Summary
discussing the party presentation rule
Summary of this case from Maltez v. N.J. Transit Rail OperationsOpinion
Civil Action 20-4457 (MEF) (JRA)
06-25-2024
Zequi Wu, et al., Plaintiffs, v. GSX Techedu Inc., et al., Defendants.
OPINION AND ORDER
Michael E. Farbiarz, U.S.D.J.
Table of Contents
I. Background
A. Allegations
B. The Lawsuit
C. Procedural History
D. The Motion
E. The Court's Approach
II. Legal Standards
A. Pleading Standards
B. Substantive Standards
1. Section 10(b)
2. Section 20(a)
III. Misrepresentations
A. Student Enrollment
1. Witnesses
a) CW-3
i. Detail
ii. Sources
iii. Corroboration
iv. Conclusion and Other Allegations
b) CW-2
c) CW-1
d) CW-5
e) CW-6
f) Conclusion
2. Additional Information
3. Conclusion
B. Revenue
C. Spring 2020 Statements
IV. Scienter
A. General Legal Standards
B. Pleading Scienter: Three Questions
C. Analysis of the Three Questions
1. Red Flags
a) In General
b) In This Case
2. Position in the Company
3. Importance to the Company
D. Conclusion
V. Loss Causation
A. In General
B. In This Case
VI. Conclusion
Investors bought stock in an online education company, and its stock price fell.
The investors then sued the company and some of its senior executives. They allege that certain public statements about the company were inaccurate, mainly because the company was falsely inflating both student enrollment and revenue numbers -- - and that when the truth about these false statements was eventually revealed, it caused the stock price to fall.
The company and its executives now move to dismiss. The motion is denied in large part and granted to a limited extent.
I. Background
A. Allegations
The relevant allegations for now are as follows.
GSX Techedu Inc. (“the Company”) is an internet-based educational platform; it sells online classes to students. See Second Amended Complaint (“Complaint”) ¶ 2.
The Company, it is alleged, made it look like it had many more students than it really did.
How?
Mainly by paying for large numbers of “bots” to pose as bona-fide students, and to attend Company classes. See id. at ¶¶ 3, 5. (A “bot” is a software application that runs automated tasks. See id.) The bots' attendance was publicly reported by the Company, as if the bots were real students attending Company classes. See id.
What was the point of doing this?
The Company, it is alleged, secretly used its own money to pay the phony students' enrollment fees --- and then publicly reported those fees as revenue. See id. at ¶¶ 5, 32, 72.
This inflated revenue stream made the Company more enticing to potential investors. See id. at ¶¶ 5, 8.
After all, the Company seemed to be attracting more students (but these were really bots) who were paying enrollment fees (but this was really Company money). See id. at ¶¶ 32, 72, 130, 136-39.
As part of this alleged scheme, the Company made a range of false public statements. Revenue, for example, was overstated, and so were student enrollment numbers. See id. at ¶¶ 195-280.
B. The Lawsuit
Pressing the allegations set out above, some investors sued. The investors (from here, “the Plaintiffs”) sued on behalf of a putative class of people who owned shares in the Company during the relevant period. See id. at ¶¶ 1, 13-16.
The lead Plaintiff is Yang Renbin. The remaining named Plaintiffs are Robert Angeline, Zequi Wu, Corey Hays, and Alexandre Tazi. All of them, as noted, were Company investors.
The Plaintiffs sued the Company, plus two of its leaders --- the Chief Executive Officer (“CEO”) and the Chief Financial Officer (“CFO”). See id. at ¶¶ 19-20. (From here, the Company, the CEO, and the CFO are collectively “the Defendants.”).
The Defendants are: GSX Techedu Inc. (the Company); Larry Xiangdong Chen (the CEO); and “Nan” Shen (the CFO).
The Plaintiffs sued under the Exchange Act of 1934. See id. at ¶ 1. Their claims are in two counts.
The first count invokes Section 10(b) of the Act, and Rule 10b-5 promulgated by the Securities and Exchange Act Commission. See id. at ¶ 367.
“Rule 10b-5 implements section 10(b) of the Securities Exchange Act.” Riley v. Simmons, 45 F.3d 764, 773 n.9 (3d Cir. 1995); see also Globis Cap. Partners, L.P. v. Stonepath Grp., Inc., 241 Fed.Appx. 832, 835 (3d Cir. 2007) (the Act's “[S]ection 10(b) . . . is enforced through Rule 10b-5”).
The second count invokes 20(a) of the Act. See id. at ¶ 372.
C. Procedural History
The Plaintiffs' First Amended Complaint was dismissed for failure to state a claim. See Wu v. GSX Techedu Inc., 2023 WL 2207422 (D.N.J. Feb. 24, 2023).
This Court, per Judge Salas, issued a thorough opinion. It held that certain claims failed to adequately allege a misrepresentation, see id. at *5-11, and that scienter was not adequately alleged as to other claims. See id. at *11-15.
About 4 months after Judge Salas issued her decision, this case was re-assigned to the undersigned.
In light of Judge Salas's opinion, the Plaintiffs filed a new Complaint. See Second Amended Complaint (April 25, 2023). The new Complaint (referred to here as “the Complaint”) added some allegations, and dropped some others. It also shifted how it made use of certain allegations, and supplemented the allegations as to scienter in light of issues that has been identified by Judge Salas.
D. The Motion
The Defendants have moved to dismiss the Complaint under Rule 12(b)(6) of the Federal Rules of Civil Procedure.
The motion is now before the Court.
E. The Court's Approach
To analyze the motion, the Court begins by briefly laying out the relevant legal standards. See Part II.
The Court then considers the three reasons that, per the Defendants, the Plaintiffs' claims must be dismissed.
The Defendants argue that the Plaintiffs have not adequately alleged (1) a misrepresentation, (2) scienter, or (3) loss causation.
But the Court concludes that these arguments are by and large not persuasive --- in Part III (misrepresentation), in Part IV (scienter), and in Part V (loss causation).
In light of this conclusion, the motion to dismiss is largely denied. See Part VI.
II. Legal Standards
Start with the legal standards that a Complaint must meet to survive a motion to dismiss.
These fall into two categories. There are pleading standards, see Part II.A, which indicate what a plaintiff must do to hit its marks. And these are substantive standards, see Part II.B, which indicate what the marks are.
A. Pleading Standards
In deciding a motion to dismiss under Federal Rule of Civil Procedure 12(b)(6), as this one is, the Court is “required to accept as true all factual allegations in the complaint and draw all inferences from the facts alleged in the light most favorable to [the plaintiff].” Phillips v. County of Allegheny, 515 F.3d 224, 228 (3d Cir. 2008).
Motions to dismiss are assessed as follows.
First, the Court “must tak[e] note of the elements [a] plaintiff must plead to state a claim.” Connelly v. Lane Constr. Corp., 809 F.3d 780, 787 (3d Cir. 2016) (quoting Ashcroft v. Iqbal, 556 U.S. 662, 675 (2009)).
The elements are “note[d]” just below, in Part II.B.
Second, the Court must identify those allegations in the complaint that are merely conclusory, and set them to one side as irrelevant to the analysis. See id.
And third, the Court must determine whether the remaining allegations “plausibly give rise to an entitlement to relief.” Connelly, 809 F.3d at 787 (quoting Iqbal, 556 U.S. at 679).
These standards apply across the board, to all Rule 12(b)(6) motions.
But the bar is higher here.
That is because two added pleading strictures apply. First, Federal Rule of Civil Procedure 9(b), because this case sounds in fraud. And second, the Private Securities Litigation Reform Act of 1995 (“the PSLRA”).
[Plaintiffs who] allege fraud . . . “must state with particularity the circumstances constituting fraud or mistake.” Fed.R.Civ.P. 9(b). Moreover, . . . the PSLRA imposes greater particularity requirements concerning alleged material misrepresentations and scienter. A complaint must “specify each statement alleged to have been misleading, the reason or reasons why the statement is misleading, and, if an allegation . . . is made on information and belief . . . all facts on which that belief is formed.” 15 U.S.C. § 78u-4 (b)(1). Concerning scienter, a complaint must “state with particularity facts giving rise to a strong inference that the defendant acted with the required state of mind.” 15 U.S.C. § 78u-4 (b)(2)(A). The PSLRA's heightened standard exists “to curb frivolous, lawyer-driven litigation, while preserving investors' ability to recover on meritorious claims.”
Tellabs, Inc. v. Makor Issues & Rts, Ltd., 551 U.S. 308, 322 (2007).Fan v. StoneMor Partners LP, 927 F.3d 710, 714-15 (3d Cir. 2019); see also In re Rockefeller Ctr. Props., Inc. Sec. Litig., 311 F.3d 198, 224 (3d Cir. 2002) (holding that Rule 9(b) and the PSLRA “impose independent, threshold pleading requirements that, if not met, support dismissal apart from Rule 12(b)(6)”); see also Hacker v. Elec. Last Mile Sols. Inc., 687 F.Supp.3d 582, 588 (D.N.J. 2023).
B. Substantive Standards
As noted, see Part I.B, the Plaintiffs press two counts under the Exchange Act of 1934.
1. Section 10(b)
Count One is brought under Section 10(b) of the Exchange Act. See Complaint ¶ 367.
A Section 10(b) claim “has six elements: (i) a misrepresentation or omission of material fact; (ii) scienter; (iii) a connection with the purchase or sale of a security; (iv) reliance; (v) economic loss; and (vi) loss causation.” City of Warren Police & Fire Ret. Sys. v. Prudential Fin., Inc., 70 F.4th 668, 679 (3d Cir. 2023); see also Hacker, 687 F.Supp.3d at 587.
2. Section 20(a)
Count Two is pressed under Section 20(a) of the Exchange Act. See Complaint ¶ 372.
“Section 20(a) of the Exchange Act imposes joint and several liability upon one who controls a violator of Section 10(b).” In re Suprema Specialties, Inc. Sec. Litig., 438 F.3d 256, 284 (3d Cir. 2006); see also Janus Cap. Grp., Inc. v. First Derivative Traders, 564 U.S. 135, 146 (2011).
“[T]he plaintiff must prove that one person controlled another person or entity and that the controlled person or entity committed a primary violation of the securities laws.” Suprema, 438 F.3d at 284; see also Rahman v. Kid Brands, Inc., 736 F.3d 237, 247 (3d Cir. 2013).
III. Misrepresentations
The Defendants' first argument: the Complaint does not properly plead a misrepresentation or omission of material fact, and this dooms both Count One and Count Two. See Motion to Dismiss at 21.
In particular, the argument goes, three main sets of misrepresentations are not properly pled.
The Court considers this argument here, with each of the three sets of misrepresentations assessed below.
A. Student Enrollment
Start with the alleged false statements as to the Company's student enrollment data.
The Complaint points to a variety of student enrollment statements made in 2019 and 2020 (“Enrollment Statements”).
From 2019, the Enrollment Statements were made: in initial public offering materials (see Complaint ¶¶ 195-99); in Securities and Exchange Commission filings (see id. at ¶¶ 20006, 212-17, 222-25); and orally by the CEO and the CFO during earnings calls (see id. at ¶¶ 206-11, 218-21).
The Enrollment Statements from 2020 are, in essence, more of the same. They were made in Securities and Exchange Commission filings (see Id. at ¶¶ 234-37, 247-52, 263-68, 275-80); Company press releases (see id. at ¶¶ 261-62, 269-72); oral statements from the CEO and the CFO during earnings calls (see id. at 23846); and oral statements by the CEO and CFO in other public contexts (see id. at ¶¶ 253-60).
The Enrollment Statements each provide a fairly specific recounting of the number of students enrolled in Company classes, including statements as to enrollment growth. See, e.g., id. at ¶ 224 (“GSXrepresented that total enrollment in its courses increased from 79,632 students in 2017 to 767,102 students in 2018”) (cleaned up); id. at ¶ 234 (“GSX claimed that total enrolments increased 290.2% year-over-year to 1,120,000 students”); id. at ¶ 239 (“[t]otal enrollments . . . hit a record high of 1.12 million, which was 3.9x that of the same period of 2018”).
“GSX” is shorthand for GSX Techedu Inc. Recall that it is referred to in this Opinion and Order as “the Company.” See footnote 2.
Per the Plaintiffs, these Enrollment Statements were false, and materially so. See, e.g., ¶¶ 225, 235, 240.
False because “student” enrollment means enrollment by human students --- but some of the Company's “students” were in fact botsposing as people. See id.
Recall that a “bot” is a software application that runs automated tasks. See id. at ¶ 3.
And materially false --- because the Company's student enrollment numbers were not off by a little here and there, but by a large amount. See id.
In response, the Defendants argue that there are insufficient allegations that a substantial percentage of the Company's enrollment numbers were based on bot enrollments. See Motion to Dismiss at 28, 36.
But this argument is not persuasive.
It runs aground on extensive and detailed allegations from confidential witnesses who worked for or with the Company, see Part III.A.1, and also on a third-party report, see Part III.A.2.
1. Witnesses
The Plaintiffs put forward various allegations as to false student enrollment numbers based on information sourced to unnamed confidential witnesses. See Complaint ¶¶ 63-95.
These allegations are taken up below, confidential witness by confidential witness.
As part of doing so, the Court determines whether, as to each witness, there is a need to “discount” the weight given to that witnesses' allegations, or to disregard the allegations entirely. See generally Instit. Invs. Grp. v. Avaya, Inc., 564 F.3d 242, 263 (3d Cir. 2009) (when evaluating the allegations of a confidential witnesses, courts must consider certain “criteria,” and “[i]f anonymous source allegations are found wanting with respect to these,” then the allegations “must [be] discount[ed]” to an appropriate extent, or disregarded).
a) CW-3
Begin with the allegations from Confidential Witness-3, or “CW-3.” See Complaint ¶¶ 75-78.
