Carpenters Pension Trust Fund v. Allstate Corporation

In the United States Court of Appeals For the Seventh Circuit ____________________ No. 19-1830 IN RE: ALLSTATE CORPORATION SECURITIES LITIGATION CARPENTERS PENSION TRUST FUND FOR NORTHERN CALIFORNIA, et al., Plaintiffs-Appellees, v. ALLSTATE CORPORATION, et al., Defendants-Appellants. ____________________ Appeal from the United States District Court for the Northern District of Illinois, Eastern Division. No. 1:16-cv-10510 — Robert W. Gettleman, Judge. ____________________ ARGUED SEPTEMBER 18, 2019 — DECIDED JULY 16, 2020 ____________________ Before KANNE, HAMILTON, and BARRETT, Circuit Judges. HAMILTON, Circuit Judge. The district court certified a plaintiff class in this securities fraud case against Allstate Cor- poration. We granted leave for defendants to pursue this in- terlocutory appeal of that order under Federal Rule of Civil 2 No. 19-1830 Procedure 23(f). The class certification presents several chal- lenging questions about how to apply the “Basic” fraud-on- the-market presumption of reliance in the wake of a series of more recent Supreme Court decisions. Established in Basic, Inc. v. Levinson, 485 U.S. 224 (1988), the fraud-on-the-market presumption allows plaintiffs to avoid proving individual reliance upon fraudulent misrepre- sentations and omissions. Instead, plaintiffs may prove that the given securities traded in efficient markets in which prices reflect all publicly available information, including misrepre- sentations, and all investors were thus entitled to rely on that public information and pricing. Id. at 246–47. That makes se- curities fraud cases better suited for class certification. Evidence supporting or refuting the Basic presumption of reliance is often relevant to three other closely related issues in a securities fraud case—materiality, loss causation, and transaction causation. Recent Supreme Court decisions on those issues pose a difficult challenge at the class certification stage. A district court deciding whether to certify a plaintiff class may not use the evidence to decide loss causation then, Erica P. John Fund, Inc. v. Halliburton Co., 563 U.S. 804 (2011) (Halliburton I), and may not use the same evidence to decide materiality then, Amgen Inc. v. Connecticut Retirement Plans and Trust Funds, 568 U.S. 455 (2013). Those questions are left for the merits. Yet to decide class certification using the Basic pre- sumption, a court must consider the same evidence if the de- fense offers it to show the absence of transaction causation, also known as price impact. Halliburton Co. v. Erica P. John Fund, Inc., 573 U.S. 258 (2014) (Halliburton II). No. 19-1830 3 These precedents require a district court to split some very fine hairs. In this case, the district court granted class certifi- cation after admitting, but without engaging with, the evi- dence that defendants offered to defeat the Basic presump- tion, an expert opinion that the alleged misrepresentations had no impact on the stock price. The judge concluded that the issue was tied so closely to the merits that he should not decide it on class certification. We understand that view. The Supreme Court has long warned the lower federal courts not to confuse class certification decisions with the merits, e.g., Ei- sen v. Carlisle & Jacquelin, 417 U.S. 156, 177 (1974), and the court may not consider materiality and loss causation at the class certification stage. Under Halliburton II, however, the court’s approach was based on a legal error, so we must vacate for reconsideration. Class certification may well be appropriate here, but the dis- trict court must decide at the class stage the price impact issue posed by the defendants’ price impact evidence and plaintiffs’ rebuttal. The court may not defer that question for the merits. We also affirm the district court’s adding a new class repre- sentative and, by agreement of the parties, direct a modifica- tion of any class certification to limit the class to buyers of the defendant’s common stock rather than any other securities. In Part I, we summarize the alleged fraud, the defendants’ response, and the district court’s order granting class certifi- cation. In Part II, we set out the standard for our review of the class certification order, including the need for factfinding. In Part III, we apply Rule 23(b)(3)’s predominance requirement for certifying plaintiff classes in securities fraud cases, the Basic presumption, and the Halliburton/Amgen trilogy at the heart of this appeal, and then set out guidance for remand. In 4 No. 19-1830 Part IV, we affirm the district court’s order adding a new pro- posed class representative, and in Part V we briefly note the parties’ and our agreement that the proposed class definition must be limited to buyers of Allstate’s common stock. I. Factual and Procedural Background A. The Alleged Fraud and the Defense Response In early 2013, Allstate announced a new growth strategy in its auto insurance business: attracting more new customers by “softening” its underwriting standards. At the time, All- state disclosed that this approach could cause “some pres- sure” on its auto claims “frequency”—that is, new and poten- tially riskier customers might file more auto claims. Allstate CEO Thomas Wilson said that the company was aware of this potential and would monitor it and adjust business practices accordingly. Allstate and the plaintiffs agree on this much. Two years later, Allstate’s stock price dropped by more than 10 percent on August 4, 2015, immediately after Allstate announced that the higher claims rates it had experienced for three quarters had been fueled at least in part by the com- pany’s recent growth strategy, and that the company was “tightening some of our underwriting parameters.” Plaintiffs contend that the risk Allstate had flagged had materialized almost from the start of the new strategy. In re- quired SEC disclosures and investor conference calls, plain- tiffs say, Allstate executives said falsely at first that claim fre- quency trends had been “extremely favorable,” when claims in fact were spiking. Later, plaintiffs assert, when it became clear to the market that claim frequency had increased, All- state misled the market by falsely attributing the increases to other factors such as higher-than-usual precipitation and No. 19-1830 5 miles driven rather than the actual cause, the company’s growth strategy of taking on riskier business. These misrep- resentations were intentional, class plaintiffs say, because All- state analyzed its claim frequency data and its relationship to both internal and external factors so closely that its senior ex- ecutives would have been aware of the increases and their causes. The August 3, 2015 announcement prompted the sharp stock price drop because, as plaintiffs see things, All- state finally came clean and admitted that its aggressive growth strategy, not bad weather or more driving, had been to blame all along. Allstate tells a very different story. It says that those who understand the insurance business know that relaxed under- writing standards can often lead to increases in claims fre- quency. Allstate says that the market understood the risks of its growth strategy when it announced it in 2013. Any result- ing increase in claims frequency—to the extent not caused by external factors, which Allstate claims it was the first among its peers to identify and address—was a trade-off predictable both to the company and to the market. Any strategic adjust- ments were likewise encompassed by Allstate executives’ 2013 promise to “monitor” and to stay on top of its underwrit- ing parameters to ensure that this growth strategy in fact in- creased profitability. B. The District Court’s Class Certification In seeking class certification under Rule 23(b)(3), plaintiffs invoked the widely used Basic presumption to help show that common issues predominate over individual ones. To show the element of reliance in their fraud claims, plaintiffs offered evidence that Allstate stock trades in large, public, efficient markets, so that any false information defendants introduced 6 No. 19-1830 into the market could be presumed to have been baked in to the market price for Allstate stock. Under Basic, that presump- tion avoids the need for individual plaintiffs to prove they re- lied on particular false statements. 