Opinion
16-MD-2704 (JPO)
12-15-2023
IN RE: INTEREST RATE SWAPS ANTITRUST LITIGATION This Document Relates to All Cases
OPINION AND ORDER
J. PAUL OETKEN, DISTRICT JUDGE
This multidistrict litigation involves antitrust claims brought against investment banks that were dealers in the market for interest rate swaps. Plaintiffs, a putative class of investors, assert that Defendants unlawfully conspired to boycott and otherwise undermine trading platforms that would have supplied investors with more competitive prices for interest rate swaps. Plaintiffs move for class certification under Federal Rule of Civil Procedure 23(b)(3). For the reasons that follow, the motion is denied.
I. Background
The Court assumes familiarity with the factual background of this matter as set forth in the prior opinions in this case and therefore recites only facts that are particularly relevant to this Opinion and Order. See In re Interest Rate Swaps Antitrust Litig. (IRS III), No. 16-MD-2704, 2019 WL 1147149 (S.D.N.Y. Mar. 13, 2019) (ECF No. 731); In re Interest Rate Swaps Antitrust Litig. (IRS II), No. 16-MD-2704, 2018 WL 2332069 (S.D.N.Y. May 23, 2018) (ECF No. 390); In re Interest Rate Swaps Antitrust Litig. (IRS I), 261 F.Supp.3d 430 (S.D.N.Y. 2017) (ECF No. 237).
Where the Court relies on documents that have been filed under seal, the Court has concluded that the parties' interests in continued sealing of the portions referenced in this Opinion and Order are insufficient to overcome the presumption of public access to judicial documents. See Lugosch v. Pyramid Co. of Onondaga, 435 F.3d 110, 119-20 (2d Cir. 2006).
This case centers on a claim of an antitrust conspiracy among Defendants-large investment banks that are dealers of interest rate swaps (“IRSs”)-to boycott trading platforms that would permit anonymous, “all-to-all” trading of IRSs. (ECF No. 748 (“Compl.”) ¶ 4.) Plaintiffs-various entities that entered into IRS transactions with Defendants during the class period-contend that Defendants unlawfully conspired to prevent the development of such platforms. (See id. ¶¶ 4, 37-39.) Specifically, Plaintiffs contend that Defendants' alleged conspiracy prevented the growth of all-to-all trading platforms that would have provided price benefits to investors unavailable in the current over-the-counter model, in which investors trade IRSs through direct bilateral communications with dealers. (Id. ¶¶ 6-9.)
An IRS is a financial derivative that permits two parties to trade one future stream of interest rate-based cash flows for another. Typically, one party to an IRS pays cash flows based on a fixed interest rate, while the counterparty pays cash flows based on a floating interest rate, keyed to a benchmark rate like the London Interbank Offered Rate (“LIBOR”). Historically, the investment banks have been the exclusive market makers or liquidity providers in the IRS market. In that role, banks profit by correctly pricing the IRSs that they buy or sell. Because the floating rate is usually keyed to LIBOR, the negotiation occurs over the fixed rate that will be paid. For a given type of IRS, the dealer sets a “bid” price (the fixed rate at which it will purchase the IRS) and an “ask” price (the fixed rate at which it will sell the IRS). For any given IRS transaction, then, the profit that accrues to the dealer is the “spread” between the purchase price (either the “bid” or “ask”) and the “true” market value of the contract-the theoretical rate at which the swap would have been an exchange of two equal income streams-which, although not directly observable, in most cases falls somewhere midway between the “bid” and “ask” prices set by the dealer. See generally IRS I, 261 F.Supp.3d at 443.
Plaintiffs allege that Defendants, in their role as market makers, conspired to stop the IRS market from developing in ways that would make the market more competitive. (Compl. ¶ 6.) One such development would have been the rise of robust anonymous, all-to-all trading platforms, in which investors could engage in transactions with other investors without the direct involvement of a dealer. (Id. ¶ 10.) Plaintiffs assert that Defendants' collusion successfully stopped meaningful development of such platforms, resulting in inflated spreads on IRSs. (Id. ¶¶ 18-20.)
Plaintiffs filed an initial complaint on November 25, 2015. On June 2, 2016, the United States Judicial Panel on Multidistrict Litigation transferred all related matters to this Court for coordinated or consolidated pretrial proceedings. (ECF No. 1.) On July 26, 2017, the Court resolved an initial motion to dismiss, sustaining the antitrust claims for the years 2013 to 2016 but dismissing the claims for prior years. See IRS I, 261 F.Supp.3d at 501. Subsequently, leave to amend the complaint to restore the pre-2013 claims was denied twice. See IRS II, 2018 WL 2332069, at *9; IRS III, 2019 WL 1147149, at *1.
Plaintiffs have now moved for class certification under Federal Rule of Civil Procedure 23(b)(3). (ECF Nos. 722, 869.) Defendants oppose certification. (ECF Nos. 815, 891.) Both sides have also filed motions to exclude the other side's expert reports. (ECF Nos. 817, 874, 877.)
