodic overcharges is evident among international cartels.103 Median international cartel overcharges were an unequalled 53.0% prior to 1920. During the interwar period these cartels attained only average levels of price effectiveness; median overcharges fell by 31% compared to before 1920. Given the poor economic conditions of the 1930s, the profits generated by these cartels may have been satisfactory to the cartelists. But overcharges continued to decline by 43% in 19471989 and finally by 60% in 20002013 relative to pre-1920 levels. In fact, prior to 20002013, international cartel episodes had exceeded domestic ones in every period by large margins, whereas in the most recent 14 years their positions had reversed!104 Regression analysis confirms that, after changes in mix of types of cartels are taken into account, cartel overcharges were significantly lower after 1919 than before 1920 (Connor & Bolotova, 2006, p. 1133). During the interwar period, overcharges were six to seven percentage points (about 20%) below the 17701919 reference period. During 19461989, overcharges were 811 percentage points lower than the reference years. Finally, after 1989 the era of strongest antitrust enforcement overcharge rates declined about 40% below the pre-1920 reference period. While the temporal decline in cartel overcharges is undeniable, the historical forces responsible have not been pinned down and they may not be irreversible.105 The rigor and geographic spread of antitrust enforcement seems to be the most natural candidate as the principal factor responsible, but other forces may be contributing.106 It is a challenge to explain the downward trends for some types of cartels. Besides the possible influence of the spread of effective anticartel enforcement, several alternative hypotheses may be put forward. Perhaps the application of more sophisticated quantitative methods by researchers in recent decades systematically yield lower estimates of price effects than the earlier studies that relied on simpler before-and-after comparisons. Perhaps expected profit rates in cartelized industries declined as the impacts
Cartel Overcharges
293
of globalization were felt in formerly protected markets, and those companies that join cartels are satisfied with smaller percentage increases from collusion. Industry mix could provide an explanation. The sample drawn from the earlier periods tends to contain more minerals and metals conspiracies, whereas the later estimates have a higher proportion of chemical, construction, and services firms represented. Construction and services have historically returned very low profit margins. Because the most recent periods contain a higher proportion of cartels that were caught by antitrust authorities, the more recent estimates may be drawn from a population of cartels that is relatively incompetent in hiding their activities; similarly, the greater antitrust scrutiny in the United States from 1940 and from Europe since the 1960s could prompt cartelists to refrain from full monopoly pricing increases so as to reduce the chances of detection. Some of these hypotheses will be investigated below. There are significant differences in the height of overcharges when the sample is split according to three cartel characteristics: national or international in membership, bid-rigging or classic price-fixing conduct, and sanctioned or unsanctioned cartels history. In the aggregate and for all time periods, highest markups are associated with international membership, classic price-fixing methods, and no history of official sanctions (Fig. 6). The patterns evident from these tabulated overcharges have been verified by a more formal statistical analysis (Connor & Bolotova, 2006). International-Membership Cartels The median overcharge for national cartels is 18.2%, whereas for international cartels it is 25.1% (38% higher). Measured by the mean averages, international overcharges are 56% and national are 35% (Table 7). Regression analyses verify that international cartels have overcharges about 45% higher than domestic schemes (Bolotova, 2009, Table 4). The strongest categorical pattern is that until the 2000s, in every historical period international cartels have had higher overcharge rates than domestic cartels (Table 5). Up to the 1990s, international cartels were on average twice as effective in raising prices than “national” cartels (cartels that fixed prices in one country and export cartels comprised of firms from single countries). This is not so surprising in the preWorld War II era because most of the prewar sample of national cartels operated in the United States and achieved quite low overcharges.107 But the fact that the differences persisted in the postwar period is somewhat unexpected. Besides antitrust-enforcement considerations, the greater pricing power demonstrated by international
294
JOHN M. CONNOR
Table 7. Average Measure
Mean Average Overcharges, by Type.
Membership
Legal Status
National International Found guilty
Bid Classic Buyers’ Rigging Price Cartels Legal Fixing
All Types
Percent Mean episodic, as reporteda Mean episodic, corrected for competitive affected salesc Mean effective (nonzero) episodes Mean effective episodes, corrected for competitive affected salesc Mean peak positive overcharges
34.6
56.1
48.7
48.6
54.5
47.2
43.6
48.7
52.9
127.8
91.6
91.1
109.7
89.4
77.3
94.9
37.8
58.7
50.5
51.6
56.0
50.6
45.5
51.8
63.1
142.1
102.0
106.6
127.3
102.4
83.5
107.5
69.0
121.7
108.5
75.9
53.9
114.1
33.1
103.5
a
The arithmetic mean. If they report their method at all, the large majority authors appear to divide the dollar overcharge by the cartel’s dollar sales during the collusive period (“affected sales”), which leads to under-reporting of the overcharge rate. Other authors do not have access to the affected sales of the cartel and instead use total market sales, which in general causes even a greater under-reporting of the overcharge rate. b The divisor is corrected for the inclusion of collusion-inflated sales. No adjustment is made for possible inclusion of fringe firms’ sales. c Suppose OV% is the conventional computation of the overcharge rate (see note a). The true overcharge rate TOV% = OV%/(100 − OV%). Note that authors that employ prices directly to derive the overcharge rate do not require this correction. Sources: Connor (2014: Appendix Tables 1 and 2), which summarize Price Fixing Overcharges Master Data Set, spreadsheet dated October 2013.
agreements may reflect a greater degree of freedom from threat of entry than for geographically more localized cartels. International cartels in all eras tended to attract members that controlled the lion’s share of production in all the regions of the world with modern production facilities. Also, international cartels by their very nature deal with internationally tradable commodities, mostly homogeneous producer intermediates with relatively
Cartel Overcharges
295
low long-distance transportation costs. Finally, international cartels can more easily engage in third-degree price discrimination among national markets than cartels organized within a single geographic market. In the 19902005 period, the superior pricing power of international schemes ebbed. The median overcharge fell to an historical low of 24.4%. In a sharp break from the first five periods, overcharges of international cartels averaged only 16% higher than national ones. The reasons for the convergence of national and international cartel markups are difficult to divine.108 Bid-Rigging Schemes A somewhat similar difference may be seen in the higher median overcharges for classic price fixing over bid rigging. In the sampled cartels, classic price-fixing conduct led to 32% higher median overcharges than observed for bid-rigging methods. Bid-rigging cartels often are organized to exploit tenders for government public-works projects. Some economists have hypothesized that government buyers are less competent in detecting rigged bids than are professional industrial buyers.109 Relatively few international cartels engage primarily in bid rigging, so this conduct category may be confounded with geographic extent or industry type (most are found in construction). The apparent lower overcharges arising from bid rigging may be an illusion. Regression analyses suggests that overcharges from bid rigging are no different from classic price fixing (Bolotova, 2009; Connor & Bolotova, 2006). This finding has policy significance, because it undermines an assumption of the U.S. Sentencing Guidelines, which impose higher penalties for bid rigging. Most other governments have no stated guidelines imposing extra fines for bid rigging, but there is evidence that bid rigging of government tenders is treated more harshly (Connor, 2009c). Sanctioned versus Unsanctioned Cartels The difference in median overcharges for “legal” versus guilty cartels is rather small; punished cartels achieve lower median overcharges overall, though not in most time periods (Table 5). Regression analysis verifies that there is no significant difference in overcharges by legal status (Bolotova, 2009, Table 4). Contrasting cartels according to their legal status may shed light on “sample selection bias,” an important methodological issue in cartel studies. Many cartel studies depend on samples of convicted cartels, and critics of these studies have asserted that cartels discovered through government
296
JOHN M. CONNOR
investigations or sued by private plaintiffs are as a group inept compared to cartels that either had no fear of sanctions or remained clandestine. “… [I]t is not known whether cartels that find themselves in court are unsuccessful or merely unlucky” (Carlton & Perloff, 1990, pp. 216217). In particular, an influential study by Asch and Seneca (1975) finds that price fixers that were caught in 19581967 were significantly less profitable during collusion than a control group of unprosecuted firms.110 Lower profitability ought to go hand in hand with relatively low overcharges. The data in Table 5 suggest a resolution of this paradoxical finding. U.S. cartels punished in the time period covered by the Asch and Seneca study (53 episodes in 19461975) were indeed relatively inept: their median overcharges of 17.3% are lower than any of the “guilty” cartels in any of the seven time periods. Moreover, their sample appears to have been drawn disproportionately from domestic bid-rigging conspiracies, the categories that throughout history have generated the lowest overcharges. While a more precise analysis is needed, it appears that the Asch and Seneca study may itself be flawed by sample selection bias. Buyers’ Cartels Blair and Harrison (2010) argue that monopsony and oligopsony are topics often given short shrift by economics and rarely addressed by the courts, in part because of the mistaken belief that if buyer power forces down prices below competitive level then consumers must benefit. In fact, if buyers explicitly collude on the price of a procured input, then an undercharge111 is likely to be imposed on suppliers that are symmetric to the antitrust damages created by overcharges on buyers from sellers’ cartels (ibid., pp. 157163). In both cases, industry output contracts from the level that would be seen in purely competitive or noncooperative oligopsonistic procurement markets and allocative inefficiency is created. Blair and Harrison (2010) valiantly attempt to convince readers that buyers’ cartels are “… far more prevalent than many have recognized” (ibid., pp. 114). Restricting their purview to cases brought in U.S. courts112 or documented in publications by American economists, by my count, they assemble a sample of 24 documented buyers’ cartels. Drawing upon a slightly older version of this work’s price-fixing overcharges data, Jing Liu (2011) statistically analyzed the prevalence and unique economic characteristics of buyers’ cartels. She finds four notable differences. Throughout history, only 5.5% of all cartel price effects were undercharges by buyer groups, but that ratio had risen from practically 0% to above 8% after 1990 (ibid., Table 1). While sellers’ cartels
297
Cartel Overcharges
are mainly in manufacturing, buyers’ cartels are preponderantly discovered in the food, tobacco, raw materials, and services industries (ibid., Table 4). Buyers’ cartels are much more likely to be domestic bid-rigging schemes than other cartels (ibid., Figures 810). Finally, the average price effects of buyers’ cartels are 33% weaker than those of sellers cartels (ibid., Table 11). Enlarging upon her work, I find that scholars have published studies on 70 cartel undercharges; that these comprise 4.6% of the sample; and that median undercharges are about 19.8%, the lowest type of cartel overcharges (Tables 6 and 8). Market Structure Overcharges are a measure of group (multilateral) market power exercised. A long tradition of empirical research in industrial economics has demonstrated a strong association between market power and several dimensions of market structure. For example, high seller market concentration raises sellers’ power, while buyer concentration lowers it. While information on market structure is difficult to obtain (particularly for older cartels), Bolotova (2009) managed to construct a sample of 156 international cartels discovered between 1990 and 2005, which include five measures of market structure (Table 2). These variables have as a group strong power to explain variation in overcharges.113 Bolotova’s regression results demonstrate that the cartel supply share (close to seller market concentration) is Table 8. Number of Peak Positive Overcharge Estimates, by Year and Type. Cartel Episode End Date
Membership
Legal Status
Bid Classic Rigging Price National International Found Legal Fixing guilty
Buyers’ All Cartels Types
Number Before 1890 1890–1919 1920–1945 1946–1973 1974–1989 1990–1999 2000–2013 All years
17 54 16 24 26 16 29 182
2 30 72 19 25 119 58 325
4 25 31 23 44 130 79 336
15 59 57 20 7 5 8 171
3 0 1 14 22 18 32 90
14 84 87 29 29 117 55 417
0 0 1 3 3 7 13 27
17 84 88 43 51 135 87 507
Sources: Connor (2014: Appendix Tables 1 and 2), which summarize Price Fixing Overcharges Master Data Set, spreadsheet dated October 2013.
298
JOHN M. CONNOR
positively related to overcharges and buyer concentration negatively related, as expected (Table 5). Furthermore, inequality of size among the cartel members (the leading firm’s market share) lowers overcharges. Two other structural variables were not statistically significant. There are historical examples of cartels that ended because of the growth of fringe production. In such cases, one would expect supplier concentration to decline after collusion ends. The vitamin C cartel of the 1990s is one well-documented case (Connor, 2007b). However, that may not apply to cartels that were broken up through enforcement actions. A recent study by Levenstein, Sivadasan & Suslow (2011, p. 12) examines the levels and changes in an importer-based proxy for supplier concentration.114 While their sample is limited to seven global organic chemicals cartels in the 1990s that ended because of antitrust actions, they find that supply is very highly concentrated in all cases. A key regression analysis shows that three to four years after the breakup year, in six out of seven cases there was no significant decline in concentration (ibid., Table 8). Unsuccessful Cartel Episodes It is worth noting that there are relatively few unsuccessful cartels in the data set. Only about 6% of the overcharges indicate that an analyst judged an episode to have produced no significant effect on market prices, even though the members had established an agreement in principle to fix prices. I do not wish to make too much of this percentage. It may understated because of selection bias in the studies relied upon. Injurious cartels may be inherently more interesting to analysts, because they are more policy relevant or the results more publishable than those about incompetent cartels. Not counting failures to discover a feasible contract, my intuition is that the true proportion of unsuccessful cartels (discovered and undiscovered) is likely to be higher than 6%.
Size Distribution of Overcharges The cartel fining guidelines of several jurisdictions, notably the EU and the United States, are based on formulas that are multiples of assumed overcharges (or proxies thereof). Given the interest in the factual foundations of the U.S. and EU Sentencing Guidelines applied to cartel sanctions, an examination of the size distribution of the overcharge estimates ought to be of interest. Fig. 9 classifies the average estimates into eight size categories. Because the U.S. Sentencing Guidelines are predicated on the assumption
299
Cartel Overcharges 25
Percent
20 15 10 5
us
9 0 20
10
0.
pl
9.
.9
019
-9 9
.9 .0 80
-7 9
.9 .0 60
-5 9
.9 -3 9
.0 40
.9 .0 30
.0 20
-1 9 .0 10
Fig. 9.
-2 9
.9
9 19. 0.
Ze
ro
0
Size Distribution of Episodic Overcharges.
that the average cartel has a 10% overcharge, that break point is of special interest. The discussion of Table 9 focuses on the effective cartels (nonzero overcharges). Perhaps the most striking result is that 60% of the cartel episodes have overcharges above 20%.115 The remaining episodes have overcharges less than 20%. The mean overcharge of these episodes is 12%. These are the episodes imagined to be typical by the creators of the U.S. Sentencing Guidelines. By contrast, the 60% of the cartel episodes with overcharges of 20% or higher have a mean overcharge of 79.7%, which is eight times the level assumed by the authors of the U.S. Guidelines. The Guidelines were designed to deter recidivism, but even if one makes the wildly optimistic assumption that the probability of detection is 100%, fiveeighths of the cartel episodes in the sample would have been under-deterred. Looking in Detail at Extreme Observations As is shown in Table 6, 3% or 4% of the sample of episodic overcharges has overcharges of 200% or higher. These are rates about which an anonymous reviewer and previous readers have expressed incredulity. Because they strongly affect the mean sample value,116 I will examine them in detail. Are the high-overcharge cartels from unique historical periods? Were the data or analyses of poor quality? Do their other traits differ significantly from the rest of the sample?
300
JOHN M. CONNOR
Table 9. Percentage Rangea
Mean Average Episodic Overcharges by Size Category. Number of Observations
Mean Average
Total
Number b
Zero or less 0.19.9c 10.019.9 20.029.9 30.039.9 40.059.9 60.079.9 80.099.9 100.0199.9 200 plus Total
92 239 345 250 181 192 81 27 72 50 1540
Distribution of Observations Nonzero
Percent 0 5.4 14.5 24.4 34.2 48.4 67.9 88.8 136.6 563.9 48.7d
6.0 15.5 22.4 16.2 11.8 12.5 5.3 1.8 4.7 3.3
6.4 16.5 23.8 17.3 12.5 13.3 5.6 1.9 5.0 3.5
100
100
a
Point estimates or midpoints of ranges. Undercharges are converted to positive numbers. c Four estimates of “weak cartels” are assumed to be 1% overcharges. d For effective cartels (those with positive overcharges), the mean average is 58.9%. Sources: Connor (2014: Appendix Tables 1 and 2), which summarize Price Fixing Overcharges Master Data Set, spreadsheet dated October 2013. b
To answer these questions, Table 10 isolates the 50 largest episodic overcharge observations and lists their essential characteristics. It also includes in the last column my subjective qualitative evaluation of the reliability of the estimates something I have refrained from doing elsewhere in this report.117 The high-overcharge cartels tend to be drawn from older cases; their average beginning year is 1949 versus 1962 for the remaining effective observations. Another difference is that the high-overcharge cartels were on average two years more durable than the remaining cartels: 10.0 years as compared to 8.1 years. I would not ascribe the high estimated price effects to poorly executed analyses. Fully 75% of the grouped observations in Table 10 were rated from very good to excellent analyses. In terms of overall industrial mix, the high-overcharge cartels look very much the same as their lower overcharge counterparts: a few raw materials, some services, but mostly manufactured intermediate inputs. However, closer look reveals that a large proportion of the cartelized products were new products in great demand as essential industrial inputs with few or no
222.0 214322 800.0836.5 243.01329.0 886.0 429.0
19241939
19281936 19281936 19281936 19281941
201.8203.0
19221940
19241939
430.0
1919
Electric light bulbs, world price Electric light bulbs, world price [4] Tungsten carbide [2] Tungsten carbide [3] Tungsten carbide [2] Tungsten carbide
254.0
19191949
Phosphate rock, US, world exports Rare books auction, bidding ring, country estate in Surrey, UK Sulfur, global [2]
243.0 257.0
19121918 19131922
433620
223.5 227.0 233.0
18901894 18991814 19001908
1908
350.0
18781881
Percent
Overchargea
Radium, global Raisins, US
Telephone service, home and office, NY City [2]
Cordage, sisal or hemp, Eastern US Borax, European based Steel tubes, US Steel, barbed wire, US
Years
Berge (1944) DOJ prosecutor’s book Government of Canada (1945) research report U.S. court decision Suslow (2005) refereed academic journal
UK Monopolies Commission (1951)
MacKie-Mason and Pindyck (1987) academic book chapter Stocking and Watkins (1946) classic book
Porter (1992) refereed article
Pierce (1913), Holt (1907) books Jones (1921) popular book, academic author Jenks and Clark (1928) popular book, academic authors Demarest (1910), an opinionated book on evils of monopoly by an author considered one of the pioneering “muckrakers” Government of Canada (1945) research report Jenks and Clark (1929) popular book, academic authors Dick (1992) refereed article
Dewing (1913) book
Source of Estimate(s)b
Very good Excellent Very good Excellent
Very good
Excellent
Excellent
Excellent
Excellent
Perhaps poor yardstick if Bell Co. had no competition Excellent Very good
Very good Very good Very good
Very good
Quality Assessmentc
Summary of the Characteristics of the 50 Highest-Overcharge Observations, by Year Cartel Began.
Cartel Market [Number of Estimates above One]
Table 10. Cartel Overcharges 301
19021990d
Cable, high-voltage power, Germany, experimental laboratory market [3] Coconut oil, Philippines
200.0 Infinity 211.0 200.0
1974
19901905
19901995
19952002
19962005
Uranium metal, U.S. market
Banks, credit-card interchange fees, Spain Banks, debit-card interchange fees, Canada Tobacco, leaf, procurement, Italy Currency conversion fees, charge cards, US
19962003
200.0
1974
378.0
244.0
480.0
739.0
1964
Natural gas pipeline bid to California “El Paso”
Source of Estimate(s)b
(Continued )
Lande and Davis (2007)
Complaint January 22, 2002
Carbo´-Valverde, Chakravorti, and Rodriguez-Fernandez (2011) working paper Carbo´-Valverde, Chakravorti, and Rodriguez-Fernandez (2011) working paper EC decision October 20, 2005
Davis and Garce´s (2009) advanced textbook on damages methods U.S. Congress (1977) report
Buschena and Perloff (1991) refereed academic paper Cassady (1967) academic book
Government of Canada (1945) research report Stocking and Watkins (1946) classic book MacKie-Mason and Pindyck (1987) academic book chapter 1255.04918.0 Fonseca and Normann (2012) refereed academic paper
302.0310.0 200.5612.0 239.9
Percent
Overcharge
Antiques auction, UK (one week) Uranium metal, U.S. market
1959
19361941 19361941 19511970
Years
Tungsten carbide [2] Tungsten carbide [2] Mercury
Cartel Market [Number of Estimates above One]
a
Table 10.
Suit has not been successful, so estimate in doubt Excellent
Very good
Very good
May not incorporate the highest scientific standards Very good
Very good
Very good
Excellent
Excellent
Excellent Excellent Excellent
Quality Assessmentc
302 JOHN M. CONNOR
400.0 270.0 577.3e 32.6
20062009 19491958
19621969
470.0
20042006
2005
200.0
20002001
348.5
450.0
20002002
20052012
1800.0 800.0
1998 19992001
Duration 8.1 years
Yoon (2009) reporting on KFTC analysis Duration 10.0 years
Antitrust Division DOJ (2008) report on increase in fuel surcharge increase compared to spot jet fuel prices Jenny (2010) newspaper opinion piece by wellinformed economist Bhatia (2006) government report
Taiwan FTC (2001) report to OECD
Connor (2003) working paper
FTC (1998) widely cited, well-documented report Guersent (2004) report by EC expert
Very good 39% excellent, 36% very good Not rated
Good
Good
Very good
Excellent Good but method a bit vague Possibly questionable yardstick Good
b
Similar estimates from similar sources are sometimes combined in one row. Single source may provide alternative models or methods, hence multiple estimates. c The author’s subjective assessment weighed according to the quality of the data employed, care used in applying the method of analysis, reputation of the authors (if known), and evidence of care in presentation of results (including peer or editorial review). d This cartel, when discovered by the German Federal Cartel Office, had archives extending back to 1902, but as Germany’s current version of its competition law was enacted only in 1958, the overcharge analysis covers only this latter period. e Using the mid-ranges, the average for the 28 very good- and excellent-rated cases is 522%. Sources: Connor (2014: Appendix Tables 1 and 2), which summarize Price Fixing Overcharges Master Data Set, spreadsheet dated October 2013.
a
Potash exports from Canada, Russia, and Belarus River boats, Phnom Penh to Siem Reap, Cambodia Glass, flat, Korea Average 50 high-overcharge cases 1398 other effective cases
Antianxiety drugs, US Euro-zone fees, banks in DE and NL Mobile telephone fees, UK and Germany Distribution, liquefied petroleum gas (LPG), southern Taiwan Air passengers, transatlantic routes, USUK
Cartel Overcharges 303
304
JOHN M. CONNOR
practical substitutes and that near-monopoly supply conditions obtained Connor (2014: Appendix Table 2). Shipowners relied almost exclusively on hemp cordage for their rigging in the late 19th century. With the use of natural manures, farmers worldwide have become dependent on phosphate and potash for fertilization of crops. Radium was highly prized as a novel illuminant for instruments in 19121918 when world production was dominated by a duopoly. Incandescent light bulbs were also quite new consumer products in 19221941 in many parts of the world, and a global cartel effectively created territorial monopolies almost everywhere except Japan. The tungsten carbide cartel was a U.S.-German territorial duopoly for what was then patented and the hardest machine coating material available for four decades.118 (Note also the large number of mutually supporting independent studies of high overcharges.) I conclude that that are no reasons to exclude the very high overcharges from the sample. They are high for reasons consistent with economic reasoning: very inelastic demand combined with duopoly or very tightly organized, monopolistic cartels and no threat of entry.
Peak Overcharges So far only the episodic overcharges have been examined those that refer to the mean price change over all or most of a price-fixing episode. Fig. 10 and Tables 11 and 13 display over 400 peak price effects attained
Fig. 10.
S
l
PE TY AL
L
’C rs Bu ye
ic
e
Fi
xi
ar te
ng
gi ng Pr
Bi
d
R ig
ga l Le
ui lty G
na l at io
In te rn
N
at
io na
l
1.90 1.80 1.70 1.60 1.50 1.40 1.30 1.20 1.10 1.00
Median Ratio of Peak to Episodic Overcharges.
305
Cartel Overcharges
Table 11.
Median Average Peak Overcharge Estimates, by Year and Type, Effective Cartel Episodes.
Cartel Membership Legal Status Bid Episode End Riggingb National International Found Legal Date guilty
Classic Price Fixing
Buyers’ All Cartels Types
Median percenta Before 1890 18901919 19201945 19461973 19741989 19901999 20002013 All years
55.5 33.6 48.0 45.9 27.4 23.7 30.1 33.3
114.5 85.0 72.0 53.0 74.0 50.0 45.0 60.5
46.8 33.3 52.5 49.0 29.4 49.0 50.0 45.0
64.0 71.7 72.0 45.6 315c 16.7 30.5 67.0
21b 50b 42.6 27.5 44.0 40.0 28.2
65 51.3 69.0 59.0 70.0 48.9 18.2 52.9
430.0 7.6 42.8 11.3 21.9 7.6 10.2
59.5 51.3 67.0 49.0 31.0 48.9 38.8 50.0
a
Medians of the point estimates or, where appropriate, of the midpoint of range estimates. Only four peak observations before 1946, so comparisons with classic price fixing are problematic. c Three of the four estimates from the global Mercury cartel only; hazardous to compare with the guilty cartels. Sources: Connor (2014: Appendix Tables 1 and 2), which summarize Price Fixing Overcharges Master Data Set, spreadsheet dated October 2013. b
by cartels the maximum, usually brief markups observed for one week, one month, one quarter, or one year of an episode, depending on the price series available. It is not always clear from a source whether a price effect being reported is episodic or peak; if it is vague, the effect is classified as peak. It is well known that collusive arrangements typically generate price changes that fall short of what a pure monopolist in a blockaded market would set in order to obtain maximum profits. Tacit collusion generally results in prices that are above, but closer to competitive levels than to monopoly levels. While overt collusion should be somewhat more effective than tacit collusion at raising prices ceteris paribus, information failures, potential competition, and cheating also typically result in submonopoly price effects. Because the peak periods are generally too brief for significant changes in the structure of the industry to change, the observed peak overcharges are measures of the short-run market power exercised by cartels when the market structure conditions are closest to optimal and the discipline of the members is at its most cohesive.119 Thus, the peak price effects
306
JOHN M. CONNOR
are instructive about the potential harm that cartels can cause when they are unfettered by coordination problems. From Fig. 10 it is apparent that on average the peak overcharges are 6080% above the episodic overcharges for all types of cartels except buyers’ cartels. Table 11 shows the median peak overcharges in detail over time and across types of effective cartels. The highest median peak cartel markups are from the interwar period.120 For all types of collusion, there is no trend in peak effectiveness over time.121 There is a slight decline over time in peak overcharges of international cartels and a weak positive trend for bid rigging.122 The absence of significant time trends for peak overcharges reinforces the idea that these are proxies for monopoly overcharges. The pattern of peak overcharges across cartel types is similar to that for the effective episodic overcharges (Table 11): (1) in all time periods, international cartels were able to reach higher levels of peak price effectiveness than the “national” cartels on average 86% higher; (2) peak markups are also higher (68% higher) for legal cartels than for sanctioned ones; and (3) cartels that fixed prices or production levels are 85% more harmful as bid-rigging agreements, both overall and in each of the seven time periods. Table 13 provides calculations of how much higher median peak overcharges are compared the median episodic overcharges. Generally speaking, the peaks are about 5070% of the episodic markups. There are no noteworthy trends in these ratios over time. However, the ratio for international cartels are far lower than domestic schemes, and lower for bid rigging than for classic price fixing. These ratios have a couple of interesting interpretations. First, a high peak/average ratio is a rough indicator of price instability during a conspiracy.123 Second, the ratios may be regarded as inverse indicators of pricing efficiency. An efficient cartel is one that has achieved episodic prices that are close to the profit-maximizing (monopoly) price. That is, low ratios may be interpreted as cartels that achieved few operational problems or external challenges from customers or fringe producers. If this latter interpretation is correct, then international cartels and bid-rigging arrangements are relatively efficient. These hypotheses await formal tests.
Overcharges by Location of Cartel Law-makers and antitrust enforcement officials may be interested in the locus of decision-making by the cartels in the sample. Fig. 10 and Table 12
307
Cartel Overcharges
Table 12.
Average Episodic Overcharge Estimates, by Geographic Concepts.
Location of Cartel Members’ Headquarters or Region of Operation
Number of Estimates
Median Mean Overcharge Overcharge
Mean Positive Overcharge
Percent
a
Membership composition USA and Canada Multiple nations in Western Europe (EU)b Single nations in Western Europe Asia and Oceania Africa, Latin America, and Eastern Europe Global (companies from two or more continents) Where collusion took placea USA and Canada Multiple nations in Western Europe (EU)a Single nations in Western Europe Asia and Oceania Africa, South America, and Eastern Europe Global (two or more continents) Geographic extent of collusion Nonglobal: Cross-nationalc Single national: of which local/regional Total
405 184
21.0 29.2
40.8 49.9
43.9 52.8
275 140 50
16.1 20.0 19.4
62.3 41.9 21.3
67.4 44.5 22.6
480
27.5
51.9
54.2
512 141
22.0 25.0
38.3 38.1
40.2 39.8
292 146 61
16.1 20.4 20.0
60.5 37.9 23.4
65.4 40.1 23.4
383
30.4
65.6
71.6
1152 559 976 286
20.1 29.0 20.0 18.3
43.1 56.8 44.1 36.8
44.0 60.8 46.7 39.5
1535
23.0
48.7
51.8
a
Export cartels that drew their membership from one nation or region are counted in that geographic area. However, many national-membership cartels affected world trade; hence, their “market location” is global. b Cartels that operated across several nations of the 27-Member European Union, most of them discovered and convicted by the European Commission. c A high proportion of these cartels are either global (69%) or EU-wide (23%). Sources: Connor (2014: Appendix Tables 1 and 2), which summarize Price Fixing Overcharges Master Data Set, spreadsheet dated October 2013.
classify episodes according to the location of the cartel’s headquarters or the place of residence of the great majority of the cartel’s corporate members. In many cases, corporate membership mix corresponds to a cartel’s geographic field of operations, which is examined next.124
1.67
All years
2.24
1.55 1.42 1.82 1.26 1.70 1.99 2.23
International
Membership
1.93
1.92 1.36 1.17 1.74 1.47 1.99 2.46
Found guilty
Bid Riggingb
2.23
3.64 1.63 1.96 2.55 0.78 1.66 1.42
1.30 1.47 3.00 1.38 2.33 2.22
Ratio of mediansa
Legal
Legal Status
1.99
2.88 1.17 1.86 2.47 3.26 1.96 0.87
Classic Price Fixing
Ratio of Peak/Episodic Effective Overcharges, by Year and Type.
b
Medians of the point estimates or, where appropriate, of the midpoint of range estimates. Excludes zero estimates. Only four peak observations before 1946. = Not available. Sources: Tables 6 and 11.
a
2.69 1.35 2.40 2.44 1.62 1.25 1.29
National
Before 1890 18911919 19201945 19461973 19741989 19901999 20002013
Cartel Episode End Date
Table 13.
