Behrend v. Comcast Corp.

PRECEDENTIAL UNITED STATES COURT OF APPEALS FOR THE THIRD CIRCUIT __________ No. 10-2865 __________ CAROLINE BEHREND; STANFORD GLABERSON; JOAN EVANCHUK-KIND; ERIC BRISLAWN v. COMCAST CORPORATION; COMCAST HOLDINGS CORPORATION; COMCAST CABLE COMMUNICATIONS, INC.; COMAST CABLECOMMUNICATIONS HOLDINGS, INC.; COMCAST CABLE HOLDINGS, LLC, Appellants __________ On Appeal from the United States District Court for the Eastern District of Pennsylvania (D.C. No. 2-03-cv-06604) District Judge: Honorable John R. Padova __________ Argued on January 11, 2011 Before: FISHER, JORDAN and *ALDISERT, Circuit Judges. (Filed: August 23, 2011) Darryl J. May BALLARD SPAHR LLP 1735 Market Street, 51st Floor Philadelphia, PA 19103-75999 Michael S. Shuster [ARGUED] Sheron Korpus KASOWITZ, BENSON, TORRES & FRIEDMAN LLP 1633 Broadway, 21st Floor New York, NY 10019 Attorneys for Appellants Samuel D. Heins Vincent J. Esades David R. Woodward Jessica N. Servais HEINS MILLS & OLSON, P.L.C. 310 Clifton Avenue Minneapolis, Minnesota 55403 Anthony J. Bolognese * Subsequent to oral argument, Judge Aldisert replaced Judge Ambro on the panel. The case was not reargued because the replacement Judge exercised his right to decide the case on the basis of the brief, the record and a transcript of the original oral argument. 2 Joshua H. Grabar BOLOGNESE & ASSOCIATES, LLC Two Penn Center Plaza 1500 JFK Boulevard, Suite 320 Philadelphia, PA 19102 Barry C. Barnett [ARGUED] Daniel H. Charest Stephen L. Shackelford, Jr. SUSMAN GODFREY L.L.P. 901 Main Street, Suite 5100 Dallas, Texas 75202-3775 Joseph Goldberg FREEDMAN BOYD HOLLANDER GOLDBERG & IVES 20 First Plaza, Suite 700 Albuquerque, NM 87102 Attorneys for Appellees __________ OPINION OF THE COURT __________ ALDISERT, Circuit Judge. In 2008 this Court handed down the seminal case of In re Hydrogen Peroxide Antitrust Litigation, 552 F.3d 305 (3d Cir. 2008), which outlines the standards a district court should apply in deciding whether to certify a class. This appeal by Comcast requires us to decide if the District Court for the 3 Eastern District of Pennsylvania properly satisfied Hydrogen‘s directions in determining that questions of fact or law common to class members predominate sufficiently to satisfy Rule 23(b)(3) of the Federal Rules of Civil Procedure. Appellants contend that the District Court exceeded a proper exercise of discretion and that its findings of fact were clearly erroneous. For the reasons that follow, we hold that the Court did not exceed its permissible discretion in determining that Plaintiffs established by a preponderance of evidence that they would be able to prove through common evidence (1) class-wide antitrust impact (higher cost on non-basic cable programming), and (2) a common methodology to quantify damages on a class-wide basis. Accordingly, we will affirm. I. A. ―For the rational study of the law the black- letter man may be the man of the present, but the man of the future is the man of statistics and the master of economics.‖ Oliver Wendell Holmes, Jr., The Path of the Law, 10 Harv. L. Rev. 457, 469 (1897). Beginning in 1998, Defendants Comcast Corporation, Comcast Holdings Corporation, Comcast Cable Communications, Inc., Comcast Cable Communications Holdings, Inc., and Comcast Cable Holdings, LLC (collectively ―Comcast‖) engaged in a series of transactions that increased Comcast‘s share of the multichannel video programming distribution services offered in the Philadelphia 4 Designated Market Area (―Philadelphia DMA‖).1 Comcast contracted with competing cable providers to either acquire them or to ―swap‖ cable systems it owned in areas outside the Philadelphia DMA for cable systems within the Philadelphia DMA. These transactions form the ―Cable System Transactions,‖ involving the ―Transaction parties.‖2 As a 1 ―A DMA is a specific media research area that is used by Nielsen Media Research to identify television stations whose broadcast signals reach a specific area and attract the most viewers. DMA boundaries are widely accepted and used by all types of companies to target and keep track of advertising.‖ Steak n Shake Co. v. Burger King Corp., 323 F. Supp. 2d 983, 986 n.2 (E.D. Mo. 2004). 2 The District Court set forth the Cable System Transactions:  The April 1998 acquisition of Marcus Cable and its 27,000 cable subscribers located in Harrington, Delaware, which is part of the Philadelphia DMA.  The June 1999 acquisition of Greater Philadelphia Cablevision, Inc., a subsidiary of Greater Media, Inc., and its 79,000 cable subscribers located in Philadelphia.  The January 2000 acquisition of Lenfest Communications, Inc. and more than 1.1 million cable subscribers located in Berks, Bucks, Chester, Delaware, and Montgomery counties in Pennsylvania, and New Castle County in Delaware.  The January 2000 acquisition of Lenfest‘s ownership interests in Garden State Cablevision L.P. and its 212,000 customers located in Atlantic, Burlington, Camden, Cape May, Cumberland, Gloucester, Mercer, and Salem counties in New Jersey, which is part of the Philadelphia DMA. 5 result of the Cable System Transactions, Comcast‘s share of subscribers in the Philadelphia DMA allegedly increased from 23.9 percent in 1998 to 77.8 percent by 2002, settling at 69.5 percent in 2007. See Behrend v. Comcast Corp., 264  The December 2000 swap agreement with AT & T, wherein Comcast obtained cable systems and approximately 770,000 subscribers, including subscribers located in Eastern Pennsylvania (Berks and Bucks counties) and New Jersey. In exchange, AT & T obtained cable systems and approximately 700,000 Comcast subscribers located in Chicago and elsewhere around the country.  The January 2001 swap agreement with Adelphia Communications Corp., wherein Comcast obtained cable systems and approximately 464,000 subscribers located primarily in the Philadelphia area and adjacent New Jersey areas. In exchange, Adelphia received Comcast‘s cable systems and subscribers located in Palm Beach, Florida and Los Angeles, California.  The April 2001 swap agreement with AT & T, wherein Comcast obtained cable systems and approximately 595,000 subscribers, including subscribers located in Pennsylvania and New Jersey.  The August 2006 swap agreement with Time Warner in connection with the Adelphia bankruptcy, wherein Comcast obtained cable systems and approximately 41,000 subscribers in the Philadelphia DMA.  The August 2007 acquisition of Patriot Media and its 81,000 cable subscribers located in New Jersey, within the Philadelphia DMA. Behrend v. Comcast Corp., 264 F.R.D. 150, 156 n.8 (E.D. Pa. 2010). 6 F.R.D. 150, 160 (E.D. Pa. 2010) (setting forth Plaintiffs‘ expert‘s calculations as to Comcast‘s market share). Plaintiffs, six non-basic cable television programming services customers of Comcast, brought a class action antitrust suit against Comcast in 2003. They alleged violations of section 1 of the Sherman Act, 15 U.S.C. § 1, for ―imposing horizontal territory, market and customer allocations by conspiring with and entering into and implementing unlawful swap agreements, arrangements or devices,‖ and section 2 of the Sherman Act, 15 U.S.C. § 2, on theories of monopolization and attempted monopolization.3 App. 00232-243 (Third Am. Compl.). The Complaint alleged anticompetitive conduct in the Philadelphia area and the Chicago area. As only the alleged conduct in Philadelphia is before us, we focus on the nature of the class and the allegations in Philadelphia. The proposed class included: ―All cable television customers who subscribe or subscribed at any time since December 1, 1999, to the present to video programming services (other than solely to basic cable services) from Comcast, or any of its subsidiaries or affiliates in Comcast‘s 3 Section 1 of the Sherman Act states: ―Every contract, combination in the form of trust or otherwise, or conspiracy, in restraint of trade or commerce among the several States, or with foreign nations, is declared to be illegal.‖ 15 U.S.C. § 1. Section 2 states: ―Every person who shall monopolize, or attempt to monopolize, or combine or conspire with any other person or persons, to monopolize any part of the trade or commerce among the several States, or with foreign nations, shall be deemed guilty of a felony . . . .‖ Id. § 2. 7 Philadelphia cluster.‖ App. 00217; see id. (excluding from the class ―governmental entities, Defendants, Defendants‘ subsidiaries and affiliates and this Court‖). The Philadelphia cluster is composed ―of the areas covered by Comcast‘s cable franchises, or any of its subsidiaries or affiliates, located in the following counties: Berks, Bucks, Chester, Delaware, Montgomery and Philadelphia, Pennsylvania; Kent and New Castle, Delaware; and Atlantic, Burlington, Camden, Cape May, Cumberland, Gloucester, Mercer and Salem, New Jersey.‖ See Behrend, 264 F.R.D. at 191.4 The Complaint alleged that Comcast had perpetrated an anticompetitive ―clustering scheme.‖ To clarify its contentions we pause to define two key terms. ―Clustering‖ refers to a ―strategy whereby cable [Multi-System Operators (―MSOs‖)] concentrate their operations in regional geographic areas by acquiring cable systems in regions where the MSO already has a significant presence, while giving up other holdings scattered across the country. This strategy is accomplished through purchases and sales of cable systems, or by system ‗swapping‘ among MSOs.‖ Implementation of the Cable Television Consumer Prot. & Competition Act of 1992, 22 F.C.C. Rcd. 17791, 17810 n.134 (2007) (citation omitted). An ―overbuilder‖ is a company that builds and offers customers a competitive alternative where a telecommunications company already operates. According to the Complaint, Comcast eliminated competition by (1) acquiring competitors in the Philadelphia market and 4 The ―Philadelphia cluster‖ and the ―Philadelphia DMA‖ are separate terms. The Philadelphia DMA includes the cluster counties as well as the counties of Lehigh and Northampton, Pennsylvania. See App. 03614, 03795. 8 (2) swapping with competitors cable systems and subscribers outside of the Philadelphia market for cable systems and subscribers within the Philadelphia market. The Complaint also alleged that Comcast engaged in conduct intended to exclude competition from overbuilder RCN Telecom Services, Inc. (―RCN‖), by denying it access to ―Comcast Sportsnet,‖ requiring contractors to enter non-compete agreements, and inducing potential customers to sign up for long contracts with special discounts and penalty provisions in the areas where RCN intended to overbuild. App. 00235- 239. As a result of its clustering, Comcast allegedly harmed the class by eliminating competition, raising entry barriers to potential competition, maintaining increased prices for cable services at supra-competitive levels, and depriving subscribers of the lower prices that would result from effective competition. App. 00241-242. In other words, Comcast subscribers allegedly pay too much for their non- basic video programming cable service. B. On May 3, 2007, after extensive motions practice, see App. 00148-172 (listing 194 docket entries prior to certification), the District Court certified the proposed class. App. 00354. It determined that Plaintiffs had met the requirements of Rule 23(a) of the Federal Rules of Civil Procedure (numerosity, commonality, typicality, and adequacy). App. 00366-372. It held also that Plaintiffs had met the predominance and superiority requirements of Rule 23(b). App. 00373-387. We denied on June 29, 2007, Comcast‘s 23(f) petition seeking interlocutory review. 9 The Court also certified the Chicago class‘s claims, but stayed them pending the outcome of the Philadelphia class. App. 00177, 00179.5 Following our decision in Hydrogen Peroxide, 552 F.3d 305, the District Court granted in part Comcast‘s motion to reconsider its Philadelphia certification decision (the Court denied without prejudice consideration of the Chicago class certification, again pending the outcome in Philadelphia). App. 00437-439. It vacated only the portion of the certification decision that addressed Rule 23(b)‘s predominance requirement. The Court scheduled a hearing on the issue of predominance as it related to (1) antitrust impact, and (2) methodology of damages. The District Court held an evidentiary hearing on October 13-15 and 26, 2009. During the four-day hearing, the Court heard live testimony from fact and expert witnesses, considered 32 expert reports, and examined deposition excerpts, as well as many other documents. Following the hearing, the Court issued to the parties a series of questions related to antitrust impact and damages methodology, and heard argument on November 16, 2009, to address its specific questions. 5 Plaintiffs‘ counsel also filed a complaint in the District of Massachusetts on behalf of a ―Boston cluster.‖ That case was transferred to the Eastern District of Pennsylvania, and has been stayed pending resolution of the Chicago cluster claims. See App. 00179. 10 On January 7, 2010, the District Court recertified the Philadelphia class, and issued an amended class certification order on January 13, 2010. The Court reaffirmed and incorporated its May 2007 certification as to numerosity, commonality, typicality, and adequacy (Rule 23(a)), as well as superiority (Rule 23(b)(3)). App. 00029. On the disputed issue of predominance, the Court held that Plaintiffs had demonstrated by a preponderance of the evidence that: (1) questions of law and fact common to the members of the class predominated; (2) the relevant geographic market could be the Philadelphia Designated Market Area; (3) the class could establish antitrust impact on the theory that Comcast‘s clustering through the swaps and acquisitions deterred overbuilder competition; (4) the models and analyses of Plaintiffs‘ damages expert, Dr. James McClave, were common evidence available to measure and quantify damages on a class-wide basis; and (5) the class could establish antitrust impact through common evidence applicable to all class members. App. 00030. In certifying the class, however, the District Court narrowed the class‘s various theories of class-wide impact to a single theory: Proof of antitrust impact relative to such claims shall be limited to the theory that Comcast engaged in anticompetitive clustering conduct, the effect of which was to deter the entry of overbuilders in the Philadelphia DMA. App. 00032. The Court accompanied its order with an 81-page memorandum opinion containing its analysis of the evidence presented at the evidentiary hearing. Behrend v. Comcast 11 Corp., 264 F.R.D. 150 (E.D. Pa. 2010). The Court summarized its opinion as follows: Having rigorously analyzed the expert reports, as well as the testimony presented by the parties during a four-day evidentiary hearing, we conclude that the class has met its burden to demonstrate that the element of antitrust impact is capable of proof at trial through evidence that is common to the class rather than individual to its members, and that there is a common methodology available to measure and quantify damages on a class-wide basis. Id. at 154. Comcast filed a Rule 23(f) petition to appeal on January 27, 2010. While that petition was pending, Comcast moved for summary judgment. The class responded, and Comcast filed a reply on June 4, 2010. We granted Comcast permission to appeal on June 9, 2010. The motion for summary judgment remains pending in the District Court. II. The District Court had jurisdiction pursuant to 28 U.S.C. §§ 1331 and 1337. We have jurisdiction pursuant to 28 U.S.C. § 1292(e) and Rule 23(f) of the Federal Rules of Civil Procedure. ―We review a class certification order for abuse of discretion, which occurs if the district court‘s decision rests upon a clearly erroneous finding of fact, an errant conclusion 12 of law or an improper application of law to fact.‖ Hydrogen Peroxide, 552 F.3d at 312 (citation and quotations omitted). We review de novo whether an incorrect legal standard has been used. Id. (citation omitted). For a district court‘s finding of fact to be clearly erroneous, the standard is high. ―Clearly erroneous‖ has been interpreted to mean that a reviewing court can upset a finding of fact, even if there is some evidence to support the finding, only if the court is ―left with the definite and firm conviction that a mistake has been committed.‖ United States v. U.S. Gypsum Co., 333 U.S. 364, 395 (1948). This means that ―[i]t is the responsibility of an appellate court to accept the ultimate factual determination of the fact-finder unless that determination either (1) is completely devoid of minimum evidentiary support displaying some hue of credibility, or (2) bears no rational relationship to the supportive evidentiary data.‖ Krasnov v. Dinan, 465 F.2d 1298, 1302 (3d Cir. 1972). Especially pertinent to the issue before us, the Supreme Court has explained: This standard plainly does not entitle a reviewing court to reverse the finding of the trier of fact simply because it is convinced that it would have decided the case differently. . . . In applying the clearly erroneous standard to the findings of a district court sitting without a jury, appellate courts must constantly have in mind that their function is not to decide factual issues de novo. If the district court‘s account of the evidence is plausible in light of the record viewed in its entirety, the court of appeals may not reverse it even though convinced that had it 13 been sitting as the trier of fact, it would have weighed the evidence differently. Where there are two permissible views of the evidence, the factfinder‘s choice between them cannot be clearly erroneous. Anderson v. Bessemer City, 470 U.S. 564, 573-574 (1985) (quotations and citations omitted); accord PA Prison Soc‘y v. Cortes, 622 F.3d 215, 231 (3d Cir. 2010). III. Comcast raises three principal arguments on appeal, urging us to overturn the District Court‘s certification order on the grounds that: (1) the Court‘s finding that the class can establish class-wide antitrust impact through common evidence was incorrect for various reasons; (2) the District Court exceeded its discretion in accepting Plaintiffs‘ proposed methodology for damages calculation; and (3) the Court‘s certification of a per se antitrust claim was clear error. In response, Plaintiffs defend in all respects the District Court‘s certification decision. We first outline the Rule 23 legal framework and then analyze each of Comcast‘s contentions. Rule 23 of the Federal Rules of Civil Procedure allows a class action if certain requirements are met. First, the class must meet the ―prerequisites‖ of Rule 23(a): numerosity, commonality, typicality, and adequacy. Second, the class must fit one of the Rule 23(b) types of classes. Here, Plaintiffs seek certification under Rule 23(b)(3), which requires (1) ―that the questions of law or fact common to class members predominate over any questions affecting only individual members,‖ and (2) ―that a class action is superior 14 to other available methods for fairly and efficiently adjudicating the controversy.‖ Rule 23(b)(3). These requirements are known as predominance and superiority. The district court must conduct a ―rigorous analysis‖ of the evidence and arguments in making the class certification decision. Hydrogen Peroxide, 552 F.3d at 318. The analysis requires ―a thorough examination of the factual and legal allegations‖ and ―may include a preliminary inquiry into the merits.