Plaintiffs seek certification of the following class:
All persons or entities who, from January 1, 2013 to the present [(the “class period”)], entered into one or more fixed-floating IRS, overnight index swaps, single-currency basis swaps, or forward rate agreements with the Dealer Defendants, or their respective affiliates, in the United States and its territories.(ECF No. 723 at 2.)
II. Legal Standard
To obtain certification of a class pursuant to Rule 23(b)(3), a plaintiff must satisfy the numerosity, commonality, typicality, and adequacy of representation requirements of Rule 23(a). Fed.R.Civ.P. 23(a). A plaintiff must also meet two additional showings: “predominance, i.e., law or fact questions common to the class predominate over questions affecting individual members, and superiority, i.e., class action is superior to other methods.” In re Initial Pub. Offerings Sec. Litig., 471 F.3d 24, 32 (2d Cir. 2006) (citing Fed.R.Civ.P. 23(b)(3)). The party seeking class certification must “affirmatively demonstrate” compliance with each of those requirements “through evidentiary proof.” Comcast Corp. v. Behrend, 569 U.S. 27, 33 (2013) (internal quotation marks and citation omitted).
Federal Rule of Evidence 702 grants an expert witness testimonial latitude unavailable to other witnesses, provided that “the expert's scientific, technical, or other specialized knowledge will help the trier of fact to understand the evidence or to determine a fact in issue,” “the testimony is based on sufficient facts or data,” “the testimony is the product of reliable principles and methods,” and “the expert's opinion reflects a reliable application of the principles and methods to the facts of the case.” Fed.R.Evid. 702. The “proponent of expert testimony has the burden of establishing by a preponderance of the evidence that the admissibility requirements of Rule 702 are satisfied.” United States v. Williams, 506 F.3d 151, 160 (2d Cir. 2007).
Although there is “some question of whether a Daubert analysis is appropriate at the class-certification stage,” the “heavy weight of authority militat[es] towards a Daubert inquiry at class certification.” Passman v. Peloton Interactive, Inc., __F.Supp.3d __, 2023 WL 3195941, at *4 (S.D.N.Y. 2023) (internal quotation marks and citation omitted); see also In re LIBOR-Based Fin. Instruments Antitrust Litig., 299 F.Supp.3d 430, 470-71 (S.D.N.Y. 2018). Still, a court's Daubert inquiry is “guided by the purpose for which the evidence is introduced”-in this case, “establishing the various class certification requirements.” In re LIBOR, 299 F.Supp.3d at 471. In other words, at this stage, the “question is not whether a jury at trial should be permitted to rely on the expert's report to find facts as to liability, but rather whether the Court may utilize it in deciding whether the requisites of Rule 23 have been met.” Id. (internal quotation marks and citation omitted); see also In re Aluminum Warehousing Antitrust Litig., 336 F.R.D. 5, 29 (S.D.N.Y. 2020).
III. Discussion
Defendants oppose the motion for class certification primarily on the ground that Plaintiffs have failed to meet Rule 23(b)(3)'s requirement of predominance. (ECF No. 815 at 20-63; ECF No. 891 at 2-16.) Defendants also oppose class certification on the grounds, among others, that Plaintiffs' proposed class is overbroad and that Plaintiffs fail to meet Rule 23(a)'s requirements of typicality and adequacy. Finally, as part of their disagreement about the suitability of class certification, both sides move to exclude the other side's expert reports under Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993).
The Court turns first to the Daubert motions before resolving the issue of class certification. Ultimately, because the Court determines that individual issues predominate over questions common to the class, Plaintiffs' motion for class certification is denied.
A. Daubert Motions
Plaintiffs move to strike the expert reports of Christopher L. Culp and Peter C. Reiss. (ECF Nos. 874, 877.) Defendants move to strike the expert report of Mark Grinblatt. (ECF No. 817.) The motion to strike Reiss's expert report is denied.
The Court need not resolve Defendants' motion to strike the expert report of Mark Grinblatt. Even affording full consideration to Grinblatt's report, the Court determines that Plaintiffs' motion for class certification should be denied. The Court also need not resolve Plaintiffs' motion to strike the expert report of Christopher L. Culp, as it does not rely on Culp's report to deny Plaintiffs' motion. Accordingly, these motions are denied as moot.
Reiss is a Professor of Economics at the Stanford Graduate School of Business. His fields of expertise include econometrics and financial markets, and his research has explored measurements of bid-ask spreads in dealer and customer markets. (ECF No. 816-2 ¶¶ 1-5.) In his report, Reiss opined that he has identified some IRS transactions included within the class definition that were not harmed by the alleged antitrust conspiracy. To reach that conclusion, Reiss calculated the spread for each IRS in his data set (a subset of all swaps in the class definition) by taking the difference between the actual rate paid on the swap and the swap's “true rate,” estimated by using an average of the bid and ask quotes provided by Bloomberg (the “mid”). (Id. ¶¶ 64, 154.) Reiss's calculations show that a substantial percentage of trades were executed at negative or zero spreads. (Id. ¶ 154.) This is significant, Reiss avers, because such trades could not have been harmed by the antitrust conspiracy. (Id. ¶ 147.) Reiss confirmed this finding with other mids from alternative sources. (Id. ¶¶ 157-60, 165.)
