Trial Plans, Sampling/Representative Evidence, Bifurcation, and Bellwether Trials

A client engages you to defend against a new employment class action, in an area of law that you know well. The legal issues seem clear enough, with well-established principles for both sides to argue liability, damages and class treatment. But then discovery commences and you realize that the case is more complicated than it appeared at first blush. And the further along the case progresses, the more you realize that, if the plaintiffs’ proposed class becomes certified (and also is not decertified before trial), both sides will have some big challenges, especially if it’s a jury trial. What will the plaintiffs attempt to do at trial, and how do they think they will prove it? Will the plaintiffs try to prove their claims through sampling or, in the case of FLSA claims, “representative evidence”? Are there good grounds to use bifurcation to break up the case by issue or plaintiff or employment location, to make the trial more manageable and less confusing to the jury? And should you agree to a bellwether trial, in which one issue, or one named plaintiff or worksite, will serve as a “test case” that will establish some basic principle(s) to govern the resolution of the remaining claims in the case?

Many management-side counsel do not give attention to these issues early enough in a case, perhaps on the assumption that most cases settle before trial. But you should raise all of these issues as early as possible, first considering what would be in the best interest of your client and the court, and then making proposals to the plaintiffs’ counsel to see if they agree. These issues could be raised as early as the Rule 26(f) conference, since agreement on them (or, alternatively, some guidance from the court) could permit the parties to frame their discovery and pretrial actions accordingly. Ultimately, pursuant to Federal Rule of Civil Procedure 1, the goal is to develop a plan for the most efficient way of resolving what otherwise might turn out to be a morass of a trial.

This paper discusses four tools for managing the trial of an employment class action (or Secretary of Labor enforcement action) with an eye toward how the claims will be adjudicated at trial. They are:  1) asking the court to require the plaintiffs to file a trial plan, in which they inform the defendant and the court how they intend to prove their case in court; 2) the use of sampling and “representative evidence” to establish class-wide liability and/or damages based on evidence that is obtained from a relatively small portion of the workforce; 3) bifurcation of the trial pursuant to Rule 42(b), such that certain issues or the claims of one or a few of the named plaintiffs are adjudicated before others; and 4) the use of one or more “bellwether” trials to establish certain principles that will provide a consensus for settlement.          

Submission of a Trial Plan By The Plaintiffs 

Federal Rule of Civil Procedure 16(e) states that a “court may hold a final pretrial conference to formulate a trial plan ….” But for good reason, plaintiffs and courts do not always wait until the final pretrial conference to consider this issue. In class actions, “[a]n increasing number of courts require a party requesting class certification to present a ‘trial plan’ that describes the issues likely to be presented at trial and tests whether they are susceptible of class-wide proof.” Fed. R. Civ. P. 23, Advisory Committee Note (2003 amendments). See In re Hydrogen Peroxide Antitrust Litig., 552 F.3d 305, 320 (3d Cir. 2008); Mwantembe v. TD Bank, N.A., 268 F.R.D. 548 n.44 (E.D. Pa. 2010) (same).

Trial plans have been used in many employment class actions.¹ See, e.g., Lowery v. Circuit City Stores, Inc., 158 F.3d 742, 754 & n.2 (4th Cir. 1998) (discussing trial plan in African-Americans’ Title VII and § 1981 action; plaintiffs’ trial plan proposed a Stage I for jury to decide classwide liability and punitive damages, along with liability and compensatory damages on the named plaintiffs’ individual claims; Stage II would determine compensatory damages for class members; Drake v. Steak N Shake Ops., Inc., 286 F. Supp. 1040, 1050 (E.D. Mo. 2017) (employees brought putative collection action seeking overtime wages under FLSA and Missouri law and filed a trial plan proposing a liability phase and a damages phase, with the first phase involving testimony from two named plaintiffs and 5-7 other collective action members addressing their number of hours and job duties; second phase on damages would include expert testimony; court found the trial plan sufficient, but would consider later requests for modification from the employer); Thompson v. Bruister and Assocs., Inc., 967 F. Supp. 2d 1204, 1210, 1217 (M.D. Tenn. 2013) (cable technicians who pursued an FLSA action for minimum wages and overtime filed a trial plan proposing that not more than “60 representative plaintiffs” who worked out of several of the employer’s offices would testify, who would be selected randomly from among the named, discovery, and non-discovery plaintiffs); Colindres v. Quietflex Mfg., 235 F.R.D. 347, 377 (S.D. Tex. 2006) (discussing trial plan in Title VII and § 1981 action brought by present and former Latino employees alleging pattern or practice of discriminatory treatment and retaliation; plaintiffs’ trial plan proposed a Stage I that would determine classwide liability and injunctive/declaratory relief and punitive damages; Stage II would determine individual back pay awards); Burrell v. Crown Central Petroleum, Inc., 197 F.R.D. 284, 292 (E.D. Tex. 2000) (discussing trial plan in Title VII and § 1981 action alleging race and sex discrimination, in which they proposed to certify and try only classwide liability, followed by individual jury trials on damages; trial plan convinced court that class action was not superior to other methods for adjudicating claims).

