Siddharth Hariharan, et al v. Adobe Systems, Inc., et al

Filing 16

Filed (ECF) Petitioner Google Inc. reply to answer to 23f petition. Date of service: 11/25/2013. [8878284]--[COURT UPDATE: Edited docket text to reflect correct ECF event. Resent NDA. 11/26/2013 by RY] (DMF)

Case No. 13-80223
UNITED STATES COURT OF APPEALS
FOR THE NINTH CIRCUIT

IN RE HIGH-TECH EMPLOYEE ANTITRUST LITIGATION

Petition for permission to appeal from the United States District Court
Northern District of California
The Honorable Lucy H. Koh, Presiding
Case No. 5:11-2509-LHK

REPLY IN SUPPORT OF PETITION FOR LEAVE TO APPEAL
UNDER FEDERAL RULE OF CIVIL PROCEDURE 23(F)
ROBERT A. VAN NEST, #84065
DANIEL PURCELL, #191424
EUGENE M. PAIGE, #202849
JUSTINA SESSIONS, #270914
KEKER & VAN NEST LLP
633 Battery Street
San Francisco, CA 94111-1809
Telephone: [phone]
Facsimile: 415 397 7188
Attorneys for Defendant and Petitioner
Google Inc.

Additional counsel listed on signature
page

TABLE OF CONTENTS
Page
I.

The Importance Of These Unsettled Issues Is Clear And Undisputed. ........... 3

II.

The District Court Approved An Averaging And Aggregation Approach That Was Explicitly Designed To Gloss Over Individualized
Issues................................................................................................................ 5

III.

The Rules Enabling Act And Due Process Violations Are Significant
And Fully Preserved. ....................................................................................... 9

CONCLUSION ........................................................................................................ 10

i

TABLE OF AUTHORITIES
Cases

Page(s)

Amgen Inc. v. Conn. Retirement Plans & Trust Funds,
133 S. Ct. 1184 (2013)................................................................................. 3, 4
Comcast Corp. v. Behrend, 133 S. Ct. 1426 (2013) ......................................1, 2, 4, 7
Ellis v. Costco, 657 F.3d 970 (9th Cir. 2011) ............................................................ 9
In re Graphics Processing Units Antitrust Litig.,
253 F.R.D. 478 (N.D. Cal. 2008) .................................................................... 3
In re Rail Freight Fuel Surcharge Antitrust Litig.,
725 F.3d 244 (D.C. Cir. 2013) ....................................................................... 10
Leyva v. Medline Industries, 716 F.3d 510 (9th Cir. 2013) .................................5, 10
Mazza v. Am. Honda Motor Co., 666 F.3d 581 (9th Cir. 2012) .............................. 10
Paige v. California, 291 F.32d 1141 (9th Cir. 2002) ................................................ 7
United States v. Williams, 504 U.S. 36 (1992) .......................................................... 9
Wal-Mart Stores, Inc. v. Dukes, 131 S. Ct. 2541 (2011) ..................................passim
Statutes and Rules

Page(s)

Fed. R. Civ. P. 23 ................................................................................................... 2, 7
Fed. R. Civ. P. 23(b)(2).............................................................................................. 9
Other Authorities

Page(s)

Richard A. Nagareda, Class Certification in the Age of Aggregate Proof,
84 N.Y.U. L. Rev. 97 (2009) ........................................................................... 3