485 U.S. at 246–47. Allstate opposed certification, arguing that the Basic pre- sumption should not apply. Allstate offered evidence that it claimed “sever[ed] the link between the alleged misrepresen- tation and either the price received (or paid) by the plaintiff, or his decision to trade at a fair market price.” 485 U.S. at 248. Allstate contends that the market knew that its growth strat- egy would likely result in increased claims frequency, so that the market could not have relied on its alleged failures to dis- close either this risk or its actual occurrence. Plaintiffs charac- terize this position as a truth-on-the-market defense, which Amgen held may not be decided on class certification. Allstate characterizes its argument as showing a lack of price impact under Halliburton II. The district court characterized the dispute as “hotly con- tested and merits-based.” It therefore granted plaintiffs’ mo- tion for class certification while declining to find disputed facts on Allstate’s defense that there was no price impact, say- ing that the defense “essentially and improperly would re- quire this court [the district court] to decide disputed material issues of fact underlying plaintiff’s case.” The district court certified a plaintiff class under Federal Rule of Civil Proce- dure 23(b)(3) consisting of “all persons who purchased All- state Securities between October 29, 2014 and August 3, 2015, inclusive and who were damaged thereby.” On appeal, Allstate argues that class certification should be either vacated or denied outright. We can take outright de- No. 19-1830 7 nial off the table now. Much of plaintiffs’ evidence and analy- sis seems compelling and could easily support class certifica- tion. We also agree with the district court that Allstate’s price impact theory looks very much like the prohibited defenses of no materiality or “truth on the market.” As we read the Su- preme Court’s opinions together, however, we conclude that the close similarity does not allow a district court to avoid a price impact defense at the class certification stage. We try to explain below how to analyze this issue without, as it were, “thinking about a pink elephant,” i.e., without paying atten- tion to the obvious implications for the merits. II. Standard of Review We review the district court’s grant of class certification for an abuse of discretion. Arreola v. Godinez, 546 F.3d 788, 794 (7th Cir. 2008). “If, however, the district court bases its discre- tionary decision on an erroneous view of the law or a clearly erroneous assessment of the evidence, then it has necessarily abused its discretion.” Messner v. Northshore University HealthSystem, 669 F.3d 802, 811 (7th Cir. 2012) (vacating denial of class certification), citing Cooter & Gell v. Hartmarx Corp., 496 U.S. 384, 405 (1990); accord, Ervin v. OS Restaurant Services, Inc., 632 F.3d 971, 976 (7th Cir. 2011) (reversing denial of class certification). The requirements for class certification are not merely pleading requirements. Parties seeking class certification must prove that they can actually satisfy them. Comcast Corp. v. Behrend, 569 U.S. 27, 33 (2013); Messner, 669 F.3d at 811. If the parties dispute factual issues that are material under Rule 23, a court must “receive evidence … and resolve the disputes before deciding whether to certify the class.” Szabo v. Bridge- port Machines, Inc., 249 F.3d 672, 676 (7th Cir. 2001). 8 No. 19-1830 Complicating matters in cases like this, the same evidence may be relevant at both the class certification and merits stages. And notwithstanding Eisen and the general rule that the court should not decide the merits when deciding class certification, the Supreme Court has also taught that merits questions may be considered “to the extent—but only to the extent—that they are relevant” in applying the Rule 23 re- quirements. Amgen, 568 U.S. at 466, citing Wal–Mart Stores, Inc. v. Dukes, 564 U.S. 338, 351 n.6 (2011); see also General Tel- ephone Co. of Southwest v. Falcon, 457 U.S. 147, 160 (1982). III. The Predominance Requirement in Rule 10b-5 Class Actions A. Rule 23(b)(3) Predominance and the Elements of a Rule 10b- 5 Claim The focus in this appeal is the Rule 23(b)(3) requirement that “questions of law or fact common to class members pre- dominate over any questions affecting only individual mem- bers, and that a class action is superior to other available methods for fairly and efficiently adjudicating the contro- versy.” The predominance requirement is “stringent” but is “readily met in certain cases alleging consumer or securities fraud or violations of the antitrust laws.” Amchem Products, Inc. v. Windsor, 521 U.S. 591, 609, 625 (1997) (emphasis added). The Rule 23(b)(3) predominance requirement inherently requires the court to engage with the merits of the case, yet without deciding the merits. To decide predominance, the court must understand what the plaintiffs will need to prove and must evaluate the extent to which they can prove their case with common evidence. “In other words, a court weigh- ing class certification must walk a balance between evaluating evidence to determine whether a common question exists and No. 19-1830 9 predominates, without weighing that evidence to determine whether the plaintiff class will ultimately prevail on the mer- its.” Bell v. PNC Bank, N.A., 800 F.3d 360, 377 (7th Cir. 2015) (emphases added). We recognize the contradiction built into the standard. The judge must examine the evidence for its co- hesiveness while studiously ignoring its bearing on merits questions, even in cases much simpler than this one. In a securities fraud case under section 10(b) of the Securi- ties Exchange Act, 15 U.S.C. § 78j(b), and SEC Rule 10b-5, 17 C.F.R. § 240.10b-5, here are the elements for cases involving publicly traded securities: (1) a material misrepresentation (or omission); (2) scienter, i.e., a wrongful state of mind; (3) a connection with the purchase or sale of a security; (4) reliance, often referred to in cases involving public securities markets (fraud-on-the-market cases) as “transaction causation”; (5) economic loss; and (6) “loss causation,” i.e., a causal connection be- tween the material misrepresentation and the loss. Dura Pharmaceuticals, Inc. v. Broudo, 544 U.S. 336, 341–42 (cita- tions and emphases omitted); accord, e.g., Schleicher v. Wendt, 618 F.3d 679, 682 (7th Cir. 2010). “When a large, public company makes statements that are said to be false,” allegations of securities fraud are particu- larly well-suited to class adjudication. See Schleicher, 618 F.3d at 681. We analyze these six elements in two groups—the first 10 No. 19-1830 three and the last three—to illustrate both why this is so and the central role the Basic presumption plays in both groups. On a Rule 10b-5 claim, plaintiffs will succeed or fail on the merits of the first three elements based