In addition, at least one court has likewise found that a trial plan is appropriate in an FLSA enforcement action to address before trial deficiencies in the Secretary of Labor’s use of representative evidence:

DSHS argues that the Court should “summarily reject DOL’s proposed representative testimony or set an early hearing to determine its validity.” The Court agrees with DSHS to the extent that DOL appears to be trying to fit a square peg into a round hole or, at the very least, has failed to identify how the disclosed witnesses will adequately represent all of the claimants. At some point, the Court will request a trial plan, but only after sufficient notice and an opportunity to be heard on the subject.

Solis v. State of Washington Dep’t of Soc. And Health Servs., No C08-5479 BHS, 2016 WL 879166, at * 6 (W.D. Wash. Mar. 8, 2016).  That case involved just two categories of jobs, unlike many other employment class actions that raise issues as to multiple jobs, pay practices, and/or plant environments. See also Dole v. Haulaway Inc., 723 F. Supp. 274 (D.N.J. 1989) (noting court’s regret that it did not require DOL to address how it would prove its case through representative evidence before trial and explaining how failure to do so led to need for trial adjournment and re-scheduling). 

Issues with sampling and the use of representative evidence can both be addressed through a trial plan, as well as how plaintiffs or the DOL intend to use an expert’s opinion as part of their proofs.  This latter consideration bears pretrial scrutiny beyond a Daubert motion. If the plaintiffs do not file a trial plan, the alternative may be to ask the court for a “mini trial” on certain issues to determine if the claims can be proved on a class-wide basis. For example, in one private FLSA collective action seeking compensation for donning and doffing activities based on a combination of a time study and calculation of the difference between employees’ time clock punch times and paid time (“gap time”), the court presided over a multi-day evidentiary hearing and then decertified the collective action, finding that the plaintiffs could not meet their burden of showing that their theory of noncompliance could be adjudicated through representative evidence.  See Lugo v. Farmer’s Pride, Inc., 737 F. Supp. 2d 291, 309-310; 316 (E.D. Pa. 2010) (finding expert’s gap time analysis had “little probative value” and his time study provided “no direct insight regarding the inherent adequacy or inadequacy of the compensation provided by defendant for donning and doffing” because of differences among departments). 

In the class action context, the specter of testimony from hundreds of class members argues against trying all such claims together unless there is some other manner of proving the claims collectively (such as representative evidence, discussed infra).  As the Sixth Circuit has reasoned in a non-employment case:

The district court took testimony from more than three hundred class members in an effort to obtain a purportedly representative sample of the representations and communications made by GM. That it was necessary to do so strongly suggests to us that class-wide relief was improper.

Sprague v. Gen. Motors Corp., 133 F.3d 388, 399 (6th Cir. 1998).  And in Broussard v. Meineke Discount Muffler Shops, Inc., 155 F.3d 331 (4th Cir. 1998), ten franchisees obtained certification of a nationwide class, and during a seven-week trial offered, among other evidence, 171 taped excerpts from Meineke’s representatives.  The Fourth Circuit, citing Sprague, reversed class certification, noting:  “We are struck by the sheer number of separate statements that were put before the jury to prove a ‘common’ message, and find the Sprague court’s rationale for refusing class certification in a similar situation persuasive”.  Id. at 341.

These same principles apply to employment claims.  Thus, for example, in Vega v. T-Mobile USA, Inc., 564 F.3d 1256 (11th Cir. 2009), a putative class action involving claims for breach of employment contact for sales commissions, the Eleventh Circuit vacated class certification and recommended that district courts make it a usual practice to require trial plans:

We do not mean to say that submission of a trial plan by the plaintiff is necessarily a prerequisite, as a matter of law, for a finding of superiority in every case. Nonetheless, a plaintiff seeking class certification bears the burden of establishing each element of Rule 23, which includes superiority in Rule 23(b)(3) cases, and courts must consider how a case will be tried as part of the superiority assessment. Accordingly, the proposal of a workable trial plan will often go a long way toward demonstrating that manageability concerns do not excessively undermine the superiority of the class action vehicle. Moreover, there is a direct correlation between the importance of a realistic, clear, detailed, and specific trial plan and the magnitude of the manageability problems a putative class action presents. We therefore recommend that district courts make it a usual practice to direct plaintiffs to present feasible trial plans, which should include proposed jury instructions, as early as practicable when seeking class certification.