The main bulk of the Complaint's CW-3's allegations:
CW-3 was employed by a third-party brushing firm in Beijing from January 1, 2019 through December 31, 2019. In this role, she was regularly paid commissions by GSX to enroll in GSX's courses with fake or assumed identities and to write positive reviews. CW-3 explains that “GSX management” had a contract with this third-party brusher, pursuant to which CW-3 and her colleagues would assume the identities of “and conduct activities as if [they] were real students.” CW-3 would “buy classes, write positive reviews, and join in the large classes to boost up [GSX]'s head counts” with harvested user data from various social media platforms, including WeChat . . . GSX paid the third-party brusher 50 RMB per enrollment.
GSX planned its third-party brushing months in advance. Each quarter, GSX prepared and sent a “detailed brushing plan” to the third-party brusher. These plans specified the dates, times, course titles, tuition rates, and instructors for whom the third party was to focus its brushing efforts. CW-3 and her colleagues referenced and used these plans daily to ascertain the GSX courses in which to enroll and then attend, and the GSX instructors about whom to write positive reviews.Id. at ¶¶ 76-77.
“Brushing” means different things in different contexts. It is not defined in the Complaint. One common understanding of brushing: the use of deceptive methods, typically online, to burnish (“brush”) a company's reputation (for example, by posting false product reviews) or sales numbers (for example, by sending people packages they did not order).
This is the Company. See footnotes 2, 6.
As noted above, see Part III.A.1, information from confidential witnesses sometimes needs to be “discount[ed].” Avaya, 564 F.3d at 263.
Here, the Court concludes that CW-3's allegations can be taken as allegations generally are at this stage --- as true, see, e.g., Wood v. Moss, 572 U.S. 744, 755 n.5 (2014), and at face value.
Subject to a small exception, see Part III.A.1.iv, discounting is not necessary because CW-3's allegations are detailed, see Part III.A.1.a.i; rest on solid-enough sources of information, see Part III.A.1.a.ii; and are corroborated in meaningful part, see Part III.A.1.a.iii.
i. Detail
Take, first, the level of “detail[]” of the above-quoted CW-3 allegations. See generally Avaya, 564 F.3d at 263 (courts must “consider the detail provided by the confidential sources” in determining whether to “discount[]” or disregard their allegations) (cleaned up); accord, e.g., Rahman, 736 F.3d at 244; Cal. Pub. Emps.' Ret. Sys. v. Chubb Corp., 394 F.3d 126, 147 (3d Cir. 2004).
The above-quoted allegations are not sparse, vague, or generic. They are highly detailed.
CW-3's allegations are about as granular as allegations of confidential witnesses that the Third Circuit has, in other cases, held to be detailed-enough.
See Avaya, 564 F.3d at 249-50, 260-264 (holding that confidential witness allegations were detailed enough when: they reflect information about the relevant company's inability to compete with specific competitors, specific price reductions the company was forced to accept, and the specific clients who received discounts); Prudential Fin., Inc., 70 F.4th at 692 (holding that confidential witness allegations were detailed enough when: the employee alleged that the relevant company was aware “that [the life insurance segment of its business] as a whole was ‘performing poorly' and that the company attributed that poor performance to ‘negative mortality experience in the legacy [life insurance policies the company acquired]'” and “those developments caused [the company] to discuss as early as May 2019” that it would result in large costs); cf. Chubb Corp., 394 F.3d at 150 n.14 (seeming to embrace as “sufficient” the level of detail in In re Cabletron, 311 F.3d 11, 31 (1st Cir. 2002), in which confidential witnesses provided information as to “tractor-trailers in the factory yard, equipment ‘borrowed' from employees' desks to be fraudulently processed, and the unusual activity in the warehouse as products were shuttled back and forth,” plus “the name and address of one employee who is said to have stashed goods in the garage of his home,” and indications “that [a specific person] told him to do so”).
And on the flip side, CW-3's allegations are markedly more detailed than confidential witness allegations that have been held wanting for insufficient specificity by the Third Circuit.
See Rahman, 736 F.3d at 245, 245 n.12 (holding that confidential witness allegations were not detailed enough when the allegations were that “things weren't on the up and up,” including a “strange feeling things weren't right,” and “[the subsidiary] had been ‘singularly uncooperative in complying with Sarbanes Oxley requirements'”) (cleaned up); OFI Asset Mgmt. v. Cooper Tire & Rubber, 834 F.3d 481, 495-97 (3d Cir. 2016) (holding that that confidential witness allegations were not detailed enough when the allegations were that the relevant company “had ‘apparently very little control over [a joint venture],' that ‘[a different company] ‘pretty closely controlled [the joint venture],' and that [the joint venture's] independent financial system resulted in [the company] being ‘closed off' from [the joint venture's] financial information”).
In short: CW-3's allegations check the box --- they are sufficiently “detail[ed],” such that discounting is not necessary. Avaya, 564 F.3d at 263.
This conclusion makes sense in light of one of the key underlying reasons why confidential witness “detail[]” matters in the first place.
To see the point, step back for a moment.
In securities fraud cases, plaintiffs must often meet both the pleading standards set in place by the United States Congress (in the PSLRA) and the Federal Rules of Civil Procedure (in Rule 9(b)). These are demanding, and require a good deal of “particularity.” See Tellabs, Inc., 551 U.S. at 313; In Re Exxon Mobil Corp. Sec. Litig., 500 F.3d 189, 191 n.2 (3d Cir. 2007).
But when it comes to particularity, confidential witness allegations start off at a deficit.
Anonymity means that a basic piece of information --- whose allegations are these? --- is always missing.
And that missing dollop of particularity can be a practical concern.
Look to two examples.
First, there is a possible danger that anonymous accusations, from a person or entity does not stand behind by name, can sometimes be less reliable. In turn, less reliable information can spin off more in the way of meritless lawsuits --- and it was in part to cut down on such lawsuits that the particularity pleading standards were put in place. See, e.g., Rowinski v. Salomon Smith Barney Inc., 398 F.3d 294, 298 (3d Cir. 2005) (the particularly requirement “include[es] more stringent pleading requirements to curtail the filing of meritless lawsuits”) (cleaned up); Haliburton Co. v. Erica P. John Fund, Inc., 573 U.S. 258, 277 (2014); Winer Family Tr. v. Queen, 503 F.3d 319, 335 (3d Cir. 2007); In re Rockefeller Ctr. Props., Inc., 311 F.3d at 216 (particularity standards are designed, in part, to “reduce[] the number of frivolous suits brought solely to extract settlements”) (cleaned up).
This is, for example, a major theme of Fourth Amendment law. See, e.g., Florida v. J.L., 529 U.S. 266, 270 (2000) (“[u]nlike a tip from a known informant whose reputation can be assessed and who can be held responsible if her allegations turn out to be fabricated, . . . an anonymous tip” requires additional indicia of reliability) (cleaned up); Alabama v. White, 496 U.S. 325, 329 (1990) (“the veracity of persons supplying anonymous tips is by hypothesis largely unknown, and unknowable” and courts must therefore look for additional indicia of reliability when relying on anonymous sources) (cleaned up).
A second example: without knowing who is saying something, it can be that much harder to put together a well-investigated defense --- and by doing so to differentiate on an information rich basis, even before discovery, between meritless lawsuits and substantial one. But facilitating such analysis --- “giv]ing] defendants notice of the claims against them,” In re Rockefeller Ctr. Props., Inc., 311 F.3d at 216 (cleaned up) --- is a basic goal of heightened pleasing standards.
Bottom line: confidential witnesses start out a step behind as to particularity --- and that can have practical consequences (including the two set out just above) that are at odds with why the particularity requirements were established in the first place.
To close the gap, the federal courts have required additional information from confidential witnesses.
Two points. First, the Third Circuit has drawn a clear through line from (a) the heightened requirements of Rule 9 and the PSLRA to (b) the need for additional information when it comes to confidential witnesses. See e.g., Avaya, 564 F.3d at 261 (“a complaint can meet the [particularity] pleading requirement [of the PSLRA] [when relying on confidential sources] by providing sufficient documentary evidence and/or a sufficient description”); City of Edinburgh Council v. Pfizer, Inc., 754 F.3d 159, 168 n.11 (3d Cir. 2014) (“[w]e apply the PSLRA's heightened pleading requirements to confidential witness allegations by evaluating the detail provided by the confidential sources, the sources' basis of knowledge, the reliability of the sources, the corroborative nature of other facts alleged, including from other sources, the coherence and plausibility of the allegations, and similar indicia”) (cleaned up); Prudential Fin., Inc., 70 F.4th at 680, 691 (similar); Rahman, 736 F.3d at 243-44 (similar). And second, the Third Circuit has also tied together the relevant bodies of law, sending their development down similar courses. Rule 9(b) and the PSLRA require more detail. See, e.g., Suprema, 438 F.3d 256; City of Cambridge Ret. Sys. v. Altisource Asset Mgmt. Corp., 908 F.3d 872, 879 (3d Cir. 2018). And the question of sufficient detail is also a critical one for assessing confidential witnesses. See, e.g., Avaya, 564 F.3d at 261 (“[w]e consider the detail provided by the confidential sources”) (cleaned up); City of Edinburgh Council, 754 F.3d at 168 n.11 (same). In a similar vein, Rule 9(b) and the PSLRA often require plaintiffs to lay some of their cards on the table, as to how they know the things they say they do. See 15 U.S.C. § 78u-4(b)(1) (under the PSLRA, “if an allegation regarding the statement or omission is made on information and belief, the complaint [must] state with particularity all facts on which that belief is formed”); Merrill Lynch, Peirce, Fenner & Smith Inc. v. Dabit, 547 U.S. 71, 81-82 (2006). And when it comes to confidential witnesses --- there is also a special focus on the witnesses' sources of information. See Avaya, 564 F.3d at 261 (for confidential witnesses courts consider “the sources' basis of knowledge”); Pfizer, Inc., 754 F.3d at 168 n.11 (similar); Prudential Fin., Inc., 70 F.4th at 680, 691 (similar); Rahman, 736 F.3d at 243-44 (similar).
But if one of the animating concerns of the law in this area is that confidential witnesses may not, without more, measure up to the applicable PSLRA/Rule 9(b) particularity standards --- then confidential witness allegations should plainly be sufficient when they do meet PSLRA/Rule 9(b) standards.
Here, that is the case.
Under the PSLRA and Rule 9(b), allegations must spell out the “who, what, when, where, and how of the events at issue.” United States ex rel. Bookwalter v. UPMC, 946 F.3d 162, 176 (3d Cir. 2019); accord, e.g., In re Rockefeller Ctr. Props., Inc. Sec. Litig., 311 F.3d at 217.
And so it stands to reason, and the Court holds, that when a confidential witness meets the who-what-when-where-and-how test generally required by the PSLRA and 9(b) --- then there is no basis to “discount[]” that witness's information on the ground that is not detailed enough.
The Court offers no views on the converse situation --whether a confidential witness provides a “detail[ed]” enough testimony to not be discounted, see Avaya, 564 F.3d at 261, if the witness does not meet the who-what-when-where-and-how test.
As to CW-3, that test is easily satisfied.
The Complaint sets out allegations as to what was done by CW-3 and her colleagues (pretending to be “real students” and buying up spots in Company classes, see Complaint ¶ 76); how it was done (using “harvested” data, see id., and operating pursuant to a “detailed” quarterly plan, see id. at ¶ 77); why it was done (for commission payments, see id. at ¶ 76); and where it was done (from Beijing, see id.) and when it was done (for all of 2019, see id.).
In a nutshell: there is more than enough detail here, and CW-3's allegations do not need to be discounted for being too spare. This is clear from a look to the confidential witness detail mustered in analogous cases. And it is clear based on consideration of the underlying concerns that animate the law in this area.
ii. Sources
Turn now to the second step in the assessment of CW-3. This is the “basis of [her] knowledge.” Avaya, 564 F.3d at 263; accord, e.g., Rahman, 736 F.3d at 244; Chubb Corp., 394 F.3d at 147.
As to what was quoted above, see Part III.A.1.a, CW-3's allegations grow directly out of what she personally did. See Complaint ¶¶ 76-77. Her allegations describe her work, see id., and the directions she received as part of it. See id. at ¶ 77.
Indeed, if called to testify at a trial, the large bulk of CW-3's allegations would seem to be admissible --- they appear to have a “foundation” in things she saw and did. See generally Fed. R. Evid. 602 (a “witness may not testify to a matter unless evidence is introduced sufficient to support a finding that the witness has personal knowledge of the matter”); see also Brooks v. Am. Centennial Ins. Co., 327 F.3d 260, 268 (3d Cir. 2003) (discussing the need for an evidentiary foundation for a witness's testimony).
In short, CW-3's allegations rest on sturdy-enough ground --direct knowledge.
They do so to roughly the same extent as confidential witness allegations that have been determined to be well-enough “source[d].” Avaya, 564 F.3d at 263.
See Avaya, 564 F.3d at 266 (holding that confidential witness allegations were sufficiently sourced when the plaintiffs adequately alleged “how or why their confidential sources would have the knowledge alleged,” by providing their specific positions and job responsibilities at the company in question, when they worked for the company and obtained the relevant information, and that their knowledge was firsthand) (cleaned up); Rahman, 736 F.3d at 241, 245 (holding that confidential witness allegations were sufficiently sourced when the witness worked as a Sarbanes Oxley compliance consultant for the company, and was therefore in a position to know that a subsidiary was uncooperative in complying with Sarbanes Oxley); Prudential Fin., Inc., 70 F.4th at 691-92 (holding that confidential witness allegations were sufficiently sourced when the source “regularly attended[ed] . . . forecast meetings with the actuarial, capital, and financial teams” for the life insurance segment of the business during the relevant time period, and that supported the allegation that “it was discussed in forecast meetings” that the life insurance segment of the business was “performing poorly” due to “negative mortality experience[s]”).