0.47
1.00 1.19 0.91 0.49 1.30 0.43
Buyers’ Cartels
2.00
2.70 1.39 1.82 2.18 1.55 2.04 1.94
All Types
308 JOHN M. CONNOR
309
Cartel Overcharges
Cartels may be composed of member companies with headquarters in only one country or one continent; many contemporary cartels are “virtual” joint ventures with no permanent addresses. On the other hand, many early 20th-century cartels established secretariats with professional staffs in London, Zurich, or similar locations. In more recent decades, trade associations or management consulting firms have assisted with cartel operation. In these cases, the geographic locus is easy to identify. Cartels with corporate members from multiple regions are more difficult to classify, but if a supramajority of the companies were headquartered entirely in North America, Western Europe, or Asia, the cartel is categorized in one continent. Global cartels are international cartels that fixed prices on two or more continents; nearly all global cartels aimed at controlling prices in at least Western Europe, North America, and East Asia.
Geographic Spread and Price Effects Here I analyze whether cartel overcharges vary due to the geographic scope of their pricing conduct. Four categories of geographic scope are employed. From most extensive to least, they are: (1) Global (pricing schemes designed to affect two or more continents, (2) Cross-National (price effects in multiple countries in one continent or in world trade), (3) National (price effects intended for only one national market or a portion of it), and (4) Local/ Subnational (a small geographic area, such as one or a few municipalities, counties, or regions of a single state). All local cartels are also national cartels, but not the reverse. Otherwise, the categories are nonoverlapping. Geographic spread of collusion makes a difference in episodic price performance (Fig. 11). Using the Global cartels as the numeraire, the data show that Cross-National cartels achieve 5% lower median and 1315% lower mean overcharge rates. Single-Nation cartels fare substantially worse, with rates 3335% below their Global counterparts. Finally, Local/Subnational cartels face the greatest challenges in raising prices; relative to Global types, small-area cartels generate margins that are 4045% lower than Global. The lesson is clear: Cartelists that are fortunate enough to co-opt all the world’s suppliers into a price agreement are far more likely to profit handsomely than are firms trying to rig bids on a municipal tender. There may be many explanations for this disparity, but the superior ability to global cartelists to deal with entry by fringe suppliers and to exploit geographic price discrimination must rank high on the list. Because the Cross-National cartels suffer little in price-raising ability on average, operating across
310
JOHN M. CONNOR 35 30
Percent
25 20 15 10 5 0 Local & Subnational
Fig. 11.
Whole Nations
Cross-National
Global
Median Average Overcharges by Geographic Extent of Pricing.
customs unions like NAFTA and the EU seems to keep many fringe producers at bay because of distance or trade barriers. Grouping cartels by geographic regions produces parallel results (Fig. 12). Those that operated in only one Western European country have on average the lowest overcharges; cartels in single nations in the ROW were slightly more profitable with median overcharges around 20%. Cross-national cartels those managed across North America,125 the EU, or other adjacent nations have significantly higher overcharges than the single-nation cartels. But those organized across continents were as a group the most successful. In general, cartels able to organize themselves over broader geographical areas were able to achieve higher price effects than those in smaller zones.
Overcharges and Duration The price-fixing overcharges data set includes information on duration for each cartel episode. It is very likely the largest data set on cartel duration ever assembled. An earlier, smaller version of these data were analyzed by Zimmerman (2005) and more recently by Abrantes-Metz and Connor (2009). The relationship of overcharge rates to episodic duration seems tenuous. I examined many plots of the two variables for various time periods and for various types of cartel conduct. Generally, overcharge rates were found to be unrelated to variation in either time or collusive duration. However,
311
Cartel Overcharges 35 30
Percent
25 20 15 10 5 0 Single Africa, Lat. Asia and Nations, America, & Oceania W. Europe E. Europe
USA and Canada
Multiple Nations, W. Europe
Global
Median Average Overcharges by Geographic Location of Pricing.
Fig. 12.
LN (Months)
Logarithm of Duration of Contemporary Global Cartels (1990–2013): Trend is Rising 5.0
R2 = 0.1108
4.0
R2 = 0.1638
3.0 2.0 1.0
Months
0.0 1994
Linear (Months) Poly(Months)
2004 Year Cartel Episode Ended (all observations)
Fig. 13.
2014
Duration of Global Cartels Increasing, 19942013.
for a subsample of 352 contemporary global price-fixing episodes, duration is rising over time (Fig. 13), while overcharge rates for these global conspiracies were holding firm. This implies a need for greater antitrust priority for this class of cartels.
312
JOHN M. CONNOR
Overcharges and Market Size The real affected sales of discovered cartels are becoming progressively larger.126 A commentary in the U.S. Sentencing Guidelines asserts that there is an inverse relationship between the size of affected sales and the height of the overcharges achieved by cartels (USSG, 1987). This commentary implies that judges are authorized to approve fines for criminal price fixing by cartels with large affected sales that are smaller per dollar of affected sales than for members of cartels with small affected sales. No conceptual or empirical justification is provided for this assertion in the Guidelines themselves.127 Moreover, subsequent empirical evidence does not support a positive market sizeovercharge connection.128 Bolotova’s (2009) regression analysis of a large sample of modern international cartels finds that affected sales is unrelated to cartel overcharges (Table 5).
DECISIONS OF ANTITRUST AUTHORITIES Economists versus Courts or Commissions Are there systematic differences between overcharge estimates made by economists the ones reported above and those resulting from a judicial process? The answer to this question is important for the policy relevance of the present study. If the estimates taken from social science publications differ significantly from the conclusions of juries, judges, or commissions, then policy makers may be skeptical of the overcharge estimates in this study as guides to assessing current anticartel enforcement or proposing changes in such enforcement. A survey of final verdicts of U.S. courts in collusion cases finds that 25 collusive episodes had a median average episodic overcharge of 21.6% and a mean overcharge of 30.0% (Connor & Lande, 2005).129 The 9 cases that reported peak overcharges produce a median peak overcharge of 71.4% and a mean peak overcharge of 130%. All but 5 found that the cartel had raised prices by more than 10%. Due to the small number of final verdicts, it would not be meaningful to analyze these verdicts in even smaller groups. By comparison, the 327 estimates for North American cartels had a median episodic overcharge of 21.0% and mean overcharge of 38.8%. Thus, the median averages from both sources are extremely close, but the mean is slightly higher from the economic studies.
313
Cartel Overcharges
25
Percent
20 15 10 5 0 Single Nations, W. Europe
Fig. 14.
USA and Canada
Asia and Africa, Lat. European All Oceania America, Commission Authorities E. Europe
Median Average Overcharges form Antitrust Authorities’ Decisions.
Fig. 14 and Table 14 combine the U.S. court survey above with other episodic overcharge estimates derived from cartel decisions by other antitrust authorities.130 There are 485 such observations from 38 antitrust authorities 32% from analyses of guilty findings of U.S. and Canadian courts, 24% from decisions of the European Commission that imposed fines on EU-wide cartels, 20% from commissions ruling on cartels that operated in single European nations, 20% from Asian and Oceania antitrust authorities, and 3.5% from the ROW. Besides U.S., Canadian, and EC decisions, there are relatively large numbers of observations from decisions by the UK Monopolies Commission in the 1950s and 1960s and the Fair Trade Commissions of Korea, Japan, and Taiwan. Most of the decisions are from decisions that fined international cartels discovered since 1990. Texts of most of these decisions can be found on the Web sites of the authorities or in various searchable law archives (LexisNexis, WestLaw, the Official Journal of the European Communities, EUR-Lex, and the like). In some cases, press releases or press summaries contained sufficient information to calculate an overcharge, but more commonly an analyst used the product definition, affected sales, and conspiracy dates in the opinion and applied this information to prices from a third party to calculate an estimate. As in the case of U.S. final verdicts, only a small minority of available decisions contain the appropriate quantitative data.131 The median episodic overcharge from the 320 authority-decision-related estimates is 20.0%, and the mean is 40.8% (Table 14). The median and
314
JOHN M. CONNOR
Table 14.
Cartel Episodic Overcharges Derived from Decisionsa of Antitrust Authorities.
Location: Antitrust Authority
North America: US, 18981911 US, 19481973 US, 19801999 US, 20002013 US, 18982013 Canada, 19452013 European Union:b European Commission, 19741990 European Commission, 19911999 European Commission, 20002013 Nations of Western Europe: Belgium Denmark France Germany Hungary Iceland Italyc Netherlands Norway Poland Portugal Slovakia Spain Sweden Switzerland United Kingdom: UK Monopolies Commission, 19511957 OFT and other UK, 19902013 Asia and Oceania: Australia China India Indonesia Israel Japan Korea Pakistan Taiwan Turkey
Number of Observations (Episodes)
Median
Mean
156 9 8 13 120 150 6 117 11 19 86 96 1 3 16 10 2 2 14 3 5 1 4 1 6 3 2 31 24
17.7 22.5 16.8 18.0 16.9 17.7 37.3 25.1 25.0 22.5 26.7 17.2 21.0 12.0 19.5 11.5 13.8 50.1 75.0 8.8 9.0 28.0 62.5 24.9 15.0 8.3 78.6 16.9 13.4
48.4 31.2 135.5 155.3 31.1 47.5 71.1 32.2 31.7 22.9 34.5 33.0 21.0 16.0 22.1 20.5 13.8 50.1 83.8 38.9 17.1 28.0 53.1 24.0 13.3 12.2 78.6 63.1 74.3
7 95 4 2 2 6 1 15 44 5 13 2
20.6 20.0 10.5 21.1 42.3 46.1 120.0 28.0 17.9 24.2 25.0 115.0
24.7 36.8 10.1 21.1 42.3 72.9 120.0 25.2 31.5 29.9 46.7 115.0
Percent
315
Cartel Overcharges
Table 14. Location: Antitrust Authority
Vietnam Africa, Latin America, Eastern Europe: Brazil Egypt Latvia Lithuania Mexico South Africa Total
(Continued )
Number of Observations (Episodes)
Median
Mean
1 17
20.0 25.0
20.0 32.7
1 3 1 2 2 4
11.3 20.9 2.7 27.8 32.6 18.7
11.3 25.6 2.7 27.8 32.6 20.1
485
20.0
40.8
Percent
a
Most decisions have a single estimate reported by or interpreted by one person, but several decisions have alternative estimates (or models) by single authors, and some have single estimates by multiple authors. b This is shorthand for collusion across two or more of the 27 nations that form the EU today. c Nine observations are from the two Infant Formula cases. I am informed by an economist familiar with these cases that while the authority strongly suspected overt collusion, it could not find documentary proof. Nevertheless, the participants were fined. Sources: Connor (2014: Appendix Tables 1 and 2), which summarize Price Fixing Overcharges Master Data Set, spreadsheet dated October 2013.
mean overcharges in Table 14 are close to the full-sample median of 23.0% (Table 5) and mean of 48.7% (Table 9), respectively. Overcharges from all jurisdictions are negatively skewed.132 Moreover, the relative geographic pattern in Table 14 is parallel to that of Table 12; that is, median overcharge rates are highest for multiple-nation EU cartels, lowest for singlenation European cartels, and about 20% for all the other continents.133 In three jurisdictions, there are enough observations to examine changes over time. In North America, median overcharges from before 1990 are slightly higher than from subsequent periods. Similarly, the UK Monopolies Commission’s reports produced very cautious estimates compared to more recent UK cases. The EC’s recent decisions suggest higher overcharges than earlier ones. I conclude that, on the whole, estimates of the height of overcharges developed from decisions of antitrust authorities around the world differ little from estimates derived from other economic studies. The overall median overcharge of the 485 legal decisions (20.0%) is about 15% lower than the remaining sample estimates.
316
JOHN M. CONNOR
Cartels Targeted by Class Actions In American and other common-law regimes, jurisprudence and counsel for private plaintiffs are commonly regarded as complementary to the anticartel efforts of the federal and state attorneys general (Lande, 2010, p. 9). This follows from the increasingly outdated view that private suits “followon” after criminal convictions have made the private suits relatively easy to prosecute. In fact, almost half of U.S. private damages suits do not follow DOJ convictions (Connor, 2012; Lande & Davis, 2006, 2008). So, what kind of cartels do plaintiffs sue these days? I informally analyzed the overcharges of the cartels that resulted in the 50 largest price-fixing settlements from class actions in North America during 19902012 (Connor, 2012, Table 1). They are ranked according to the amount of cash settlements received by direct or indirect purchasers in any jurisdiction (expressed in $2012). Of these 50 largest, 258 overcharge estimates are available for 41 cartels. The median overcharges of these 41 cartels were 29.4%, or 30% higher than all guilty cartels during the same period (Table 5). Except for being slightly less durable (7.2 years) than the average cartel (8.2 years), there were no other obvious differences between cartels that settled and other contemporary cartels. Hence, on balance, cartels generating larger dollar injuries are being targeted by plaintiffs.
SUMMARY AND IMPLICATIONS Summary This article’s major goal is to collect and analyze the largest possible body of serious, quantitative estimates of price-fixing and bid-rigging overcharge rates. From several hundred publications dating to 1888, I assembled 2044 such estimates that belong to 532 cartels functioning during the past three centuries. The primary finding is that the median134 episodic cartel overcharge for all types of cartels over all time periods is 23.0%. It is lower for cartels with solely domestic membership (18.2%), higher for international cartels (25.1%), and highest of all for global cartels (30.4%). Overcharges from courts and commissions are slightly lower than from social science analyses. Cartel overcharges are skewed to the high side, pushing the mean overcharge for all successful cartels to 52%. The “peak” cartel
317
Cartel Overcharges
overcharges in the sample are typically almost double those of the long-run averages.135 This article’s findings are generally consistent with the few, more limited works that comment on cartel overcharges.136 Six previously published economic studies with samples ranging from 5 to 38 overcharges report a simple average median overcharge of 28% of affected sales. Moreover, the social science and legal sources yield generally similar estimates. The authors’ professions, types of publications, years of publication, intensity of peer review, and analytical estimation methods incorporated in the sample vary greatly. There is some indication that estimates prepared from the yardstick method are higher than other approaches and that estimates appear in government publications are lower than others.137 Otherwise, however, extensive examinations of variation in overcharge rates across such categories give no reason to regard any subset of the sample as inherently biased or unreliable.
Implications for Economics The great majority of economists, whether swayed by collusion theory or by empirical evidence, roundly condemn cartels. Yet, there is a small minority view among industrial-organization economists that there is little evidence that cartels injure the markets in which they operate. Cartels, they believe, rarely raise prices significantly above noncollusive levels. Moreover, even if cartel price effects are significant, cartels are such fragile coalitions that the harm to the allocative functions of markets is negligible. Finally, they believe that the negative static allocative effects of cartels are counterbalanced by two forces: improved static productiveness and increases in the dynamic performance of cartelized industries through increased productivity growth.138 In sum, these critics dismiss the importance of the cartel phenomenon and, by implication, the relevance of economic cartel studies. In this section, I briefly respond to the empirical validity of these criticisms.139 In defending the value of empirical studies of cartels, I must once again mention a great limitation of the behavioral social sciences: one cannot observe the unobservable. Increasingly, since the middle of the 20th century, most cartel managers have gone to great lengths to hide their illegal joint ventures from public view. Consequently, in the past several decades, empirical studies of cartels have been limited to analyzing samples of discovered, punished cartels. “… [W]e know a great deal about cartels that
318
JOHN M. CONNOR
get caught, but very little about those that escape detection” (Carlton & Perloff, 2005, p. 127). These samples may not be representative of the population of all cartels. Effective clandestine cartels may well have better managers, greater endurance, and superior financial returns than the putatively inept discovered cartels. Equally plausible are scenarios in which secret cartels are fragile coalitions that escape the notice of buyers and regulators by keeping their price effects modest; that is, undetected cartels are relatively ineffective in generating large profits. Several responses can be made to concerns about “sample selection bias” (nonrepresentative cartel samples). First, this survey’s sample is unparalleled in its extraordinarily large amount of data spanning centuries. The historical depth of the sample suggests that time is not a source of potential bias. Large numbers of the cartels in the present study operated in legal environments with little or no fear of prosecution or severe monetary penalties: they predate the current era of high penalties.140 Second, the present study distinguished between overcharges of “guilty” cartels versus unsanctioned cartels. These categories mimic, however imperfectly, discovered versus undiscovered cartels. There is no great difference in overcharge rates between the two categories. Third, there are still legal cartels to be studied today. Samples of legally registered export cartels and governmentsponsored cartels also tend to find evidence of positive price effects. Fourth, controlled laboratory market experiments find ample evidence of strong price effects when the conditions are correctly specified. Fifth, even if undiscovered cartels are indeed different from detected cartels, international discovered cartels are the most appropriate sample for studying the influence of competition laws. In short, the absence of sample selection bias seems just as likely as its presence. Beliefs about the fragility of collusive conduct are driven by cartel theories that focus on the profit incentive that individual cartel members have to cheat on price agreements. While this incentive is undeniable, so is the creativity of cartels that create credible punishment mechanisms.141 The empirical reality is that durable cartels are observed in the great preponderance of quantitative studies. Duration is bimodal, with a large number lasting less than a year and the remainder much longer lasting (Levenstein & Suslow, 2006, pp. 4445). Median duration of cartels is typically five to seven years, but the median life of international and global cartels is higher, probably because of smaller fringe competition and higher profits from geographic price discrimination. Moreover, the mean cartel duration is higher than the median because some cartels last for many decades.142 Finally, although data constraints are especially severe, most recent economic analyses of investment efficiency or productivity change do not
319
Cartel Overcharges
support a sanguine view of cartels on this score.143 One intensely studied phenomenon is legal German commodity cartels. Regression analyses of output of a large sample of German coal-mining firms determined that productive efficiency did not change when they joined the Ruhr cartel during 18831913 (Burhop and Lu¨bbers, 2008). Audretsch (1989) collected information on a large number of post-1945 German cartels; he showed that cartel formation resulted in lower output, not lower costs.144 Blanckenburg and Geist (2011) analyzed data from the German cement cartel of 19812002; during collusion this cartel experienced significant price increases, changes in price dispersion, and declines in cost efficiency. Another heavily researched natural market experiment is the temporary introduction of government-supervised cartels in the United States under the National Industrial Recovery Act (NIRA) in 1933. Although Bittlingmayer (1995) found no output changes due to cartelization, this finding is not supported by several other studies. After mid-1934 when federal compulsion flagged, many of the cartels fell apart; however, those that were able to implement open-price filing (Krepps, 1997), those industries where firms had symmetric costs (Alexander, 1994), and those cartels with viable self-enforcement mechanisms (Alexander, 1997) did experience output contraction.145 One might think that higher profits from collusion might result in increased industry investment. Peters (1989) and Steen and Sørgard (1999) do observe this in two cartels. However, Connor (2008, p. 205) displays internal capacity data for the lysine cartel that shows more plant investment before and after collusion than during collusion. Levenstein and Suslow (2006, p. 85) conclude that the effects of national cartel policies have no clear effects on national economic productivity and development. However, rigorous empirical research on the dynamic effects of cartelization is just beginning.
Implications for Public Policy The results of the survey have significant policy implications. First, there is a minority view among antitrust writers that there is little evidence that cartels raise prices significantly for a period long enough to justify anticartel laws and, especially, contemporary cartel penalties.146 Consequently, they argue for the repeal or scaling back of the fines or damages that result from collusion. This survey’s results, which are based upon an extraordinarily large amount of data spanning a broad swath of history of all types of private cartels, sharply contradict these views. In fact, the data suggest the
320
JOHN M. CONNOR
opposite. Mean overcharges are several times as high as the average level presumed by the U.S. Sentencing Commission (i.e., 10% of sales) and similar guidelines of other antitrust authorities. Generally speaking, sanction guidelines aim at optimal deterrence of cartel formation (Connor & Lande, 2012). More specifically, antitrust enforcement generally seeks general deterrence rather than specific deterrence. Hence, rules for imposing cartel fines ideally combine a proxy for a cartelist’s antitrust damages (typically its affected commerce) with some average multiplier of cartel harm.147 It is not clear which of the many concepts of “average” are the most appropriate for an antitrust authority to employ in designing effective and transparent sanction guidelines. What is clear is that the median averages discussed in this article are inappropriately conservative guides to cartel fines. Alternative and perhaps superior mean averages are given in Table 7. Mean episodic overcharges are more than double the respective median averages. Moreover, if authors failed to compute overcharges with competitive sales instead of actual sales,148 then the mean overcharges attained by cartels were around 100% much higher than the medians of 2325%. Second, the relative injuriousness of bid rigging is sensitive to the measure of central tendency employed. Compared to other forms of collusion, median bid rigging overcharges were generally 25% lower, but mean episodic bid-rigging overcharges were 1124% higher than classical price fixing. These results suggest that antitrust sanctions’ guidelines should not necessarily treat bid rigging per se more harshly than other forms of collusion. Third, international cartels are typically more destructive of competitive market forces than domestic conspiracies. Connor and Lande (2012) propose raising the overcharge presumption for U.S. fines to 20% for hardcore cartels.149 Despite the evident increases in cartel detection rates and the size of monetary fines and penalties in the past decade, a good case can be made that current global anticartel regimes are under-deterring (Bush et al., 2004; Connor, 2005). Global cartels especially are more difficult to detect, have less fear from entry of rivals, achieve higher levels of sales and profitability, and systematically receive weaker corporate antitrust sanctions than comparable domestic cartels. Base fines of 20% of cartelists’ affected commerce, even when adjusted by significant culpability multipliers,150 will do little to deter most of these cartels. Fourth, the preponderance of the evidence suggests that when cartels collapse because of the direct intervention of antitrust authorities, prices
321
Cartel Overcharges
both in the short run and long run typically do decline. Nor does antitrust enforcement that suppresses collusion seem to have adverse effects on either static or dynamic industrial efficiency. See, for example, the research in Buccirossi, Ciari, Duso, Spagnolo, and Vitale (2013) showing that competition-law enforcement directly spurs total factor productivity growth. When the effects of private suits are factored in, it is clear that the U.S. court system is already shouldering the bulk of the world’s burden of punishing international cartels and their managers; moreover, more severe prison sentences for executives have relatively little additional deterrence power (Connor & Lande, 2012). This survey suggests that overcharges generated by cartels discovered in most jurisdictions are higher than North America-centered cartels. Consequently, barring multilateral antitrust treaties, anticartel laws, and fine-setting practices abroad are in even greater need of strengthening. The surge in EU cartel fines (by both the EC and the EU’s National Competition Authorities) and the rising intensity of enforcement in 50 more jurisdictions since 2000 will marginally improve cartel deterrence. But with virtually no private rights of action or incarceration outside North America, total penalties are likely to remain suboptimal for quite some time (Connor, 2010). One sanguine development is that for most types of cartels there are secular reductions seen in cartel markups observed. Because the post-1990 era has been the period with by far the highest level of fines imposed, this decrease is consistent with the theory of optimal deterrence. It also suggests that the recent worldwide trend toward the intensification of cartel penalties has ameliorated cartel injuries. If procedures for calculating criminal fines correspond more closely to the actual levels of cartel overcharges, monetary sanctions against price fixing will more closely provide optimal deterrence.
NOTES 1. I eschew the term “success” used by many authors of cartel studies, because it connotes the financial performance of price-fixing activity from the point of view of the cartel managers or the participating companies. “Effectiveness,” on the other hand, seems to be consistent with (and inversely related to) the social welfare perspective embedded in economics. 2. Longevity, also called duration, measures the life span of a cartel or, if it has more than one, the length of time of one episode. Some researchers use the term stability synonymously with duration, but more commonly it refers to the absence
322
JOHN M. CONNOR
of price wars or other reversions to competitive conduct during a cartel’s time span. Stability is perhaps equivalent to low variation in a cartel’s “discipline,” where discipline may be measured by how close a cartel’s selling prices are to its target prices. Efficiency can refer to static allocative efficiency or, rarely, to technical efficiency or dynamic efficiency. 3. The undercharge from a buyer’s cartel is symmetrically defined as a price effectuated by buyer power of an input purchased by companies acting as a cartel. For details, see a fine legal-economic treatise on monopsony and oligopsony (Blair & Harrison, 2010). 4. Private cartels are those not protected by treaties or sovereignty, and “hardcore” is overt price-setting or quantity-setting conduct. Such cartels are subject to the most severe penalties. 5. The term overcharge is little used in economic discourse. For example, the magisterial New Palgrave lists it nowhere (Eatwell, Milgate, & Newman, 1987). In contrast, a basic handbook on antitrust damages prepared by a committee of lawyers and economists has a long chapter devoted entirely to overcharges (ABA, 2010, chapter 7). However, the overcharge rate has close correlate in the well-known economic measure of market power, the Lerner Index (discussed in Note 58). 6. Antitrust damages are legal remedies for persons (natural or business) that are injured by prohibited anticompetitive conduct of other persons (ABA, 2010, p. 3). While many remedies can be ordered by courts, the most common are monetary payments to compensate victims for their losses (“make them whole”). These are also known as objective, special, or single damages. Injunctive relief in the form of constraints on future conduct by the defendants is sometimes seen. Authorities may also impose punitive costs on the perpetrators, for example, the portion of treble damages awarded above single damages. Overcharges incurred by buyers are only partial damages. Potential buyers who reduce or eliminate their purchases are also injured, but the latter effect is not an overcharge. Economists refer to this consumer loss as the deadweight loss. Courts generally do not regard the harm inflicted on buyers priced out of the market compensable harm because it is difficult to identify these particular victims and because of the presumed difficulty of accurately calculating the deadweight loss. (However, the State of Mississippi’s antitrust law does allow for harm to the State’s economy, which might reasonably be equated with the deadweight loss). A solution to this conundrum would seem to be for courts to allocate additional cy pres awards of 1020% of the value of recoveries. (See Connor and Lande (2010) for the derivation of these percentages.) 7. The benchmark is referred to as the “but-for price” the market equilibrium price that would have been observed were it not for the overtly collusive conduct of the sellers. The benchmark may be the purely competitive price, or it may be a somewhat higher price generated by legal tacit collusion by companies in an oligopolistic industry. 8. It is also easy to convert the (incorrect) ratio of overcharge to affected sales (OV/AS) to the correct one. Let OR be the overcharge rate. Then OR = 1/(1 − OV/ AS). For example, if the overcharge is $5 and affected sales $10, then the true OR is 1/(15/10) = 1/0.5 = 100%. This shows that if reported overcharge rates are computed using affected sales, the true overcharge rates are being under-reported.