Plaintiffs challenge Reiss's report on three grounds, but none succeeds. The first is that Reiss's reliance on the mids from Bloomberg requires exclusion because Reiss “did nothing to assess the reliability of using the Bloomberg composite mids as a true midpoint value of an IRS.” (ECF No. 878 at 11.) As an initial matter, Plaintiffs' own expert witness, Grinblatt, used Bloomberg mids as a crucial input in his econometric model. (See ECF No. 872-3 ¶ 378 (“As an estimate of, or proxy for, the ‘true rate,' I use the mid-point of Bloomberg's data ....”).) That fact alone suggests that the data is acceptable in the industry and therefore available for use by Defendants' expert witness. See, e.g., In re Rail Freight Fuel Surcharge Antitrust Litig., 292 F.Supp.3d 14, 67 (D.D.C. 2017) (“Plaintiffs make no argument that the transaction data upon which [the defendants' expert] built his exercises are irrelevant or unreliable-nor could they, because it is the same data upon which [plaintiffs' expert] built his models.”); Marketquest Grp., Inc. v. BIC Corp., No. 11-CV-618, 2018 WL 1756117, at *3 (S.D. Cal. Apr. 12, 2018) (“It appears that [the plaintiff's expert's] reliance on this data is acceptable in the industry, though, given that [the defendant's] expert uses the very same data.”).
Plaintiffs object that Grinblatt did not use the data for a “similar purpose,” and that Grinblatt used the Bloomberg mids merely as an “imperfect proxy input to, along with other regression variables, estimate an IRS's value.” (Id. at 4-5 & n.4.) But that use is not for a fundamentally different purpose: Grinblatt's purpose, as described by Plaintiffs themselves, was to estimate the true value of an IRS (which would inform whether an IRS was traded at a zero or negative spread). Both Plaintiffs' expert and Defendants' expert therefore used the mids to “buil[d] [their] models” to estimate IRS spreads. In re Rail Freight, 292 F.Supp.3d at 67. The fact that Grinblatt used non-Bloomberg data for one purpose does not preclude his use of Bloomberg data for that very same purpose.
Even if Plaintiffs did somehow use the Bloomberg data for a different purpose, Plaintiffs themselves provide evidence that the Bloomberg data is reliable, and they fail to identify any fundamental flaws that would justify the exclusion of Reiss's expert report. As Grinblatt explains, Bloomberg is a “widely used third-party vendor of IRS quote data relied on by investors, dealers, regulators and academics to make trade decisions, conduct research studies, and learn about the IRS market.” (ECF No. 725-1 App. 3 ¶ 3.) Grinblatt's report also describes how the specific data at issue is generated: “In its capacity as a commercial data-provider, Bloomberg maintains intra-day data during the class period for what it describes as ‘composite' bid and ask quotes for IRS with certain contractual specifications ....” (Id. ¶ 217.) Grinblatt then goes on to explain that Bloomberg has described its composite data as “showing the best (or highest) bid rate and the best (or lowest) ask rate at each given point in time across all active, contributed indicative quotes from a set of dealers whose quotes have previously been screened by Bloomberg for reliability.” (Id.)
Plaintiffs cannot now turn around and argue that the Bloomberg data is wholly unreliable. The question for Daubert purposes is not whether the expert employed the best possible input data to form her testimony, but whether “the testimony is based on sufficient facts or data.” Fed.R.Evid. 702; see also U.S. Bank Nat'l Ass'n v. PHL Variable Life Ins. Co., 112 F.Supp.3d 122, 131 (S.D.N.Y. 2015) (an expert is “permitted to rely on facts, opinions, and data not of the expert's own making . . . even if those facts, opinions, and data are otherwise inadmissible”). The Bloomberg data at issue provides relevant estimates, and Plaintiffs have not shown that Reiss's use of Bloomberg's mids are so “speculative,” “conjectural,” or “unrealistic and contradictory” that they render his estimate of spreads fundamentally unreliable. Zerega Ave. Realty Corp. v. Hornbeck Offshore Transp., LLC, 571 F.3d 206, 214 (2d Cir. 2009) (internal quotation marks and citation omitted). Accordingly, Defendants' objection goes to weight, not admissibility. See U.S. Bank Nat'l Ass'n, 112 F.Supp.3d at 131.