ii

Plaintiffs avoid grappling with the unsettled issues of class-certification law
that the district court acknowledged. Instead, plaintiffs quote liberally from the certification order as if its mere existence—and volume—validated its reasoning.
They recite evidence that no-cold-call agreements existed and may have affected a
few of the 60,000 class members in some of the 2400 job categories. But plaintiffs
have no meaningful answer to the reasons why leave to appeal should be granted.
Plaintiffs do not dispute that this class cannot be certified unless antitrust
impact is a common issue, because determining impact individually would defeat
predominance. Nor do they dispute that the district court endorsed averaging as a
stand-in for a common method to prove impact and damages for each class member. Averaging ignores individual differences among class members, however, and
cannot determine whether all, or only some, were injured. Thus, an average result
does not “resolve an issue that is central to the validity of each one of the claims in
one stroke.” Wal-Mart Stores, Inc. v. Dukes, 131 S. Ct. 2541, 2551 (2011).
Contrary to plaintiffs’ contention, manifest error is not the “only” identified
basis for the petition (Resp. 4). Plaintiffs have invoked no controlling authority after Dukes and Comcast that permits the nimble use of aggregation and averaging
methods, like those deployed here, that assume the common impact that they must
have a common means of proving. And the citations plaintiffs do provide show
that—as the petition explained (at 3, 8-10, 20 & n.6)—these fundamental issues of

1

class-action law are unsettled in this Circuit after Dukes and Comcast.
Plaintiffs insist that these high-technology defendants used rigid pay structures reminiscent of the civil service. They invent a Department of Justice finding
(absent from SER687-89) that the no-cold-call agreements reduced employee
compensation. But plaintiffs do not address the absurdity of their expert’s results
in light of the minuscule (and unchanged) hiring between defendants before, during, and after the agreements. For example, Intel’s only agreement was with
Google, which had a fundamentally different business and little demand for Intel
employees: in 12 years (2001-2012), only 102 Intel employees left for Google,
most of them during the class period. See ER1095(291). Yet plaintiffs’ model purports to show that limiting cold calls from this single, unfruitful source injured
each of Intel’s 30,000 employees in exactly the same way, resulting in an undifferentiated damages estimate of hundreds of millions of dollars. Plaintiffs cannot explain (much less measure) how a restraint on one recruiting method by a sliver of
the market could affect every member of this widely diverse class in the same way.
The certification order approves the use of the class device when there is no
possibility that class litigation will result in accurate and reliable determination of
each class member’s right to relief. That approach makes a mockery of the commonality and predominance standards of Rule 23 and flatly violates the limitations
of the Rules Enabling Act. Review by this Court is therefore warranted.

2

I.

The Importance Of These Unsettled Issues Is Clear And Undisputed.
Plaintiffs do not dispute that the district court’s approach, if allowed to

stand, would bring about a sea-change in the law of class actions: any class could
be certified so long as it contains enough injured persons to show some average effect. This shortcut to class certification could be repurposed for every area of law.
See Richard A. Nagareda, Class Certification in the Age of Aggregate Proof, 84
N.Y.U. L. Rev. 97, 103 (2009) (repeatedly cited with approval in Dukes).
A. Plaintiffs cannot dispute that the district court was correct to label “the
legal standard with respect to the predominance inquiry” as “somewhat unsettled”
and “not altogether clear.” ER822(19). Indeed, the court acknowledged that district
courts across the country are divided over class certification in wage-suppression
antitrust cases. See id. n.7. But neither the district court nor plaintiffs acknowledge
the order’s square conflict with Judge Alsup’s rejection of averaging as a stand-in
for common proof. See In re Graphics Processing Units Antitrust Litig., 253
F.R.D. 478, 494 (N.D. Cal. 2008).
The district court remarked that “intervening authority such as Amgen [Inc.
v. Conn. Ret. Plans & Trust Funds, 133 S. Ct. 1184 (2013),] has refined the predominance inquiry since the cases cited by Defendants were decided,” ER822(19)
n.7, but that misapprehension of Amgen highlights the need for clarification by this
Court. Taking their cue from the district court and the Seventh Circuit (ER822-