‖ Id. at 317 (quoting Newton v. Merrill Lynch, Pierce, Fenner & Smith, Inc., 259 F.3d 154, 166, 168 (3d Cir. 2001)). We explained in Hydrogen Peroxide the permissible extent of any inquiry into the merits: [T]he requirements set out in Rule 23 are not mere pleading rules. The court may delve beyond the pleadings to determine whether the requirements for class certification are satisfied. . . . An overlap between a class certification requirement and the merits of a claim is no reason to decline to resolve relevant disputes when necessary to determine whether a class certification requirement is met. Some uncertainty ensued when the Supreme Court declared in Eisen v. Carlisle & Jacquelin, 417 U.S. 156, 177 (1974), that there is ―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.‖ Only a few years later, in addressing whether a party may bring an interlocutory appeal when a district court 15 denies class certification, the Supreme Court pointed out that ―the class determination generally involves considerations that are ‗enmeshed in the factual and legal issues comprising the plaintiff‘s cause of action.‘‖ [Coopers & Lybrand v.] Livesay, 437 U.S. [463,] 469 [(1978)] (quoting Mercantile Nat‘l Bank v. Langdeau, 371 U.S. 555, 558 (1963)). As we explained in Newton, 259 F.3d at 166- 69, Eisen is best understood to preclude only a merits inquiry that is not necessary to determine a Rule 23 requirement. Other courts of appeals have agreed. 552 F.3d at 316-317 (quotations and citations omitted).6 Accordingly, at the class certification stage, we are precluded from addressing any merits inquiry unnecessary to making a Rule 23 determination. Id. Further, any findings for the purpose of class certification ―do not bind the fact-finder on the merits.‖ Id. at 318. Plaintiffs bear the burden of establishing each element of Rule 23 by a preponderance of the evidence. Id. at 320 (―[T]o certify a class the district court must find that the evidence more likely than not establishes each fact necessary to meet the requirements of Rule 23.‖) (citation omitted). The 6 The Supreme Court confirmed our interpretation of the Rule 23 inquiry in Wal-Mart Stores, Inc. v. Dukes, 131 S. Ct. 2541 (2011). See id. at 2551, 2552 n.6 (stating that ―[f]requently [the Rule 23] ‗rigorous analysis‘ will entail some overlap with the merits of the plaintiff‘s underlying claim,‖ but Eisen still prohibits ―a merits inquiry for any other pretrial purpose‖). 16 court must also examine critically expert testimony on both sides and may be persuaded by either side as to whether a certification requirement has been met. Id. at 323. Indeed, ―[w]eighing conflicting expert testimony at the certification stage is not only permissible; it may be integral to the rigorous analysis Rule 23 demands.‖ Id. The parties dispute whether Plaintiffs have met the predominance requirement. Predominance ―tests whether proposed classes are sufficiently cohesive to warrant adjudication by representation.‖ Amchem Prods., Inc. v. Windsor, 521 U.S. 591, 623 (1997). It ―is a test readily met in certain cases alleging consumer or securities fraud or violations of the antitrust laws,‖ id. at 625, but a court may not relax its certification analysis as to each element of Rule 23, see Hydrogen Peroxide, 552 F.3d at 322. To assess whether common or individual issues predominate, a district court must examine the nature of the evidence and ―formulate some prediction as to how specific issues will play out . . . .‖ Id. at 311 (quotations and citations omitted). Reviewing a district court‘s certification of a class, we examine the elements of the class‘s claims ―through the prism‖ of Rule 23. Id. (quoting Newton, 259 F.3d at 181). The elements of the claims before us are (1) a violation of the antitrust laws (here, sections 1 and 2 of the Sherman Act), (2) individual injury resulting from that violation, and (3) measurable damages. See id. Individual injury, also known as antitrust impact, ―is critically important for the purpose of evaluating Rule 23(b)(3)‘s predominance requirement because it is an element of the claim that may call for individual, as opposed to common, proof.‖ Id. At the class certification stage, Plaintiffs‘ burden is ―to demonstrate that 17 the element of antitrust impact is capable of proof at trial through evidence that is common to the class rather than individual to its members.‖ Id. at 311-312. IV. Comcast devotes much of its energy to contending that the District Court exceeded its discretion in holding that Plaintiffs had established common evidence of antitrust impact. It attacks this issue in two ways: first, that the District Court failed to apply the correct legal standard for determining the relevant geographic market, and, second, that the District Court made clearly erroneous factual findings by relying on Plaintiffs‘ expert for proof of class-wide antitrust impact. We address each contention in turn. A. Before the District Court, Plaintiffs contended that the relevant geographic market was the Philadelphia Designated Market Area, whereas Comcast countered that it was each individual‘s household. The District Court agreed with Plaintiffs. Behrend, 264 F.R.D. at 160. Plaintiffs‘ expert, Dr. Michael Williams, provided seven bases to support the conclusion that the relevant geographic market was the Philadelphia DMA. The District Court set forth each basis, as well as Comcast‘s counterarguments. 264 F.R.D. at 157-160. The Court stated that Comcast‘s focus on the individual household was not supported by the record, and that setting such a small market would be ―impractical and inefficient.‖ 264 F.R.D. at 160. Instead, the Court noted that the alleged conduct centered on Comcast‘s attempt to acquire substantially all of the cable systems in the Philadelphia 18 DMA, and that the Federal Communications Commission aggregates relevant geographic markets in which customers face ―similar competitive choices.‖ 264 F.R.D. at 160. The Court concluded, ―[T]he record evidence shows that consumers throughout the DMA can face similar competitive choices and suffer the same alleged antitrust impact resulting from Comcast‘s clustering conduct in the Philadelphia DMA.‖ 264 F.R.D. at 160. Comcast contends that the Court failed to articulate or apply the correct legal standard. According to Comcast, the geographic market is defined in terms of consumer demand substitutability—the area in which a buyer may look for the goods or services he seeks. Because an individual can choose only among providers offering video programming services to his household, Comcast asserts that the geographic market must be the household. Comcast contends additionally that the Court improperly credited Dr. Williams‘s seven bases for the geographic market because it later rejected three of the seven theories, and that the Court‘s two stated reasons for accepting the geographic market were irrelevant and erroneous. Plaintiffs respond at three levels. First, they contend that they need not define the relevant geographic market: per se claims do not require defining the geographic market, and they offered direct evidence of market power, thereby relieving them of the obligation to define the relevant geographic market. Second, Plaintiffs state that the District Court used the commercial realities test to determine the relevant geographic market and did not ignore demand substitutability. Third, according to Plaintiffs, Comcast cannot demonstrate clear error in the Court‘s factual 19 determination that ―consumers throughout the [Philadelphia] DMA can face similar competitive choices.‖ See Behrend, 264 F.R.D. at 160. B. We will affirm the District Court‘s conclusion that the Philadelphia DMA is a relevant geographic market ―susceptible to proof at trial through available evidence common to the class.‖ 264 F.R.D. at 160. The relevant geographic market is a component of substantive antitrust law. For antitrust claims analyzed through the rule of reason, plaintiffs must demonstrate that the defendant possessed market power in the relevant geographic market. See Pa. Dental Ass‘n v. Med. Serv. Ass‘n of Pa., 745 F.2d 248, 260 (3d Cir. 1984). For per se claims, plaintiffs need not establish a geographic market. See In re Ins. Brokerage Antitrust Litig., 618 F.3d 300, 316-317 (3d Cir. 2010) (explaining that some prohibited practices can be conclusively presumed to unreasonably restrain competition). Additionally, ―direct proof of monopoly power does not require a definition of the relevant market.‖ See Broadcom Corp. v. Qualcomm Inc., 501 F.3d 297, 307 n.3 (3d Cir. 2007). Defining the relevant geographic market, however, is an issue of the merits. See, e.g., Borough of Lansdale v. Phila. Elec. Co., 692 F.2d 307 (3d Cir. 1982) (addressing on appeal whether jury verdict should be set aside because of allegedly erroneous definition of relevant geographic market). At the class certification stage, a court need only be satisfied that issues—including the definition of a geographic market—will 20 be capable of proof through evidence common to the class. See Hydrogen Peroxide, 552 F.3d at 311; IIA Phillip E. Areeda et al., Antitrust Law ¶ 398b (3d ed. 2007) (describing that at the class certification stage the plaintiffs‘ expert typically concludes that ―any significant economic issues underlying the class representative‘s antitrust claims, including but not limited to issues regarding market definition . . . will be analyzed and proven through the use of common data and evidence that would be used to prove the claims of the other members of the proposed Class‖) (emphasis added). If the plaintiffs allege per se claims, they may still need to persuade the district court that, in the event defining the relevant geographic market becomes necessary, it is capable of common proof. See Areeda et al., supra, ¶ 398b. The inquiry before the District Court, therefore, was whether Plaintiffs could demonstrate by a preponderance of the evidence that they would be able to establish a relevant geographic market capable of proof common to the class. The District Court concluded it was: ―We conclude that Dr. Williams‘ geographic market definition is susceptible to proof at trial through available evidence common to the class.‖ 264 F.R.D. at 160. The parties dispute whether the District Court properly defined the relevant geographic market—Comcast contends it erred as a matter of law, and Plaintiffs respond they need not establish a geographic market. These are merits arguments, which are not properly before us. Our review is limited to whether the Court exceeded its discretion in determining that the class could establish through common proof that the relevant geographic market could be the Philadelphia DMA. We conclude it did not, legally or factually. 21 C. First, we perceive no legal error in the District Court‘s reasoning. Procedurally, it conducted the required ―rigorous analysis‖ by examining in depth the expert opinions on both sides and setting forth its conclusions. See Hydrogen Peroxide, 552 F.3d at 317, 320. Substantively, the Court determined that ―the record evidence shows that consumers throughout the DMA can face similar competitive choices and suffer the same alleged antitrust impact resulting from Comcast‘s clustering conduct in the Philadelphia DMA.‖ 264 F.R.D. at 160. Comcast contends that the Court failed to apply the consumer demand substitutability test, which defines the relevant geographic market as ―that area in which a potential buyer may rationally look for the goods or services he seeks.‖ Gordon v. Lewiston Hosp., 423 F.3d 184, 212 (3d Cir. 2005) (citing Pa. Dental Ass‘n, 745 F.2d at 260). We determine otherwise: the Court‘s analysis of the relevant geographic market for purposes of class certification comported with our precedent. ―[I]dentification of the relevant geographic market is a matter of analyzing competition.‖ Borough of Lansdale v. Phila. Elec. Co., 692 F.2d 307, 311 (3d Cir. 1982). Defining it ―is a question of fact to be determined in the context of each case in acknowledgment of the commercial realities of the industry being considered.‖ Gordon, 423 F.3d at 212 (quoting Borough of Lansdale, 692 F.2d at 311). In these decisions of our Court, one of which has commanded our attention for almost thirty years, we relied on two Supreme Court cases to develop this standard: United States v. Grinnell Corp., 384 U.S. 563, 576 (1966), which held that the relevant geographic market under the Sherman Act was ―not the several local 22 areas which the individual stations serve, but the broader national market that reflects the reality of the way in which they built and conduct their business,‖ and Tampa Electric Co. v. Nashville Coal Co., 365 U.S. 320, 327, 332 (1961), which defined the relevant geographic area for § 3 of the Clayton Act, 15 U.S.C. § 3, as ―the market area in which the seller operates, and to which the purchaser can practicably turn for supplies‖ or as the area in which suppliers ―effectively compete.‖ In another Clayton Act case, the Supreme Court stated: ―The geographic market selected must, therefore, both correspond to the commercial realities of the industry and be economically significant. Thus, although the geographic market in some instances may encompass the entire Nation, under other circumstances it may be as small as a single metropolitan area.‖ Brown Shoe Co. v. United States, 370 U.S. 294, 336-337 (1962) (quotations and citations omitted); see Grinnell, 384 U.S. at 572 (citing Brown Shoe as analogous to determining the relevant market for the Sherman Act). D. The District Court‘s determination—that consumers ―face similar competitive choices‖ in the Philadelphia DMA as a result of Comcast‘s alleged clustering conduct—is consistent with the above standards because it considers both where a buyer may rationally look for goods and the commercial reality of the industry. Comcast‘s insistence that the geographic market must be the individual household (as the only place where a consumer can ―comparison shop‖) ignores that the geographic market must be ―economically significant,‖ Brown Shoe Co., 370 U.S. at 336-337, and may be premised on ―the commercial realities of the industry 23 being considered,‖ Borough of Lansdale, 692 F.2d at 311, the area where suppliers ―effectively compete,‖ Tampa Electric Co., 365 U.S. at 332, or the broader market reflecting the reality of conducting business, Grinnell, 384 U.S. at 576.7 We therefore discern no legal error in the District Court‘s analysis. E. 7 We note additionally the tension between the concept of a ―geographic market‖ and Comcast‘s conclusion that ―the relevant geographic market . . . is each class member‘s residence.‖ Appellants‘ Br. 15. As of 2009, Philadelphia County alone had over 560,000 households. See U.S. Census Bureau, Philadelphia County QuickFacts, http://quickfacts.census.gov/qfd/states/42/42101.html. Nationwide, in 2010 there were over 117 million households. See U.S. Census Bureau, America‘s Families and Living Arrangements: 2010, http://www.census.gov/population/www/socdemo/hh- fam/cps2010.html (Table AVG1). Taken at face value, Comcast‘s assertion that there are millions of geographic markets in the Philadelphia DMA (or over one hundred million geographic markets nationwide for multichannel video programming distributors) renders the phrase ―geographic market‖ nonsensical. Perhaps for this reason, our research revealed no case—nor does Comcast provide one— in which a geographic market has been set at the individual household level. Cf. Brown Shoe, 370 U.S. at 337 (―Thus, although the geographic market in some instances may encompass the entire Nation, under other circumstances it may be as small as a single metropolitan area.‖). 24 Second, we recognize ample evidence in the record supporting the District Court‘s factual findings underpinning its market determination, which precludes us from reversing those findings as clearly erroneous. See, e.g., EBC, Inc. v. Clark Bldg. Sys. Inc., 618 F.3d 253, 273 (3d Cir. 2010) (―We will not reverse ‗[i]f the district court‘s account of the evidence is plausible in light of the record viewed in its entirety‘ even if we would have weighed that evidence differently.‖ (quoting Anderson, 470 U.S. at 573-574)). The Court cited Dr. Williams‘s seven bases for drawing the geographic market as the Philadelphia DMA. Behrend, 264 F.R.D. at 157-160. Although it rejected three of those bases, the remaining four tended to show that Comcast‘s clustering had anticompetitive effects in the Philadelphia DMA by deterring overbuilders from entering the Designated Market Area, and that the industry itself used DMAs to focus its competition. Additional evidence in the record, reviewed in detail below, demonstrated that clustering results in fewer competitors and higher cable prices for the entire market. This evidence belies Comcast‘s claim that there is no change at the individual level when Comcast aggregates surrounding franchises. Simply put, the District Court determined by a preponderance of the evidence that, when addressed on the merits, the class may be able to prove through common evidence that the relevant geographic market is the Philadelphia DMA. This determination did not exceed the Court‘s permissible discretion. To the extent Comcast reads the Court‘s opinion as actually fixing the relevant geographic market, we note that its determination was made solely for the 25 purposes of class certification and will not be binding on the merits. See Hydrogen Peroxide, 552 F.3d at 318.8 8 The Concurring and Dissenting Opinion (―Concurrence- Dissent‖) faults the parties, the District Court and this Opinion for using ―equivocally‖ the phrase ―relevant geographic market.‖ Slip Concurring-Dissenting Op. at 8. Specifically, it asserts that this Opinion ―assumes . . . that the class is properly defined to cover the Philadelphia DMA . . . .‖ Id. at 10. The Concurrence-Dissent misunderstands an important distinction: as noted supra footnote 4, Plaintiffs have alleged a ―class region‖ (to borrow from the Concurrence-Dissent‘s terminology) of a ―Philadelphia cluster,‖ which is distinct from the contested relevant geographic market of the ―Philadelphia DMA.‖ Our ―assumption‖ concerning the ―class region‖ is an uncontested piece of Plaintiffs‘ case: Comcast appeals only the precise issue of whether the District Court applied a correct legal standard in determining that the substantive antitrust geographic market could be established by evidence common to the class, not whether the ―Philadelphia cluster‖ is an appropriate ―class region.‖ See Appellants‘ Br. at 15 (labeling the issue as: ―The District Court Failed To Apply The Correct Legal Standard In Its Ruling On Plaintiffs‘ Geographic Market Definition‖); id. at 20 (summarizing that ―the alleged geographic market accepted by the district court is wholly divorced from the legal standard for determining the correct geographic market‖). Accordingly, when the Concurrence- Dissent states, ―A compelling argument could be made . . .,‖ Slip Concurring-Dissenting Op. at 11 (emphasis added), it goes beyond our role as a reviewing court by raising and addressing an argument not before us. See, e.g., AT & T v. F.C.C., 582 F.3d 490, 495 (3d Cir. 2009) (―An appellant 26 V. Comcast hinges its next line of arguments on the District Court‘s final certification: ―Proof of antitrust impact relative to such claims shall be limited to the theory that Comcast engaged in anticompetitive clustering conduct, the effect of which was to deter the entry of overbuilders in the Philadelphia DMA.