Plaintiffs similarly challenge Reiss's reliance on mids taken from sources other than Bloomberg, such as Defendants' reported Pre Trade Marks (“PTMs”), which are estimated marks that dealers are required by law to report. (See ECF No. 878 at 1.) Plaintiffs argue that Reiss has “no understanding of how Defendants' PTMs are generated and made no effort to evaluate why the PTMs substantially differ from each other.” (Id. at 7.) That argument also fails, however, as Reiss had a general understanding of how they were generated, and he knew that the mids were at least used by the dealers themselves as estimates of the “true rate” of IRSs. (ECF No. 816-2 ¶¶ 40-42.) Reiss explained that the PTM “represents the dealer's assessment of the value of the swap,” providing probative evidence of the value of the IRSs. (Id. ¶ 40). For example, Reiss's report contains comments from one major market participant explaining that the PTMs represent an estimate of the value of the swap calculated using “proprietary computer valuation models that are used by [the participant] to prepare its own financial books and records” and “relevant mid-market price data inputs” at the time of estimation. (Id. ¶ 41.) “[I]t is not uncommon for experts to rely on interviews with third parties in forming their opinions,” just as Reiss did here. Better Holdco, Inc. v. Beeline Loans, Inc., __ F.Supp.3d __, 2023 WL 2711417, at *9 (S.D.N.Y. 2023) (quoting Williams v. Illinois, 567 U.S. 50, 109 (2012) (Thomas, J., concurring)).
Nor does it make a difference, contrary to Plaintiffs' argument, that different dealers submitted different PTMs for the same swaps. Again, the data relied on by Reiss need not be perfectly accurate to be informative, and Plaintiffs appear to concede that PTMs constitute data prepared in the ordinary course of business that are routinely relied upon by regulators and market participants. Moreover, Reiss uses the PTM data largely to corroborate conclusions that he independently reached by relying on the Bloomberg mids. (ECF No. 816-2 ¶ 161.) Accordingly, where objections focus “almost entirely on [an expert's] factual inputs and assumptions rather than his methodology,” as they do here, such objections “go to the probative value of [the expert's] testimony, not its admissibility.” SourceOne Dental, Inc. v. Patterson Cos., Inc., No. 15-CV-5440, 2018 WL 2172667, at *6 (E.D.N.Y. May 10, 2018).
Finally, Plaintiffs challenge Reiss's reliance on record evidence to conclude that market participants sometimes traded at spreads at or below zero. The record contains significant evidence of such a phenomenon from testimony and documents from knowledgeable market participants. (ECF No. 816-2 ¶¶ 121-46.) Plaintiffs contend that those statements “fail[] to specify what ‘mid' the authors were alluding to,” meaning that Reiss could not know whether the authors were referring to the “true economic value of an IRS” or “some other concept of a ‘mid,'” such as “the dealers' subjective view of mid-market value.” (ECF No. 878 at 14.) As an initial matter, Plaintiffs quibble with only a subset of the market participants' statements, and Reiss's opinion relies on testimony and documents from a wide range of participants beyond the challenged statements. (See ECF No. 892 at 10 & n.6.)
Plaintiffs' objection also fails on its merits. Reiss's assumption that the speakers were referring to the fair market value of the IRS-rather than “some other concept of a ‘mid,'” which Plaintiffs do not attempt to specify-is a sensible assumption in the context, especially when Plaintiffs give no evidence to the contrary. As a result, that assumption is not “so unrealistic and contradictory as to suggest bad faith,” and any argument that such “assumptions are unfounded go to the weight, not the admissibility, of the testimony.” Zerega, 571 F.3d at 214 (quoting Boucher, 73 F.3d at 21). Moreover, even if Plaintiffs are right that the participants' statements were about their subjective views, such views would still provide relevant evidence for the likelihood that certain IRSs were traded at zero or negative spreads. As a result, the Court cannot say at this point that Reiss's testimony would not “help the trier of fact to understand the evidence or to determine a fact in issue.” Fed.R.Evid. 702. Apart from that objection, Plaintiffs do not challenge the significant record evidence from market participants that supports Reiss's opinion that IRSs sometimes trade at zero or negative spreads.
Because all of Plaintiffs' objections to Reiss's expert report at most go to weight and not admissibility, the motion to exclude Reiss's report is denied.
B. Class Certification
To certify a class, the Court must conduct a “rigorous analysis” to determine that Rule 23's requirements have been satisfied. Comcast Corp., 569 U.S. at 33 (quoting Wal-Mart Stores, Inc. v. Dukes, 564 U.S. 338, 351 (2011)). Here, that includes Rule 23(b)(3)'s requirement of predominance, which “tests whether proposed classes are sufficiently cohesive to warrant adjudication by representation.” Amchem Prods. Inc. v. Windsor, 521 U.S. 591, 623 (1997).
The Court's analysis begins and ends with the issue of predominance. “Because the Court concludes that Plaintiffs have not adequately demonstrated predominance, it need not address the other [Rule 23] requirements.” Hunter v. Time Warner Cable Inc., No. 15-CV-6445, 2019 WL 3812063, at *9 (S.D.N.Y. Aug. 14, 2019). In the antitrust context, the predominance requirement means that Plaintiffs “must demonstrate that the elements of their underlying claims can be proven by common evidence.” Dial Corp. v. News Corp., 314 F.R.D. 108, 114 (S.D.N.Y. 2015). The elements of an antitrust claim are “(1) a violation of antitrust law; (2) injury and causation; and (3) damages.” Cordes & Co. Fin. Servs., Inc. v. A.G. Edwards & Sons, Inc., 502 F.3d 91, 105 (2d Cir. 2007) (internal quotation marks and citation omitted).