Cartel Overcharges
323
9. The but-for sales might also be output under Cournot or some other reasonable noncooperative oligopolistic conduct, which would also be considerably smaller than collusive sales. In the lysine case, the conspirators twice reverted to prices that were slightly below the long-run marginal cost of the industry leader. 10. The follow-on U.S. private damages litigation frequently adds time to the period of time mention in DOJ plea agreements. Also, in international cartel cases, the durations in decisions of other antitrust authorities tend to be longer than the durations for the same cartels negotiated in U.S. plea agreements. 11. An overt collusive agreement is a contract that is the result of observable, explicit communication between the parties. The contract may be a written document, a verbal unwritten agreement, a “handshake” (or “gentlemen’s”) agreement, a cryptic or encoded message, or even simply body language (a “wink and a nod”). In some cultures, silence at the conclusion of a meeting at which consistent proposals were made may indicate a consensus agreement. In a jurisdiction with no antitrust laws or one that provides an industry exemption, the contracts may be publicized and may be enforceable in a court of law. In jurisdictions with anticartel laws, such contracts are usually hidden and are enforced only by the cartel members themselves. The need for self-enforcement of a secret agreement is the unique economic feature of contemporary cartels. 12. Customers are direct buyers and they are usually industrial buyers, but overcharges normally will be passed in whole or in part to final consumers as indirect buyers. If cartels improve technical or dynamic efficiency, this may offset the buyers’ losses. The EU and some other jurisdictions permit innovation cartels in those rare occasions when the fruits of innovation passed on to consumers outweigh the static losses. 13. Technically, as a matter of economic and statistical principles, collusion can and does affect prices in ways other than a correctly measured overcharge. Keep in mind that Pm and Pc are ordinarily prices averaged over the collusive period for several hours or several decades. However, there is a burgeoning literature that focuses on the dispersion of prices that result from collusive conduct (Connor, 2005). In statistics, the mean average is but the first of four “moments” (or formulas) that statistically describe a sample of prices; the other higher order moments are variance, skewness, and kurtosis. Cartels, for example, tend to reduce price variance and increase skewness. Theoretically, cartels can significantly affect price dispersion without creating an overcharge, but empirical works show that changes in mean prices are usually accompanied by changes in dispersion (Connor, 2004d; Connor, Bolotova, & Miller, 2008; Blanckenburg, Geist, & Khodlodilin, 2010). Analyses of price-dispersion effects have promise in the detection of cartels and in proof of antitrust damages. 14. In large U.S. markets for manufactured products, the deadweight loss is typically one-fifth to one-tenth as large as the overcharge, and the two losses are highly correlated (Peterson and Connor, 1995). Lande and Connor (2012, pp. 457461) determined that from the few good studies available, the ratio was more in the 320% range. 15. This picture is simplified. Real-world distribution channels may be lengthened if there are multiple sales from distributor to distributor, the cartel members may sell their products as components to other manufacturers for final assembly, or the channel may be foreshortened by manufacturerdistributor integration. Or, the
324
JOHN M. CONNOR
chain may be much shorter than the example above, if, for example, consumers buy directly from cartelists via Internet sites. 16. Linear demand and supply curves, a homogeneous product, constant returns to scale, and fixed proportions in input use. See Harris and Sullivan (1979). In an extreme case of a monopolistic wholesaler and retailer, the pass-through rate from a manufacturing cartel to consumers is 0.5 × 0.5 = 0.25 or 25%. If the chain of sellers in the vertical distribution system is long, then a pass-through rate below one will shrink greatly before it reaches the consumer. If the distributors are competitive and the product is highly differentiated like cigarettes, then consumers could bear a 120% overcharge or higher. 17. An early (2004) version of the present study is cited (Levenstein & Suslow, 2006, note 96). 18. The dominant U.S. textbook in the 1990s devoted 15 pages to cartels (Scherer & Ross, 1990, pp. 235248, 258). Its market successor, about the same total length, spends 13 pages (Carlton & Perloff, 2004, pp. 128131, 140145, 148150). 19. I exclude, of course, antecedents of this article by the present author. 20. In a personal communication to the author in 2006, Terry Calvani (former Commissioner of the U.S. FTC and of the comparable Irish competition authority) commented on the release of my first working paper on overcharges, saying: “[M]uch of what we thought we knew about cartel overcharges was largely ‘urban legend’.” 21. Criminal filings are made in cases of per se, covert, intentional conspiracies by participants who are aware of the probable anticompetitive consequences (Hovenkamp, 1999, pp. 585586). More than 95% of all naked cartel cases are brought as criminal actions, but a small number of such cases are, at the discretion of the DOJ, filed as civil matters. 22. The USSC Guidelines start with a base fine double the 10% presumed overcharge and use it in conjunction with the assigned base Offence Level for antitrust offenses. They adjust this offense level by a number of factors, such as whether bid rigging and other aggravating factors were involved, and by mitigating factors as well. This adjustment results a pair of “culpability multipliers” that are between 0.75 and 4.0. The product of the base fine (20% of the affected commerce) and the culpability multipliers results in the fine range that is to be imposed on a cartel member. Thus, the fine range recommended for convicted cartelists is at its lowest 15% and at its highest 80% of affected sales. These fine ranges usually are adjusted downward for cooperation or as a part of the Division’s leniency program. The USSC’s Commentary also notes that “In cases in which the actual monopoly overcharge appears to be either substantially more or substantially less than 10%” it might not employ the 20% base fine. But in practice, the DOJ almost always uses the figure of 20% of affected commerce as their starting point in their criminal fine calculations. 23. See U.S. Sentencing Commission Guidelines For the United States Courts, 18 U.S.C. Section 2R1.1, Bid-Rigging, Price Fixing or Market-Allocation Agreements Among Competitors, Application Note 3. 24. See United States v. Hayter Oil Co., 51 F.3d 1265, 1277 (1995). The court noted: “The offense levels are not based directly on the damage caused or profit made by the defendant because damages are difficult and time consuming to
Cartel Overcharges
325
establish. The volume of commerce is an acceptable and more readily measurable substitute … .” 25. In a statement to the Commission, Assistant Attorney General Ginsburg stated that “the optimal fine for any given act of price-fixing is equal to the damage caused by the violation divided by the probability of conviction … such a fine would result in the socially optimal level of price-fixing, which in this case is zero” (USSG, 1986, p. 14). He stated his judgment that “price fixing typically results in price increases that has harmed the consumers in a range of 10 percent of the price …” and that these violations had no more than 10% chance of detection (ibid., p. 15). Connor and Lande (2012) comment extensively on the appropriate detection probability for cartels and the other standard assumptions of the simple optimal deterrence model. For example, they consider the implications of risk-loving behavior of cartel managers or corporate cartelists in place of the usual assumption of risk neutrality (ibid., pp. 432455), and the implications of the present value of expected future monopoly profits and cartel penalties rather than nominal values (ibid.). 26. Although the Gallo study covers only U.S. fines, cartel fines in other jurisdictions were negligible before 1990. 27. Adler and Laing are correct that the fining standards of the DOJ do not compute fines simply as a function of damages, but rather as a function of the company’s affected commerce, which is loosely related to damages. However, these authors do not document their claim that antitrust fines are harsher than other corporate crimes. In recent years, corporate fines for fraud and environmental crimes have greatly eclipsed antitrust fines. 28. Denger appeals primarily to an increase in settlement rates in treble-damage direct-purchaser suits to establish the unfairness of the high fines imposed on corporate price fixers, an increase that, he believes, cannot be explained by increases in overcharge rates. He cites about 8 domestic U.S. law cases that settled for 24% of sales in the 1970s and one international case in 2001 that settled for 1820% (pp. 34). It is argued later that settlements are inappropriate evidence of overcharges. 29. Klawiter contrasts enforcement powers in the late 1990s with the clearly suboptimal maximum fine of $10 million available to the DOJ in the 1970s and 1980s. 30. Japan’s Antimonopoly Law was seriously weakened after 1953 by a perceived need for centralized industrial planning. However, it has been reinvigorated since the 1980s by the growing influence of the country’s consumer organizations and a new appreciation of the efficiency benefits of more intense market competition. Taiwan, South Korea, and other East Asian countries have adopted aspects of Japan’s antitrust law. 31. The practice at the time was for the Council of Ministers to appoint an Advisory Committee comprised of Commission civil servants to develop a report on proposed regulations of administrative practices. Although these regulations were essentially EEC laws, the Parliament had no role at the time. The Commissioner of Competition (a German) is often credited with drafting Regulation 17. 32. Rule 17 was amended in 2004, but these provisions were unaffected. 33. Until recently, under Section 45 of Canada’s Competition Act, fines were limited to C$10 million, but foreign price-fixing conspiracies can be prosecuted under Section 46, which has no fine limit (Low, 2004, p. 17).
326
JOHN M. CONNOR
34. In 1986, the Assistant Attorney General for Antitrust, Douglas Ginsburg, estimated that the enforcers catch less than 10% of all cartels (USSG, 1986, p. 15). If he is correct, optimal penalties for cartels should be more than 10-fold damages. See also the illustration of detection probability in Landes (1983, p. 115, fn. 1). The percentage of cartels that are caught and proven guilty is probably higher since the mid-1990s (Miller, 2009). There is, however, neither evidence nor speculation that it exceeds 33%, either historically or at present (Connor & Lande, 2012, Table 3). 35. Mandatory since their inception, the U.S. Sentencing Guidelines became advisory in January 2005. 36. For larger price-fixed markets “… ten percent is almost certainly too high” (Cohen & Scheffman, 1989, p. 343). They arrive at this conclusion in part by relying on evidence of price-fixing settlements rather than awards made after trial; because settlements are the result of bargaining under uncertainty, reliance upon settlements biases overcharge estimates downward. However, their article is internally contradictory. It cites 710 overcharge observations. Despite the downward bias, the median is in the range of 814%. 37. See Baker (2003), Werden (2003), and Kwoka (2003). According to Kwoka (2003, note 2), Crandall and Winston’s earlier drafts “… endorsed consideration of outright appeal of the antitrust laws.” 38. Severity for nonglobal cartels with international membership is similar but lower in every jurisdiction (Table 11). 39. If adjusted for inflation and the time value of money, the 40% figure would be reduced by 2040%. 40. This section summarizes a fuller treatment of the conceptual development of the concept of and empirical study of cartels by economists contained in Connor (2014), which is itself a revised and expanded version of Connor (2007b, pp. 6572, 129135). 41. Almost half of the publications seemed promising, but ultimately contained no useful information. 42. The Bibliography section lists about 780 sources with useful information about private cartels. The 514 unique citations used for quantitative overcharge estimates are listed in Connor (2014: Appendix Table 2). 43. The exclusive attention to the theories of perfect competition monopoly, and perhaps monopolistic competition (and the absence of oligopoly) prior to the 1920s is illustrated by the dominance of the English language microeconomics textbook of Alfred Marshall (1890). However, a few oligopolistic topics are treated in Marshall’s largely empirical Industry and Trade (1919). 44. Although Cournot’s oligopoly model was published in 1838, it was more than 100 years before it was rediscovered (Martin, 2005, pp. 2627). 45. An issue among economists up to the 1940s was whether cartels raised average prices in a manner consistent with monopolies or whether cartels simply stabilize price movements with no net increase in prices. Liefmann was in the minority that accepted profit maximization as the goal of a cartel. 46. Unlike most of his colleagues, who believed that price or output stabilization were the objectives, Liefmann accepted that raw materials cartels typically did raise prices. However, Liefmann considered the price effects of industrial cartels an open question. While most of his contemporaries considered such calculations impossible,
Cartel Overcharges
327
Liefmann took the position that precision was difficult because of simultaneous changes in demand and supply. The lack of attention to estimates of price effects may also have resulted from an absence of cartel suits in German courts. 47. An interesting exception is the book on Australian trusts by Wilkinson (1914), which grew out of that colony’s 1906 federal competition law modeled on the U.S. Sherman Act (Shanahan & Round, 2008). However, the law’s requirement that collusive conduct had to be proven to have been to the “detriment of the public” led to confusion in the courts. 48. The current world champion for endurance is the Indo-Ceylon-Pakistan Shipping Conference, which was established in 1875 and dissolved by the Competition Commission of India in October 2008 a life of 134 years (Connor, 2009b). 49. Stocking and Watkins had access to the results of a number of major investigations. The Temporary National Economic (or “Kilgore”) Committee published its hearings a few years before their books were published (U.S. Congress 19381940). Other Congressional committees investigated the munitions industry and patent pools. The authors also had information on U.S. criminal prosecutions by the Justice Department of more than 40 international cartels. 50. Stocking appears to have had overall leadership of the team. George W. Stocking was a professor at the University of Texas during 19261947. He was appointed as the economic advisor to the new U.S. Attorney General Thurman Arnold in 1938. Stocking served as the co-chair of the Consent Decree Section of the DOJ through at least 1943 (Mueller, 2007, pp. 187188). It was in the early 1940s that the DOJ investigated the many international cartels that would be formally indicted by the DOJ in 19441948. As there were few, if any, economists employed by the DOJ, Stocking played a role something like the first Chief Economist of the DOJ. Stocking mentored many students who became leaders in the fields of industrial-organization economics, including my mentor Willard F. Mueller (Anonymous, 1976; Marion, 2007). 51. Technically, because one of the defendants was British American Tobacco, the 1911 conviction of American Tobacco et al. was the first U.S. prosecution of an international cartel. 52. I found 22 of these reports had useful overcharges estimates. 53. When the guilty pleas were received in the Philadelphia U.S. District Court in early 1961, nearly every daily newspaper in the United States placed the events on their front page. 54. The conspiracies are notable for their duration (up to 40 years), the as-yet unsurpassed size of the sales involved ($7 billion per year in the late 1950s), the large number of well-known companies involved (General Electric, Westinghouse, etc.), the size of the fines imposed (over $2 million), the size of the damage awards in three trials and private settlements ($400$500 million) from more than 1,900 suits, and the imposition for the first time of significant prison sentences for several top executives. 55. Works about the conspiracy include at least six monographs (Bane, 1973; Epstein & Newfarmer, 1980; Herling, 1962; Smith, 1963; Sultan, 1975; U.S. Congress, 1965). In addition, three journal articles were devoted to the cartels (Finkelstein & Levanbach, 1983; Kuhlman, 1972; Lean, Ogur, & Rodgers, 1985).
328
JOHN M. CONNOR
56. I do not include OPEC in this survey because it was created and enforced by a multilateral treaty organization. 57. George W. Stocking wrote a nontechnical study in 1970 of the oil industry, Middle East Oil, that his biographer calls “prophetic” (Anonymous, 1976, p. 454). 58. The Lerner Index is also computed by starting with the dollar overcharge in the numerator, just as one calculates the overcharge rate, except that the Lerner Index is measured by dividing the overcharge by the monopoly price instead of the competitive benchmark price. That is, the Lerner Index is a margin on the collusive selling price, while the overcharge is a markup on the competitive benchmark price. Thus, for the same cartel the Lerner Index is a smaller number than the overcharge ratio, though the differences are small for small overcharges. The Lerner Index is L = (P − C)/P, where P is the observed market price and C is the but-for or competitive price. Because C is equal to marginal cost in competitive equilibrium, L is also a profit margin on sales. L is zero in perfectly competitive markets and has a maximum value of one. The monopoly overcharge is a markup: MO = (P − C)/C. MO is also zero in perfectly competitive markets, but can approach positive infinity when C is very small. Because P is always greater than or equal to C, MO is greater than L whenever L is positive. If the but-for scenario is perfect competition, the simple algebraic substitution allows one to express MO as a function of L, viz., MO = L/(1 − L). Alternatively, L = MO/(MO + 1). If, however, the but-for state of competition is effective noncooperative oligopoly, then the overcharge conversion will overstate the Lerner Index (Boyer & Kotchoni, 2012). For that reason, we include Lerner Indexes in the sample of overcharges without conversion. This will cause averages of overcharges to be understated. 59. In addition to journal articles, this study draws upon numerous working papers of economists, many of which became journal papers. 60. He does not say, but Sultan’s work may have arisen from a consulting project for the defendants. 61. Of the leading textbooks in industrial organization, Carlton and Perloff (1990) devote more space to cartels than most almost 50 pages out of 852 total pages. This work mentions by name 60 cartels, most of them interwar, international cartels. Other textbooks have far fewer numbers of cartels cited. 62. Thirty-nine observations are Lerner Indexes, not overcharges. If competition is assumed and the indexes are converted to overcharges, the averages raise to 38.9% and 53.2%, respectively. 63. Of the published cartel studies that I found from the periods when cartels operated openly (and for some export cartels up to the present time), about half were discarded because they contained no usable price data. In the Private International Cartels data set, which is comprised entirely of discovered cartels since 1990, for only about one-third of overcharges can be obtained or computed. 64. Of the 360 cartels operating in 2004, 17% were permitted to set conditions of sale, 66% could set domestic quantities or prices, and 15% were export cartels. 65. For example, the “nitrogen” cartel is in fact dry salts of nitrogen used as fertilizer, not the gaseous form. Some authors wrote loosely of a 19th-century “fertilizer” cartel that mined guano in Chile or later mined potash in Germany. The hugely successful “vitamins” cartel is best regarded as a series of overlapping collusive ventures, each of which focused on one of 16 nonsubstitutable products.
Cartel Overcharges
329
66. There is no limit on the types of goods and service cartelized. Even spiritual services can become cartels (Cabolis, Axarloglou, & Chrissidis, 2012). 67. Episodes are in principle different from phases of cartels that give rise to cartels instability. Episodes mark changes in cartel organization, whereas stability is measured by changes in the degree of cartel discipline or cohesiveness. 68. In the rare instances where a cartel kept the market price absolutely constant for the whole episode, the two overcharge concepts will be the same number. 69. A few markets were cartelized by both types; typically, a domestic cartel was expanded to respond to foreign competition. The potash cartel is one example; originally German, it became international shortly after World War I because after World War I potash mines in Lorraine became part of France. A joint FrancoGerman scheme was established to regulate world exports. Thus, after 1918 the two jointly administered national potash cartels became counted as international; however, the earlier pre-1918 domestic German episodes are classified as national. 70. Of course, if an export cartel is composed of companies drawn from two or more countries, then this cartel is categorized in this study as international. Some contemporary export cartels registered in Germany contain companies from several European nations. Price-fixing export cartels maintain the fiction that their activities do not affect prices in the “home country.” Most export cartels cooperate on merchandising or other nonprice matters. For a survey of export cartels, see Levenstein and Suslow (2004). 71. In Europe, bid rigging is generally referred to as collusion involving “tenders.” 72. Only a couple of cartels were oligopsonies. 73. Counted in this category are criminal convictions; adverse decisions of the UK Monopolies Commission, which made recommendations to the government similar to consent decrees; adverse decisions of the European Commission and similar civil authorities; and those cartels that paid court-approved damages. A few unfinished probes by antitrust authorities are placed in this category because 96% of these investigations yield convictions. Since 1990, virtually all the cartels in the sample are guilty; prior to 1990 the ratio is below 60%. 74. Because of the multiple dimensions that must be assessed, it is not unusual for experts to differ on the dates of cartel episodes. 75. Sometimes authors report monetary overcharges along with affected sales, in which case a true calculation of the percentage overcharge can be made (i.e., one that calculates the denominator by subtracting the dollar overcharge from affected sales). More commonly, authors provide a percentage overcharge that is understated because they divide the overcharge by total affected sales during the episode. Readers often are in the dark as to which method of calculation is used. 76. I have preserved these ranges in the appendix tables of Connor (2014), but have used the midpoints of the ranges for the tables in this article. The median ranges, if any, are quite narrow. 77. One rather rough estimate from the 4th century BCE Ancient Greece is also included. For a narrative of this interesting case, see Connor (2007c, pp. 3233). 78. They are also convenient to chart changes in the historical views toward cartels and in methods of analysis. For example, the constant-cost method was popularized around 1890, and econometric modeling of overcharges in the early 1970s.
330
JOHN M. CONNOR
79. There were written laws against price fixing in ancient times (Assyria, for example), in 15th- and 18th-century England, and in revolutionary France. None is known to have been effective against private hard-core cartels. 80. But few economists. The first time the Supreme Court took notice of economists was in the 1925 Maple Flooring decision (Kovacic & Shapiro, 2000, p. 47). 81. Even in recent decades, however, there is a notable absence of empirical publications by European economists working out of European research institutions. Obviously, there are many European analysts, mostly lawyers by training, located in EU and national antitrust authorities’ bureaucracies and performing cartel studies, but few of them publish outside of their governments’ official organs. 82. Raised to $100 million in April 2004; maximum prison sentences rose from 3 to 10 years. 83. UK coal-cartels studies include Ashton and Sykes (1964), Levy (1927), Sweezy (1938), Hausman (1980), and Tan (2003). 84. Records of taxes paid on “sea coal” in London go back to 1213 (Levy, 1927, p. 9). 85. Mine owners who sent coal by coastal ships from Newcastle to London controlled this cartel. The number of mines was quite large at times. Coal was mined in many parts of Britain, but high land transportation costs conferred a monopoly on the Vend over a wide range of delivered London prices. When railroads from the Midlands reached London in the early 1840s, the Newcastle owners’ transportationcost advantage disappeared. 86. When the UK, Germany, and the EEC began requiring registration of cartels in the 1950s, hundreds came forth in each jurisdiction. 87. In the 1970s, many U.S. states passed laws releasing bids on requests for proposals under open record laws. The U.S. Freedom of Information Act and similar national laws elsewhere opened up valuable, large data sets on government tenders. See Hansen (1985) and Athey, Levin, and Seira (2011). 88. They are four early episodes of UK copper smelting (17871867), coal lightermen in London (17001729), a UK books auction (1919), military gunpowder (18511862), power equipment in Japan (19311939), and cast-iron pipes in the United States (18951896). 89. Although there is a dip in 19461990, the correlation between the number of episodic observations per year and a linear time trends highly positive. 90. I do not include national cartels that were fostered by governments (some governments even compelled all the companies in an industry to join) in this data set; likewise, I exclude many international commodity-stabilization schemes that were regulated by government ministries under parliamentary laws or came about because of a multilateral treaty. The second tea cartel in the 1930s, which was authorized by several parliaments of the British Empire and regulated by the Colonial Office, is one example of a “public” cartel. However, I do include a few international cartels with one or more members consisting in part of governmentappointed committee members, government-owned corporations, or governmentsanctioned national cartels, if they were formed by a voluntary agreement among the members. An example is the sugar cartel in the late 1930s. Many of the European export cartels also created national monopolies for their members.
Cartel Overcharges
331
91. U.S. companies apparently believed that patent pooling with foreign firms was legal; others joined cartels indirectly through controlled overseas subsidiaries. U.S. courts judged these and other subterfuges illegal. 92. This ratio may be deceptively high. Many durable cartels straddled eras that bridged shifts in public attitudes or antitrust enforcement. Almost all the sanctioned-cartel observations prior to 1890 derive from the Newcastle Vend, which was not “punished” until the 1830s when a British Parliamentary committee issued an unfavorable report but no further consequences. Later in the 19th century, Parliament again passed laws making coal price fixing illegal, but no monetary sanctions were levied. Similarly, the U.S. Anthracite Coal cartel operated for four decades before it was indicted and convicted, with weak remedies imposed. 93. Export cartels that in theory did not affect the jurisdiction’s commerce were permitted in the United States from 1918 and in most other nations throughout the 20th century. Today, less than one-third of all countries permit export cartels, and many that have an antitrust exemption appear ready to repeal the loophole (Levenstein & Suslow, 2004b). 94. In the last decade, announcements of probes, guilty pleas, and fines on cartelists are more and more to be found in convenient Internet sites and through Internet search engines than formerly. 95. One highly critical response to the sampling methods employed in this article falls into this category. Ehmer and Rosati (2009) state: “Many of [Connor’s] estimates are taken from the works of historians, political scientists, and journalists … rather than from economic studies published in refereed economics journals” (p. 2). They then state that because I have not rejected such publications of “lesser quality” [sic], the sample is fundamentally unreliable, biased, and inflated. They sampled about 10% of the larger overcharge estimates and found one episodic overcharge that was incorrectly computed. It has been removed from this article. 96. A very small number of omitted studies and the reasons for rejection are given in Connor (2014: Appendix Table 3). 97. Alternatively, one might infer that analysts may have increasingly employed techniques that have won court approval as forensically reliable (see Connor, 2004a). 98. The same estimates sometimes appear in multiple publications (see Bibliography). Here I count only the total number of books, articles, and reports that contain one or more original estimates. The undercharges are entered as positive numbers. 99. By “an estimate,” I mean to include a point estimate, single range, or the midpoint of a range. 100. Note that 1536 + 455 + 53 = 2044. Most episodic tables have only 1535 observations because one overcharge is infinity. 101. “Successful” cartels are those with nonzero overcharges (Table 6). In the earliest reports of this research, the median average was 25.0%, but as more observations were added, mostly from cartels ending in 19902013, the median has declined. 102. International membership is also bimodal, rising sharply after 1880 and falling from a plateau after 1940 and repeating this pattern after 1989.
332
JOHN M. CONNOR
103. It is rather odd that the notable surge in discovered international cartels after 1990 came at a time when the profit incentives for cartel formation were at an historic low (Connor, 2003). Of course, if profits declined in the 1980s and 1990s, it is possible that the percentage increase in expected cartel profits may have been at an historic high point. Uctum (1998) presents evidence of just such a decline in the United States, Canada, Germany, and Japan from the 1950s or 1960s to the 1990s. 104. I do not often use exclamation marks in professional writing. This is a most curious phenomenon that demands an explanation. Several experienced plaintiffs’ antitrust attorneys have conveyed to me privately their surprise at the historically low overcharges being estimated by economist experts in high-profile global damages cases. 105. Note that antitrust enforcement was suspended in the United States in 1933, in Germany after the Weimar Republic, and in Japan after the mid-1950s. 106. Globalization (through freer international trade and foreign direct investment) does not seem to be a strong alternative explanation. Most cartels appear in manufacturing. The rise of Asia as the world’s new center of gravity for manufacturing may have played a subtle role in international cartels. Most of these cartels discovered after the mid-1990s (but organized typically from the early 1980s or later) contained non-Chinese Asian companies. It is possible that these firms were more likely to cheat or, more likely in my view, were more likely to have lower long-term profit goals (before and during collusion). Chinese firms have been the biggest spoilers of international cartels since about 1990; if they should become joiners rather than remaining on the fringe, cartel formations will rise. 107. Few international cartels in 19001945 had U.S. corporate members. Those U.S. companies that did join international conspiracies may have believed that they had structured their participation in international cartels in ways that would not run afoul the Sherman Act. 108. One possibility is the rise in exports of manufactures from China. Prior to 2005, there is no example of a Chinese company joining an international cartel. 109. Cohen and Scheffman (1989, p. 345) also cite low normal profits and declining demand. 110. The authors interpret their results in two ways. Firms are more likely to collude when industry conditions cause profits to decline, or cartels that are relatively ineffective at raising prices are also inept at hiding their illegal conduct and, consequently, the most likely to be detected and indicted by the antitrust authorities. 111. Oddly, this term does not appear in Blair and Harrison’s book. They stick to the more rigidly formal economic jargon of a “Buyer Power Index.” They do not present more than one or two examples of empirical buyer-power estimates. 112. In some cases, plaintiffs were denied standing or lost their cases. 113. Market structure variables are far stronger explanatory variables than industry type or geographic location (Bolotova, 2009, Table 5). In a broader sample of cartel episodes, industry was the strongest explanatory group. This is evidence that industry variables capture variation in the structure of supply. 114. The Herfindahl Index is computed from national import values for several years before and after the cartel broke up. It may be understated because in some
Cartel Overcharges
333
cases (probably uncommon) two cartel members may each have plants in the same exporting country and because domestic production in the importing country is ignored. 115. Note that from a legal perspective, each episode is an actionable offense. For the highest overcharges, the implied own-price elasticities of demand are very large. One of the highest overcharges (800%) is for tungsten carbide, for which General Electric had a monopoly in the United States in 19271937. This newly developed material was sold at $453/lb to most customers and at $360/lb to a few favored buyers; up to 1927, Krupp sold it at $50/lb in the United States and during 19271937 at $45$50/lb in Europe (Stocking & Watkins, 1948, p. 132). These numbers imply that the U.S. elasticity of demand was 81.564.8. 116. The reviewer suggested the 200% break point as worthy of special attention. The mean average of all 1,447 episodic nonzero overcharges is 51.02%. When the 49 overcharges of 200% or higher are excluded, the mean average drops to 32.57%, or by 36%. (The median average is very little affected: it falls from 24.8% to 24.0%.) 117. My assessment is based upon a combination of what I know about the quality of the price data available, craftsmanship in applying the method of overcharge analysis, professional reputation of the authors or organization responsible (if known), and evidence of balance in presentation of results (including peer or editorial review). 118. Tungsten carbide was simultaneously invented by General Electric and Krupp Steel in the early 1920s. Only industrial diamonds are harder, but natural diamonds were prohibitively expensive for most industrial applications until artificial diamonds were first marketed in the 1950s. 119. Peak price changes may well be affected by short-run shifts in demand. Exogenous, unanticipated shifts in demand may exaggerate the peak price changes. However, in some cases, these shifts are endogenous. Especially when a wellfinanced cartel felt free to announce a new agreement that buyers perceived as likely to be effective, “panic buying” often ensued, which amplified the purely collusive effect on prices. 120. Approximately one-fifth of the 413 observations available for Table 11 refer to interwar cartels, which have been well studied by economic historians who often had available public commodity-exchange prices. Forty-two percent of the observations on peak prices are for episodes ending after 1989. 121. The correlation of episode end year with peak overcharge for all 413 observations is not significantly different from zero. 122. The correlation over time (the end year of each episode) for international cartels is r = −0.102 and for national cartels r = +0.070; for bid-rigging schemes, there is a weak positive time trend (r = +0.085); but for guilty cartels, legal cartels, and classic price-fixing cartels, there is no time trend. 123. These ratios could be relevant for assessing whether cartels intend to maximize profits through price increases (as most economists assume) or whether the goal is to control variation in their output or prices. Apologists for cartels, particularly those writing about international cartels during the Great Depression, tended to assert that cartels did not aim to raise prices so much as stabilize prices (Marlio, 1947; Pyndyck, 1979). There is little evidence in Table 13 that the interwar,
334
JOHN M. CONNOR
international cartels achieved greater price stability than those ending before or after the interwar period. 124. The major exception is export cartels, which are categorized in their country or region of origin but set prices in the “rest of the world.” 125. Defined here is the United States and Canada, but could include Mexico in many cases because of the absence of formal tariff barriers. Unfortunately, until recently, the Mexican federal antitrust commission did not often prosecute international cartels. Connor and Bolotova (2006) confirm that North American cartels and single nations of Western Europe as a whole have significantly lower overcharges. 126. This can be shown for the past few decades, but it is difficult to know whether this statement applies to cartels throughout the 20th century. 127. The original testimonies about the USSGs are unpublished (U.S. Sentencing Commission, 1987). However, the few empirical studies of cartels with information on price effects available to the Sentencing Commission in 1986 (e.g., Asch & Seneca, 1975; Fraas & Greer, 1977; Hay & Kelley, 1974; Posner, 1976). Eckbo (1976) and Griffin (1989) do not link the prices to cartel size. 128. Appropriate data to test this proposition are contained in Connor (2003, Tables A.1A.12). This working paper develops affected sales and overcharge data for a modest sample of modern international cartels: approximately 92 pairs of such data are available. Sales are in current U.S. dollars and generally fall into the decade of the 1990s. Correlation statistics were calculated for a number of subsamples. The first sample of 50 cartels examined the largest geographic market for each cartel; the coefficient correlating sales and overcharge rates was not significantly different from zero (r = −0.105). To see whether extreme observations might unduly affect the result, I repeated the experiment but dropped first all cartels with $5 billion in sales or more and second all cartels with overcharges of 65% or higher; in both cases, r became closer to zero (−0.065 and +0.019, respectively), which indicates that extreme observations do not account for the low correlations found. Finally, I examined geographic subgroups of the cartels: global, U.S., EU, and other single national markets. The correlations for these four samples varied from −0.17 to +0.24, none statistically significant. 129. Robert Lande and research assistants under his direction in 2004 calculated these figures. Less than 1% of all U.S. published court opinions on price-fixing damages contain both the dollar damages and the affected sales of a cartel. For a discussion of the merits of examining only final verdicts, see Connor and Lande (2005). 130. Sometimes a published decision will mention explicitly an overcharge figure, but more commonly court and commission decisions need to be interpreted. For example, a decision may mention in passing the price series upon which it relied to determine the severity of a sentence, and that series is then interpreted by an economist. The DOJ, U.S. FTC, and federal courts are counted as one authority. However, the EC and the EU NCAs are counted as separate authorities. 131. Guilty pleas and sentencing memoranda of the DOJ and Canadian Competition Bureau almost never mention damages. The EC has fined more than 100 cartels since 1969, but the full decisions are not always published, publication can be delayed for up to five years, and only a small proportion include price data. EC decisions yielded usable information on product definition, affected sales, geographic area, dates of the conspiracy, or other helpful information for 75 episodes.
