BROWN, Circuit Judge:
Over the last decade, the four major freight railroads imposed rate-based fuel surcharges on shipments over their tracks. Although the practice had existed for some time, it proliferated and intensified early last decade. Suspecting foul play, a group of shippers who paid these surcharges brought an antitrust suit accusing the freight railroads of engaging in a price-fixing conspiracy. They also sought and obtained certification of a class including all similarly situated shippers who paid these surcharges during the relevant period. The freight railroads now seek, via interlocutory appeal, to undo class certification, the crux of their argument being that separate trials are needed to distinguish the shippers the alleged conspiracy injured from those it did not. Satisfied that this case is among the rare instances in which interlocutory review of a certification decision is warranted, we exercise our discretion to hear this appeal.
Four companies account for nearly 90% of rail freight traffic: BNSF Railway Co. (BNSF); CSX Transportation, Inc. (CSX); Norfolk Southern Railway Co. (NS); and Union Pacific Railroad Co. (UP). See In re Rail Freight Fuel Surcharge Antitrust Litig. (Fuel Surcharge I), 587 F.Supp.2d 27, 29 (D.D.C.2008). In some regions, the railroads' networks overlap. In others, tracks may belong almost exclusively to a single railroad. A sizable percentage of shipping traffic over the four railroads' tracks is "interline," i.e., serviced by multiple railroads,
To offset fuel costs, freight railroads often include fuel surcharges on top of the base rates they charge their customers. These fuel surcharges have traditionally taken two forms. Mileage-based fuel surcharges raise total rates in proportion to shipping distances. Rate-based fuel surcharges, by contrast, depend on a prearranged "strike" or "trigger" price. When fuel prices are below the trigger price, no fuel surcharge supplements the base rate. But once fuel prices exceed the trigger price, a surcharge is imposed as a function of the base rate. Together, the fuel surcharge and base rate constitute the total rate paid (sometimes called the "all-in" rate).
Rate-based fuel surcharges were not unheard of at the start of the new millennium, but neither were they the norm. That all changed by the mid-2000s, when fuel surcharge provisions became ubiquitous, governing the vast majority of the defendants' shipments. At the same time, the defendants sharpened the surcharges' sting, with all four dropping their trigger prices between March 2003 and March 2004. Not all shippers were affected, though. Some had entered into so-called legacy contracts with the defendants before this period, thereby guaranteeing they would be subject to fuel surcharge formulae that predated the later changes.
The heyday of the rate-based fuel surcharge did not last. Eventually, the Surface Transportation Board (STB) put an end to the practice with respect to common carrier traffic within its regulatory authority. See Rail Fuel Surcharges, Ex Parte No. 661, 2007 WL 201205 (S.T.B. Jan. 25, 2007). The STB was especially troubled by the disconnect between the purported rationale for the fuel surcharges — fuel cost recovery — and the formula's dependence on base rates, which need not reflect the marginal fuel costs of a particular shipment. See id. at *4. The decision did not, however, directly implicate those shippers whose traffic was governed by bilateral contract. See id. at *10.
A flurry of antitrust class actions against the four major freight railroads ensued, all of which were ultimately consolidated before the district court. See Fuel Surcharge I, 587 F.Supp.2d at 29. While several sets of plaintiffs were part of the consolidated proceedings, this case deals with those eight plaintiffs
Class certification is far from automatic. As recognized by the district court, "Rule 23 does not set forth a mere pleading standard. A party seeking class certification 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." Wal-Mart Stores, Inc. v. Dukes, ___ U.S. ___, 131 S.Ct. 2541, 2552, 180 L.Ed.2d 374 (2011); see Fuel Surcharge II, 287 F.R.D. at 22. Oftentimes, this inquiry resembles an appraisal of the merits, for "it may be necessary for the court to probe behind the pleadings before coming to rest on the certification question." Gen. Tel. Co. of Sw. v. Falcon, 457 U.S. 147, 160, 102 S.Ct. 2364, 72 L.Ed.2d 740 (1982). Such was the case here. While the defendants opposed class certification on a number of grounds, much of the debate centered on the predominance requirement and whether the plaintiffs could show, through common evidence, injury in fact
The plaintiffs' case for certification hinged on two regression models prepared
Before addressing the merits, we pause to consider a thorny threshold question. Should we be exercising our appellate jurisdiction over this case in the first place? Class certification is, after all, not a final decision of the sort we typically review on appeal from the district court. See 28 U.S.C. § 1291. Certain interlocutory decisions do qualify for immediate appellate review. See id. § 1292. But in the case of class certification, that review is discretionary, not automatic. See FED. R.CIV.P. 23(f) & advisory committee's note; see also 28 U.S.C. § 1292(e) (creating jurisdiction over interlocutory appeals in those circumstances in which the Supreme Court has prescribed rules providing for such review).
