Trends in Class Certification

As class certification standards have continued to develop in the antitrust context during the past year, questions of predominance have remained at the forefront, although the numerosity requirement has also once again been put to the test. This chapter places these issues of class certification in context by tracing the standards of certification and discussing the evolution of the ‘rigorous analysis’ requirement now required by federal courts. It then spotlights notable decisions from the past year that have grappled with challenges to the sufficiency of plaintiffs’ statistical models used to demonstrate the preponderance of class-wide questions, principally on the issue of showing class-wide harm, and with the number of putative class members required to be sufficiently numerous for certification.

Standards for class certification

Claims of anticompetitive conduct often involve allegations of harm to a large number of market participants. It is no surprise, then, that such claims are often brought by way of Rule 23 of the Federal Rules of Civil Procedure, which allows for the collective representation of a large group of plaintiffs allegedly injured by the same conduct. Indeed, antitrust disputes continue to account for a significant share of all active class actions. Setting aside class actions relating to covid-19, which accounted for approximately 6 percent of all class actions in 2021, antitrust class actions were the sixth most common type of case litigated on a class-wide basis, although their share of total class actions has been falling in the past few years, as has the total number of antitrust cases filed (see chart below^[1]).^[2]

Although class actions are commonplace in antitrust litigation, the Supreme Court has long considered them to be the exception to the general rule that parties may litigate only on their own behalf.^[3] Only if a proposed representative demonstrates that he or she is part of a class whose members possess the same interest and suffered the same injury may a court allow a class action to proceed.^[4] To that end, Rule 23(a) requires a putative class to satisfy four requirements:

• the class must be sufficiently numerous that the joinder of all members would be impracticable (the ‘numerosity’ requirement);
• the lawsuit must raise questions of law or fact common to the putative class (the ‘commonality’ requirement);
• the representative plaintiffs’ claims^[5] must be typical of the claims of the class (the ‘typicality’ requirement); and
• the representative parties must show that they will fairly and adequately protect the interests of the class (the ‘adequacy of representation’ requirement).^[6]

A proposed class must also satisfy Rule 23(b), which contemplates three types of classes of plaintiffs actions. A class satisfies Rule 23(b)(1) if it demonstrates either that separate actions ‘would create a risk of [] inconsistent or varying adjudications with respect to individual class members that would establish incompatible standards of conduct for the party opposing the class’^[7] or ‘would be dispositive of the interests of the other members not parties to the individual adjudications or would substantially impair or impede their ability to protect their interests’.^[8] (The latter class action is known as a ‘limited funds’ action.^[9]) Under Rule 23(b)(2), a class action may also be maintained if plaintiffs seek only injunctive relief.^[10] Most commonly in the antitrust context, however, plaintiffs seek certification under Rule 23(b)(3), which allows a class of plaintiffs seeking monetary damages who demonstrate that common questions of law or fact predominate over any questions affecting only individual class members (the ‘predominance’ requirement), and that litigation by class action is superior to litigating individual claims (the ‘superiority’ requirement).

The courts’ evolving approach to applying class certification standards

For decades, plaintiff classes faced a relatively low hurdle to certification under Rule 23(b)(3). Guided by the Supreme Court’s holding in Eisen v Carlisle & Jacquelin^[11] that ‘nothing in either the language or history of Rule 23 . . . 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’,^[12] district courts routinely eschewed defendants’ efforts to defeat class certification by reference to purported deficiencies in plaintiffs’ theories of liability. This was so even after 1982, when the Supreme Court held in General Telephone Company of the Southwest v Falcon^[13] (Falcon) that district courts must conduct a ‘rigorous analysis’ of the Rule 23 factors before certifying a class, and that ‘sometimes it may be necessary for the court to probe behind the pleadings before coming to rest on the certification question’.^[14]

But exactly how far district courts could pull back the curtain on the merits of a suit at the certification stage remained unsettled. In 2004, for example, one court held that it would be an ‘injustice’ to require plaintiffs to establish the elements of Rule 23 by a preponderance of the evidence when those elements are ‘enmeshed’ in the merits.^[15] Another court observed in 2007 that the circuits were splite regarding the district court’s role in resolving a ‘battle of the experts’ at class certification.^[16]

