Over the last decade, US courts have been on a steady march toward requiring increased economic rigor in demonstrating that impact and damages from alleged anticompetitive conduct can be established with information and methods common to a proposed class. The US Supreme Court’s decision in Comcast Corp. v. Behrend continued this march by clarifying that plaintiffs’ economic experts are required to go beyond assertions and show that their proposed econometric analyses would produce persuasive evidence at the class certification stage that is consistent with the plaintiffs’ theories of liability.1 The requirement to “show—don’t tell” applies equally to defendants’ experts. In decisions post-Comcast, analyses conducted by defendants’ experts have been most persuasive to courts when grounded in rigorous econometrics, as opposed to merely pointing out weaknesses in the plaintiffs’ theory or model. The increased focus on econometric analyses has necessarily led to a “battle of the experts” at the class certification stage of a case. Econometric models put forth by either side are subject to vigorous debate about statistical significance, functional forms, structural breaks, and appropriate estimation procedures. This trend has increased the analytical work that economists, parties, and the court must perform at the class certification stage. The challenge for courts is to determine whether the predictions of an econometric model persuasively show classwide impact or instead reveal a spurious result that is the statistical equivalent of random chance. While economists have a number of techniques to test if the results of a model are economically valid, not all of those tools are equally accessible or persuasive to a court. By Christine Siegwarth Meyer, Lauren J. Stiroh, and Claire (Chunying) Xie Antitrust, Vol. 30, No. 2, Spring 2016. © 2016 by the American Bar Association. Reproduced with permission. All rights reserved. This information or any portion thereof may not be copied or disseminated in any form or by any means or stored in an electronic database or retrieval system without the express written consent of the American Bar Association. www.nera.com 2 In the discussion below, we examine a selection of court decisions on class certification since Comcast with a view to evaluating whether there are certain themes for when a court appears to give weight to plaintiffs’ arguments or is persuaded by defendants’ counterarguments. We focus on a particular subset of specification testing that examines the propensity for the plaintiffs’ model, unaltered,2 to yield faulty predictions, which we believe is particularly effective in persuading a court that an econometric model can be capable of establishing classwide impact or, alternatively, is irredeemably flawed and cannot support a theory of classwide impact or harm. A faulty prediction can arise when an econometric model inappropriately finds impact or overcharges in populations where the plaintiffs’ theory of harm would predict that no impact should exist, or finds no effect where the allegations indicate one should be present. Our examination of the decisions post-Comcast suggests that courts may be most persuaded that a model is unreliable for showing impact and damages at trial if the defendant can empirically demonstrate that the model as constructed yields faulty predictions. Similarly, showing that the predictions of a model hold for different recognizable subclasses or class periods may help plaintiffs persuade the court that the predictions of their model are sufficiently reliable to warrant class certification. Typical Economic Analysis Used by Plaintiffs’ Experts at the Class Certification Stage To establish classwide damages using a common method, economists working on behalf of plaintiffs typically construct an overcharge model that compares the allegedly harmed class members during the conspiracy period to a benchmark, which is claimed to be unaffected by the alleged behavior. Examples of benchmarks include a time period when the alleged behavior is not in effect, or a reference group of customers that may have purchased products from the defendants during the class period but are expected to be unaffected by that behavior under the plaintiffs’ theory.3 Overcharge models typically consist of a regression analysis explaining prices that include an indicator variable for the class members during a conspiracy period. In an appropriately specified regression model, the estimated coefficient for the indicator variable measures the average impact of the alleged anticompetitive conduct on the class. If the regression results suggest that the estimate for the indicator variable is statistically significant and has a sign that is consistent with plaintiff’s theory of harm, then plaintiffs’ experts typically conclude that the class members