on a common set of evidence, at least where the securities are traded in large, pub- lic, and efficient markets. Companies issuing such securities ordinarily disseminate information about their past, current, and expected future performance through channels that reach the market as a whole. Here, for example, plaintiffs base their fraud claims on statements made by Allstate and its execu- tives in public SEC filings, quarterly reports disseminated to the public, and conference calls with analysts from leading in- vestment firms. The falsity and materiality of these represen- tations (element one) and whether Allstate executives made any misrepresentations with scienter (element two, see Ernst & Ernst v. Hochfelder, 425 U.S. 185, 197 (1976); Schleicher, 618 F.3d at 681) are merits questions. At class certification, the is- sue is not whether plaintiffs will be able to prove these ele- ments on the merits, but only whether their proof will be com- mon for all plaintiffs, win or lose. A case built on public state- ments to markets is based on common evidence on these ele- ments. The third element of the 10b-5 claim, a connection to the purchase or sale of a security, will also rest on common evi- dence in class actions against public companies. Though class members will have bought and sold securities on different No. 19-1830 11 dates, price information for publicly traded securities is com- mon and readily available.1 The fourth, fifth, and sixth elements—reliance, economic loss, and loss causation—are closely related to each other, and for reliance and loss causation, the question of common proof can be more complex. The statute that now expressly author- izes private securities fraud litigation, 15 U.S.C. § 78u-4, ena- bles plaintiffs to recover damages based on their economic losses. In its simplest form, a plaintiff’s economic loss is the difference between the amount she paid to buy the security (higher than it should have been, in successful 10b-5 cases) and the amount she received when she sold it. See, e.g., Dura Pharmaceuticals, 544 U.S. at 344–45. For publicly traded secu- rities, individual loss can be a simple arithmetic calculation using common evidence about the security’s price move- ments over the relevant time. B. The Basic Presumption at Class Certification A sharp drop in share price alone is not enough for a class to be certified. Rather, 15 U.S.C. § 78u−4(b)(4) requires the plaintiff to prove reliance, also referred to as loss causation, i.e., “that the act or omission of the defendant alleged to violate this chapter caused the loss for which the plaintiff seeks to re- cover damages.” (Emphases added.) 1Some aspects of this element require individualized proof, but they “can be resolved mechanically. A computer can sort them out using a da- tabase of time and quantity information.” Schleicher, 618 F.3d at 681. The information populating that database will be evidence common to all class members. 12 No. 19-1830 For proof of reliance, the Supreme Court endorsed in Basic the fraud-on-the-market theory, in which “reliance is pre- sumed when the statements at issue become public.” Ston- eridge Investment Partners, LLC v. Scientific-Atlanta, 552 U.S. 148, 159 (2008). The Basic presumption provides a practical way for plaintiffs to prove reliance through common, class- wide evidence in the context of modern securities markets where millions of shares change hands daily without the “face-to-face transactions contemplated by early fraud cases.” Basic, 485 U.S. at 243–44. The Basic presumption of reliance is based on the efficient market hypothesis: “the market price of shares traded on well-developed markets reflects all publicly available information, and, hence, any material misrepresen- tations.” Id. at 246.2 As a result, if the securities in question trade on an efficient market, then the market itself provides the causal connection between a misrepresentation and the price of the stock. “The price both transmits the information and causes the loss.” Schleicher, 618 F.3d at 682. The Basic presumption shifts the re- liance inquiry from whether an individual plaintiff relied on particular representations in buying or selling the security to whether all individuals trading in a given security during a given time period “relied on the integrity of the price set by the market.” Basic, 485 U.S. at 226. 2The efficient capital markets hypothesis has been criticized, but in Halliburton II, the Supreme Court rejected arguments to overrule Basic. 573 U.S. at 271–72. Whatever the empirical merits of critiques of the efficient market hypothesis, see, e.g., Donald C. Langevoort, Theories, Assumptions, and Securities Regulation: Market Efficiency Revisited, 140 U. Pa. L. Rev. 851 (1992), as a matter of law it remains the foundation for fraud-on-the- market claims. No. 19-1830 13 The fourth, fifth, and sixth elements of a 10b-5 claim are thus intertwined legally, conceptually, and factually. But the Supreme Court has taught that these elements must be con- sidered at different stages of the case. To certify a class under Rule 23(b)(3), a plaintiff must show the ability to use common evidence of reliance, i.e., to use the Basic presumption. Loss causation, on the other hand, must be entirely reserved for the merits. Halliburton I, 563 U.S. 804 (2011). Even when plaintiffs show that the securities trade in effi- cient markets, the Basic presumption is rebuttable. “Any showing that severs the link between the alleged misrepresen- tation and either the price received (or paid) by the plaintiff, or his decision to trade at a fair market price, will be sufficient to rebut the presumption of reliance.” Basic, 485 U.S. at 248. In the latter category, defendants can try to show that plaintiffs did not in fact rely on the integrity of the market price when they traded or that the securities did not in fact trade in an efficient market. Id. at 249. Basic also allows defendants to show that their alleged misrepresentations did not actually affect the market price in two ways that are difficult to distinguish from the merits of the plaintiff’s claims. First, if the “‘market makers’ were privy to the truth” about information allegedly concealed, or sec- ond, if “news of [the allegedly concealed truth] credibly en- tered the market and dissipated the effects of the misstate- ment,” the causal connection between the alleged fraud and the market price would be broken. Id. at 248–49. Under the first option, the defense shows that only true information was impounded in the market price at the time of purchase; the second option does the same by the time of sale. 14 No. 19-1830 As the Court later explained, “an inflated purchase price will not itself constitute or proximately cause the relevant eco- nomic loss” because that causation is demonstrated only when no alternate causes have intervened. Dura Pharmaceuti- cals, 544 U.S. at 342. The second rebuttal option under Basic demonstrates the close relationship between reliance and loss causation. Both inquiries focus on whether an intervening cause disrupted the connection between a false statement and a trade relying on the assumption that the false statement was factored into the market price. C. The Halliburton/Amgen Trilogy In a series of decisions from 2011 to 2014, the Supreme Court grappled with the conceptual and evidentiary overlap between the Basic presumption of reliance and other elements of 10b-5 claims in deciding on class certification. The three key cases are Erica P. John Fund, Inc. v. Halliburton Co., 563 U.S. 804 (2011) (Halliburton I); Amgen Inc. v. Connecticut Retirement Plans and Trust Funds, 568 U.S. 455 (2013); and Halliburton Co. v. Erica P. John Fund, Inc., 573 U.S. 258 (2014) (Halliburton II). Together, they pose the central problem in this appeal. Halliburton I vacated the denial of class certification in a securities fraud case. The Court held that securities fraud plaintiffs need not “prove loss causation in order to obtain class certification.” 