In 2008, the Third Circuit^[17] articulated a more muscular understanding of the district courts’ gatekeeping role in conducting a ‘rigorous analysis’ in the antitrust context, holding that (1) a class may be certified only upon a showing by a preponderance of the evidence (and not merely a ‘threshold’ showing that some courts had required) that the Rule 23 requirements are met, (2) the district court’s role at certification is to resolve all relevant factual or legal disputes, even if they overlap with the merits, and (3) in making its decision, the court must consider expert testimony, no matter which party offers it.^[18]

The Supreme Court was not far behind. In 2011, the Court held in Wal-Mart Stores, Inc v Dukes, an employment discrimination action, that a plaintiff ‘must affirmatively demonstrate his compliance with the Rule – that is, he must be prepared to prove that there are in fact sufficiently numerous parties, common questions of law or fact, etc.’.^[19] That such proof may overlap with the merits, as the Court held in Falcon ‘cannot be helped’.^[20]

Two years later, the Supreme Court appeared to soften its stance in Amgen Inc v Connecticut Retirement Plans & Trust Fund,^[21] holding that Rule 23(b) ‘requires a showing that questions common to the class predominate, not that those questions will be answered, on the merits, in favor of the class’.^[22] It further explained that ‘Rule 23 grants courts no license to engage in free-ranging merits inquiries at the certification stage. Merits questions may be considered to the extent – but only to the extent – that they are relevant to determining whether the Rule 23 prerequisites for class certification are satisfied’.^[23] Nevertheless, it held the following month in Comcast Corp v Behrend that courts must ‘take a close look at whether common questions predominate over individual ones’, and reiterated the district court’s burden to conduct a ‘rigorous analysis’.^[24] Notably, that analysis requires consideration of whether ‘any model supporting a plaintiff’s damages case [is] consistent with its liability case, particularly with respect to the alleged anticompetitive effect of the violation’.^[25]

The Supreme Court once again addressed class certification in Tyson Foods, Inc v Bouaphakeo (Tyson Foods),^[26] holding that plaintiffs may sometimes rely on statistical sampling to establish that common questions of liability predominate. The plaintiffs – Iowa meat-processing employees – sought to prove that, including time spent donning and doffing protective gear, they worked more than 40 hours per week and were entitled to overtime pay. In the absence of individualized data, the Court found that the plaintiffs satisfied the predominance requirement by applying an average donning and doffing time from a sample of employees to the class as a whole. The Court, while careful to avoid prescribing general rules as to the use of such representative sampling to establish class-wide liability, held that the soundness of the plaintiffs’ sampling was a question common to the class, and one more properly addressed at summary judgment.^[27] Although it sided with the plaintiffs, the Court reaffirmed the district courts’ need ‘to give careful scrutiny to the relation between common and individual questions in a case’.^[28]

The circuits have followed suit. Although the Supreme Court has not explicitly ruled on burdens of proof at the class certification stage, circuits now widely agree that a plaintiff must prove compliance with Rule 23 by a preponderance of the evidence, even if the proof overlaps with that plaintiff’s ultimate theory of liability.^[29] Nevertheless, decisions issued in the past couple of years demonstrate that the full contours of the ‘rigorous analysis’ standard, and how it applies to antitrust claims, are far from settled.

Frequent issues in antitrust class certification

The question of predominance in Rule 23(b)(3) putative class actions has been the focus of many antitrust cases, and there has been a continuation of the trend in 2021 as was observed in the previous year.^[30] Although there were perhaps fewer landmark decisions than in the past, the Ninth Circuit’s en banc decision affirming certification in Olean Wholesale Grocery Cooperative, Inc v Bumble Bee Foods LLC (Olean) was significant. In Olean and others, statistical and econometric modeling has continued to be central at the certification stage as parties offered dueling methods of attempting to prove or undercut predominance. In addition, courts continue to navigate their role as gatekeepers to certification when defendants identify uninjured plaintiffs allegedly swept up by plaintiffs’ models, or when defendants otherwise criticize those models for papering over individualized differences among class members. Separately, although less frequently litigated, the numerosity standard continued to develop last year, with the Fourth Circuit weighing in on the inquiry required by Rule 23(a)(1).