Discretionary does not mean arbitrary. Choosing whether to exercise jurisdiction over an interlocutory appeal from a certification decision turns on more than what we had for breakfast. According to our case law, three situations warrant immediate review. The first of these arises when the decision to certify is "questionable" and is accompanied by a "deathknell" — i.e., it places "substantial pressure on the defendant to settle independent of the merits of the plaintiffs' claims." In re Lorazepam & Clorazepate Antitrust Litig., 289 F.3d 98, 102, 105 (D.C.Cir.2002). The second situation occurs when the certification decision "presents an unsettled and fundamental issue of law relating to class actions, important both to the specific litigation and generally, that is likely to evade end-of-the-case review." Id. at 105. Thirdly and finally, we will grant interlocutory review of a certification decision that is "manifestly erroneous." Id. Absent "special circumstances," these three categories constitute the sole bases for interlocutory review. Id. at 106.
The categories are mutually reinforcing, not exclusive. A certification decision may warrant immediate review under multiple theories. See, e.g., Szabo v. Bridgeport Machs., Inc., 249 F.3d 672, 675 (7th Cir. 2001). And, where a single ground for interlocutory appeal might otherwise be shaky, the confluence of multiple rationales may fortify our decision — the sort of "special circumstances" contemplated by our case law. Cf. In re Veneman, 309 F.3d 789, 795 (D.C.Cir.2002) (holding out the possibility that a fundamental issue of first impression unlikely to evade end-of-case review may nevertheless qualify for interlocutory review as a special circumstance). We have such a hybrid rationale here. Even if the amount involved does not sound a death knell for the defendants, it is still astronomical. Recent decisions of
We start with the death knell question. While the case law is neither precise nor particularly informative, a few lessons may still be gleaned. We know, for instance, that defendants invoking the death-knell rationale must go beyond "mere assertions." In re Lorazepam, 289 F.3d at 108. It is not enough to claim irresistible pressure to settle and call it a day. Nor do we deal in absolutes: "what might be `ruinous' to a company of modest size might be merely unpleasant to a behemoth." Waste Mgmt. Holdings, Inc. v. Mowbray, 208 F.3d 288, 294 (1st Cir.2000). Above all, death-knell cases are uncommon, for courts have discouraged them except in those rare instances in which "the grant of class status raises the cost and stakes of the litigation so substantially that a rational defendant would feel irresistible pressure to settle." Prado-Steiman ex rel. Prado v. Bush, 221 F.3d 1266, 1274 (11th Cir.2000).
Under normal circumstances, we would take this opportunity to inject some clarity and detail the elements that sound a death knell. Unfortunately, many of the financial particulars are under seal. We will note, however, that the plaintiffs demand a vast sum in damages, which, because this is an antitrust case, are subject to trebling. See Clayton Act § 4, 15 U.S.C. § 15(a). The sheer magnitude is eye-catching, but even more important is the context. Despite the defendants' size and market position, liability of this magnitude could threaten their financial stability. According to the defendants' uncontested representations, at least some would risk a damages award that "would wipe out a substantial portion of their market capitalization." Pet'rs' Br. 23.
The plaintiffs are skeptical, citing the defendants' own financial disclosures. NS, for example, has stated in a filing with the Securities and Exchange Commission, "We do not believe that the outcome of these proceedings will have a material effect on our financial position, results of operations, or liquidity." NS, Quarterly Report (Form 10-Q), at 13 (filed Apr. 24, 2013). But this rosy prognostication is not a function of NS's limitless resources; in the sentence just before, NS also revealed, "We believe the allegations in the complaints are without merit and intend to vigorously defend the cases." Id.; see also UP, Quarterly Report (Form 10-Q), at 33 (filed Apr. 18, 2013) ("We believe that these lawsuits are without merit, and we will vigorously defend our actions. Therefore, we currently believe that these matters will not have a material adverse effect on any of our results of operations, financial condition, and liquidity."). Implicit in these representations is the defendants' hope they will prevail on appeal from the district court's certification decision. Of course, the defendants' belief in the availability of interlocutory review in this case does not make it so. But for that matter, the plaintiffs' logic similarly depends on an appeal to the defendants' faith in their own solvency. Public filings may offer useful guidance to the death-knell inquiry when they actually discuss a company's ability to satisfy a judgment, but not when they merely speculate on a defendant's prospects for success on the merits. In this case, it is the arithmetic that matters.