were harmed by the allegedly anticompetitive behavior and that the average effect of the harm is quantified by the estimated coefficient.4 However, the predictions of the overcharge model are faulty if the same indicator variable in the same model is found to be significant for groups that, under the plaintiffs’ theory, were unharmed by the alleged behavior or insignificant for groups that allegedly suffered harm. Faulty predictions can arise for various reasons, including misspecified regression models. Model misspecification arises from incorrect functional forms, omission of relevant market factors, or issues in the data, among others.5 www.nera.com 3 An empirical demonstration of faulty predictions, using the plaintiffs’ own model, is one way for defendants to demonstrate clearly and persuasively the weaknesses of a plaintiffs’ analysis. Conversely, the court may give limited weight to criticisms of the plaintiffs’ model without an empirical showing of faulty predictions. On the other side of the v., plaintiffs may be able to establish the robustness of their model by showing that the results hold for a variety of subclasses or subperiods or by showing that variations in the predicted outcomes of their model for different groupings of class members are consistent with the plaintiffs’ theory of liability and classwide harm. Defendants’ Use of Faulty Prediction Analyses at the Class Certification Stage Defendants in recent cases have used faulty prediction analysis to try to persuade the court of the weaknesses of the plaintiffs’ analysis. In some cases, the court rejected such an analysis and found the plaintiffs’ analysis sufficient for the purpose of class certification; in others, the defendants’ analysis appeared to resonate with the court. These decisions highlight important lessons about effective economic evidence on classwide impact and damages. Courts Persuaded by Faulty Prediction Analysis In re Rail Freight Fuel Surcharge Antitrust Litigation In Rail Freight, 6 the plaintiffs alleged that four rail freight carriers conspired to increase freight prices through fuel surcharge programs and sought to certify a class of direct purchasers during a class period from 2003 to 2008. The plaintiffs’ expert proposed two regression models: a common factor model and a damages model. The discussion here focuses on the latter as the defendants’ argued that it could generate faulty predictions. To demonstrate common impact, the plaintiffs’ expert controlled for factors purportedly common to prices of all rail shipments and reported to have found a structural break in the relationship between fuel prices and freight rates at the start of the class period. The plaintiffs’ expert argued that the damages model was a plausible method for establishing classwide injuries. The defendants contended that the plaintiffs’ damages model suffered several infirmities, and in its decision regarding class certification, the appellate court focused on one of them— namely the “false positives” argument. Specifically, the defendants found that the plaintiffs’ model yielded positive damages when it was applied to customers whose rates were set by contracts signed before the commencement of the alleged anticompetitive behavior (i.e., the legacy shippers) and thus, under the plaintiffs’ theory, should not have been affected by the alleged conspiracy under the plaintiffs’ theory. Because the plaintiffs’ damages model found anticompetitive effects when, under the plaintiffs’ allegations, none should exist, the defendants argued that the model generated false positives for the legacy shippers. www.nera.com 4 Based on the plaintiffs’ model’s propensity towards producing false positives, the appellate court vacated the district court’s class certification decision, finding that the court could not assess the validity of the conclusions from the plaintiffs’ damages model in light of this failing.7 In re High-Tech Employee Antitrust Litigation In High-Tech, 8 the plaintiffs alleged that seven Silicon Valley high-tech companies conspired to limit the use of “cold calling” as a recruitment method and thereby restricted employee mobility and suppressed employee compensation. The plaintiffs sought certification of two classes: one that included all types of the defendants’ employees during the class period and another that was confined to the defendants’ employees who “work[ed] in the technical, creative and/or research and development fields.”9 The plaintiffs’ expert proposed three econometric and statistical analyses, and this discussion focuses on the expert’s compensation movement charts. Specifically, the plaintiffs’ expert plotted the salaries for the top ten positions at two of the seven defendant firms, and found co-movements in the salaries for those positions within each firm. Because only a subset of the classes