Statistical and econometric modeling to show class-wide antitrust injury

One area of recent focus has been the degree to which district courts scrutinize plaintiffs’ methodologies for determining class-wide harm. This raises special considerations in antitrust cases, where plaintiffs must demonstrate not only class-wide injury (i.e., antitrust injury)^[31] but also that the harm at issue resulted from the antitrust violation (i.e., antitrust impact), requiring them to show that injury consistent with their theory of liability is capable of proof on a class-wide basis.^[32]

In this regard, and in the wake of Tyson Foods, courts have typically accepted the use of statistical modeling as a method of proof of antitrust impact.^[33] Although some courts have cautioned that ‘certification [should not be] automatic every time counsel dazzle the courtroom with graphs and tables’,^[34] there is widespread agreement that ‘plaintiffs are permitted to use estimates and analysis to calculate a reasonable approximation of their damages’ to establish antitrust impact at the certification stage.^[35] Regression models, which ‘control for the effects of the differences among class members and isolate the impact of the alleged antitrust violations on the prices paid by class members’, have been viewed particularly favorably, with the Ninth Circuit recently endorsing them as ‘a generally reliable econometric technique’ to prove class-wide antitrust impact.^[36]

The Ninth Circuit is not alone. In December 2020, a court in the Eastern District of Virginia approved of an econometric model purporting to isolate the effects of an alleged price-fixing conspiracy by comparing prices during the alleged conspiracy with those that prevailed during a ‘benchmark’ period unaffected by anticompetitive behavior.^[37] Although the court recognized that the plaintiffs may hone their model and adjust their variables through discovery, it was satisfied at the class certification stage that the methodology they presented could reasonably be used to determine class-wide antitrust impact.^[38]

Such regression modeling has also reached near ubiquity in class actions in the pharmaceutical space, particularly with respect to allegations that brand and generic drug manufacturers conspired to delay entry of generic versions of brand-name drugs. In the In re Ranbaxy multidistrict litigation,^[39] direct purchaser plaintiffs, end payor plaintiffs^[40] and third-party payors brought claims against Ranbaxy for allegedly delaying entry of three generic drugs by wrongfully acquiring exclusivity periods from the US Food and Drug Administration (FDA).^[41] In support of claims that they paid artificially inflated prices for the generic drugs, the direct purchaser and end payor plaintiffs offered similar regression models that compared the monthly average prices of the at-issue drugs with the average prices in a ‘but-for’ world without the alleged anticompetitive conduct.^[42] Siding with the plaintiffs, the court approved of these ‘yardstick’ models as ‘widely accepted methods of proving antitrust injury and damages on a class-wide basis’.^[43]

The presence of uninjured plaintiffs in putative classes

Although, as discussed above, it is uncontroversial that statistical evidence can supply the necessary showing of antitrust impact in recent years, courts have taken a harder look at that evidence, particularly when defendants have offered competing models or theories to undermine the reliability of plaintiffs’ statistical evidence or to drive a wedge between that evidence and the plaintiffs’ theory of liability. Although some courts still reserve battles of the experts for the jury,^[44] most now consider it necessary to resolve such disputes as part of their ‘rigorous analysis’ at the certification stage.

One common theme in attacks on plaintiffs’ modeling efforts has been the presence of uninjured plaintiffs in a proposed class. The issue has become a proving ground for the predominance requirement and has sparked a debate among courts about the degree to which statistical models should be scrutinized at certification. In the past decade, courts have refused to certify a class containing more than a de minimis number of uninjured plaintiffs, reasoning that the need to identify those uninjured plaintiffs will overshadow questions common to the class.^[45] The presence of uninjured plaintiffs has also implicated questions of Article III standing,^[46] as well as Seventh Amendment and due process concerns, with some defendants arguing that the inclusion of unidentified, uninjured plaintiffs in a certified class deprives them of a meaningful opportunity to contest each plaintiff’s injury and forces them to pay for more harm than the alleged anticompetitive conduct may have caused.^[47]