And on the other hand, CW-3's allegations are rooted in personal knowledge and experience to a much greater extent than confidential witness allegations that have been found not to measure up in other cases.
See Rahman, 736 F.3d at 245 (holding that confidential witness allegations lacked an adequate basis when witnesses did not attend the meetings they provided information about and did not explain how they knew managers took certain steps); Chubb Corp., 394 F.3d at 148-49 (holding that confidential witness allegations lacked sufficient sources when brief job descriptions, such as “senior customer service team leader,” did not show how the confidential witnesses could “possess information that the standard commercial business was succeeding or failing on a national level”); see also Martin v. GNC Holdings, Inc., 757 Fed.Appx. 151, 154 (3d Cir. 2018).
The long and short of it: CW-3's above-quoted allegations easily clear the “sources of information” hurdle, and do not need to be discounted.
iii. Corroboration
Third and finally, courts are required to kick the tires to an extent --- to assess the confidential witness allegations from an external perspective (are they corroborated by other information?) and from an internal perspective (do the allegations hang together in a common sense way?). See Avaya, 564 F.3d at 261 (holding that courts should consider “the corroborative nature of other facts alleged, including from other sources, [and] the coherence and plausibility of the allegations”).
This corroboration-and-coherence test is met here.
CW-3's allegations dovetail closely with allegations from other confidential witnesses here. See, e.g., Part III.A.1.b., Part III.B. That provides ample corroboration. See generally Avaya, 564 F.3d at 261 (discussing the ways in which confidential witness statements corroborate each other); Prudential Fin., Inc., 70 F.4th at 692 (same); Glazer Cap. Mgmt., L.P. v. Forescout Techs., Inc., 63 F.4th 747, 772-3 (9th Cir. 2023) (same).
And from the perspective of “experience and common sense,” Iqbal, 556 U.S. at 664, there is nothing about the above-quoted CW-3 allegations that is so incoherent or hard-to-believe that the allegations must be discounted.
iv. Conclusion and Other Allegations
The allegations from CW-3 are detailed, see Part III.A.1.a.i, well-enough sourced, see Part III.A.1.a.ii, and corroborated, see Part III.A.1.a.iii.
They do not need to be discounted. They can be taken as is.
In addition to those already discussed, the Complaint sets out some more CW-3 allegations:
According to CW-3, in 2019, the third-party brusher's employees, including CW-3, accounted for approximately 40% of all enrollments in GSX's K-12 courses, and wrote approximately 35% of the positive reviews for GSX's K-12 instructors....As part of the scheme, when the Company sent its brushing plans to the third-party each quarter, GSX concurrently transferred the full amount of the tuition fees, together with commissions, for the agreed-upon fake enrollments.Complaint ¶¶ 76, 78.
As to “brusher[],” see footnote 8.
Recall that GSX is the Company. See footnotes 2, 6.
To determine if these need to be discounted, move through the three-part test sketched out above.
First, these allegations are relatively “detailed.” Avaya, 564 F.3d at 263.
Second, these allegations are coherent, and link up to other information in the Complaint. See, e.g., Part III.A.1.a, III.A.2; see generally Avaya, 564 F.3d at 261 (discussing the need for both corroboration and coherence).
As to the third piece of the analysis: are CW-3's allegations --- about “bot” percentages, and the Company's quarterly payments --- well enough sourced?
The best inference is: yes. Those allegations seem to flow from what CW-3 knows directly, through her work.
As an initial matter, employees are often given at least some sense of how their work fits into the bigger picture. To prioritize tasks, for example, it is useful for an employee to know whether they are servicing a client who is relying on them heavily (“40% of all enrollments”) or to a lesser extent.
In addition, CW-3 and her colleagues were allegedly using bots to, among other things, mimic the behavior of real students. It would be surprising if such employees did not have at least a rough sense of which students were real (60%) and which were bots (the 40% that were faked).
And all the more so because bots allegedly engaged in discernible “bot-like” behavior, including things like making identical comments on classes. See Part III.A.1.d. It stands to reason that people professionally deploying bots can readily tell the difference between those who are bots and those that are not. And once that distinction is top of mind --- it is straightforward enough for CW-3 to come to a rough conclusion as to how many of the students were, in fact, bots deployed by her company.
All of this adds up to an inference of solid-enough sourcing for CW-3's allegations.
But an alternative inference is possible. CW-3 may have learned of the 40% number in a second-hand way --- for example, as a piece of workplace gossip. Cf. Chubb Corp., 394 F.3d at 148 (the Court is “left to speculate whether the anonymous sources obtained the information they purport to possess by firsthand knowledge or rumor”).
Given this possible inference, the value of the CW-3 allegations set out in this section must be trimmed back to an extent.
The Court discounts them a bit, and affords them less weight on the scale than the other CW-3 allegations.
b) CW-2
Zero in now on the next confidential witness, “CW-2.”
CW-2 allegedly worked at the Company, see Complaint ¶ 67, and the key allegations associated with her are as follows:
• The Company began using bots to bump up enrollment in 2015. See Complaint ¶ 68.
• To what extent? CW-2 provided an example: in classes of five students, 500 bot-generated users would be enrolled in the class as “students,” see id. at ¶ 68, and these bots then bought new Company class enrollments, signed in and out of classes, and sent messages to other students. See id.
• Accomplishing this apparently required a good deal of computing power, and more than half of the Company's Beijing headquarters were occupied with servers, each of which controlled around 1,000 bots. See id. at ¶ 70.
These allegations do not need to be discounted.
This is subject to the third point in footnote 20.
First, they are plainly “detailed,” well beyond the baseline established by the leading cases. See Avaya, 564 F.3d at 249- 50, 260-264; Rahman, 736 F.3d at 245, 245 n.12; Chubb Corp., 394 F.3d at 150 n.14; Cooper Tire & Rubber, 834 F.3d at 495-97; Prudential Fin., Inc., 70 F.4th at 692; see generally Part III.A.1.a.i (discussing the law in this area).
Second, these allegations appear to be rooted in solid-enough sources. CW-2's knowledge allegedly flows directly from her work. She is said to have been a Company engineering manager who supervised Company software engineers. See Complaint ¶ 67. And she worked out of the Company's Beijing headquarters, see id., where she could presumably observe the size of the server farms alluded to in the Complaint.
This degree of “sourc[ing]” has been held to pass muster. See Avaya, 564 F.3d at 266; Rahman, 736 F.3d at 245; Chubb Corp., 394 F.3d at 148-49; Prudential Fin., Inc., 70 F.4th at 691-92; GNC Holdings, Inc., 757 Fed.Appx. at 154; see generally Part III.A.1.a.ii (discussing the law in this area).
And third and finally: CW-2's allegations are amply corroborated. They are buttressed by other allegations in this case. See Complaint ¶¶ 63-65 (describing rooms full of bot-phones in the Beijing headquarters), id. at ¶¶ 60-61 (report concluding 70% of users were bots). This is more than enough. See Avaya, 564 F.3d at 261; Prudential Fin., Inc., 70 F.4th at 692; Forescout Techs., Inc., 63 F.4th at 772-3; see generally Part III.A.1.a.iii (discussing the law in this area).
Three notes. First, CW-2 also alleges that the Company paid a third-party entity to “brush” enrollment numbers. See Complaint ¶ 72. This allegation exerts somewhat less of a pull than the allegations cited in the text. It is strongly corroborated --- for example, by CW-3's testimony. See id. at ¶ 76; see generally Avaya, 564 F.3d at 261; Prudential Fin., Inc., 70 F.4th at 692; Forescout Techs., Inc., 63 F.4th at 772-73. But the source of this information is not crystal clear. See generally Avaya, 564 F.3d at 266; Rahman, 736 F.3d at 245; Chubb Corp., 394 F.3d at 148-49; Prudential Fin., Inc., 70 F.4th at 691-92; GNC Holdings, Inc., 757 Fed.Appx. at 154. Second, there are also some CW-2 allegations as to which Judge Salas held that the basis of CW-2's knowledge is, as pled, too little to go on. See Wu, 2023 WL 2207422, at *7. The Court does not consider those allegations. Third, Judge Salas previously determined that CW-2's allegations should be discounted to an extent because she was not interviewed by the Plaintiffs, see Complaint ¶ 348, and because the Complaint does not say when she stopped working at the Company. See Wu, 2023 WL 2207422, at *7. The Court hews to Judge Salas's holdings, see Musacchio v. United States, 577 U.S. 237, 244-45 (2016) (“The law-of-the-case doctrine generally provides that when a court decides upon a rule of law, that decision should continue to govern the same issues in subsequent stages in the same case.”) (cleaned up), and applies the discount she identified.
c) CW-1
Move on now to the next confidential witness, “CW-1,” who allegedly worked for the Company. See Complaint ¶ 63.
The key CW-1 allegations:
• CW-1 harvested data from WeChat accounts to set up fake accounts for “students,” see id. at ¶ 65, and reprogramed electronic devices, mainly with the social media data. See id.
• These devices then functioned as bots --- and enrolled in Company courses, logged into them, and wrote online reviews of the classes. See id.
• CW-1 was paid directly by the Company for her work, sketched out above. See id. at ¶ 64.
• The Beijing headquarters of the Company contained “mobile phone rooms” that contained “many thousands” of phones packed on shelves. Id. (The inference: these were used to enroll fake students at scale.).
“WeChat is a Chinese multi-purpose messaging, social media and mobile payment app.” Complaint ¶ 26.
These allegations do not need to be discounted.
First, they are “detailed” enough. See Avaya, 564 F.3d at 24950, 260-264; Rahman, 736 F.3d at 245, 245 n.12; Chubb Corp., 394 F.3d at 150 n.14; Cooper Tire & Rubber, 834 F.3d at 495-97; Prudential Fin., Inc., 70 F.4th at 692; see generally Part III.A.1.a.i (discussing the law in this area).
Second, CW-1's “sources” are that she allegedly was an engineer who worked out of the Company's Beijing headquarters, where she personally programmed bot-phones. See Complaint ¶ 63. This is more than enough to go on. See Avaya, 564 F.3d at 266; Rahman, 736 F.3d at 245; Chubb Corp., 394 F.3d at 148-49; Prudential Fin., Inc., 70 F.4th at 691-92; GNC Holdings, Inc., 757 Fed.Appx. at 154; see generally Part III.A.1.a.ii (discussing the law in this area).
And third, CW-1's allegations are strongly corroborated --- by other, similar strands of allegations that they are tightly braided together with. Compare, e.g., Complaint ¶ 64 (CW-1: describing rooms full of many thousands of phones in the Beijing headquarters) with id. at ¶¶ 69-70 (CW-2: describing rooms full of servers in the Beijing headquarters); compare, also, e.g., id. at ¶ 65 (CW-1: implying bot-phones enrolled in classes) with id. at ¶ 69 (CW-2: alleging that bot-controlled devices assigned a mobile phone number enrolled in classes).
This is more than enough corroboration. See Avaya, 564 F.3d at 261; Prudential Fin., Inc., 70 F.4th at 692; Forescout Techs., Inc., 63 F.4th at 772-3; see generally Part III.A.1.a.iii (discussing the law in this area).
d) CW-5
Take the next confidential witness, “CW-5,” who allegedly enrolled in one of the Company's online classes in September 2019. See Complaint ¶ 80. Upon enrollment, she was given access to two group text accounts --- one purportedly for students enrolled in the class, and one for their parents. See id. The class was due to meet for 4 hours, but it stopped short after one hour. See id. at ¶ 81. The allegations:
On the first day of class, the instructor taught for about one hour, after which “there was nothing” in the classroom, so CW-5 sent a message to the student and parent WeChat groups, inquiring whether any others were suddenly unable to view the course content. In the parent group, two users, each purporting to be parents of enrolled students, responded within seconds of each other with identical messages, explaining that it was the “Self-Study Period.”
Approximately ten minutes after the class period ended, at around 12:10 p.m., individuals purporting to be students and parents, each “with different WeChat IDs sent out as many as five groups of almost identical praise” for the instructor in both the student group and in the parent group.
For example, two “different” individuals in both groups said: “Teacher Song really hit the nail on the head, love it,” using identical Chinese characters in the same order.
In another example, two “different” individuals in both groups said: “I always wanted to learn from root words like this, Teacher Song did very well. Perfect, looking forward to tomorrow's class,” using identical Chinese characters in the same order:
In another example, two “different” individuals in both groups said: “I enjoy this methodology for memorizing words. Doing more with less, I like it,” using the same Chinese characters in the same order
In another example, two “different individuals in both groups said: “The teacher speaks very well, very helpful for children. Great!,” using the same Chinese characters in the same order.
According to CW-5, each of the class sessions she attended followed the same routine: an instructor would teach for about one hour and then leave the platform. Ten minutes after the class was scheduled to end, different users in both the student and parent groups would post favorable reviews of the lesson and of the instructor using identical grammatical syntax.Id. at ¶¶ 81-87 (excluding screenshots provided in these paragraphs).
These allegations can be taken as is. There is no need for a mark down.