Cartel Overcharges
335
Similar comments apply to the other authorities’ press releases, newsletters, or formal decisions. 132. With three or more observations, the mean is higher than the median. 133. Although not shown, decisions regarding global cartels resulted in higher overcharge estimates than other types or locations of cartels. 134. All medians presented in this section incorporate all relevant zero estimates and omit peak results unless otherwise mentioned. 135. If one assumes that the peak markups are the result of a cartel having achieved something close to monopoly price levels, then the lower episodic overcharges imply that historical cartels are constrained by substitutes, fear of entry, internal discord, or other factors that frustrate optimization. This is a common finding in studies that measure the degree of monopoly power. 136. All of the relevant estimates in the seven works are incorporated in the sample assembled for this article. 137. Two other types (historical case studies and government reports) tended to be low. 138. Such beliefs seem to arise from theoretical modes of collusion that typically do not allow for communication or contain unrealistically strict assumptions. For example, Telser (1985) has a proof that joint sales agencies improve efficiency, but in fact few contemporary cartels forward vertically integrate. Those in the past that did create sales agencies tend to have longer duration (Levenstein & Suslow, 2006, p. 69). 139. There are theoretical models that prove the possibility of efficiency improvements under overt collusion. 140. In the United States, corporate penalties for cartel conduct were light until the early 1990s, and prison sentences for individuals likewise (Connor, 2004b, 2009b). The EU only began imposing serious cartel fines in the late 1980s and still has no managerial penalties. Outside these jurisdictions, significant penalties appeared in only the past 10 years. There is a good case to be made that even today global cartels function with impunity. 141. The development of infinitely repeated games demonstrates the wide range of conditions over which collusion can persist indefinitely. 142. There are four cartels that endured from 96 to 134 years. Two were ended by antitrust agencies and two by entry. 143. An oft-cited study that the author suggests shows that cartels can be efficient is Dick’s (1992a) study of 16 legal U.S. export cartels. Of the 16, he finds that 6 either lowered prices, raised output exported, or both; of these, only 3 were “efficiency-seeking.” Three cartels raised prices, and seven had insignificant or conflicting effects either way. This seems to be an almost random outcome. See also Gu¨nster, Carree, and van Dijk (2011). 144. For studies that purport to find that late 19th century governmentcontrolled German coal, iron and steel cartels were efficiency enhancing; see Troesken (1989), Kinghorn (1996), and Kinghorn and Nielsen (2004). 145. The aggregate impact of the NIRA codes on U.S. durable-goods manufacturing output was at least negative 10% (Taylor, 2002, p. 8). In a later paper using more disaggregated data, Taylor (2010) finds that about one-fourth of the industries with the most variable production displayed output increases associated with
336
JOHN M. CONNOR
efficiency enhancement (many were dairy products); the remaining three-fourths experienced the expected output reductions. 146. A paper by Crandall and Winston (2003) disparages the effectiveness of antitrust laws and enforcement. It is answered well by Baker (2003) and Werden (2003). Connor (2004c) also criticizes Crandall and Winston’s reliance on a slim sample of facts concerning cartels. 147. While most jurisdictions adopt a single percentage multiplier (within a stipulated range) as a starting point, others have categorical multipliers. The United States uses 10% and 20% to calculate a range. The EC chooses a single number between 15 and 30, depending on gravity. The JFTC has a much higher percentage for manufacturers than for retailers. Connor and Lande (2008) proposed a single percentage that was double for international cartelists compared to domestic cartels. 148. I do not know what share of estimates this correction ought to be to. In most cases when working with dollar overcharges, authors did not reduce affected commerce by the amount inflated by collusion. However, authors computing overcharges with prices need no such correction. So, these figures are to be regarded as upper limits. 149. As an anonymous reviewer of an article derived from material in this article suggested that such changes need to be considered alongside appropriate levels for private settlements. These recommendations are particularly complicated by corporate leniency programs and by the joint fining policies of overseas antitrust authorities for international conspiracies. 150. For a variety of factors, however, very few firms actually pay a fine amounting to 20% or more of the amount of commerce affected. Most violators have their fines reduced by 6080% of the maximums. 151. Some readers have overlooked this selection criterion. For example, Bergman (2008) has said the following: Connor’s results are based on all estimates of price effects that he has been able to find, irrespective of the quality of the underlying analysis … [M]any of the studies are unsubstantiated claims by competition authorities.
“Serious” studies are identified primarily by form of publication. Books, monographs, academic journals, and government publications that are written by professionals and show attention to detail nearly always make the cut. Working papers by scholars that seem to have publication potential are included. Newspaper articles, editorials, and opinion pieces; essays in popular magazines; and blogs are rarely included, unless they happen to be cited approvingly by academic experts. Statements by antitrust officials about overcharges are included only if their methods are explained and are methods normally accepted by U.S. courts. To my knowledge, other than a well-regarded OECD (2003) survey, there are no unfiltered assertions by antitrust officials or any other parties to cartel legal suits in this survey. I have, however, omitted a very small number of egregiously methodologically flawed studies. (See Connor (2014: Appendix Table 3) for the brief list of excluded studies and the reasons for their exclusion.) Admittedly, and by design, seriousness and professionalism is not a high bar, but when authors attempt to pick a sample of studies according to some subjective criterion of “quality,” excluding data points opens them to reasonable suspicion of
Cartel Overcharges
337
tilting results to fit their inevitable prior beliefs (and lose friends). Indeed, Bergman’s analysis itself may be criticized for basing his paper on a sample of 13 overcharges when so many others were available; he displays a strict preference for econometric studies that I have argued may be the counsel of perfection in practical competition enforcement (Connor, 2007c). Meta-analysis is one appropriate method for dealing ex post with heterogeneous quality; minimal, harmless heterogeneity has been detected in this article (see Connor & Bolotova, 2006). 152. In the United States, bringing criminal indictments for only hard-core cartels is a matter of custom, not law. Some hard-core cartels are brought as civil matters because prosecutors judge that the criminal burden of proof cannot be met. Since the 1980s, the EU and most other civil-law jurisdictions have abandoned requiring an effects test and now follow more or less the same conspiracy approach used in common-law countries (Joshua & Jordan, 2003). 153. Wallace and Edminster (1930, Appendix A) provide a convenient chronology of most government-sponsored export-control monopolies through the late 1920s. The Japanese camphor monopoly of 1899, the Italian citric acid monopoly of 1910, the Greek currant monopoly of 1895, and the New Zealand kauri-gum monopoly of 1927 are examples of clearly public, government-managed cartels. 154. In some cases, particularly in the early 1930s, the earlier phases of an international cartel were controlled by national producers’ organizations of private firms that negotiated voluntary quota reductions; when cheating threatened the effectiveness of the cartel, colonial or metropolitan governments stepped in to pass mandatory supply-control legislation. The early phase of the cartel I deem private, but not the latter. 155. Unless available in translation, I have mostly confined this survey to English language sources. Many antitrust authorities now translate their press releases, decisions, and annual reports into English; moreover, members and some nonmembers submit summaries of their annual reports in English to the OECD. The preponderance of sources published after 1945 is explained by the growth of the field of industrial-organization economics and the passage of effective anticartel legislation worldwide. 156. One might speculate as to why this is so. The supply of well-trained industrial economists in Europe is unlikely to be an explanation. The principal European organization for industrial economists (EARIE) equally active in sponsoring meetings the past decade than its U.S. counterpart (IOS), and the EARIE meetings had a good proportion of empirical and legal-economic papers. The structure of academic departments at European and Asian universities may be one explanation of the paucity of useful studies. Compared to U.S. departments of economics, European departments tend to be smaller (perhaps falling below the threshold necessary for collaborative teamwork on large-scale data sets), more focused on IO theory, and have different expectations for Ph.D. dissertations. Perhaps a more important factor is the inability of academics to obtain access to the price data needed to calculate overcharges. Civil antitrust damages cases are unusual in Europe, so the little work being done on cartel overcharges is done in-house by antitrust authorities. Unlike North America, there is little mobility between the staffs of European antitrust authorities and universities or think tanks. Finally, a survey of European and North American industrial-organization economists reveals that there are very few attitudinal differences between the two groups on economic
338
JOHN M. CONNOR
theory, but the former were less inclined to expect economists to influence competition policies (Aiginger, McCabe, Mueller, & Weiss, 2001). 157. The term “antitrust authority” has gained currency in recent years to cover any national or supranational government agency empowered to enforce criminal or civil antitrust laws or competition-law rules. Thus, it encompasses the U.S. DOJ Antitrust Division, the Fair Trade Commissions of many nations, the EC, and the many administrative authorities modeled on the aforementioned. Courts supervising antitrust trials or damages litigation are acting as antitrust authorities. 158. If a credible study of a cartel concludes that it was “ineffective,” I have coded this comment as a zero price effect and included this observation in the averages. Likewise, conclusions that the impact of collusion was “overwhelmed” by natural market forces are interpreted as a zero overcharge. However, vague conclusions that a cartel episode was “effective” in controlling prices are not tabulated in the quantitative summaries. 159. The influences of types of publications and methods of computation are formally analyzed in Connor and Bolotova (2006). 160. I have confined journalists’ accounts of cartels primarily to book-length treatments of cartels, in the belief that such monographs are in-depth accounts of a cartel collected from many sources, some of them anonymous, over a period of time sufficient for the author to provide a balanced account of conflicting claims. Books by journalists typically do not focus on the quantitative economic aspects of the case at hand, so in practice there are relatively few overcharges drawn from these sources in the present study. I rarely include overcharge estimates embedded in newspaper or magazine articles, though some specialists may judge such assertions to be sufficiently reliable to include in their published studies. For example, Elzinga (1984) cites Demaree (1969), and Carlton and Perloff (1990) cite Smith (1963). 161. The danger with this method is that the product yardstick may be a substitute for the cartelized product, and, hence, price-responsive to a cartel overcharge. 162. These data are often proprietary facts revealed during the discovery phase of litigation or submitted to an antitrust authority under compulsory legal processes. In addition to transaction prices of the defendants, production and marketing costs of details of business contracts may be handed over on a confidential basis. 163. Either a dummy variable is included for the assumed collusive period, or the model can forecast or backcast benchmark prices from a noncollusive period. 164. On the other hand, if a cartel operated during a span in which cost conditions (input prices, expansion of capacity, inventories, and technology) were steady and demand conditions (consumer preferences, disposable income, available substitutes, and the like) did not shift, then elaborate econometric models and the more traditional methods will yield the same overcharges. For durable cartels, constancy of all these factors is unlikely. 165. This event may be marked by the signatures of the cartelists on a written contract, by the adoption of a verbal agreement and handshakes all around, or by some similar less formal method of communication. 166. Some early writers were fuzzy about this notion, but Sweezy (1938) and his successors like Eckbo (1976) and Griffin (1989) were meticulous in identifying temporal episodes carefully.
Cartel Overcharges
339
167. This event may be marked by the signatures of the cartelists on a written contract, by the adoption of a verbal agreement and handshakes all around, or by some similar less formal method of communication. 168. Beginning dates may be reported by government antitrust authorities as earlier than the true dates because the standard of proof is high or because they are only interested in dates after their laws take effect. In the United States, the DOJ can make plea bargaining more expeditious by moving forward a provable starting date as a concession to a defendant. Frequently, follow-on private plaintiffs are able to secure damages from a longer episode than that written in a plea agreement. 169. The main source of tips is disaffected directors or employees of cartel participants, the secondary source is outgrowths of other investigations (including Amnesty-Plus applications), and the tertiary source is customer complaints. 170. See the discussion in Sullivan and Grimes (2006, pp. 165233), which shows that in per se cases the plaintiff does not have to prove whether prices rose (or even whether defendants had market power). The issue of whether prices rose can be an element of a rule of reason case, but rule of reason cases do not give rise to criminal fines, so are not the subject of this article. 171. What the documents say is that the percentage of what the defendant and the Government both agree is the amount of damages that prosecutors could prove beyond a reasonable doubt had a criminal trial been held. This is a higher standard of certainty than economic statistical reasoning can usually provide. 172. The most useful sites were: The European Commission (EC); the Australian Competition and Consumer Commission (ACCC); the Canadian Competition Bureau (CCB); the German Bundeskartellamt (BKA); the Fair Trade Commissions of Japan, Korea, and Taiwan; and the competition authorities of Finland, Sweden, Norway, Denmark, the Netherlands, France, Italy, Portugal, and Israel. Many of these authorities seem committed to reminding taxpayers of precisely how harmful the cartels they ensnared have been. 173. Many of the verdicts found were only expressed in monetary amounts, which could only be translated into percentages if trade sources could be found for the often narrowly defined cartelized products. Other decisions gave good or at least minimally acceptable price change data for the affected markets. 174. I read almost 100 EC decisions that imposed fines on cartels (listed in Burnside (2003, Annex 1) and others published since 2003). The UK Monopolies Commission also released detailed reports, and I read about 40 of the ones that declared the cartel was “not in the public interest.” 175. Occasionally, the commission reported an absolute overcharge, and the size of affected sales needed to be estimated. 176. When the exchange mechanism is the double oral auction, buyers pay prices slightly below the full monopoly price (Plott, 1989, p. 1143). 177. However, like monopoly, both buyers’ and sellers’ cartels showed weaker price effects when the trading system was a double auction. 178. Engel also presents two other pricing efficiency indexes, but little is lost by focusing on only MI. 179. MI = (P − Pc)/(Pm − Pc), where P is the observed average outcome price, Pm the monopoly price, and Pc competitive benchmark price. The same indexes can be computed for quantity experiments. MI cannot be converted into Lerner or overcharge indexes.
340
JOHN M. CONNOR
180. However, MI does not decline from N = 4 to N = 5, and the pattern for more than five sellers is irregular. 181. By “anonymous” is meant that sellers face a computerized demand curve, which seems to me equivalent to a large number of buyers. With face-to-face human buyers (i.e., small numbers of buyers), MI is below 14%.
ACKNOWLEDGMENTS Anonymous reviewers of RLE, John Kwoka, and numerous comments on two previously posted working papers resulted in a large number of constructive suggestions, for which I am grateful. My students Jeff Zimmerman, David Ubilava, and Yuliya Bolotova assisted in coding early versions of the Master Data spreadsheet summarizing these data. Numerous users of this spreadsheet have provided corrections and additions.
BIBLIOGRAPHY The publications below include all those from which overcharge estimates were taken or derived plus noteworthy milestones in the intellectual development of empirical cartel studies. ABA. (2010). Proving antitrust damages: Legal and economic issues (2nd ed.). Chicago, IL: ABA Section of Antitrust Law, American Bar Association. Abrantes-Metz, R., & Connor, J. M. (2009, April 35). Determinants of cartel duration. Paper presented at the 7th International Industrial Organization Conference, Boston, MA. Abrantes-Metz, R., Froeb, L., Geweke, J., & Taylor, C. (2006, May). A variance screen for collusion. International Journal of Industrial Organisation, 24, 467486. Abrantes-Metz, R. M., Villas-Boas, S. B., & Judge, G. (2011). Tracking the LIBOR rate. Applied Economics Letters, 18, 893899. Adams, R. F., & Bock, G. M. (1980). Damages in horizontal price fixing cases. Antitrust Law Journal, 49, 141152. Adelman, M. A. (19721973, Winter). Is the oil shortage real? Foreign Policy, 71, 69107. Adler, H., & Laing, D. J. (1997, March). Explosion of international criminal antitrust enforcement. The Corporate Counselor, 11, 1. Adler, H., & Laing, D. J. (1999). As corporate fines skyrocket. Business Crimes Bulletin, 6, 1. AGCM. (2009, May 5). Battery recycling: Antitrust authority imposes total of Euro 13.3 million in fines over two anti-competitive arrangements. Press release, Autorita Garente della Concorrenza e del Mercato, Rome. Retrieved from http://www.agcm.it/eng/index. htm
Cartel Overcharges
341
Aiginger, K., McCabe, M., Mueller, D. C., & Weiss, C. (2001). Do American and European industrial organization economists differ? Review of Industrial Organization, 19, 383405. Aiginger, K., & Pfaffermayr, M. (1997). Looking at the cost side of monopoly. The Journal of Industrial Economics, 45, 245267. Albaek, S., Møllgaard, P., & Overgaard, P. B. (1997). Government-assisted oligopoly coordination? A concrete case. Journal of Industrial Economics, 45, 429443. Alexander, B. (1994). The impact of the national industrial recovery act on cartel formation and maintenance costs. Review of Economics and Statistics, 76, 245254. Alexander, B. (1997). Failed cooperation in heterogeneous industries under the national recovery administration. Journal of Economic History, 57, 322344. Alexander, C. R., Arlen, J., & Cohen, M. A. (2000, December 29). Toward an empirical foundation for federal sentencing practice: An evaluation of corporate sentencing data. Economic Analysis EAG Discussion Paper No. 00-12. Antitrust Division, U.S. Department of Justice, Washington, DC. Alexander, D. L. (1988). An empirical test of monopoly behavior: An application to the hardwood case. Applied Economics, 20, 11151127. Allen, G. C. (1923). An eighteenth century combination in the copper-mining industry. Economic Journal, 33, 7485. Andrews, E. B. (1889). The late copper syndicate. Quarterly Journal of Economics, 3, 508516. Anonymous. (1976). In memoriam: George W. Stocking, 18921975. American Economic Review, 66, 453454. Antitrust Division. (2008). Congressional submission: FY 2009 performance budget. Washington, DC: U.S. Department of Justice. Retrieved from http://www.usdoj.gov/ jmd/2009justification/pdf/fy09-atr.pdf Appelbaum, E. (1979). Testing price taking behavior. Journal of Econometrics, 9, 283294. Arnold, T. W. (1946). Cartels or free enterprise? Pamphlet No. 103. Washington, DC: U.S. Public Affairs Committee. Asch, P., & Seneca, J. J. (1975). Characteristics of collusive firms. Journal of Industrial Economics, 23, 223237. Ashenfelter, O. C., & Graddy, K. (2002, May). Art auctions: A survey of empirical studies. Discussion Paper No. 3387. Center for Economic Policy Research. Retrieved from www.cepr.org Ashenfelter, O. C., & Graddy, K. (2005). Anatomy of the rise and fall of a price-fixing conspiracy: Auctions at Sotheby’s and Christie’s. Journal of Competition Law and Economics, 1, 320. Ashton, T. S., & Sykes, J. (1964). The coal industry of the eighteenth industry (2nd ed.), Manchester: Manchester University Press. Asker, J. (2008, July). A study of the internal organization of a bidding cartel. Working Paper. Stern School of Business, New York University, New York, NY. Retrieved from http://pages.stern.nyu.edu/~jasker/stamps070628.pdf Athey, S., Levin, J., & Seira, E. (2011). Comparing open and sealed bid auctions: Evidence from timber auctions. Quarterly Journal of Economics, 126, 207257. ATTR. (various dates). Antitrust and trade regulation report. Serial publication. Audretsch, D. B. (1989, Fall). Legalized cartels in West Germany. Antitrust Bulletin, 34, 579600. Autorite´ de la Concurrence. Decisions by date. Retrieved from http://www.autoritedelacon currence.fr
342
JOHN M. CONNOR
Bajari, P., & Summers, G. (2002). Detecting collusion in procurement auctions. Antitrust Law Journal, 70, 143170. Bajari, P., & Ye, L. (2003). Deciding between competition and collusion. Review of Economics and Statistics, 85, 971989. Baker, J. B. (1989). Identifying cartel policy under uncertainty: The U.S. steel industry, 19331939. Journal of Law and Economics, 32, S47S76. Baker, J. B. (2003, Fall). The case for antitrust enforcement. Journal of Economic Perspectives, 17, 2750. Baker, J. B., & Rubinfeld, D. L. (1999). Empirical methods in antitrust litigation. American Law and Economics Review, 1, 386435. Baker, W. E., & Faulkner, R. P. (1993). The social organization of conspiracy: Illegal networks in the heavy electrical equipment industry. American Sociological Review, 58, 837860. Baldwin, L. H., Marshall, R. C., & Richard, J.-F. (1997). Bidder collusion at forest service timber sales. Journal of Political Economy, 105, 657695. Bane, C. A. (1973). The electrical equipment conspiracies: The treble damage actions. New York, NY: Federal Legal Publications. Banerji, A., & Meenakshi, J. V. (2004a). Buyer collusion and efficiency of government intervention in wheat markets in Northern India: An asymmetric structural auctions analysis. American Journal of Agricultural Economics, 86, 236253. Banerji, A., & Meenakshi, J. V. (2004b, September). Millers, commission agents and collusion in grain auction markets: Evidence from basmati auctions in North India. Working Paper No. 129. Centre for Development Economics, Department of Economics, Delhi School of Economics. Retrieved from http://www.cdedse.org/pdf/work129.pdf Barbeau, N. (2008). Medical bodies in probe into price fixing. Sunday Tribune, February 17, p. E1. Barbezat, D. (1989). Competition and rivalry in the international steel cartel, 19261933. Journal of Economic History, 49, 435447. Barbezat, D. (1990, March 2325). International cooperation and domestic cartel control: The international steel cartel, 19261932. Business and economic history. Papers presented at the 36th annual meeting of the Business History Conference, Baltimore, MD, Second Series (Vol. 19, pp. 1416), College of William and Mary, Department of Economics Business History Conference, Williamsburg, VA. Barbezat, D. (1994). Structural rigidity and the severity of the German depression: The AVI and the German steel cartels, 19251932. Explorations in Economic History, 31, 479500. Barbezat, D. (1996). The comptoire de sidurgique de France, 19301939. Business History Review, 70, 517540. Barjot, D. (Ed.). (1994). International cartels revisited (18801980). Caen, France: EditionDiffusion du Lys. Barker, T. C., & Harris, J. R. (1993). A Merseyside town in the industrial revolution: St. Helens, 17501900. London: Cass. Barnett, P. G., Keeler, T. E., & Hu, T. W. (1995). Oligopoly structure and the incidence of cigarette excise taxes. Journal of Public Economics, 57, 457470. Beaton-Wells, C., & Brydges, N. (2008). The cardboard box cartel case. Australian Business Law Review, Vol. 36. Retrieved from http://papers.ssrn.com/sol3/papers.cfm?abstract_ id=1406267 Beaton-Wells, C., & Ezrachi, A. (Eds.). (2011). Criminalising cartels: Critical studies of an international regulatory movement. London: Hart.
Cartel Overcharges
343
Becker, G. S. (1968). Crime and punishment: An economic approach. Journal of Political Economy, 76, 169217. Beckmann, M. (2004). Art auctions and bidding rings: Empirical evidence from German auction data. Journal of Cultural Economics, 28, 125141. Bennett, J. M. W. (2011, November 9). In re: Iowa ready-mix concrete antitrust litigation: Memorandum opinion and order regarding plaintiffs’ unopposed motion for an award of attorneys’ fees, the reimbursement of expenses, and incentive awards for class representatives (Case No. C 10-4038-MWB). U.S. District Court, Northern Iowa, Western Division. Benni, A. S., Lammers, C., Marlio, L., & Myer, A. (1930). Review of the economic aspects of several international agreements. Geneva: Economic Committee, League of Nations. Berge, W. (1944). Cartels: Challenge to a free world (1st ed., 2nd printing), Washington, DC: Public Affairs Press. Bergenstock, D. J., & Maskulka, J. M. (2001, May). The de beers story: Are diamonds forever? Business Horizons 44, 3744. (De Beers Botswana Mining Company). Bergman, M. A. (2008). Quis custodiet ipsos custodes? Or measuring and evaluating the effectiveness of competition enforcement. De Economist, 156, 387409. Bernheim, B. D. (2002, May 24). Expert report of B. Douglas Bernheim. In Re Vitamins Antitrust Litigation, MDL No. 1285. U.S. District Court for the District of Columbia. Berry, S., & Briggs, H. (1988). A non-parametric test of a first-order Markov process for regimes in a non-cooperatively collusive industry. Economics Letters, 27, 7377. Besanko, D., & Sperber, D. F. (1990). Are treble damages neutral? American Economic Review, 80, 870887. Beyer, J. C. (2000). Affidavit of John C. Beyer. In Re Vitamin Antitrust Litigation, MDL No. 1285. Beyger, S. (2012). Cartel in the Indian cement industry: An attempt to identify it. Discussion Paper No. 2012-18. Economics (E-Journal). Retrieved from http://papers.ssrn.com/ sol3/papers.cfm?abstract_id = 2035307 Bhatia, G. R. (2006). Combating cartel in markets Issues & challenges. New Delhi: Competition Commission. Bhuhan, S., & Lopez, R. (1997). Oligopoly power in the food and tobacco industries. American Journal of Agricultural Economics, 79, 10351049. Bittlingmayer, G. (1982). Decreasing average cost and competition: A new look at the addyston pipe case. Journal of Law and Economics, 25, 201229. Bittlingmayer, G. (1995). Output and stock prices when antitrust is suspended: The effects of the NIRA. In F. S. McChesney & W. F. Shughart II (Eds.), The causes and consequences of antitrust: The public choice perspective. Chicago, IL: University of Chicago Press. Bizzari, S. N. (2003, August). Methyl methacrylate: CEH marketing research report. Palo Alto, CA: SRI International. BKA. (2008, February 20). Bundeskartellamt verhaengt Millionenbussen gegen Drogereiartikelhersteller. Press release. Retrieved from http://www.bundeskartellamt. de/wDeutsch/aktuelles/2008_02_20.php Blair, J. M. (1976). The control of oil. New York, NY: Pantheon. Blair, R., Mak, J., & Bonham, C. (2007). Collusive duopoly: The effects of the Aloha and Hawaiian airlines’ agreement to reduce capacity. Antitrust Law Journal, 74, 409438. Blair, R. D., & Harrison, J. L. (2010). Monopsony in law and economics. Cambridge: Cambridge University Press.
344
JOHN M. CONNOR
Blanckenburg, K., & Geist, A. (2011). Detecting illegal activities: The case of cartels. European Journal of Law and Economics, 32, 1533. Blanckenburg, K., Geist, A., & Khodlodilin, K. A. (2010, May). The influence of collusion on price changes: New evidence from major cartel cases. DIW Discussion Paper No. 1004. Deutsches Institut fuer Wirtschaftsforschung, Berlin. Blanckenburg, K., Geist, A., & Khodlodilin, K. A. (2012). The influence of collusion on price changes: New evidence from major cartel cases. German Economic Review, 13, 245256. Bloch, K. (1932). On German cartels. Journal of Business, 5, 213222. Block, M. K., & Nold, F. C. (1981). The deterrent effect of antitrust enforcement. Journal of Political Economy, 89, 429445. Blum, U. (2009). The East German cement cartel. Eastern European Economics, 45, 528. Boies, D. (2004). Courting justice. New York, NY: Miramax Books. Bolotova, Y. V. (2008, December). Does the potato supply management program work? A case of the Idaho potato industry. SSRN Working Paper No. 1175202. Retrieved from SSRN: http://ssrn.com/abstract = 1175202 Bolotova, Y. V. (2009). Cartel overcharges: An empirical analysis. Journal of Economic Behavior & Organization, 70, 321341. Bolotova, Y., Connor, J. M., & Miller, D. (2005, April 9). The impact of collusion on price behavior: Empirical results from two recent cases. Paper presented at the 3rd International Industrial Organization Conference, Atlanta, GA. Bolotova, Y., McIntosh, C. S., Muthusamy, K., & Patterson, P. E. (2008). The impact of coordination of production and marketing strategies on price behavior: Evidence from the idaho potato industry. International Food and Agribusiness Review, 11, 128. Bolotova, Y., McIntosh, C. S., Patterson, P. E., & Muthusamy, K. (2010). Is stabilization of potato price and supply effective? Empirical evidence from Idaho. Agribusiness: An International Journal, 26, 177201. Boone, J. (2011, February 8). Private Communication to John M. Connor. Boone, J., & van Leuvensteijn, M. (2010, December). Measuring competition using the profit elasticity: American sugar industry, 18901914. SSRN Working Paper No. 1721520. Retrieved from http://ssrn.com/abstract=1721520 Borgeson, N. S. (1999, May). Amino acids: CEH marketing research report. Palo Alto: SRI International. Bos, P. V. F. (2006). International scrutiny of payment card systems. Antitrust Law Journal, 73, 739777. Bosch, J., & Eckard, E. W. (1991). The profitability of price fixing: Evidence from stock-market reactions to federal indictments. Review of Economics and Statistics, 73, 309317. Boyer, M., & Kotchoni, R. (2012, May 31). How much do cartels typically overcharge? Center for Interuniversity Research and Analysis on Organization (CIRANO) Scientific Publications No. 2012s-15. Retrieved from http://ssrn.com/abstract = 2076055 Brander, J. A., & Ross, T. W. (2006). Estimating damages from price-fixing. Canadian Class Action Review, 3, 335–369. Bra¨nnlund, R. (1989). The social loss from imperfect competition The case of the swedish pulpwood market. Scandinavian Journal of Economics, 91, 689704. Brannman, L. E., & Klein, J. D. (1992). The effectiveness and stability of highway bid-rigging. In Empirical studies in industrial organization: Essays in honor of Leonard W. Weiss: Studies in industrial organization (Vol. 16). Norwell, MA: Kluwer Academic.
Cartel Overcharges
345
Briggs, H. (1992). The effects of concentration and cartels on nineteenth-century railroad freight rates. In Empirical studies in industrial organization: Essays in honor of Leonard W. Weiss: Studies in industrial organization (Vol. 16). Norwell, MA: Kluwer Academic. Bright, A. A. (1949). The electric lamp industry: Technological change and economic development from 1800 to 1947. New York, NY: Macmillan. Brown, J. H. (2004, March). Jeremiah Jenks: A pioneer of industrial organization? Journal of the History of Economic Thought, 26, 6989. Bruland, K., & O’Brien, P. (Eds.). (1998). From family firms to corporate capitalism. Oxford: Clarendon Press. Buccirossi, P., Ciari, L., Duso, T., Spagnolo, G., & Vitale, C. (2013, September). Competition policy and productivity growth: An empirical assessment. Review of Economics and Statistics, 95, 13241336. Bullock, C. J. (1901). Trust literature: A survey and criticism. Quarterly Journal of Economics, 15, 167217. Bullock, C. J. (1905). Pools, trusts and corporations. Boston, MA: Ginn & Co. Burhop, C., & Lu¨bbers, T. (2008, April). Cartels, managerial incentives, and productive efficiency in German coal mining, 18811913. Bonn: Max Planck Institute (later published in the Journal of Economic History, 2009, Vol, 69, pp. 500527). Burnside, A. (2003, April 24). European cartel enforcement and investigations. Speech at the annual spring meeting of the Antitrust Section of the American Bar Association, Washington, DC. Buschena, D. E., & Perloff, J. M. (1991). The creation of dominant firm market power in the coconut oil export market. American Journal of Agricultural Economics, 73, 10001008. Bush, D., Connor, J. M., Flynn, J. J., Ghosh, S., Grimes, W., Harrington, J. E., Jr., Semeraro, S. (2004, March 15). Brief Amici Curiae No. 03-724 in the Supreme Court of the United States, F. Hoffmann-LaRoche et al. v. Empagran et al. Cabolis, C., Axarloglou, K., & Chrissidis, N. (2012, April 13). A cartel that lasts for centuries: The case of the eastern orthodox church indulgences. Yale ICF Working Paper No. 06-13. Retrieved from SSRN: http://ssrn.com/abstract = 2083870 Calvert, A. F. (1913). A history of the salt union. London: Effingham Wilson. Canada. (1945). Canada and international cartels: An inquiry into the nature and effects of international cartels and other trade combinations. Ottawa: Combines Investigation Commission. Carlton, D. W., Bamberger, G. E., & Epstein, R. J. (1995). Antitrust and higher education: Was there a conspiracy to restrict financial aid? Rand Journal of Economics, 26(1), 131147. Carlton, D. W., & Perloff, J. M. (1990). Modern industrial organization. Glenview, IL: Scott Foresman. Carlton, D. W., & Perloff, J. M. (2005). Modern industrial organization (4th ed.). Boston, MA: Pearson. Carmen, H. F., & Pick, D. H. (1990). Orderly marketing for lemons: Who benefits? American Journal of Agricultural Economics, 72, 347357. Caron, F. (1988). Cartels et Fusions en France: 1914–1945. In H. Pohl (Ed.), Wettbewerbsbeschraenkungen auf Internationalen Maerkten [Competition Restrictions in International Markets]. Zeitschrift fuer Unternehmensgeschichte. [Special Edition of the Journal of Corporation History.], 46. Wiesbaden, Germany.