563 U.S. at 807. The Court was careful to distinguish loss causation from the related yet distinct concept of “transaction causation” that the Court has long held is synonymous with reliance under Rule 10b-5. See id. at 812, citing Dura Pharmaceuticals, 544 U.S. at 341–42, citing in turn Basic, 485 U.S. at 248–49. This is Basic’s “fundamental premise—that an investor presumptively relies on a misrepresentation so long as it was reflected in the market No. 19-1830 15 price at the time of the transaction.” Halliburton I, 563 U.S. at 813. To invoke this presumption, plaintiffs must show that “the alleged misrepresentations were publicly known … , that the stock traded in an efficient market, and that the relevant transaction took place between the time the misrepresentations were made and the time the truth was revealed.” Id. at 811 (internal quotation marks omitted). This is distinct, the Court explained, from “[t]he fact that a subsequent loss may have been caused by factors other than the revelation of a misrepresentation,” which bears “no logical connection to the facts necessary to establish the efficient market predicate to the fraud-on-the-market theory.” Id. at 813. Halliburton I also distinguished between “loss causation” and “price impact.” The defendant had argued that the real question was whether plaintiffs had demonstrated “price im- pact” as required for their fraud-on-the-market theory, or “whether the alleged misrepresentations affected the market price in the first place.” Id. at 814. The Court defined price im- pact as “the effect of a misrepresentation on a stock price” and rejected Halliburton’s “wishful interpretation” of the Court of Appeals’ loss causation requirement as the “theory … that if a misrepresentation does not affect market price, an investor cannot be said to have relied on the misrepresentation merely because he purchased stock at that price.” Id. Though Halli- burton I did not endorse or reject the need to prove or disprove price impact at the class certification stage, it firmly distin- guished between price impact and loss causation. Id. In short, after Halliburton I the reliance inquiry needed to focus on the mix of factors that caused the purchase of the security in ques- tion, not on any later drop in price leading to plaintiffs’ eco- nomic losses. 16 No. 19-1830 Two years later in Amgen, the Supreme Court affirmed a grant of class certification, holding that the defense was not entitled to litigate the issue of materiality at the class certifica- tion stage. Amgen resolved the core tension that arises from including the first 10b–5 element, a material misrepresenta- tion or omission, in the Basic presumption aimed at reliance at class certification. The Court recognized that “materiality is not only an element of the Rule 10b-5 cause of action; it is also an essential predicate of the fraud-on-the-market theory.” Amgen, 568 U.S. at 466. On this foundation, the defense had tried to litigate materiality to defeat class certification in Amgen: “Because immaterial information, by definition, does not affect market price[, the defense argued], it cannot be re- lied upon indirectly by investors who, as the fraud-on-the- market theory presumes, rely on the market price’s integrity.” Id. at 466–67. The Amgen Court rejected this effort. The Court agreed that materiality was an “essential predicate” of fraud-on-the- market reliance, but it explained that “the pivotal inquiry is whether proof of materiality is needed to ensure” the pre- dominance of common questions of law or fact. 568 U.S. at 467. The Court reasoned that materiality, as an objective ques- tion, will always be proved through common evidence, and that “the failure of proof on the element of materiality” “would not cause individual reliance questions to overwhelm the questions common to the class” but “would end the case” on the merits for all plaintiffs. Id. at 467–68. In fact, the Court noted, a failure to prove materiality bars even individual re- covery under Rule 10b-5, let alone class certification. Id. at 474. Amgen therefore approved the district court’s choice to disre- gard the defense evidence offered to rebut plaintiffs’ evidence in support of the Basic presumption, saying that a “truth-on- No. 19-1830 17 the-market” defense “is a matter for trial” (or summary judg- ment). Id. at 481–82. After rejecting defense efforts to rebut the Basic presump- tion in both Halliburton I and Amgen, the Court returned the next year in Halliburton II to the role of price impact evidence at the certification stage. After the remand ordered in Halli- burton I, the district court had granted class certification and the Fifth Circuit had affirmed. In Halliburton II, the case re- turned to the Supreme Court, which again vacated and re- manded on class certification. The defense argued that Basic should be overruled. The Court first said it was leaving Basic intact. 573 U.S. at 271−72. The Court then considered other ways for defendants to argue that alleged false statements had no price impact. First, the Court noted, this evidence can always be intro- duced at the merits stage. Id. at 280–81. Second, the Court con- firmed that “defendants may introduce price impact at the class certification stage, so long as it is for the purpose of coun- tering a plaintiff’s showing of market efficiency, rather than directly rebutting the [Basic] presumption,” noting that plain- tiffs often use price impact evidence to demonstrate market efficiency, which is needed to invoke the Basic presumption in the first place. Id. at 280. The class plaintiffs had urged the Court to restrict district courts’ use of price impact evidence at the certification stage. The Court made clear that the defense is entitled to offer evi- dence of a lack of price impact at the class certification stage: “While Basic allows plaintiffs to establish th[e] precondition [of price impact] indirectly, it does not require courts to ignore a defendant’s direct, more salient evidence showing that the 18 No. 19-1830 alleged misrepresentation did not actually affect the stock’s market price and, consequently, that the Basic presumption does not apply.” Id. at 282. The challenge lies in the fact that both reliance and loss causation overlap the materiality of the alleged misrepresen- tations. Judge Trauger captured the problem nicely: At the heart of this confusing area of the case law is the fact that all three concepts addressed—loss causation, materiality, and price impact—are, in essence, slightly different takes on the same fundamental question: Did a statement matter? As a result, evidence relevant to each issue is likely also to be relevant to the others. … Taking a piece of evidence and placing it in any of the three boxes, to the exclusion of the others, would be an artificial and logically questionable exercise. Grae v. Corrections Corp. of America, 330 F.R.D. 481, 498 (M.D. Tenn. 2019) (in wake of Halliburton II, reconsidering denial and granting class certification). Hence the challenge: how can a district court deciding class certification (a) decide whether reliance can be proven by common evidence without (b) delving too far into the merits of the materiality or falsity of the representations at issue, while