‖ App. 00032. According to Comcast, the District Court made clearly erroneous findings of fact by relying on Plaintiffs‘ expert, Dr. Williams, in support of the certified theory of antitrust impact. The District Court considered in great detail the arguments presented by both sides. It rejected three of Plaintiffs‘ four theories of class-wide impact. Behrend, 264 F.R.D. at 166 (rejecting theory of direct broadcast satellite (―DBS‖) foreclosure); id. at 177-178 (rejecting benchmark theory); id. at 181 (rejecting bargaining power theory). Nonetheless, it accepted that Plaintiffs could establish class- wide antitrust impact on the theory of clustering and its impact on overbuilder competition. After detailing the evidence put forth by both sides, id. at 166-174, the Court concluded that ―the Class has met its burden to demonstrate that the anticompetitive effect of clustering on overbuilder competition is capable of proof at trial through evidence that is common to the class,‖ id. at 174. The Court found that through the model of Plaintiffs‘ expert, Dr. Williams, and the empirical studies conducted by governmental agencies and waives an argument in support of reversal if he does not raise that argument in his opening brief.‖), rev‘d on other grounds, 131 S. Ct. 1177 (2011). 27 private researchers, the class had shown that the presence of an overbuilder constrains cable prices, and that Comcast engaged in conduct designed to deter the entry of overbuilders in the Philadelphia DMA. Id. at 174. It found unpersuasive the conclusions of Comcast‘s expert, Dr. David J. Teece, that overbuilding is not a successful business model. Id. at 174-175. A. On appeal, Comcast constructs a four-tiered argument to support its objections. First, it contends that Plaintiffs cannot show class-wide antitrust impact based on potential overbuilding by any of the ―Transaction parties.‖9 According to Comcast, the evidence demonstrated there was no actual competition between the Transaction parties; Plaintiffs therefore must show that the challenged conduct eliminated potential competition. In Comcast‘s view, the record evidence reflects that no Transaction parties had taken any affirmative steps to overbuild and, consequently, there was no potential competition to eliminate. Second, Comcast contends that Plaintiffs identified only RCN Telecom Services, Inc., as attempting to overbuild in the Philadelphia DMA. The evidence establishes, according to Comcast, that RCN was not going to overbuild as a result of its own financial woes, not as a result of any alleged activity on the part of Comcast. 9 As detailed supra note 2, the ―Transaction parties‖ are the parties that Comcast acquired or with which it swapped cable systems, which include: Marcus Cable; Greater Philadelphia Cablevision, Inc.; Lenfest Communications, Inc.; AT&T; Adelphia Communications Corp.; Time Warner; and Patriot Media. 28 Third, as the argument goes, because there was no record evidence demonstrating actual or potential competition, the theoretical opinions indicating otherwise rendered by Plaintiffs‘ expert, Dr. Williams, were clearly erroneous. Comcast disputes at many levels Dr. Williams‘s methodology and results in his ―market structure‖ and ―market performance‖ opinions. Summed up, Comcast contends that theoretical expert opinions are no replacement for market facts, the record evidence showed no actual or potential overbuilding (as addressed in the first two contentions), and therefore any reliance on the expert opinions for evidence of anticompetitive behavior was clearly erroneous. Fourth, Comcast adds that any evidence of anticompetitive conduct specific to Delaware County could not serve as evidence of class-wide impact for the Philadelphia cluster. B. Plaintiffs respond to each level of Comcast‘s position. First, citing many portions of the record, they assert that there is ―overwhelming‖ record evidence that Comcast‘s clustering of the Philadelphia DMA deterred and reduced overbuilding competition, resulting in antitrust impact (higher cable prices) for all class members. According to the class, the record demonstrates: clustering deters overbuilding, the swaps and acquisitions eliminated competition, Multi-System Operators (―MSOs‖) actually do overbuild one another, Comcast and other MSOs look to one another‘s prices to set their own, and the MSOs chose affirmatively not to compete. The class adds that Comcast is raising a merits argument by asking the Court to consider the ―potential competition‖ doctrine. Second, Plaintiffs contend that Comcast raises a merits issue by asking the Court to examine whether Comcast‘s conduct in fact 29 prevented RCN from overbuilding in more areas than it did. In any event, they state that the record evidence demonstrates RCN had the intent and capital to overbuild the Philadelphia market. Third, Plaintiffs state that Dr. Williams‘s theoretical model plainly shows common evidence of class-wide impact; Comcast‘s contention that Dr. Williams‘s opinions do not prove antitrust impact is one for the jury to decide on the merits. Fourth, the evidence related to Delaware County ―adds to and illustrates‖ the common evidence of Comcast‘s anticompetitive clustering conduct. VI. We begin the analysis of these contentions by focusing on the precise inquiry: Plaintiffs‘ burden at the class certification stage is not to prove the element of antitrust impact, although in order to prevail on the merits each class member must do so. Instead, the task for plaintiffs at class certification is to demonstrate that the element of antitrust impact is capable of proof at trial through evidence that is common to the class rather than individual to its members. Hydrogen Peroxide, 552 F.3d at 311-312 (emphasis added). Many of Comcast‘s contentions ask us to reach into the record and determine whether Plaintiffs actually have proven antitrust impact. This we will not do. Instead, we inquire whether the District Court exceeded its discretion by finding that Plaintiffs had demonstrated by a preponderance of the 30 evidence that they could prove antitrust impact through common evidence at trial. This dispute therefore is evidentiary. When facts are at issue, the District Court exceeds its discretion in certifying a class only if its findings are clearly erroneous. Id. at 312. Comcast bears a heavy burden in convincing us that the District Court‘s factual findings were clearly erroneous. See Anderson, 470 U.S. at 573-574 (―If the district court‘s account of the evidence is plausible in light of the record viewed in its entirety, the court of appeals may not reverse it . . . .‖); Krasnov, 465 F.2d at 1302 (―It is the responsibility of an appellate court to accept the ultimate factual determination of the fact-finder unless that determination either (1) is completely devoid of minimum evidentiary support displaying some hue of credibility, or (2) bears no rational relationship to the supportive evidentiary data.‖). Comcast has not carried its burden. Plaintiffs provided evidence at the certification hearing that tended to show that Comcast‘s clustering (through swaps and acquisitions) reduced competition, deterred the entry of overbuilders, and resulted in higher cable prices for the entire class. This evidence displays ―some hue of credibility‖ and bears a rational relationship to the Court‘s finding. See Krasnov, 465 F.2d at 1302. For example, one of Plaintiffs‘ experts, Dr. Williams, concluded after a detailed analysis that, inter alia, Comcast‘s clustering increased its market share and, consequently, its market power, thereby raising barriers to entry for other multichannel video programming distributors and resulting in higher cable rates for all members of the class. App. 03599- 31 3600; see also Behrend, 264 F.R.D. at 166-171 (providing in great detail the analyses, evidentiary support, and conclusions of Dr. Williams). Dr. Williams also cited to Federal Communications Commission reports, Government Accountability Office reports, and academic research, all of which indicated that reducing competition by clustering leads to higher cable rates. App. 03663-3668. Another expert, Dr. Hal Singer, used extensive record evidence to analyze how Comcast‘s clustering denied overbuilders access to the Philadelphia DMA. App. 03501-3529. Dr. Singer concluded that Comcast‘s actions allowed it to foreclose competitors and elevate prices. App. 03450. He also referenced multiple studies—both governmental and private, some of which overlapped with those referenced by Dr. Williams—that concluded that cable prices are lower when overbuilder competition is present. App. 03537-3548. Also in the record are specific instances of Multi-System Operators attempting to overbuild one another around the country. See Appellees‘ Br. 27 n.17 (citing 13 distinct examples in the record of MSOs overbuilding one another). All of this evidence demonstrates that Comcast‘s alleged clustering conduct indeed could have reduced competition, raised barriers to market entry by an overbuilder, and resulted in higher cable prices to all of its subscribers in the Philadelphia Designated Market Area. Based on this evidence, we determine that the antitrust impact Plaintiffs allege is ―plausible in theory‖ and ―susceptible to proof at trial through available evidence common to the class.‖ Hydrogen Peroxide, 552 F.3d at 325; see also In re Linerboard Antitrust Litig., 305 F.3d 145, 158 (3d Cir. 2002) (holding that common issues predominated sufficient for class certification when plaintiffs allegedly ―were all affected by 32 the increased price‖ they paid for linerboard). We are satisfied that the District Court‘s findings were supported by the evidence and were not clearly erroneous. Comcast protests that the record demonstrates that there was no actual or potential competition among the Transaction parties. In light of the above record evidence, however, Comcast‘s interpretation of the evidence does not render the District Court‘s findings clearly erroneous. Comcast remains free to make these arguments to the jury. VII. Comcast‘s other contentions are equally unpersuasive. There is conflicting evidence as to the role Comcast played in RCN Telecom Services, Inc.‘s decision to not overbuild further in the Philadelphia DMA. Plaintiffs highlight record evidence that RCN had the intent and capital necessary to overbuild the Philadelphia market. Appellees‘ Br. 34-35. Comcast contends instead that RCN faced financial woes, as a result of which it abandoned its plans to overbuild. Appellants‘ Br. 24-28. The District Court credited Plaintiffs‘ explanation: ―What Dr. Teece considers ‗unlikely,‘ Dr. Singer considers to be the common evidence of antitrust impact, namely that RCN was stymied in its efforts by Comcast‘s predatory behavior.‖ Behrend, 264 F.R.D. at 175. Again, we are satisfied that the District Court‘s finding was not clearly erroneous. ―Where there are two permissible views of the evidence, the factfinder‘s choice between them cannot be clearly erroneous.‖ Anderson, 470 U.S. at 574. Here there are two permissible views of the evidence and we will not disturb the District Court‘s finding. 33 Similarly, Comcast contends that Dr. Williams‘s analysis and methodology was flawed for various reasons, including the allegation that it was unsupported by any actual evidence. We disagree. As detailed above, there was ample evidence that clustering conduct can deter entry of overbuilders and result in higher cable prices. Dr. Williams and Dr. Singer examined evidence specific to Comcast‘s activities in the Philadelphia market, as well as numerous independent studies on the effects of cable clustering, to reach their conclusions. Comcast cites various cases for the proposition that ―expert theory is not a substitute for market facts.‖ See, e.g., Brooke Grp. Ltd. v. Brown & Williamson Tobacco Corp., 509 U.S. 209, 242 (1993) (expert opinion rendered unreasonable by indisputable record facts); In re New Motor Vehicles Canadian Export Antitrust Litig., 522 F.3d 6, 27 (1st Cir. 2008) (expert analysis unfinished and ―purely conclusory‖); In re Baby Food Antitrust Litig., 166 F.3d 112, 135 (3d Cir. 1999) (―An expert opinion based on . . . meager superficial information . . . is highly speculative, unreliable, and of dubious admissibility.‖). Although expressing a correct legal precept, those cases addressed situations in which the experts largely failed to tie their theories to any evidence; the precept therefore does not apply to this case in which the experts‘ theories were based on and correlated to other record evidence. Comcast also asserts that every individual had one or two options from which to choose cable and that consequently only the name of the provider changed, not the number of options. This assertion completely overlooks the nature of the claims of the class: by clustering, Comcast was able to deter the entry of overbuilders, which resulted in higher prices for all non-basic Comcast subscribers. And 34 Plaintiffs provided evidence that clustering can have this effect. In short, the District Court‘s task was to weigh expert testimony and make a determination, Hydrogen Peroxide, 552 F.3d at 323, and we discern no error in the Court‘s determination that Dr. Williams‘s analysis demonstrated that class-wide antitrust impact was susceptible to common proof. As to Comcast‘s remaining contention that the District Court erred by crediting as evidence of class-wide impact the alleged conduct targeted at RCN Telecom Services, Inc., in Delaware County, we agree with the class that the alleged conduct is relevant to establishing class-wide impact. We have explained that ―courts must look to the monopolist‘s conduct taken as a whole rather than considering each aspect in isolation.‖ LePage‘s Inc. v. 3M, 324 F.3d 141, 162 (3d Cir. 2003) (en banc) (citing Cont‘l Ore Co. v. Union Carbide & Carbon Corp., 370 U.S. 690, 699 (1962)). Alleged specific conduct aimed at preventing the entry of an overbuilder anywhere in the Philadelphia DMA supports Plaintiffs‘ allegations of Comcast‘s ability to maintain supra- competitive prices for the entire market. VIII. At bottom, Comcast misconstrues our role at this stage of the litigation. Comcast would have us decide on the merits whether there was actual or potential competition among the Transaction parties, the reason RCN Telecom Services, Inc., abandoned the Philadelphia market, and whether Plaintiffs‘ experts proved antitrust impact. We are not the jury. Although in Hydrogen Peroxide we heightened the inquiry a district court must perform on the issue of class certification, nothing in that opinion indicated that class certification hearings were 35 to become actual trials in which factual disputes are to be resolved. Indeed, as we explained in Hydrogen Peroxide, a district court may inquire into the merits only insofar as it is ―necessary‖ to determine whether a class certification requirement is met. 552 F.3d at 316. Eisen still precludes any further inquiry. See Eisen, 417 U.S. at 178 (―[T]he question is not whether the plaintiff or plaintiffs . . . will prevail on the merits, but rather whether the requirements of Rule 23 are met.‖ (quoting Judge Wisdom‘s holding in Miller v. Mackey Int‘l, Inc., 452 F.2d 424, 427 (5th Cir. 1971))); Hydrogen Peroxide, 552 F.3d at 317 (―Eisen is best understood to preclude only a merits inquiry that is not necessary to determine a Rule 23 requirement.‖). We allow preliminary merits inquiries when necessary for Rule 23 because of the potentially ―decisive effect on litigation‖ of a certification decision, Newton, 259 F.3d at 167, but those inquiries remain limited and non-binding on the merits at trial, Hydrogen Peroxide, 552 F.3d at 318. Nothing in Hydrogen Peroxide requires plaintiffs to prove their case at the class certification stage; to the contrary, they must establish by a preponderance that their case is one that meets each requirement of Rule 23. 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.10 10 Indeed, recent scholarship uniformly has expressed concern over the trend towards converting certification decisions into mini trials. See Joshua P. Davis & Eric L. Cramer, Antitrust, Class Certification, and the Politics of Procedure, 17 Geo. Mason L. Rev. 969, 970 (2010) (contending that the ―judicial tendency to impose requirements at class certification‖ serves no legitimate purpose and risks violating the Seventh Amendment); Michael J. Kaufman & John M. Wunderlich, 36 In sum, we hold that the District Court‘s determination—that Plaintiffs have demonstrated by a preponderance of the evidence that they can establish class- wide antitrust impact through common evidence—did not exceed its discretion. IX. To satisfy another portion of the predominance requirement, Plaintiffs must establish that the alleged damages are capable of measurement on a class-wide basis using common proof. See Hydrogen Peroxide, 552 F.3d at 311, 325-326; cf. Newton v. Merrill Lynch, Pierce, Fenner & The Unjustified Judicial Creation of Class Certification Merits Trials in Securities Fraud Actions, 43 U. Mich. J.L. Reform 323, 323 (2010) (stating that judicial resolution of merits at the certification stage precludes victims from obtaining redress, infringes on the Seventh Amendment, and serves no legitimate policy concerns); Steig D. Olson, ―Chipping Away‖: The Misguided Trend Toward Resolving Merits Disputes as Part of the Class Certification Calculus, 43 U.S.F. L. Rev. 935, 940 (2009) (intensifying the Rule 23 analysis is inconsistent with the Rule itself and highly inefficient); J. Douglas Richards & Benjamin D. Brown, Predominance of Common Questions – Common Mistakes in Applying the Class Action Standard, 41 Rutgers L.J. 163, 169 (2009) (contending, inter alia, that requiring the district court to determine by a preponderance whether plaintiffs‘ proposed proof is actually correct or incorrect would ―substitute a court‘s own evaluation of key merits questions for that of the jury‖). 37 Smith, Inc., 259 F.3d 154, 187 (3d Cir. 2001) (stating that the ―Herculean task‖ of calculating individual damages from hundreds of millions of different transactions ―counsels against finding predominance‖). The District Court concluded that Plaintiffs, through their expert Dr. McClave, provided a damages model based on a common methodology available to measure and quantify damages on a class-wide basis. 264 F.R.D. at 191. Comcast assails that determination as an abuse of discretion. A. The District Court examined the methodology, conclusions, and criticisms of the experts on both sides, before providing its conclusions. 264 F.R.D. at 181-191. (Comcast does not contest that the Court performed the ―rigorous analysis‖ required by Hydrogen Peroxide.) Because on appeal Comcast renews the arguments it made to the District Court, we set forth each side‘s position in the District Court and the Court‘s response. Plaintiffs‘ damages expert, Dr. McClave, concluded that the prices in the Philadelphia market were consistently and substantially higher than the prices in areas of effective competition. 