346
JOHN M. CONNOR
Cassady, R. (1967). Auctions and auctioneering. Berkley, CA: University of California Press. Carbo´-Valverde, S., Chakravorti, S., & Rodriguez-Fernandez, F. (2011, March 7). Regulating two-sided markets: An empirical investigation. Retrieved from http://santiagocarbo. com/CCR07Mar2011.pdf# Cawson, A. (1985). Organized interests and the state. London: Sage. CBC. (various dates). Press release of the Competition Bureau of Canada, Ottawa, Canada. CCI. (2012, July 30). In re: On the alleged cartelization of cement manufacturers (p. 82). New Delhi: Competition Commission of India. Retrieved from http://www.cci.gov.in/index. php?option = com_content&task = view&id = 150 CFTC. (1998). Order instituting proceedings pursuant to. Sections 6(c) and 6(d): In the Matter of Sumitomo Corporation. Washington, DC: Commodities Futures Trading Commission. Retrieved from http://www.cftc.gov/ogc/oporders98/ogcfsumitomo.htm Chakravorti, S. (2010). Externalities in payment card networks: Theory and evidence. Review of Network Economics, 9(2), 99134. Christie, W. G., Harris, J., & Schultz, P. (1994). Why did NASDAQ market makers stop avoiding odd-eighth quotes? Journal of Finance, 49, 18411860. Ciaretta, A. (2001). Empirical analysis on the effects of cartels on prices and real sales in Sweden from 1976 to 1990. Unpublished. University of Alicante. Clabault, J. M., & Block, M. (1981). Sherman Act Indictments, 19551980. Unpublished statistical compilation of fines as proportions of volume of commerce (From Cohen and Scheffman, 1989, note 12). Clark, B. C. (1895). The cordage industry, Chapter 74. In C. M. Depew (Ed.), One hundred years of american commerce: 17951895. New York, NY: D. O. Haynes & Co. Clark, R., & Houde, J.-F. (2011, February 19). Collusion with asymmetric retailers: Evidence from a gasoline price-fixing case. Unpublished working paper. Retrieved from http:// www.indiana.edu/∼econdept/workshops/Spring_2011_Papers/ClarkHoudeCollusion.pdf Clarke, J. L., & Evenett, S. J. (2002, December). The deterrent effects of national anti-cartel laws: Evidence from the international vitamins cartel. Working Paper 02-13. AEI-Brookings Joint Center for Regulatory Studies, Washington, DC. Clarke, J. L., & Evenett, S. J. (2003). The deterrent effects of national anticartel laws: Evidence from the international vitamins cartel. Antitrust Bulletin, 48, 289726. Clay, K., & Troesken, W. (2003). Further tests of static oligopolistic models: Whiskey, 18821898. Journal of Industrial Economics, 51, 151166. Cohen, M. A. (1989a). The role of criminal sanctions in antitrust enforcement. Contemporary Policy Issues, 7, 3646. Cohen, M. A. (1989b). Corporate crime and punishment: A study of social harm and sentencing practices in the federal courts, 19841987. American Criminal Law Review, 26, 605660. Cohen, M. A., Ho, C.-C., Jones, E. D., III, & Schleich, L. M. (1988). Organizations as defendants in Federal Court: A preliminary analysis of prosecutions, convictions, and sanctions, 1984–1987. Whittier Law Review, 10, 103–124. Cohen, M. A., & Scheffman, D. T. (1989). The antitrust sentencing guideline: Is the punishment worth the costs? Journal of Criminal Law 27, 331366. Coleman, D. C. (1969). International Consortia. In G. Jones (Ed.), Coalitions and collaboration in international business (pp. 5178). Aldershot: Edward Elgar. Comanor, W. S., & Schankerman, M. A. (1976). Identical bids and cartel behavior. Bell Journal of Economics, 7, 281286.
Cartel Overcharges
347
Comision National de la Competencia (CNC). (2011, October 24). CNC fines 47 construction companies for bid rigging bids in public works tenders. Press release. Madrid, Spain. Committee on the Judiciary. (1921, June 220). Authorizing association producers of agricultural producers: Hearings on H. R 2373. Washington, DC: U.S. Senate. Conference Board. (2010, October 1). Saskatchewan in the spotlight: Acquisition of potash corporation of Saskatchewan Inc. Risks and opportunities. Conference Board of Canada. Connor, J. M. (1998). The global citric acid conspiracy: Legal-economic lessons. Agribusiness: An International Journal, 14, 253259. Connor, J. M. (2001a). Global price fixing: “Our customers are the enemy”. Boston, MA: Kluwer Academic. Connor, J. M. (2001b). “Our customers are our enemies:” The lysine cartel of 19921995. Review of Industrial Organization, 18, 521. Connor, J. M. (2003, November). Private international cartels: Effectiveness, welfare, and anticartel enforcement. Staff Paper No. 03-12. Department of Agricultural Economics, Purdue University, West Lafayette, IN. Connor, J. M. (2004a). Global cartels redux: The amino acid lysine antitrust litigation (1996). In J. E. Kwoka & L. White (Eds.), The antitrust revolution (4th ed.), Oxford: Oxford University Press. Connor, J. M. (2004b, September). Global antitrust prosecutions of modern international cartels. Journal of International Competition and Trade, 4, 239267. Connor, J. M. (2004c, January). A comment on Crandall and Winston, American Antitrust Institute Web site. Retrieved from http://www.antitrustinstitute.org/Archives/291.ashx Connor, J. M. (2004d, May). Collusion and price dispersion. Staff Paper No. 04-14. Department of Agricultural Economics, Purdue University, West Lafayette, IN. Retrieved from http://agecon.lib.umn.edu/cgi-bin/view.pl Connor, J. M. (2005, May). Collusion and price dispersion. Applied Economics Letters, 12, 335338. Connor, J. M. (2005a). Price-fixing overcharges: Legal and economic evidence. SSRN Working Paper No. 787924. Retrieved from http://ssrn.com/abstract=787924 Connor, J. M. (2007a, January). Price-fixing overcharges: Legal and economic evidence, Chapter 4, In J. B. Kirkwood (Ed.), Research in law and economics (Vol. 22, pp. 59153). Oxford: Elsevier. Connor, J. M. (2007b). Global price fixing. Studies in industrial organization (2nd updated and Rev. ed., Vol. 26). Heidelberg: Springer. Connor, J. M. (2007c, October). Forensic economics applied to price-fixing overcharges. Journal of Competition Law & Economics, 4(1), 3159. Connor, J. M. (2008, July). Global cartels redux: The amino acid lysine antitrust litigation (1996). In J. E. Kwoka & L. White (Eds.), The antitrust revolution (5th ed.), Oxford: Oxford University Press. Connor, J. M. (2009a, August 31). Cartels & antitrust portrayed: Private international cartels from 1990 to 2008. AAI Working Paper No. 09-06. Retrieved from http://ssrn.com/ abstract = 1467310 Connor, J. M. (2009b, July). Private international cartels full data. Excel spreadsheet (a proprietary data set with estimates prepared by the author from convicted modern international cartels; data and methods are detailed in Connor (2014: Appendix Tables 1 and 2). Connor, J. M. (2009c, May 21). Governments as cartel victims. AAI Working Paper No. 09-03. Retrieved from http://ssrn.com/abstract = 1412463
348
JOHN M. CONNOR
Connor, J. M. (2010, December). Recidivism revealed: Private international cartels 19902009. Competition Policy International, 6(2), 101128. Retrieved from http:// ssrn.com/abstract = 1688508 Connor, J. M. (2011). Cartels portrayed, a 21-year perspective: U.S. vs. EC: Who’s winning the prosecution race? AAI Working Paper No. 11-03 (posted August 2, 2011). Connor, J. M. (2011a, Summer). Problems with prison in international cartel cases. T. Calvani & T. H. Calvani (Eds.), Antitrust Bulletin: Symposium: Cartel Penalties: Effective Deterrents or License Fees? 56, 311344. Connor, J. M. (2011b, September 28). Cartel detection and duration worldwide. Competition Policy International Antitrust Chronicle, 2, 210. Retrieved from https://www. competitionpolicyinternational.com/cartel-detection-and-duration-worldwide Connor, J. M. (2012, December). Private recoveries in international cartel cases worldwide: What do the data show? Competition Policy International Antitrust Chronicle, 1, 224. Retrieved from http://www.antitrustinstitute.org/∼antitrust/sites/default/files/Working PaperNo12-03.pdf Connor, J. M. (2013). Cartel fine severity and the European commission. European Competition Law Review, 34, 5877. Retrieved from http://ssrn.com/abstract = 2249010 Connor, J. M. (2014). Price-fixing overcharges: Legal and economic evidence (3rd ed.). SSRN Working Paper. Forthcoming. Connor, J. M., & Bolotova, Y. (2006). A meta-analysis of cartel overcharges. International Journal of Industrial Organization, 24, 11091137. Retrieved from http://papers.ssrn. com/sol3/papers.cfm?abstract_id = 788884 Connor, J. M., Bolotova, Y., & Miller, D. J. (2008, November). The impact of collusion on price behavior: Empirical results from two recent cases. International Journal of Industrial Organization, 26, 12901307. Retrieved from http://ssrn.com/abstract = 991908 Connor, J. M., & Helmers, C. G. (2007, January). Statistics on modern private international cartels. Working Paper No. 07-01. American Antitrust Institute, Washington, DC. Retrieved from http://www.antitrustinstitute.org/recent2/567.pdf Connor, J. M., & Lande, R. H. (2005, Winter). How high do cartels raise prices? Implications for optimal cartel fines. Tulane Law Review, 80, 513570. Connor, J. M., & Lande, R. H. (2008, July). Cartel overcharges and optimal cartel fines, Chapter 88. In W. D. Collins (Ed.), Issues in competition law and policy (Vol. 3, pp. 22032218). Chicago, IL: Section of Antitrust Law of the American Bar Association. Retrieved from SSRN: http://ssrn.com/abstract = 1285455 Connor, J. M., & Lande, R. H. (2012, December). Cartels as rational business strategy: New data demonstrates that crime pays. Cardozo Law Review, 34(2), 428491. Retrieved from http://ssrn.com/abstract = 1917657 Conseil Concurrence. (2003, March, 31). Decision No. 03-D-17 of March 31, 2003, relative to anticompetitive practices observed in the motorway fuel distribution market. Paris: Conseil de la concurrence. Conseil Concurrence. (2005, December 15). Decision No. 05-D-69 of December 15, 2005, relative to anticompetitive practices observed in the roadworks sector in the Seine-Maritime de´partement. Paris: Conseil de la concurrence. Conseil Concurrence. (2007, May 6). Decision No. 075-D-15 of 9 May 2007, relative to anticompetitive practices observed in the public tenders for secondary schools in Ile-de-France region. Paris: Conseil de la concurrence.
Cartel Overcharges
349
Constantine, L. (2009). Priceless: The case that brought down the visa/mastercard bank cartel. New York, NY: Kaplan Books. Crampton, P., & Palfry, T. (2000). Collusive bidding: Lessons from the FCC spectrum auctions. Journal of Regulatory Economics, 17, 229252. Crandall, R. W., & Winston, C. (2003, Fall). Does antitrust policy improve consumer welfare? Assessing the evidence. Journal of Economic Perspectives, 17, 126. Crespi, J. M., & Chacon-Cascante., A. (2004). Do U.S. marketing orders have much market power? An examination of the Almond Board of California. Agribusiness: An International Journal, 21, 116. Curtis, R. E. (1931). Trusts and economic control: A book of materials. New York, NY: McGraw-Hill. Daggett, R. S., & Freedman, D. A. (1985). Econometrics and the law: A case study in the proof of antitrust damages. In L. M. Le Cam & R. A. Olshen (Eds.), Proceedings of the Berkeley conference in honor of Jerzy Neyman and Jack Kiefer (Vol. 1). New York, NY: Wadsworth. Daily News Egypt. (2008). Scandals of 2008. Daily News Egypt, December 23. Retrieved from http://www.thefreelibrary.com/Scandals+of+2008%3A+Monopoly,+murder+and+ mayhem.-a0191062270 Dalkir, S. (2006). Near discoveries and half punishments against cartels can be self-defeating. Iktisat, Is¸letme ve Finans (Economics, Money and Finance), 21(242), 522. Dalkir, S. (2007, January 12). Personal communication on the Turkish cement cartel. Dana, D. E. (1922, November). Jones’ the trust problem in the United States. Quarterly Journal of Economics, 28, 175182. Davidow, J. (1972). Proof of purchases and damages by public buyers of price-fixed goods. The Antitrust Bulletin, 17, 363378. Davies, S., Coles, H., Olczak, M., Pike, C., & Wilson, C. (2004, July). Benefits from competition: Some illustrative UK cases. DTI Economics Paper No. 9. Department of Trade and Industry, UK Government, London. Retrieved from http://www.dti.gov.uk/files/ file13299.pdf. Davies, S., & Majumdar, A. (2002, June). The development of targets for consumer savings arising from competition policy. Economic Discussion Paper No. 4/OFT 386. Report prepared for the Office of Fair Trading, London. Davis, J. P., & Lande, R. H. (2012). Towards an empirical and theoretical assessment of private antitrust enforcement. University of San Francisco Law Research Paper No. 2012-17. Retrieved from http://ssrn.com/abstract = 2132981 Davis, P., & Garce´s, E. (2009). Quantitative techniques for competition and antitrust analysis. Princeton, NJ: Princeton University Press. de Jong, H. W. (2009a). Economists from the German language area (Nineteenth and Twentieth Centuries), Chapter 3. In H. W. de Jong & W. G. Shepherd (Eds.), Pioneers of industrial organization. Cheltenham: Edward Elgar. de Jong, H. W. (2009b). Market theory in the low countries, Chapter 4. In H. W. de Jong & W. G. Shepherd (Eds.), Pioneers of industrial organization. Cheltenham: Edward Elgar. de Jong, H. W., & Shepherd, W. G. (Eds.). (2007). Pioneers of industrial organization: How the economics of competition and monopoly took shape. Cheltenham: Edward Elgar. de Roos, N. A. (2001, October). Collusion with a competitive fringe: An application to vitamin C. Unpublished manuscript, Department of Economics, Yale University, 30 pp. de Roos, N. A. (2004a). Model of collusion timing. International Journal of Industrial Organization, 22, 351387.
350
JOHN M. CONNOR
de Roos, N. A. (2004b, December). Examining models of collusion: The market for lysine. Unpublished manuscript, 49 pp. de Roos, N. A. (2006, November). Examining models of collusion: The market for lysine. International Journal of Industrial Organization, 24, 10831107. De Rousiers, P. (1927). Cartels trusts and their development. Geneva: League of Nations. Demaree, A. T. (1969). How judgment came for the plumbing conspirators. Fortune, December, pp. 9699 and 175182. Denger, M. L. (2003, April). Too much or too little, a summary of discussion. American Bar Association’s Antitrust Remedies Forum, Washington, DC. Depew, C. M. (Ed.). (1895). One hundred years of American commerce (Vol. 2). New York, NY: D. O. Haynes. Devos, G. (1994). International cartel agreements in the zinc industry. In D. Bardot (Ed.), International cartels revisited, 18801980 (pp. 143151). Caen, France: Editionsdiffusion du Lys. Dewing, A. S. (1913). A history of the national cordage company. Cambridge, MA: Harvard University Press. Dick, A. R. (1992a). Are export cartels efficiency-enhancing or monopoly-promoting? Research Law and Economics, 15, 89127. Dick, A. R. (1992b). The competitive consequences of Japan’s export cartel associations. Journal of Japanese and International Economics, 6, 275298. Dick, A. R. (1996). When are stable cartels stable? Journal of Law and Economics, 39, 241283. DOJ. (1990, October 24). Government’s memorandum in aid of sentencing: United States v. Baer, Crim. No. 90-0314 (HHG). Washington, DC: District Court of the District of Columbia. DOJ. (various dates, 19912008). Press releases, plea agreements, etc. by the Antitrust Division of the U.S. Department of Justice. Retrieved from http://www.usdoj.gov/atr/ index.html Domeratzky, L. (1928). The international cartel movement: Trade information Bulletin No. 556. Washington, DC: U.S. Department of Commerce. Donovan, K. (2005). Goliath: The trials of David Boies. New York, NY: Pantheon. Drewery. (1996). Global container market: Prospects and profitability in a high growth era. London: Drewery Shipping Consultants. Duffy, S. P. (2009, April 13). Instant appeal certified in chocolate price-fixing case. Legal Intelligencer, 239, 1. Duggan, J. E., & Narasimham, G. V. L. (1981). Price fixing in the titanium industry: A time series analysis. In Proceedings of the business and economic statistics section. American Statistical Association, Washington, DC. Eatwell, J., Milgate M., & Newman P. (Eds.). (1987). The new Palgrave: A dictionary of economics (Vols. 14), London: Macmillan. EC. (various dates of decisions). Decisions of the European Commission. RAPID. Eckbo, P. L. (1976). The future of world oil. Cambridge, MA: Ballinger. Economic and Financial Section. (1927). Electrical industry. Geneva: League of Nations. Edgerton, C. E. (1897). The wire-nail association of 18951896. Political Science Quarterly, 12, 246272. Edwards, C. D. (1944). Economic and political aspects of international cartels. Washington, DC: U.S. Government Printing Office.
Cartel Overcharges
351
Edwards, C. D. (1946). Cartelization in Western Europe. Washington, DC: Bureau of Intelligence and Research, U.S. Department of State. Edwards, C. D. (1966). Trade regulations overseas: The national laws. New York, NY: Oceana Publications. Edwards, C. D. (1967). Control of cartels and monopolies: An international comparison. Oceana Publications. Ehmer, C., & Rosati, F. (2009, November). Science, myth, and fines: Do cartels typically raise prices by 25%? Concurrences No 4-2009, 15. Eichenwald, K. (2000). The informant: A true story. New York, NY: Broadway Books. Elliott, W. Y., May, E. S., Rowe, J. W. F., & Wallace, D. H. (1937). International control in the non-ferrous metals. New York, NY: Macmillan. Ellison, G. (1994). Theories of cartel stability and the joint executive committee. RAND Journal of Economics, 25, 3757. Ellison, G. (2003, November 12). Personal communication to J. M. Connor about Ellison’s 1994 JEC study. Elzinga, K. G. (1970). Predatory pricing: The case of the gunpowder trust. The Journal of Law and Economics, 13, 223240. Elzinga, K. G. (1984, Spring). New developments on the cartel front. Antitrust Bulletin, 29, 326. Elzinga, K. G., & Wood, W. C. (1988). The costs of the legal system in private antitrust enforcement. In L. J. White (Ed.), Private antitrust litigation: New evidence, new learning. Cambridge, MA: MIT Press. Engel, C. (2007). How much collusion? A meta-analysis of oligopoly experiments. Journal of Competition Law and Economics, 3, 491549. Epstein, B. (1971). Politics of trade in power plants. London: Trade Policy Research Center. Epstein, B., & Newfarmer, R. S. (1980). International Electric Association: A continuing cartel. Report to the subcommittee on oversight and investigations, House Committee on Interstate and Foreign Commerce, U.S. Congress, USGPO, Washington, DC. Erickson, W. B. (1976). Price-fixing conspiracies: Their long-term impact. Journal of Industrial Economics, 24, 189202. Erutku, C., & Hildebrand, V. A. (2010). Conspiracy at the pump. Journal of Law and Economics, 53, 223237. Eswaran, M. (1997). Cartel unity over the business cycle. Canadian Journal of Economics, 30, 644672. Euromonitor. (2013). Euromonitor International: Passport search (online library research service, frequent source of sales and prices, mostly consumer goods and services). European Commission (EC). (date of decision). Commission decision of … [date given]. Brussels. Retrieved from www.europa.eu.int Evans, D. S., & Mateus, A. M. (2011, June 27). How changes in payment card interchange fees affect consumers fees and merchant prices: An economic analysis with applications to the European Union. Retrieved from SSRN: http://ssrn.com/abstract = 1878735 or http:// dx.doi.org/10.2139/ssrn.1878735] Evenett, S. J., Levenstein, M. C., & Suslow, Y. V. (2001). International cartel enforcement: Lessons from the 1990s. World Economy, 24, 12211246. Fallenius, C. A. (2001). A Swedish example The petrol case. In Fighting cartels Why and how? Proceedings of a conference sponsored by the Konkurrensverket, Elanders Graphic Systems, Stockholm.
352
JOHN M. CONNOR
Feinstein, J. S., Block, M., & Nold, F. (1985). Asymmetric information and collusive behavior in auction markets. American Economic Review, 75, 441460. Fejø, J. (2001). Deterrent effect of competition law on hard core cartels. In Fighting cartels Why and how? Stockholm: Konkurrensverket. Fetter, F. A. (1931). The masquerade of monopoly. New York, NY: Harcourt Brace. Filson, D., Keen, E., Fruits, E., & Borcherding, T. (2001). Market power and cartel formation: Theory and empirical test. Journal of Law and Economics, 44, 465–480. Fine, B. (1990). Economies of scale and a featherbedding cartel? A reconsideration of the interwar British coal industry. Economic History Review (second series), 43, 438449. Fink, C., Mattoo, A., & Neagu, I. C. (2002). Trade in international maritime services: How much does policy matter? World Bank Economic Review, 16, 8188. Finkelstein, M., & Levanbach, H. (1983). Regression estimates of damages in price-fixing cases. Law and Contemporary Problems, 46, 145169. Fisher, F. (1980). Multiple regression in legal proceedings. Columbia Law Review, 80, 702. Fisher, F. M., & Cootner, P. H. (1972). An econometric model of the world copper industry. Bell Journal of Economics and Management Science, 3, 570591. Fisher, L. A., & Owen, A. D. (1981). An econometric model of the U.S. aluminum market. Resources Policy, 150160. Fleming, G. (2000). Collusion and price wars in the Australian coal industry during the late nineteenth century. Business History, 42, 4770. Fog, B. (1956). How are cartel prices determined? Journal of Industrial Economics, 5, 1623. Fo¨lster, S., & Peltzman, S. (1997). The social cost of regulation and the lack of competition in Sweden. In R. B. Freeman, R. H. Topel, & B. Swedenborg (Eds.), The welfare state in Sweden: Reforming the Swedish model. Chicago, IL: University of Chicago Press. Fonseca, M. A., & Normann, H.-T. (2011, May). Explicit vs. tacit collusion: The impact of communication in oligopoly experiments. Discussion Paper 95. Duesseldorf Institute for Competition Economics (DICE), Duesseldorf. Fonseca, M. A., & Normann, H.-T. (2012). Explicit vs. tacit collusion The impact of communication in oligopoly experiments. European Economic Review, 56, 17591772. Fourie, F. C., & Smith, A. (1994). The South African cement cartel: An economic evaluation. South African Journal of Economics, 62, 123143. Fourie, F. C., & Smith, A. (1995). The South African cement cartel: Reply. South African Journal of Economics, 63, 91102. Fraas, A. G., & Greer, D. F. (1977). Market structure and price collusion: An empirical analysis. Journal of Industrial Economics, 26, 2144. Frank, N., & Lademann, R. P. (2010). Economic evidence in private damage claims: What lessons can be learned from the German cement cartel case? Journal of European Competition Law & Practice, 1, 360366. French, B. C., & Nuckton, C. F. (1982). An empirical analysis of economic performance under the marketing order for raisins. American Journal of Agricultural Economics, 64, 581593. Friedman, M. (1974). FEO and the gas line. Newsweek, March 4, p. 71. Frode, S. (2006, December 4). Cement and Lerner Index: Personal email communication. Froeb, L. M., Koyak, R. A., & Werden, G. J. (1993). What is the effect of bid-rigging on prices? Economics Letters, 42, 419423.
Cartel Overcharges
353
Froeb, L. M., Koyak, R. A., & Werden, G. J. (1994). Bid rigging against the government. Owen Manager, 15, 111. Froeb, L. M., Koyak, R. A., & Werden, G. J. (1995a). Effects of price-fixing and bid-rigging conspiracies. Retrieved from www.antitrust.org/economics/conspiracyeffects.html, downloaded February 24, 1999. Froeb, L. M., Koyak, R. A., & Werden, G. J. (1995b). Japan’s price-fixing and bid-rigging conspiracies. Retrieved from www.antitrust.org/economics/conspiracyeffects.html, downloaded February 24, 1999. FTC. (1948). Report of the federal trade commission on international electrical equipment cartels. Washington, DC: U.S. Federal Trade Commission. FTC. (1998, December 21). Illegal conduct resulted in price increases of up to 3000 percent for two anti-anxiety drugs, FTC alleges; agency to seek $120 million in consumer refunds. Press release. Retrieved from http://www.ftc.gov/opa/1998/12/mylanpv. shtm Fugate, W. L. (1973). Foreign commerce and the antitrust laws (2nd ed.). Boston, MA: Little Brown. Funderburk, D. R. (1974). Price fixing in the liquid asphalt industry: Economic analysis versus the “hot document”. Antitrust Law & Economic Review, 7, 6174. Gaban, E. M., & Domingues, J. O. (2012). Antitrust law in Brazil: Fighting cartels. Alphen aan den Rijn: Wolters Kluwer. Gallet, C. A. (1997). Cyclical fluctuations and coordination in the US steel industry. Applied Economics, 29, 279285. Gallo, J. C., Dau-Schmidt, K. G., Craycraft, J. L., & Parker, C. J. (1994). Criminal penalties under the Sherman Act: A study in law and economics. Research in Law and Economics, 16, 2571. Gately, D. (1984). A ten-year retrospective: OPEC and the world oil market. Journal of Economic Literature, 22, 11001114. Gaver, K. M., & Zimmerman, J. L. (1977). An analysis of competitive bidding on BART contracts. Journal of Business, 50, 279295. Gavin, A. (2004). Email to John M. Connor dated October 7. Genesove, D., & Mullin, W. P. (1998). Testing static oligopoly models: Conduct and cost in the sugar industry, 18901914. Rand Journal of Economics, 29, 355377. Genesove, D., & Mullin, W. P. (2001). Rules, communication, and collusion: Narrative evidence from the sugar institute case. American Economic Review, 91, 379398. Gerber, D. (1998). Law and competition in twentieth century Europe: Protecting prometheus. Oxford: Oxford University Press. Gillingham, J. (1994). An ill-fitting jacket: The United States and the international steel cartel. In D. Bardot (Ed.), International cartels revisited, 18801980 (pp. 129142). Caen, France: Editions-diffusion du Lys. Gilo, D., & Spiegel, Y. (2005). The credit card industry in Israel. Review of Network Economics, 4, 266289. Ginsberg, D. H. (1986, July 15). Statement of the Assistant Attorney General for Antitrust, U.S. Department of Justice. Hearings Before the United States Sentencing Commission Concerning Alternatives to Incarceration. Ginsberg, D. H., & Wright, J. D. (2010). Antitrust sanctions. Competition Policy International, 6, 339.