The parties dispute, among other things, whether the predominance requirement is met with respect to the second element: antitrust “impact.” See Dial Corp., 314 F.R.D. at 114-15 (“To satisfy the Rule 23(b)(3) predominance requirement, Plaintiffs must demonstrate that class-wide injury or ‘impact' is capable of proof at trial through evidence that is common to the class rather than individual to its members.”). Antitrust “impact” requires a showing that the plaintiff suffered “some injury as a result of defendants' violation of the antitrust law.” In re Indus. Diamond Antitrust Litig., 167 F.R.D. 374, 382 (S.D.N.Y. 1996). “It is important to note at the outset that ‘impact' (or ‘injury-in-fact') and ‘damages' are two distinct elements of an antitrust claim-injury-in-fact is whether the plaintiffs were harmed and damages quantify by how much.” In re Currency Conversion Fee Antitrust Litig., 264 F.R.D. 100, 115 (S.D.N.Y. 2010) (internal quotation marks and citation omitted).
Accordingly, Plaintiffs argue “that common, class-wide proof exists that all putative class members paid more than they would have in the but-for world.” Dial Corp., 314 F.R.D. at 115 (emphasis added); see also Sykes v. Mel S. Harris & Assocs. LLC, 780 F.3d 70, 82 (2d Cir. 2015) (explaining that to show predominance, plaintiffs must “show that they can prove, through common evidence, that all class members were . . . injured by the alleged conspiracy” (emphasis added) (internal quotation marks and citation omitted)). Defendants disagree, contending, among other things, that large numbers of the relevant IRS trades were unharmed by the alleged conspiracy because they were executed at spreads that were less than or equal to zero. These “below-zero” and “at-zero” trades could not have been harmed by the antitrust conspiracy, Defendants argue, because “no one offers to trade at a zero spread or a negative spread in anonymous trading.” (ECF No. 815 at 32.) Accordingly, Defendants contend, the presence of these at- and below-zero trades renders a showing of class-wide antitrust impact impossible.
Defendants provide several potential reasons why at- or below-zero trading would occur. (See ECF No. 815 at 32 & nn.55-56.) For one, dealers can be willing to absorb losses on individual trades to develop relationships that can benefit their banks in the longer run. (See ECF No. 816-1 ¶¶ 83-85.) For another, dealers sometimes offer spreads at or below zero to trade out of unbalanced risk positions. (See ECF No. 815 at 28 & n.46; ECF No. 816-1 ¶ 85.)
Plaintiffs claim that no below-zero trades occurred and cite a report by Mark Grinblatt, Plaintiffs' expert, to support their claim. In response, Peter Reiss, Defendants' expert, concludes that Grinblatt's own analysis indicates that twenty-three percent of the trades in Grinblatt's data set were executed at or below zero. (See ECF No. 815 at 32; ECF No. 816-2 ¶ 154.) Reiss also adduces various statements in the record from market participants indicating that IRS trades were executed at or below zero spreads. (See ECF No. 815 at 30-31 (citing ten representative statements).) In short, we are presented with a dispute between two experts that-for purposes of the Rule 23 analysis-must be resolved at this stage of the litigation. See In re LIBOR, 299 F.Supp.3d at 471 (“[T]hough a Daubert motion is an improper venue in which to take sides in a ‘battle of the experts' offered by competing parties, disputes between experts must be resolved if necessary to the Rule 23 analysis.” (citations omitted)).
Plaintiffs have failed to persuasively rebut Defendants' strong showing that there were numerous at- or below-zero trades during the class period. To counter Reiss's findings, Plaintiffs rely exclusively on Grinblatt's analysis, which indicates that there were no IRS trades with zero or negative spreads. (See ECF No. 869 at 24-29.) But as Defendants point out, Grinblatt's methodology is fundamentally unsuited to determining whether such trades occurred. Recall that the spread is calculated for a given IRS by taking the difference between the actual rate at which the IRS was traded and the IRS's “true” rate, which is the rate at which the swap would have been an exchange of two equal income streams. Both parties' experts agree that although the actual rate at which each IRS was traded is known, the “true” rate for the IRS cannot be directly observed. To calculate an IRS's spread, then, one must choose an estimator of the true rate. Grinblatt did so by means of a regression. A crucial input into Grinblatt's regression is the midpoint of the ask and bid rates, taken from Bloomberg data, for each IRS in his data set. (ECF No. 872-3 ¶¶ 377-378.) The other input variables include various characteristics of IRS trades: whether the swap was traded on or off a swap execution facility, whether the swap was cleared, the year in which the transaction took place, the swap's notional value, the swap's tenor, and the trading frequency of the contract. (ECF No. 872-1 ¶ 235.) From that model, Grinblatt calculated an estimated “true rate” for each category of IRS. (Id. ¶¶ 228-34.)