3

26(19-22)), plaintiffs suggest that Amgen relieves them of their duty under Dukes
to show that their averaging and aggregation methods can resolve the claim of each
class member. But Amgen did not silently alter the Dukes standard or relieve class
proponents from their obligation to demonstrate that the aspects of a claim that can
be proved through common evidence predominate over those that cannot.
Amgen addressed a unique presumption that all buyers of a security traded in
an efficient market have relied on any material misrepresentation. See 133 S. Ct. at
1195-96. But Amgen does not mean that every issue becomes a merits issue whenever the plaintiffs average the evidence. The questions of antitrust impact and
damages can have different answers for different plaintiffs.1 As Comcast confirms,
Amgen does not relieve plaintiffs of the need to demonstrate a means of commonly
proving those points. See Comcast Corp. v. Behrend, 133 S. Ct. 1426, 1431 (2013)
(antitrust impact); id. at 1433 (damages).
The confusion and disarray in the lower courts over the interplay among
Comcast, Dukes, and Amgen, and the limited guidance from this Court, reinforce
the need for review. Among the issues requiring clarification is Dukes’ rule that
common questions under Rule 23 are those that generate a “common answer” to a
“crucial question.” 131 S. Ct. at 2552. In a passage that plaintiffs try to obscure

1

Here, for example, some employees may have missed concrete opportunities as a
result of the challenged agreements. Contrary to the district court’s conclusion,
their cases would not “rise[] and fall[] with their common evidence.” ER888(85).
4

(Resp. 4-5), the district court characterized “the critical question” as “whether
common questions predominate over individual questions—not whether plaintiffs
could show common answers to those questions.” ER825(22). That misapprehension likely underlies the court’s confusion of averages with common answers.
B. Comcast underscores the importance of rigorously scrutinizing proposed
means of measuring individual class members’ damages through common evidence. Unlike Leyva v. Medline Industries, 716 F.3d 510, 514 (9th Cir. 2013),
where individual damages could be calculated using arithmetic and records in the
defendant’s “computerized payroll and time-keeping database,” plaintiffs’ expert
here conceded that his model could not measure any individual class member’s
damages. Dist. Ct. Dkt. 308-1, at 23-24. Plaintiffs propose instead to produce only
an aggregate sum impact. The district court’s approval of that method is flatly inconsistent with Comcast, and reinforces the need for this Court’s review.
II.

The District Court Approved An Averaging And Aggregation Approach
That Was Explicitly Designed To Gloss Over Individualized Issues.
A. The district court approved plaintiffs’ blending of class members into a

single equation to make dispositive individualized issues magically disappear. By
aggregating data across all defendants, plaintiffs’ approach assumes the common
answers it purports to prove. It assumes that the impact on employees of a company that entered into one agreement (as did four defendants) was the same (adjusted
for age, seniority, and such) as the impact on employees of a company that entered

5

into two (Pixar) or three (Apple and Google). See Dist. Ct. Dkt. 65 ¶ 108. It assumes a film company and a computer chip maker would experience the same impact from different single agreements. It assumes that a safety engineer would suffer the same impact as a web designer. It assumes that employees who might not
have received retention incentives because of lower performance or skills would be
affected the same as top performers and mission-critical employees. And the common impact factor uses compensation actions by any one defendant to determine
undercompensation with regard to all class members.
Plaintiffs admit that they designed their aggregation and averaging models to
work only at the collective level because “the individual data is likely to be dominated by forces that operate at the individual level.” ER1151(347); see Pet. 15. Put
another way, the inherently individualized nature of employee compensation made
it impossible to devise a means of commonly proving impact as Dukes requires.
Plaintiffs’ expert, Dr. Leamer, admitted that “inherent noise in the individual level
data” would “drown out” the supposedly rigid “internal pay structures” on which
plaintiffs’ entire theory depends. ER874(71).2 Plaintiffs now boast that, by combining data for disparate class members, “the idiosyncratic parts get averaged out.”
2

In defending the district court’s disregard for the collapse of Dr. Leamer’s model
when it was disaggregated even at the company level (Resp. 14), plaintiffs forget
Leamer’s admission that he concealed his own results from disaggregating his
model by defendant. Dist. Ct. Dkt. 308-1, at 360-361. No wonder: disaggregation
altered the sign of the effect, absurdly indicating that the agreements caused overcompensation at some employers. ER1010-11(206-07).
6

