In Olean Wholesale Grocery Co-op Inc. v. Bumble Bee Foods LLC, No. 19-56514 (9th Cir. Apr. 6, 2021), the Ninth Circuit clarified several class certification issues that have been heavily in dispute in antitrust class actions for more than a decade:

  • Plaintiffs must demonstrate the Rule 23 predominance requirement by a preponderance of evidence, joining precedent set in several other circuits;
  • Statistical or representative evidence can be used to establish predominance;
  • A more than de minimis  number of uninjured putative class members defeats predominance; and
  • District courts must resolve factual disputes among competing economic statistical models before finding predominance and certifying a class.


Plaintiffs alleged price fixing by the three largest American producers of packaged tuna that collectively account for over 80% of branded packaged sales in the United States. Plaintiffs alleged defendants conspired to inflate prices through price fixing, limited promotional activity, and the exchange of confidential information. Two of the three defendants, Bumble Bee and StarKist, pleaded guilty to criminal price fixing charges after the civil action was filed, and the third defendant, Chicken of the Sea, later admitted to price fixing and cooperated with the criminal investigation.

Plaintiffs proposed three classes of purchasers: a direct purchaser class (“DPP”), a commercial food service product class, and an end payer class (made up of end user consumers). To establish predominance for these proposed classes, plaintiffs submitted expert testimony that relied on statistical analyses for the average overcharge for the classes and a regression analysis for the percentage of injured class members. Plaintiffs' expert concluded that defendants' alleged conduct caused injury to 95.4% of the members of the direct purchaser class. Defendants' experts Dr. John Johnson and Dr. Laila Haider objected to this analysis, opining that an average estimated overcharge incorrectly assumed that every class member was injured in the same way. Defendants' expert Dr. Johnson conducted an overcharge analysis unique to each class member and concluded that only 72% of the proposed DPP class members had been injured, not 95.4% as plaintiffs opined. Both parties' experts conducted similar analyses for the other two proposed classes with similar results.

The district court certified the classes, concluding that defendants' challenges to plaintiffs' methodology for establishing predominance in the class were “ripe for use at trial” but “not fatal to a finding of classwide impact.” Defendants petitioned for review under Rule 23(f), which a motions panel granted. Defendants made two arguments on appeal: first, that the representative evidence presented by plaintiffs' expert – referring to the use of average overcharges across the class – could not be used to establish predominance; and second, that the representative evidence did not in fact establish predominance because a significant percentage of the class, up to 28% in defendants' experts view, may not have suffered an injury.

The Ninth Circuit first explained that as a general matter, representative evidence can be used to establish predominance. The Court noted that using representative evidence at class certification is “nothing new.” The Supreme Court held in Tyson Foods that representative evidence can be reliable, ruling that whether a particular representative sample is sufficient to establish classwide liability should be determined based on the facts and proposed use in a particular case. Representative evidence must still be “scrutinized with care and vigor,” because both courts and jurors can over-rely on statistical evidence about which they lack knowledge. Nonexperts can tend to defer to the statisticians presenting such evidence, so a court must be wary of overreliance on statistical evidence to ensure it is sufficiently scrutinized. In the instant case, the Ninth Circuit determined that plaintiffs' use of representative evidence here could be reliable.

The Ninth Circuit identified three factors to consider when determining whether the representative evidence can be used to show predominance. First, the Court considered whether the offered representative evidence can be used to establish liability in an individual suit, because if it could not be relied on in an individual suit, representative evidence cannot establish predominance at class certification. The Court concluded that each class member could have relied on the models presented by plaintiffs' expert, which was sufficient for this factor. Second, the Court considered whether, consistent with the rule the Supreme Court set forth in Comcast, the plaintiffs' model and representative evidence effectively tie defendants' conduct to the claimed injury, meaning the evidence is consistent with plaintiffs' underlying theory of liability. In the instant case, the Court concluded that there was a sufficient nexus between the representative evidence and the price-fixing theory to satisfy this factor. And third, the Court asked whether any averaging assumptions “mask” individual differences among class members, which would make the evidence inadequate to establish that common questions predominate. The Court concluded, while direct purchasers individually negotiated prices, the alleged price fixing “could have raised the baseline price at the start of negotiations” for even the largest retailers with the greatest bargaining power, so any individualized differences between the class members did not render the model automatically unreliable to show classwide overcharge.