264 F.R.D. at 181. His econometric analysis demonstrated that the alleged antitrust impact was class-wide, because the prices were elevated above competitive levels across all class members and for the entire time period. Id. For his methods, Dr. McClave constructed ―but-for‖ prices against which to compare the prices Comcast charged in the Philadelphia DMA. ―But-for‖ prices are those that would have existed absent the alleged anticompetitive conduct. To construct the ―but-for‖ prices, he first selected comparable 38 ―benchmark‖ counties around the country by applying two ―screens‖ to determine whether the counties represented a level of competition similar to what Comcast would have faced in the Philadelphia market absent its alleged anticompetitive conduct. It is important to understand these two screens. The first screen—the ―market share screen‖ or ―40% screen‖—required that the county have a Comcast subscriber penetration rate of less than 40%. App. 03410. Dr. McClave chose 40% because it represented the approximate midpoint of Comcast‘s penetration rate in the Philadelphia DMA (between approximately 20% in 1998 and 60% from 2003 through 2008). He chose this number also because it allowed for growth during the class period but focused on markets where Comcast was likely to have less market power than it does in the Philadelphia market. Id. The second screen—the ―Direct Broadcast Satellite screen‖, or ―DBS screen‖—required that the county be in a Designated Market Area where the penetration level for Alternative Delivery Systems (which essentially includes DBS, but also master antenna systems and multipoint distribution systems) was at or higher than the national average of Alternative Delivery Systems penetration rates in Comcast markets.11 Using data 11 Dr. McClave used Alternative Delivery Systems penetration rates as a proxy to measure Direct Broadcast Satellite penetration rates. App. 03410 n.11. Comcast‘s expert, Dr. Chipty, referred to the screen as the ―DBS screen,‖ and as measuring DBS penetration rates. App. 03834. The parties and District Court have continued using the DBS terminology. Although the screen technically measured Alternative Delivery Systems penetration rates, we will use the parties‘ terminology and refer to it as the ―DBS 39 from the counties that fit the two screens, Dr. McClave performed a multiple regression analysis to compare actual prices in the Philadelphia DMA to the estimated ―but-for‖ prices. He then applied the overcharge percentage to the relevant revenue obtained by Comcast for expanded basic service in the Philadelphia market during the class period to reach a final conservative estimated overcharge value: $875,576,662. Comcast‘s experts, Dr. Teece and Dr. Tasneem Chipty, contested several parts of Dr. McClave‘s methodology, and questioned his results. 264 F.R.D. at 183. First, they challenged both benchmark screens used by Dr. McClave. Regarding the ―DBS screen,‖ Dr. Teece asserted that Dr. McClave erroneously chose the higher national Direct Broadcast Satellite penetration rate, instead of the lower regional rate predicted by Plaintiffs‘ experts Dr. Singer and Dr. Williams. The District Court rejected the critique, stating that Dr. McClave ―used his national average DBS penetration screen as a descriptor of typical competitive market conditions,‖ and was not attempting to predict the Direct Broadcast Satellite penetration rate of the Philadelphia DMA. Id. at 184. Regarding the ―market share screen,‖ Dr. Chipty contended that because Comcast was present in only a few counties in 1999, its actual market share was much higher in the counties where it was and 0% where it was not; as a result, the less-than-40% penetration rate provided an inappropriate screen. App. 03833. The District Court rejected the criticism as unsupported by the record, stating that Dr. Chipty should have presented evidentiary data to show that screen‖ and as measuring Direct Broadcast Satellite penetration rates. 40 40% was an incorrect midpoint estimate or average rate. 264 F.R.D. at 184. The Court also noted that the 40% screen was supported by the evidence as Comcast‘s approximate share of the Philadelphia DMA at the midpoint of the class period. Id. at 184 n.43. Second, Dr. Chipty faulted Dr. McClave‘s model for failing to consider properly demographic variables among the counties: specifically, for omitting the variables of population density and the number and type of households. The District Court credited as well-supported Dr. McClave‘s response as to why he omitted population density: it is correlated with medium household income (which he included) and using it as well as household income would create confounding and unreliable results. 264 F.R.D. at 185-186. Additionally, according to Dr. McClave, adding it would mask the effects of anticompetitive influences because higher population density results in lower costs per subscriber. Id. at 185. The Court noted that Dr. Chipty‘s use of population density as a variable resulted in it being positive and statistically significant in one model but negative and statistically significant in another. Id. at 186. Moreover, the Court added that Federal Communications Commission and Government Accountability Office studies included population density but found it was not a statistically significant variable. Id. Third, Dr. Chipty criticized Dr. McClave‘s model for comparing list prices for expanded basic cable in the Philadelphia DMA against the benchmark counties. She opined that Dr. McClave‘s model did not take into account the significant number of promotions and discounts offered to Comcast customers. Id. at 187. Dr. Chipty offered several rebuttal models that included population density and 41 discounted prices, which resulted in significantly lower or even negative damages. The Court rejected Dr. Chipty‘s models as ―suffer[ing] significant flaws.‖ Id. at 188, 189. It stated that Dr. McClave‘s model accounted for discount prices in the formula (not model) when he multiplied anticompetitive overcharge by Comcast‘s relevant revenues (because Comcast receives revenue only for prices charged, the revenue side of the formula by definition includes discount prices). Accordingly, by adding discount prices to the model as well, Dr. Chipty‘s model doubly counted the discount. The Court also noted that, as Dr. McClave explained, more than 80% of Comcast‘s customers pay list price for expanded basic cable, and discounts from list prices are temporary (after which they return to list price). As to another of Dr. Chipty‘s models, which calculated damages through direct calculations instead of multiple regression, the Court rejected it in the words of Dr. McClave as a ―novel and non-standard formula for calculating damages.‖ Id. at 189. Fourth, the District Court rejected Dr. Chipty‘s attempt to impeach Dr. McClave‘s model by using it to calculate damages for basic cable prices, instead of expanded basic cable. Id. at 190. The Court explained that Dr. McClave‘s model aimed to analyze only expanded basic cable, because Comcast alters its prices at the expanded level, so ―any application of the McClave model to [basic cable prices] explains nothing.‖ Id. Comcast does not contest that ruling. Fifth and finally, the Court asked the parties after the hearing how to interpret Dr. McClave‘s damages model if it credited at least one, but not all, of Dr. Williams‘s four theories of antitrust impact. Id. It determined that Dr. McClave‘s damages model was still viable, even if it rejected 42 some theories of antitrust impact, explaining that Dr. McClave selected benchmarks to isolate the effect of anticompetitive conduct, and that his use of the DBS screen was ―entirely unrelated‖ to Dr. Williams‘s DBS foreclosure theory. Id. The Court concluded that Dr. Williams‘s theories of antitrust impact were not relevant to Dr. McClave‘s methods of choosing benchmarks because ―[a]ny anticompetitive conduct is reflected in the Philadelphia DMA price, not in the selection of the comparison counties.‖ Id. at 191. B. Comcast contends that the District Court exceeded its discretion in accepting Plaintiffs‘ proposed damages calculation methodology. Its arguments are recast versions of those rejected by the District Court. First, Comcast contends that Dr. McClave‘s damages theory was based on all of Plaintiffs‘ alleged anticompetitive effects, but the District Court rejected three of Plaintiffs‘ four theories. Because Dr. McClave stated that his model was based on the cumulative effect and could not isolate damages for individual theories of harm, according to Comcast the District Court erred in accepting the damages model. Second, Comcast asserts that the economic assumptions underlying the damages model lack foundation in the record evidence. According to Comcast, both screens employed by Dr. McClave are factually unsupported and economically unsound: the ―DBS penetration screen‖ because the Court rejected Dr. Williams‘s Direct Broadcast Satellite foreclosure theory, and the ―market share screen‖ because it bears no relation to the conditions that would have existed in the Philadelphia region but for the complained-of conduct. Third, Comcast contends that the 43 damages model is flawed because it fails to include population density as a variable, and because it calculates damages based on list prices, which fails to consider the discounted prices that some subscribers actually pay.12 Plaintiffs remind us that the District Court already thoroughly considered and rebutted each of the points that Comcast now raises. As to the specific contentions, first, the class asserts that the District Court explicitly held that Dr. McClave‘s model was suitable for calculation of damages on all or individual theories of liability. Second, the class emphasizes that the damages model provides a methodology that can establish damages on a class-wide basis using common proof, and that Comcast ignores the proper inquiry at class certification and instead prematurely attacks the merits of the model. As a result, Comcast‘s arguments concerning the benchmarks miss the point. Third, the class asserts that Dr. McClave had ample justification to omit population density as a variable, and that the damages model incorporates discount prices. X. 12 Following the Supreme Court‘s decision in Wal-Mart, Comcast added that Dr. McClave‘s damages model, like the expert model in Wal-Mart, could be ―safely disregard[ed].‖ See Wal-Mart, 131 S. Ct. at 2554. We disagree. 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. 44 We pause to identify the forest for the trees. If allowed to proceed to trial, the class must establish that the injury it suffered from the violation of the antitrust laws is measurable. See Hydrogen Peroxide, 552 F.3d at 311; see also Newton, 259 F.3d at 188 (―Proof of injury (whether or not an injury occurred at all) must be distinguished from calculation of damages (which determines the actual value of the injury).‖). The usual measure in an overcharge case ―is the difference between the illegal price that was actually charged and the price that would have been charged ‗but for‘ the violation multiplied by the number of units purchased.‖ Areeda et al., supra, ¶ 392a. 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.‖ Story Parchment Co. v. Paterson Parchment Paper Co., 282 U.S. 555, 563 (1931) (―[W]hile the damages may not be determined by mere speculation or guess, it will be enough if the evidence show the extent of the damages as a matter of just and reasonable inference . . . .‖); see also Stelwagon Mfg. Co. v. Tarmac Roofing Sys., Inc., 63 F.3d 1267, 1273 (3d Cir. 1995) (citing Story Parchment and explaining that ―damage issues in these cases are rarely susceptible to the kind of concrete, detailed proof of injury which is available in other contexts‖). The inquiry for a district court at the class certification stage is whether the plaintiffs have demonstrated by a preponderance of the evidence that they will be able to measure damages on a class-wide basis using common proof. See Hydrogen Peroxide, 552 F.3d at 325. Some variation of damages among class members does not defeat certification. See 7AA Charles Alan Wright et al., Federal Practice and Procedure § 1781 (3d ed. 2005) (stating for antitrust class 45 certification that ―it uniformly has been held that differences among the members as to the amount of damages incurred does not mean that a class action would be inappropriate.‖). Complex and individual questions of damages, however, weigh against finding predominance. Compare Newton, 259 F.3d at 187 (reasoning that having to examine proof of the circumstances of hundreds of millions of individual transactions counseled against finding predominance), with Linerboard, 305 F.3d at 157-158 (determining that, in contrast to Newton, all purchasers were affected by the increased price). As the Court of Appeals for the First Circuit explained: It is true that the validity of plaintiffs‘ theory is a common disputed issue. It will be for the fact finder to decide whether this theory is persuasive. At the class certification stage, however, the district court must still ensure that the plaintiffs‘ presentation of their case will be through means amenable to the class action mechanism. We are looking here not for hard factual proof, but for a more thorough explanation of how the pivotal evidence behind plaintiff‘s theory can be established. If there is no realistic means of proof, many resources will be wasted setting up a trial that plaintiffs cannot win. In re New Motor Vehicles Canadian Export Antitrust Litig., 522 F.3d 6, 29 (1st Cir. 2008) (citation omitted); see also Areeda et al., supra, ¶ 331 (explaining for the issue of 46 damages that ―courts will not permit class actions unless they can devise a practical means for their litigation‖).13 13 In response to the Concurrence-Dissent‘s position that Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993), applies at the stage of class certification, see Slip Concurrence-Dissent Op. at 16, we make two observations. First, as the Opinion acknowledges, ―in neither the District Court nor before us‖ did Comcast raise this issue, id. at 17 n.18, and it is therefore not properly before us. Second, although the Supreme Court recently hinted that Daubert may apply for evaluating expert testimony at the class certification stage, it need not turn class certification into a mini-trial. Wal-Mart, 131 S. Ct. at 2553-54. We understand the Court‘s observation to require a district court to evaluate whether an expert is presenting a model which could evolve to become admissible evidence, and not requiring a district court to determine if a model is perfect at the certification stage. This is consistent with our jurisprudence which requires that at class certification stage, we evaluate expert models to determine whether the theory of proof is plausible. Hydrogen Peroxide, 552 F.3d at 324. ―[I]f such impact is plausible in theory, it is also susceptible to proof at trial through available evidence common to the class. When the latter issue is genuinely disputed, the district court must resolve it after considering all relevant evidence.‖ Id. at 325. When plaintiffs present multiple models created by expert witnesses that can show common evidence and those models are based on data, a district court does not have to determine which model should be used at the time of class certification. Linerboard, 305 F.3d at 155. Here, the District Court likely determined that Dr. McClave‘s model could be refined between the time 47 On appeal, the inquiry narrows. Because the District Court held that Plaintiffs had established they could measure damages through common proof, we examine whether that determination was beyond the Court‘s discretion. Having identified the forest of law, we proceed to scrutinize the timber that Comcast faults as rotted. A. Comcast contends that Dr. McClave‘s model cannot isolate damages for individual theories of harm, and that it therefore cannot distinguish between lawful and unlawful competition. Comcast cites Coleman Motor Co. v. Chrysler Corp., 525 F.2d 1338, 1353 (3d Cir. 1975), and Concord Boat Corp. v. Brunswick Corp., 207 F.3d 1039, 1057 (8th Cir. 2000). In both cases, following adverse jury verdicts, the courts held that the experts‘ theories of damages were ―speculation‖—not ―just and reasonable inferences‖— because the models did not distinguish between the effects of lawful and unlawful competition. In Coleman, we quoted the guidepost of Story Parchment: ―The rule which precludes the recovery of uncertain damages applies to such as are not the certain result of the wrong, not to those damages which are definitely attributable to the wrong and only uncertain in respect of their amount.‖ Coleman, 525 F.2d at 1353 (quoting Story Parchment, 282 U.S. at 562). We are not persuaded by Comcast‘s argument. To measure damages, Dr. McClave used screens to select and when class certification was granted and trial so as to comply with Daubert. 48 average benchmark counties against which to compare the actual Philadelphia market. The screens themselves were not intended to calculate damages, but instead to construct an estimated competitive ―but-for‖ Philadelphia market (a market absent the alleged anticompetitive conduct). For example, although the screens incorporated Direct Broadcast Satellite penetration rates, those rates were included to estimate typical competitive market conditions, not to calculate liability for the foreclosure of DBS competitors.14 The model then calculates damages by comparing actual prices to the constructed ―but-for‖ market. Differences between actual prices and ―but-for‖ prices reflect anticompetitive impact. In other words, the model calculates supra-competitive prices regardless of the type of anticompetitive conduct. Further, the model uses standard econometric methodology to calculate damages. See generally Areeda et al., supra, ¶ 394 (detailing the basic steps in calculating antitrust damages). Indeed, as Dr. McClave highlighted, Comcast‘s expert Dr. Chipty employed the same methodological approach—identify a suitable benchmark and employ multiple regression analysis to control for differences—to estimate damages on a class-wide basis. App. 04041 (―Dr. Chipty and I agree that the application of 14 The Concurrence-Dissent misreads this observation as addressing the on-the-merits validity of the DBS screen. Slip Concurring-Dissenting Op. at 25-27. We address Comcast‘s contention regarding the merits of the DBS screen, however, infra Part X. This observation indicates simply that the exclusion of the DBS foreclosure theory of liability does not render Dr. McClave‘s damages methodology incapable of calculating damages on a class-wide basis if the class can prove that Comcast engaged in anticompetitive behavior. 49 multiple regression analysis to compare Philadelphia to a suitable benchmark is an appropriate methodology that can be applied on a classwide basis to quantify the amount of economic damages in this case.‖). As a result, if the class proves at trial that Comcast engaged in anticompetitive behavior, it can use the constructed ―but-for‖ market to measure the anticompetitive impact on the class members. At the class certification stage we do not require that Plaintiffs tie each theory of antitrust impact to an exact calculation of damages, but instead that they assure us that if they can prove antitrust impact, the resulting damages are capable of measurement and will not require labyrinthine individual calculations. Cf. Newton, 259 F.3d at 187. We are satisfied that Plaintiffs‘ damages model meets this burden.15 15 The Concurrence-Dissent states that Dr. McClave‘s damages theory can establish damages only in the five counties where RCN attempted to overbuild. This concern misses the central theory of Plaintiffs‘ case: by deterring the entry of overbuilders through clustering, Comcast allegedly maintained higher prices across the entire market area. Dr. McClave‘s damages model appropriately reflected a ―but-for‖ world by accounting for overbuilding only in the five counties where RCN attempted to overbuild, and his resulting calculations showed that—taking the limited actual overbuilding into account—―the Philadelphia DMA market prices were elevated above the but-for prices in every county- year combination.