Id. at 1279 n.20 (citations omitted).  The Eleventh Circuit then went on to state that, despite individualized claims and defenses that were proposed to be tried, the named plaintiff  “does not appear to have given meaningful consideration to how this case … would be tried. …  By putting off these issues until trial, or blinking them entirely, both [plaintiff] and the district court ran an unnecessarily high risk of introducing needless and avoidable complexity into an already complex case.”  Id. at 1279-80.

A number of state courts, especially in California, have adopted a similar view of the utility of trial plans to ensure the viability of trying class claims.  See Duran v. U.S. Bank Nat'l Assn., 59 Cal. 4th 1, 27, 325 P.3d 916, 930 (2014) (“Class certification is appropriate only if [the individual questions common in misclassification claims] can be managed with an appropriate trial plan.”); McCleery v. Allstate Ins. Co., 37 Cal.App.5th 434 (2019) (denial of class certification warranted where plaintiffs’ proposed trial plan failed to address individualized issues and deprived defendants the ability to assert defenses); Payton v. CSI Electrical Contractors, Inc., 27 Cal.App.5th 832 (2018) (lower court was within its discretion to reject plaintiff’s proposed trial plan for lack of specifics on procedural tools and management of individualized issues).

Proving Class Claims Through Sampling and Representative Evidence

It has become more common in the last few decades for plaintiffs to attempt to use statistical sampling as a basis for extrapolating liability or damages in class actions. These issues can arise at a variety of stages, including initial class certification, proposed trial plans, Daubert motions, summary judgment, motions to decertify, and motions during and after trial. In three cases, the United States Supreme Court has reviewed the use of statistical evidence at the class certification stage and trial. The principles announced in these cases have had a significant effect on lower-court decisions.

In Wal-Mart Stores, Inc. v. Dukes, 564 U.S. 338, 366–67 (2011), the Supreme Court ruled that Title VII gender discrimination plaintiffs cannot establish liability and damages for a sample set of class members and then apply the percentage of “valid” claims and the “average” damages award to arrive at the entire class judgment (what the Court called “trial by formula”) because doing so would violate the Rules Enabling Act, 28 U.S.C. § 2072. Such an approach manufactures commonality for the class certification stage and trial by masking individual differences that are essential to accurately determine class members’ claims, without affording the employer the right to present statutory defenses through individualized proceedings.

In Comcast Corp. v. Behrend, 569 U.S. 27 (2013), the Supreme Court considered a regression analysis offered by antitrust plaintiffs at the class certification stage to prove class-wide antitrust impact. The plaintiffs originally had pursued four theories of antitrust injury, but the district court had disallowed three of them, leaving only one viable theory of recovery in the case. The plaintiffs’ regression model calculated damages for the entire class, but the model did not isolate damages from any one of the four theories of antitrust impact. The majority ruled that the plaintiffs’ damages model failed to show that damages could be measured on a class-wide basis because it did not measure only those damages attributable to the sole surviving theory of antitrust impact accepted for class treatment. “[A]t the class-certification stage (as at trial), any model supporting a ‘plaintiffs’ damages case must be consistent with its liability case.” Id. at 35 (citations omitted).

The Supreme Court also considered the use of statistical evidence for Rule 23 wage-hour overtime claims in Tyson Foods, Inc. v. Bouaphakeo. 577 U.S. 442 (2016).  In that case, the parties agreed that the same standards of proof also applied to a smaller FLSA collective action at issue in the same trial. “Whether and when statistical evidence can be used to establish classwide liability will depend on the purpose for which the evidence is being introduced and on ‘the elements of the underlying cause of action.’”  Id. at 455. The Supreme Court allowed the plaintiff class to use their expert’s time study based on sampling to prove the average amount of uncompensated time because, the Court reasoned, any employee would be entitled to use the same type of evidence to prove his or her individual liability claim. Due to the federal labor policy of the FLSA to not punish employees for an employer’s failure to keep accurate records, the Supreme Court allowed the employees to use the time-study average, in conjunction with individualized time records for time worked on the line, to prove their damages by a just and reasonable inference, which the employer could then rebut with more precise information.  See Anderson v. Mt. Clemens Pottery Co., 328 U.S.  680 (1946).  At the same time, however, the Supreme Court cautioned in Bouaphakeo:  “This is not to say that all inferences drawn from representative evidence in an FLSA case are ‘just and reasonable.’  Representative evidence that is statistically inadequate or based on implausible assumptions could not lead to a fair or accurate estimate of the uncompensated hours.”  577 U.S. at 459 (citing Mt. Clemens Pottery). The employer in Bouaphakeo had not made a Daubert challenge to the plaintiffs’ time-study expert, and this language essentially encourages employers to raise such a challenge in all future wage-hour cases where the plaintiffs intend to rely upon an expert’s measurements.