would have received the cold calls absent the alleged conspiracy, in seeking to establish classwide harm, the plaintiffs’ expert used the co-movements found in his compensation movement charts to support the argument that the wages of class members tended to move together over time and in response to common factors. The defendants’ expert expanded the plaintiffs’ plots to the top 25 positions at each of the seven defendant firms and found that, in any given year, compensation for those positions moved in different directions, which contradicted the plaintiffs’ argument that wages tended to move together over time. The court did not give weight to the plaintiffs’ compensation movement charts in supporting their wage rigidity argument. By expanding the plaintiffs’ compensation movement charts to more positions and all of the defendants, the defendants showed that the plaintiffs’ co-movement analysis generated faulty predictions and that salaries did not necessarily move together over time at the defendant firms. The court ruled that it could not rely on the plaintiffs’ co-movement analysis in reaching its decisions on class certification. On the other hand, as we discuss in a later section, the court was not persuaded by the defendants’ rebuttal to one of the other analyses used by the plaintiffs’ expert to establish common impact. In the decision that considered the compensation movement charts and the defendants’ rebuttal to those charts, the court denied class certification but allowed the plaintiffs leave to amend. Food Lion, LLC v. Dean Foods Co. In Food Lion, 10 the plaintiffs—a group of retail sellers of processed milk—alleged that the milk wholesalers increased the prices of milk though non-compete agreements, and sought to certify a class of purchasers, excluding schools and school districts, of fresh white fluid milk sold by the defendants. To estimate damages, the plaintiffs’ expert proposed a regression model that examined the relationship between the price of milk and demand- and supply-side factors that would explain fluctuations in milk prices. Using the estimated model, the plaintiffs’ expert estimated a positive overcharge on the price of processed milk as a result of the alleged conspiracy. www.nera.com 5 The defendants contended that the plaintiffs’ model suffered several infirmities, including that the plaintiffs’ damages model yielded false positives. Specifically, the defendants argued that a significant portion of the processed milk at issue was under oral “cost-plus” pricing agreements, and two of the three components that determined those “cost-plus” contracts were beyond the effects of the alleged conspiracy. The defendants argued that the plaintiffs’ damages model was unreliable because it estimated positive damages for those contracts that were unrelated to the alleged conspiracy. The court was unconvinced by the plaintiffs’ arguments that those “cost-plus” pricing agreements were affected by the conspiracy, and the court agreed with the defendants that the plaintiffs’ damages model was unreliable because of its propensity to yield false positives. Defendants’ Critiques of Plaintiffs’ Models that Courts Did Not Find Persuasive IL Fornaio (America) Corp. v. Lazzari Fuel Co. In IL Fornaio, 11 the plaintiffs alleged that a group of mesquite lump charcoal distributors allocated customers in order to charge prices above the competitive level. The plaintiffs sought to certify a class of restaurants and other food service providers that directly purchased mesquite lump charcoal from the defendants. To determine classwide overcharges, the plaintiffs’ expert assumed a single geographic market for mesquite lump charcoal in the United States and proposed to estimate damages by comparing the actual prices to the but-for prices simulated in a market absent the alleged conspiracy. The defendants argued that the plaintiffs’ analysis erroneously assumed a single market for mesquite lump charcoal in the United States and argued that deliveries to end users and deliveries to wholesalers, as well as deliveries to various geographic locations, should have been considered as separate markets. The defendants also argued that they only competed in deliveries to wholesalers, not deliveries to end users. However, the defendants did not demonstrate empirically that the plaintiffs’ model could not be used to identify common impact under the alternative market definition or whether it would yield faulty predictions. The court opined that the defendants’ criticism of the plaintiffs’ market definition was insufficient to disprove the plaintiffs’ overcharge model for the purpose of class certification. Ultimately, the court ruled that predominance was met. In re Nexium (Esomeprazole) Antitrust Litigation In Nexium, 12 the end-payor plaintiffs alleged that AstraZeneca and several generic pharmaceutical