‖ App. 03412. Additionally, the Concurrence-Dissent apparently takes up the mantle of an additional Comcast expert and raises multiple arguments against Dr. McClave‘s damages model not addressed by 50 Additionally, the cases that Comcast offers are distinguishable on multiple grounds. Most to the point, those cases considered the merits of experts‘ theories following adverse jury verdicts; here, we address only whether Plaintiffs have provided a method to measure and quantify damages on a class-wide basis. We have not reached the stage of determining on the merits whether the methodology is a just and reasonable inference or speculative. And, to the extent Comcast worries about distinguishing between lawful and unlawful conduct, Dr. McClave‘s damages methodology does not suffer from the defects present in those cases because it constructs a competitive ―but-for‖ world that includes lawful competition, not a hypothetical one bereft of both lawful and unlawful competition. See Concord, 207 F.3d at 1056-1057 (model was ―mere speculation‖ because it ignored inconvenient evidence, failed to account for external market events, and did not incorporate economic reality of market); Coleman, 525 F.2d at 1352-1353 (model premised on hypothetical world without even lawful competition).16 Comcast‘s experts at the District Court level nor advanced by Comcast on appeal. We must limit our review to the issues presented by Appellants and Appellees. We are not permitted to embark on an intellectual adventure of our own. 16 Comcast adds that because overbuilding occurs at the franchise level, Dr. McClave‘s county-to-county metric cannot calculate damages if the jury finds that only some (if any) franchises were impacted. First, Dr. McClave indicated that franchises within counties often have identical or nearly identical pricing, which assuages Comcast‘s concern. See App. 03409. Second, Comcast is attempting again to redefine the relevant market: inasmuch as Plaintiffs have established 51 B. Comcast‘s remaining arguments contest specific parts of Dr. McClave‘s damages methodology. These contentions are a renewal of those it made to the District Court, each of which the Court rejected. For those determinations to be beyond the Court‘s discretion, Comcast must convince us that the Court‘s acceptance of the pieces of Dr. McClave‘s methodology was clearly erroneous. At the outset, we agree with the class that the heart of Comcast‘s arguments are attacks on the merits of the methodology that have no place in the class certification inquiry. Even if we were to overrule as clearly erroneous the District Court‘s findings on all four contested pieces of Dr. McClave‘s methodology—i.e., modify both of Dr. McClave‘s screens,17 add population density as a variable, and incorporate Dr. Chipty‘s proposed method for calculating that the relevant geographic market can be the Philadelphia DMA, see supra Part IV.A, their damages model passes muster at this stage of the proceedings. 17 The Concurrence-Dissent—unlike Comcast‘s experts, Comcast‘s lawyers and the District Court—identifies a ―third screen.‖ Slip Concurring-Dissenting Op. at 22. Again, this ―screen‖ was not raised by the parties before us and we do not address it (we doubt additionally that it is a screen: the two screens were used to select benchmark counties, whereas the presence of overbuilders was an identification attached to the already-selected benchmark counties for purposes of performing a multiple regression analysis, see App. 03412, 03421 (Corrected McClave Decl.)). 52 discounts—only the final amount of estimated damages would change. See App. 03082 (Hr‘g Ex.) (chart demonstrating differing damages amounts based on different model specifications, including Dr. Chipty‘s suggested specifications); App. 04557 (Dr. McClave Supplemental Decl.) (damages remain class-wide and substantial even using Dr. Chipty‘s proposed methodology, after correcting for two obvious errors). Comcast‘s assertions do not impeach the District Court‘s ultimate holding that damages are capable of common proof on a class-wide basis. See Behrend, 264 F.R.D. at 191; see also In re Scrap Metal Antitrust Litig., 527 F.3d 517, 535 (6th Cir. 2008) (―Indeed, we have never required a precise mathematical calculation of damages before deeming a class worthy of certification.‖). All of the cases Comcast proffers examine damages models on their merits following adverse jury verdicts. For reasons explained above, these cases do not address the question at the class certification stage. Because Comcast‘s contentions do not cast doubt on the District Court‘s holding that Plaintiffs will be able to measure class-wide damages through a common methodology, we decline to consider them further. See Hydrogen Peroxide, 552 F.3d at 317 (describing the Supreme Court‘s rule prohibiting consideration of the merits if not ―necessary‖ for purposes of Rule 23) (citing Eisen, 417 U.S. at 177). Plaintiffs have provided a common methodology to measure and quantify damages on a class-wide basis. The District Court acted within its discretion in so finding.18 18 The Concurrence-Dissent expresses its additional concern over using mathematical averages across the Philadelphia DMA, given the potential variation among the franchise 53 areas. Once again, this concern is notably absent from Comcast‘s briefing (except as already addressed above regarding the screens and demographic variables). Nor does the Concurrence-Dissent grapple with the abuse-of-discretion standard of review we must apply to the District Court‘s acceptance of Dr. McClave‘s damages model. We also note in passing that the Concurrence-Dissent overstates the degree of dissimilarity among the franchise areas. It recognizes that Dr. McClave‘s model examines actual prices on a county-by- county level, see Slip Concurring-Dissenting Op. at 39-40 n.36, but fails to note, as Dr. McClave explained: ―Many franchises within counties often have identical or nearly identical pricing. More price variability, and thus from an econometric perspective more information about prices and their determinants, is obtained by aggregating prices at the county level.‖ App. 03409 (Corrected McClave Decl.). Not even Comcast‘s expert contested this reasoning. See App. 03831 (Chipty Decl.); App. 03954 (Chipty Rebuttal Report). Finally, to the extent the Concurrence-Dissent questions the appropriateness of using county-level statistics to measure damages across the entire Philadelphia DMA, we observe that this question was contested strenuously and repeatedly by the experts on both sides at the District Court level. See App. 03410 (Corrected McClave Decl.) (explaining choice of market share screen); App. 03833 (Chipty Decl.) (contesting market share screen); App. 04066 (McClave Rebuttal Decl.) (defending market share screen); App. 03961 (Chipty Rebuttal Report) (disputing screen again); App. 04262 (McClave Reply Decl.) (responding to Dr. Chipty‘s criticisms of screen). After reviewing the reports and hearing careful examination of the experts on this point, the District Court found that Dr. McClave‘s 40% county-level market-share 54 XI. The District Court certified the class for resolution of four claims. Comcast contends that the District Court erred by certifying the following claim: Whether Defendants conspired with competitors, and whether Defendants entered into and implemented agreements with competitors, to allocate markets, territories, and customers for cable television services; and whether such conduct is a per se violation, or whether it constitutes a restraint of trade in violation of Section 1 of the Sherman Act, 15 U.S.C. § 1. App. 00031 (emphasis added). According to Comcast, the District Court lacked any legal authority to certify a per se claim based on the class‘s allegations. screen was ―supported by the evidence‖ and that Dr. Chipty‘s rebuttal was not supported by appropriate data. 264 F.R.D. at 184. Through a clearly erroneous lens, we may not reverse a District Court‘s factual finding if we would weigh the evidence differently; instead, the Court‘s finding must be implausible in light of the record, Anderson, 470 U.S. at 573- 574, or completely devoid of minimum evidentiary support displaying some hue of credibility, Krasnov, 465 F.2d at 1302. The Concurrence-Dissent breezes past this formidable standard of review to reach its own factual finding. 55 This is a merits issue beyond the scope of our Rule 23(f) jurisdiction. Comcast misconstrues the District Court‘s certification order. The Court certified the class and stated that one of the questions to be litigated is whether there has been a per se violation. It did not declare that a per se violation had occurred. Appeals taken pursuant to Rule 23(f) do not furnish the proper vehicle to address the merits of Plaintiffs‘ antitrust claims. See McKowan Lowe & Co. v. Jasmine, Ltd., 295 F.3d 380, 390 (3d Cir. 2002) (describing the ―scrupulous‖ limits of Rule 23(f) jurisdiction). Comcast appeals from the District Court‘s determination that questions of law or fact common to class members predominate, which was the only issue before the District Court. See App. 00029 (District Ct. Certification Order) (―The only class certification element that remained in dispute was the requirement of Fed. R. Civ. P. 23(b)(2) that common issues of law and fact predominate.‖). Comcast itself stipulated as much. See App. 00436 (Comcast Letter to the District Ct., Mar. 25, 2009) (―With respect to the issues to be addressed in a new class certification motion, Comcast is prepared to stipulate that the only issues to be resolved are those of antitrust impact and methodology of damages . . . .‖). Comcast‘s request to have us declare on the merits that Plaintiffs cannot establish a per se antitrust violation is beyond the scope of the certification decision from which Comcast appeals pursuant to Rule 23(f). Accordingly, we do not reach this contention. ***** We have considered carefully all the contentions presented by the parties. Plaintiffs have demonstrated that this case can proceed as a class action. Comcast has not carried its burden to convince us otherwise. Accordingly, we will 56 AFFIRM in all respects the District Court‘s Order certifying the class. 57 Behrend, et al. v. Comcast Corporation, et al., No. 10-2865 JORDAN, Circuit Judge, concurring in the judgment part and dissenting in part I agree with the Majority‟s conclusion, though not its reasoning, with respect to the question of antitrust impact, and I therefore join in holding that the District Court did not abuse its discretion when it determined that Plaintiffs could establish antitrust impact through evidence common to a class comprising Comcast cable television customers in the Philadelphia DMA.1 But because I conclude that damages cannot be proven using evidence common to that entire class, I would vacate the certification order to the extent it provides for a single class as to proof of damages, and I would remand the case to the District Court to consider whether the class can be divided into subclasses for the purpose of proving damages. I therefore respectfully dissent in part.2 1 I adopt the defined terms, such as “DMA,” as used in the Majority opinion. 2 Although the Majority opinion decides the question of certification for a single class comprising Comcast customers in the Philadelphia DMA, it should be noted that its decision will become a template for resolving similar class certification questions pending in cases involving the Chicago and Boston media markets (see Slip Op. at 10 & n.5), and in all likelihood it will be cited in other lawsuits against cable television service providers (cf. App. at 3652 (Williams Dec.) (explaining that, as part of Comcast‟s swaps and acquisitions, “Adelphia received Comcast‟s cable systems and subscribers 1 As the Majority explains, Plaintiffs‟ claims have three elements, (1) an antitrust violation, (2) antitrust impact, and (3) damages (see Slip Op. at 17 (citing In re Hydrogen Peroxide Antitrust Litig., 552 F.3d 305, 311 (3d Cir. 2008)).3 In pursuing its motion to decertify the initial class, however, Comcast effectively conceded that there was predominance with respect to the element of an antitrust violation, stipulating that it was contesting only “the Rule 23(b) issues of predominance of the common issues of (1) antitrust impact and (2) methodology of damages.” (App. at 438.) When the located in Palm Beach, Florida and Los Angeles, California”)). Thus, the problems in the Majority‟s reasoning will have practical repercussions far beyond this case. I therefore write not only because I cannot join the Majority in permitting Plaintiffs to pursue damages on a class-wide basis, but also to provide a counterpoint to the Majority‟s analysis for future consideration. 3 Plaintiffs make separate claims for violation of both § 1 and § 2 of the Sherman Act, but each of those claims contains the three elements described above, with only the nature of the particular antitrust violation differing. Compare Hydrogen Peroxide, 552 F.3d at 311 (listing the elements of a § 1 claim as “(1) a violation of the antitrust laws – here, § 1 of the Sherman Act, (2) individual injury resulting from that violation, and (3) measurable damages”), with Am. Bearing Co. v. Litton Indus., 729 F.2d 943, 948 (3d Cir. 1984) (listing the elements of a § 2 claim as “(1) an antitrust violation, in this case a violation of section 2 of the Sherman Act; (2) fact of damage or injury; and (3) measurable damages”). 2 District Court granted Comcast‟s motion,4 it accepted that stipulation and instructed the parties that, moving forward, they “need only address these discrete issues.” (Id.) On appeal, after the District Court once more certified a class, Comcast has again limited its arguments to addressing predominance as to impact and damages. We are therefore faced with two related questions: First, whether the District Court abused its discretion by holding that, as required by Federal Rule of Civil Procedure 23(b)(3), common issues of law or fact predominate with respect to the question of antitrust impact, and, second, whether the District Court abused its discretion by likewise holding that common issues of law or fact predominate with respect to the question of damages.5 4 Because Comcast had moved to decertify the class entirely before stipulating to all issues other than the predominance questions described above, the District Court, which construed the motion to decertify as a motion for reconsideration, granted the motion only with respect to those predominance issues and denied it with respect to all other issues. (App. at 437.) 5 While not expressed, the requirement that there must be predominance with respect to both antitrust impact and damages appears to be accepted by the parties and the Majority, and I likewise accept that predominance is issue specific. See, e.g., Hydrogen Peroxide, 552 F.3d 305 at 311 (“We examine the elements of plaintiffs‟ claim through the prism of Rule 23,” to determine whether “proof of the essential elements of the cause of action requires individual treatment.” (internal quotation marks omitted)); Newton v. Merrill Lynch, Pierce, Fenner & Smith, Inc., 259 F.3d 154, 3 The Majority opinion skillfully lays out the legal requirements for predominance and the standard under which we must review the District Court‟s decision, and there is no need to repeat that legal background. I emphasize, however, the instruction from Hydrogen Peroxide that the question of predominance hinges on whether the elements of a class claim are “capable of proof at trial through evidence that is common to the class rather than individual to its members.” 552 F.3d at 311-12. With that requirement in mind, I address the contested elements in turn. 172 (3d Cir. 2001) (“To determine whether the claims alleged by the putative class meet the requirements for class certification, we must first examine the underlying cause of action … . If proof of the essential elements of the cause of action requires individual treatment, then class certification is unsuitable.”) Of course, where only some elements of a claim require individual treatment, while others can be litigated collectively, it may be appropriate to certify a class for those elements that can be treated collectively, while certifying subclasses or requiring individual treatment for those that cannot. See, e.g., FED. R. CIV. P. 23(b)(4), advisory committee‟s notes (explaining that application of Rule 23(c)(4)‟s provision allowing “that an action may be maintained as a class action as to particular issues only” may be appropriate where, for instance, liability can be proven class wide, but damages cannot); In re Nassau Cnty. Strip Search Cases, 461 F.3d 219, 231 (2d Cir. 2006) (remanding to district court with instructions to certify a class for liability and to consider whether to also certify for damages or to, alternatively, certify subclasses for damages). 4 I. Whether Antitrust Impact Can Be Proven Using Evidence Common To The Class In seeking class certification, Plaintiffs initially presented four theories of antitrust impact.6 The District Court rejected three of them,7 leaving Plaintiffs with only a single theory of antitrust impact: that Comcast‟s clustering 6 Those theories were: (1) that Comcast‟s high market share resulting from clustering made it profitable for Comcast to deny Comcast SportsNet to DBS providers, which lowered DBS penetration rates and allowed Comcast to raise prices; (2) that Comcast‟s clustering reduced “benchmark competition” (the ability of customers to compare service and prices among competing providers), which allowed Comcast to raise prices; (3) that Comcast‟s market power increased its bargaining power vis-à-vis content providers, which allowed it to raise prices for its services; and (4) that Comcast‟s clustering deterred competition from overbuilders, allowing Comcast to raise prices. 7 The District Court rejected the theory that clustering reduced DBS penetration because it found that Comcast‟s denial of Comcast SportsNet to DBS providers predated and was unrelated to clustering. It rejected the theory that clustering reduced benchmark competition because Plaintiffs had provided no evidence that television consumers actually engaged in benchmark competition. It rejected as “wholly unsupported” the theory that increased bargaining power vis- à-vis content providers increased prices. 5 reduced overbuilding8 and, therefore, increased prices. Like the Majority, I see no abuse of discretion in the District Court‟s holding that antitrust impact may be proven using evidence that clustering reduced overbuilding and so caused increased prices. Thus, I agree with my colleagues in the Majority that the element of antitrust impact is at least capable of proof on behalf of some class of consumers. The more complicated question, as I see it, is whether antitrust impact is capable of proof for a class encompassing all Comcast customers in the Philadelphia DMA, through the use of common evidence. 9 On that issue too I agree with the Majority‟s holding that the District Court was within its discretion to conclude that the Philadelphia DMA is the 8 “Overbuilding,” as the Majority explained, is where a second cable provider – the “overbuilder” – “builds and offers customers a competitive alternative where a telecommunications company already operates.” (Slip Op. at 9.) The existing provider is often referred to as the “incumbent” provider. 