still (c) reserving loss causation entirely for the merits phase? We are obliged to follow all three cases, and we must read them together. A district court deciding whether the Basic presumption applies must consciously avoid deciding materiality and loss causation. Halliburton I and Amgen require that much. At the same time, a district court must be No. 19-1830 19 willing to consider evidence offered by the defense to show that the alleged misrepresentations did not actually affect the price of the securities. Halliburton II requires that. And yes, the same evidence is likely to have obvious implications for the off-limits merits issues of materiality and loss causation. Halliburton II teaches, however, that a district court may not use the overlap to refuse to consider the evidence. The court must still consider the evidence as relevant to price impact (also known as transaction causation). Plaintiffs seeking class certification need not offer direct evidence of price impact. Halliburton II, 573 U.S. at 279. But Halliburton II gave defendants half a loaf. The defense is enti- tled to make “[a]ny showing that severs the link between the alleged misrepresentation and either the price received (or paid) by the plaintiff, or his decision to trade at a fair market price,” and such a showing “will be sufficient to rebut the pre- sumption of reliance.” Basic, 485 U.S. at 248. This showing may include direct evidence demonstrating that the alleged misrepresentations had no impact on the stock price. Hallibur- ton II, 573 U.S. at 279–80. Indeed, “an indirect proxy should not preclude direct evidence when such evidence is availa- ble.” Id. at 281. The logical corollary is that although plaintiffs need not initially introduce direct evidence of price impact, they may choose to do so as a means of responding to (or an- ticipating) a defendant’s direct rebuttal evidence. The crucial challenge for the district court is to decide only the issues the Supreme Court has said should be decided for class certification while resisting the temptation to draw what may be obvious inferences for the closely related issues that must be left for the merits, including materiality and loss cau- sation, as required by Halliburton I and Amgen. 20 No. 19-1830 Finally, the appropriate focus of the inquiry into “the ele- ment of reliance in a private Rule 10b-5 action [is] transaction causation, not loss causation.” Halliburton I, 563 U.S. at 812 (ci- tations and quotations omitted). At class certification, plain- tiffs need not “show that a misrepresentation that affected the integrity of the market price also caused a subsequent eco- nomic loss.” Id. (emphasis in original). Price impact evidence may be relevant to both the transaction- and loss-causation inquiries. As noted, in an efficient market, “[t]he price both transmits the information and causes the loss.” Schleicher, 618 F.3d at 682. Such evidence will likely address the drop in price at the end of a class period that is usually the centerpiece of the plaintiffs’ case. But in deciding whether to certify a plain- tiff class, a district court must consider that information as in- direct evidence relevant to transaction causation, not as direct evidence for or against loss causation. The analysis looks backward to the time of purchase—to whether all purchasers can be said to have “relied on the integrity of the price set by the market,” Basic, 485 U.S. at 226—not to what may or may not have happened after that. The district court here made a legal error by embracing Amgen at the expense of Halliburton II—a tempting way of more cleanly managing price impact evidence—rather than engaging in the messier but required process of simultane- ously complying with the instructions from the Supreme Court in both cases. We must therefore vacate the class certi- fication and order and remand for further consideration of ev- idence relevant to price impact. We can draw a few conclu- sions about the requirements for that consideration. No. 19-1830 21 D. Guidance for Remand 1. The Scope of the Evidence The first pragmatic question at stake here is the scope of the evidence that district courts are permitted and required to admit at the class certification stage when securities fraud plaintiffs invoke the fraud-on-the-market theory. The Basic line of cases imposes few if any limits. Recall that Basic itself allows defendants to make “Any showing that severs the link between the alleged misrepresentation and either the price re- ceived (or paid) by the plaintiff, or his decision to trade at a fair market price.” 485 U.S. at 248 (emphasis added). And Hal- liburton II specifically endorsed the use of both direct and in- direct evidence of price impact. 573 U.S. at 283. Allstate here does not seek to introduce additional evidence; it only takes issue with whether and how that evidence was evaluated. The district court appropriately admitted Allstate’s desired evi- dence: an economist’s report analyzing price impact. One concurring opinion in Halliburton II noted that “[a]dvancing price impact consideration from the merits stage to the certification stage may broaden the scope of discovery available at certification.” Id. at 284 (Ginsburg, J., joined by Breyer, and Sotomayor, JJ., concurring). We agree, and this point deserves emphasis because of its implications for man- aging discovery. Given the significant and growing overlap between the evidence at stake at the certification and merits stages, district courts may well choose not to bifurcate discov- ery at all in putative fraud-on-the-market securities class ac- tions.3 3 The Manual for Complex Litigation recognizes that a strict separa- tion between class and merits discovery can be artificial and that it is well 22 No. 19-1830 2. Managing the Basic Presumption The major securities precedents of the past decade have confirmed that the fraud-on-the-market presumption en- dorsed in Basic creates a burden-shifting framework. We agree with the Second Circuit in interpreting this dimension of Basic. See Waggoner v. Barclays PLC, 875 F.3d 79, 96–104 (2d Cir. 2017). As a threshold matter, we agree with Waggoner that Federal Rule of Evidence 301 “imposes no impediment to our conclusion that [once plaintiffs have made a prima facie showing] the burden of persuasion, not production, to rebut the Basic presumption shifts to defendants.” Id. at 103.4 within the district court’s discretion not to bifurcate discovery on certain substantive issues: Discovery relevant only to the merits delays the certifica- tion decision and may ultimately be unnecessary. Courts often bifurcate discovery between certification issues and those related to the merits of the allegations. Generally, discovery into certification issues pertains to the require- ments of Rule 23 and tests whether the claims and de- fenses are susceptible to class-wide proof; discovery into the merits pertains to the strength or weaknesses of the claims or defenses and tests whether they are likely to succeed. There is not always a bright line between the two. Courts have recognized that information about the nature of the claims on the merits and the proof that they require is important to deciding certification. Arbitrary insistence on the merits/class discovery distinction some- times thwarts the informed judicial assessment that cur- rent class certification practice emphasizes. Federal Judicial Center, Manual for Complex Litigation, Fourth, § 21.14 Precertification Discovery 256 (2004). 