Resp. 16. But Rule 23 does not permit a method of supposed common proof that
does not explain and account for each class member’s situation, but instead simply
averages out the differences. Guidance from this Court is needed.
The district court recognized that the method it approved “may have masked
some of the individual variations within each job title.” ER874(71). Under a proper
analysis (and in a typical putative employee class comprising just one or a few job
titles), individual variations within a job title are sufficient to preclude certification.
See Pet. 11 n.4. Here the in-title variations are multiplied 2400 times over.
B. Wishing away Comcast and Dukes, plaintiffs quote this Court’s 2002
holding—outside the class certification context—that “it is a generally accepted
principle that aggregated statistical data may be used where it is more probative
than subdivided data.” Paige v. California, 291 F.3d 1141, 1148 (9th Cir. 2002)
(quoted in Resp. 16). Paige permitted the aggregation of statistics from multiple
examination sittings to assess whether the test had a discriminatory impact on a racial group. Id. at 1144. That does not suggest that aggregate statistics can prove individual injury—especially in light of the holding in Dukes that the plaintiff must
demonstrate “that the class members have suffered the same injury.” 131 S. Ct. at
2551 (emphasis added; internal quotation marks omitted). And Comcast has made
clear that any proposed methodology for presenting classwide proof must in fact be
susceptible to classwide application. 133 S. Ct. at 1433.

7

C. Plaintiffs have nothing to say about the undisputed evidence refuting the
critical premise of the certification order—that defendants’ “company-wide compensation structures … left little scope for individual variation.” ER860(57). The
compensation range within a single job classification routinely exceeded 50% of
the low salary (e.g., ER1454), and the compensation trajectories of individuals
within a single job classification were literally all over the map. See ER1250(472);
Pet. 16. Plaintiffs ultimately insist that the mere existence of compensation structures and overall compensation budgets—i.e., of nonrandom compensation—is
enough to permit average effects to substitute for common proof of impact and
damages.3 Plaintiffs also point to Google’s across-the-board change in compensation structure—nearly two years after the class period—in response to employee
surveys indicating a strong preference for base salary over bonus and equity.
ER1488-89. That change occurred in a setting where startups were actually hiring
Google employees at a high rate, not merely cold calling them. Google lost more
employees to Facebook alone in a single year (2010) than it lost to all defendants
from 2005 to the present—and it lost none to other defendants before 2005. Compare SER865 n.69 with ER1095. Nothing suggests that any defendant would have
implemented across-the-board increases in response to cold calls from companies
3

What the district court cited as evidence that any decrease in cold-calling would
affect all 60,000 class members (ER841-42(38-39)) reflected only counteroffers
for a few “targeted” top performers and a general effort to pay employees well
enough so that the best ones would not leave.
8

that were hiring few of its employees.
III.

The Rules Enabling Act And Due Process Violations Are Significant And
Fully Preserved.
Plaintiffs contend (Resp. 18) that defendants failed to preserve a challenge

under the Rules Enabling Act and Due Process Clause. Not so. This issue was
pressed (SER716) and passed upon below (ER 889(86)). Cf. United States v. Williams, 504 U.S. 36, 41 (1992).
Plaintiffs do not defend the district court’s manifestly erroneous holding that
the Rules Enabling Act’s strictures apply only to Rule 23(b)(2) classes. Plaintiffs
instead insist that the foreclosed defenses here are no more individualized than in
any other case. Resp. 19-20. But use of undifferentiated averages and aggregate data to resolve whether class members were injured deprives defendants of any opportunity to show that individuals (e.g., low-performers, or those whose skills had
no interest for a party subject to a no-cold-call agreement) were not injured at all.
Citing Seventh Circuit cases, plaintiffs maintain that “the possibility that a
few class members might not have been injured does not preclude class certification in antitrust cases.” Resp. 19. But both the Supreme Court and this Court have
held otherwise: class members must “have suffered the same injury.” Dukes, 131
S. Ct. at 2551 (internal quotation marks omitted; emphasis added); Ellis v. Costco,
657 F.3d 970, 981 (9th Cir. 2011) (quoting Dukes, 131 S. Ct. at 2552). Where one
set of class members—in this case at most a few, if any—has suffered an antitrust