In 2018, the plaintiffs in In re Asacol^[48] were dealt a blow when the First Circuit held that the proposed class from the plaintiffs’ economic model swept up too many purchasers who were not injured by the defendants’ conduct. That case concerned allegations that drug manufacturers conspired to delay market entry of generic versions of an ulcerative colitis treatment. At the certification stage, the defendants argued that certain ‘brand loyalists’ would not have switched to the generic drug even if it had been introduced earlier, and therefore did not suffer cognizable injury from the allegedly delayed entry of the generic.^[49] The district court certified a class of two subsets of direct purchasers, but the First Circuit reversed, holding that the district court failed to conduct a sufficiently rigorous analysis of the plaintiffs’ methodology for determining antitrust impact. Under First Circuit precedent, the presence of a de minimis number of uninjured plaintiffs does not categorically defeat a finding of predominance.^[50] In In re Asacol, however, the court found that as many as 10 percent of the defined class’s members were uninjured,^[51] which exceeded the de minimis threshold.^[52] In the absence of an administratively feasible mechanism to weed them out of the class, the court held that the plaintiffs had failed to carry their burden that common questions predominated.^[53]

The DC Circuit in In re Rail Freight Fuel Surcharge^[54] sided with the First Circuit the following year when it upheld a district court’s denial of certification to a class of direct purchasers who accused the four largest freight railroads in the United States of conspiring to fix fuel prices. As in In re Asacol, the DC Circuit held that the plaintiffs’ expert’s damages model, even if reliable in attempting to show an average overcharge to the class, failed to show class-wide injury because the plaintiffs’ modeling identified 2,037 members of the proposed class (or 12.7 percent) as uninjured, exceeding a de minimis amount.^[55] Furthermore, because the plaintiffs’ model did not have a winnowing mechanism, the Rail Freight court upheld the district court’s denial of certification.^[56]

A panel of the Ninth Circuit appeared to follow in the footsteps of In re Asacol in April 2021 when it embraced the de minimis limit on uninjured class members.^[57] The defendants in Olean had admitted that they conspired to fix the price of canned tuna.^[58] Having established antitrust liability, the plaintiffs secured certification of three different classes of purchasers. Reversing the district court, the Ninth Circuit panel found that although the plaintiffs’ economic model classified only 5.5 percent of the direct purchaser plaintiffs as uninjured, the defendants’ model suggested that the proportion was as high as 28 percent.^[59] Citing In re Asacol, it held that the ‘rigorous analysis’ standard required the district court to resolve whether the plaintiffs’ class in fact included as many uninjured plaintiffs as the defendants had predicted, even if that question overlapped with the merits.^[60]

The defendants’ victory was short-lived. In August 2021, the Ninth Circuit vacated the panel decision in Olean^[61] and, in April 2022, it revived the certified classes in an en banc opinion.^[62] The divided court held that common questions may predominate over individualized ones even if a proposed class contains more than a de minimis number of uninjured plaintiffs.^[63] The majority declined to read either In re Asacol or Rail Freight as adopting a per se rule precluding certification where the class contains more than a de minimis number of uninjured class members.^[64] Instead, it understood those decisions as urging a case-by-case examination, and as requiring a finding that the predominance requirement is unsatisfied only if ‘the need to identify uninjured class members “will predominate and render an adjudication unmanageable”’.^[65] In this regard, the Olean majority appears to view its decision as being in harmony both with In re Asacol and Rail Freight on the one hand and with Messner on the other hand.^[66] In its 2012 Messner decision, the Seventh Circuit counseled against certification of a class if it contains a ‘great many’ uninjured plaintiffs – a flexible standard that turns on the facts of each case.^[67] Although it remains to be seen how district courts will navigate their respective circuit’s stance on uninjured plaintiffs, Olean may foreshadow a more unified case-by-case approach to predominance.^[68]

In the end, however, the Olean majority did not wade far into the debate about when the presence of uninjured plaintiffs in a proposed class may or may not defeat a finding of predominance. Departing from the 2020 panel’s decision, the majority held that the defendants’ model did not actually identify any uninjured plaintiffs, and that there was no dispute that the defendants’ conspiracy affected the entire packaged tuna industry nationwide.^[69] With the undisputed conspiracy as a backdrop, the court found that the plaintiffs’ model isolated the effect of that conspiracy, and that it was therefore capable of proving the plaintiffs’ ‘simple one-step theory’ that the conspiracy raised the baseline price of tuna for all buyers.^[70]

