Students for Fair Admissions, Inc. v. President and Fellows of Harvard College et al

Filing 571

Opposition re 546 MOTION in Limine filed by President and Fellows of Harvard College. (Attachments: # 1 Exhibit 1, # 2 Exhibit 2, # 3 Exhibit 3, # 4 Exhibit 4, # 5 Exhibit 5, # 6 Exhibit 6, # 7 Exhibit 7, # 8 Exhibit 8, # 9 Exhibit 9, # 10 Exhibit 10)(Waxman, Seth)

UNITED STATES DISTRICT COURT FOR
THE DISTRICT OF MASSACHUSETTS
BOSTON DIVISION
STUDENTS FOR FAIR ADMISSIONS, INC.,
Plaintiff,
v.
PRESIDENT AND FELLOWS OF HARVARD
COLLEGE (HARVARD CORPORATION),

Civil Action No. 1:14-cv-14176-ADB

Defendant.

HARVARD’S OPPOSITION TO SFFA’S MOTIONS IN LIMINE

The Court’s order denying summary judgment (Dkt. 566) reflects the need for numerous
factual disputes to be resolved at trial—including, as relevant here, the dispute between
Harvard’s and SFFA’s statistical experts and disputes concerning a set of documents prepared by
individuals within Harvard’s Office of Institutional Research (“OIR”). SFFA’s motions in
limine effectively seek to foreclose the Court’s full consideration of those disputes by keeping
important evidence out of the record. The Court should deny SFFA’s motions and permit the
full airing of evidence that will be necessary for the fair adjudication this case warrants.
I.

SFFA’S MOTION NO. 1 IS MOOT
In light of the Court’s ruling on summary judgment (Dkt. 566), and SFFA’s

representation that its founder and President, Edward Blum, “has no relevant information to offer
at this trial” (Mot. 1), Harvard and SFFA have agreed to remove Mr. Blum from their respective
witness lists. SFFA’s first motion in limine, to preclude Harvard from calling Mr. Blum as a
witness, is therefore moot.
II.

SFFA’S MOTION NO. 2, TO PRECLUDE EVIDENCE OR ARGUMENT REGARDING
APPLICATION FILES AND THE PERSONAL RATING, SHOULD BE DENIED
One of the principal disagreements between the statistical experts in this case is whether

the personal rating assigned by admissions officers reflects bias against Asian-American
applicants. SFFA’s expert, Dr. Peter Arcidiacono, attempted to analyze three ratings that
applicants receive from admissions officers: the academic, extracurricular, and personal ratings.
His regression models found that Asian-American ethnicity was associated with stronger
academic and extracurricular ratings, but slightly weaker personal ratings, than the data in the
model would otherwise suggest. Dr. Arcidiacono then hypothesized—without any basis for the
obvious inconsistency—that factors outside the data explained the estimated positive effect of
Asian-American ethnicity on the academic and extracurricular ratings, but not the estimated

1

negative effect on the personal rating. Dr. Card points out the inconsistency in Dr. Arcidiacono’s
interpretations of the three models and argues that, in fact, factors outside the data explain all
three effects. As the Court recognized in denying summary judgment, the resolution of that
dispute (and many others) between Dr. Card and Dr. Arcidiacono will be a key issue for trial.
SFFA’s second motion in limine attempts to pretermit the Court’s consideration of both
experts’ views, by seeking to exclude Dr. Card’s opinion that factors outside the data—as
opposed to bias against Asian-American applicants—explain Dr. Arcidiacono’s regression
results for the personal rating. SFFA tries to justify that extraordinary request on the premise
that Dr. Card’s opinion relies on “portions of applications Harvard refused to produce,” such as
applicants’ essays and recommendation letters. That is incorrect. SFFA’s motion fails because
Dr. Card’s opinion relies not on information in undisclosed application files—which Dr. Card of
course never reviewed—but on documents, testimony, and data to which SFFA had equal access.
1.

SFFA’s motion starts with an attempt to relitigate the Court’s discovery rulings.