companies entered into reverse payment settlements that allegedly delayed generic entry and resulted in end payers paying a higher price for the pharmaceutical drug, Nexium. The end-payor plaintiffs sought to certify a class that included consumers, third-party payors, and union plan sponsors that purchased or provided reimbursements for Nexium. www.nera.com 6 Because generic entry was delayed by agreement until May 2014, actual price and sales data did not exist for Nexium’s generic counterpart at the time that the end-payor plaintiffs moved for class certification. To calculate the overcharges of the alleged conspiracy, the plaintiffs’ expert proposed to use a “yardstick” approach that compared the price of Nexium to the average post-generic entry price of drugs in the same therapeutic category. The defendants argued that, because the plaintiffs’ econometric model measured the average but-for prices for Nexium, the model erroneously estimated positive damages for certain class members who likely did not suffer any harm. Such class members included, for example, consumers who would have purchased the branded drug after the generic entry, insured consumers who would have the same co-pays for the branded and the generic drugs, and participants of AstraZeneca’s coupon program. To rebut this critique, the plaintiffs’ expert showed quantitatively that the incidence of uninjured class members was small in terms of total sales of Nexium. The court was satisfied with the plaintiffs’ approach at least in part because it found that the defendants did not reliably quantify the prevalence of third-party payors and consumers who—through rebates, contracts, and brand-loyalty purchasing—suffered no antitrust injury. The court opined that, without knowing the prevalence of uninjured members, the defendants’ concern that the plaintiffs’ approach erroneously predicted positive damages for uninjured members was insufficient to undermine the plaintiffs’ showing of common impact. In re High-Tech Employee Antitrust Litigation In High-Tech, 13 the plaintiffs accused seven Silicon Valley high-tech companies of restricting employee mobility and suppressing employee compensation through anti-solicitation agreements. As discussed above, the plaintiffs’ expert proposed three econometric and statistical analyses for the purpose of class certification, one of which was a regression to estimate the alleged under-compensation to the putative classes. Specifically, the plaintiff’s expert proposed to regress wages on indicator variables to capture the effect of the antisolicitation agreements, while controlling for common employee characteristics and defendant-specific factors that may also explain variations in wages. The plaintiffs argued that the model could be used to estimate the average or net under-compensation at each firm caused by the alleged conspiracy. The defendants’ expert contended that the plaintiffs’ model was misspecified, and that, once corrected, the model would have generated different results. Specifically, the defendants’ expert added additional defendant-specific controls and alternative defendant-specific performance variables to the plaintiffs’ model and found that the plaintiffs’ results did not hold after those alterations. However, the court did not give weight to the defendants’ argument that the plaintiffs’ model was misspecified or that an appropriately specified model would have led to different conclusions. Specifically, the court opined that 28 of the 42 additional defendant-specific controls that the defendants’ expert added to the plaintiffs’ model were related to the alleged conspiracy and that the inclusion of those variables could have diluted the observed impact of the alleged conspiracy, artificially lowering the estimated coefficient of the indicator variable. www.nera.com 7 In re Processed Egg Products Antitrust Litigation: All Direct Purchaser Actions In Processed Egg Products-All Direct Purchaser Actions, 14 the plaintiffs alleged that the major egg producers in the United States engaged in conspiracies to limit the supply of eggs by limiting flock size and thereby increased the prices of shell eggs and other egg products. The plaintiffs sought to certify a class of direct purchasers of shell eggs and other egg products. To estimate the effect of the conspiracy on the flock size of egg layers, the plaintiffs’ expert proposed using a regression analysis that compared the flock size before the commencement of the alleged conspiracy to the flock size during the conspiracy period. The plaintiffs’ expert found that the but-for flock size would have been higher than the actual flock size during the conspiracy period. The defendants argued that the plaintiffs’ model measured damages for some purchasers who were under cost-plus contracts and whose prices were set based on