9

injury and another set has not, the class members have not suffered “the same injury.” To proceed as a class, plaintiffs must “prove, through common evidence, that
all class members were in fact injured by the alleged conspiracy”; averaging together some class members who were injured with others who were not does not
suffice. In re Rail Freight Fuel Surcharge Antitrust Litig., 725 F.3d 244, 252 (D.C.
Cir. 2013) (emphasis added). And, regardless of the Seventh Circuit’s view, this
Court recognizes that “no class may be certified that contains members lacking Article III standing.” Mazza v. Am. Honda Motor Co., 666 F.3d 581, 594 (9th Cir.
2012).
The district court’s embrace of averaged and aggregated data acts as a procedural device to deprive defendants of their substantive defenses, resulting literally in the “Trial by Formula” that Dukes rejected, 131 S. Ct. at 2561. None of this
implies a “ban on math” or that “a class can never be certified for damages purposes.” Resp. 19-20. Damages often may be determined by applying a common method to each member’s individual circumstances. E.g., Leyva, 716 F.3d at 514. But
no class can be certified under an approach that obliterates individualized issues to
sweep in a substantial number of plaintiffs whose injury is doubtful, individualized, and not susceptible to determination through common evidence.
CONCLUSION
The petition should be granted and the certification order reversed.

10

KEKER & VAN NEST LLP

Dated: November 25, 2013
By:

/s/ Robert A. Van Nest
Robert A. Van Nest
Daniel Purcell
Eugene M. Paige
Justina Sessions
633 Battery Street
San Francisco, CA 94111
Telephone: [phone]
Facsimile: [phone]
Edward D. Johnson
Lee H. Rubin
Donald M. Falk
MAYER BROWN LLP
Two Palo Alto Square
3000 El Camino Real, Suite 300
Palo Alto, CA 94306-2112
Telephone: [phone]
Facsimile: [phone]
Attorneys for Defendant and Petitioner
GOOGLE INC.

Dated: November 25, 2013

O’MELVENY & MYERS LLP

By: /s/ Michael F. Tubach
Michael F. Tubach
George Riley
Christina J. Brown
Two Embarcadero Center, 28th Floor
San Francisco, CA 94111
Telephone: [phone]
Facsimile: [phone]
Attorneys For Defendant and Petitioner APPLE INC.

11

Dated: November 25, 2013

JONES DAY

By: /s/ David C. Kiernan
David C. Kiernan
Robert A. Mittelstaedt
Craig A. Waldman
555 California Street, 26th Floor
San Francisco, CA 94104
Telephone: [phone]
Facsimile: [phone]
Attorneys for Defendant and Petitioner ADOBE SYSTEMS, INC.

Dated: November 25, 2013

MUNGER TOLLES & OLSON, LLP

By: /s/ Gregory P. Stone
Gregory P. Stone
Bradley S. Phillips
Gregory M. Sergi
John P. Mittelbach
355 South Grand Avenue, 35th Floor
Los Angeles, CA 90071-1560
Telephone: [phone]
Facsimile: [phone]
Attorneys for Defendant and Petitioner INTEL
CORPORATION.
ATTESTATION: The filer attests that concurrence in the filing of this document
has been obtained from all signatories.

12

Case No. 13-80223
CERTIFICATE OF SERVICE
I hereby certify that I electronically filed the foregoing with the Clerk of the
Court for the United States Court of Appeals for the Ninth Circuit by using the appellate CM/ECF system on November 25, 2013.
I certify that all participants in the case are registered CM/ECF users and
that service will be accomplished by the appellate CM/ECF system.

s/ Donald M. Falk

.