4 We do not believe our holding here, following the Second Circuit, is fundamentally inconsistent with that of the Eighth Circuit in IBEW Local No. 19-1830 23 In granting class certification here, the district court held in effect that plaintiffs had made at least a prima facie show- ing sufficient to invoke the Basic presumption. On remand, the burdens of production and persuasion will shift to All- state. Allstate evidently believes that its expert report meets its burden of production. The district court should assess whether Allstate has met its burden of persuasion by a pre- ponderance of evidence, see Arkansas Teachers Ret. Sys. v. Gold- man Sachs Grp., Inc., 879 F.3d 474, 485 (2d Cir. 2018), taking into account plaintiffs’ rebuttal reports and additional evi- dence challenging Allstate’s showing. “It would be incon- sistent with Halliburton II to require that plaintiffs meet this evidentiary burden while allowing defendants to rebut the Basic presumption by simply producing some evidence of market inefficiency, but not demonstrating its inefficiency to the district court.” Waggoner, 875 F.3d at 100 (emphasis in original). After all, Basic said that “[a]ny showing that severs the link” would be sufficient to rebut the presumption, 485 U.S. at 248 (emphasis added), not that mere production of ev- idence would defeat the presumption. See also Merritt B. Fox, Halliburton II: It All Depends on What Defendants Need to Show to Establish No Impact on Price, 70 Bus. Law. 437, 448 n.27 (2015). 98 Pension Fund v. Best Buy Co., 818 F.3d 775, 782 (8th Cir. 2016), which cited Federal Rule of Evidence 301 only for the following proposition: “We agree with the district court that, when plaintiffs presented a prima facie case that the Basic presumption applies to their claims, defendants had the burden to come forward with evidence showing a lack of price impact. See Fed. R. Evid. 301 (‘the party against whom a presumption is directed has the burden of producing evidence to rebut the presumption’).” 24 No. 19-1830 With the evidence admitted and the burdens allocated, the district court must then make findings needed to decide class certification while resisting the temptation to draw even ob- vious inferences on topics that are forbidden at this stage: ma- teriality and loss causation. The court must assess evidence that may speak directly to the forbidden merits inquiries of materiality and loss causation, while evaluating it only for what it reveals about the core Basic inquiry of transaction cau- sation. The heart of the factual dispute between Allstate and the class plaintiffs is the proper characterization of the evidence contained in the report submitted by Allstate’s expert, Lucy Allen. The Allen report makes two primary claims about the nine statements plaintiffs alleged to be misrepresentations. First, Allen said that she found no statistically significant in- crease in Allstate’s stock price following any of the alleged misrepresentations, from which she argues that the state- ments had no price impact. Allen Rpt. at 1, 16. Second, Allen said that: the alleged misrepresentations could not [i.e., as a matter of logic] have had price impact because Allstate’s growth strategy, and the fact that the Company’s growth strategy was expected to cause higher claims frequencies, was publicly disclosed in the Company’s conference calls prior to the alleged Class Period, was covered in analyst reports on the Company published prior to and at the beginning of the alleged Class Period and, in an efficient market, would have already been impounded into Allstate’s stock price. Allen Rpt. at 1. In other words, Allstate’s position is that because the mar- ket at all times had correct information, the later statements by Allstate that plaintiffs treat as corrective disclosures could not have caused any concurrent price reactions. Plaintiffs con- tend, in turn, that Allstate had at best disclosed only potential risks, but upon numerous occasions chose not to inform the market that these dangers were in fact being realized. Plain- tiffs therefore characterize the Allen report as a truth-on-the- market defense forbidden by Amgen. The concept of “price impact” boils down to the question of whether an alleged misrepresentation “actually affect[ed] the market price” of the security in question. Halliburton II, 573 U.S. at 269, citing Basic, 485 U.S. at 248–49. The question of which factors affected the market price of a security could be asked in theory with respect to any given date. If asked with respect to the day the plaintiff sold, the question looks very much like one of loss causation. This is why the Supreme Court has held that the relevant temporal focus upon class certification is at the time of purchase—that is, price impact as an essential mechanism of “transaction causation.” Hallibur- ton I, 563 U.S. at 812, citing Dura Pharmaceuticals, 544 U.S. at 341–42. Data from later times may be relevant to this inquiry, but only insofar as they help the district court determine the information impounded into the price at the time of the initial transaction. To explain, consider a simplified model of price impact. The stock price of a company is x on January 1 and remains at x through the end of the month. On January 31, the company makes a material misrepresentation about, say, its growth strategy that is received enthusiastically by the market. On February 1, assuming an efficient market, the stock price 26 No. 19-1830 shoots up, say to 1.25x. On March 1, the company makes a corrective disclosure, saying that the January 31 statement was false and that the company had never had any intention to pursue that strategy. On March 2, the stock price immedi- ately returns to x. No other information about the company enters the market during this period. Anyone who purchased the stock during February and held the stock past March 1 would have been injured in the amount of 0.25x. The misrep- resentation caused both the transaction and the loss via the mechanism of price. The March 1 statement and ensuing price drop are the best evidence available of the impact of the Janu- ary 31 statement on the price. They are direct evidence as to loss causation and indirect evidence as to transaction causation for buyers who purchased between the January 31 and March 1 statements. Real allegations of securities fraud are never so simple, of course. In this case, for example, plaintiffs allege the “inflation maintenance” version of the theory. We endorsed this theory in Glickenhaus & Co. v. Household Intern., Inc., 787 F.3d 408 (7th Cir. 2015), and affirm its viability again now. In the real-world market, as opposed to our simple example above, stock prices respond to many different sources of information, often in- cluding both good and bad news about the company, and truths as well as the alleged falsehoods. Sustaining an infla- tion maintenance theory requires plaintiffs “to prove … that the defendants’ false statements caused the stock price to re- main higher than it would have been had the statements been truthful,” even if the price itself does not change by a single cent. Id. at 419. We have observed that a direct approach to this question is difficult “because it requires knowing a counterfactual: No. 19-1830 27 what the price would have been without the false statement.” Id. at 415. The stock price may even decline after a false state- ment, but be inflated nonetheless “because the price might have fallen even more” if the full extent of the bad news were known. Id. For this reason, price reaction (the simple move- ment of the price in response to a given statement) is quite different from the legal concept of price impact. Accordingly, the Allen report’s finding that a lack of price reaction