During discovery, SFFA sought production of 1,600 application files for the classes of 2018 and
2019. Dkt. 284; Dkt. 285. 1 The Court concluded that “[r]equiring Harvard to redact and
produce 1600 files at [that] time [was] disproportionate to the demonstrated need.” Dkt. 326 at
2. The Court instead directed Harvard to produce 80 application files of its own choosing from
the class of 2018 cycle, another 80 from the class of 2019 cycle, and 160 files chosen by SFFA
from each cycle—a total of 480 files. Id. But although it denied SFFA’s request for additional
files, the Court invited SFFA to “renew [its] request once it … reviewed the application files”

1

SFFA originally moved to compel production of 6,400 application files. Dkt. 64; Dkt.
65. The Court denied SFFA’s motion as moot with leave to renew on October 30, 2015 (Dkt.
113), and SFFA filed a renewed motion seeking production of 1,600 application files on April
10, 2017 (Dkt. 284; Dkt. 285).

2

that Harvard produced. Id. SFFA never accepted that invitation. If SFFA thought it needed
additional files, it should have availed itself of the opportunity to request them, rather than using
its dissatisfaction with the Court’s discovery rulings as an excuse for an eleventh-hour attempt to
exclude expert testimony on a key issue in this case.
2.

SFFA’s attempt to parlay its failure to request additional discovery into a basis for

excluding Dr. Card’s opinions fails because, despite SFFA’s accusations, neither Harvard nor Dr.
Card is relying on unproduced application files. Contrary to SFFA’s argument (Mot. 5), Dr.
Card is not relying on a “positive inference” from unproduced application files “to assume” that
information in those files explains applicants’ personal ratings. Rather, Dr. Card’s opinion on
this issue is that Dr. Arcidiacono’s model of the personal rating cannot support an inference that
the personal rating reflects bias against Asian-American applicants—in other words, that Dr.
Arcidiacono’s results are instead attributable to factors that cannot be included in a statistical
model. See Mot. Ex. 2 (Card Rebuttal) ¶ 7 (Dr. Arcidiacono’s model is “not capable of reliably
determining whether the personal rating is in fact ‘biased’”). That is, of course, the same
conclusion Dr. Arcidiacono reached with respect to his models of the academic and
extracurricular ratings. Dr. Card reached that conclusion for two main reasons, neither of which
had anything to do with information in unproduced application files.
First, Dr. Card explained that the personal rating reflects a great deal of information that
cannot be quantified and thus cannot be included in a statistical model. See, e.g., Ex. 1 at
271:11-275:11 (Card Dep.). Dr. Arcidiacono himself agreed that “the personal rating … is
difficult to measure directly[.]” Ex. 2 at 37 (Arcidiacono Report). Dr. Card reached that
conclusion based not on information in undisclosed application files but on sources as available
to SFFA as they are to Harvard, including extensive deposition testimony that the personal rating

3

relies on non-quantifiable information such as essays, the content of recommendation letters (not
just the numerical ratings assigned to some of those letters), and written comments from alumni
or staff interviews (not just the numerical ratings from those interviews). 2 See, e.g., Ex. 3 ¶ 147
& n.122 (Card Report); Card Rebuttal ¶ 40 & nn.30-31. 3
Second, Dr. Card corroborated his interpretation of Dr. Arcidiacono’s results by
analyzing the same data to which SFFA had access. SFFA asserts (Mot. 3) that “there is no
evidence in the record that Asian-American applicants fare more poorly” on factors that inform
the personal rating. But Dr. Card found that, on average across the entire pool of applicants,
Asian-American applicants are less strong than White applicants across factors that are
quantified and that inform the personal rating—which suggests they are also less strong across
unquantifiable factors that inform the personal rating, and thus that the results of Dr.
Arcidiacono’s personal rating regression are likely due to those unquantifiable factors. See Card
Rebuttal ¶¶ 28-30, 46-50, 52-53. Dr. Card also observed that Dr. Arcidiacono’s own analysis
shows that “non-academic variables that are important to the personal rating (like the teacher,
guidance counselor, and alumni interview ratings) shift the unexplained gap towards zero.” Id.
¶ 51. He explained that, given the trend established by Dr. Arcidiacono’s analysis, “it is likely
that more non-academic variables would have the same effect” if it were possible to quantify
such variables and thus include them in the statistical model. Id.