The reason: in light of the legal principles that have now been discussed a number of times, see Part III.A.1.a.i - Part III.A.1.a.iii, CW-5's allegations are (a) sufficiently detailed; (b) well-sourced (she was, for example, allegedly a student in the classes she described, see Complaint ¶ 80), and (c) amply corroborated (including by alleged screen captures of group texts she was in, which are embedded in the Complaint, see id. at ¶¶ 84-85.).
e) CW-6
Finally, consider confidential witness “CW-6,” a teacher who taught online chemistry classes for the Company for a bit more than a month during 2019, alongside 200 other teachers. See id. at ¶ 91. Per CW-6:
CW-6 and the tutors routinely assumed alternative identities to purchase promotional courses to inflate enrollment numbers. CW-6 and the tutors did this using the WeChat credentials of former “one-time” promotional-course purchasers who did not continue their enrollments. CW-6 and the other tutors took screenshots of these sham transactions and sent them to the current “one-time” customers, claiming falsely that there “were only a few
spots left” in the given course, hoping to induce more demand. Moreover, to “liven up the atmosphere” among customers, CW-6 and the other tutors routinely “brushed” course reviews in the WeChat groups, at [the Company's] direction.Id. at ¶ 93.
This is not especially telling information. It covers only a narrow period of time, and one branch location. See Wu, 2023 WL 2207422, at *8.
But there is otherwise no reason to reduce its value here. It is detailed. See generally Part III.A.1.a.i. It is based on personal knowledge. See generally Part III.A.1.a.ii. And it is corroborated by other allegations. See generally Part III.A.1.a.iii; compare, e.g., Complaint ¶ 69-71, 73 (CW-2: describing how teachers brushed student enrollment and remotely controlled bot behavior in classes) with id. at ¶¶ 92-93 (CW-6: discussing how teachers brushed student enrollment and brushed course reviews).
f) Conclusion
As set out above, some of the allegations of the confidential witnesses need to be discounted. In some instances, for example, it is not clear how a particular confidential witness knows something. See, e.g., Part III.A.1.b. And one of the confidential witnesses was not interviewed by the Plaintiffs. See footnote 20.
But these are here-and-there limitations, not across-the-board problems. By and large, the confidential witness allegations do not need to be discounted. They must generally be accepted at face value. See generally Wood, 572 U.S. at 755 n.5 (“[i]n ruling on a motion to dismiss, we have instructed, courts ‘must take all of the factual allegations in the complaint as true'”) (quoting Iqbal, 556 U.S. at 678).
Taken together, see Tellabs, Inc., 551 U.S. at 322, the confidential witnesses paint an unmistakable picture of allegedly systematic and large-scale use of bots by the Company to generate fake student enrollments.
Company employees say so. A customer says so. And so does the employee of a third-party entity --- who was allegedly paid commission to gin up class enrollments, using bot students posing as real ones. See generally Chubb Corp., 394 F.3d at 155 (“Citing to a large number of varied [confidential witness] sources may in some instances help provide particularity, as when the accounts supplied by the sources corroborate and reinforce one another.”).
As to Company employees: they allege systematic harvesting of online information to create fake “student” profiles --- and then enrolling the fake students in Company classes. See Complaint ¶ 93 (CW-6), ¶ 65 (CW-1). They allege, at least at one moment in time, a ratio of fake:real students of 100:1. See id. at ¶ 68 (CW-2). And they allege that Company headquarters was physically dominated by servers, see id. at ¶ 70 (CW-2), and “mobile phone rooms” that contained “many thousands” of phones packed on shelves, see id. at ¶ 64 (CW-1) --- presumably to be controlled by bots programmed to seem like students, and to enroll in classes.
As to the customer, she alleges being in a Company class --- and being presented with all-but unmissable indications that many of the other “people” in her class were bots. For example, when a Company teacher cut short a class by about three hours, surprisingly positive messages were then sent into “student” and “parent” group chats. The messages were from different phones --- but used the same words. See id. at ¶ 84 (“I always wanted to learn from root words like this, Teacher Song did very well. Perfect, looking forward to tomorrow's class[.]”).
And finally, the employee of a third-party entity alleges that she and her colleagues were paid to generate fake student enrollments, using a detailed blueprint supplied by the Company. The Company, it is alleged, “prepared and sent” a plan that “specified the dates, times, course titles, tuition rates, and instructors for whom the third party was to focus its . . . efforts. CW-3 and her colleagues referenced and used these plans daily to ascertain the GSX courses in which to enroll and then attend[.]” Id. at ¶ 77 (CW-3).
In sum: through confidential witnesses, the Complaint sets out detailed allegations that the Company used bots on a widespread and systematic basis to markedly increase student enrollment numbers.
2. Additional Information
As to inflated student enrollment numbers, the Complaint folds an additional allegation into the mix.
In particular, the Complaint cites a published third-party report (“Report-1”) that analyzed Company-supplied data --- and concluded that more than 70% of the Plaintiffs' students were, in fact, bots. See id. at ¶¶ 34-35.
Report-1 is a May 18, 2020 report issued by Muddy Waters Capital.
Report-1, and the other reports discussed later by the Court, are attached as exhibits to the motion to dismiss. They are incorporated by reference into the Complaint, and the Plaintiffs do not dispute their authenticity. They may be assessed here. See, e.g., Santomenno ex rel. John Hancock Tr. v. John Hancock Life Ins. Co., 768 F.3d 284, 291 (3d Cir. 2014).
The Defendants argue that, essentially for two reasons, Report-1's data-analysis should be put to the side, and not relied on here by the Court. See Motion to Dismiss at 23, 26, 28, 31.
First, the Defendants say that Report-1 was prepared by a short seller, see id. at 23, and that some courts have held that short seller reports should be given an especially hard look --and may sometimes need to be marked down. See Valdes v. Kandi Tech. Grp., Inc., 2024 WL 1348697, at *4 (E.D.N.Y. Mar. 29, 2024); Ng v. Berkley Lights, Inc., 2024 WL 695699, at *9 (N.D. Cal. Feb. 20, 2024); Saskatchewan Healthcare Emps. Pension Plan v. KE Holdings Inc., 2024 WL 775195, at *22-23 (S.D.N.Y. Feb. 26, 2024); In re DraftKings, Inc. Sec. Litig., 650 F.Supp.3d 120, 154 (S.D.N.Y. 2023); In re EHang Holdings Ltd. Sec. Litig., 646 F.Supp.3d 443, 459-60 (S.D.N.Y. 2022); In re Hebron Tech. Co., Ltd. Sec. Litig., 2021 WL 4341500, at *13 (S.D.N.Y. Sept. 22, 2021); Miao v. Fanhua, Inc., 442 F.Supp.3d 774, 801 (S.D.N.Y. 2020); cf. In re Bofl Holding, Inc. Sec. Litig., 977 F.3d 781, 797 (9th Cir. 2020).
Short sellers sell stock they do not yet own. See GFL Advantage Fund, Ltd. v. Colkitt, 272 F.3d 189, 196 (3d Cir. 2001). Usually, they accomplish this by borrowing shares from a broker for a set price or fee. See id. The short seller is required to purchase the same or equivalent number of shares to repay the broker, at a certain later time. See id. This allows short sellers to make a profit if the stock price declines between the two transactions. See id. “In other words, short sellers are betting that the stock price will decline between the time they sell the borrowed stock and the time they must ‘cover,' i.e., purchase replacement shares to repay the borrowed stock.” Id.
The Courts assumes arguendo, solely for present purposes, that the hard look approach is the right one.
The Third Circuit has not yet weighed in on this question. And not all courts have treated short seller reports with special skepticism. See, e.g., In re Eros Int'l PLC Sec. Litig., 2021 WL 1560728, at *9 (D.N.J. Apr. 20, 2021); McIntire v. China MediaExpress Holdings, Inc., 927 F.Supp.2d 105, 12324 (S.D.N.Y. 2013) (similar); Ho v. Duoyuan Glob. Water, Inc., 887 F.Supp.2d 547, 564 (S.D.N.Y. 2012) (similar); Lewy v. Sky People Fruit Juice, Inc., 2012 WL 3957916, at *13 (S.D.N.Y. Sept. 10, 2012) (similar); In re China Educ. All., Inc. Sec. Litig., 2011 WL 4978483, at *4 (C.D. Cal. Oct. 11, 2011) (similar).
For hard look courts, the question of whether and to what extent to discount short seller reports requires considering the same questions that are asked of confidential witnesses --- questions discussed above, see Part III.A.1., about level of detail, sourcing, and corroboration. See Berkley Lights, Inc., 2024 WL 695699, at *9; KE Holdings Inc., 2024 WL 775195, at *22-23; In re EHang Holdings Ltd. Sec. Litig., 646 F.Supp.3d at 459-60; In re Hebron Tech. Co., Ltd. Sec. Litig., 2021 WL 4341500, at *13; Miao, 442 F.Supp.3d at 801.
But here, a detail-sources-corroboration inquiry does not suggest any need to trim back the weight accorded to Report-1.
Report-1 is detailed. It is 25 pages, and it is granular. It lays out the number of sign-ins, users, and classes that were analyzed. See Report-1 at 3. It describes with specificity the information on users that Report-1 was able to obtain and work through --- user name, name, class join time, exit time, class ID, and IP address. See id. at 13-14. And it explains both why certain categories of behavior are likely to be indicative of bots, see id. at 5, 7-8, 10, and exactly how many Company “students” fell into each of these categories. See id. at 5-10.
As to the alleged sources that Report-1 says it is based on --- they are solid. The authors of Report-1 paid for Company classes, see id. at 13-14, 17-20, and got access to a large trove of data, see id., for users in over 200 classes. See id.
For one platform, they were able to access this data by logging into the Company's classes, selecting certain items in developer view, and then looking for a certain value for each class. By copying this value, they were able to download it into a zip file. See id. at 17-19. For the second platform, they downloaded an app on their phone and then configured the phone to send data through an intercepting HTTP proxy, which allowed them to watch messages running between the phone and the Company server. This allowed the authors of Report-1 to discover a path to the class archives, which they were able to download for each class. See id. at 19-20.
As to corroboration, the Report's core conclusion is that a large percentage of the Company's students were, in fact, bots. See Complaint ¶ 35. This is corroborated by the various confidential witnesses --- whose allegations point to that same conclusion. See Part III.A.1.
There is another kind of asserted corroboration here, but it is less telling. The Plaintiffs hired an expert who allegedly looked to the same data as the authors of Report-1, analyzed that data using the same methods, and came to essentially the same conclusions. See Complaint ¶¶ 34-35. Replication of this sort is meaningful. See generally S. Schmidt, Shall We Really Do It Again? The Powerful Concept of Replication is Neglected in the Social Sciences, Review of General Psychology 13 (2009). But here, as in many cases, replication is just table stakes, the price of admission --- an analysis generally needs to be reasonably replicable to be relied on to any extent. But the fact that a successful do-over is necessary does not mean that the do-over is corroborative. The law typically envisions a kind of triangulation, where multiple lines of allegations (here: data analysis of the kind undertaken in Report-1 --- plus allegations from confidential witnesses) all suggest the same conclusion (here: that many of the Company's students were in fact bots). Replication and triangulation are not the same thing. See generally Munafd MR, Davey Smith G, Comment, Robust Research Needs Many Lines of Evidence, Nature, Jan. 23, 2018.
In short: even assuming arguendo that short seller reports should sometimes be given a hard look and discounted, there is no need for such discounting here. This is because discounting is only potentially needed when a short seller report is wanting in terms of detail, sources and corroboration --- but Report-1 is not.
The Defendants' second argument is that, discounted or not, Report-1's analysis is flawed --- and it therefore cannot support the conclusion that most of the Company's students were actually bots.
But this, too, is not persuasive.
The Defendants argue that the data set used in Report-1 is too small, and therefore may not be representative. See Motion to Dismiss at 31-32; Int'l Bhd. of Teamsters v. United States, 431 U.S. 324, 339 n.10 (1977) (“[Considerations such as small sample size may, of course, detract from the value of . . . evidence”).
But Report-1 relies on information from over 200 classes, with data from 54,000 users, and 463,217 sign-in records from January to March 2020. See Complaint ¶ 38.
In absolute terms, these are not small numbers.
And not in relative terms, either. For one of the years in question, the Company is said to have enrolled 2,743,000 students, see id. at ¶ 234, with an average enrollment (at least for one quarter) of 1,400 students per class. See id. at ¶ 218. This implies a total of around 1,959 classes allegedly offered, and looking at data from 200 classes (as Report-1 allegedly did, see Report-1 at 14) is to look at information from around 10.2% of the Company's classes.That is a meaningful sample size. See Roscoe, J.T., Fundamental Research Statistics for the Behavioural Sciences (1975 2d ed.); Carmen R. Wilson VanVoorhis & Betsy L. Morgan, Understanding Power and Rules of Thumb for Determining Sample Sizes, 3 Tutorials in Quantitative Methods for Psych. 43, 50 (2007). And it is certainly not so small a sample size that Report-1 can be pushed aside at this stage, on the theory that it is entirely unreliable. See generally Wood, 572 U.S. at 755 n.5 (“[i]n ruling on a motion to dismiss, . . . courts ‘must take all of the factual allegations in the complaint as true'”) (quoting Iqbal, 556 U.S. at 678).
Another short seller report, quoted in the Complaint, suggests that the Company offered 1,752 classes in the first quarter of 2020. See Citron Research April 14, 2020 Report (“Report-2”) at 12. That implies Report-1 sampled around 11.4% of all of the Company's classes.