354
JOHN M. CONNOR
Goldberg, V. P. (1991). The international salt puzzle. Research in Law and Economics, 14, 3941. Goldscmidt, B. (1989, June 5). How it all began in Canada – The role of French scientists. Proceedings of the 29th annual conference of the Canadian nuclear society: Special symposium: 50 years of nuclear fission in review 1. Retrieved from http://www.cns-snc.ca/ history/fifty_years/goldschmidt.html. Downloaded on 9/9/06. Gommesen, H. (2003, December 3). Cartels in Denmark, personal communication. Goyder, D. G. (1998). EC competition law (3rd ed.). Oxford: Clarendon. Graebner, W. (1973). Great expectations: The search for order in bituminous coal, 18901917. Business History Review, 48, 4972. Granitz, E., & Benjamin, K. (1996). Monopolization by “raising rivals’ costs”: The standard oil case. Journal of Law and Economics, 39, 147. Grant, H., & Thille, H. (2001). Tariffs, strategy and structure: Competition and collusion in the Ontario petroleum industry, 18701880. Journal of Economic History, 61, 390413. Grat van den, H. (2005). The parliamentary enquiry on fraud in the Dutch construction industry. Crime, Law, and Social Change, 44, 133151. Graubert, J. D. (2003, April). Too much or too little, a summary of discussion. American Bar Association’s Antitrust Remedies Forum, Washington, DC. Gray, E. (1982). The great uranium cartel. Toronto: McClelland & Stewart. Great Britain Board of Trade. (1944, 1946). Survey of international cartels and internal cartels. London: Board of Trade (two volumes). Greer, D. F. (1984). Industrial organization and public policy (2nd ed.), New York, NY: Macmillan. Griffin, J. M. (1989). Previous cartel experience: Any lessons for OPEC? In R. K. Lawrence & J. Marquez (Eds.), Economics in theory and practice: An eclectic approach. Boston, MA: Kluwer Academic. Griffin, J. M., & Neilson, W. S. (1994). The oil price collapse and afterwards: What does game theory add? Economic Inquiry, 32, 543561. Groenewegen, P. (1987). “Political economy” and “economics”. In J. Eatwell, et al., The new palgrave (Vol. 3). London: Macmillan. Groner, J. (2000, October 2). Chalk up a few wins for microsoft. The National Law Journal, A4. Grossman, P. Z. (1996). The dynamics of a stable cartel: The railroad express, 18511913. Economic Inquiry, 34, 220236. Grossman, P. Z. (Ed.). (2004). How cartels endure and how they fail: Studies of industrial collusion. Cheltenham: Edward Elgar. Grout, P. A., & Sonderegger, S. (2005, March). Predicting cartels, office of fair trading. Discussion Paper (OFT 773). Office of Fair Trading, London. Guersent, O. (Ed.). (2004). The fight against international cartels of the European Commission: 19 Decisions in 19 Months between July 2001 and December 2003. EC DirectorateGeneral for Competition (2001), Brussels. Retrieved from http://europa.eu.int/comm/ competition/antitrust, downloaded March. Guilfoil, J. D. (1967, June). Damage determination in private antitrust suits. Notre Dame Lawyer, 42, 647679. Gu¨nster, A., Carree, M., & van Dijk, M. A. (2011, December). Do cartels undermine economic efficiency? Retrieved from http://scholar.google.com/scholar?sciupd = 1&hl = en&as_ sdt = 0,22
Cartel Overcharges
355
Gupta, B. (2001a). The international tea cartel during the great depression, 19291933. Journal of Economic History, 61, 144159. Gupta, B. (2005). Why did collusion fail? The indian jute industry in the inter-war years. Business History, 47, 532552. Gupta, S. (2001b). The effect of bid-rigging on prices: A study of the highway construction industry. Review of Industrial Organization, 19, 453467. Haan, M. A., Schoonbeek, L., & Winkel, B. M. (2009). Experimental results on collusion: The role of information and communication. In J. Hinloopen & H.-T. Normann (Eds.), Experiments and competition policy (pp. 933). Cambridge: Cambridge University Press. Haber, L. F. (1971). The chemical industry 19901930. Oxford: Clarendon Press. Hahn, V., Normann, H.-T., & Der Fall des, H.-T. (2001). Kabelherstellerkartells Zur Bedeutung von Wettbewerbsrecht und Marktkra¨ften. Schmollers Jahrbuch, 121, 407426. Hajivassilou, V. A. (1989). Testing game-theoretic models of price-fixing behavior. Discussion Paper No. 935. Cowles Foundation for Research in Economics, New Haven. Haney, L. H. (1919a, March). Price fixing in the United States during the war I. Political Science Quarterly, 34, 104126. Haney, L. H. (1919b, June). Price fixing in the United States during the war II. Political Science Quarterly, 34, 262289. Haney, L. H. (1919c, September). Price fixing in the United States during the war III. Political Science Quarterly, 34, 434453. Hansen, R. G. (1985). Empirical testing of auction theory. American Economic Review, 75, 156159. Harding, C., & Joshua, J. (2003, 2010). Regulating cartels in Europe: A study of legal control of corporate delinquency. Oxford: Oxford University Press. Harrington, J. E. (2004). Post-cartel pricing during litigation. Journal of Industrial Economics, 52, 517–533. Harrington, J. E. (2007). How do cartels operate? Boston, MA: now. Harrington, J. E. (2012). Comment on antitrust sanctions. Competition Policy International Journal, 6, 4153. Harris, R. G., & Sullivan, L. A. (1979). Passing on the monopoly overcharge: A comprehensive policy analysis. University of Pennsylvania Law Review 128, 269. Hasagawa, S. (1994). International cartels and the Japanese electrical machinery industry. In D. Bardot (Ed.), International cartels revisited, 18801980 (pp. 243252). Caen, France: Editions-diffusion du Lys. Haucap, J., Heimeshoff, U., & Schultz, L. M. (2010). Legal and illegal cartels in Germany between 1958 and 2004. DICE Discussion Paper No. 08. Du¨sseldorf Institute for Competition Economics. Hausfeld, M. D. (2003, October 31). Class plaintiffs’ motion for preliminary approval of settlement between class plaintiffs and defendants Mitsui & Co., Mitsui & Co. (U.S.A.), Inc., and Bioproducts, Inc. Washington, DC: U.S. District Court for the District of Columbia. Hausman, W. J. (1980). A model of the london coal trade in the eighteenth century. Quarterly Journal of Economics, 94, 114. Hausman, W. J. (1984). Cheap coals or limitation of the vend? The london coal trade, 17701845. Journal of Economic History, 44, 321328. Hay, G. A., & Kelley, D. (1974). An empirical survey of price-fixing conspiracies. Journal of Law and Economics, 17, 1338.
356
JOHN M. CONNOR
Hayashi, F. (2010, April). Public authority involvement in credit and debit card markets: Various countries. Kansas City: Federal Reserve Bank of Kansas City. Retrieved from www.justice.gov/atr/cases/f269100/269181.pdf Heath, J. (1963). Still not enough competition? Restrictive business practices re-examined. London: Institute of Economic Affairs Ltd. Heaulme, M. (1948). L’Evolution commerciale des houille`res du Nord et du Pas de Calais, 19191939. Ph.D. Dissertation, University of Paris. Hendricks, K., & Porter, R. H. (1988). An empirical study of an auction with asymmetric information. American Economic Review, 98, 867883. Hendricks, K., & Porter, R. H. (1989). Collusion in auctions. Annales d’Economie et de Statistique, 15/16, 217230. Hendricks, K., & Robert, H. (1988). An empirical study of an auction with asymmetric information. American Economic Review, 78, 865883. Henley, A. (1988). Price formation and market structure: The case of the interwar coal industry. Oxford Bulletin of Economics and Statistics, 50, 263278. Herfindahl, O. C. (1959). Copper costs and prices: 18701957. Baltimore: Johns Hopkins University Press. Herling, J. (1962). The great price conspiracy: The story of the antitrust violations of the electrical industry. Washington, DC: Robert D. Luce. Hexner, E. (1943). The international steel cartel. Chapel Hill, NC: University of North Carolina Press. Hexner, E. (1946). International cartels. Chapel Hill, NC: University of North Carolina Press (with the assistance of Adelaide Walters). Hirst, F. W. (1905). Monopolies, trusts, and kartells. London: Methuen. Holloway, S. K. (1988). The aluminum multinationals and the bauxite cartel. New York, NY: St. Martin’s. Holmes, W. F. (1994). The southern farmers’ alliance and the jute cartel. Journal of Southern History, 60, 5980. Holt, B. W. (1907). The relationship of the protective tariff to the trusts. Publications of the American Economic Association (Vol. 8, 3rd Series), 212228. Hong Kong Monetary Authority. (2007, February). Testing for collusion in the Honk Kong banking sector: Research memorandum 03-2007. Retrieved from http://www.info.gov. hk/hkma/eng/research/RM03-2007.pdf Hovenkamp, H. (1999). Federal antitrust policy (2nd ed.). St. Paul, MN: West Group. Howard, J. H., & Kaserman, D. (1989). Proof of damages in construction industry bid-rigging cases. Antitrust Bulletin, 34, 359393. Hoxby, C. M. (2000). Benevolent colluders? Working Paper No. 7754. NBER, Cambridge, MA. Hudson, H. (1890). The southern railway and steamship association. Quarterly Journal of Economics, 5, 7094. Hu¨eschelrath, K., Mu¨ller, K., & Veith, T. (2012a, May). Concrete shoes for competition: The effect of the German cement cartel on market price. Discussion Paper No. 12-035. ZEW Centre for European Economic Research, Mannheim, Germany. Retrieved from http://ftp.zew.de/pub/zew-docs/dp/dp12035.pdf Hu¨eschelrath, K., Mu¨ller, K., & Veith, T. (2012b, May). Estimating damages from price-fixing: The value of transaction data. Discussion Paper No. 12-036. ZEW Centre for European Economic Research, Mannheim, Germany. Retrieved from http://ftp.zew.de/ pub/zew-docs/dp/dp12036.pdf
Cartel Overcharges
357
Hu¨eschelrath, K., & Veith, T. (2011, December). Cartel detection in procurement markets. Discussion Paper No. 11-066. ZEW Centre for European Economic Research, Mannheim, Germany. Retrieved from http://papers.ssrn.com/sol3/papers.cfm?abstract_ id = 1983280 ICN. (2005, June). Building blocks for effective anti-cartel regimes (Vol. 1). ICN Working Group on Cartels, Bonn, Germany. Retrieved from http://www.internationalcompetitionnetwork.org/uploads/library/doc346.pdf ICN. (2010, March). Anti-cartel enforcement manual: Chapter 4. ICN Working Group on Cartels. Retrieved from http://www.internationalcompetitionnetwork.org/workinggroups/current/cartel/manual.aspx Indian Express. (2008). MRTP pulls up ACC, Lafarge, Birla for cement price rigging. Indian Express, March 3. Ippolito, R. A., & Masson., R. T. (1978). The social cost of government regulation of milk. Journal of Law and Economics, 21, 3365. Isaacson, W. (2008). Investigation of Chinese cartels affecting U.S. Commerce. Unpublished presentation. Jacquemin, A., Nambu, T., & Dewez, I. (1981). A dynamic analysis of export cartels: The Japanese case. The Economic Journal, 91, 685696. Jacquemin, A., & Slade, M. E. (1989). Cartels, collusion, and horizontal merger. In R. Schmalensee & R. D. Willig (Eds.), Handbook of industrial organization (Vol. 1), Amsterdam: Elsevier. Jarkarta Post. (2007). KPPU rules against temasek. Jarkarta Post, November 20. Jenks, J. W. (1888). The Michigan salt association. Political Science Quarterly, 3, 7898. Jenks, J. W. (1900, 1901, 1903). The trust problem. Garden City, NY: Doubleday, Page (1st, 2nd, and 3rd editions). Jenks, J. W., & Clark., W. E. (1917, 1929). The trust problem. Garden City, NY: Doubleday, Doran (4th and 5th editions). Jenny, F. (2003, February 7). A tale of three cartels. Presentation at the 2003 forum on International Competition Law, Section of Antitrust Law, American Bar Association, New York, NY. Jenny, F. (2005, November 1418). Cartels and collusion in developing countries. Paper at the 5th annual UNCTAD conference [on] restrictive business practices, Antalya, Turkey. Jenny, F. (2010). Global potash trade and competition. The Financial Times (India), November 25. Jeon, S., & Shin, K. (2005, April). Economic analysis of competition effects of international electrodes cartel on Korean import market. Paper at the 3rd International Industrial Organization Conference, Atlanta, GA. JFTC. (2004, May 19). Basic idea behind the planned major revision of the antimonopoly law. Tokyo: Fair Trade Commission. Retrieved from www2.jftc.go.jp/kaisei.htm, downloaded August 2004. [Kishi, Nobuko. Translation of JFTC (2004)] Johnsen, D. B. (1991). Property rights to cartel rents: The Socony-Vacuum story. Journal of Law and Economics, 34, 177203. Jones, E. (1914). The anthracite coal combination in the United States. Cambridge, MA: Harvard. Jones, E. (1921). The trust problem in the United States. New York, NY: Macmillan. Jones, G. (Ed.). (1993). Coalitions and collaboration in international business. Aldershot: Edward Elgar. Josephson, M. (1962). The robber barons. San Diego, CA: Harcourt Brace.
358
JOHN M. CONNOR
Joshua, J. (2009, August). Single continuous infringement of Article 81 EC: Has the commission stretched the concept beyond the limit of its logic? European Competition Journal, 5, 451477. Joshua, J., & Rodriguez, E. A. (2010). E pluribus unum: Rolling up cartel infringements in EC competition law. The European Antitrust Review. London: Global Competition Review. Retrieved from http://www.globalcompetitionreview.com/reviews/19/the-europeanantitrust-review-2010. Joshua, J. M., & Jordan, S. (2003). Combinations, concerted practices and cartels: Adopting the concept of conspiracy in European community competition law. Northwestern Journal of International Law & Business, 24, 647. Joskow, P. (1976). Commercial impossibility, the uranium market and the Westinghouse case. Journal of Legal Studies, 6, 119176. Just, R. E., & Chern, W. S. (1980). Tomatoes, technology, and oligopsony. Bell Journal of Economics, 11, 584602. Kamerschen, D. R., & Morgan, J. E. (2004). Collusion analysis of the Alabama asphalt market. Applied Economics, 36, 690. Kamita, R. Y. (2010, May). Analyzing the effects of temporary antitrust immunity: The Aloha-Hawaiin immunity agreement. Journal of Law and Economics, 53, 238261. Kemp, T. (1972). The French economy, 19131939. London: Longman. KFTC. (2001, June 7). Harms of cartels and sanctions against them: Report to the OECD. Seoul: Republic of Korea. KFTC. (2003, April 25). The KFTC imposes surcharges on the international cartel of vitamin companies. Press release. Korea Fair Trade Commission, Seoul. KFTC. (2008). Korea fair trade commission. Annual report. Retrieved from www.ftc.go. kr/eng Kikkawa, T. (1994). International cartel and domestic cartel in Japan: The case of the 1932 gasoline agreement. In D. Bardot (Ed.), International cartels revisited, 18801980 (pp. 233241). Caen, France: Editions-diffusion du Lys. Kinghorn, J. R. (1996). Kartells and cartel theory: Evidence from early twentieth century German coal, iron and steel industries. Essays in Economic and Business History, 14, 339363. Kinghorn, J. R., & Nielsen, R. (2004). A practice without defenders: Price effects of cartelization. In P. Z. Grossman (Ed.), How cartels endure and how they fail: Studies of industrial collusion. Cheltenham: Edward Elgar. Kirby, M. W. (1973). The control of competition in the British coal-mining industry in the thirties. Economic History Review (2nd series), 26, 273284. Kleinwa¨chter, F. (1883). Die Kartelle: Ein Beitrag zur Frage der Organisation der Volkswirtschaft. Innsbruck, Austria. Kobayashi, B. H. (2001). Antitrust, agency, and amnesty: An economic analysis of criminal enforcement of the antitrust laws against corporations. George Washington Law Review, 69, 715. Komninos, A. & a multi-jurisdictional team of lawyers, with economic assistance of Beckert, W., van Damme, E., Dewatripont, M., Franks, J., ten Kate, A., Legros, P., & with legal assistance from Borum, G. C., Czapracka, K., Dawes, A., Garcia-Nieto, J., InnesStubb, S., Janda, I., Killick, J., … Toth, A. (2009, December). Quantifying antitrust damages: Study prepared for the European Commission, Oxera. Retrieved from http:// www.unipr.it/arpa/defi/econlaw/econlaw%20non%20collegati/Oxera(2009).pdf
Cartel Overcharges
359
Kovacic, W. E., & Shapiro, C. (2000). Antitrust policy: A century of economic and legal thinking. Journal of Economic Perspectives, 14, 4360. KPMG. (2003). Aalborg, A˚rhus, Odense, og Københavns Kommuner: Rapport om erstatningssopgørelser I forbindelser med kommunernes handel med Alstom Power Flowsystems A/S I perioden 19911998, Copenhagen. Krepps, M. 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, 151154. Kronstein, H. (1973). The law of international cartels. Ithaca: Cornell University Press. Kudo, A. (1994). The Japanese strategy of the international dyestuffs cartel. In D. Bardot (Ed.), International cartels revisited, 18801980 (pp. 215222). Caen, France: Editionsdiffusion du Lys. Kuhlman, J. M. (1972). Theoretical issues in the estimation of damages in private antitrust actions. Antitrust Bulletin, 17, 548558. Kwoka, J. E. (1977). Pricing under federal milk market regulation. Economic Inquiry, 5, 367384. Kwoka, J. E. (1997). The price effect of bidding conspiracies: Evidence from real estate “knockouts”. Antitrust Bulletin, 42, 503516. Kwoka, J. E. (2003, March). The attack on antitrust policy and consumer welfare: A response to Crandall and Winston. Working Paper No. 03-008. Department of Economics, Northeastern University, Boston, MA. Labys, W. C., & Hewheler, J. F. (1974). Survey of commodity demand and supply elasticities. Memorandum No. 48. Research Division, Geneva, UNCTAD. Lamer, M. (1957). The world fertilizer economy. Stanford, CA: The Food Research Institute, Stanford University Press. Lande, R. H. (1993). Are antitrust “treble” damages really single damages? Ohio State Law Journal, 54, 117174. Lande, R. H. (2004). Why antitrust damage levels should be raised. Loyola Consumer Law Review, 16, 326. Lande, R. H. (2010, March 5). Comparative deterrence from private enforcement and criminal enforcement of the U.S. Antitrust Laws. SSRN Working Paper 1565693. Retrieved from http://ssrn.com/abstract=1565693. (later published in Brigham Young University Law Review (2011): 101–175). Lande, R. H., & Davis, J. P. (2006, November 8). An evaluation of private antitrust enforcement: 29 case studies. Unpublished manuscript. Lande, R. H., & Davis, J. P. (2007, December 10). Benefits from private antitrust enforcement: An analysis of forty cases. Paper presented at the AAI symposium on future of private antitrust enforcement, Washington, DC (published in law journal 2008). Retrieved from http://www.antitrustinstitute.org/archives/files/AAI%20Benefits%20from%20 Private%20AT%20Enforcement%20-%20Lande%20Davis%20Report%20-%2012. 10.07_121220070944.pdf Lande, R. H., & Davis, J. P. (2008). Benefits from private antitrust enforcement: An analysis of forty cases. University of San Francisco Law Review, 42, 879. Landes, W. H. (1983). Optimal sanctions for antitrust violations. University of Chicago Law Review, 50, 652678. Lanzillotti, R. F. (1996). The great school milk conspiracies of the 1980s. Review of Industrial Organization, 11, 413458. Lanzillotti, R. F. (2000). Great school milk conspiracies revisited: Rejoinder. Review of Industrial Organization, 17, 343353.
360
JOHN M. CONNOR
Leach, D. F. (1994). The South African cement cartel: A critique of fourie and smith. South African Journal of Economics, 62, 254279. Lean, D. F., Ogur, J. D., & Rodgers, R. P. (1985). Does collusion pay … does antitrust work? Southern Economic Journal, 51, 828841. LeCraw, D. J. (1977). The uranium cartel. Business Quarterly, 7783. Lee, I. K. (1999). Non-cooperative tactic collusion, complementary bidding and incumbency premium. Review Industrial Organization, 15, 115134. Lee, I. K., & Hahn, K. (2002). Bid-rigging in auctions for Korean public-works contracts and potential damage. Review of Industrial Organization, 21, 7388. Leffler, K. (2012, February 9). United States of America v. AU Optronics Corp. et al. (U.S. District Court, Northern District of California): Expert testimony of Keith Leffler. Trial transcript, pp. 32743506. Leiden University. (2002). Interwar cartel database. Historical Institute, Leiden, Netherlands. Retrieved from http://www.let.leidenuniv.nl/history/rtg/cartels/bal2.html, downloaded September 26. Leitzinger, J. (1983). Regression analysis in antitrust cases: Opening the black box. The Philadelphia Lawyer, 20, 1. Levenstein, M., Sivadasan, J., & Suslow, V. (2008, May 17). Effect of collusive agreements on international trade: Evidence from an international cartel. Paper at the International Industrial Organization Conference, Arlington, VA. Retrieved from http://papers.ssrn. com/sol3/papers.cfm?abstract_id = 1628822 Levenstein, M., Sivadasan, J., & Suslow, V. (2011). The effect of competition on trade patterns: Evidence from the collapse of international cartels. Ross School of Business Paper No. 1147. Retrieved from SSRN: http://ssrn.com/abstract = 1628822 Levenstein, M., Suslow, V., & Oswald, L. (2003, February). International price-fixing cartels and developing countries: A discussion of effects and policy remedies. Working Paper No. 9511. National Bureau of Economic Research, Cambridge, MA. Levenstein, M. C. (1995). Price mass production conquers the pool: Firm organization and the nature of competition in the nineteenth century. Journal of Economic History, 55, 575611. Levenstein, M. C. (1997). Wars and the stability of collusion: A study of the pre-World War I bromine industry. Journal of Industrial Economics, 45, 117137. Levenstein, M. C. (2000). Strategic consumer behavior and cartel stability: The potash industry 18701914. Ann Arbor, MI: mimeo. Levenstein, M. C., Sivadasan, J., & Suslow, V. Y. (2011, February). The effect of competition on trade patterns: Evidence from the collapse of international cartels. Working Paper No. 02-001. University of Michigan, Ross School of Business. Levenstein, M. C., & Suslow, V. Y. (2002, January). What determines cartel success? Working Paper No. 1147. University of Michigan Business School (updated in 2006 below). Levenstein, M. C., & Suslow, V. Y. (2003). Contemporary international cartels and developing countries: Economic effects and implications for competition policy. Antitrust Law Journal, 71, 801852. Levenstein, M. C., & Suslow, V. Y. (2004a). Cartel contract duration: Empirical evidence from international cartels. In P. Z. Grossman (Ed.), How cartels endure and how they fail: Studies of industrial collusion. Cheltenham: Edward Elgar. Levenstein, M. C., & Suslow, V. Y. (2004b, November 10). The changing international status of export cartel exemptions. Retrieved from http://ssrn.com/abstract = 618201
Cartel Overcharges
361
Levenstein, M. C., & Suslow, V. Y. (2006, March). What determines cartel success? Journal of Economic Literature, 64, 4395. Levenstein, M. C., & Suslow, V. Y. (2010, September). Breaking up is hard to do: Determinants of cartel duration. Ross School of Business Paper No. 1150. Retrieved from SSRN: http://ssrn.com/abstract = 1676968 Levitt, S. D., & Venkatesh, S. A. (1998, June). Drug-selling gang’s finances. NBER Working Paper No. 6592. National Bureau of Economic Research. Levitt, S. D., & Venkatesh, S. A. (2000, August). An economic analysis of a drug-selling gang’s finances. Quarterly Journal of Economics, 115, 755789 (later published as a book chapter). Levy, H. (1927). Monopoly, cartels and trusts in british industry (2nd ed.). London: Cass & Co. (reprinted 1968). Lieber, J. B. (2000). Rats in the grain: The dirty tricks and trials of archer daniel midland. New York, NY: Four Walls Eight Windows. Liefmann, R. (1897). Die Unternehmerverbande, Konventionen, Kartelle. Liefmann, R. (1932). Cartels, concerns and trusts. New York, NY: Dutton. Lister, L. (1960). Coal: Concerted practices, in Europe’s coal and steel community: An experiment in economic union (pp. 251281). New York, NY: Twentieth Century Fund. Liu, J. (2011, February 25). Buyers’ cartels: An empirical study of prevalence and economic characteristics. Honor’s thesis, Department of Agricultural Economics, Purdue University, West Lafayette, IN. Loescher, S. M. (1980, Winter). Economic collusion, civil conspiracy, and treble-damage deterrents: The sherman act breakthrough with southern plywood. Quarterly Review of Economics and Business, 20, 635. Lord, R. A., & Farr, W. K. (2003, Spring). Collusion and financial leverage: An analysis of the integrated steel mill industry. Financial Management, 127148. Lorenz, C. (2006, March). Screening markets for cartel detection Collusive marker in the CDF cartel audit. Muenster University Working Paper. Institute of Public Economics, Muenster, Germany. Low, D. M. (2004, May 1718). Cartel enforcement, immunity, and jurisdiction: Some recent canadian developments. Speech at the International Bar Association, Rome, Italy. Low, D. M., & Halladay, C. W. (2011). Redesigning a criminal cartel regime: The Canadian conversion, Chapter 4. In C. Beaton-Wells & A. Ezrachi (Eds.), Criminalising cartels: Critical studies of an international regulatory movement. London: Hart. Lozowick, A. H., Steiner, P. O., & Miller, R. (1968). Law and quantitative multivariate analysis: An encounter. Michigan Law Review, 66, 1641–1678. Luebbers, T. (2009, March). Is cartelization profitable? A case study of the Rhenish Westfalian coal syndicate, 18931913 (Preprint 2009/9). Bonn: Max Planck Institute. Ma, T.-C. (2005a). An empirical test of the collusive equilibrium in a quantity-setting supergame. Unpublished paper. National Kaohsiung University of Applied Sciences, Taiwan. Ma, T.-C. (2005b, May). Strategic investment and excess capacity: A study of the Taiwanese flour industry. Journal of Applied Economics, 1, 153170. MacAvoy, P. W. (1965). The economic effects of regulation. Cambridge, MA: MIT Press. MacGregor, D. H. (1927). International cartels. Geneva: League of Nations. MacKie-Mason, J. K., & Pindyck, R. S. (1987). Cartel theory and cartel experience in international minerals markets. In R. L. Gordon, H. D. Jacoby, & M. B. Zimmerman (Eds.), Energy: Markets and regulation. Cambridge, MA: MIT Press.
362
JOHN M. CONNOR
Madhavan, A., Masson, R. T., & Lesser, W. H. (1994). Cooperation for monopolization? An empirical analysis of cartelization. Review of Economics and Statistics, 76 161–175. Mailmath, G. J., & Zemsky, P. (1991). Collusion in second price auctions with homogeneous bidders. Games and Economic Behavior, 3, 467486. Marion, B. (2007). Willard Fritz Mueller, pp. 23234 in De Jong and Shepherd, 2007. Mariuzzo, F., & Walshy, P. P. (2009, April). Deterministic demand cycles in cartel price data: The joint executive committee (18801886). Paper presented at the International Industrial Organization Conference, Boston, MA. Markham, J. W. (1952). Competition in the Rayon industry. Cambridge, MA: Harvard University Press. Marlio, L. (1930). Les ententes industrielles. Revue de Paris, 1, 823852. Marlio, L. (1947). The aluminum cartel. Washington, DC: Brookings Institution. Marshall, A. (1890). Principles of economics. London: Macmillan. Marshall, A. (1919). Industry and trade. London: Macmillan. Marshall, R. C., & Marx, L. M. (2008). Cartel price announcements: The vitamins industry. International Journal of Industrial Organization, 26, 762802. Marshall, R. C., & Marx, L. M. (2012). The economics of collusion: Cartels and bidding rings. Cambridge, MA: The MIT Press. Martin, S. (2002). Advanced industrial economics (2nd ed.). Malden, MA: Blackwell. Martin, S. (2007). Remembrance of things past: Antitrust, ideology, and the development of industrial economics. Working Paper. Economics Department, Purdue University (December 2005), West Lafayette, IN (revised and republished as Chapter 2 in The Political Economy of Antitrust). Martinez, A. P. (2007, October). Crushed rock cartel case. International Competition Network Workshop. Marvel, H. P., Netter, J. M., & Robinson, A. M. (1987). Price fixing and civil damages: An economic analysis. Stanford Law Review, 40, 561. Mason, C. (2004). The art of the steal: Inside the Sotheby’s-Christie’s auction house scandal. New York, NY: G.P. Putnam’s Sons. Maurer, H., & Mirow, K. (1982). Webs of power: International cartels and the world economy. Boston, MA: Houghton Mifflin. McClave, J., Rothrock, P., & Ailstock, J. P. (1978). Computerized statistical investigation of the Florida bread market. Southeastern Anti-Trust Review, 1, 1354. McCrosty, W. H. (1907). The trust movement in British industry: A study of business organization. London: Longmans. McGee, J. S. (1960). Ocean freight rate conferences and the American merchant marine. University of Chicago Law Review, 27, 191314. McMillan, J. (1991). Dango: Japan’s price-fixing conspiracies. Economic and Politics, 3, 201218. McMillan, J. (2002). Reinventing the bazaar: A natural history of markets. New York, NY: W. W. Norton. Mehmood, J. (1999). Monopoly control authority asks cement makers to deposit Rs. 4 billion excess charge in Bait-ul-Maal. Business Recorder, February 22. Mehra, S. K. (2004). More is less: A law-and-economics approach to the international scope of private antitrust enforcement. Temple Law Review, 77, 4770. Merchant, E. O. (1920). The government and the newsprint manufacturers. Quarterly Journal of Economics, 34, 313328.