2

See, e.g., Ex. 4 at 164:11-165:2 (McGrath 2015 Dep.); Ex. 5 at 245:18-246:20, 248:6-24
(Fitzsimmons Dep.); Ex. 6 at 359:16-360:9 (McGrath 2017 Dep.); Ex. 7 at 80:3-18 (Banks
Dep.); Ex. 8 at 60:14-20 (Walsh Dep.); Ex. 9 at 121:2-6 (Bever Dep.).
3

Dr. Card also noted that, according to Dr. Arcidiacono’s own model, factors in the data
explain almost twice as much about the academic rating as about the personal rating—making it
“indefensible” for Dr. Arcidiacono to attribute the positive effect of Asian-American ethnicity in
his academic rating model, but not the negative effect in his personal rating model, to factors
outside the data. Card Rebuttal ¶¶ 39, 42.

4

Ignoring all of this, SFFA offers a grossly misleading account of Dr. Card’s opinions.
For example, SFFA accuses Dr. Card (Mot. 5) of having “admitted he only believes race does
not influence the personal score because Harvard’s Dean of Admissions assured him it did not.”
In reality, Dr. Card’s extensive review of documents and testimony, as well as the applications
that were produced, supports his conclusion that the results of Dr. Arcidiacono’s personal rating
regression reflected not any actual effect of race but rather the effect of factors that cannot be
captured in a statistical model.
3.

SFFA also attempts to argue (Mot. 3-4) that Dr. Card’s conclusion was legally

“impermissible.” As a threshold matter, SFFA’s arguments speak only to the weight that the
Court should afford Dr. Card’s analysis; they supply no basis to limit the scope of Dr. Card’s
testimony. In any event, SFFA misstates the law when it suggests that Dr. Card cannot criticize
Dr. Arcidiacono’s regression for its inability to account for the key, unquantifiable factors that
inform the personal rating. Far from suggesting that Dr. Card should be precluded from offering
his opinion, the Supreme Court’s decision in Bazemore v. Friday, 478 U.S. 385 (1986), on which
SFFA relies, recognized that where a regression does not control for critical factors—as is true of
Dr. Arcidiacono’s attempt to analyze the personal rating—it may be “so incomplete as to be
inadmissible as irrelevant.” Id. at 400 n.10; see also Palmer v. Shultz, 815 F.2d 84, 101 (D.C.
Cir. 1987) (“the relevance of a factor to the selection process” may be “so obvious that the
defendants, by merely pointing out its omission, can defeat the inference of discrimination
created by the plaintiffs’ statistics”). The First Circuit has likewise observed that, “[w]hile a
litigant need not include every conceivable independent variable in [a regression] analysis, the
major independent variables must be considered, … not only to render the statistical study

5

probative, but also to save it from being dismissed as irrelevant.” Wessmann v. Gittens, 160 F.3d
790, 805 n.8 (1st Cir. 1998) (citations omitted). 4
In short, Dr. Card is not relying on information in application files that the Court
determined Harvard was not required to produce, and his opinion appropriately points out crucial
defects in Dr. Arcidiacono’s analysis that make his conclusions unreliable—defects that are
central to the dispute between the experts. SFFA is of course free to cross-examine Dr. Card
regarding his opinions at trial, but it has shown no basis to exclude those opinions from the
courtroom. SFFA’s motion in limine is no more than a thinly veiled attempt to prevent the
fundamental weaknesses in its expert’s work from being exposed.
III.