That said, specific facts relating the defendants' wealth and liquidity to the magnitude of the damages they face and the litigation choices they would be forced to make are sparse. The defendants have generally let the sheer size of the demanded damages award speak for itself. But although the defendants could have done a better job establishing the existence of a death knell, they have adduced just enough facts to satisfy the standard in this case. The threat to NS's market capitalization is considerable, and according to the defendants, certification exposes them to potential liability so massive it would exceed their cumulative adjusted net income for all of 2003-2011. Tellingly, the plaintiffs dispute none of these facts.
Even when the numbers are dispositive, though, a death knell alone does not warrant interlocutory review. Not only must certification push litigants inexorably toward settlement, but the certification decision must itself be "questionable."
Meeting the predominance requirement demands more than common evidence the defendants colluded to raise fuel surcharge rates. The plaintiffs must also show that they can prove, through common evidence, that all class members were in fact injured by the alleged conspiracy. See Amchem Prods., Inc. v. Windsor, 521 U.S. 591, 623-24, 117 S.Ct. 2231, 138 L.Ed.2d 689 (1997). Otherwise, individual trials are necessary to establish whether a particular shipper suffered harm from the price-fixing scheme. That is not to say the plaintiffs must be prepared at the certification stage to demonstrate through common evidence the precise amount of damages incurred by each class member. Messner v. Northshore Univ. HealthSystem, 669 F.3d 802, 815-16 (7th Cir.2012); see Wal-Mart, 131 S.Ct. at 2558. But we do expect the common evidence to show all class members suffered some injury.
With this in mind, we proceed to the defendants' claim the certification decision is "questionable." Of the medley of attacks on the plaintiffs' ability to satisfy the predominance requirement, we focus on one: according to the defendants, Rausser's damages model is defective. The model purports to quantify the injury in fact to all class members attributable to the defendants' collusive conduct. But the same methodology also detects injury where none could exist. When applied to shippers who were subject to legacy contracts — i.e., those shippers who, during the Class Period, were bound by rates negotiated before any conspiratorial behavior was alleged to have occurred — the damages model yields similar results. If accurate, this critique would shred the plaintiffs' case for certification. Common
Rausser contested the legal significance of this criticism of his damages model, but he conceded it measured overcharges to legacy shippers and class members alike. The district court opinion, for its part, regarded Rausser's damages model as essential to its certification decision, for "neither his common factor model nor his damage model in isolation attempts to prove common injury-in-fact. Rather, the result of the damage model must be viewed as the final step in the body of evidence ... presented to show that injury-in-fact is capable of common proof." Fuel Surcharge II, 287 F.R.D. at 69 (internal quotation marks omitted). It did not, however, address the defendants' concern that the damages model yielded false positives with respect to legacy shippers. See id. at 66-70. Together, Rausser's concession and the district court's silence are sufficient to render the certification decision questionable under the death-knell rationale for interlocutory review — particularly when combined with Comcast Corp. v. Behrend, ___ U.S. ___, 133 S.Ct. 1426, 185 L.Ed.2d 515 (2013), which clarified the law of class actions after the district court certified the class.
After the district court's decision to certify — after the parties even submitted their briefing in this appeal — the Supreme Court decided Comcast Corp. v. Behrend, whose analysis informs the case before us. Behrend also involved certification of an antitrust class action under Rule 23(b)(3) based on regression modeling. See 133 S.Ct. at 1430-31. The plaintiffs in Behrend had proposed four theories of antitrust impact, all but one of which the district court rejected. See id. at 1431. The plaintiffs' sole ground, however, for asserting that damages could be calculated on a classwide basis was a model that assumed the validity of all four theories. See id. Not good enough. Predicating class certification on a model divorced from the plaintiffs' theory of liability, the Court held, indicates a failure to conduct the rigorous analysis demanded by Rule 23. See id. at 1433. Without some alternative model to turn to, predominance could not be shown: "Questions of individual damage calculations will inevitably overwhelm questions common to the class." Id. Rejected was the view of the Court of Appeals that "attacks on the merits of the methodology ... have no place in the class certification inquiry." Behrend v. Comcast Corp., 655 F.3d 182, 207 (3d Cir.2011).