Before this trilogy of Supreme Court cases, there was no majority rule on whether statistical sampling could be used in class actions to prove liability and damages, and most cases discussing the issue contained scant analysis. Those courts that did allow statistical sampling generally emphasized its efficiency, while courts rejecting it typically did so on due process grounds or because they concluded that doing so impermissibly permitted the class to shift the burden of proof or avoid proving the elements of the claim on a class-wide basis. Certain types of cases—such as antitrust cases and certain employment claims—were generally regarded as being more receptive to statistical sampling, while other types of cases—such as mass personal injury claims (e.g., asbestos or tobacco cases)—were generally regarded as being inappropriate for statistical sampling.  See, e.g., Tex. Dep't Hous. Cmty. Affairs v. Inclusive Cmtys. Project, Inc., 576 U.S. 519, 543 (2015) (“A plaintiff who fails to allege facts at the pleading stage or produce statistical evidence demonstrating a causal connection cannot make out a prima facie case of disparate impact.”); Aberman v. Bd. of Educ. of City of Chicago, 242 F. Supp. 3d 672, 685 (N.D. Ill. 2017) (noting that disparate impact discrimination plaintiffs can “establish causation by offering statistical correlation evidence demonstrating that a specified employment practice of the defendant has a disproportionately negative effect on members of the plaintiff’s protected class”) (internal quotation marks omitted).   

One interesting wage-hour case involving “sampling” of employees is Reich v. IBP, Inc., Civ. No. 88-2171, 1996 WL 137817 (D. Kan. Mar. 21, 1996).  In that “donning and doffing” case, the DOL represented the interests of 23,580 employees.  It sent them all a questionnaire.  Of the knife-using employees (the only ones who performed compensable activities), 4,331 returned the survey and DOL randomly selected 378 (8.7%) for deposition.  DOL took the time estimates from those depositions and presented an average to the court, which ruled on damages as trier-of-fact because the employer had not demanded a jury trial. Despite this Herculean project, the court rejected DOL’s proffered number for a variety of reasons, including the following:  “Even ignoring the employees’ personal interest in the case, we question the employees’ ability, under the circumstances, to accurately recall the amounts of time involved in each activity.  Moreover, even assuming the employee’s estimates are accurate to a degree, they are still estimates of the actual time involved, rather than the reasonable time necessarily required.  In addition, our review of the record reveals several instances where time for noncompensable activities could not be separated from the aggregate time estimates given by employees.”  Id. at *3.  The court went on to explain other reasons why such time estimates are faulty, including the large number of clear overstatements of time and relatively few understatements of time.  Id. at *4.

After the Supreme Court’s opinions on the use of statistics in class actions in the above-mentioned three cases, however, there appear to be a few general rules for using statistics to prove liability on a class-wide basis. These rules apply at all phases of the case, including the class certification stage and trial: (1) the ability to offer statistics on a liability determination will depend on the elements of the underlying cause of action and whether the statistics are reliable in proving or disproving a common answer to one or more of those elements; (2) statistical sampling to prove liability will not be allowed if individualized proofs could explain away a common conclusion to be drawn from the statistics; (3) the population must be sufficiently similar, or the statistics must be based on a sufficient measure and subsets of the population so that the results are probative of the class-wide answer without making implausible assumptions; (4) if the statistics could be used by an individual plaintiff to prove his or her claim in an individual action, then they likewise may be offered in a class action; (5) if such evidence is allowed to be offered by the plaintiffs, the defendant still has a right to offer individualized evidence in support of its defenses in addition to raising a generalized challenge that would cause the entirety of the plaintiffs’ class-wide proofs to fail on a common basis; and (6) once a class action is tried, it can be decertified on the basis of the statistical proofs only if no reasonable juror could believe there is an approximately common answer supplied by the statistics.

As for the use of statistics to prove damages, the Supreme Court’s decisions also suggest a few general rules. Again, these rules apply at all phases of the case: (1) if the underlying substantive law requires a separate inquiry into damages for each class member, the defendant may challenge class-wide damages proceedings on that basis; (2) class-wide damages cannot be determined by merely calculating the average damages from a subset of the class and then applying that average to the entire class; (3) a model for class-wide damages must allow a just and reasonable inference and must not be speculative or arbitrary; (4) the statistics must purport to measure the effect of conduct (and only the conduct) that is allegedly unlawful and creates liability; and (5) a defendant might waive its right to the foregoing rules (e.g., by failing to raise a Daubert challenge, by offering certain jury instructions, by opposing individualized determinations, or by agreeing to a general verdict form).