In a related vein: was the sample a random one? The Plaintiffs say so, see Complaint at ¶ 40, but it is not clear why --- Report-1 does not itself say it was based on a randomized sample. Rather, Report-1 explains that the sample that it analyzed was equally divided between two of the Company's platforms, and that the Company classes that were analyzed spanned various different grade levels and subject matters. See Report-1 at 14. This suggests the possibility of some degree of randomized sampling, though how much is unclear. But for now, there is no need to dwell on this point. The Defendants make no argument as to randomization. Rather, they make a one sentence point: “[the author of Report-1] did not explain how it selected classes for its survey or provide sufficient information to enable the reader to assess whether its sample was representative.” Motion to Dismiss at 31-32. But Report-1 does say how it selected its classes, as noted just above in this note. And the undeveloped reference to a “representative” sample is not enough to suggest Report-1 is not randomized enough to rely on at this stage.
Moreover, even if the sample size was somewhat on the small side --- that would mainly have the effect of widening the margin for error and pushing down on the confidence level. See, e.g., Sergey Dorofeev & Peter Grant, Statistics for Real Life Sample Surveys, 3-5 (2006); Howard F. Stetter, Statistical Sampling Techniques, 918, 921-22 (1962).
But that is unlikely to make a bottom-line difference here. Report-1 concludes that 70% of the Company's students were bots. See Complaint ¶ 37. Even if this were off by 10% or 20%, or even by half --- the number of bots would still be very large, and certainly large enough to be material. Cf. In re Adams Golf, Inc. Sec. Litig., 381 F.3d 267, 275 (3d Cir. 2004) (holding that a misstatement that effected around 2% of golf clubs sold over the course of one quarter was not immaterial as a matter of law); Ganino v. Citizens Utils. Co., 228 F.3d 154, 163, 166 (2d Cir. 2000) (holding that a misstatement of aftertax income by 11.9% was sufficiently material); Litwin v. Blackstone Grp., L.P., 634 F.3d 706, 722 (2d Cir. 2011) (holding a misstatement about potentially adverse trends that could impact 22.6% of total assets was sufficiently material); cf. SEC Staff Accounting Bulletin No.99, 64 Fed.Reg. 45150, 45151-2 (1999) (noting that while as a general “rule of thumb” a financial misstatement of less than 5% is not material, the Financial Accounting Standards Board believes “there are numerous circumstances in which misstatements below 5% could well be material”).
The Defendants also press other arguments as to why Report-1 should not be considered. Their strongest one is that some of those enrolled in Company classes should not have been categorized as bots on the basis of their being linked to a user that exhibited bot-like behavior, either by (1) “join[ing] class at the exact same moment” as the bot-like user, or (2) “shar[ing] the same distinct [internet protocol] address” as the bot-like user. Motion to Dismiss at 30, 32. But the Plaintiffs' counter-argument, rooted in their expert's report, see Complaint ¶ 36, is certainly plausible enough. As to (1): perfectly synchronized logins, down to the second, by multiple devices are unlikely to occur frequently by coincidence. And as to (2): phones that are used by a bot once are likely to be bot-controlled in the future. See Brief in Opposition at 26 n.7; Lutz v. Portfolio Recovery Assocs., LLC, 49 F.4th 323, 334 (3d Cir. 2022) (“pleading[s] receive[] the benefit of reasonable inferences at the motion-to-dismiss stage”); Bagic v. Univ. of Pittsburgh, 773 Fed.Appx. 84, 88-89 (3d Cir. 2019) (“[w]hen considering a motion to dismiss, a district court cannot weigh competing inferences and forgo drawing a reasonable one in the plaintiff's favor”).
3. Conclusion
The Complaint contains ample allegations that a very large percentage of the Company's enrolled students were, in fact, bots --- such that the Company's publicly-announced student enrollment numbers were not only false, but materially so.
At one point, the Defendants contend there was no actionable misrepresentation here, because the Company disclosed that rogue employees might engage in brushing or fraud. See Motion to Dismiss at 21-22; see generally In re Aetna, Inc. Sec. Litig., 617 F.3d 272, 282 (3d Cir. 2010) (“cautionary language, if sufficient, renders the alleged omissions or misrepresentations immaterial as a matter of law”). But the allegations here are of widespread, systematic, and intentional use of bots to increase the Company's student enrollment figures. This is a long way from what is conveyed by a disclosure as to a generic internal controls risk, and even one that spotlights rogue employees.
B. Revenue
Turn, now, to another set of allegedly false statements --- as to the revenue earned by the Company (“Revenue Statements”).
The Revenue Statements were also made in 2019 and 2020.
From 2019, the Revenue Statements were made in initial public offering materials (see Complaint ¶¶ 198-199); in Securities and Exchange Commission filings (see id. at ¶¶ 202-05, 212-15, 222-23); and orally by the CEO and the CFO during earnings calls (see Id. at ¶¶ 206-09, 218-21).
The Revenue Statements from 2020 were in: Securities and Exchange Commission filings (see id. at ¶¶ 226-233, 247-250, 263-66, 275-278); Company press releases (see id. at ¶¶ 261-62, 269-72); oral statements from the CEO and the CFO during earnings calls (see id. at ¶¶ 241-44); and oral statements by the CEO and CFO in other public contexts (see id. at ¶¶ 253-60).
According to the Complaint, the Revenue Statements were false, and materially so. See, e.g., id. at ¶¶ 203, 248, 264, 272.
The Defendants argue the revenue misstatements are not properly pled. See Motion to Dismiss at 35, 37-39.
The Court disagrees, and concludes they were.
As noted above, there are ample allegations of systematic efforts by the Company to use bots to greatly inflate the number of enrolled students in Company classes. See Part III.A.
And it stands to reason that the Company's revenues are a function of the number of students it enrolls.
And there are, in any event, specific allegations to that effect. See Complaint ¶¶ 27-32.
Because of this, there is a strong inference that all-but inescapably runs from (a) allegedly systematic and large overstatements of student enrollment numbers to (b) overstatements of revenue numbers.
That inference is propped up here by allegations from CW-2, a Beijing-based engineering manager for the Company, see Complaint ¶ 67, that bot enrollments were reported by the Company as revenue. See id. at ¶ 72.
In addition, the Complaint contains allegations from another third-party report (“Report-2”). See id. at ¶¶ 96-100. Report-2's authors bought and enrolled in Company classes, see Report 2 at 11; tracked the comments in the classes, see id. at 12; and discerned that the classes seemed to have very few actual students, see id. at 12, 27-28 --- too few to generate the revenue numbers publicly reported by the Company, see id. at 13.
The Defendants say Report-2 is a short seller report. See Motion to Dismiss at 8. As noted above, some courts mark down the value of such reports. See Part III.A.2. But no discount is warranted here. First, Report-2 is detailed --- mustering highly specific information about those classes that were tracked, see Report-2 at 23-26, per-class revenue calculations, see id. at 13, and information from actual Company revenue reports, see id. at 3, 9, 13. Second, Report-2 is predicated on sufficiently solid sources --- allegedly direct observations, gleaned from enrolling in Company classes and from working to identify the bona fide students in them. See id. at 23-26. And third, Report-2's conclusion that there were few real students in the relevant classes is well-corroborated, by extensive allegations from the various confidential witness that many of the Company's students were in fact bots. See Part III.A.
How to explain the discrepancy?
That is where CW-2's allegations come in. Bot enrollments, per CW-2, were allegedly being counted as Company revenue. And that helps to explain what Report-2 documented: revenues that were too high given the apparent numbers of real, human student enrollments.
Bottom line: the Complaint lays out more-than-sufficient allegations that the Revenue Statements are actionable.
The Revenue Statements were false, because they allegedly reflected “revenue” that was not, in fact, bona fide revenue.
For sources explaining why the Company's paying third parties to generate revenue could not count as Company revenue, see, for example, J. Randolph Mallek, et al., Revenue Recognition: Fundamental Principles, Bloomberg Tax & Acct. Portfolio 51002nd; Raj Gnanarajah, Cong. Rsch. Serv., R43811, Cash Versus Accrual Basis of Accounting: An Introduction, 19 (2014). The Complaint's allegations suggest the Company was engaging in a practice closely adjacent to round-tripping. See generally Gould v. Winstar Comm'cn, 692 F.3d 148, 155 (2d Cir. 2012) (“‘[r]ound-tripping' typically refers to reciprocal agreements, involving the same products or services, that lack economic substance but permit [both] parties to book revenue and improve their financial statements” and is a way companies may misrepresent their revenue); Tchrs. Ret. Sys. of LA v. Hunter, 477 F.3d 162, 178 (4th Cir. 2007) (round trip transactions are “improper because the parties book revenues even though the transactions ‘wash out' without any economic substance”); see also Suprema, 438 F.3d at 265 (round-trip sales scheme led to a misrepresentation of a large portion of the company's revenue). But whether the revenue inflation here is or is not best understood as akin to round-tripping, there is no doubt that allegedly fraudulent practices to materially inflate revenues are potentially actionable under the Exchange Act of 1934. See generally In re Cabletron Sys., Inc., 311 F.3d at 24 (fraudulent practices to inflate revenue, such as reporting false sales, leading to improperly recognizing revenue of between $20-30 million per quarter “certainty add up to fraudulent and material misstatements”); Aviva Partners LLC v. Exide Tech., 2007 WL 789083, at *16 (D.N.J. Mar. 13, 2007) (“Published earnings figures that are allegedly false or misleading are the types of statements actionable under Section 10(b) and Rule 10b-5.”); see also In re Cognizant Tech. Sols. Corp. Sec. Litig., 2018 WL 3772675, at *22 (D.N.J. Aug. 8, 2018).
Were the alleged misstatements material?
The Court's conclusion: yes.
The very large scale of the Company's alleged misrepresentations as to student enrollments virtually guarantees material misstatements as to revenue numbers. This is because the latter were based on the former.
And more directly: Report-2 concludes that Company revenue was misstated by up to 70%, and it does so based on a methodology that, at least at this stage, is sensible enough. See Report-2 at 13.
Against the conclusion that the Revenue Statements were materially misstated, the Defendants offer two main counterarguments.
First, they suggest that fake student enrollments did not lead to inflation of revenue data. See Motion to Dismiss at 36-39.
What, then, were all of the alleged bots for? The theory would seem to be that bots made Company classes feel busy and vibrant. That was their purposes, not to send enrollment fees back to the Company.
It is not crystal clear that the Defendants are indeed making this argument, see id. at 36-37, but the Court proceeds as if they do.
But this argument is, among other things, directly contradicted by the allegations from CW-2 set out just above. See Complaint ¶ 72 (“GSX reported these enrollments as ‘revenues'”). And “[i]n ruling on a motion to dismiss . . . courts ‘must take all of the factual allegations in the complaint as true.'” Wood, 572 U.S. at 755 n.5 (quoting Iqbal, 556 U.S. at 678).
Second, the Defendants contend that Report-2 is not reliable, because, among other things, it was based on a too-small sample size. See Motion to Dismiss at 38 n.20.
But this is not persuasive. Report-2 is based on the Report's authors buying seats in over 20% of the Company classes offered during a particular time period. See Report-2 at 12. This is not too small a sample size to reason from. See Part III.A.2.
C. Spring 2020 Statements
Look now to a final set of allegedly false statements, said to have been made by the Company and by Company leaders in April, May, and June of 2020 (“Spring 2020 Statements”).
Consider these one at a time.
First, on June 3, the Company denied certain allegations contained in a “short seller” report. See Complaint ¶ 273.
The Defendants contend this denial was not alleged in a sufficiently specific way, see Motion to Dismiss at 40 n.22, and this Court (per Judge Salas) previously so held. See Wu, 2023 WL 2207422, at *11.
“[W]hen a court decides upon a rule of law, that decision should continue to govern the same issues in subsequent stages in the same case.” Bailey v. Viacom Inc., 435 Fed.Appx. 85, 91 (3d Cir. 2011) (citing Arizona v. California, 460 U.S. 605, 618 (1983)) (cleaned up).
That is this case. Judge Salas “decide[d]” that the June 3 alleged misstatement was not pled with sufficient particularity. And that is controlling here, because the Plaintiffs, given the opportunity to re-plead, have not added any new allegations.
Second, on April 8, 2020 the Company CEO allegedly said: “I think if [the author of Report-1] analyses our data seriously, there is a high probability that [they] will not be so stupid. The level and IQ of the [Report-1 author] is quite high.” See Complaint ¶ 253.
But as pled, this statement is too indefinite to go on.
“Data,” for example, might be a reference to information about student enrollments and revenues. But “data” might also be something else entirely --- the various reports here contain information about subjects other than enrollments and revenues.
For example, Report-3 said that the Company was not doing as well as its competitors, that its teachers were inexperienced, that the Company's founders were planning their exits, that shareholders were selling their stock, and that management's background presented a risk to the Company. See Report-3 at 28, 41, 47, 50, 52.
And saying that there is a “high probability” that someone else will “not be so stupid” if they do a “serious[]” analysis is, more than anything, hard to understand --- let alone to take as a backwards-looking endorsement of the accuracy of Company enrollment and revenue numbers.
Third, on April 9, 2020 the CEO and CFO allegedly held a teleconference with investors. See Complaint ¶ 255. During this teleconference, they made a number of statements. See id. at ¶¶ 255-59. To the extent those alleged statements included specific assertions about Company enrollment or revenue data, those statements are actionable here. They are particular. And the Court has already discussed the various allegations that suggest Company enrollment and revenue information were materially defective. See Part III.A.; Part III.B.
Fourth, on April 15, 2020 the Company allegedly issued a press release that said it “firmly denied the false and ungrounded allegations raised in [Report-2].” Complaint ¶ 261.
The allegation that the Company's revenue was inflated by 70% is the dominant allegation in Report-2. See Report 2 at 1, 8, 1020, 23-33.
Other portions of Report-2 focused on how the Company's growth was “too good to be true” and briefly address fake students in Company classes. See id. at 2-10, 21-22.