9 The geographic scope of the class is actually defined as Comcast‟s Philadelphia cluster, which, as noted by the Majority, excludes the DMA counties of Lehigh and Northampton. As Dr. Chipty explains, those are the two counties in which Comcast has no presence (see App. at 3795 & n.12 (Chipty Reply Dec.)), and, therefore, they would be excluded from the class regardless of its geographic scope. For ease of reference, I refer to the class as encompassing the Philadelphia DMA, rather than the Philadelphia Cluster, recognizing that those DMA counties in which there are no Comcast customers are not included in the class. 6 appropriate geographic region within which antitrust impact can be proven with common evidence. I do not agree, however, with the Majority‟s reasoning in support of that conclusion. Much confusion has been caused in this case by the conflation of two distinct concepts: the antitrust concept of “relevant geographic market,” which has traditionally been defined as the smallest area within which a monopolist can exercise market power,10 and the class action concept of a “class definition,” which gives the parameters of a set of plaintiffs as to whom the elements of a claim can be proven using common evidence.11 Because, in this case, the class 10 For example, the Federal Trade Commission defines “relevant geographic market” as the region in which a hypothetical monopolist “would impose at least a [small but significant nontransitory price increase] on some customers in that region” without “this price increase [being] defeated by substitution away from the relevant product or by … customers in the region travelling outside it to purchase the relevant product.” FEDERAL TRADE COMMISSION, HORIZONTAL MERGER GUIDELINES 14-15 (2010). Cf. PHILLIP E. AREEDA & HERBERT HOVENKAMP, FUNDAMENTALS OF ANTITRUST LAW § 5-30 (2010) (“[T]he relevant inquiry” for identifying a geographic market is “how far [customers] are willing to travel in order to avoid paying the defendant monopoly prices.”). 11 See In re DVI, Inc. Sec. Litig., 639 F.3d 623, 639 n.22 (3d Cir. 2011) (explaining that, pursuant to Rule 23(c)(1)(B), the class definition describes both “which individuals and entities are included” and the “claims, issues 7 definition includes a geographic component, the term “relevant geographic market” has been used equivocally by the parties, the District Court, and the Majority to describe both the area affected by antitrust impact and the area within which potential class members reside – the latter area being what I will call, for lack of a better term, the “class region.”12 The problem with that equivocal usage is that the relevant geographic market and the class region are not necessarily coterminous. Even if we assume that, within the Philadelphia DMA, there are many distinct geographic markets that are relevant for antitrust purposes, as Comcast argues, that does not mean that Plaintiffs cannot prove, by common evidence, or defenses to be treated on a class basis”). While Rule 23(c)(1)(B) does not expressly state that the class should include only those for whom the defined claims can be proven by common evidence, it is apparent that any class must be defined in a manner consistent with all Rule 23 requirements, including commonality and predominance. Cf. id. at 639 & n.22 (explaining that the question of whether there was predominance when it was alleged that some members of a proposed class “would be unable to demonstrate loss causation,” was an issue of “which individuals and entities are included in the putative class … primarily relevant to class definition”). 12 The class region will not necessarily be the same with respect to each element of a class‟s claims. In fact, even in this case, the class region differs with respect to antitrust impact and damages because, for the reasons I identify infra Part II(B), antitrust impact can be proven using common evidence across a wider region than damages can be. 8 that Comcast‟s acts caused antitrust impact within all of them. As a theoretical matter, class proof can cover multiple relevant geographic markets, and, indeed, other Courts of Appeals have so held. See, e.g., In re Sugar Antitrust Litig., 559 F.2d 481, 483-84 (9th Cir. 1977) (rejecting the argument that a class could not be certified “where the antitrust claims involve a variety of geographic and product markets”); Windham v. Am. Brands Inc., 539 F.2d 1016, 1018 (4th Cir. 1976) (holding that a district court abused its discretion in refusing to certify an antitrust class that encompassed “11 different geographic markets”). While the relevant geographic market and the class region are conceptually distinct,13 the Majority, like the District Court, initially attempts to identify the class region in terms of the relevant geographic market. Unlike the District Court, however, the Majority decides that because “[d]efining the relevant geographic market … is an issue of the merits,” the question of the relevant geographic market is “not properly before us.” (Slip Op. at 20-21.) The Majority is correct that defining the relevant geographic market is not a task we need to undertake at this stage, but that is not because the task takes us into the merits. It is rather because, regardless of whether there are one or many relevant geographic markets associated with the 13 That is not to say that a class region and a relevant geographic market will always be different. An antitrust violation may often affect people in only a single geographic market, in which case the relevant geographic market and the class region would be in essence the same. 9 Philadelphia DMA, the question before us at this juncture is whether there is some class, in this case defined geographically, that can be shown, through common evidence, to have experienced elevated prices as a result of reduced overbuilding because of Comcast‟s clustering. Should that region include only those franchise areas involved in the Cable System Transactions?14 Should it include only those franchise areas in which RCN was licensed to overbuild, but did not? Should it encompass the Philadelphia DMA or some lesser or greater area? The Majority does not ask those questions, but, instead, after determining that Plaintiffs can attempt to prove that the relevant geographic market is the Philadelphia DMA, the Majority assumes that that also means that the class is properly defined to cover the Philadelphia DMA and, therefore, that Plaintiffs can prove by common evidence that clustering reduced overbuilding and increased prices throughout the DMA. (See, e.g., Slip Op. at 51-52 n.16 (dismissing Comcast‟s argument that overbuilding should be analyzed at the franchise level because “Plaintiffs have established that the relevant geographic market can be the Philadelphia DMA”).) Fortunately, what the Majority assumes, namely that the Philadelphia DMA is the 14 The Cable System Transactions are, as described by the Majority, the transactions through which Comcast “clustered” its franchise areas by “contract[ing] with competing cable providers to either acquire them or to „swap‟ cable systems it owned in areas outside the Philadelphia DMA for cable systems within the Philadelphia DMA.” (Slip. Op. at 5.) 10 appropriate class region for proving antitrust impact, is supportable. A compelling argument could be made that the class should consist only of those people living in franchise areas where RCN was licensed to overbuild, because only those franchise areas that would otherwise have been overbuilt could have been affected by the elimination of that overbuilding.15 Because RCN was licensed to overbuild only five of the eighteen Philadelphia DMA counties (see, e.g., 15 At least, Plaintiffs have provided no evidence that persons outside of franchise areas that would otherwise have been overbuilt can be affected by the elimination of that overbuilding. Dr. Williams opines that, where some parts of a franchise area are overbuilt, the overbuilding can affect prices in other parts of that same franchise area that are not overbuilt. (App. at 3704-14 (Williams Dec.) (explaining that where competing cable companies have “alternating franchise areas,” overbuilding by one company into portions of the competitor‟s adjacent franchise area can affect prices in the portion of the overbuilt franchise area “that remain monopolized”).) As a theoretical matter, it is also plausible that, when one franchise area has been overbuilt, the threat of further expansion by that overbuilder could put downward pressure on prices in nearby franchise areas. If such an effect is described in the multitude of expert opinions, however, the parties have not identified it. Moreover, even if there is such an effect, it would likely be attenuated by distance. It seems doubtful that overbuilding in, for instance, Bucks County, Pennsylvania would influence prices in Kent County, Delaware. 11 App. at 3640 (Williams Dec.); App. at 4284-85 (Singer Reply Dec.)), that would suggest limiting the class region to those five counties.16 Nonetheless, both Dr. Williams and Dr. Singer opined that, had RCN successfully overbuilt the five counties in which it was already licensed, it would have continued overbuilding into the remainder of the Philadelphia DMA. (App. at 4285 (Singer Reply Dec.) (“[H]ad RCN entered the five counties that it intended to … it is likely that RCN would have expanded its footprint beyond those five counties into geographically contiguous areas throughout the Philadelphia DMA.”); App. at 4306 (Williams Reply Dec.) (“RCN likely would have continued to pursue its strategy of building into other areas in the Philadelphia DMA adjacent to its existing cable infrastructure, beyond the five counties.”).) The District Court relied on those statements in holding that Plaintiffs had shown that the anticompetitive effects of clustering could be proven throughout the Philadelphia DMA. Behrend v. Comcast Corp., 264 F.R.D. 150, 174-75 (E.D. Pa. 2010). Though one may be skeptical that RCN would have overbuilt even the five counties in which it was licensed, let alone the remainder of the Philadelphia DMA, it was not clearly erroneous for the District Court to accept that the prospect of overbuilding throughout the DMA was capable of proof. Consequently, it was not an abuse of discretion for the District Court to hold that Plaintiffs could show, by common evidence, the antitrust impact of clustering throughout the 16 Given that the Class‟s whole theory is rooted in the premise that Comcast‟s clustering deterred overbuilding, it is no small matter that RCN – the only entity licensed to overbuild anywhere in the Philadelphia DMA – was licensed to overbuild in just five of the eighteen counties. 12 Philadelphia DMA. Accordingly, while I do not agree with the Majority‟s reasoning, I agree that the District Court was within its discretion in determining that an appropriate class region for proving antitrust impact is the Philadelphia DMA.17 17 The Majority responds to my efforts to identify the class region by stating that I have “misunderst[ood] an important distinction,” namely that Plaintiffs have identified a “„class region‟ … of a „Philadelphia cluster‟ which is distinct from the contested relevant geographic market of the “„Philadelphia DMA.‟” (Slip Op. at 26 n.8.) I do acknowledge that distinction. However, that does not speak to the point because, in spite of that distinction, there remains an equivocal use of the term “relevant geographic market.” That equivocation is evidenced by the Majority‟s statement – in response to Comcast‟s suggestion that franchise areas might be the appropriate class region for damages – that “Comcast is attempting to redefine the relevant market: inasmuch as Plaintiffs have established that the relevant geographic market can be the Philadelphia DMA … their damages model passes muster.” (Slip. Op. at 51-52 n.16.) The Majority also asserts that there is no question about the class region because Comcast does not dispute the class region but disputes only the relevant geographic market. (Slip Op. at 26 n.8) That is not correct. While Comcast does not use the term “class region,” Comcast and its experts plainly argue that the scope of the class is too broad, and they dispute the District Court‟s conclusion that antitrust impact can be proven by common evidence across the Philadelphia DMA. (See, e.g, App. at 3923 (Teece Reply Dec.) (“[E]ven if RCN would have overbuilt all five counties entirely in the 13 but-for world, this would not be sufficient to conclude that the impact of the challenged conduct would have affected all Comcast customers in the Philadelphia DMA.”); id. at 3922 (“I have seen no evidence that RCN ever intended to build out the entire Philadelphia DMA.”); Appellants Br. at 33 (arguing that Dr. Williams‟s models do not show that clustering “deterred overbuilding … in a manner affecting all class members”); id. at 24-25 (noting that RCN was licensed in only five counties and arguing that Plaintiffs cannot prove that RCN would have entered the Philadelphia DMA). While I do not agree with Comcast‟s effort to define the class region by reference to the relevant geographic market (any more than I agree with the Majority‟s conflating of those concepts), to say that Comcast does not dispute the contours of the class region is not accurate, as the foregoing citations indicate. However, even if Comcast had not disputed the class region, it would still be appropriate for us to address it. The Majority faults me for, in its view, addressing problems not raised by Comcast, which the Majority asserts are, therefore, waived. (See, e.g., slip op. at 43-44 n.15 (“[T]he Concurrence-Dissent … raises multiple arguments … not addressed by Comcast‟s expert … . We must limit our review to the issues presented by Appellants and Appellees.”). But “there can be no waiver … of the Judge‟s duty to apply the correct legal standard. … This is particularly true in the class action context, where „the district court acts as a fiduciary who must serve as a guardian of the rights of absent class members.‟” In re Cmty. Bank of N. Virginia, 622 F.3d 275, 302 n.20 (3d Cir. 2010) (quoting United States v. Ali, 508 F.3d 136, 144 n.9 (3d Cir. 2007) and In re General Motors Corp. Pick-Up Truck Fuel Tank Prods Liab. Litig., 55 14 F.3d 768, 785 (3d Cir. 1995)). Thus, where Comcast has raised the issues of whether there is predominance with respect to antitrust impact and damages, we are required to “apply the correct legal standard,” – which is to determine whether those elements can, in fact, be proven using evidence common to the class – even if that requires us “„to conduct [our] own thorough [R]ule 23[b] inquiry.‟” Id. (quoting Sitrman v. Exxon Corp., 280 F.3d 554, 563 n.7 (5th Cir. 2002)). By disregarding the problems I have endeavored to identify, the result is an overly broad class definition and, to the extent any legitimate claims are proven, a likely dilution of recovery. Our fiduciary responsibility to absent class members requires that we ensure compliance with the provisions of Rule 23, especially those “„designed to protect absentees by blocking unwarranted or overbroad class definitions.‟” Id. at 291 (quoting Amchem Prods., Inc. v. Windsor, 521 U.S. 591, 620 (1997)); cf. Tri-M Group, LLC v. Sharp, 638 F.3d 406, 416 (3d Cir. 2011) (“[T]he waiver principle is only a rule of practice and may be relaxed whenever the public interest or justice so warrants.”). Moreover, we must be cognizant of “the pivotal status of class certification in large-scale litigation,” which is “often the defining moment in class actions (for it may sound the „death knell‟ of the litigation on the part of plaintiffs, or create unwarranted pressure to settle nonmeritorious claims on the part of defendants).” Hydrogen Peroxide, 552 F.3d at 310 (internal quotation marks omitted). Pointing out analytical problems central to the certification question is no frolic and detour. It is our obligation. 15 II. Whether Damages Can Be Proven Using Evidence Common To The Class I part ways with the Majority entirely, however, when it comes to class-wide proof of damages. The only evidence supporting Plaintiffs‟ claim that damages can be proven using evidence common to the class is the expert opinion of Dr. McClave. But, as detailed hereafter, Dr. McClave‟s testimony is incapable of identifying any damages caused by reduced overbuilding in the Philadelphia DMA. Consequently, his testimony is irrelevant and should be inadmissible at trial, pursuant to Federal Rule of Evidence 702 and Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993), as lacking fit. Thus, it cannot constitute common evidence of damages.18 18 Although we have never explicitly held that expert testimony must satisfy Daubert at the class certification stage, it is implicit in both Supreme Court precedent and our precedent. In Wal-Mart Stores, Inc. v. Dukes, the Supreme Court recently expressed its “doubt” about a district court‟s conclusion that “Daubert did not apply to expert testimony at the certification stage of class-action proceedings.” 131 S. Ct. 2541, 2553-54 (2011). In Hydrogen Peroxide, we explained that “opinion testimony should not be uncritically accepted as establishing a Rule 23 requirement merely because the court holds the testimony should not be excluded, under Daubert or for any other reason.” 552 F.3d at 323. Inherent in that statement is the conclusion that a court could, at the class certification stage, exclude expert testimony under Daubert. 16 Even without the guidance of Dukes and Hydrogen Peroxide, simple logic indicates that a court may consider the admissibility of expert testimony at least when considering predominance. A court should be hard pressed to conclude that the elements of a claim are capable of proof through evidence common to a class if the only evidence proffered would not be admissible as proof of anything. I recognize, of course, that in neither the District Court nor before us did Comcast describe its challenge to certification as a challenge to the admissibility of Dr. McClave‟s testimony. Nonetheless, while it did not use the language of Daubert, the substance of Comcast‟s challenge was that Dr. McClave‟s damages testimony was irrelevant and, therefore, did not fit the case. (See, e.g., Appellants‟ Br. at 37 (“Dr. McClave admitted that his damages model takes all of the anticompetitive effects of all of the complained-of conduct as a whole, and therefore cannot isolate damages attributable to specific conduct or effects.”); id. at 42 (“Dr. McClave‟s DBS penetration screen is substantively invalid because it bears no relation to the competitive conditions that would have prevailed in the Philadelphia region.”); id. at 43 (“Dr. McClave‟s „market share‟ screen is likewise invalid because it bears no relation to the competitive conditions that would have prevailed in the Philadelphia region.”). The Majority protests my invocation of Daubert, but, regardless of whether we frame the issue as a question of fit under Daubert or simply ask whether the District Court abused its discretion by relying on irrelevant evidence, we are effectively asking the same question. I have chosen the terminology of Daubert because it is particularly apt for describing the difficulty created by the change in Plaintiffs‟ 17 theory of impact and the consequent disconnect between that altered theory and Dr. McClave‟s expert report. The short of it is, Dr. McClave‟s model no longer fits the case. This observation is not, as the Majority fears, either an invitation or a demand for mini-trials in conjunction with class certification motions. I note here as well my disagreement with the Majority‟s claim that, at the class certification stage, we need only “evaluate expert models to determine whether the theory of proof is plausible.” (Slip Op. at 47 n.13.) The Majority supports that position by quoting Hydrogen Peroxide’s statement that “„if such impact is plausible in theory, it is also susceptible to proof at trial through available evidence common to the class.