That approach, however, sheds little to no light on whether any individual trades were conducted at or below zero spreads. Grinblatt's model makes a crucial assumption: that all trades with the same characteristics-including tenor, notional value, trading intensity, and year-share the same “true” rate. (See ECF No. 816-2 ¶ 64.) That assumption “smooths out” what is otherwise considerable variation in each individual IRS's mid, which Reiss extensively documents in his opinion. (See id. ¶¶ 121-46.) Grinblatt's approach therefore raises the specter of false positives, as under his approach, there are many IRSs that should qualify as at- or below-zero swaps but appear to be above-zero swaps in Grinblatt's model once his smoothed-out “true” rate is substituted. Reiss has attempted to calculate the number of false positives in Grinblatt's data set by calculating the percentage of trades that have a zero or negative spread if the individual trade's “true” mid value is used, but instead have a positive spread if Grinblatt's smoothed-out rate is used. That false positive rate, according to Reiss, is a significant twenty-three to twenty-nine percent. (See ECF No. 815 at 32; ECF No. 816-2 ¶ 154; ECF No. 891 at 8.)
Plaintiffs contend that reliance on “raw” mids to calculate individual spreads is unreliable. According to Plaintiffs, mids are “just estimates, and not precise measures of the ‘true value' of a swap.” (ECF No. 869 at 25.) As a result, because mids are subject to “measurement error,” Plaintiffs argue that the relevant analysis requires a regression analysis that econometrically aggregates otherwise noisy mids and “assigns a spread to each transaction that reflects its many economic attributes.” (ECF No. 872-3 ¶¶ 378-389.)
Grinblatt disagrees that regression analysis is a type of “averaging.” (ECF No. 872-3 ¶ 389.) But Grinblatt acknowledges that his methodology assigns a “true rate” to each IRS that “reflects its many economic attributes, its timing, and other relevant factors.” (Id.) His approach, in other words, assigns all IRSs with certain characteristics a specific “true rate” that smooths out what is otherwise considerable variation within that kind of IRS's mids.
Plaintiffs' defense of their model might make sense if the purpose of the model is to calculate class-wide damages. After all, an approach that uses noisy mids to calculate each individual IRS's spread could conceivably lead to more measurement error than an approach that aggregates mids by category of IRS and calculates spreads on a category-by-category basis. (See id. ¶ 381.) But that approach is less helpful if the point of the model is to determine whether any individual IRSs were traded at zero or negative spreads in the first place. If the “true rate” for a category of IRSs is calculated by econometrically aggregating mid data for that class, any individual IRSs that were traded at a negative spread will be treated as if they were traded at the estimated spread for that category of IRS.
To be sure, Plaintiffs are correct that using mids to determine whether any individual IRSs were traded at zero or negative spreads has its own issues. As Grinblatt rightly observes, mids are derived from market participants' subjective estimates of the true rate, rendering them subject to some measurement error. (Id. ¶ 380.) Further, mids also occasionally suffer from data recording errors. (Id. ¶ 381.) It is conceivable, then, that some fraction of Reiss's false positives are not actually false positives at all, but rather above-zero spread transactions for which the mid provides an incorrect measurement of the “true rate.”
Nevertheless, a false positive rate of twenty-three to twenty-nine percent is difficult to explain away as statistical noise-which is, in essence, what Plaintiffs ask this Court to do.
Additionally, the fact that estimates of mids may suffer from some measurement and recording errors should work in both directions, and Plaintiffs have not explained why such inaccuracies work only to overestimate, but not underestimate, the proportion of IRSs traded with zero or negative spreads. More fundamentally, though, it is Plaintiffs' burden to “affirmatively demonstrate” compliance with Rule 23 “through evidentiary proof.” Comcast Corp., 569 U.S. at 33 (internal quotation marks and citation omitted). Although Plaintiffs have taken issue with Reiss's findings, they have provided no persuasive showing of their own about the frequency at-or below-zero trades. Plaintiffs offer Grinblatt's expert report and his observations that there are almost no at- or below-zero trades, but for the reasons explained above, his report does not suffice. Absent such a showing, class certification should be denied. Id. at 34.
Plaintiffs' alternative arguments for demonstrating predominance also fail. For example, Plaintiffs contend that “any uninjured trades can be identified through simple arithmetic and excluded.” (ECF No. 869 at 26.) In so doing, Plaintiffs invoke the possibility of a “manageable, individualized process at or before trial” to identify at- or below-zero trades. In re Foreign Exch. Benchmark Rates Antitrust Litig., 407 F.Supp.3d 422, 434 (S.D.N.Y. 2019) (internal quotation marks and citation omitted). But Plaintiffs' own model cannot be used for such a process because, as explained above, it essentially assumes the relevant conclusion. Reiss's approach could potentially identify such trades, but that approach requires mid data for each individual IRS-data that, as Defendants point out, does not currently exist for more than ninety percent of the IRSs traded during the class period. (See ECF No. 815 at 47-48; ECF No. 891 at 14.)