“[F]irmly den[ying]” that was a misstatement. See Part III.B. (holding that the allegations the Company was inflating revenue were adequately pled).
And it was a material misstatement, too --- it would have had the effect of lulling investors into thinking there was nothing to see in Report-2, and that they should move on. See Avaya, 564 F.3d at 247-48 (holding misstatements are actionable, including from CFO in the face of questioning, about company being “on track” to meet financial goals when it was not); In re Pareteum Sec. Litig., 2021 WL 3540779, at *8-9, *11 n.2 (S.D.N.Y. Aug. 11, 2021) (holding misstatements actionable, including a denial of short seller reports); In re Signet Jewelers Ltd. Sec. Litig., 389 F.Supp.3d 221, 231 (S.D.N.Y. June 11, 2019) (holding misstatements actionable when “[i]n the face” of allegations of “rampant sexual harassment . . .
Defendants sought to reassure the investing public that Signet did not, in fact, have a toxic workplace. They did so by including representations in their periodic SEC filings that the company expressly ‘denies the allegations.' . . . As alleged, their word was not truthful.”) (cleaned up); City of Brockton Ret. Sys. v. CVS Caremark Corp., 2013 WL 6841927, at *1-4 (D.R.I. Dec. 30, 2013) (statement “denying that the company had lowered prices for some of its . . . customers ‘because of a lack of service,' or that the company had ‘an issue with its computer systems'” were misrepresentations when the allegations demonstrated the company had problems with service and the integration of its systems).
Fifth, on May 19, 2020 the Company issued a press release that “refuted” Report-1. See Complaint ¶ 269.
Sixth, on May 29, 2020 the Company issued a press release that “refuted” a follow-up to Report-1, and that provided further support for Report-1's conclusion. See id. at ¶ 271.
The allegation that the Company was using bots to inflate enrollment numbers is the dominant allegation in Report-1. See Report-1 at 2-14, 17-25. In light of the Court's conclusion as to Company bot usage in Part III.A, these two “refut[ations]” are actionable.
D. Conclusion
The Complaint is laced with strong-enough allegations that Company enrollment and revenue numbers were systematically and materially inflated. See Part III.A, III.B.
Given this, the Plaintiffs have properly pled that the Enrollment Statements, see Part III.A, and the Revenue Statements, see Part III.B, were materially false, and that some of the Spring 2020 Statements were, too. See Part III.C.
IV. Scienter
As to the actionable statements here, see Part III.D, were these made with scienter, as the law requires?
To answer this scienter question, the allegations must be considered on a “holistic[]” basis, Tellabs, Inc., 551 U.S. at 326, looking to the full body of information in the Complaint. See also Avaya, 564 F.3d at 267-68 (“The pertinent question is whether all of the facts alleged, taken collectively, give rise to a strong inference of scienter, not whether any individual allegation, scrutinized in isolation, meets that standard.”) (cleaned up) (emphasis added); Hacker, 687 F.Supp.3d. at 589.
The Court undertakes that analysis below, and concludes that when the allegations the Plaintiffs added to the current version of the complaint are considered --- the scienter inference is strong enough, and markedly more persuasive than any competing inference.
The stepping-off point for this analysis is a description of the legal standards that govern, set out in Part IV.A and Part IV.B. The analysis then gets underway in Part IV.C.
A. General Legal Standards
For a Section 10(b) claim, the “required state of mind is scienter --- the intent to deceive, manipulate, or defraud either knowingly or recklessly.” Pamcah-UA Loc. 675 Pension Fund v. BT Grp. PLC, 2021 WL 3415060, at *1 (3d Cir. Aug. 5, 2021); see also Hacker, 687 F.Supp.3d. at 588; see generally Ernst & Ernst v. Hochfelder, 425 U.S. 185 (1976).
Here, the argument is that the Defendants were reckless. See Brief in Opposition at 9.
Recklessness is “not merely simple, or even inexcusable negligence, but an extreme departure from the standards of ordinary care[.]” Avaya, 564 F.3d at 267 n.42 (cleaned up); see also Hacker, 687 F.Supp.3d. at 589.
Because scienter is a state of mind, it must often be established indirectly. It has to be “inferred.” See Hacker, 687 F.Supp.3d. at 589.
To be legally sufficient, an inference of scienter must be “strong.” 15 U.S.C. § 78u-4(b)(2)(A). “A complaint adequately pleads a strong inference of scienter ‘only if a reasonable person would deem the inference of scienter cogent and at least as compelling as any opposing inference one could draw from the facts alleged.'” In re Hertz Glob. Holdings Inc, 905 F.3d 106, 114 (3d Cir. 2018) (quoting Tellabs, 551 U.S. at 324). The standard is demanding. It “is not easy to allege.” Pamcah-UA Loc. 675 Pension Fund, 2021 WL 3415060, at *1; see also Hacker, 687 F.Supp.3d. at 589.
B. Pleading Scienter: Three Questions
What a plaintiff needs to adequately allege is scienter.
But how does a court go about discerning whether the scienter bar has been cleared in a given case?
To begin answering that question, first take a step back.
The named defendants in many securities fraud cases include a company's senior executives. There are many reasons why. Among them: senior executives are the people who are generally authorized to speak publicly for the company, and it is allegedly false public statements about a company that are at the core of securities law prohibitions.
But it can be difficult to determine whether a senior executive has scienter.
One reason, especially in the context of larger companies, is that senior executives are senior. They exercise a relatively wide span of control, almost by definition. And this can make it difficult at the motion to dismiss stage, before there has been discovery, to sort between senior executives who had scienter and those that did not.
A CEO may have said something false --- but without knowing it was false. Perhaps because accurate information on that particular topic did not squeak its way through the dense crush of the CEO's focus on a large number of other pressing concerns. Or perhaps because accurate information simply failed to percolate its way up through what may be a sprawling company hierarchy.
Or on the other hand, the CEO may have said something false --and did know it. Because, for example, it was about a subject the CEO was laser-focused on --- a subject that the CEO prioritized even beyond other competing concerns and as to which the CEO was especially knowledgeable, requiring frequent and extensive updates from all corners and levels of the organization.
These are stylized possibilities. But they underscore the difficulty of making an accurate assessment of whether there is a strong inference of scienter at the motion to dismiss stage, before the peek behind the curtain that discovery allows.
To address this difficulty, federal law focuses on a set of questions that can help distinguish between instances in which an allegedly false statement was made with scienter and instances in which it was not.
Three of those questions follow.
These three questions are just introduced here. The fuller substance of them is discussed more fully below, in Parts IV.C.1, Parts IV.C.2, and Parts IV.C.3.
First, was the issue as to which the statement was made especially important to the Company?
This question is often discussed in terms of the “core operations” doctrine. See, e.g., Avaya, 564 F.3d at 269, 272; Rahman, 736 F.3d at 246; S. Ferry LP, No.2 v. Killinger, 542 F.3d 776, 784-85 (9th Cir. 2008), or in terms of the sheer scale of the issue. See, e.g., Matrix Cap. Mgmt. Fund, LP v. BearingPoint, Inc., 576 F.3d 172, 184 (4th Cir. 2009); Garfield v. NDC Health Corp., 466 F.3d 1255, 1267-68 (11th Cir. 2006); Adams v. Kinder-Morgan, Inc., 340 F.3d 1083, 1106 (10th Cir. 2003); Rocker Mgmt., L.L.C. v. Lernout & Hauspie Speech Prod. N.V., 2005 WL 1366025, at *8 (D.N.J. June 8, 2005); P. Schoenfeld Asset Mgmt. LLC v. Cedant Corp., 142 F.Supp.2d 589, 609 (D.N.J. 2001).
Second, was the issue as to which the statement was made especially important to the particular senior executive whose alleged scienter is in play?
And third, was there something that would have helped to cut through the noise --- to pique the executive's focus on the statement by suggesting a real possibility that it was false?
This question is often discussed in terms of the metaphor of “red flags.” See, e.g., Doshi v. Gen. Cable Corp., 823 F.3d 1032, 1039 (6th Cir. 2016); City of Taylor Gen. Emps. Ret. Sys. v. Astec Indus., Inc., 29 F. 4th 802, 813-14 (6th Cir. 2022); Novak v. Kasaks, 216 F.3d 300, 308 (2d Cir. 2000).
“Yes” answers to these three questions can help to support the conclusion that a false statement was made with scienter.
These questions are taken up below, in reverse order.
C. Analysis of the Three Questions
Before getting to the analysis, a final preliminary matter.
The Defendants do not seek to differentiate between the scienter of the parties. They speak of scienter as a whole, and do not address the possibility that the scienter of certain Defendants might be on a different footing than others'. See Motion to Dismiss at 11-17.
This approach may well make sense.
On the allegations here, the scienter of the Company CEO and the Company CFO (two of the Defendants) would likely be imputed to the Company (the third Defendant). See In re ChinaCast Educ. Corp. Sec. Litig., 809 F.3d 471, 476 (9th Cir. 2015); Makor Issues & Rights, Ltd. v. Tellabs, Inc., 513 F.3d 702, 707-08 (7th Cir. 2008); Adams, 340 F.3d at 1106-07. So trying to tease apart CEO/CFO scienter from Company scienter might not have been worth the candle for the Defendants.
And in a similar vein, the allegations as to the CEO's scienter and the CFO's scienter are largely baked together in this case. See Complaint ¶ 206 (CEO and CFO together presenting second quarter 2019 financial results on August 2019 earnings call); id. at ¶ 218 (CEO and CFO together presenting third quarter 2019 financial results on November 2019 earnings call); id. at ¶ 255 (CEO and CFO jointly held teleconference to address “market concerns” on April 9, 2020).
Trying now to separate out the ingredients --- by arguing, for example, that the CEO had scienter but the CFO did not --- is hard to make work.
But one way or another, the bottom line is the same: the Defendants do not seek to differentiate between the scienter of one Defendant and the scienter of another.
The Court will follow the Defendants' lead.
The Defendants “know what is best for them,” and deciding what arguments to “advance,” is fundamentally their call. United States v. Sineneng-Smith, 590 U.S. 371, 375-76 (2020) (cleaned up); see also, e.g., United States v. Samuels, 808 F.2d 1298, 1301 (8th Cir. 1987) (R. Arnold, J., concurring in denial of rehearing en banc).
Accordingly, the Court proceeds here on the assumption that the allegations in the Complaint that support or do not support scienter should be applied across-the-board --- to assess the alleged mental state of each of the three Defendants, for all the misstatements.
With that preliminary note out of the way, begin the analysis, with a look at each of the three questions described in Part IV.B, starting with the last one, which is often discussed using the “red flags” metaphor.
1. Red Flags
a) In General
Red flags can come in different types. Two are relevant here.
A first type of red flag: allegations that a person has been directly presented with information that suggests a strong possibility that a statement the person later makes is materially false or misleading.
See, e.g., Matrixx Initiatives, Inc. v. Siracusano, 563 U.S. 27, 49 (2011) (scienter when company made certain statements after receiving direct reports as to adverse drug effects that tended to suggest that the statements might not be accurate); Avaya, 564 F.3d at 268-69 (scienter allegations would have been “foritif[ied]” by allegations of a “particular document or conversation that would have informed” relevant people that certain statements they made may not have been accurate); Martin, 757 Fed.Appx. at 154 (no scienter, and noting that “[t]he complaint contains no allegations that [relevant people] . . . knew that GNC's DMAA-replacement products may have contained ingredients banned by the FDA, or that they received any report that banned substances may be included in replacement supplements”); Astec Indus., Inc., 29 F. 4th at 813-14 (holding that scienter was adequately alleged when statements of the CEO were “contradict[ed] [by] internal reports” and that the court “need not view” his optimistic statements as “mere ignorance”); City of Dearborn Heights Act 345 Police & Fire Ret. Sys. v. Align Tech., Inc., 856 F.3d 605, 620 (9th Cir. 2017) (“particularized allegations that defendants had ‘actual access to the disputed information' may raise a strong inference of scienter”); Elam v. Neidorff, 544 F.3d 921, 929 (8th Cir. 2008) (“allegations that a defendant made materially misleading statements, while in possession of conflicting information, support a strong inference of scienter”); Teamster Local 445 Freight Div. Pension Fund v. Dynex Cap. Inc., 531 F.3d 190, 196 (2d Cir. 2008) (noting “knowledge of access to information” that suggests public statements were inaccurate supports scienter); cf. BT Grp. PLC, 2021 WL 3415060, at *2 (suggesting that scienter allegations might have been on different footing if a particular telling email had actually been received by the defendants); OFI Asset Mgmt., 834 F.3d at 499 (noting, in the securities fraud context, the highly limited inferences that can be drawn from a needle in a haystack --- “a single phrase buried within a filing that encompassed dozens of pages”). Note that the force of the inference of scienter may depend, in part, on how much time elapsed between getting the information and making the statement. See Avaya, 564 F.3d at 271-72; Doshi, 823 F.3d at 1039.
A second sort of red flag is both less direct and less telling.
This type of red flag is “a signal of possible danger around the curve.” Hacker, 687 F.Supp.3d at 591. It provides a reason to investigate further before making a statement, to “perceive[] possible trouble ahead, and slow[] down.” Id. It can be reckless within the meaning of the securities laws to ignore this sort of red flag, especially when it is “bright,” because what is conveys is unmistakable, and “big,” because it relates to something especially important. Id.
For an auditor, missing paperwork might be a red flag of this kind, pushing the auditor to study a company's financial records that much more deeply. See Suprema, 438 F.3d at 280.