‟” (Id. (quoting Hydrogen Peroxide, 552 F.3d at 325).) That quotation is better understood, however, if one includes the first half of the quoted sentence, which states that “the question at class certification is whether, if such impact is plausible in theory, it is also susceptible to proof at trial through available evidence common to the class.” 552 F.3d at 325 (emphasis added). Thus, Hydrogen Peroxide does not suggest that we need only “evaluate expert models to determine whether the theory of proof is plausible,” as the Majority claims. To the contrary, Hydrogen Peroxide instructs that, even where a theory is plausible, “the question at class certification is whether” that plausible theory is susceptible to common proof. Id. If the only common proof offered is inadmissible expert testimony, then Plaintiffs have not met their burden of showing that the theory – plausible or not – is capable of common proof. 18 Our precedent explains that Rule 702 and Daubert impose three requirements for admission of expert testimony: the expert must be qualified, the expert‟s methodology must be reliable, and the expert‟s proffered testimony must fit the particular case. See United States v. Schiff, 602 F.3d 152, 172 (3d Cir. 2010). Testimony fits when it “„is sufficiently tied to the facts of the case that it will aid the jury in resolving a factual dispute.‟” Id. at 173 (quoting United States v. Downing, 753 F.2d 1224, 1242 (3d Cir. 1985)). Like any relevancy determination, the question of fit is reviewed for abuse of discretion. United States v. Ford, 481 F.3d 215, 217-18. Here, Dr. McClave‟s opinion fails the requirement of “fit” because it is disconnected from Plaintiffs‟ only viable theory of antitrust impact, i.e., reduced overbuilding, and, thus, the proffered expert testimony cannot help the jury determine whether reduced overbuilding caused damages.19 It was, consequently, an abuse of discretion for the District Court to consider Dr. McClave‟s opinion as demonstrating that damages could be proven using evidence common to the class. As explained by the Majority, Dr. McClave arrived at his damages calculation by comparing actual cable prices in the Philadelphia DMA to prices in benchmark counties outside the Philadelphia DMA. By making those comparisons, Dr. McClave sought to identify the “but for” price of cable – that is the price that would have prevailed in the Philadelphia DMA but for the alleged anticompetitive 19 I need not, and do not, question whether Dr. McClave is qualified as an expert or whether his methodology is reliable. 19 conduct of Comcast. (App. at 3407 (McClave Dec.).) For that comparison to be relevant, however, Dr. McClave‟s benchmark counties must reflect the conditions that would have prevailed in the Philadelphia DMA in the absence of any impact from that conduct. (Cf. App. at 719 (McClave Cross) (stating that the goal of his bechmarking model was to identify “counties that reflect characteristics that one would find absent … the effects of [Comcast‟s alleged anticompetitive] conduct”).) And because the only surviving theory of antitrust impact is that clustering reduced overbuilding, for Dr. McClave‟s comparison to be relevant, his benchmark counties must reflect the conditions that would have prevailed in the Philadelphia DMA but for the alleged reduction in overbuilding. In all respects unrelated to reduced overbuilding, the benchmark counties should reflect the actual conditions in the Philadelphia DMA, or else the model will identify “damages” that are not the result of reduced overbuilding, or, in other words, that “are not the certain result of the wrong.” Story Parchment Co. v. Paterson Parchment Paper Co., 282 U.S. 555, 562 (1931); see also, e.g., Coleman Motor Co. v. Chrysler Corp., 525 F.2d 1338, 1353 (3d Cir. 1975) (“„The rule which precludes the recovery of uncertain damages applies to such as are not the certain result of the wrong.‟” (quoting Story Parchment, 282 U.S. at 562)); Amerinet, Inc. v. Xerox Corp., 972 F.2d 1483, 1494 (8th Cir. 1992) (same); Broan Mfg. Co. v. Associated Distrib., Inc., 923 F.2d 1232, 1235 (6th Cir. 1991) (same). Dr. McClave‟s benchmark counties fail in that regard because he formulated his model at a time when Plaintiffs had four separate theories of antitrust impact, and so he did not select his benchmark counties to isolate the impact of reduced overbuilding. He chose them, as one would expect, to reflect 20 the impact of other conditions in addition to reduced overbuilding. Consequently, as described in greater detail below, once the District Court rejected Plaintiffs‟ other theories of antitrust impact – leaving only the reduced- overbuilding theory – Dr. McClave‟s model no longer fits Plaintiffs‟ sole theory of antitrust impact and, instead, produces damages calculations that “are not the certain result of the wrong.” Story Parchment, 282 U.S. at 562.20 A. Dr. McClave’s Benchmark Counties Do Not Reflect “But For” Conditions in the Philadelphia DMA To identify his benchmark counties, Dr. McClave used three “screens.” First, he screened for counties where Comcast‟s market share was “less than 40%,” because that figure identified “markets where Comcast is likely to have less market power than it has acquired in the Philadelphia market.” (App. at 3410 (McClave Dec.).) Second, he screened for counties where DBS penetration21 was “at or above the national average” because “DBS … penetration 20 Whether Dr. McClave‟s opinion would have fit had the District Court allowed Plaintiffs to pursue all four of their theories of antitrust impact is irrelevant at this point. 21 As noted by the Majority, “DBS” stands for “direct broadcast satellite” television service. (Slip Op. at 27.) Dr. McClave actually used penetration rates for all alternative delivery systems (“ADS”), rather than just DBS systems. He opined, however, and the parties seem to agree, that “ADS is a proxy for DBS penetration rates.” (App. at 3410.) 21 was allegedly constrained by the anticompetitive behavior of Comcast.” (Id.) Third, having identified counties in which Comcast‟s share was less than 40 percent and DBS penetration was above the national average, Dr. McClave screened for overbuilding, identifying “each of the benchmark counties … as either overbuilt or not overbuilt.” (App. at 3411-12 (McClave Dec.).) While those screens might, if properly employed, have helped identify relevant benchmark counties in a case involving antitrust impacts beyond limited overbuilding, they fail to identify the “but for” conditions that are relevant to what is now the only impact of Comcast‟s allegedly anticompetitive conduct, namely the deterrence of overbuilding. They, therefore, cannot help identify damages caused by that impact. I examine the screens in reverse order. 1. The Overbuilt Counties Screen While there are several problems in Dr. McClave‟s opinion that reflect the lack of fit, nothing demonstrates it with more certainty than this: For thirteen of the eighteen counties in the Philadelphia DMA, Dr. McClave‟s opinion does not even attempt to show that there were elevated prices resulting from reduced overbuilding. In fact, he assumes that there was no such effect. As noted above, after identifying his benchmark counties using the market share and DBS penetration screens, Dr. McClave used a third screen to divide those counties into two groups, identifying “each of the benchmark counties … 22 as either overbuilt or not overbuilt.”22 (App. at 3411-12 (McClave Dec.).) Having done so, Dr. McClave estimated “but for” competitive prices, by comparing, on a county by county basis, prices in the eighteen actual Philadelphia DMA counties to prices in either the “overbuilt” or “not overbuilt” benchmark counties, and – crucially – he did so “assum[ing] that only the five counties that RCN indicated it planned to enter as an overbuilder would have been overbuilt.” (App. at 3412 (McClave Dec.).) At the outset, therefore, it is clear that Dr. McClave assumed that elevated prices resulting from 22 The Majority notes that the overbuilding screen is not mentioned by the parties or the District Court. (Slip Op. at 52 n.17.) While it is true that the parties do not use the terminology “overbuilding screen,” the District Court did indeed describe the concept to which I have given that label. See Behrend, 264 F.R.D. at 182 (“Once a county qualified as a benchmark for a particular year by satisfying [the DBS penetration and market share screens], it was examined to determine whether or not it had been significantly overbuilt.”). Whether one uses the “screen” terminology is not what is important. Dr. McClave, in fact, does not describe any of the benchmarking criteria as “screens,” which is a term that appears to have been only later applied to his methods. Regardless of the terminology, the fact remains that Dr. McClave did screen for overbuilding in an attempt to account for elevated prices resulting from reduced overbuilding. Thus, that screen cannot be ignored in any “rigorous analysis,” Hydrogen Peroxide, 552 F.3d at 318, of whether damages resulting from reduced overbuilding can be proven by common evidence. 23 reduced overbuilding would be present in only five of the eighteen Philadelphia counties. Dr. McClave then explained that, after making his calculations, “the overbuilt factor indicate[d] lower prices [in his model] in counties where the overbuilding factor [was] present.” (App. at 3422 (McClave Dec.) (emphasis added).) Thus, Dr. McClave‟s model assumes that elevated prices from reduced overbuilding could be present only in the five counties “that RCN indicated it planned to enter,” and the model did, in fact, identify elevated prices from reduced overbuilding only in those counties. (App. at 3412, 22 (McClave Dec.).) For the remaining counties, while there may be some uncertainty as to what exactly caused any elevated prices, this much is certain: the elevated prices identified by Dr. McClave in those thirteen counties were, according to Dr. McClave himself, the result of something other than reduced overbuilding. Consequently, any “damages” identified by Dr. McClave with respect to those thirteen counties are “uncertain damages … [that] are not the certain result of [reduced overbuilding],” and “may be substantially attributable to lawful competition.” Coleman Motor, 525 F.2d at 1353 (quoting Story Parchment, 282 U.S. at 562). Because Plaintiffs have been limited by the District Court to an overbuilding theory of antitrust impact, any price elevation resulting from a source other than reduced overbuilding is simply irrelevant. Thus, not only have Plaintiffs failed to show that damages can be proven using evidence common to the class, they have failed to show, for thirteen counties in the Philadelphia DMA, that damages can be proven using any evidence whatsoever – common or otherwise. Perhaps, in those other counties, there is a way to show damages resulting from reduced overbuilding, but, if so, 24 Plaintiffs have not identified it. As the burden lies with Plaintiffs to establish predominance, that alone should compel us to vacate the District Court‟s certification order with respect to class-wide proof of damages.23 2. The DBS Penetration Screen Dr. McClave screened for counties where DBS penetration was at or above the national average because “DBS … penetration was allegedly constrained by the 23 The Majority states that, in criticizing Dr. McClave‟s model for identifying overbuilding damages in only five counties, I have “misse[d] the central theory of Plaintiffs‟ case: by deterring the entry of overbuilders through clustering, Comcast allegedly maintained higher prices across the entire market area.” (Slip Op. at 50-51 n.15.) This misperceives my reasoning. I understand Plaintiffs‟ theory but have pointed out that the theory, as altered by the District Court‟s ruling, no longer matches Dr. McClave‟s opinion. More precisely, Plaintiffs‟ claim is that by reducing overbuilding “Comcast allegedly maintained higher prices across the entire market area,” (id.) whereas Dr. McClave attempts to show that, by reducing overbuilding, Comcast maintained higher prices in only the “five counties that RCN indicated it planned to enter as an overbuilder,” (App. at 3412 (McClave Dec.)). The Majority notes that this particular problem with Dr. McClave‟s damages theory was not identified by Comcast, but we ought note overlook significant problems with the class certification simply because they are ones we have identified rather than ones to which our attention has been directed. 25 anticompetitive behavior of Comcast.” (App. at 3410 (McClave Dec.).) Using that screen would have been appropriate if, as Plaintiffs originally argued and as Dr. McClave was originally informed, DBS penetration had been constrained by Comcast‟s anticompetitive conduct. But, as the District Court explicitly held, Plaintiffs failed to tie “Comcast‟s clustering activity in the Philadelphia DMA to reduced DBS penetration.” Behrend, 264 F.R.D. at 165. Consequently, there is no evidence in the record suggesting that DBS penetration in the Philadelphia DMA was in any way affected by Comcast‟s allegedly anticompetitive conduct. Rather, the District Court found that, while DBS penetration in Philadelphia was well below the national average, the cause of that reduced penetration – Comcast‟s refusal to distribute Comcast SportsNet through DBS providers – “occurred prior to the class period,” is “unrelated to clustering,” is “based upon valid business considerations” and is “specifically permitted” by the FCC. Id. Therefore, while DBS penetration in the Philadelphia DMA is below the national average, the cause of that reduced rate predated and is unrelated to Comcast‟s clustering and, thus, even in the absence of Comcast‟s allegedly anticompetitive conduct, DBS penetration in the Philadelphia DMA would be no different than the below average rate that has actually prevailed. As a result, any benchmark county used to identify “but for” conditions should use the actual DBS penetration rate from the Philadelphia DMA. Dr. McClave, nonetheless, used the much higher national average 26 rate,24 which identified benchmark counties in which cable prices were lower than in counties having DBS penetration similar to that in the Philadelphia DMA.25 Because Dr. McClave then calculated damages by comparing prices in those benchmark counties (with national average DBS penetration and, therefore, lower prices) to actual prices in the Philadelphia counties (with below national average DBS penetration and, therefore, higher prices), at least a portion of Dr. McClave‟s damages calculation results from the Philadelphia DMA having below national average DBS penetration. Since the cause of the below national average DBS penetration in the Philadelphia DMA is “unrelated to clustering,” is “based upon valid business considerations,” and is “specifically permitted” by the FCC, id., that reduced DBS penetration is the result of lawful competition, and, it follows, “[t]he damage figures advanced by [Dr. McClave] may be substantially attributable to lawful competition.” Coleman Motor, 525 F.2d at 1353. The Majority responds to this flaw only by stating that the DBS penetration screen was “included to estimate typical competitive market conditions, not to calculate liability for 24 According to Dr. McClave, national average DBS penetration during the six year period for which he calculated damages averaged 24.17%, whereas actual DBS penetration in the Philadelphia DMA averaged 12.77%. (App. at 3411 (McClave Dec.).) 25 The District Court discussed extensively the evidence that “DBS competition constrains cable prices.” Behrend, 264 F.R.D. at 163-65. 27 the foreclosure of DBS competitors.” (Slip Op. at 49.) That explanation misses the mark. In identifying benchmark counties for use in a damages analysis, the goal is not to identify “typical competitive market conditions.” The goal is, and must be, to identify the conditions that would have existed “but for” Comcast‟s alleged anticompetitive conduct. In this case, even in the “but for” hypothetical world, the Philadelphia DMA would not have been typically competitive. Rather, given the District Court‟s findings, there is no question that, as a result of Comcast‟s lawful competition, DBS penetration in the Philadelphia DMA would have been well below that present in a typical competitive market. Thus, by comparing Philadelphia to benchmark counties having the much higher national average DBS penetration, Dr. McClave‟s model wrongly “calculate[s] liability for the foreclosure of DBS competitors,” (id.) imposing damages based on the prices that would have prevailed had Comcast not lawfully foreclosed DBS competition. 3. The Market Share Screen Dr. McClave screened for counties where Comcast‟s market share was “less than 40%,” because that figure represented the midpoint between Comcast‟s 20 percent share before the class period and its 60 percent share during the class period and so identified “markets where Comcast is likely to have less market power than it has acquired in the Philadelphia market.” (App. at 3410 (McClave Dec.).) Under Plaintiffs last viable theory of antitrust impact, however, while Comcast‟s market share is relevant to the question of whether there has been any reduction in overbuilding, it is not relevant – at least not in isolation – to 28 determining the damages caused by that reduction. Instead, the relevant market share is the share that would have been held by any incumbent in the “but for” hypothetical world. As an illustration of that point, consider a hypothetical county with two equally sized franchise areas. Assume that, prior to the class period, Comcast had a 100 percent share of one franchise area and that AT&T had a 100 percent share of the other, so that each had a 50 percent share of the county as a whole. Assume further that, as part of its clustering efforts, Comcast acquired AT&T‟s franchise area so that, today, Comcast has a 100 percent share of the entire county. To test the theory that clustering reduces overbuilding, a comparison between Comcast‟s current 100 percent share of the county and the 50 percent share that Comcast would have had but for its clustering would surely be relevant in determining whether clustering effected any reduction in overbuilding. Next, assume that, after making that comparison, Plaintiffs could show that, had no clustering taken place, RCN would have overbuilt 20 percent of each of the two franchise areas, so that, in the “but for” world, RCN would have a 20 percent share in each franchise area, and Comcast and AT&T would each have an 80 percent share in their respective franchise area. Pursuant to Plaintiffs‟ only theory – that increased overbuilding decreases prices – any damages in that scenario arise solely from the difference between RCN‟s 20 percent share in the “but for” franchise areas and RCN‟s zero percent shares in the current franchise areas. The damages resulting from that foregone overbuilding are the same whether, in the “but for” world, the remaining 80 percent of the franchise in question would have been controlled by Comcast or by AT&T. It follows, therefore, 29 that once the antitrust impact of Comcast‟s clustering – i.e., the reduction in overbuilding – has been identified and accounted for as part of an overbuilding screen, any market share screen applied to isolate the “but for” conditions that would have prevailed in the Philadelphia DMA should screen not just for Comcast‟s share, but for the share of whatever incumbent would have been present but for the clustering. 