production costs and thus were not affected by fluctuations in the supply of eggs. However, the defendants did not run the plaintiffs’ regression models using the data from those cost-plus contracts. To rebut the defendants’ critique, the plaintiffs argued that those cost-based contracts were set during the conspiracy periods and that documentary evidence suggested that the prices under those contracts were linked to the spot prices of egg products, which were affected by the conspiracy under the plaintiffs’ theory. The court agreed with the plaintiffs and concluded that the cost-plus contracts could still be affected by the alleged conspiracy.15 Lessons Learned A critique of the assumptions underlying the plaintiffs’ model appears to be most persuasive to a court considering whether to certify a class when the critique is supported by empirical analysis of the faulty predictions that the misspecified model generates. In the extreme case, the model might find impact when, in fact, there was none. Failing to demonstrate the quantitative effect of a critique involving faulty predictions may cause the court to give it less weight. Demonstrating that a model with different regression specifications—which inherently changes the assumptions of the plaintiffs’ econometric model—yields different predictions appears to carry less weight in persuading a court that the plaintiffs’ model is too deeply flawed to support class certification. For example, in IL Fornaio, the court did not give weight to the defendants’ assertions of distinct geographic markets with differences in marketplace dynamics that were not supported by empirical evidence, and, as such, were not sufficient to prevent certification of the class. In Nexium, the court found that the defendants failed to establish that the concern about uninjured class members was sufficiently extensive to warrant not certifying the class.16 On the other hand, in Rail Freight and High-Tech (regarding the plaintiffs’ compensation movement charts), the defendants were able to show quantitatively faulty predictions resulting from the plaintiffs’ analysis, and the courts found the defendants’ arguments persuasive in both cases. www.nera.com 8 Rebuttal with quantitative analyses appears to be most persuasive if the defendants’ expert uses the plaintiffs’ own model to demonstrate faulty predictions and if the faulty predictions analysis is grounded in case-specific facts and economic theory. In Rail Freight, the defendants applied the plaintiffs’ econometric model to a group of consumers who, according to the plaintiffs’ theory of the case, should have been unaffected by the alleged conspiracy, and found impact on those consumers. This analysis brought the reliability of the plaintiffs’ model into question, as the court opined that there was “no way of knowing the overcharges the damages model calculated for class members was any more accurate than the obviously false estimates” that the model produced.17 In High-Tech, the defendants expanded the plaintiffs’ method to plot the wages of various job titles over time and found model predictions that contradicted the plaintiffs’ theory. The court found this analysis to be important in its decision not to certify the broad classes that the plaintiffs originally proposed. On the other hand, it appears to be less effective when the defendants use an alteration of the plaintiffs’ model to show that the model is misspecified and generates faulty predictions, especially if case-specific facts and economic theory do not offer a stronger justification for the defendants’ alterations than for the plaintiffs’ model. For example, in High-Tech (regarding the plaintiffs’ regression analysis to calculate under-compensation), the defendants altered the plaintiffs’ regression analysis to demonstrate the sensitivity of the plaintiffs’ results to inclusions of additional explanatory variables. However, the court was not persuaded by the defendants’ argument because it found that the defendants’ alteration suffered several model specification issues that could lead to erroneous results. Empirically Testing Faulty Predictions As discussed above, the court gives more weight to the analysis that empirically shows that the unaltered plaintiffs’ model yields faulty predictions. Thus, it may be useful for the plaintiffs and the defendants to empirically test whether an econometric model is prone to faulty predictions. We outline two key elements in conducting such tests: identifying a group that is uninjured under the plaintiffs’ theory and correctly interpreting regression results. Identifying an Uninjured Group Under the Plaintiffs’ Theory As a first step in finding faulty predictions, it is often necessary to identify a group that suffered or did not suffer harm under the plaintiff’s theory using case-specific facts. For example, in Rail