after the nine statements at issue indicates that they had no price im- pact does not actually resolve the legal issue of price impact. We affirm the district court’s recognition of plaintiffs’ inflation maintenance theory here.5 5 Our view on this point comports with that of the Eleventh Circuit, which has explained: A “fraud on the market” occurs when a material misrep- resentation is knowingly disseminated to an information- ally efficient market. Basic, 485 U.S. at 247. Just as an effi- cient market translates all available truthful information into the stock price, the market processes the publicly dis- seminated falsehood and prices it into the stock as well. See id. at 241–42, 243–44, 246–47. The market price of the stock will then include an artificial “inflationary” value—the amount that the market mistakenly attributes to the stock based on the fraudulent misinformation. So long as the falsehood remains uncorrected, it will continue to taint the total mix of available public information, and the mar- ket will continue to attribute the artificial inflation to the stock, day after day. If and when the misinformation is finally corrected by the release of truthful information (of- ten called a “corrective disclosure”), the market will re- calibrate the stock price to account for this change in in- 28 No. 19-1830 But this leaves the second core dispute over the Allen re- port’s findings: the claim that the alleged misrepresentations could not have had a price impact because they were not news to the market, as demonstrated, in part, by later stock price movements and analyst reports. In Glickenhaus, we acknowl- edged that “[t]he best way to determine the impact of a false statement is to observe what happens when the truth is finally disclosed and use that to work backward, on the assumption that the lie’s positive effect on the share price is equal to the additive inverse of the truth’s negative effect. (Put more simply: what goes up, must come down.)” 787 F.3d at 415. In essence, we take Allen’s argument to be that because nothing came down after the alleged corrective disclosures, nothing can have gone up in the first place. Yet that argument is difficult for us to square with the 10 percent price drop on August 4, 2015, and the Allen report offers little on that score. On remand, the district court may take into account expert findings with regard to “ex post price distortion,” or “[w]hether the stock price responds when the [alleged] fraud is revealed to the market,” only as backward-looking, indirect evidence of the core question here—“ex ante price distortion” as a constituent part of transaction causation, or “whether formation, eliminating whatever artificial value it had at- tributed to the price. That is, the inflation within the stock price will “dissipate.” FindWhat Investor Grp. v. FindWhat.com, 658 F.3d 1282, 1310 (11th Cir. 2011). In keeping with this analysis, the FindWhat court held “that the se- curities laws prohibit corporate representatives from knowingly peddling material misrepresentations to the public—regardless of whether the statements introduce a new falsehood to the market or merely confirm misinformation already in the marketplace.” Id. at 1290. No. 19-1830 29 stock price [is] distorted at the time that the plaintiff trades.” Jill E. Fisch, The Future of Price Distortion in Federal Securities Fraud Litigation, 10 Duke J. Const. L. & Pub. Pol’y 87, 94 (2015).6 As the district court noted, separating this argument from the kind of truth-on-the-market defense proscribed by Amgen’s holding on materiality cuts extraordinarily fine. We see this case as a question of scope and specificity. Allstate claims that its broad statements made at a high level of generality—that profitability could decrease as a result of its strategic decision, disclosed to the market, to soften underwriting standards—encompassed any subsequent auto claim frequency spikes that may or may not have happened or that may or may not have been timely disclosed to the 6 Both sides’ experts here have submitted event studies, as is typical in securities litigation. Indeed, since Basic, event studies have come to be treated as the sine qua non for proving or disproving price impact and loss causation. See Michael J. Kaufman & John M. Wunderlich, Regressing: The Troubling Dispositive Role of Event Studies in Securities Fraud Litigation, 15 Stan. J.L. Bus. & Fin. 183, 208 (2009). “Event studies may help, but there is no reason in the class certification inquiry to limit evidence to those, espe- cially in ‘confirmatory lie’ cases. Courts should be open to all probative evidence on that question—qualitative as well as quantitative—aided by a good dose of common sense.” Donald C. Langevoort, Judgment Day for Fraud-on-the-Market: Reflections on Amgen and the Second Coming of Halli- burton, 57 Ariz. L. Rev. 37, 56 (2015). Econometrics, finance, and securities law experts have criticized the methods used in event studies prepared for litigation, and they caution courts to think carefully about how such study designs and findings often do a poor job of answering the legal questions at stake. See, e.g., Alon Brav & J.B. Heaton, Event Studies in Se- curities Litigation: Low Power, Confounding Effects, and Bias, 93 Wash. U. L. Rev. 583, 585–87 (2015); Jill E. Fisch, Jonah B. Gelbach, & Jonathan Klick, The Logic and Limits of Event Studies in Securities Fraud Litigation, 96 Tex. L. Rev. 553, 616 (2018). 30 No. 19-1830 market. Plaintiffs counter that there is a meaningful difference between knowing of a possible risk and knowing that the danger has in fact been realized. For plaintiffs, the more general representations that Allstate made do not encompass the more specific representations it should have made— especially where, as plaintiffs argue, those representations were not merely vague but actively misleading. Again, the question at class certification is not the truthfulness or materiality of any of Allstate’s representations with regard to these questions, but whether they are susceptible of common proof, and the level of specificity of the information the market would have understood the price of Allstate’s common stock to transmit at the time of the purchase transaction. Accordingly, we vacate the class certification and remand for further proceedings because the Supreme Court has made clear that factfinding as to whether common issues predomi- nate is not only proper but necessary at the class certification stage. The Basic presumption is the linchpin of plaintiffs’ pre- dominance argument, so the district court must find relevant facts as to whether they may invoke that presumption. IV. Adding a New Class Representative Before granting class certification, the district court granted plaintiffs’ motion for leave to amend their complaint to add an additional class representative, the City of Provi- dence Employee Retirement System, known as Providence ERS. Dkt. No. 105 (Sept. 12, 2018). Allstate argues in this ap- peal that granting leave to amend was an abuse of discretion because the new class representative’s claims are barred by the two-year statute of limitations in 28 U.S.C. § 1658(b). Plaintiffs respond that the new named class representative No. 19-1830 31 was entitled to rely on American Pipe tolling, so that its claims were already brought before the court in a timely way. See American Pipe & Construction Co. v. Utah, 414 U.S. 538 (1974). Under American Pipe, the addition of Providence ERS as a named representative was a routine application of Rule 15 and an essential step in managing a class action. The issue is a legal one, and