The federal circuit courts disagree about whether plaintiffs may rely on damage models that use a uniform measure of harm derived from an average or a regression analysis. Some of these decisions came before the Supreme Court’s trilogy, and others are more recent. For a more complete discussion of the case law, see Michael Mueller & Jason Beach, A Practitioner's Guide to Class Actions, Chapter 20, Statistical Sampling as a Basis for Extrapolating Liability and Damages (American Bar Association Tort Trial & Insurance Practice Section, 3d edition, 2021).

As noted above, however, there is a special rule for FLSA collective action claims (and DOL enforcement actions) where an employer has failed to keep adequate records.  See Mt. Clemens Pottery, supra.  A wide range of types of evidence might be offered by plaintiffs or the Secretary as so-called “representative evidence” in an effort to shift the burden onto the employer at trial.  This includes 1) hourly employee testimony; 2) management witnesses’ testimony; 3) expert testimony (often a time-study, but also including things such as geolocation tracking surveys of employees or truck in-out or dispatch records); 4) employee declarations, if they qualify for admission under a hearsay rule such as Fed. R. Evid. 801(d)(2)(D); 5) videotape evidence of employee activities (such as surveillance camera footage); 6) time-clock or punch-in/out records; and 7) pay policies and other corporate records that are being offered to prove employee duties or estimates of time worked. See, e.g., Su v. East Penn Mfg. Co., Civ. No. 18-1194, 2023 WL 7336368, at **4-8 (E.D. Pa. Nov. 7, 2023).

There is no clear definition of what qualifies as “representative evidence,” and in a trial that determination is one for the jury, just as the “preponderance of the evidence” question is always one for the jury.  But typically a court will look to whether the evidence being offered is qualitatively capable of proving something extending beyond the employee or group from which it is drawn such that it tends to prove the same thing for the entire class.  There is also a question of whether the evidence being offered is quantitatively sufficient. In Mt. Clemens Pottery, the Supreme Court upheld the evidence as being “representative” when eight current and former employees testified on behalf of approximately 300 employees.  See Mt. Clemens Pottery Co. v. Anderson, 149 F.2d 461, 462 (6th Cir. 1945), aff’d in relevant part and rev’d in other part, 328 U.S.  680.  That amounted to 2.7 percent of the workforce, which has become a loose litmus test of sorts for the minimum amount of employee testimony that is required.  Courts have rejected employees’ testimony as insufficiently “representative” where the percentage is lower than this, and have even rejected much higher percentages where the witnesses who testified did not adequately cover the range of subgroups at issue.  See Reich v. S. Md. Hosp., Inc., 43 F.3d 949 (4th Cir. 1995) (remanding with observation that 1.6% of employees “constitutes the lowest percentage by far in any reported case using representative testimony under the proof scheme of Anderson v. Mt. Clemens Pottery Co.”); Reich v. Gateway Press, Inc., 13 F.3d 685, 702 (3d Cir. 1994) (in case where 22 of 70 employees, or 31.4%, had testified, remanding with directive that trial court should “carefully scrutinize” the evidence before allowing or denying back wages to non-testifying employees); Secretary of Labor v. DeSisto, 929 F.2d 789, 793 (1st Cir. 1991) (one of 244 employees, or 0.41%, was inadequate to represent employees holding a variety of positions at different locations); Marshall v. Truman Arnold Distrib. Co., 640 F.2d 906, 910–11 (8th Cir. 1981) (upholding back wage award to 17 of 25 employees who testified, or 68%, but remanding for further proceedings as to amount and extent of work performed by 8 non-testifying employees).

Bifurcation of Issues to Make Trial More Manageable

Pursuant to Federal Rule of Civil Procedure 42(b), “For convenience, to avoid prejudice, or to expedite and economize, the court may order a separate trial of one or more separate issues, claims, crossclaims, counterclaims, or third-party claims.”  However, “When ordering a separate trial, the court must preserve any federal right to a jury trial.”  Id.

Courts often bifurcate trials to avoid the presentation of unnecessary evidence when the need for the second “phase” is predicated on the outcome of the first.  In the employment context, there are two (or perhaps three) natural dividing points:  one is between the liability and damages determinations, and another is between a determination of liability and punitive damages or other related concepts such as, in FLSA cases, the employer’s state of mind (willfulness) and liquidated damages (good faith).  See, e.g., Goldman v. Radioshack Corp., No. Civ.A.03-0032, 2005 WL 1155751, at *1 (E.D. Pa. May 13, 2005) (deeming bifurcation into liability and damages / willfulness phases “appropriate in this [FLSA and state law wage] case because it will promote judicial economy by limiting the presentation of lengthy evidence which may be irrelevant if the jury does not find RadioShack liable”); McGrath v. City of Philadelphia, 864 F. Supp. 466, 490 (E.D. Pa. 1994) (noting that parties in FLSA case agreed to bifurcate the issues of liability and damages); contra Lowe v. Philadelphia Newspapers, Inc., 594 F. Supp. 123, 125 (E.D. Pa. 1984) (citations omitted) (bifurcation is appropriate “where the court concludes that separate liability and damage trials further convenience or avoid prejudice.”; denying bifurcation in § 1981 race discrimination case because damages evidence was involved in proof of liability).  See generally cases discussed supra in section about trial plans.