Thus, Plaintiffs have not sufficiently demonstrated that class-wide impact is “capable of proof at trial through evidence that is common to the class.” Dial Corp., 314 F.R.D. at 114-15. While Plaintiffs attempt to characterize these problems as relevant only to damages and not to the issue of class certification (see, e.g., ECF No. 869 at 26, 64), they are indeed relevant to the antecedent inquiry of whether the class members are dissimilar enough, due to the presence of uninjured trades, to defeat predominance. Courts routinely “requir[e] that statistical models offered by plaintiffs show all class members suffered some injury” as part of the predominance inquiry. Nat'l ATM Council, Inc. v. Visa Inc., 2023 WL 4743013, at *6 (D.C. Cir. July 25, 2023) (internal quotation marks and citation omitted). Here, however, because of the individual inquiry required to evaluate whether each IRS traded at or below zero spreads, “any class trial would . . . devolve into thousands of mini-trials to determine eligibility for class members.” Hunter v. Time Warner Cable Inc., No. 15-CV-6445, 2019 WL 3812063, at *12 (S.D.N.Y. Aug. 14, 2019); see also In re Indus. Diamond Antitrust Litig., 167 F.R.D. 374, 384 (S.D.N.Y. 1996) (rejecting showing of predominance because the court “would be required to scrutinize each transaction to ascertain whether the purchaser paid a supracompetitive price”). Class certification is therefore inappropriate.
Plaintiffs also disclaim the necessity of econometric evidence at all, arguing that classwide antitrust impact has been sufficiently demonstrated via their recitation of record evidence and the relevant academic literature. (See ECF No. 869 at 11.) In a case like this one, however, in which different plaintiffs bought different products at different prices, having a class-wide method that shows that “damages can be proven using a method common across the class” is “critical to the class proponent's ability to demonstrate that common issues will predominate.” William B. Rubenstein, 6 Newberg and Rubenstein on Class Actions § 20:64 (6th ed. 2023). Therefore, when class members “purchased different products or were subject to different prices or competitive conditions,” it is difficult to show predominance by reference to economic theory or to generalizations in the record about the market. See id. § 20.52; see also In re Hydrogen Peroxide Antitrust Litig., 552 F.3d 305, 325 (3d Cir. 2008) (“[T]he question at [the] class certification stage is whether, if [antitrust] impact is plausible in theory, it is also susceptible to proof at trial through available evidence common to the class.”); In re Wholesale Grocery Prods. Antitrust Litig., No. 09-MD-2090, 2012 WL 3031085, at *10 (D. Minn. July 25, 2012) (“Plaintiffs cannot rely on theory alone [for antitrust impact], they must produce a method of confirming that theory matched reality for each class member because they bear the burden of proving impact.”). Given the nature of the relevant market, Plaintiffs' appeal to the economic literature and to expert reports containing only generalized analysis, such as Dr. Darrell Duffie's report, falls short. (See ECF No. 723 at 48-55; ECF No. 869 at 11-24.) And even if Plaintiffs are correct that such generalized analysis could suffice in some cases to demonstrate predominance, here, Defendants have identified empirical evidence of at- or below-zero trades that persuasively rebuts such generalized evidence and requires more from Plaintiffs to justify class certification.
Finally, Plaintiffs argue that even if Defendants have successfully identified transactions that could not have been harmed by the antitrust conspiracy, they have not successfully identified investors who were unharmed. (ECF No. 869 at 27-28.) This distinction matters, Plaintiffs contend, because antitrust injury occurs if an investor is overcharged even on a single transaction. (Id. (citing cases).) Under this logic, to successfully raise a challenge to Plaintiffs' showing of predominance, Defendants must show that there are investors in the class that exclusively traded at a spread at or below zero.
It is possible that some investors received lower prices on IRSs than they would have received in the but-for world. (ECF No. 815 at 57-59.) Reiss finds support for this hypothesis in Grinblatt's data (again, applying trade-specific mids rather than the outputs of Grinblatt's regression). (See id. at 32 (noting that “29 percent [of the trades in Grinblatt's data set] took place at spreads lower than” the spreads Dr. Grinblatt calculated in the but-for world absent Defendants' conduct); see also ECF No. 816-2 ¶ 154.) If so, then Plaintiffs' predominance argument may suffer from yet another flaw. Like other courts in this District, this Court is “skeptical . . . that a single impacted payment is sufficient to establish antitrust injury” in the presence of offsetting pricing benefits for the same product in the same market arising from the same antitrust violation. In re LIBOR, 299 F.Supp.3d at 594. The “question of [antitrust] injury [is] whether [the plaintiff] is worse off than it would be if the market were free of anticompetitive forces.” IQ Dental Supply, Inc. v. Henry Schein, Inc., 924 F.3d 57, 64 (2d Cir. 2019). This does not mean that all effects of the anticompetitive conduct, such as “distant second-order effect[s],” should be evaluated in the inquiry. In re LIBOR, 299 F.Supp.3d at 594; see also Kottaras v. Whole Foods Mkt., Inc., 281 F.R.D. 16, 25 (D.D.C. 2012) (“The issue here, however, is not whether to offset harms in one market with benefits in another, but whether to count benefits against harms in the same market .... [T]he cases that do indicate that the benefits must be factored into assessments of harm.”). Under the logic of the Second Circuit in IQ Dental Supply, though, the impact analysis should consider both harms and benefits caused by the antitrust conspiracy in the same market and for the same product.