26 26 Again, this is not to say that Comcast‟s market share, in particular, will never be relevant. As just discussed, it is highly relevant for determining antitrust impact. Moreover, it might have been relevant to damages had the District Court not excluded three of Plaintiffs‟ theories of antitrust impact. In fact, the market share screen appears to be another relic of the Plaintiffs‟ having initially presented four theories of impact. One of those theories was that Comcast‟s increased market share increased its bargaining power and allowed it to reduce prices, Behrend, 264 F.R.D. at 178-81, and a second was that Comcast‟s increased market share reduced the ability of consumers to engage in benchmark pricing by comparing Comcast‟s prices to the prices of other cable providers in the region, id. at 175-78. Had either of those theories survived the class certification process, it might have made sense for Dr. McClave to screen for Comcast‟s market share, because, under those theories, Comcast‟s market share directly impacted price. But the District Court rejected those theories, allowing Comcast to proceed only on a theory that clustering reduced overbuilding. Under that theory, what is relevant is the market share of all incumbent cable providers vis-a-vis overbuilders. 30 Because Dr. McClave‟s model already assumes that there has been a reduction in overbuilding and screens for it, the relevant market share for damages purposes is the share of the market maintained by any incumbent – regardless of the identity of the particular incumbent. By calculating the appropriate market share screen using only Comcast‟s average share throughout the Philadelphia DMA, Dr. McClave has ignored any market share that, in the “but for” hypothetical world, would have been maintained by an incumbent other than Comcast. For franchise areas where Comcast was not present prior to the class period, Dr. McClave should have calculated damages by comparing Comcast‟s current share to the “but for” share that would have been held by any incumbents Comcast replaced. Because he instead effectively calculated damages by comparing Comcast‟s current share to Comcast‟s zero percent share prior to the class period,27 he unfairly suppressed the relevant incumbent share and artificially inflated the damages calculation. 27 I say he “effectively calculated damages” that way because Dr. McClave did not actually make a franchise by franchise comparison, which, as discussed infra Part II(B), is itself problematic. He instead calculated Comcast‟s market share by averaging its share throughout the Philadelphia DMA. But, because he included in that average Comcast‟s zero percent share in the franchises in which it had not been present prior to the class period, instead of including the share held by the incumbent Comcast replaced, it is fair to say that he effectively calculated damages by comparing Comcast‟s actual share in those franchise areas to Comcast‟s zero percent share prior to the class period. 31 Because none of Dr. McClave‟s screens reflect the conditions that would have prevailed in the Philadelphia DMA “but for” any reduction in overbuilding, the damages Dr. McClave calculated are “not the certain result of the wrong,” Story Parchment, 282 U.S. at 562. Accordingly, Dr. McClave‟s opinion cannot help a jury determine damages, and so would be inadmissible at trial for lacking fit. Because Dr. McClave‟s opinion is the only evidence Plaintiffs have offered to meet their burden of showing that damages can be proven using evidence common to the class, I would vacate the District Court‟s class certification with respect to class- wide proof of damages.28 28 The Majority suggests that any problems with Dr. McClave‟s screens are “attacks on the merits of the methodology that have no place in the class certification inquiry,” because, “[e]ven if we were to overrule as clearly erroneous the District Court‟s findings on all four contested pieces of Dr. McClave‟s methodology – i.e., modify both of Dr. McClave‟s screens … only the final amount of estimated damages would change.” (Slip Op. at 52.) I disagree. First, the problems I have identified with Dr. McClave‟s screens call into question not only the amount of damages but also whether there are any means of proving damages at all in thirteen of the eighteen Philadelphia DMA counties. See supra Part II(A)(1). Second, if Dr. McClave‟s model does not presently constitute a relevant means of calculating class- wide damages, to say that the model might be fixed, for example by “modify[ing] both of Dr. McClave‟s screens,” (Slip Op. at 52), is no better than saying that Plaintiffs have made “a threshold showing” of predominance or shown a sufficient “intention to try the case in a manner that satisfies 32 B. Damages Are Not Capable of Being Proven By Evidence Common to the Entire Class While my thoughts thus far have focused on why Plaintiffs have not met their burden of showing that damages can be proven using evidence common to the class, none of the problems I have noted are necessarily irreparable. That is, Dr. McClave could conceivably redesign his model to address overbuilding throughout the Philadelphia DMA, to use actual DBS penetration rates, and to screen for the market share of all incumbents, not just Comcast. Nevertheless, there remains an intractable problem with any model purporting to calculate damages for all class members collectively. Central to Dr. McClave‟s damages model is the conclusion that the price of cable television service in any given franchise area is affected by the relative market shares of at least three entities: overbuilders, DBS providers, and incumbent cable providers. All else being equal, for example, the predominance requirement” – both of which are insufficient under Hydrogen Peroxide. 552 F.3d at 321 (internal quotation marks omitted). Plaintiffs have the burden of establishing predominance and, until they have actually proffered a model that shows how damages can be calculated on a class-wide basis, they have not met that burden – particularly when the only evidence they have offered should be entirely inadmissible. The Majority‟s willingness to overlook the debilitating flaws in Dr. McClave‟s model in an effort to avoid an “attack on the merits,” is precisely the kind talismanic invocation of “concern for merits-avoidance” that Hydrogen Peroxide forbids. Id. at 317 n.17. 33 areas that are overbuilt will have lower prices than areas that are not overbuilt, and areas with high DBS penetration will have lower prices than areas with low DBS penetration. For that reason, Dr. McClave‟s model identifies benchmark counties by screening for the relative market shares of those three entities.29 While I do not accept the manner in which Dr. McClave has measured the relative shares of those entities in the “but for” Philadelphia DMA, I accept the premise that the relative shares have significant influence on the price of cable television service. If price does vary with the changes in relative share within a franchise area, however, it is hard to see how those 650 franchise areas30 can simply be treated as average for purposes of proving damages. The record indicates that, on the contrary, the “but for” market shares of overbuilders, DBS providers, and incumbent providers would vary, sometimes significantly, from franchise area to franchise area. Addressing overbuilding first, RCN – the only party licensed to overbuild any part of the Philadelphia DMA – was licensed to overbuild in only five counties. (App. at 3640 (Williams Dec.); App. at 4284-85 (Singer Reply Dec.).) 29 Or, at least, he screens for overbuilders, DBS providers, and a single incumbent provider – Comcast. I have already identified, supra Part II(A)(3), why he should instead screen for incumbent share. 30 Dr. Besen, one of Comcast‟s experts, reports that there are 649 unique franchise areas in the Philadelphia DMA. (App. at 3782 (Besen Reply Dec.).) 34 While Plaintiffs‟ experts have opined that, had RCN successfully overbuilt those five counties, it would have continued overbuilding elsewhere, (App. at 4284-85 (Singer Reply Dec.)), any overbuilding into the other parts of the Philadelphia DMA would, it seems clear, have come later than the overbuilding of the five licensed counties. Thus, while some franchise areas might have been overbuilt early in the class period, other franchise areas would likely never have been overbuilt at all or have been overbuilt only later in the class period. There might, for instance, in the “but for” world be some franchise areas that were 50 percent overbuilt for the entire class period and other franchise areas that were only 5 percent overbuilt and only for a single year, or perhaps not overbuilt at all. That means that, both throughout the Philadelphia DMA and throughout the class period, there would probably be very significant variation in the “but for” level of overbuilding from franchise area to franchise area. Consider next DBS penetration. Dr. McClave testified that the DBS penetration rate he used for the Philadelphia DMA was an average for the DMA, but he also said that it was his understanding that “DBS penetration varies across the cluster here” and that it was “possible that some of the counties in the Philadelphia DMA in fact have penetration that‟s above the national median.” (App. at 729-30 (McClave Cross).) Thus, according to Dr. McClave, not only does DBS penetration vary across the Philadelphia DMA, but the variation is pronounced enough that some parts of the Philadelphia DMA have above national average DBS penetration despite the fact that the Philadelphia DMA, as a whole, has DBS penetration at only half the national average. Because DBS penetration was unaffected by Comcast‟s alleged anti-competitive conduct, see supra Part II(A)(2), 35 DBS penetration in the “but for” Philadelphia DMA would likewise vary significantly from one franchise area to another. Finally, with respect to the incumbents‟ market share, the record gives little information regarding what the share of any non-Comcast incumbent would be in the “but for” world. We do know, though, that Comcast‟s share prior to clustering varied markedly from franchise area to franchise area. (See, e.g., App. at 3833 (Chipty Dec.) (stating that Comcast “had a zero percent share of housing units in the majority of counties” and, therefore, that “Comcast‟s share in the counties in which it was present was substantially higher than [its average market share]”); App. at 733 (McClave Cross) (testifying that, at the beginning of the class period, Comcast was present in “maybe half, maybe less of the counties” and that its share “in the counties where [it was] present” was probably higher than its average share)). And, where the other two components of market share – DBS penetration and overbuilding31 – vary from one franchise area to another, it becomes a near mathematical certainty that the remaining portion of the franchise held by incumbent cable providers must likewise vary.32 31 While there may be other “alternative delivery systems” that have a limited share of the market, Dr. McClave includes those providers in his DBS penetration screen, see supra note 21, and they are, therefore, accounted for. 32 It is possible, of course, that the variation in DBS and overbuilder shares could be such that the combined total of the two is the same in different franchise areas, and, therefore, it is not a true mathematical certainty that 36 The wide variation in the relative market shares evidenced by the record makes it hard to imagine a means of calculating class-wide damages. Even if Dr. McClave‟s benchmarks were not problematic, to say that Comcast‟s “but for” share of the market throughout the Philadelphia DMA would be, on average, 40% is about as meaningful as saying that “with one foot on fire and the other on ice, I am, on average, comfortable.”33 Given that the three major factors identified as influencing price – overbuilding, DBS penetration, and incumbent share – vary widely within the franchise areas across the DMA, and given further that Comcast prices its cable service at the franchise level, (see App. at 716), I have difficulty accepting that it is appropriate to ignore those differences and take an average across the counties of the DMA.34 incumbent share must also vary. That that would occur across the 650 franchise areas, however, seems implausible in the extreme. 33 Sometimes attributed to Mark Twain, the actual source of this quote is unknown. 34 The Majority asserts that “concern over using mathematical averages across the Philadelphia DMA … is notably absent from Comcast‟s briefing ….” (Slip Op. at 53- 54 n.18.) But it is not absent. In fact, Comcast criticizes Dr. McClave‟s screens by explaining that: prior to the Transactions[,] Comcast did not operate in the majority of franchise areas in the DMA. By contrast, in the franchise areas where Comcast did operate, it is undisputed that 37 This primary flaw in Dr. McClave‟s methodology – using a single set of assumptions for the entire Philadelphia DMA – cannot be fixed merely by altering his model. It seems to me that no model can calculate class-wide damages because any damages – such as they may be – are not distributed on anything like a similar basis throughout the DMA.35 Rather, where some class members might reside in a Comcast‟s market share was significantly higher than 40%. Thus, the pre-class “20%” market share Dr. McClave employed in the creation of his screen is a mirage, arrived at solely through the artifice of averaging Comcast‟s greater-than-40% share of markets where it did operate with its “0% share” in hundreds of markets where it was not even present. (Appellants Br. at 43.) That criticism precisely mirrors my own. Because of the significant variation in the market makeup from franchise area to franchise area, DMA-wide averages are not reliable. See also infra n.35. 35 This is not simply a case where there might be some variation in the amount of damages from one class member to another that can be ignored in order to gain the benefit of class treatment. Instead, due to the wide variations in the market makeup of the franchise areas across the DMA, proving damages will require some factual inquiry into the relative market shares of overbuilders, DBS providers, and incumbents in individual franchise areas (or perhaps, as subsequently noted in this dissent, groups of franchise areas). The Majority, quoting Wright‟s Federal Practice and 38 franchise area that would have been 50 percent overbuilt for the entire class period and other class members might reside in a franchise area that would have been only 5 percent overbuilt and only for a single year, or not overbuilt at all, it strains credulity to believe that the damages suffered by those individuals would all be the same as a result of reduced overbuilding. Yet Dr. McClave‟s model treats them as though they are the same,36 as would any model attempting to calculate damages on an average class-wide basis. Procedure, suggests that those differences in damages should not affect the certification process because “„it uniformly has been held that differences among the members as to the amount of damages incurred does not mean that a class action would be inappropriate.‟” (Slip. Op. at 45-46 (quoting CHARLES ALAN WRIGHT ET AL., FEDERAL PRACTICE AND PROCEDURE § 1781 (3d ed. 2005)).) I agree with the quoted statement, but would point also to the following quote from the same treatise: “Rather, the question of damages can be severed from that of liability and tried on an individual basis.” Id. Thus, neither Wright – nor any authority I can find – suggests that where there are wide differences in damages from one class member to another, those differences can be ignored. Wright suggests instead that those differences can be accounted for by considering liability on a class-wide basis but damages on a more individualized basis – consistent with what I propose. 36 I recognize that Dr. McClave‟s model does not treat all franchise areas exactly the same, because he uses actual prices on a county-by-county level and, as a result, calculates a separate “but for” price for each county. (App. at 3424-26 (McClave Dec.).) But, while he uses actual prices on a 39 The variation in conditions within the nearly 650 franchise areas in the Philadelphia DMA means that the issue of damages is more fractured than a single class can accommodate. I do not suggest that there necessarily would need to be 650 subclasses. It may well be that subclasses could be created encompassing groups of multiple franchise areas having similar demographics. See, e.g., Eisen v. Carlisle & Jacquelin, 391 F.2d 555, 566 (2d Cir. 1968) (explaining that where “differences among the class members county-by-county level, he calculates the “but for” prices using the same benchmark counties for the entire Philadelphia DMA (again, excepting Lehigh and Northampton, where Comcast has no presence). He treats the DMA as though the “but for” conditions would have been the same throughout. The Majority states that this criticism “overstates the degree of dissimilarity among the franchise areas” because I fail to note that “[m]any franchise areas within counties often have identical or nearly identical pricing.” (Slip Op. at 54 n.18 (quoting App. 3409 (McClave Dec.).) It may well be that there are similarities allowing for grouping of franchise areas, as I note in suggesting the possibility of subclasses. More fundamentally, however, the problem with Dr. McClave‟s opinion is not that it fails to account for variations in actual prices in the Philadelphia DMA, but that it fails to account for variations in the “but for” conditions that would have existed within the Philadelphia DMA. By so doing, the model is unable to distinguish between persons living in areas that may have been highly overbuilt and who, thus, would have suffered substantial damages, and persons living in areas that may never have been overbuilt and who, thus, would have suffered no damages. 40 bear only on the computation of damages,” it “can be adequately handled … [by] divid[ing] the class into appropriate subclasses”). Whether that would necessitate the creation of so many subclasses as to defeat the benefit of class treatment is something I do not venture to conclude on this record. But I would remand the case to the District Court for consideration of the feasibility of subclasses. III. Conclusion For the foregoing reasons, I would vacate the District Court‟s certification order to the extent it provides for a single class as to proof of damages and remand the case for the District Court to address whether Dr. McClave‟s model could, in fairness, be revised to accurately reflect the conditions that would have existed in the Philadelphia DMA in the absence of any reduction in overbuilding caused by clustering. I would further ask the District Court to consider whether the class certified for proving antitrust impact can be divided into appropriate subclasses for purposes of proving damages. 41