Cartel Overcharges
363
Messerlin, P. A. (1990, December). Antidumping regulations or procartel law? The EC chemical case. Working Paper No. WPS397. International Economics Department, The World Bank (April) (later published in The World Economy 13). Retrieved from http://wwwwds.worldbank.org/servlet/WDSContentServer/WDSP/IB/1990/04/01/000009265_ 3960929174031/Rendered/PDF/multi0page.pdf Michels, R. K. (1928). Cartels, combines and trusts in post-war Germany. New York, NY: Columbia University Press. Mihm, M. M. (2003, May 13). Decision in re: High fructose corn syrup antitrust litigation, Gray & Co. v. Adm et al., 97-1203. MDL No. 1087 and Master File No. 95-1477. U.S. District Court for the Central District of Illinois, Peoria Division. Miller, N. H. (2009, June). Strategic leniency and cartel enforcement. American Economic Review, 99, 750768. Miyajima, H. (1994). Strategic intervention against international cartels: The case of the Japanese chemical industries in the interwar period. In B. Dominique (Ed.), International cartels revisited, 18801980 (pp. 223232). Caen, France: Editions-diffusion du Lys. Møllgaard, P. (2006, June). Assessment of damages in the district heating pipe cartel. Working Paper No. 10-2006. Department of Economics, Copenhagen Business School. Montant, G. (2004). The effectiveness of the Nord-Pas-de-Calais coal cartel during the interwar period: A research note. Explorations in Economic History, 41, 300307. Montero, J. P., & Guzman, J. I. (2005, September). Welfare-enhancing collusion in the presence of a competitive fringe. Working Paper No. 298. Pontifical University of Chile. Morrison, C. J. (1990). Market power, economic profitability, and productivity growth measurement: An integrated structural approach. NBER Working Paper No. 3355. National Bureau of Economic Research. Morrison, C. J. (1993). Productive and financial performance in U.S. manufacturing industries: An integrated structural approach. Southern Economic Journal, 69, 376392. Morrison, S. A., & Winston, C. (1996). Causes and consequences of airline fare wars. Brookings Papers on Economic Activity: Microeconomics, 85123. Morse, A. B., & Hyde, J. (2003, April). Estimation of cartel overcharges: The case of Archer Daniels Midland and the market for lysine. Staff Paper No. 00-8. Department of Agricultural Economics, Purdue University (2000), West Lafayette, IN (revised version presented at the 1st International Industrial Organization Conference, Boston, MA). Morton, S., & Fiona, M. (1997). Entry and predation: British Shipping cartels 18791929. Journal of Economics and Management Strategy, 6, 679724. MRA. (2013). Market resaerch.com Academic (online library research service, frequent source of market sales estimates, mostly producer goods). Mueller, W. F. (2007). George ward stocking, pp. 187189 in De Jong and Shepherd 2007. Mueller, W. F., & Parker, R. C. (1992). Bakers of washington cartel: Twenty-five years later. Review of Industrial Organization, 7, 7582. Mund, V. (1960). Identical bid prices. Journal of Political Economy, 68, 150169. Nelson, J. P. (1993). Comparative antitrust damages in bid-rigging cases: Some findings from a used vehicle auction. Antitrust Bulletin, 38, 369394. New York State Attorney General. (2006, January 18). Agreement between the attorney general of the state of New York and American International Group. Retrieved from http://www. ins.state.ny.us/acrobat/oag0209061.pdf
364
JOHN M. CONNOR
Newell, E. (1998). “The irremediable evil”: British copper smelters’ collusion and the cornish mining industry, 17251865, in Bruland and O’Brien. Newfarmer, R. S. (1978). The international market power of TNCs (UNCTAD/ST/MD/13). New York, NY: United Nations. Newman, P. C. (1948). Key German cartels under the Nazi regime. Quarterly Journal of Economics, 62, 576595. Newmark, C. M. (1988). Does horizontal price fixing raise price? A look at the bakers of Washington. Journal of Law and Economics, 31, 469484. Normann, H.-T., & Tan, E. (2013, September). Effects of different cartel policies: Evidence from the German power-cable industry. Discussion Paper No. 108. DICE, Duesseldorf, Germany (forthcoming Industrial and Corporate Change 2014). Normann, H.-T., & Tan, E. (2005, October). The effects of cartel policy: Evidence from the German power-cable industry. Working Paper. Royal Holloway College, University of London. Notz, W. F. (1920). International private agreements in the form of cartels, syndicates, and other combinations. Journal of Political Economy, 28, 658679. Notz, W. F. (1929). Representative international cartels, combines, and trusts: Trade promotion series No. 81. Washington, DC: U.S. Department of Commerce. OECD. (1974). Export cartels. Paris: Organization of Economic Co-Operation and Economic Development. OECD. (1976). Collusive tendering. Paris: Organization of Economic Cooperation and Development. OECD. (2001, October). Summary of cartel cases described by invitees (CCNM/GF/COMP (2001)4). Paris: OECD. OECD. (2002, 9 April). Report on the nature and impact of hard core cartels and sanctions against cartels under national competition laws (DAFFE/COMP (2002) 7). Organization of Economic Co-Operation and Development, Paris (a/k/a “The Second Report on Cartels”). Retrieved from http://www.oecd.org/dataoecd/16/20/2081831.pdf OECD. (2003a). Competition law and policy: 20022003 annual reports. Organization of Economic Co-Operation and Development, Paris. Retrieved from http://browse.oecd bookshop.org/oecd/pdfs/browseit/2403011E.PDF (National antitrust authorities’ annual reports appended; site goes back to 19941995). OECD. (2003b, June 1113). Case of cartel agreement in the construction sector. OECD workshop on competition policy, Vilnius, Lithuania. Retrieved from www.konkuren.lt/ English/cases_article_5_vilnius_case.html OECD. (2005a). Evaluations of the actions and resources of competition authorities (DAF/ COMP(2005)30. Paris: Organization of Economic Co-Operation and Development. OECD. (2005b, December 15). Hard core cartels: Third report on the implementation of the 1998 recommendation. Paris: Organization of Economic Co-Operation and Development. OECD. (2005c). Competition law and policy in Brazil. Paris: Organization of Economic Co-Operation and Development. OECD. (2006a). Roundtable on prosecuting cartels without direct evidence of agreement: Contribution from Italy (DAF/COMP/GF/WD(2006)5). Paris: Organization of Economic Co-Operation and Development. OECD. (2006b). Roundtable on prosecuting cartels without direct evidence of agreement: Contribution from Argentina (DAF/COMP/GF/WD(2006)7. Paris: Organization of Economic Co-Operation and Development.
Cartel Overcharges
365
OECD. (2006c). Roundtable on prosecuting cartels without direct evidence of agreement: Contribution from Lithuania (DAF/COMP/GF/WD(2006)9. Paris: Organization of Economic Co-Operation and Development. OECD. (2006d). Roundtable on prosecuting cartels without direct evidence of agreement: Contribution from Turkey (DAF/COMP/GF/WD(2006)17. Paris: Organization of Economic Co-Operation and Development. OECD. (2006e). Roundtable on prosecuting cartels without direct evidence of agreement: Contribution from Chinese Taipei (DAF/COMP/GF/WD(2006)20. Paris: Organization of Economic Co-Operation and Development. OECD. (2006f). Roundtable on prosecuting cartels without direct evidence of agreement: Contribution from Peru (DAF/COMP/GF/WD(2006)27. Paris: Organization of Economic Co-Operation and Development. OECD. (2006g). Roundtable on prosecuting cartels without direct evidence of agreement: Contribution from Romania (DAF/COMP/GF/WD(2006)30. Paris: Organization of Economic Co-Operation and Development. OECD. (2006h). Roundtable on prosecuting cartels without direct evidence of agreement: Contribution from Korea (DAF/COMP/GF/WD(2006)38. Paris: Organization of Economic Co-Operation and Development. OECD. (2007). Competition and efficient usage of payment cards DAF/COMP (2006) 32. Paris: Organization of Economic Co-Operation and Development. Officer, M. S., Ozbas, O., & Sensoy, B. A. (2010). Club deals in leveraged buyouts. Journal of Financial Economics, 98(2), 214–240. OFT. (2003). MasterCard interchange fees OFT634. London: UK Office of Fair trading. Ohio OIG. (2010, January 6). Report of investigation. Columbus, OH: Office of the Inspector General, State of Ohio. Osiatynski, J. (Ed.). (1996). Collected works of michael kalecki (Vol. 6). Oxford: Clarendon. Oualid, W. (1938). International raw commodity price control. Paris: International Studies Conference, League of Nations. Page, W. (1980). Antitrust damages and economic efficiency: A study in law and economics. University of Chicago Law Review, 47, 1539. Page, W. H. (General Editor). (1996). Proving antitrust damages: Legal and economic issues. Chicago, IL: Section of Antitrust Law, American Bar Association. Parker, A. A. (1972). Measuring damages in federal treble damages actions. Antitrust Bulletin, 17, 497517. Passer, H. C. (1953). The electrical manufacturers 18971900: A study in competition, entrepreneurship, technical change, and economic growth. Cambridge, MA: Harvard University Press. Paxton, R. O. (1992). The calcium carbide case and the decriminalization of industrial ententes in France, 19151926. In P. Fridenson (Ed.), The French home front, 19141918. Berg Publishers. PCA. (various dates). Press releases of the Portugal competition authority. Retrieved from http://www.autoridadedaconcorrencia.pt/Aconcorrencia_C03.aspx?cat = Cat_Processos Decisoes_OutrosProceDecisoes&idmenu = 2 Peck, M. J. (1961). Competition in the aluminum industry 19451958. Cambridge, MA: Harvard University Press. Pesendorfer, M. A. (2000). Study of collusion in first-price auctions. Review of Economic Studies, 67, 381411.
366
JOHN M. CONNOR
Peterman, J. L. (1979). The international salt case. The Journal of Law and Economics, 22, 351364. Peters L. L. (1989). Managing competition in german coal, 18931913. Journal of Economic History, 49, 419434. Peterson, E. B., & Connor, J. M. (1995). A comparison of oligopoly welfare loss estimates for US food manufacturing. American Journal of Agricultural Economics, 77: 300–308. Pfeiffer, R. A. C. (2005). Recent aspects of hard core prosecution in Brazil. Presentation at the third Latin American competition forum, Madrid. Phillips, A. (1972). Price fixing in UK manufacturing industries 19531956. In K. Cowling (Ed.), Market structure and corporate behaviour. London: Gray-Mills. Pierce, F. (1913). The trusts resulting from the protective tariff and leading to socialism, Chapter 2. The tariff and the trust (pp. 4788). New York, NY: Macmillan. Pierce, R. J. (2000). Essay: Antidumping law as a means of facilitating cartelization. Antitrust Law Journal, 76, 725. Piotrowski, R. (1933). Cartels and trusts. London: George Allen & Unwin. Pirrong, S. C. (1992). An application of core theory to the analysis of ocean shipping markets. Journal of Law and Economics, 35, 89131. Plott, C. R. (1989). An updated review of industrial organization: Applications of experimental methods, Chapter 19. In R. Schmalensee & R. D. Willig (Eds.), Handbook of industrial organization (Vol. 2, pp. 11091176). Amsterdam: North-Holland. Plummer, A. (1934, 1951). International combines in modern industry. London: Pittman. Porter, R. H., & Zona, J. D. (1993). Detection of bid rigging in procurement markets. Journal of Political Economy, 101, 518538. Porter, R. H., & Zona, J. D. (1999). Ohio school milk markets: An analysis of bidding. RAND Journal of Economics, 30, 263288. Porter, R. H., & Zona, J. D. (2004). Bidding, bid rigging, and school milk prices: Ohio v. Trauth (1994). In J. E. Kwoka & L. J. White (Eds.), The antitrust revolution (4th ed.). New York, NY: Oxford University press. Porter, R. H. A. (1983). Study of cartel stability: The joint executive committee, 18801886. Bell Journal of Economics and Management, 14, 301314. Porter, R. H. A. (1992). Book review: Anatomy of an auction: Rare books at Ruxley Lodge, 1919. Journal of Political Economy, 100, 433436. Porter, R. H. A. (2001). Detecting collusion among bidders in auction markets. In Fighting cartels why and how? Stockholm: Konkurrensverket. Porter, R. H. A. (2005). Detecting collusion. Review of Industrial Organization, 26, 147167. Posner, R. A. (1975). The social costs of monopoly and regulation. Journal of Political Economy, 83, 807828. Posner, R. A. (1976). Antitrust law: An economic perspective. Chicago, IL: University of Chicago Press. Posner, R. A. (2001). Antitrust law. (2nd ed.). Chicago, IL: University of Chicago Press. Posner, R. A. (2004, February 4). Personal communication to John M. Connor. Powers, N. J. (1991). Marketing order impacts on farm-retail price spreads: Comment. American Journal of Agricultural Economics, 73, 507510. Prain, R. (1976). Copper: Anatomy of an industry. London: Mining Journal Books Limited. Prais, S. J. (1974, December). The electric lamp monopoly and the life of electric lamps. Journal of Industrial Economics, 23, 153158. Pribram, K. (1935). Cartel problems. Washington, DC: Brookings.
Cartel Overcharges
367
Price, R. M., & Bonett, D. G. (2002). Distribution-free confidence intervals for difference and ratio of medians. Journal of Statistical Computational Simulation, 72, 119124. PriceWaterhouseCoopers. (2002). Alternative calculation of damages based on ALSTOM’s economic profit. Hellerup, Denmark. Procurement News. (2010, May 4). Spain: Spanish competition authority fines three companies $11.6 million for bath gel cartel. Retrieved from www.tendersinfo.com Pursiainen, H., Ma¨lko¨nen, V., Saxell, T., Tukiainen, J., & Siikanen, M. (2011, June). Asfalttikartellin kunnille aiheuttamien vahinkojen arviointi Loppuraportti: VATT policy. Helsinki, Finland: Government Institute for Economic Research (VATT). Retrieved from report 8. http://www.vatt.fi/file/vatt_publication_pdf/valm.rap.8.pdf Pursiainen, H., Saxell, T., & Tukiainen, J. (2012, March). Asfalttikartellin vahinkojen arviointi vastauksia kritiikkiin. VATT Working Paper No.19. Government Institute for Economic Research (VATT), Helsinki, Finland. Retrieved from http://www.vatt.fi/file/ vatt_publication_pdf/muistiot_19.pdf Pyndyck, R. S. (1979). The cartelization of world commodity markets. American Economic Review, 69, 154158. Read, R. (1993). Cartels and internalization in the eighteenth-century copper industry. In J. Brown & M. B. Rose (Eds.), Entrepreneurship, networks and modern business. Manchester: Manchester University Press. Reader, W. J. (1993). Imperial chemical industries: A history (1975). In G. Jones (Ed.), Coalitions and collaboration in international business (pp. 283316). Aldershot, UK: Edward Elgar. Rees, R. (1993). Collusive equilibrium in the great salt duopoly. Economic Journal, 103, 833848. Reimann, G. (1942). Patents for Hitler. New York, NY: Vanguard Press. Renda, A., van den Bergh, R. J., & Pardolesi, R. (2007, December 21). Making antitrust damages actions more effective in the EU: Welfare impact and potential scenarios: Final report (Contract DG COMP/2006/A3/012). Brussels, Rotterdam, and Rome: Centre for European Policy Studies, Erasmus University Rotterdam, and Luiss Guido Carli. Retrieved from http://ec.europa.eu/competition/antitrust/actionsdamages/files_ white_paper/impact_study.pdf#page=1. Reuter, P. (1993). The cartage industry in New York. Crime and Justice, 18, 149201. Richardson, G. B. (1966). Pricing of heavy electrical equipment. Bulletin of the Oxford University Institute of Economics and Statistics, 73, 7392. Ripley, W. Z. (Ed.). (1916). Trusts, pools and corporations (rev. ed.). Boston, MA: Ginn. Ro¨ller, L.-H., & Steen, F. (2003). On the workings of a cartel: Evidence from the Norwegian cement industry. Discussion Paper No. 25/03. Norwegian School of Economics and Business Administration, Bergen, Norway. Ro¨ller, L.-H., & Steen, F. (2006). On the workings of a cartel: Evidence from the Norwegian cement industry. American Economic Review, 96, 321338. Ross, T. (2005). Affidavit of Thomas Ross in Glen Ford et al. v. F. Hoffmann-La Roche et al. Ontario Superior Court of Justice, Court File No. 00-CV-202080CP. Rothwell, R. P. (1921). The mineral industries. New York, NY: McGraw-Hill (previous annual volumes from 1893). Rowe, J. W. F. (1936). Markets and men. New York, NY: Macmillan. Rubinfeld, D. L., & Steiner, P. O. (1983). Regression estimates of damages in price-fixing cases. Journal of Law and Contemporary Problems, 46, 68144.
368
JOHN M. CONNOR
Russo, F., Schinkel, M. P., & Gu¨nster., A. (2010). European Commission decisions on competition: Economic perspectives on landmark antitrust and merger cases. Cambridge: Cambridge University Press. Russo, F., Schinkel, M. P., Gu¨nster, A., & Carree, M. (2007). European Commission decisions on competition: Landmark antitrust and merger cases from an economic point of view. Amsterdam: ACLE (draft of September). Salop, S. C., & White, L. J. (1988). Private antitrust litigation: Introduction and Framework. In L. White (Ed.), Private Antitrust Litigation. Cambridge, MA: Massachusetts Institute of Technology Press. Salvo, A. (2007, February). Inferring conduct under the threat of entry: The case of the Brazilian cement industry. SSRN Working Paper 868543. http://ssrn.com/abstract = 868543 Samkeppniseftirlitid. (2005, January 29). The oil cartel in Iceland. Report of the Icelandic Competition Authority, Reykjavik, Iceland (Translation September 23, 2005). Retrieved from www.samkepni, downloaded October 28, 2005. Sarkar, G. K. (1972). The world tea economy. Bombay: Oxford University Press. Sarkar, S. (1996, July). Did the U.S. gypsum industry really collude? Some evidence using tests on non-nested hypotheses. Journal of Quantitative Economics, 12, 119. Sauvy, A., & Hirsch, A. (1984). Histoire Economique de la France entre les Deux Guerres. Paris: Economica. Saxell, T., Tukiainen, J., & Siikanen, M. (2011). Asfalttikartellin valtiolle aiheuttamien vahinkojen arviointi: Valmisteluraportit 9 (pp. 121). Helsinki, Finland: Government Institute for Economic Research (VATT). Retrieved from http://www.vatt.fi/file/vatt_ publication_pdf/valm.rap.9.pdf Sayous, A. E. (1902). Cartels and trusts in Holland in the seventeenth century. Political Science Quarterly, 17, 381393. Schalk, W. (1891). Over Ondernemers-Verenigingen (Kartels en Trusts). Leiden: Leiden University. Schaub, A. (2001). Cartel enforcement by the European Commission. In Fighting cartels Why and how? Stockholm: Konkurrensverket. Scherer, F. M. (1996). Industry structure, strategy, and public policy. New York, NY: HarperCollins. Scherer, F. M. (2007, June). Class actions in the U.S. experience: An economist’s perception. Research Working Paper RWP07-028. Kennedy School of Government, Harvard University, Cambridge. Scherer, F. M., & Ross, D. (1990). Industrial market structure and economic performance (3rd ed.). Boston, MA: Houghton Mifflin. Schinkel, M. P. (2009). Interview by ZEMBLA TV. The Dutch construction cartel: documentaries on CD. VARA/NPS. Schinkel, M. P. (2010, October 15). Market oversight games. Amsterdam Center for Law & Economics. Working Paper No. 2010-11. Retrieved from: SSRN: http://ssrn.com/ abstract = 1692733 Schmitt, N., & Weder, R. (1998). Sunk costs and cartel formation: Theory and application to the dyestuff industry. Journal of Economic Behavior & Organization, 36, 197220. Schro¨ter, H. G. (1988). Cartels as a form of concentration in industry: The example of the international dyestuffs cartel from 1927 to 1939, reprinted in Jones (1993), pp. 109140.
Cartel Overcharges
369
Schro¨ter, H. G. (1994). The international potash syndicate. In D. Bardot (Ed.), International cartels revisited, 18801980. Caen, France: Editions-diffusion du Lys. Schro¨ter, H. G. (1999). Cartelization and decartelisation in Europe, 18701995: Rise and decline of an economic institution. Journal of Economic History, 25, 129153. Scott, F. A. (2000). Great school milk conspiracies revisited. Review of Industrial Organization, 17, 325341. Seager, H. R., & Gulick, C. A. (1929). Trust and corporation problems (second printing). New York, NY: Harper. Shanahan, M. P., & Round, D. K. (2008). Serious cartel conduct, criminalization, and evidentiary standards: The Coal Vend case of 1911 in Australia. Paper presented at the European Business History Association Conference, Bergen (subsequently in Business History, 51(2009): 875–906). Retrieved from http://scholar.google.com/scholar?hl=en& q=serious+cartel+coal+australia&btnG=&as_sdt=1%2C22&as_sdtp= Sheer, A., & Ho, C. (1989, August 16). Preliminary review of sentences imposed on organizations in 1988: Memorandum to the U.S. Sentencing Commission. Shepard, L. (1986). Cartelization of the CaliforniaArizona orange industry, 19341981. Journal of Law and Economics, 29, 83123. Shepard, W. G. (1985). The economics of industrial organization (7th ed.). Homewood, IL: Richard O. Irwin. Shin, K. (2002). The treatment of market power in Korea. Review of Industrial Organization, 21, 113128. Sjostrom, W. (1989). Collusion in ocean shipping: A test of monopoly and empty core models. Journal of Political Economy, 75, 11601179. Sjostrom, W. (1991). An empirical test of monopoly behavior: An application to the hardwood case: A comment. Applied Economics, 23, 339341. Smith, A. (1776). The wealth of nations. New York, NY: Modern Library (reprinted 1937). Smith, D. C. (1970). History of papermaking in the United States (16911969). New York, NY: Lockwood. Smith, R. A. (1963). Corporations in crisis. Garden City, NY: Doubleday. Sørgard, L. (2007, February 22). Cartels investigations in Norway. Seminar hosted by the Norwegian Competition Authority, Bergen, Norway. Spar, D. (1994). Cooperative edge: The internal politics of international cartels. Ithaca: Cornell University Press. Spar, D. (2006, Summer). Markets: Continuity and change in the international diamond market. Journal of Economic Perspectives, 20, 195. Spratling, G. R. (2001). Detection and deterrence. George Washington Law Review, 69, 817823. Sproul, M. F. (1993). Antitrust and prices. Journal of Political Economy, 101, 741754. Staley, E. (1937). Raw materials in peace and war. New York, NY: Council on Foreign Relations. STAT-USA. (various dates). On-line U.S. trade data. Washington, DC: U.S. Department of Commerce. Steen, F., & Sørgard, L. (1999). Semicollusion in the Norwegian cement market. European Economic Review, 43, 17751796. Stevens, W. S. (1912a). The powder trust, 18721912. Quarterly Journal of Economics, 26, 444481.
370
JOHN M. CONNOR
Stevens, W. S. (1912b). The dissolution of the powder trust. Quarterly Journal of Economics, 27, 202207. Stevens, W. S. (1912c). A group of trusts and combinations. Quarterly Journal of Economics, 26, 593644. Stewart, J. B. (2001). Bidding war. The New Yorker, October 15, p. 158. Stigler, G. J., & Kindahl, J. K. (1970). The behavior of industrial prices. New York, NY: Columbia University Press. Stiglitz, J. (1998). Globalization and its discontent. New York, NY: New Press. Stockder, A. H. (1932). Regulating an industry: The Rhenish-Westphalian coal syndicate 18931929. New York, NY: Columbia University Press. Stocking, G. W. & Watkins, M. W. (1946). Cartels in action. New York, NY: Twentieth Century Fund. Stocking, G. W. & Watkins, M. W. (1948). Cartels or competition? New York, NY: Twentieth Century Fund. Stocking, G. W. & Watkins, M. W. (1951). Monopoly and free enterprise. New York, NY: Twentieth Century Fund. Stucke, M. E. (2006). Morality and antitrust. Columbia Business Law Review, 2006, 443538. Sultan, R. G. M. (1974, 1975). Pricing in the electrical oligopoly (Vols. 1 and 2), Boston, MA: Harvard Business School. Sullivan, L. A., & Grimes, W. S. (2006). The law of antitrust (2nd ed.). Thompson West. Sunneva˚g, K. J. (2007). Personal communications, dated February 2 and March 1. Suslow, V. Y. (1986). Estimating monopoly behavior with competitive recycling: An application to Alcoa. RAND Journal of Economics, 17, 389403. Suslow, V. Y. (2001, October). Cartel contract duration: Empirical evidence from international cartels. Unpublished manuscript that updates a 1988 Hoover Institution Working Paper, University of Michigan Business School. Suslow, V. Y. (2005). Cartel contract duration: Empirical evidence from inter-war international cartels. Industrial and Corporate Change, 14, 705744. Swann, D., O’Brien, D. P., Maunder, W. P. J., & Howe, W. S. (1974). Competition in British industry: Restrictive practices legislation in theory and practice (Vol. 26). London: Allen & Unwin. Sweezy, P. M. (1938). Monopoly and competition in the English coal trade 15501850. Cambridge, MA: Harvard University Press. Symeonidis, G. (2002). The effects of competition: Cartel policy and the evolution of strategy and structure in British industry. Cambridge, MA: MIT Press. Tan, E. S. (2003, October 25). Economic regulation and rent protection: A study of the coal cartel in the first industrial revolution: WP03-23. Ann Arbor: Michigan Business School. Tan, E. S. (2009). Market structure and the coal cartel in early nineteenth-century England. Economic History Review, 62, 350365. Taylor, C. R. (2003, October 15). Impacts of captive supplies on the cash market price for slaughter cattle: Updated. Report prepared for Pickett, et al, v. Iowa Beef Packers, Inc., Civil Action No. 96-A-1103-N, U.S. District Court, Middle District of Alabama Northern Division. Taylor, C. R. (2013, March). The world’s cartelized fertilizer industry. AAI Working Paper 13-XX. American Antitrust Institute, Washington, DC. Taylor, J. E. (2002). The output effects of government-sponsored cartels during the new deal. Journal of Industrial Economics, 50, 110.
Cartel Overcharges
371
Taylor, J. E. (2010). The welfare impact of collusion under various industry characteristics: A panel examination of efficient cartel theory. B.E. of Economic Analysis and Policy, 10, 127. Taylor, J. H. & Yokell, M. D. (1979). Yellowcake: The international uranium cartel. New York, NY: Pergamon. Taylor, W. G. L. (1905). The world’s prosperity in the year 1903. Journal of Political Economy, 13, 217236. Teece, D. J., Sunding, D., & Mosakowski, E. (1993). Natural resource cartels. In A. V. Kneese, J. L. Sweeney (Eds.), Handbook of natural resource and energy economics (Vol. 3, pp. 1131–1166). Amsterdam: North-Holland. Telser, L. G. (1985). Cooperation, competition, and efficiency. Journal of Law and Economics, 28, 271295. TFTC. (date of decision). Decisions of the Taiwan fair trade commission. Taipei, Taiwan. TFTC-OECD. (2001). OECD global forum on competition: Contribution on competition (CCNM/GF/WD(2001)16). Paris: OECD. Thompson, G. D., & Lyon, C. C. (1989). Marketing order impacts of farm-retail price spreads: The suspension of prorates on California-Arizona navel oranges. American Journal of Agricultural Economics, 71, 647660. Thompson, J. S., & Kaserman, D. L. (2001). After the fall: Stock price movements and the deterrent effect of antitrust enforcement. Review of Industrial Organization, 19, 329334. Tilton, M. (2004). The difference government policy makes: The case of Japan, Chapter 7. In Grossman (Ed.), How cartels endure and how they fail: Studies of industrial collusion. Cheltenham: Edward Elgar. Timberlake, E. C. (1961). The legal injury requirements and proof of damages in treble damage actions under the antitrust law. The George Washington Law Review, 30, 231284. Tosdal, H. R. (1916). The German potash syndicate. In Ripley, W. Z. (Ed.), Trusts, pools and corporations (Rev. ed., pp. 795832). Boston, MA: Ginn. Troesken, W. A. (1989). Note on the efficacy of the german steel and coal syndicates. Journal of European Economic History, 18, 595603. Tsokhas, K. (2000). The rise and decline of an international zinc and lead cartel. Australian Economic History Review, 40, 263286. Tweede, K. (2002). Tweede Kamer report of December 12, 2002 (2001-2, 28093, No. 11). The Hague, Netherlands. U.S. Antitrust Modernization Commission. (2007, April). Report and recommendations. Washington, DC. Retrieved from http://govinfo.library.unt.edu/amc/report_recommen dation/amc_final_report.pdf U.S. Congress. (19381940). Hearings of the temporary national economic committee (Vols. 111). Washington, DC: U.S. Government Printing Office. U.S. Congress. (1965). Staff study of income tax treatment of treble damages payments under the antitrust laws. Washington, DC: Joint Committee on Taxation, U.S. Government Printing Office. U.S. Congress. (19661967). Hearings on prices of quinine and quinidine. Subcommittee on antitrust, U.S. Senate Committee on the Judiciary. Washington, DC: U.S. Government Printing Office. U.S. Congress. (1977). Hearings on the international uranium cartel. Subcommittee on Oversight and Investigations, Committee on Interstate and Foreign Commerce, U.S. House of Representatives. Washington, DC: U.S. Government Printing Office.
372
JOHN M. CONNOR
U.S. Industrial Commission. (1901, December). Report on trusts and industrial combinations: Document 182: Volume XIII. Washington, DC: U.S. House of Representatives. U.S. Sentencing Commission. (1986, July 15). United States sentencing guidelines: unpublished public hearings for the first guidelines made law in 1987. Washington, DC: United States Sentencing Commission. U.S. Tariff Commission. (1939). Incandescent electric lamps (2nd Series). Report No. 133. Washington, DC: U.S. Government Printing Office. Uctum, M. (1998). Why have corporate profits declined? An international comparison. Review of International Economics, 6, 234251. UK Monopolies Commission. (1951). Report on the supply of electric lamps. London: HMSO. UK Monopolies Commission. (1952a). Report on the supply of insulated electric wires and cables. London: HMSO. UK Monopolies Commission. (1952b). Report on the supply of insulin. London: HMSO. UK Monopolies Commission. (1953). Report on the supply and export of matches and the supply of match-making machinery. London: HMSO. UK Monopolies Commission. (1955). Report on the supply and export of certain semimanufactures of copper and copper-based alloys. London: HMSO. UK Monopolies Commission. (1956a). Report on the supply of hard fibre cordage. London: HMSO. UK Monopolies Commission. (1956b). Report on the supply of certain industrial and medical gases. London: HSMO. UK Monopolies Commission. (1956c). Report on the supply of linoleum. London: HMSO. UK Monopolies Commission. (1957a). Report on the supply and export of electrical and allied machinery and plant. London: HMSO. UK Monopolies Commission. (1957b). Report on the supply of electronic valves and cathode ray tubes. London: HMSO. UK Monopolies Commission. (1959). Report on the supply of chemical fertilizers. London: HMSO. UK Monopolies Commission. (1963). Report on the supply of electrical equipment for mechanically propelled land vehicles. London: HMSO. UK Monopolies Commission. (1964). Report on the supply of wallpaper. London: HMSO. UK Monopolies Commission. (1968a). Flat glass: A report on the supply of flat glass. London: HMSO. UK Monopolies Commission. (1968b). Man-made cellulosic fibres: A report on the supply of man-made cellulosic fibres. London: HMSO. UK Monopolies Commission. (1970). Metal containers. London: HMSO. UK Monopolies Commission. (1971). Starch, glucoses, and modified starches. London: HMSO. UK Monopolies Commission. (1973). Chlordiazepoxide and diazapam: A report on the supply of chlordiazepoxide and diazapam. London: HMSO. Includes the Monopolies and Restrictive Practices Commission and the Monopolies and Mergers Commission. UK Monopolies Commission. (1981). Concrete roofing tiles. London: HMSO. UK Monopolies Commission. (1986). White salt. London: HMSO. UK Monopolies Commission. (1989). Opium derivatives. London: HMSO. UK Monopolies Commission. (1990). Plasterboard. London: HMSO. UKCC. (2001, July). BASF and Takeda Chemical Industries Ltd: A report on the acquisition by BASF AG of certain assets of Takeda Chemical Industries Ltd. London: UK Competition Commission.