Non-employment cases also provide authority on the rationale for bifurcation (or even trifurcation).  See, e.g., Franklin Music Co. v. Am. Broad. Cos., Inc., 616 F.2d 528, 538 (3d Cir. 1979) (upholding bifurcation decision where the trial “could have been considerably shorter, had the jury decided in favor of the defendants in the liability phase”); Lontex Corp. v. Nike, Inc., No. 18-5623, 2021 WL 2138621, at *2 (E.D. Pa. May 26, 2021) (bifurcating trial into liability and damages phases because “[t]here is great benefit to the jury, if it decides liability in favor of Nike, as it will not have to spend an extra three to five days hearing damages testimony”); Laboratory Skin Care, Inc. v. Limited Brands, Inc., 757 F. Supp. 2d 431, 442 (D. Del. 2010) (bifurcating liability from damages / willfulness issues because “[j]udicial resources may be conserved through bifurcation, as liability may not be found, or even if found it may simplify the subsequent damages and willfulness trial”); Tabas v. Tabas, Civ. No. 91-1355, 1996 WL 107848, at *2 (E.D. Pa. Mar. 12, 1996) (trifurcating trial into federal law, state law, and damages phases in part because “the possibility of a defense verdict on the liability phase” could “spare both sides the expense of presenting” damages evidence, despite overlapping evidence in state and federal phases); 9A Wright & Miller Fed. Prac. & Proc. § 2388 (“If a single issue could be dispositive of the case … separate trial of that issue may be desirable to save the time of the court and reduce the expenses of the parties.”).

In Bouaphakeo, discussed supra, the Supreme Court essentially encouraged employers in future wage-hour cases to request bifurcation of the damages phase from the liability determination, to avoid the possibility that uninjured employees might be included in a jury’s determination of liability and a lump-sum award to the entire workforce. In that case, the employer challenged the jury’s damages award as potentially awarding wages to class members who could not prove injury (i.e., unpaid overtime), as it was unclear from the jury’s verdict how many uncompensated minutes the jury had found.  In upholding the judgment, the Supreme Court faulted the employer for not agreeing to bifurcate liability and damages issues, which would have avoided the problem of ascertaining uninjured class members. See 577 U.S. at 461-62 (“Respondents [plaintiffs] proposed bifurcating between the liability and damages phases of this proceeding for the precise reason that it may be difficult to remove uninjured individuals from the class after an award is rendered. It was petitioner [the employer] who argued against that option and now seeks to profit from the difficulty it caused.”).  More recently, though, the Supreme Court has made a stronger statement that the federal courts lack the power to enter a judgment that includes uninjured class members.  See TransUnion LLC v. Ramirez, 594 U.S. 413, 425-31 (2021) (Article III does not give federal courts the power to award relief to uninjured persons) (citing Bouaphakeo, 577 U.S. at 466Roberts, C.J., concurring)).

An important consideration for bifurcating is whether doing so will enhance juror comprehension of the pertinent legal standards while preventing undue prejudice to a party, such as hearing evidence on damages before liability has been established. See Barr Labs., Inc. v. Abbott Labs., 978 F.2d 98, 115 (3d Cir. 1992) (upholding bifurcation where district court determined non-moving party would not suffer prejudice); Murillo v. Coryell Cty. Tradesmen, LLC, No. 15-3641, 2017 WL 2735559, at *5 (E.D. La. June 26, 2017) (bifurcating FLSA action and noting that “the issues of damages and liability here are so distinct and separable that a bifurcated trial may be had without injustice or prejudice to any party” (internal quotations omitted); Laboratory Skin Care, 757 F. Supp. 2d at 442 (bifurcating liability and damages / willfulness where “[t]he burden on the jury, and risk of juror confusion, will [] be reduced by trying these issues separately”); Lontex Corp., 2021 WL 2138621, at *1 (“[I]t will be much more straightforward for the Court and the jury, to have the jury’s answers to the liability interrogatories before any evidence is introduced as to damages.”); Goldman, 2005 WL 1155751, at *1-2 (“[B]ifurcation is appropriate because …  [t]he issue of liability and the issue of damages are legally distinct .… [T]he intermingling of liability with issues of willfulness and damages might prejudice RadioShack.”). From an employer’s perspective, postponing evidence of state of mind for a second phase could eliminate (or at least reduce) the risk that the jury will punish the employer by increasing its findings for the determined amount of uncompensated time spent on the work activities. Cf. Lontex, 2021 WL 2138621, at *2 (“This Court believes that, considering liability alone without having heard any damages testimony, the jury can still be fair to both sides and render a verdict based on the evidence.”).