Freight, the defendants argued that legacy shippers were unharmed by the conspiracy because their prices were set in contracts that were negotiated before the alleged conspiracy began. In Food Lion, the defendants argued that the oral “cost-plus” pricing agreements were established by government agencies and were unrelated to the alleged conspiracy. Because identifying a group of uninjured consumers prior to the quantitative analyses relies on case-specific facts and the plaintiffs’ theory of harm, this approach avoids identifying faulty predictions that may not be explained by case-specific facts or economic theories. Another advantage of this approach is that it does not necessarily require alterations of the plaintiffs’ model and thus any potential faulty predictions that may arise from the analyses are critiques of the model on its own terms. www.nera.com 9 Using case facts in seeking to identify a group of injured or uninjured class members under the plaintiffs’ theory is crucial in false positive analysis. For example, in Processed Egg Products, although the defendants argued that customers with cost-plus contracts should not have been injured by the alleged reduction in the supply of eggs, the plaintiffs were able to show through documentary evidence that those customers may still have been harmed under their theory of how the alleged conspiracy affected spot prices of eggs. The Necessity of Correctly Interpreting Regression Results Besides using case-specific facts to identify an uninjured group under the plaintiffs’ theory, analyses involving faulty prediction also require correct interpretation of regression results— specifically that an indicator variable in typical overcharge models estimates the average effect of an alleged conspiracy. Because, in some cases, the actual transaction price varies across class members, comparing an average but-for price derived from the overcharge model to an actual transaction price may lead to an erroneous conclusion of faulty predictions. For example, in Electronic Books, 18 the plaintiffs accused five book publishers and Apple of engaging in price fixing that allegedly resulted in an increase in prices of e-books. The plaintiffs sought certification of a class that included purchasers of e-books published by the publisher defendants during the class period (April 1, 2010 to May 21, 2012). To calculate damages, the plaintiffs’ expert proposed to conduct regression analyses using actual transaction prices, controlling for common factors that might influence an e-book’s price, to obtain an estimate of the overcharges resulting from the alleged conspiracy. The defendants’ expert argued that the plaintiffs’ model yielded faulty predictions because the average but-for prices predicted by the plaintiffs’ model were higher than actual prices for a significant number of transactions. The plaintiffs’ expert contended that the defendants’ faulty prediction analysis was erroneous because, in comparing average but-for prices to actual transaction prices, the defendants considered transactions with significant discounts as faulty predictions even though those transactions may still have been affected by the alleged conspiracy. The plaintiffs’ expert used an example to show that the but-for prices for transactions with significant discounts could be lower than the average but-for prices for all transactions on that book and that the defendants could have erroneously classified such transactions as faulty predictions by erroneously comparing the actual price to the average but-for price of all transactions, instead of comparing it to the but-for price for that transaction. In evaluating the defendants’ Daubert motions, the court found the plaintiffs’ arguments persuasive and concluded that the defendants’ faulty prediction concerns did not undermine the reliability of the plaintiffs’ model.19 This example shows that, when comparing an average but-for price derived from regression results to the entire distribution of actual prices, it is possible that the average but-for price is higher than some actual prices. As the plaintiffs’ expert demonstrated in this case, actual prices could be lower than the predicted average but-for price in the presence of discounts. However, it does not necessarily imply that the consumers who paid those actual prices were unharmed by a conspiracy under the plaintiffs’ theory. www.nera.com 10 Conclusion The examples discussed here focus on post-Comcast court rulings on challenges to plaintiffs’ economic evidence on common injury and damages based on faulty prediction analysis, i.e., when the defendants contend that the plaintiffs’ model inappropriately finds impact or overcharges in populations where none should exist, or finds no effect where the allegations indicate one should be present. These examples suggest that, when the defendants demonstrate that the unaltered plaintiffs’ model yields faulty predictions, the court is more likely to be persuaded that the plaintiffs’ model is flawed and cannot support a theory of classwide impact or harm. On the other hand, the court is more likely to find the plaintiffs’ model sufficient for the purpose of class certification when the defendants use an alteration of the plaintiffs’ model to show faulty predictions, in particular when case-specific facts and economic theory do not provide ample justification for such alterations. 