For company management, being directly exposed to certain sorts of questions and concerns can sometimes be a red flag of this sort --- even if the executive in question does not have direct evidence that the questions and concerns are well-founded.
In Avaya, for example, the Third Circuit determined that a “strong inference” of scienter had been established in part because “specific” and “focused” questions from market analysts about a major issue were put “directly and repeatedly” to a company executive --- who then answered the questions without any hedge. Avaya, 564 F.3d at 269-70.
And Avaya does not stand alone. See also Rahman 736 F.3d at 246; In re Fibrogen, Inc., 2022 WL 2793032, at *24 (N.D. Cal. July 15, 2022); In re Qualcomm Inc. Sec. Litig., 2019 WL 1239301, at *11 (S.D. Cal. Mar. 18, 2019); Roofer's Pension Fund v. Papa, 2018 WL 3601229, at *21-22 (D.N.J. July 27, 2018); KBC Asset Mgmt. NV v. 3D Sys. Corp., 2016 WL 3981236, at *5, *9, *11 (D.S.C. July 25, 2016); In re Urban Outfitters, Inc. Sec. Litig., 103 F.Supp.3d 635, 653-54 (E.D. Pa. 2015).
Bottom line: sharp, detailed, and recurring questions that challenge the company's position on important matters can strongly bear on scienter.
Why? Because it stands to reason that such questions can ordinarily be expected to be a goad to further inquiry. And if after being pushed by such questions, a person nonetheless provides false information, a possible inference is that the person (a) did not investigate before providing the false information (which might establish recklessness) or (b) investigated, determined the truth, but made a false statement anyway (which might establish intent).
b) In This Case
Pivot back now, from the legal principles sketched out just above, to this case --- and note that there were red flags here of both the first and second kinds. As to the first sort of red flag, there are five relevant reports. See Complaint ¶ 253, 255, 259, 261, 269, 271, 273. Each of these five reports claimed the Company's revenue numbers and enrollment numbers were greatly overstated. See id. at ¶¶ 272, 274. Some of the reports were very detailed. See Report-1; Report-2.
In turn, there are ample allegations that Company leaders, including the CEO and CFO, were directly aware of some of the reports. See id. at ¶¶ 253, 255, 259. (There is some likelihood that other reports were seen by the CEO and CFO, too. See id. at ¶¶ 261, 269, 271. And there is an allegation that an additional report was directly seen by the CEO. See id. at ¶ 330.)
Indeed, the Company CEO and the Company CFO, and the Company through its press releases, directly commented on the reports, see id. ¶¶ 261, 269, 271, 273 --- and allegedly expressed the sorts of opinions about the reports that can only be based on reviewing them. See id. at ¶¶ 269, 271, 273.
Bottom line: the Defendants made statements, the gist of which were that no systematic revenue or student enrollment inflation was taking place, see Part III.A, III.B --- even though they had been provided with detailed information that directly suggested precisely the opposite conclusion.
This is a firm basis from which to infer scienter.
And that inference is buttressed, at least to an extent, by red flags here of the second sort. The Complaint lays out allegations that suggest the CEO and the CFO were aware of questions related to the possibility that there were enrollment and revenue troubles. For example, the CEO and CFO held a conference call to “address market concerns” after a report was released that contained allegations of false enrollment numbers and revenue. See Complaint ¶ 255. And to cite another example: the Company responded to questions regarding its “customer acquisition strategy” in February 2020. Id. at ¶ 245.
In the face of this, there is an inference --- though not the strongest one --- the CEO and CFO either followed-up internally and inquired (in which case they should have been aware of what was allegedly afoot) or they did not (in which case they were acting in a reckless manner).
2. Position in the Company
Aside from a focus on red flags, courts routinely look to a person's position in a given company to determine whether the person's statements were made with scienter. See generally Part IV.B.
This is the approach in the Third Circuit. See e.g., Avaya, 564 F.3d at 271; In re PTC Therapeutics, Inc. Sec. Litig., 2017 WL 3705801, at *17 (D.N.J. Aug. 28, 2017). And throughout the United States. See, e.g., E. Ohman J:or Fonder AB v. NVIDIA Corp., 81 F.4th 918, 946 (9th Cir. 2023), cert. granted sub nom. Nvidia Corp. v. Ohman J, No. 23-970, 2024 WL 3014476 (U.S. June 17, 2024); Okla. Firefighters Pension & Ret. Sys. v. Six Flags Ent. Corp., 58 F.4th 195, 219 (5th Cir. 2023); Constr. Indus. & Laborers Joint Pension Tr. v. Carbonite, Inc., 22 F.4th 1, 9-10 (1st Cir. 2021); In re VeriFone Holdings, Inc. Sec. Litig., 704 F.3d 694, 704, 710 (9th Cir. 2012); Dorsey v. Portfolio Equities, Inc., 540 F.3d 333, 342 (5th Cir. 2008); In re Stone & Webster, Inc., Sec. Litig., 414 F.3d 187, 211 (1st Cir. 2005); Nursing Home Pension Fund, Local 144 v. Oracle Corp., 380 F.3d 1226, 1234 (9th Cir. 2004); Adams, 340 F.3d at 1106.
This approach makes sense.
Imagine an executive at a car manufacturer who falsely describes a new model's attributes --- the stopping power of its brakes, for example, or its turning radius and safety features. It stands to reason that such a false description is more likely to have been offered with scienter by the company engineer (who probably knows a great deal about brakes, etc.) than by the company's controller (who likely knows less about a given car's performance specs). Cf. In re VeriFone Holdings, Inc. Sec. Litig., 704 F.3d at 704, 710; In re Stone & Webster, Inc., Sec. Litig., 414 F.3d at 211.
What of the statements as to revenue and student enrollment? Were they within the Defendants' workplace bailiwick?
The Court's conclusion: yes.
As for publicly-reported revenue numbers, that was plainly within the CFO's job description. See Avaya, 564 F.3d at 271. And there are ample allegations in the Complaint that the CEO and the CFO were both focused on revenue numbers. See Complaint ¶¶ 300-01, 313-14.
As for student enrollments, more of the same: there are sufficient allegations that the CEO and the CFO were zeroed-in on them. See Id. at ¶¶ 302, 304, 306; cf. Id. at ¶ 316.
To be sure, the allegations alluded to in the preceding two paragraphs are associated with three confidential witnesses. But those allegations do not need much discounting. See generally Part III.A.1.a (describing caselaw in this area).
They are granular. See, e.g., Complaint ¶ 300 (“[the CFO] was responsible for knowing how much the company was spending on marketing, and specifically responsible for knowing and understanding the individual line items that made up the marketing budget”); id. at ¶ 306 (“CW-9's team developed a dashboard for senior executives, which they called the ‘boss dashboard.' It included flow of products, number and analysis of students, teacher evaluations, and class arrangements.”).
They are, in large part, based on close-in workplace sources. One of the confidential witnesses was allegedly the CFO's direct report for part of her time at the Company. See id. at ¶ 299. Another allegedly spoke to the CEO directly. See id. at ¶ 313. Another allegedly spoke to the CEO daily. See id. at ¶ 304.
And the confidential witness allegations are corroborated. By each others' allegations. By various statements in the Complaint that suggest hands-on involvement by the CEO and CFO in revenue/enrollment issues. See id. at ¶¶ 259, 323, 325-27, 329. And by the basic plausibility of what is alleged. There is nothing odd or surprising about a CFO (or a CEO) following revenue numbers closely, or a CEO (or a CFO) keeping a hawk's eye on the company's core customer base (enrolled students).
Some limited discounting is necessary. One of the confidential witnesses worked for the Company outside of the relevant period, and another worked for the company only for a short time during the class period. See Complaint ¶¶ 299, 304; see generally Wu, 2023 WL 2207422, at *8; Chan v. New Oriental Educ. & Tech. Grp. Inc., 2019 WL 2865452, at *10 n.7 (D.N.J. July 3, 2019); Schaffer v. Horizon Pharma. PLC, 2018 WL 481883, at *8 (S.D.N.Y. Jan. 18, 2018). But while this requires some mark down, it does not require much. There is no strong reason to think the CEO and CFO's job responsibilities changed very dramatically as between one time or another --- or that, at any point, the CEO and CFO were not focused on revenue and student enrollments. (The allegations of the fourth confidential witnesses, CW-10, are not considered by the Court. There are no allegations that CW-10 had any direct interactions with the CEO or CFO. Without this, the allegations lack a sufficient “source.” See Part III.A.1.a.ii.)
The newly-added allegations described in this Part IV.C.2 add a great deal to the mix of allegations in the Complaint, and from a “holistic[]” perspective, Tellabs, Inc., 551 U.S. at 326, they decidedly strengthen the inference here of scienter --- well beyond what was before the Court in the Plaintiffs' previous complaint.
3. Importance to the Company
Finally, in assessing whether scienter has been adequately pled, courts frequently look to whether alleged misstatements (a) relate to critical parts of a company's operations or (b) are especially large. See Avaya, 564 F.3d at 271; Rahman, 736 F.3d at 246; In re Hertz Glob. Holdings Inc., 905 F.3d at 116; Carbonite, Inc., 22 F.4th at 9; Local No. 8 IBEW Ret. Plan & Tr. v. Vertex Pharmas., Inc., 838 F.3d 76, 82-83 (1st Cir. 2016); In re VeriFone Holdings, Inc. Sec. Litig., 704 F.3d at 704, 709; N.M. State Inv. Council v. Ernst & Young LLP, 641 F.3d 1089, 1100 (9th Cir. 2011); BearingPoint, Inc., 576 F.3d at 184; Killinger, 542 F.3d at 784-85; Garfield, 466 F.3d at 1267-68; Adams, 340 F.3d at 1106; City of Philadelphia v. Fleming Cos., 264 F.3d 1245, 1261 (10th Cir. 2001); Hacker, 687 F.Supp.3d. at 599-600; Rocker Mgmt., L.L.C. v. Lernout & Hauspie Speech Prod. N.V., 2005 WL 1366025, at *8 (D.N.J. June 8, 2005); P. Schoenfeld Asset Mgmt. LLC v. Cedant Corp., 142 F.Supp.2d 589, 609 (D.N.J. 2001); In re Medtronic Inc., Sec. Litig., 618 F.Supp.2d 1016, n.19 (D. Minn. 2009).
The law in this area tracks common sense. Even amidst competing pulls on their time and attention, company leaders are more likely to be focused on the heart of the venture --- and also on any serious, large-scale problems that are eating at the company, even if they do not arise in its operational core. Moreover, because of the likely importance to investors of statements about central corporate matters, company leaders can be expected to be especially thoughtful when speaking about them. See generally Carbonite, Inc., 22 F.4th at 9-10; Fleming Cos., 264 F.3d at 1261.
Here, there can be little doubt that revenue and student enrollment were no backwaters for the Company.
Revenue is the lifeblood of any for-profit firm. And following enrollments was, in essence, an effort to follow the number of customers the Company had.
Revenue and enrollments were, in short, plainly among the Company's most fundamental metrics. And the Company acted like it --- tracking those numbers at the highest level, and with direct reporting to the CEO and CFO. See Complaint ¶¶ 301, 306, 315-16.
Moreover, the alleged “misses” here --- as to revenue numbers, for example, that are asserted to have been off by at least half, see id. at ¶ 199 --- would have been large enough to get and hold the attention of the CEO and CFO. Cf. In re Hertz Glob. Holdings Inc, 905 F.3d at 116 (collecting cases that infer scienter based on the size of restated financial statements).
D. Conclusion
The Court has analyzed scienter in light of the three questions often considered by courts in this context. See Part IV.B.
Based on this, the Court concludes that the Plaintiffs have pled a “strong inference” of scienter.
The Defendants were made directly aware of information that suggested a large risk that revenues and enrollment were vastly overstated. See Part IV.C.1. Moreover, they were aware of investor concerns about the revenue and enrollment numbers. See Part IV.C.1. These were red flags, of each of the two basic types.
In addition, the CEO and CFO could have been expected, by virtue of their positions, to be attuned to revenue and enrollment data --- and indeed the allegations are that they were, see Part IV.C.2, perhaps in part because of the obvious importance of such data to the Company, see Part IV.C.3.
Are there other inferences here, of something other than scienter? See generally Tellabs, Inc., 551 U.S. at 323-24.
The Defendants argue that the Plaintiffs' theory of scienter is not as compelling as the competing inference --- that “there was no fraud, let alone any fraud of which the Defendants were aware.” Motion to Dismiss at 17.
But as to the major note (“there was no fraud”), the Court has already held that the Plaintiffs have adequately pled fraud, see Part III, so this argument is to no avail.
And as to the minor note (there was no “aware[ness]” of fraud), the argument would presumably be that if revenue and enrollment numbers were inflated --- that was the work of lower-level Company employees, and the CEO and CFO were unaware of it.
But this is not more persuasive than the inference of scienter.
It is alleged that the CEO and CFO were repeatedly and directly presented with detailed information that suggested large-scale revenue and enrollment inflation.
Even if revenue and enrollment inflation had at some point been obscure to them --- that stopped being the case after they were allegedly presented with the various reports that detailed serious revenue/enrollment problems.
And more basically, the allegations in the Complaint paint a picture of a Company that systematically and intentionally made use of bots to drive up “student” enrollment --- to make the Company look more revenue-rich, and accordingly more attractive to investors. The Company could have operated in this way without the CEO and CFO knowing. But that is not, on balance, the more persuasive inference.