Having confirmed that plaintiffs' representative sampling method could be used to support a finding that common questions predominate, the Court then held that the district court abused its discretion in failing to “resolve any factual disputes necessary to determine whether predominance has in fact been met. . . . In other words, the threshold predominance determination cannot be outsourced to a jury.” A predominance analysis requires a court to determine whether plaintiffs' proposed classes and supporting statistical evidence includes uninjured class members, and if a substantial number of class members are uninjured, there exist individual issues unique to those members that defeat predominance.

After laying this groundwork, the Court concluded that the district court should have resolved the disputes between the competing experts. The district court did not need to wholesale accept one expert's opinion, but determining that plaintiffs' analysis was “plausibly reliable” was not sufficient. The district court needed to make a factual finding regarding whether common issues predominated: namely, the district court's decision to leave the reliability of plaintiffs' expert modeling to the jury is precisely what cannot be “outsourced to a jury.” As the Court noted, “resolving this dispute is of paramount importance to certification of the class.” Failing to resolve whether the representative evidence presented included 5.5% or 28% uninjured class members was an abuse of discretion.

The Court was then careful to note that it has not “established a threshold for how great a percentage of uninjured class members would be enough to defeat predominance,” relying instead in the oft-cited “de minimis” standard. However, the Ninth Circuit looked to other circuit courts for guidance on what can be considered de minimis. The D.C. Circuit, for example, has suggested that 5% to 6% is the “outer limits” of de minimis, while the First Circuit has found that common evidence cannot predominate in a proposed class with more than 10% uninjured class members. The Court concluded that under any of these prior cases, “the inclusion of 28% uninjured class members would unquestionably defeat predominance.” The district court's failure to meaningfully address the stark difference between plaintiffs' and defendants' statistical evidence regarding uninjured class members was an abuse of discretion.

In a dissenting opinion, Judge Hurwitz generally agreed with the majority decision, but cautioned that the court “should not legislate from the appellate bench based on [the judges'] personal concerns with the class action device.” Judge Hurwitz concluded that fact-based decisions should be left to the discretion of the district court, which is in a better position to determine what should be considered de minimis – “the critical question is not what percentage of class members is injured, but rather whether the district court can economically ‘winnow out' uninjured plaintiffs to ensure they cannot recover for injuries they did not suffer.”


The Ninth Circuit's decision is the latest among a rising tide of circuit courts to opine on the use of statistical evidence to establish predominance in antitrust class actions. We anticipate further battles over just how many uninjured class members is too many. While the Ninth Circuit was careful not to define what it considered the outer bounds of de minimis, it cited favorably cases from around the country that, when taken together, suggest that a double-digit percentage of uninjured class members is too much to satisfy the predominance requirement. The Court made clear that such disputes must be resolved by the district court instead of punting to the finder of fact at trial. In doing so, the Ninth Circuit indicated that it expects district courts to delve deeper into fact issues at class certification and resolve disputes between the experts. District courts will now have to decide how much fact finding will be enough when making predominance determinations.

In the aftermath of this decision, we could see motions to decertify from defendants in district courts within the Ninth Circuit, as well as more appeals.

The Court also left the door open door for Article III standing challenges based on the number of uninjured plaintiffs – the Court noted in a footnote that it was “skeptical” that Article III would permit certification when there are countless unnamed class members that lack standing due to no injury. The Court determined that this was an issue that it did not need to reach in this particular opinion, but issues of standing are likely to appear more often in class certification oppositions filed in the future.

One thing is clear: the importance of statistical econometric analyses at class certification, in particular by defendants to identify uninjured class members as a way to defeat predominance, is even more important now than it already was. The battle of the experts continues.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.