The Ninth Circuit’s Olean decision was one of a few opinions during the past year to address the distinction between a class that includes members who by definition could not have been harmed and a class that includes members who ultimately fail to carry their burden of proof that they were harmed. That distinction was relevant to the Seventh Circuit’s reversal of class certification in In re Opana ER Antitrust Litigation.^[71] In June 2021, the Northern District of Illinois certified a class of end payor plaintiffs and direct purchaser plaintiffs alleging an anticompetitive arrangement that delayed entry of a generic version of Opana, an opioid painkiller.^[72] On appeal, the Seventh Circuit criticized the district court for failing to address the defendants’ arguments regarding categories of uninjured plaintiffs and for failing to explain why they did not undermine a finding of predominance.^[73] It emphasized the need to distinguish between plaintiffs who may not have been injured and plaintiffs who, by definition, could not have been injured based on the plaintiffs’ theory of harm.^[74] It also instructed the district court on remand to address whether the plaintiffs’ proposed amended class definition excluded such uninjured groups.^[75]

Consistent with our observation last year,^[76] the debate about a requisite proportion of uninjured plaintiffs may have less to do with a quantitative disagreement among courts about how many uninjured plaintiffs is too many, and more to do with perceived deficiencies in the plaintiffs’ models and their relationship to the proposed classes’ liability theories.

Using averages to demonstrate varying degrees of class-wide injury

The use of average-pricing models has also been a focal point in recent predominance inquiries. This is related in large part to the issues discussed above, and defendants often argue that the presence of uninjured plaintiffs undermines the reliability of averaging. However, although the presence of too many uninjured class members appears to be treated as a binary issue, the use of average-pricing models is often cited by defendants as implicating more granular and individualized questions, such as plaintiffs’ price elasticity and negotiating power, that may operate to defeat a finding of predominance.

In challenging the direct purchaser plaintiffs’ proof of class-wide antitrust impact, the defendants in Olean argued that their use of ‘pooled’ regression models, which aggregated the sale transaction data during the conspiracy and benchmark periods, impermissibly utilized averaging assumptions that acted to ‘paper over individualized differences among the class members’.^[77] The en banc Ninth Circuit was not persuaded. It disagreed that the plaintiffs’ regression models merely applied ‘averaging assumptions’ to a complex market, and it rejected any categorical argument that a pooled regression model is incapable of showing class-wide antitrust impact.^[78] Instead, in assessing whether the district court conducted a sufficiently rigorous analysis, the Ninth Circuit looked to whether the models could reliably explain the plaintiffs’ theory of injury on a class-wide basis, and more granularly, whether the models controlled for variables that may otherwise undermine the explanatory power of those regressions.^[79] Affirming the district court’s order, it held that it was ‘both logical and plausible’ that the conspiracy raised the baseline price for all class members, and that the district court did not abuse its discretion in finding that the models sufficiently controlled for individualized variation among plaintiffs.^[80]

In contrast, in In re Pre-Filled Propane Tank Antitrust Litigation,^[81] a court in the Western District of Missouri declined to certify a class of end purchasers of propane gas tanks who alleged that the nation’s leading sellers of portable propane exchange tanks conspired to underfill the tanks without concomitantly reducing their price.^[82] As in Olean, the plaintiffs’ expert introduced a regression analysis purporting to show that the defendants’ conspiracy caused all indirect purchasers to pay supracompetitive prices for propane exchange tanks. And, as with the plaintiffs’ models in Olean, the plaintiffs’ model arrived at an average overcharge for the class. The court denied the certification motion, crediting the defendants’ model, which criticized the plaintiffs’ use of a single average overcharge estimate as wrongly attributing injury to a significant portion of the class that otherwise lacked evidence of any injury. It also cited the plaintiffs’ expert’s admission that his regression model could not account for individualized differences between retailers, such as negotiating leverage. The court also found there was no evidence that the alleged conspiracy persisted through the class period, such that the model ‘assume[d], rather than . . . assesse[d], the “common impact”’.^[83]