However, courts are more hesitant to bifurcate where doing so may require the parties to present the same evidence twice. For example, the Third Circuit has admonished to “divide issues between separate trials in such a way that the same issue is not reexamined by different juries.” In re Citizens Bank N.A., 15 F.4th 607, 620 n.15 (3d Cir. 2021); see also Idzojtic v. Pennsylvania R. Co., 456 F.2d 1228, 1230 (3d Cir. 1972) (affirming district court’s bifurcation decision where jury was able to determine the liability issue apart from the damages issue). See generally Zinno v. GEICO Gen. Ins. Co., Civ. No. 16-792, 2016 WL 6901697, *2 (E.D. Pa. Nov. 21, 2016) (denying a motion to bifurcate where evidence was overlapping and claims were similarly intertwined); RadioShack, 2005 WL 1155751, at *1 (bifurcating FLSA action into liability and damages phases because different standards and evidence applicable at each phase); Laboratory Skin Care, 757 F. Supp. 2d at 442 (bifurcating liability and willfulness / damages).

Bifurcation also may be useful and appropriate where there are numerous complex situations, such as multiple work locations at issue in the same case, each presenting different facts.  See Franklin Music Co., 616 F.2d at 538 (bifurcation was not abuse of discretion where case was complex, including four parties, 73 live witnesses, and over 300 exhibits); cf. Klimaski v. Parexel Int’l, No. Civ.A. 05-298, 2005 WL 857350, at *3 (E.D. Pa. Apr. 4, 2005) (addressing analogous question of joinder and concluding joinder was not warranted even though former employees’ claims shared same theory of retaliatory discharge and similar allegations of improper activities, because the circumstances of each employee’s termination were “factually distinct”).

Courts have also bifurcated FLSA enforcement actions brought by the Secretary.  See Donovan v. DialAmerica Mktg., Inc., 757 F.2d 1376, 1381 n.2 (3d Cir. 1985) (noting that, in FLSA action, “[c]ounsel and the court had already agreed to bifurcate the case. First, the court would determine whether the workers in question were ‘employees’ under the FLSA .… [If so,] the court would then consider whether they were entitled to backpay or other relief. Any action on the second part of the case was to be suspended until the first part was completed”); Donovan v. United Video, Inc., 725 F.2d 577, 579 (10th Cir. 1984) (FLSA action bifurcated into two phases: 1) “considering first the issue whether [employer] was obligated to pay overtime compensation to the employees in question”; and 2) “determin[ing] the amount of overtime compensation owed”); Reich v. IBP, Inc., No. CIV. A. 88-2171, 1996 WL 137817, at *1 (D. Kan. Mar. 21, 1996), aff’d sub nom. Metzler v. IBP, Inc., 127 F.3d 959 (10th Cir. 1997) (FLSA action bifurcated into two phases; in Phase I, the court determined that certain pre-shift and post-shift activities by workers at defendant’s plants were compensable; next, Phase II involved the following issues: (1) the amount of time spent per day by employees performing compensable activities; (2) whether the time involved was de minimis; (3) whether pre-judgment interest should be awarded; and (4) whether the court should permanently enjoin defendant from future violations); Donovan v. Star Bakery, Inc., 626 F. Supp. 1208, 1210 (D.P.R. 1986) (FLSA trial bifurcated, without opposition by Secretary, for sole trial on issue of liability; complaint dismissed after liability trial because employer was not deemed “enterprise” within meaning of FLSA); Marshall v. Suicide Prevention of Fla., No. WPB-76-8339-Civ-CF, 1977 WL 1766, at *1, *8 (S.D. Fla. Aug. 1, 1977) (non-jury trial bifurcated into liability and damages phases; counsel agreed that magistrate judge would “determine the issue of damages and to recommend to the court the specific amount of damages due to the various employees of the defendants”).  

Bellwether Trials        

Rule 42(b) also permits, as part of the “bifurcation” decision, a determination to try the claims of a single issue or certain class members in an effort to resolve common issues and increase the likelihood of settlement.  See In re Chevron U.S.A., Inc., 109 F.3d 1016, 1019–20 (5th Cir. 1997) (“[T]he results of such trials can be beneficial for litigants who desire to settle such claims by providing information on the value of the cases as reflected by jury verdicts. Common issues or even general liability may also be resolved in a bellwether context in appropriate cases.”); Grant Heilman Photography, Inc. v. McGraw-Hill Companies, 115 F. Supp. 3d 518, 522 (E.D. Pa. 2015) (“Bellwether trials … are recognized as an effective means for a trial judge to enhance settlement prospects or resolve common issues or claims in complex litigations.); see also In re Tylenol (Acetaminophen) Mktg., Sales Practices, and Prods. Liab. Litig., No. 12–7263, MDL No. 2436, 2015 WL 2417411, at *1 & n. 3 (E.D. Pa. May 20, 2015) (“A ‘bellwether case’ is a test case. ‘Bellwether’ trials should produce representative verdicts and settlements.”).”).