SFFA’S MOTION NO. 3, TO PRECLUDE HARVARD’S FACT WITNESSES FROM
TESTIFYING ABOUT THE OIR ANALYSES, SHOULD BE DENIED
SFFA’s third motion in limine seeks to preclude Harvard from “elicit[ing] undisclosed

expert testimony” regarding documents created by employees in Harvard’s Office of Institutional
Research (“OIR”). Mot. 6. SFFA mischaracterizes both the record and the law and seeks onesided relief that would allow it to invoke the OIR documents while stifling Harvard’s response.
The motion should be denied.
SFFA’s motion rests largely on the premise that, “[d]uring his deposition, Dean [Rakesh]
Khurana held himself out as having expertise in logit regressions by offering off-the-cuff

4

See also Cooper v. Southern Co., 390 F.3d 695, 719 (11th Cir. 2004) (“Bazemore … did
not hold that any statistical report, no matter how many critical variables were missing, should
automatically present a jury question, indicating that there are ‘some regressions so incomplete
as to be inadmissible as irrelevant.’”); Koger v. Reno, 98 F.3d 631, 637 (D.C. Cir. 1996) (“Courts
have not … understood Bazemore to require acceptance of regressions from which clearly major
variables have been omitted[.]”); Daniel L. Rubinfeld, Reference Guide on Multiple Regression,
in Federal Judicial Center, Reference Manual on Scientific Evidence 303, 314 (3d ed. 2011)
(“Failure to include a major explanatory variable that is correlated with the variable of interest in
a regression model may cause an included variable to be credited with an effect that actually is
caused by the excluded variable.”).

6

criticisms of OIR’s” supposed “conclusion.” 5 Mot. 8. But Dean Khurana was not presenting
himself as an expert witness; he was explaining his actions as a percipient witness. SFFA’s
counsel asked Dean Khurana a series of questions about whether and when he had seen the OIR
analyses, how he had reacted to them, and why—questions that he answered relying on his own
background and experiences. For example, when Dean Khurana testified that he recalled seeing
OIR tables like those shown to him at his deposition, which he described as “an examination of
… the admittance of students,” the following colloquy ensued:
Q.
…
A.
Q.
A.

… And what do you remember about this examination?
Just that I didn’t think that this was – analysis was done appropriately.
And why didn’t you think that?
There’s limitations, a lot of limitations to doing what are called fitted models like
this.

Ex. 10 (Khurana Dep.) at 253:13:20. A few minutes later, that discussion continued:
Q.

A.

When you say there are a lot of problems with a fit model such as these, what
specific problems, if any, do you recall thinking about when you looked at tables
like the ones in this report?
That models like this violate statistical and hypothesis testing.

Id. at 256:9-14. In other words, Dean Khurana commented on the OIR analyses in the context of
answering questions about how he had reacted to them and why. There is nothing impermissible
about this testimony.
SFFA’s summary judgment briefs focused heavily on how Harvard officials,
including Dean Khurana and Dean Fitzsimmons, reacted to the OIR analyses—in Dean
Fitzsimmons’s case, in 2013; and in Dean Khurana’s case, when he was shown some of
the analyses in 2014 upon assuming the deanship. See Pltf.’s Mem. in Support of its Mot.

5

In actuality, OIR did not reach any “conclusion” that Harvard was discriminating against
Asian-American applicants; nor did OIR even seek to address that question. See SJ Mem. 38;
Def.’s Mem. in Opp. to Pltf.’s Mot. for S.J. (Dkt. 435) 19-23; SJ Reply 4-5.