Again, Plaintiffs misunderstand the burden of proof. Defendants have already made a showing that the class period includes a significant number of uninjured transactions, suggesting the same may be true of investors themselves. Cf. In re Rail Freight Fuel Surcharge Antitrust Litig., 725 F.3d 244, 254 (D.C. Cir. 2013) (“[P]laintiffs are right that . . . absence of evidence is not evidence of absence,” but that argument “misapprehend[s] their burden.”). Plaintiffs-not Defendants-must “affirmatively demonstrate” compliance with the predominance requirement “through evidentiary proof,” and in the face of a contrary showing, Plaintiffs have provided no evidence that the presence or absence of uninjured class members is capable of common proof. Comcast Corp., 569 U.S. at 33 (internal quotation marks and citation omitted); see also In re Air Cargo Shipping Servs. Antitrust Litig., No. 06-MD-1175, 2014 WL 7882100, at *44 (E.D.N.Y. Oct. 15, 2014) (“The defendants contend that [there were uninjured transactions], though they do not venture to estimate how frequently....This is not their burden, they correctly note.”). Moreover, determining which investors did and did not suffer injury would similarly entail individualized inquiries for each investor. Given the breadth of the class and the number of investors involved, Plaintiffs' distinction between uninjured investors and uninjured transactions does not shield them from the problems of individualized inquiries.
Plaintiffs cite cases in which courts have rejected arguments about false positives because the defendant did not identify uninjured class members, but those cases are distinguishable because the plaintiffs either persuasively rebutted the defendant's showing that there were uninjured class members, or because the plaintiffs suggested some plausible and easily manageable basis to weed out a relatively small portion of uninjured class members. Here, in contrast, Plaintiffs have done neither: although Plaintiffs suggest in a footnote that they have identified only one uninjured class member (ECF No. 869 at 28 n.44), that calculation is based on Grinblatt's model, which assumes Plaintiffs' desired conclusion for the reasons explained above. And although Plaintiffs continue to maintain that Grinblatt's model provides a plausible basis to determine class-wide impact when taken at face value on its own terms, “courts must consider potential defenses in assessing the predominance requirement.” Myers v. Hertz Corp., 624 F.3d 537, 551 (2d Cir. 2010). As a result, this Court must “assess all of the relevant evidence admitted at the class certification stage”-including evidence from Defendants-to “determine whether each Rule 23 requirement has been met, just as the judge would resolve a dispute about any other threshold prerequisite for continuing a lawsuit.” In re IPO, 471 F.3d at 42. Accordingly, Plaintiffs have not met their “burden to show that the conspiracy impacted ‘all or virtually all' class members.” In re Air Cargo, 2014 WL 7882100, at *44.
For example, in In re Air Cargo, the “plaintiffs ha[d] submitted significant record evidence suggesting that waivers, offsets, and the other corrective measures identified by the defendants [that would result in uninjured transactions] did not commonly occur,” in addition to “submit[ting] the testimony of two experts” to the same effect. 2014 WL 7882100, at *44, *46. Similarly, in City of Philadelphia v. Bank of America Corporation, No. 19-CV-1608, 2023 WL 6160534, at *7 (S.D.N.Y. Sept. 21, 2023), the defendants “d[id] not point to evidence of systemic false positives,” and the plaintiffs countered by, among other things, “offer[ing] reasonable explanations of why [defendants'] potentially cherry-picked examples were misleading snapshots in time and might not genuinely represent false positives.”
In Messner v. Northshore Univ. HealthSystem, 669 F.3d 802, 826 & n.15 (7th Cir. 2012), for example, the Seventh Circuit acknowledged that the class involved “minor overbreadth problems” that affected at most “2.4 percent . . . of the class,” but instructed the district court “to amend the class definition as needed to correct for the overbreadth.” Similarly, in In re Lidoderm Antitrust Litig., No. 14-MD-2521, 2017 WL 679367, at *12 (N.D. Cal. Feb. 21, 2017), the court rejected a false-positives argument because “further analysis or refinement of who is in the class . . . c[ould] be readily managed.” And in Tyson Foods, Inc. v. Bouaphakeo, 136 S.Ct. 1036, 1050 (2016), the Supreme Court only suggested that the presence of uninjured class members was not fatal to certification if there were a “methodology” that would “be successful in identifying uninjured class members”-a methodology that Plaintiffs have not provided here.
In sum, Plaintiffs have failed to show that “questions of law or fact common to class members predominate over any questions affecting only individual members.” Fed.R.Civ.P. 23(b)(3). Class certification is therefore denied.
IV. Conclusion
For the foregoing reasons, Plaintiffs' motion is for class certification is DENIED.
Plaintiffs' motion to exclude the expert report of Peter C. Reiss is DENIED. Plaintiffs' motion to exclude the expert report of Christopher L. Culp is DENIED as moot. Defendants' motion to exclude the expert report of Mark Grinblatt is DENIED as moot.
The Clerk of Court is directed to close the motions at ECF Numbers 722, 817, 874, and 877.
SO ORDERED.