The district court in In re Lamictal Direct Purchaser Antitrust Litigation^[84] has also denied certification based on the plaintiffs’ use of averages. In 2020, the Third Circuit had reversed the district court’s certification of a class of direct purchasers of the brand-name anti-epilepsy drug Lamictal and a generic version, who alleged that the defendants entered into a reverse settlement agreement delaying the market entry of the brand manufacturer’s authorized generic.^[85] In reversing, the Third Circuit criticized the plaintiffs’ model, which compared average generic discounts in a but-for world with the average price paid by the plaintiffs.^[86] On remand, the district court characterized its central task as determining the ‘acceptability of averages’ and found that, although the plaintiffs’ theory of liability was sound, they failed to account for evidence that the brand manufacturer offered varying discounts in response to the generic’s market entry. Although the court accepted that, on average, the introduction of an authorized generic decreases the price of a competing generic, it held that the plaintiffs could not rely on such an average without also proving by a preponderance of the evidence that they would have received additional discounts if the authorized generic had been launched.

As with the issue of uninjured plaintiffs, courts have continued to scrutinize average-pricing models, the propriety of which depends not only on a proposed class’s theory of liability but on the degree to which they account for variation of injury between class members.

Numerosity’s return to the limelight

As in past years, numerosity is rarely a stumbling block for putative classes alleging antitrust violations. For example, the District Court of Massachusetts certified a class of 39 plaintiffs, holding that judicial economy and the plaintiffs’ geographic dispersion favored proceeding as a class action.^[87] The Northern District of Illinois certified a class of 37 plaintiffs on similar grounds, noting that there is no ‘magic number’ to establish numerosity.^[88]

Nevertheless, the Third Circuit added teeth to that requirement in 2016 when it vacated the Eastern District of Pennsylvania’s certification of a class of 22 plaintiffs.^[89] Citing circuit precedent that numerosity is generally satisfied if the potential number of plaintiffs exceeds 40, the Third Circuit clarified that a court’s numerosity analysis turns not on bright-line numerical thresholds but on whether the joinder of all interested parties would be impracticable. On remand, the Eastern District of Pennsylvania denied certification on numerosity grounds, finding both that judicial economy concerns and the plaintiffs’ ability and motivation to litigate as joint plaintiffs disfavored certification.^[90]

This past year, the Fourth Circuit followed suit when it reversed certification on numerosity grounds. In 2020, the Eastern District of Virginia had certified a class of 35 plaintiffs over the defendants’ objection that the class failed to satisfy Rule 23(a)(1).^[91] The district court found that the plaintiffs – who were direct purchasers of cholesterol medication Zetia who accused drug manufacturers Merck and Glenmark Pharmaceuticals of conspiring to keep generic versions of Zetia off the market – were sufficiently numerous to justify a class action, holding in part that the focus of Rule 23(a)(1) is whether judicial economy favors class adjudication over multiple individual trials.^[92] Citing the Third Circuit’s Modafinil decision,^[93] the Fourth Circuit reversed and held that the appropriate comparison is instead between class actions and joinder, observing that in the district court’s formulation, ‘the judicial-economy factor would always favor class certification, which is simpler to manage than individual lawsuits’.^[94] In light of what he observed as the lack of a mechanical test for determining satisfaction of numerosity, Circuit Judge Niemeyer, writing in concurrence, identified non-exclusive factors for district courts to consider, including the geographic dispersion of the putative class members, their ability and motivation to file suit absent class certification, and any difficulty in identifying specific class members.^[95]

Although the frequency with which numerosity has been litigated at the class certification stage pales in comparison with issues of predominance, the requirement has nevertheless received its share of the limelight in recent years.

Concluding remarks

Although there were perhaps fewer landmark opinions during the past year that have addressed class certification in antitrust disputes than there has been in previous years, courts have continued the trend of developing the ‘rigorous analysis’ standard. The question of predominance has continued to take center stage as courts have scrutinized plaintiffs’ ability to demonstrate – through statistical modeling that often utilizes averaging assumptions – that common questions predominate, even if a putative class contains uninjured plaintiffs. In addition, although fewer courts have grappled with the numerosity standard, it too has undergone development. Although it is difficult to predict what the future holds for these standards, the decisions during the past year suggest that their development will persist in the years ahead.

Notes