Courts have used the bellwether approach in the context of multi-employee cases.  See, e.g., Zapata v. IBP, Inc., 1997 WL 614427, *2 (D. Kan. Sept. 22, 1997) (in case where 14 Mexican-American or native-born Mexicans alleged that employer engaged in a pattern or practice of discrimination based on national origin, ancestry, ethnicity and race, court stated: “When we read Federal Rule of Civil Procedure 21 in conjunction with Rule 42(b), we find that two trials will best serve the convenience of the parties and the court and minimize expense and delay. We believe this approach is more likely than fourteen separate trials to result in a just and expeditious final disposition of the litigation.”); In re Tyson Foods, Inc., 694 F. Supp. 2d 1372, 1380 (M.D. Ga. 2010) (trying, as a bellwether trial, one test “donning and doffing” FLSA case out of consolidated MDL involving approximately 40 poultry processing plants). 

In deciding whether to order a bellwether trial, a court must consider whether it would violate the employer’s Seventh Amendment right to have a jury decide fact issues, including whether the employer owes back pay to employees at locations beyond the one that is the subject of the bellwether trial and, if so, how much.  See Cimino v. Raymark Indus., Inc., 151 F.3d 297 (5th Cir. 1998) (seminal case holding that trial by extrapolation to establish damages for all members of class violated company’s Seventh Amendment right where only some workers’ damages were established in “Phase III” trial, but the defendant had 22 sites and therefore had a right to have the amount of the legally recoverable damages for those additional workers decided by a jury); Alexandra D. Lahav, Bellwether Trials, 76 Geo. Wash. L. Rev. 576, 581 (2008) (acknowledging that courts no longer “experiment[] with binding bellwether trials … perhaps due to the influence of” Cimino); Dunson v. Cordis Corp., 854 F.3d 551, 555 (9th Cir. 2017) (explaining presumption that, unless the parties agree to be bound, bellwether trial results “will not be binding” on the parties in later proceedings but will instead be used for informational purposes only; also explaining that this is the “far more common” type of bellwether trial).  Courts have long forbidden binding bellwether trials for claims requiring individualized proof of liability and damages because the Seventh Amendment guarantees parties a jury determination thereof.  Cimino, 151 F.3d at 320 (“[W]here the issues to be separately tried are separable and distinct, the Seventh Amendment rights of the parties are preserved as to both sets of issues.”). In Cimino, the Fifth Circuit expressly disavowed earlier dicta implying that “representative bellwether verdicts could properly be used to determine individual causation and damages for other plaintiffs” because the law at issue required individual determination of causation and damages.  Id. at 318-19.

Other courts to have considered whether a bellwether trial can bind a party on a non-common issue to which that party’s claim was not evaluated have come out the same way.  See, e.g., In re TMI Litig., 193 F.3d 613, 725 (3d Cir. 1999) (“[A]bsent a positive manifestation of agreement by Non–Trial Plaintiffs, we cannot conclude that their Seventh Amendment right is not compromised by extending a summary judgment against the Trial Plaintiffs to the non-participating, non-trial plaintiff.”); Silivanch v. Celebrity Cruises, Inc., 333 F.3d 355, 359 (2d Cir. 2003) (describing procedure in which bellwether bound non-parties on common issues but not issues of proximate cause and compensatory damages, which would be tried separately for each party). 

Despite these limitations on the ability of the outcome of the first trial to bind the employer, parties should consider the efficiencies that a bellwether trial may provide.  For  example, if after one or two trials it appears likely that subsequent juries would rule the same way on liability as to different work locations or subgroups of the employee population, the parties would have a basis to enter into an informed settlement rather than continuing to try the remaining claims. 

---

[1] A WESTLAW search on October 1, 2024, located 51 cases, in the database for all federal courts, mentioning “trial plans” in Title VII, Section 1981, ADEA and FLSA cases.  Undoubtedly there are more, including cases under other employment statutes and common-law claims.

©2024. Published by the American Bar Association. Reproduced with permission. All rights reserved. This information or any portion thereof may not be copied or disseminated in any form or by any means or stored in an electronic database or retrieval system without the express written consent of the American Bar Association or the copyright holder.