1 “[A] model purporting to serve as evidence of damages in this class action must measure only those damages attributable to [the liability] theory” that was “accepted for class-action treatment by the District Court.” Comcast Corp. v. Behrend, 133 S. Ct. 1426, 1433 (2013). “[A]t the class-certification stage (as at trial), any model supporting a ‘plaintiff’s damages case must be consistent with its liability case, particularly with respect to the alleged anticompetitive effect of the violation.’” Id. (citation omitted). 2 We understand that in some cases, the parties do not necessarily agree on whether the plaintiffs’ model was “unaltered.” In this article, a model is unaltered when its functional form is not changed and when there is no addition or subtraction of explanatory variables. 3 ABA SECTION OF ANTITRUST LAW, ECONOMETRICS: LEGAL, PRACTICAL, AND TECHNICAL ISSUES 356 (Lawrence Wu ed., 2d ed. 2014). 4 Id. at 356–60. 5 Id. at 102–16. 6 In re Rail Freight Fuel Surcharge Antitrust Litig., 725 F. 3d 244 (D.C. Cir. 2013). 7 Id. at 254–55. 8 In re High-Tech Emp. Antitrust Litig., 289 F.R.D. 555 (N.D. Cal. 2013). 9 Id. at 562. 10 Food Lion, LLC v. Dean Foods Co., No. 2:07-CV-188, 2016 WL 310432 (E.D. Tenn. Jan. 25, 2016). 11 IL Fornaio (America) Corp. v. Lazzari Fuel Co., LLC, No. C 13-05197, 2014 U.S. Dist. LEXIS 25518 (N.D. Cal. Feb. 27, 2014). 12 In re Nexium (Esomeprazole) Antitrust Litig., 968 F. Supp. 2d 367 (D. Mass. 2013). 13 In re High-Tech Emp. Antitrust Litig., 289 F.R.D. 555 (N.D. Cal. 2013). 14 In re Processed Egg Prods. Antitrust Litig., 312 F.R.D. 171 (E.D. Pa. 2015). 15 The defendants raised a similar argument regarding cost-plus contracts against the class certification of indirect purchasers in this case. The plaintiffs rebutted that such cost-plus contracts were rare, but the court was unsatisfied with the plaintiffs’ argument, stating that “[p]laintiffs’ have not. . . advanced their arguments about costplus contracts beyond the theoretical level to the level of sufficient evidence to demonstrate common impact.” In re Processed Egg Products, 312 F.R.D. 124, 157 (E.D. Pa. 2015). 16 But see Vista Healthplan Inc. v. Cephalon Inc., No. 2: 06-cv-1833, 2015 U.S. Dist. LEXIS 74846 (E.D. Pa. June 10, 2015). In Vista Healthplan, the defendants argued that the putative class included a significant number of potentially uninjured members, one of which the defendants estimated to be 5% of the class. Even though the defendants’ argument was similar to the argument used in Nexium, the court in Vista was satisfied with the defendants’ argument. 17 In re Rail Freight, 725 F. 3d at 254. 18 In re Elec. Books Antitrust Litig., No. 11 MD 2293 (DLC), 2014 U.S. Dist. LEXIS 42537 (S.D.N.Y. Mar. 28, 2014). 19 Id. at *87–88. Notes About NERA NERA Economic Consulting (www.nera.com) is a global firm of experts dedicated to applying economic, finance, and quantitative principles to complex business and legal challenges. For over half a century, NERA’s economists have been creating strategies, studies, reports, expert testimony, and policy recommendations for government authorities and the world’s leading law firms and corporations. We bring academic rigor, objectivity, and real world industry experience to bear on issues arising from competition, regulation, public policy, strategy, finance, and litigation. NERA’s clients value our ability to apply and communicate state-of-the-art approaches clearly and convincingly, our commitment to deliver unbiased findings, and our reputation for quality and independence. Our clients rely on the integrity and skills of our unparalleled team of economists and other experts backed by the resources and reliability of one of the world’s largest economic consultancies. With its main office in New York City, NERA serves clients from more than 25 offices across North America, Europe, and Asia Pacific. Contact For further information and questions, please contact the author: Christine Siegwarth Meyer Senior Vice President +1 914 448 4119 email@example.com Lauren Stiroh Senior Vice President +1 914 448 4143 firstname.lastname@example.org The opinions expressed herein do not necessarily represent the views of NERA Economic Consulting or any other NERA consultant. Please do not cite without explicit permission from the author.