Two quick notes. First, the Defendants argue that scienter is less plausible here because there are no sufficient allegations as to particular motives (created by, say, stock options) to inflate revenue and enrollments. See Motion to Dismiss at 12-14. A showing of motive can help to establish scienter. See Rahman, 736 F.3d at 245-46. But it is not necessary in general, see id.; Tellabs, Inc., 551 U.S. at 325; Avaya, 564 F.3d at 277, and it is not necessary here --- because the inference of scienter is especially strong. Second, the Defendants do not develop any arguments about timing and scienter, other than as to allegations of motive. The Defendants do not, for example, seek to contend that a particular statement was not made with scienter because the strongest allegations of scienter post-date the statement.
V. Loss Causation
Where things stand: the Court has held that the Plaintiff has adequately alleged that various public statements were false, see Part III, and were made with scienter. See Part IV.
The Defendants' final argument is that the Plaintiffs have not properly pled loss causation. See Motion to Dismiss at 17-21.
That argument is taken up here.
A. In General
To make out a 10(b) claim, such as is pressed here in Count One, a plaintiff must properly plead loss causation. See 15 U.S.C. § 78u-4(b)(4); Halliburton, 573 U.S. at 267; Prudential Fin., Inc., 70 F.4th at 679; McCabe v. Ernst & Young, LLP, 494 F.3d 418, 424 (3d Cir. 2007).
At this stage of the proceedings, this is typically a matter of alleging (a) that the relevant company's share price “significantly” fell, Dura Pharmas., Inc. v. Broudo, 544 U.S. 336, 347 (2005), and (b) that the price fall was caused to a sufficient extent by information becoming known that revealed the truth about something the company had previously said (or failed to say). See id.; see also, e.g., In re DVI, Inc. Sec. Litig., 639 F.3d 623, 632 (3d Cir. 2011) abrogated on alternative grounds by Amgen Inc. v. Connecticut Ret. Plans & Tr. Funds, 568 U.S. 455 (2013); McCabe, 494 F.3d at 425, 428-29; Bouriz v. Carnegie Mellon Univ., 585 F.3d 765, 774 n.6 (3d Cir. 2009); In re Nektar Therapeutics Sec. Litig., 34 F.4th 828, 838 (9th Cir. 2022); In re Williams Sec. Litig. WCG Subclass, 558 F.3d 1130, 1137-1139 (10th Cir. 2009); Hull v. Glob. Digit. Sols., Inc., 2017 WL 6493148, at *15 (D.N.J. Dec. 19, 2017); Nat'l Junior Baseball League v. Pharmanet Dev. Grp. Inc., 720 F.Supp.2d 517, 561 (D.N.J. 2010).
Prong (b) is focused on the making of what is often called a “corrective disclosure.” See In re DVI, Inc. Sec. Litig., 639 F.3d at 638-39 abrogated on alternative grounds by Amgen Inc., 568 U.S. 455; In re Bofl Holdings, Inc. Sec. Litig., 977 F.3d at 789-91; Meyer v. Green, 710 F.3d 1189, 1196-97 (11th Cir. 2013); Public Emps. Ret. Sys. of Miss. v. Amedisys, Inc., 769 F.3d 313, 320-21 (5th Cir. 2014); Katyle v. Penn Nat'l Gaming, Inc., 637 F.3d 462, 470-73 (4th Cir. 2011); FindWhat Inv. Grp. v. FindWhat.com, 658 F.3d 1282, 1312 n.28 (11th Cir. 2011); In re Williams Sec. Litig. WCG Subclass, 558 F.3d at 1140.
The public “disclosure” of information “correct[s]” the earlier misrepresentation --- and the plaintiff then seeks to establish that the publicly-disclosed information also “correct[ed]” the stock price, pulling it back down toward what it would have been without the misrepresentation. See Meyer, 710 F.3d at 1195-96; FindWhat.com, 658 F.3d at 1314-1316.
How is the first correction (of information) yoked to the second correction (of price)? By the overarching idea that in a market economy such as ours, prices adjust up and down based on information --- and fairly quickly, in the sort of “efficient” market that most U.S. stocks trade on. See generally Basic, Inc. v. Levinson, 485 U.S. 224, 243-249 (1988); In re Merck & Co., Inc. Sec. Litig., 432 F.3d 261, 269 (3d Cir. 2005); Eugene F. Fama, Efficient Capital Market: A Review of Theory & Empirical Work, 25 J. of Fin. 383 (1970).
An example as to how this might work in practice: if a stock's price was $10 just after a statement was made, and the price falls to $6 soon after the statement is revealed to have been false (through a “corrective disclosure”) --- then $4 might be thought of as a rough starting point for thinking about the loss caused to shareholders by the statement that is now understood to have been false.
But once the analysis begins in earnest, any number of difficult possible proof problems might emerge.
There may be some conceptual issues, too. See Donald C. Langevoort, Compared to What?: Economic Evidence and the Counterfactual Difficulty, 35 J. Corp. L. 183, 184-86 (2009). But those are irrelevant here, given the arguments the parties make.
For example, $4 may overstate or understate the relevant loss -because it reflects the impact not just the corrective disclosure, but also the possible impact of other contemporaneous news. See In re DVI, Inc. Sec. Litig., 639 F.3d at 632; Meyer, 710 F.3d at 1195-96.
Or $4 might reflect not just the “price” of what is now revealed as having been a company misrepresentation --- but also the “price” of the company being revealed as the sort of company that makes misrepresentations. See Jonathan M. Karpoff, et al., The Cost to Firms of Cooking the Books, 43 J. of Fin. & Quantitative Analysis 581 (2008); Donald C. Langevoort, Basic at Twenty: Rethinking Fraud-on-the-Market, 2009 Wis. L. Rev. 151, 160 n.40 (2009).
Or another possible difficulty: perhaps the new information is not so new at all. Through a variety of other sources, the truth might have already publicly come out --- and that truth was already reflected in the stock's price by the time the asserted corrective disclosure was made. See Meyer, 710 F.3d at 1197-98; FindWhat.com, 658 F.3d at 1312 n.28, 1315-16; In re Bofl Holdings, Inc. Sec. Litig., 977 F.3d at 794; Norfolk Cnty. Ret. Sys. v. Cmty. Health Sys., Inc., 877 F.3d 687, 695 (6th Cir. 2017). In such a scenario, it does not meaningfully matter that the stock price fell around the time of the asserted corrective disclosure --- because in fact the corrective disclosure was not that. The relevant information had already come out, and its impact had presumably already been “impounded” in the price of the stock.
Imagine a company that publicly announces on a Tuesday morning that it is restating its earnings because of a previous accounting error. A week later, a newspaper article covers the restatement. If it does not provide new information, the newspaper article cannot count as corrective disclosure. This approach has sometimes been criticized, see Langevoort, Basic at Twenty, 2009 Wis. L. Rev. at 175-76 (collecting sources), but there is little doubt that it generally reflects federal law. See In re Merck & Co., Inc. Sec. Litig., 432 F.3d at 270.
B. In This Case
Here, the Defendants press one argument --- and it is a variant on the it-was-already-public argument set out in the preceding paragraph.
To see the Defendants' argument, note first that the Plaintiffs identify a range of corrective disclosures. See Complaint ¶¶ 284, 287, 291, 293. Of these, two are short seller reports about the Company--- and these two that are the focus of the Court's analysis here. See id.
Three of these asserted corrective disclosures, see Complaint ¶¶ 281, 289, 295, will not be considered here. This is because Judge Salas has already ruled that those do not count. That Court treats that decision as controlling here, under the law of the case doctrine. See Arizona, 460 U.S. at 618; Saint-Jean v. Palisades Interstate Park Comm'n, 49 F.4th 830, 836 (3d Cir. 2022); Bailey, 435 Fed.Appx. at 91. As to these disclosures, the Plaintiffs have not meaningfully tried to explain why the prior decision of Judge Salas should not control. And the allegations as to these three corrective disclosures are the same, or virtually the same --- tweaked in ways that do not address the points Judge Salas laid out. In addition, the Court also does not consider here as corrective disclosures the August 9 Report or the announcement of an SEC investigation. See Complaint ¶¶ 291, 293. Analyzing these here is unnecessary. See footnote 47.
The most familiar sort of corrective disclosure comes from a company itself. For example, a company might disclose that its prior financial statements contained material misrepresentations, because its accounting practices had been deficient. See, e.g., Hacker, 687 F.Supp.3d at 587. But a corrective disclosure might also come from a third-party source. See, e.g., Metzler Inv. GMBH v. Corinthian Colls., Inc., 540 F.3d 1049, 1063-64 (9th Cir. 2008) (newspaper article); Norfolk Cnty. Ret. Sys., 877 F.3d at 695 (lawsuit complaint); Amedisys, Inc., 769 F.3d at 322 (short seller report, newspaper article, and announcement of SEC investigation). Some courts have held that special standards need to be applied where the third-party making a corrective disclosure is a short seller. See In re Nektar Therapeutics Sec. Litig., 34 F.4th at 839-840. But while the Defendants allude to the identified corrective disclosures as “self-interested” short seller reports, Motion to Dismiss at 20, they do not argue that special standards should be applied to them.
The Defendants argue: the reports do not count as corrective disclosures, because they told the public what the public already knew, from a previous short seller report. See Motion to Dismiss at 19-21. (That previous short seller report, dated February 25, 2020, has been referred to in this Opinion as Report-3. In this Part, for ease of reference, it is called “the First Report.”)
The Defendants' argument is not persuasive.
The First Report accused the Company of faking student enrollments. See First Report at 23-26. But it did so largely based on a quick citation to an article that was, at that point, around five years old. See id. at 23. This is threadbare. It is less a disclosure of information than an accusation.
The First Report mustered one other meaningful piece of information: it said that in late 2019 a person saw 6 incidents of “students” offering “identical” reviews as to Company courses. Id. at 26.
This is potentially important information. But it is also anecdotal, and the First Report makes no effort to explain how or why this might be part of a larger pattern of false student enrollments.
All of this is in stark contrast with the information as to fake student enrollments and their impact on Company revenues set out in the other two asserted corrective disclosures the Court focuses on.
These two have been discussed in this Opinion as Report-1 and Report-2.
To see the point, zero in on one of the asserted corrective disclosures --- the May 2020 report, Report-1. Report-1 did not look to 6 students, as the First Report did. Rather, Report-1 said that it accessed and then analyzed over 460,000 class sign-in records for over 54,000 users over 200 Company classes. See Report-1 at 14.
In addition, Report-1 described its methods for determining whether a “student” user was in fact a bot. See id. at 13, 1725. It laid out in numerous charts granular information (in places to three decimal points) as to who was joining Company classes and when --- and why and to what extent those “students” appeared to be bots. See id. at 5-10, 14.
Moreover, the body of information scrutinized by Report-1 was not, in the main, public information. Indeed, the critical underlying records were apparently those that became available to paying students in Company classes, see id. at 17-20 --- plus information assertedly provided by a former Company manager. See id. at 11-12.
In addition, Report-1 showed its work to a meaningful extent, so that readers could make their own assessments. Report-1 explained how to access the underlying records it based its analysis on. See id. at 17-20. And it included an “uninterrupted segment” of the transcript of an interview with the former Company manager, describing in some detail the Company's asserted efforts to make systematic use of fake students. See id. at 11-12.
Examples from Report-1 could be multiplied. And so could examples from at least one of the other reports (Report-2, see footnotes 44, 46) identified by the Plaintiffs as a corrective disclosure.
But that is not necessary. The Plaintiffs were required to plausibly allege loss causation based on the corrective disclosures they identified. See In re Facebook, Inc. Sec. Litig., 87 F.4th 934, 954 (9th Cir. 2023); Loreley Fin. (Jersey) No. 3 Ltd. v. Wells Fargo Sec., LLC, 797 F.3d 160, 188 (2d Cir. 2015); Lormand v. US Unwired, Inc., 565 F.3d 228, 267 (5th Cir. 2009).
The Defendants contend they failed to do so, because the cited reports were no more than “repackaging” of the First Report. But even a quick read of Report-1 and Report-2 makes this argument impossible to sustain.
The information and analysis in those reports was, by comparison with the information in the First Report, more detailed, more systematic, more rigorous, and closer to first-hand --- and each of these by leaps and bounds. The reports prepared after the First Report “required extensive and tedious research involving the analysis of far-flung bits and pieces of data” --- and Report-1 and Report-2 are based mainly on non-public information. In re Nektar Therapeutics Sec. Litig., 34 F.4th at 839 (quoting In re Bofl Holding, Inc. Sec. Litig., 877 F.3d at 797). These are critical distinctions.
There can be no argument, at least at this early stage, that Report-1 and Report-2 added too little incremental information to the mix, beyond the First Report such that by the time; these reports were released, the information in them was old hat as a matter of law, and had already been impounded into the Company's stock price, absorbed and reflected from the time the First Report came out.
Loss causation has been plausibly pled. See id.; Lea v. Tai Educ. Grp., 837 Fed.Appx. 20, 28 (2d Cir. 2020).
VI. Conclusion
For the reasons set out above, the motion to dismiss is denied, except as to certain statements enumerated in Part III.C, as to which it is granted.
Two points. First, the Court's analysis has focused solely on the Plaintiffs' Section 10(b) claim. This is because the Defendants seek to dismiss the Plaintiffs' other claim, under Section 20(a) of the Exchange Act of 1934, see Part II.B.2, solely on the basis that no claim has been stated under Section 10(b). See Motion to Dismiss at 40 n.23. Given that this is the Defendants' argument, the motion to dismiss the Section 20(a) claim is granted and denied to the same extent as the Section 10(b) claim. Second, the Plaintiffs press certain arguments as to marketing expenses, and two additional corrective disclosures. Given the Court's resolution of the other theories of misrepresentation and loss causation in the Plaintiffs' favor, those are irrelevant for now and are not taken up.