Cartel Overcharges
373
Ulen, T. S. (1980). The market for regulation: The ICC from 1887 to 1920. American Economic Review, 70, 306310. UNCTAD. (2005a, September 5). A synthesis of recent cartel investigations that are publicly available (TD/RBP/CONF.6/4). Geneva: United Nations Conference on Trade and Development. UNCTAD. (2005b, September 6). Recent important cases involving more than one country (TD/RBP/CONF.6/5). Geneva: United Nations Conference on Trade and Development. USSG. (1987). United States sentencing guidelines. Washington, DC: U.S. Government Printing Office (first adopted by Congress and published in 1987, revised annually). USSG Advisory Group. (2003, October 7). Report of the ad hoc advisory group on the organizational guidelines. Washington, DC: U.S. Sentencing Commission. van Bergeijk, P. A. G., Meulmeester, P., & van der Laan, R. (2006, March 17). On the alleged invisibility of the Dutch construction cartels. Paper delivered at the Amsterdam Center for Law and Economics Workshop, University of Amsterdam, Amsterdam, Netherlands. (later published in the Journal of Competition Law and Economics, 2008). van Eden, H. (2002). Fighting hard-core cartels: Harm, effective sanctions and leniency programmes. Paris: OECD. von Beckerath, H. (1930). Modern industrial organization: An economic interpretation. New York, NY: McGraw-Hill (translated 1933). von Hirsch, A. (1989). Federal sentencing guidelines: Do they provide principled guidance? American Criminal Law Review, 27, 367390. Walker, F. (1906). The German steel syndicate. Quarterly Journal of Economics, 20, 353398. (Reprinted as pp. 833868 in Ripley, William Z. (Ed.). (1916). Trusts, pools and corporations (rev. ed.). Boston, MA: Ginn.) Wallace, B. B., & Edminster, L. R. (1930). International control of raw materials. Washington, DC: Brookings. Walters, A. (1944). The international copper cartel. Southern Economic Journal, 11, 133156. Wang, X. (2001). OECD global forum on competition: Contribution from China (CCNM/ COMP/WD(2001)10). Paris: OECD. Retrieved from http://www.oecd.org/dataoecd/40/ 61/2491690.pdf Webb, S. B. (1980). Tariffs, cartels, technology and growth in the German steel industry, 18791914. Journal of Economic History, 40, 309330. Weder, R. (1995). The Swiss dyestuffs industry. In M. J. Enright & R. Weder (Eds.), Studies in Swiss competitive advantage. Bern: Peter Lang. Wells, W. (2002). Antitrust and the formation of the postwar world. New York, NY: Columbia University Press. Werden, G. J. (1989, Summer). Price-fixing and civil damages: Setting the record straight. Antitrust Bulletin, 34, 307335. Werden, G. J. (2003, January). The effect of antitrust policy on consumer welfare: What Crandall and Winston overlook. EAG 03-2. Washington, DC: Economic Analysis Group, Antitrust Division, U.S. Department of Justice. Werden, G. J. (2004). Economic evidence on the existence of collusion: Reconciling antitrust law with oligopoly theory. Antitrust Law Journal, 71, 719800. Werden, G. J., & Simon, J. M. (1965). Why price fixers should go to prison. Antitrust Bulletin, 32, 924958. Whinston, M. D. (2006). Lectures on antitrust economics. Cambridge, MA: MIT Press.
374
JOHN M. CONNOR
White, L. J. (Ed.). (1988). Private antitrust litigation: New evidence, new learning. Cambridge, MA: MIT Press. Whitney, E. B. (1896). Brief submitted to the 6th Circuit Court of Appeals in U.S. v. Addyston Pipe and Steel Co. Wiedenfeld, K. (1927). Cartels and combines. Geneva: League of Nations. Wiggins, S. N., & Libecap, G. D. (1987). Firm heterogeneities and cartelization efforts in domestic crude oil. Journal of Law, Economics, and Organization, 3, 125. Wiley, J. S., Jr. (1987, Spring). Antitrust and core theory. University of Chicago Law Review 54, 556589. Wilkinson, H. L. (1914). The trust movement in Australia. Melbourne: Critchley Parker. Williams, M. A. (1986). An economic application of bootstrap statistical methods: Addyston pipe revisited. American Economist, 30, 5258. Wils, W. P. J. (2005, June). Is criminalization of EU competition law the answer? World Competition, 28, 117159. Wilson, M. R. (2003). Gentlemanly price fixing and its limits: Collusion and competition in the U.S. explosives industry during the civil war era. Business History Review, 77, 207234. Wood, F. (1998). Fuelling the local economy: The Fenland coal trade, 17601850, in Bruland and O’Brien. Woodall, B. (1996). Japan under construction: Corruption, politics, and public works. Berkeley, CA: University of California Press. Wrobel, S. (2007, August 7). Treasury finalizing law to break cooking gas cartel lower costs. Jerusalem Post, 17. Wurm, C. A. (1993). International Industrial Cartels, the State and Politics (1989). In G. Jones (Ed.), Coalitions and collaboration in international business (pp. 234245). Aldershot, UK: Edward Elgar. Yonhap, A. (various dates). Korean News Service. Yoon, J.-y. (2009). HanGlas, KCC admit glass price fixing. Korea Times, June 26 (press release of KFTC decision). Yu, Y. (2003, May). The impact of private international cartels on developing countries. Honor’s thesis, Department of Economics, Stanford University (Timothy Bresnahan, Advisor). Retrieved from http://www.econ.stanford.edu/academics/honors_thesis/ Theses_2003/Yu.pdf Yuspeh, L. (1976). A case for increased use of competitive procurement in the department of defense. In Y. Amhuid (Ed.), Bidding and auctioning for procurement and allocation. New York, NY: NYU Press. Zembla. (2009). Four documentaries: The Dutch construction cartel [a CD]. Zimmerman, J. E. (2005, May). Determinants of cartel Duration: A cross-sectional study of modern private international cartels. M.S. thesis, Purdue University. Zona, J. D. (2011) Structural approaches to estimating overcharges in price-fixing cases. Antitrust Law Journal, 77, 473–494.
375
Cartel Overcharges
APPENDIX: DATA SOURCES AND COLLECTION METHODS Selection Criteria I have made every attempt to locate and extract all useful information on private, hard-core cartel overcharges available from serious151 published sources. A private, hard-core cartel is one that by contemporary U.S. standards could be criminally indicted under the Sherman Act.152 Private cartels are those not protected by treaties or national sovereignty. Hardcore or “naked” cartels are those that made explicit agreements on horizontal restraints to control prices or limit quantities to be produced or sold. Price agreements may cover list prices or transaction prices; the transactions prices may be floor prices, target prices, or, if a common sales agency is employed, actual transactions prices. Prices may refer to sales of goods or services, procurement of inputs, or bids in auctions or tenders. Quantity restrictions most commonly involve fixed market shares for each participant, but may also include territorial exclusivity, customer allocations, production-capacity, or fringe-boycott agreements. Cartels that focused exclusively on collective action regarding vertical restraints, advertising, patent pooling, technical standards, R&D, and the like are not considered hard core. Classifying the sampled cartels at times requires judgment. Some cartels operated prior to 1890, after which passage of the Sherman Act made participation by U.S. companies illegal, but many 20th-century cartels headquartered in Europe predate the beginnings of effective European anticartel laws. If these cartels were not formed by means of a legally enforced government monopoly, they are generally considered private schemes.153 However, if a government simply required registration or chartering of a cartel but left its management in corporate hands, they are included in the data set. Beginning in 1918 in the United States and in most European countries in the interwar period, domestic producers were permitted to register and operate export cartels with no or minimal supervision; I consider these private cartels, unless they were compulsory by law. Similarly, if a government-owned national monopoly or commodity association voluntarily joins an international cartel, the latter may qualify as a private cartel. Thus, the mere fact that governments tolerated or turned a blind eye to cartels does not disqualify them from inclusion in the data set. However, commodity agreements known to have been initiated, actively sponsored, or intentionally protected by national sovereignty are not
376
JOHN M. CONNOR
included in this article.154 In these “public” cartels, the active involvement of governments is signaled by the signing of a treaty, government ownership of stocks or commodities, or the appointment of civil servants to cartel-management positions. There are many fine studies of such agreements, but the inclusion of government-sponsored or -enforced cartels would tend to bias upward the overcharges in the sample (Suslow, 2001). Moreover, public cartels are beyond the reach of antitrust law. With very few exceptions, this article reports on every scholarly or serious study that contained quantitative information on the price effects of hard-core private cartels. Writings by economists, political scientists, economic historians, and legal scholars are included. Written decisions or detailed reports of decisions of antitrust authorities everywhere in the world were examined. While no time limit was placed on the literature search, the large majority of the sources consulted were written after 1945.155 I have examined more than 1,000 English language books, journal articles, working papers, reports, and other shorter analyses looking for evidence of cartel price effects. Many were written primarily as historical case studies or are focused on demonstrating a new method. Some mention price effects only in passing. The great majority of the cartel studies were written by economists, typically by North American academics using cartel episodes that affected commerce in the United States or Canada. The small number of empirical studies by European or Asian academics is striking.156 In addition, countless hours were spent reading press releases, decisions, etc. at the Web sites of antitrust authorities.157 In general, I aimed at collecting the largest possible body of reasonably professional, quantitative estimates of cartel overcharges, and avoided applying possibly subjective quality screening. In the vast majority of cases, the writers themselves provided the overcharge calculations. In a small minority of cases, I made inferences from price data contained in the works, following the judgment of and the facts supplied by the author, such as dates of collusion. The bases for my inferences are briefly outlined in Connor (2007c).158 Overcharge claims appearing in newspapers, magazines, and newsletters are avoided because such assertions are usually from anonymous sources who may not be disinterested parties in an ongoing law suit or in some public policy debate, roles that may color their assertions. In some cases, overcharge estimates may originate from information in industry trade journals, but if they were cited by economists, historians, or legal scholars with some background in cartel studies, such estimates are reported in the present survey (e.g., Demaree, 1969). Estimates found in a small number of book-length, years-long investigations by journalists,
377
Cartel Overcharges
public servants, or other professional nonfiction writers are included (e.g., Berge, 1944; Taylor & Yokell, 1979). Clearly, this catholic approach to data gathering will create concerns in the minds of many readers about the reliability and precision of the overcharges. There may be substantial variation in the quality of the price data, the methods used, degrees of judicial scrutiny, and the professional orientation of the sources that could affect reliability as perceived by any individual. I noted above the lack of clarity among professional writers about the essential characteristics of the cartels until at least the 1920s. Consequently, some readers may wish to dismiss scholarship before that decade, while others will be untroubled by semantic differences. Economists may well give greater weight to writings by professionals in their own field than to opinions reached by judges, commissions, or juries, whereas legal scholars will often give greater credence to the latter. Legal professionals may have strong preferences for high court decisions over state or district courts, or they may have strong opinions about European versus American antitrust jurisprudence. Similarly, many economists might trust results published in refereed scientific journals more than other publication outlets that receive less peer scrutiny, prefer modern quantitative methods to deep historical case studies, or express skepticism about the analyses of economists writing before the Age of Game Theory. To contend with the disparate preferences of readers, I have chosen to cast my net widely, but look across the sources for evidence of systematic bias. In addition, the data are displayed across several time periods, data sources, and methods of computation so as to permit readers to choose the combinations they prefer. Indeed, the analysis of these data by source, time period, or method may provide useful insights in itself. I hope to provide the interested reader with enough information to make up his or her own mind about reliability.159
Social Science Studies The first block of sources consists of archived materials: books, monographs, reports, and refereed journal articles written by specialists in many fields: economists, historians, political scientists, lawyers, and in a few instances journalists.160 Newer publications were located by using various bibliographic search engines, by noting the references cited by authors in the works themselves, and by searching online library catalogs. These studies vary substantially in terms of depth and the degree of professional
378
JOHN M. CONNOR
commitment to the study of cartels. Some economists and historians have spent substantial portions of their careers specialized in cartel analysis, but most of the publications quoted herein are by social scientists for whom cartels were just a passing interest. There are several methods used by social scientists to derive the effects of cartels on prices. Older economic studies tended to use a rather informal method of price analysis that now comes under the rubric of the “beforeand-after method” (Connor, 2007c). That is, armed with knowledge of when overt collusion occurred, the author would compare prices during the affected period with prices before the cartel began or after it ended; in some cases, the basis of comparison would be a price war that erupted during the affected period. The base price was typically assumed to be the long-run competitive equilibrium benchmark price (now rather succinctly, if inelegantly, termed the “but-for price”). Although some were careful to take such factors into account, in many cases analysts ignored the possibility that shifts in demand or supply conditions could have caused the benchmark price during the affected period to depart systematically from the before or after price; moreover, the idea that price wars could generate unsustainably low prices was not often recognized. Some economists of the time realized the importance of averaging before or after prices for periods long enough to eliminate the influence of transitory disturbances in markets, but others were satisfied to identify one month’s prior price as the but-for price. A second way of calculating a benchmark price is the yardstick method. In this type of analysis, an economist would collect prices for analogous markets that were believed to be free from cartelization. For a localized conspiracy, the competitive yardstick could be prices in a nearby city or an adjacent state with similar demand or cost conditions. If prices before or after collusion are highly correlated, then the trend in cartel prices could then be compared to the trend in the yardstick-market prices during the collusive period. Yardstick price movements can also be constructed for a noncartelized product made in the same region that is made with the same inputs, utilizes a similar technology, and is consumed by the same customers.161 If a cartel colludes against only some of its customers, then the discounts offered to other similarly situated customers could yield a yardstick. Third, sometimes the costs of production and the margins earned by firms in the relevant lines of business may provide collateral indicators of variations in the degree of competitiveness of a firm or market. Absent significant changes in production technology, constant long-run marginal costs or constant operating margins may be assumed before, during, and
Cartel Overcharges
379
after alleged collusion; if they are not constant, collusion may be the cause. Cost-based estimates are relatively uncommon because detailed internal business records are needed. The before-and-after, yardstick, and constantmargin methods require expert judgment about the market or industry in question, but all are acceptable methods used in courts of law or commission hearings to determine the fact of injury or the amount of damages. Fourth, since the 1970s, the rigor and precision displayed in deriving estimates of cartel overcharges have made several advances (Baker & Rubinfeld, 1999). Driven by developments in oligopoly theory, statistical methods, and the increasing availability of detailed company and market data, increasingly it is econometric models of the alleged collusive market that are specified and fitted to the available data.162 Game theory has influenced contemporary concepts of collusion, the design of competition policies, and empirical modeling of oligopolies (Werden, 2004). One type of econometric modeling is an elaboration of the before-and-after method. A structural model of the market before or after the conspiracy can be estimated and used to predict the competitive benchmark price during the conspiracy (Brander & Ross, 2006, pp. 1720). A second type of econometric model can specify demand, supply, and an oligopoly model (usually Cournot or Bertrand) and fit the model to data from the collusive period (ibid., pp. 2123). An early example of this approach is Dick’s (1992a) study of 16 U.S. Webb-Pomerene cartels. The most common approach is a reduced-form model. These models usually specify the demand and supply conditions in the relevant market as a function of the observed market price before, during, and after a conspiracy; the analyst then investigates through statistical tests whether and to what extent changes in prices or output fail to respond to normal, competitive market forces (ibid., pp. 2329).163 Because these models can simultaneously incorporate multitudinous factors that explain prices, economists tend to regard overcharge estimates from such models as more accurate than analyses that depend on more informal ways of accounting for such factors.164 For this article, except for about 5% of the episodes where no method was given, I classified the remaining 1,461 episodes according to one of eight estimation methods used to derive the overcharge rates (Table A1). One of the most unusual methods (accounting for 2% of the episodic overcharges) is an historical examination of original cartel archives. This method of analysis resulted in by far the lowest mean overcharge of 10.8%; in fact, more than half of such estimates were zero. Echoing the findings of Connor and Bolotova (2006), estimates derived from a yardstick approach were the highest on average. Cost-based estimation (69 episodes) produced
380
JOHN M. CONNOR
the second-highest mean overcharges. Interestingly, the most popular method (639 episodes) – the three “straight-line” before-and-after methods – had lower-than-average mean values. But even lower were episodes derived from econometrics (289) and from legal decisions (245). Table A1.
Average Episodic Overcharge Estimates, by Method of Estimation Employed.
Method
Number of Estimates
Median Overcharge
Mean Overcharge
Mean Positive Overcharge
Percent Unavailable/none given Historical examination of cartel archives But-for price from before collusion But-for price from price war during collusion But-for price from after collusion Cost-based or constant-margin Yardstick from comparable unaffected market Econometric model prediction Legal decisiona Other quantitative (simulation, etc.)b Total
74 20
18.3 0
33.2 10.8
40.3 30.7
411
26.0
40.3
41.9
28
28.2
39.7
41.2
200
25.0
46.3
48.0
69 192
21.5 28.8
50.3 78.7
52.6 81.7
289 245 7
19.5 17.5 670
31.3 35.0 1277
33.6 35.5 1277
1535
23.0
48.7
51.8
a
No specific method mentioned by court, jury, or commission issuing decision. May be a monetary amount or a percentage. Also includes judgments about what amounts constitute adequate compensation or restitution for victims. See Table 14 for details. b Four of the seven observations (and the highest) are from Normann and Tan (2013). Sources: Connor (2014: Appendix Tables 1 and 2), which summarize Price Fixing Overcharges Master Data Set, spreadsheet dated October 2013.
The fact that some methods result in above- or below-average overcharges does have implications for accuracy, as each type of method may be associated with different mixes of cartel types, locations, or time periods. Econometric methods and legal decisions, for example, tend to be of a more recent vintage.
381
Cartel Overcharges
Episodes Like most natural phenomena, most cartels are born and die only once, and the dates of those events are known with precision. This describes one episode. The birth of a cartel (“formation”) is marked by the day a collusive agreement is adopted.165 Cartel deaths are more varied and sometimes more difficult for observers to pinpoint. Cartels can die “natural” deaths if changes in market conditions make collusion unsustainable; natural deaths may be quiet events marked by a consensus among the cartelists to close shop (e.g., if fringe entry becomes large, a new superior substitute product appears, or warfare among nations makes business as normal impossible), or they may end in convulsion (e.g., if cheating or defections become excessive or if major players engage in open warfare) (Levenstein & Suslow, 2010). Cartels may also die “unnatural” deaths if cheating or defections are caused by the presence of effective anti-cartel laws (e.g., an antitrust authority’s decision to investigate suspicious prices being charged to buyers, signals from a screening program that market prices are incredibly steady or significantly less variable, or an application by a participant in the cartel to a government leniency program). However, some cartels have led charmed lives, dying and being reborn. Indeed, some cartels are formed, disband, reform, and disband several times. Each collusive cycle is an episode. Overcharges are computed for episodes, not cartels, so the dates of those episodes are critical parameters for analysts. Consistent with most previous empirical studies of cartels, each cartel episode is treated as a unique observation.166 The reasons for analyzing episodes rather than one cartelized market over time are fairly straightforward. When a new episode appears, the cartel may have new members, a different territory, or simply a revised agreement. Pauses between episodes are often quite lengthy. Because the agreement or the players are different, in effect a new cartel is launched. Changes in these factors will generally affect price outcomes. The period between the termination of one episode and the rebirth of the next episode is known in economic game theory as a “reversion to competition.” During this interim, the cartelists cease to observe and enforce the contract, cease to have (harmonious) multilateral contacts, may engage in open warfare, and typically suffer lower prices and profits than previously. During reversion, prices may fall from near-monopoly levels to levels associated with noncooperative oligopoly (Cournot equilibrium, for example), purely competitive prices, or even subcompetitive
382
JOHN M. CONNOR
prices. Price wars are not necessarily signs of failure, rather, they may be opportunities for a cartel to reorganize and adopt better rules for price setting, profit sharing, compensation, and the like (Levenstein & Suslow, 2006). Sometimes, there are practical impediments to measuring episodic dates. The birth of a cartel (“formation”) is marked by the day the terms of a collusive agreement are adopted.167 In a forensic setting, the dates marking an episode may be obvious or uncontroversial; both sides stipulate the dates, and fines or damages can be computed with the stipulated dates. However, particularly for more durable cartels, the beginning date of an episode may be debatable, because written records have been lost or destroyed, cartel managers have retired or moved on, or memories faded.168 Ending dates should be better documented because in modern times most cartels end with publicly reported raids. However, cartel deaths are quite varied and some are difficult for observers to pinpoint. Cartels can die “natural” deaths if fringe supply grows too large, if cheating becomes excessive, or if defections (including leniency applications) occur; or cartels can die sudden “antitrust” deaths from raids resulting from tips169 to an antitrust authority (Levenstein & Suslow, 2010, Table 2). Of these causes of death, only raids can be objectively recorded. Moreover, in the case of global cartels, various antitrust authorities often cite different dates. To assist forensic economists in objectively identifying the existence and dates of collusion, a quantitative technique called a variance screen has been developed and implemented. Statistical analysis of price distributions begins with determining when the mean average price deviates from the but-for price. The mean is the first moment of the distribution, and there are three higher moments: variance, skewness, and kurtosis. Connor (1985) was probably the first to suggest the rationale for the notion that higher moments could be used to identify cartel price effects. Abrantes-Metz, Froeb, Geweke, and Taylor (2006) found that price variance declined during collusion by frozen fish sellers. Connor, Bolotova, and Miller (2008) also successfully tested the variance as a screen for cartel behavior. Blanckenburg, Geist, and Khodlodilin (2012) test for the effects of cartelization on all four moments of price distribution. Abrantes-Metz, Villas-Boas, and Judge (2011) applied Benford’s Law to demonstrate how LIBOR rates differed from the expected noncollusive distribution of digits, suggesting that bid rigging could have been detected.
383
Cartel Overcharges
Decisions of Antitrust Authorities The second big block of information includes the printed reports and Web pages of scores of antitrust agencies, lists of court and commission decisions, and multilateral organizations concerned with competition issues. Data collection began by trying to collect verdicts in collusion cases, namely, final decisions antitrust cases involving horizontal collusion, broadly defined to include bid rigging and related practices, where a judge, jury, or commission calculated the damages. Starting with the United States, in theory, researchers should be easily able to determine how high cartels raise prices by a straightforward examination of a statistically significant sample of the thousands of U.S. antitrust cases that involved cartels. However, for many decades in U.S. government cases, resolution of these numerous cases has involved fewer than 10 trials per year, most of them of individuals, not corporations. Moreover, the amount that prices changed, or even whether prices were affected at all, is not relevant to the issue of whether a defendant violated U.S. criminal antitrust law.170 In U.S. criminal antitrust cases, it is unnecessary for prosecutors to present evidence of the extent of any overcharges or undercharges. Even at the sentencing phase of criminal price-fixing trials, prosecutors rarely offer information on damages. Guilty-plea statements and sentencing memoranda often mention affected sales and culpability factors that were used to calculate the sentencing guidelines ranges. Only a few contain stipulated damages as percentages of affected sales, and these percentages are probably minimal overcharge rates.171 In civil damages cases, however, the damages awarded to a successful plaintiff are equal to three times the overcharges, so in these cases, plaintiffs must demonstrate how much prices increased or decreased due to the actions of the cartel. Finding overcharge rates in judicial decisions in civil actions also proved to be extremely difficult, because almost every private antitrust suit for damages settles or is dismissed before an overcharge can be calculated by a neutral observer and made part of the public record of the case. As a consequence, final verdicts involving cartels where a judge or jury calculated an overcharge are surprisingly rare. This approach yielded less than 30 episodic overcharges (Connor & Lande, 2005). Besides U.S. court decisions, the Web sites of many foreign antitrust authorities were examined.172 In the jurisdictions employing Common
384
JOHN M. CONNOR
Law, most cartels are sanctioned after government negotiations that result in guilty pleas or by monetary settlements with private parties out of court. When this is the method of resolution, the press releases practically never mention the degree of harm caused by the cartel. Very few cartels defend themselves in court, and very few of the trials result in published decisions that reveal the overcharges. Although judicial decisions themselves may not mention an overcharge rate, there are other ways to obtain overcharges from some of the decisions. Three sources were explored: computer-assisted searches of data bases, reading through a large number of articles and treatises on cartels and on antitrust damages, and messages to groups of knowledgeable antitrust professionals. For example, inquiries were made on the antitrust list serves of the ABA Antitrust Section, the National Association of Attorneys’ General, and of the American Antitrust Institute. Every qualifying final collusion verdict is included.173 The small sample size of overcharges from U.S. decisions is disappointing. In other legal systems, antitrust commissions hold confidential hearings to determine guilt and impose sanctions. These decisions are announced in press releases that seldom mention the extent of cartel damages. Italy, the Netherlands, and Korea are exceptions to this rule; these overcharges are collected in Connor (2003). Moreover, these antitrust authorities and some others have reported a few of their decisions and overcharge estimates to the OECD (2003). However, in some jurisdictions, a detailed report is released a year or two after the decision, and some of these reports have prices that can yield useful overcharge information, though that is not often the case.174 Additionally, commission decisions can be appealed to a court that renders a decision with a recitation of the facts of the case.175
Laboratory Market Experiments The overcharges reported in this article are derived from studies that use a wide variety of analytical methods (see Connor, 2007c). Many of these methods are not controlled scientific experiments in the strictest sense. Some come from econometric studies, which are quasi-experimental results derived from observations taken from natural markets. Since at least 1948, economists have been reporting on prices generated by controlled, small-scale laboratory market experiments. The supply sides of these games are oligopolies, and the treatments consist of changes in the number of players, supply conditions, available information, trading rules,
Cartel Overcharges
385
and seller communication protocols. Goods are almost always homogeneous and bought by anonymous buyers. All laboratory experiments allow the players to “communicate” tacitly through observed transaction prices or quantities, but a smaller number permits sellers or buyers to talk. Only the latter type opens up the possibility of cartel behavior. A classic survey of laboratory experiments with homogeneous-product monopoly and oligopoly can be found in Plott (1989, pp. 11421159). The predictions of pure monopoly theory are verified by these controlled experiments. One laboratory experiment finds that “[W]hen the monopolists post prices, market behavior is … accurately captured by monopoly theory” (ibid., p. 1144). That is, buyers end up paying the monopoly price.176 More apropos of this survey are oligopoly experiments that simulate cartels. A central conclusion of oligopoly experiments is that “market participants almost always recognize a harmony of interests” and that where direct communication is not permitted, observation of bids, offers, or transaction prices is one way that tacit agreements are realized (ibid., p. 1149). In other oligopoly experiments that allowed traders to talk among themselves (but prohibited profit sharing or side payments), traders “discussed conspiracy almost immediately and they had no difficulty in articulating an agreement” (ibid., p. 1150). When the few sellers post prices and have full information about each other (i.e., perfect monitoring within the collusive group), prices are supracompetitive (ibid., p. 1154). Long periods of interaction also facilitate collusion. Similarly, bid riggers who post their offers are able to reach infracompetitive prices (ibid., p. 1157).177 Perhaps because profit sharing, side payments, and punishment for cheaters are not allowed (all common features of cartels), collusion experiments result in prices below the full monopoly (above the monopsony) price. More recent experiments reinforce the importance of information and transparency among sellers in a cartel in achieving pricing effectiveness (Haan, Schoonbeek, & Winkel, 2009). Under tacit collusion, information about other sellers cannot be shared and sellers cannot talk to each other before or during trading. Experimental markets with tacit collusion generally result in competitive prices, except for homogeneousproduct duopolies, which usually achieve Cournot-level prices. First, access to rival sellers’ information can be collusive. If private information about current sellers’ costs or market shares is made available voluntarily, through a trade association for example, experiments with repeated games produce collusive effects. Revealing information about all past outputs and profits of rival sellers usually increases collusion.
386
JOHN M. CONNOR
Second, with posted pricing, explicit collusion among sellers who can easily detect cheating typically results in near-monopoly prices. Third, if sellers explicitly collude on list prices but buyers can also communicate and ask for secret discounts, transaction prices are still well above competitive levels; collusion ends only if sellers compete on both list and transaction prices. A meta-analysis of 154 oligopoly publications reports on the results of 512 controlled experiments that focused on the collusive price effects of sellers’ oligopolies under various treatments (Engel, 2007). These oligopoly price effects are reported using a measure of pricing efficiency that I will call the monopoly index (MI).178 The monopoly index divides the observed equilibrium overcharge by the maximum possible (monopoly) overcharge, expressed as a percentage.179 Without specifying the type of collusion, the efficiency of collusion increases with the fewness of sellers; MI is highest for duopoly experiments (MI = 62%), lower for triopolies (MI = 43%), and lower still for quadropolies (MI = 14.6%).180 (ibid., pp. 504506); similarly, the use of posted prices intensifies collusion relative to other pricing systems (ibid., pp. 537538). There are three experimental designs that shed light on collusion with overt communication. First, some experiments permit communication among sellers before trading begins (the classic Prisoners’ dilemma with “cheap talk”), and this may permit either misleading posturing or a degree of trust to develop among sellers. Collusion with prior communication seems to instill trust, because it generally results in more efficient collusion than when communication is prevented (ibid., pp. 521525). The efficiency of price collusion when communication occurs depends strongly on certain interactive factors. When the choice variable is quantity (i.e., a Cournot game in which price is an outcome, not a choice variable), cartels achieve higher pricing power (MI = 74%) than do tacitly colluding sellers (MI = 47%); experienced sellers that are allowed to talk (possibly a proxy for trust) achieve much higher pricing efficiency (MI = 73%) than inexperienced participants (MI = 24%); and price effects are stronger when sellers have good ex ante information (MI = 6164%). Second, a necessary feature of cartels is that sellers can conclude an enforceable agreement. In laboratory experiments, the availability of an enforceable agreement significantly increases price effects under certain conditions: when concentration is high (with duopoly MI = 87%), when buyers are anonymous (MI = 84%),181 when the game is Cournot (MI = 79%), when sellers are symmetric (MI = 93%), and when their capacity is unconstrained (ibid., pp. 523528). Thus, when an enforceable
Cartel Overcharges
387
agreement is concluded, high seller concentration, seller symmetry, low buyer concentration, homogeneous products, or excess capacity resulted in MIs above 70%. Third, one experiment shows the profound price effects that result when sellers can communicate after bidding begins (Fonseca and Normann, 2011). The MI with a duopoly averages 94%, and it declines when the number of sellers increases from 2 to 4 (MI = 81%), to 6 (MI = 65%), and to 8 (MI = 55%). However, compared to tacit collusion, explicit agreements result in smaller price gains under duopoly and when N = 8 than when N is 4 or 6 (ibid., p. 11). Similarly, when sellers expect a fine that is high (half of the monopoly gains), the sellers choose to cartelize more than half of the time when N = 28, whereas in a duopoly tacit collusion is chosen two-thirds of the time (ibid., p. 12). This study is unique in studying the content and purposes of messages sent between sellers; the authors conclude that explicit communications help raise prices by implementing more sophisticated pricing strategies, assisting in dispute mediation (e.g., after, a defection is observed), and, if permitted before trading begins (but not after), instilling greater trust that improves pricing effectiveness (ibid., pp. 2930). In summary, laboratory market experiments are a promising way to study cartel price effects using the utmost scientific rigor. Unfortunately, none yet incorporate most or all of the salient characteristics of real-world cartels. In particular, only a small minority of market laboratory experiments permits overt communication among suppliers. Nevertheless, when limited pre-play communication or during-play agreements are permitted, monopolistic pricing conduct is observed. Collusive prices on homogeneous goods approach monopoly levels when buyers are many and sellers are few, symmetric, experienced, have excess capacity, post their prices, and chose output as the strategic variable. While it is tempting to include the price results of market experiments, the sample reported herein covers only prices from natural markets.