7

for S.J. (Dkt. 413) 15-18; Pltf.’s Mem. in Opp. to Harvard’s Mot. for S.J. (Dkt. 449) 1011; Pltf.’s Reply Mem. in Support of Mot. for S.J. (Dkt. 510) 11-12; SJ Reply 5. To the
extent SFFA presses a similar narrative at trial, arguing that the reactions somehow show
Harvard’s intent to discriminate against Asian-American applicants, Harvard’s witnesses
should be permitted to respond—in answering questions from either SFFA’s counsel or
Harvard’s—just as Dean Khurana did at his deposition: by discussing their personal
impressions of the OIR analyses’ limitations and how those impressions affected their
reactions.
To the extent SFFA is arguing that Harvard’s fact witnesses cannot be permitted
to discuss scientific or technical concepts in conveying their reactions to the OIR
analyses, that is incorrect. As the First Circuit has explained, “Rule 26 uses the term
expert ‘to refer to those persons who will testify under Rule 702 of the Federal Rules of
Evidence with respect to scientific, technical, and other specialized matters.’” Gomez v.
Rivera Rodriguez, 344 F.3d 103, 113 (1st Cir. 2003). “That definition does not
encompass a percipient witness who happens to be an expert.” Id. Thus, “a party need
not identify a witness as an expert so long as the witness played a personal role in the
unfolding of the events at issue and the anticipated questioning seeks only to elicit the
witness’s knowledge of those events.” Id. at 113-114. That is the capacity in which
Dean Khurana and other fact witnesses would testify about their reactions to the OIR
analyses.
SFFA’s authorities are not to the contrary. First, SFFA quotes Federal Rule of
Evidence 701 as prohibiting fact witnesses from testifying “based on scientific, technical,
or other specialized knowledge”—but the rule actually refers to “scientific, technical, or

8

other specialized knowledge within the scope of Rule 702.” Fed. R. Evid. 701(c)
(emphasis added). SFFA omits the italicized phrase. As the First Circuit held in Gomez,
a fact witness’s testimony about events in which he participated is not testimony “within
the scope of Rule 702,” even if it reflects the witness’s expertise. Second, SFFA cites
Perez-Perez v. Popular Leasing Rental, Inc., 993 F.2d 281 (1st Cir. 1993), which stands
for the basic proposition that expert testimony must be disclosed. But the testimony at
issue in Perez-Perez was squarely within the scope of Rule 702; it was based on the
witness’s medical expertise and his “review of [defendant]’s medical records” after
litigation began. Id. at 286. That is not true here.
If SFFA wishes to argue that Harvard officials’ reactions to the OIR analyses
somehow show Harvard’s intent to discriminate against Asian-American applicants, then
Harvard’s fact witnesses should be permitted to respond by answering questions—from
either SFFA’s counsel or Harvard’s—about their personal experiences with the OIR
analyses, including the bases for their reactions. Alternatively, if Harvard is precluded
from eliciting any fact witness testimony regarding the OIR analyses, then the Court
should preclude both parties from eliciting testimony regarding the OIR analyses. The
Court should reject SFFA’s request to present its side of the story without response.

9

Respectfully submitted,
/s/ Seth P. Waxman
Seth P. Waxman (pro hac vice)
Paul R.Q. Wolfson (pro hac vice)
Danielle Conley (pro hac vice)
Brittany Amadi (pro hac vice)
Daniel Winik (pro hac vice)
WILMER CUTLER PICKERING
HALE AND DORR LLP
1875 Pennsylvania Ave. NW
Washington, D.C. 20006
Tel: [phone]
Fax: [phone]
[e-mail]
[e-mail]
[e-mail]
[e-mail]
[e-mail]

William F. Lee (BBO #291960)
Felicia H. Ellsworth (BBO #665232)
Andrew S. Dulberg (BBO #675405)
Elizabeth Mooney (BBO #679522)
Sarah R. Frazier (BBO #681656)
WILMER CUTLER PICKERING
HALE AND DORR LLP
60 State Street
Boston, MA 02109
Tel: [phone]
Fax: [phone]
[e-mail]
[e-mail]
[e-mail]
[e-mail]
[e-mail]

Debo P. Adegbile (pro hac vice)
WILMER CUTLER PICKERING
HALE AND DORR LLP
7 World Trade Center
250 Greenwich Street
New York, NY 10007
Tel: [phone]
Fax: [phone]
[e-mail]
Dated: October 1, 2018

Counsel for Defendant President and
Fellows of Harvard College

CERTIFICATE OF SERVICE
I hereby certify that this document filed through the CM/ECF system will be sent
electronically to the registered participants as identified on the Notice of Electronic Filing.
/s/ Seth P. Waxman
Seth P. Waxman

10