As we see it, Behrend sharpens the defendants' critique of the damages model as prone to false positives. It is now indisputably the role of the district court to scrutinize the evidence before granting certification, even when doing so "requires inquiry into the merits of the claim." 133 S.Ct. at 1433. If the damages model cannot withstand this scrutiny then, that is not just a merits issue. Rausser's models are essential to the plaintiffs' claim they can offer common evidence of classwide injury. See Fuel Surcharge II, 287 F.R.D. at 66. No damages model, no predominance, no class certification.
Recall that we may assume jurisdiction even when a certification decision does not fit neatly within one of the three categories calling for interlocutory appeal. "Special circumstances" might also prompt immediate review. In re Lorazepam, 289 F.3d at 106. While our case law has not identified what makes a circumstance special,
Lest our decision be misunderstood, we reiterate our view that "interlocutory appeals are generally disfavored as disruptive, time-consuming, and expensive" for both the parties and the courts. In re Lorazepam, 289 F.3d at 103 (internal quotation marks omitted). But this is not the ordinary case. In light of Behrend, the pressure to settle posed by the threat to the defendants' market capitalization, and the identified defect in the damages model, we grant the defendants interlocutory review under Rule 23(f) because this qualifies as even more than a death-knell situation.
Now to the merits. We have already laid out the damages model's propensity toward false positives. The plaintiffs nevertheless make several attempts at saving the model — and with it, the certification decision. They first argue that "shipments under these pre-Class Period contracts are not even part of the Class, and the relevant issue is whether Class members paid higher all-in rates following Defendants' aggressive imposition of new fuel surcharges." Resp'ts' Br. 65-66. The plaintiffs are right that the defendants' critique does not disprove the damages model's claim of classwide overcharges as a matter of logical necessity; absence of evidence is not evidence of absence.
The plaintiffs next suggest the defendants' price-fixing conspiracy predates the start of the Class Period, meaning antitrust violations may also have tainted even legacy contracts. But the plaintiffs failed to adduce specific evidence of this possibility — say, by rerunning both models from an earlier start date. The claim also runs directly counter to the district court's factual finding that "the fuel surcharge programs applied before the class period were nothing like the widespread and uniform application of standardized fuel surcharges during the class period" because "[b]efore the alleged conspiracy, defendants' differentiated fuel surcharges were subject to competition and negotiation with shippers, were less aggressive, and were applied only sporadically." Fuel Surcharge II, 287 F.R.D. at 48. This was, after all, the crux of the plaintiffs' own evidence of collusion — that there was a "structural break" at the start of the Class Period in the
Finally, the plaintiffs also place great emphasis on the standard of review, contending that the district court's endorsement of Rausser's models was a finding of fact we may only review for clear error. But there is no factual finding to which we could defer. As we already explained, the district court never grappled with the argument concerning legacy shippers. That the district court deemed the damages model workable in the face of different challenges is irrelevant. And, in any event, "while the data contained within an econometric model may well be `questions of fact' in the relevant sense, what those data prove is no more a question of fact than what our opinions hold." Behrend, 133 S.Ct. at 1433 n. 5.
Before Behrend, the case law was far more accommodating to class certification under Rule 23(b)(3). Though Behrend was grounded in what the Court deemed "an unremarkable premise," id. at 1433, courts had not treated the principle as intuitive in the past. In determining Rausser's two models are "`plausible,'" the district court understandably relied on these precedents — including the very decision the Supreme Court reversed in Behrend. Fuel Surcharge II, 287 F.R.D. at 62 (quoting Behrend v. Comcast Corp., 655 F.3d at 204 n. 13). In a similar vein, the district court looked to cases from other circuits suggesting that false positives do not indict the viability of a class, since "[c]lass certification is not precluded simply because a class may include persons who have not been injured by the defendant's conduct." Mims v. Stewart Title Guar. Co., 590 F.3d 298, 308 (5th Cir.2009); see Kohen v. Pac. Inv. Mgmt. Co., 571 F.3d 672, 677 (7th Cir.2009) (same). It is now clear, however, that Rule 23 not only authorizes a hard look at the soundness of statistical models that purport to show predominance — the rule commands it.
Mindful that the district court neither considered the damages model's flaw in its certification decision nor had the benefit of Behrend's guidance, we will vacate class certification and remand the case to the district court to afford it an opportunity to consider these issues in the first instance. See DL v. Dist. of Columbia, 713 F.3d 120, 129 (D.C.Cir.2013). We need not reach the defendants' alternate grounds for relief.
For the foregoing reasons, we vacate the district court's class certification decision and remand the case to permit the district court to reconsider its decision in light of Comcast Corp. v. Behrend.
Fuel Surcharge II, 287 F.R.D. at 12.