it is important for managing class actions fairly and in compliance with Rule 23. The issue is fully briefed, and it would be helpful to resolve it now, keeping in mind that one purpose of Rule 23(f) appeals is to develop the law of class actions. Mullins v. Direct Digital, LLC, 795 F.3d 654, 658–59 (7th Cir. 2015), citing Blair v. Equifax Check Services, Inc., 181 F.3d 832, 835 (7th Cir. 1999). Allstate argues that China Agritech, Inc. v. Resh, 138 S. Ct. 1800 (2018), now bars the addition of Providence ERS as a class representative. Allstate offers two theories. The first is that China Agritech limited American Pipe so that Providence ERS may not become a class representative after the statute of limitations would have run on its claims, absent American Pipe tolling. The second theory is that Providence ERS somehow waived its right to seek appointment as a lead plaintiff by not filing an application to do so at the outset of the case. Both theories rest on a misreading of China Agritech. The practical implications of Allstate’s position would be arbitrary and unfair, and would undermine the purposes of American Pipe tolling and the larger purposes of Rule 23. All- state proposes to prohibit any class member who has relied on American Pipe tolling from stepping up to act as a class rep- resentative after the statute of limitations would have run for filing an entirely new action based on the same events. As a practical matter, that rule would commit the fate of class 32 No. 19-1830 claims inexorably to the initial class representative, regardless of issues that might arise concerning the initial representa- tive’s ability or willingness to continue serving in that role. Allstate’s proposal would also impose arbitrary and poten- tially fatal obstacles where a district court finds it appropriate or even necessary to split a class or to create sub-classes. These arbitrary obstacles would undermine effective case manage- ment and would conflict with well-established practices and precedents. In American Pipe, the Supreme Court held that the timely filing of a class action tolls the applicable statutes of limita- tions for all persons within the scope of the class alleged in the complaint. If certification is ultimately denied, those per- sons within the scope of the proposed class may then choose to pursue individual claims either in the still-pending case or in new individual cases. 414 U.S. at 552–53; see also Crown, Cork & Seal Co. v. Parker, 462 U.S. 345, 350 (1983) (broadening American Pipe to apply to separate actions by members of pu- tative class). The American Pipe rule eliminates the need for members of the putative class to rush to court to protect their rights while class certification is still pending and uncertain in the original action. In China Agritech, the Supreme Court dealt with an entirely different statute of limitations issue for class actions: whether American Pipe tolling applies to successive attempts to file en- tirely new class actions, effectively stacking class actions in the hope that a court somewhere can be convinced to certify a class in another case, filed perhaps many years after the stat- ute of limitations has expired. The Supreme Court held in China Agritech that when class certification is denied, a mem- ber of the putative class may join the existing suit or promptly No. 19-1830 33 file an individual action, but she may not start a new class ac- tion beyond the time allowed by the statute of limitations. 138 S. Ct. at 1806. Allstate would read China Agritech much more broadly to prohibit any addition or substitution of a new class repre- sentative within the original class action after the statute of limitations period would have run, but for American Pipe toll- ing. We see no hint in the China Agritech opinion or its reason- ing that would support this proposed extension. American Pipe tolling is intended to promote efficiency and economy in litigation. 414 U.S. at 553. Prohibiting its use within the origi- nal class action to add new class representatives, whether be- cause they would be better representatives, because class def- initions are modified, because subclasses are needed, or for any other case-management reason, would arbitrarily—even randomly—undermine those goals of efficiency and econ- omy. Allstate’s reading would also undermine the benefits of American Pipe by encouraging as many individual members of the putative class to join as parties as quickly as possible. Second, we reject Allstate’s argument that Providence ERS somehow waived its ability to act as a class representative in this case by relying for a time on the original lead plaintiff to pursue the case. China Agritech cautions those interested in fil- ing their own class actions to do so early so as to prevent the stacking of separate, successive class actions. 138 S. Ct. at 1810–11. But plaintiffs who are part of the original putative class and who seek only to take on a new role in an existing action are not required to do so where, as here, the statute of limitations was already tolled on their behalf by the initial class complaint. See American Pipe, 414 U.S. at 552–55. The whole point of American Pipe tolling is that such parties are 34 No. 19-1830 entitled to watch and wait while the initial class representa- tive pursues the case. Plaintiffs here sought only to rearrange the seating chart within a single, ongoing action. What they proposed amounted to an ordinary pleading amendment governed by Federal Rule of Civil Procedure 15. Plaintiffs’ motion to add Providence ERS as a class representative was in substance a motion to amend a pleading (here, the class complaint) relat- ing back to the initial pleading within the meaning of Rule 15(c)(1). The amended complaint falls squarely within Rule 15(c)(1)(B), which allows relation back when “the amendment asserts a claim or defense that arose out of the conduct, trans- action, or occurrence set out—or attempted to be set out—in the original pleading.” The alleged fraud is the same in both pleadings. The new representative may be able to help resolve or avoid problems with another class representative or may enable certification of a modified class or subclasses. Adding Providence ERS did not impair any “interest in repose.” See Krupski v. Costa Crociere S.p.A., 560 U.S. 538, 550 (2010); accord Joseph v. Elan Motorsports Technologies Racing Corp., 638 F.3d 555, 558, 559–60 (7th Cir. 2011). By the end of the limitations period, Allstate already knew it was facing a class action. Adding Providence ERS as a class representative caused All- state no cognizable prejudice and was otherwise appropriate. V. The Class Definition Both sides have requested that we change the definition of the proposed class from “all persons who purchased Allstate Securities between October 29, 2014 and August 3, 2015, inclu- sive and who were damaged thereby,” as appears in the dis- trict court’s class certification order, to “all persons who pur- chased Allstate common stock between October 29, 2014 and No. 19-1830 35 August 3, 2015, inclusive and who were damaged thereby,” as litigated in the district court. This was likely nothing more than an inadvertent error in the order. Upon remand, if the district court recertifies the class, it should be defined to in- clude only buyers of common stock. * * * The district court’s order granting leave to amend the complaint to add Providence Employee Retirement System as class representative is AFFIRMED. The district court’s order certifying the plaintiff class is VACATED and the case is REMANDED for further proceedings consistent with this opinion.