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

Filing 65

MEMORANDUM in Support re 64 MOTION to Compel Production filed by Students for Fair Admissions, Inc.. (Caldwell, Benjamin)

IN THE UNITED STATES DISTRICT COURT
FOR THE DISTRICT OF MASSACHUSETTS
STUDENTS FOR FAIR ADMISSIONS, INC.,
Plaintiff,
v.

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

PRESIDENT AND FELLOWS OF HARVARD
COLLEGE (HARVARD CORPORATION),
REQUEST FOR
ORAL ARGUMENT

Defendant.

MEMORANDUM OF LAW IN SUPPORT OF
PLAINTIFF’S MOTION TO COMPEL PRODUCTION
Plaintiff Students for Fair Admissions, Inc. (“SFFA”) hereby respectfully submits
this motion to compel production of material responsive to certain Requests for
Production of Documents that SFFA served on Defendant, the President and Fellows of
Harvard College (“Harvard”), on May 12, 2015.1
PRELIMINARY STATEMENT
This lawsuit alleges that Harvard’s use of race in its admissions process violates
Title VI of the Civil Rights Act of 1964. See Compl. (Dkt No. 1) ¶¶ 1-9. Harvard
concedes that it considers race in admissions decisions, see Answer (Dkt No. 17) ¶ 147,
so the case turns on whether the way in which Harvard uses race satisfies strict scrutiny.

1

Counsel for the parties have conferred and attempted in good faith to resolve the issues
set forth herein, including by telephone on June 23 and email correspondence. As a result of
Harvard designating portions of Exhibit B to the Strawbridge Declaration as either “Confidential”
or “Highly Confidential – Attorneys’ Eyes Only,” sections of this memorandum which reference
that designated material have been redacted pursuant to the Stipulated Protective Order. An
unredacted version of this memorandum will be delivered to the Court in a sealed envelope along
with SFFA’s Motion to File Under Seal.

1

Plaintiff has requested three categories of relevant information and documents
that Harvard has unreasonably refused to produce. 2 First, SFFA seeks a preliminary
sample of application files from the last four admissions cycles at Harvard. SFFA’s need
for this information is obvious. Harvard uses a numerical system—under which it claims
race is not a factor—to score every applicant. Once that scoring is complete, Harvard
then uses a tiered review process in which applicants compete to advance from their
geographic region, to full-committee review, and ultimately into the admitted class.
Harvard concedes that race is a factor in determining which applicants advance through
that process, but denies that it uses race in a way that violates Title VI. SFFA therefore
must analyze application files to discover whether: (1) Harvard’s applicant scoring can be
justified based on its stated race-neutral criteria; and (2) Harvard’s selection of candidates
for advancement in its admissions process is consistent with its claim that race is not an
unlawfully significant factor in that process. Unsurprisingly, the only way to examine a
process that claims to review and evaluate application files individually is to review and
evaluate application files individually.
With that in mind, and in a good-faith effort to mitigate the potential burden of
producing the roughly 148,000 application files Harvard received during the last four
admission cycles, SFFA accepted Harvard’s offer to produce a “statistically significant
sample of redacted applications” needed “to evaluate SFFA’s claims.” Joint Statement of
April 23, 2015 (Dkt. No. 26), at 13 (emphasis added). By definition, a statistically

2

Harvard has yet to produce a single document in this case, notwithstanding SFFA’s
repeated invitations to begin the mutual exchange of documents. The parties also continue to
negotiate other discovery disputes, which are not yet ripe at this time and which SFFA hopes to
resolve without needing to resort to further motion practice.

2

significant sample is one that is large enough to ensure that reliable conclusions can be
drawn about the total population of application files from which the sample is drawn.
In a further effort to minimize Harvard’s burden of production, SFFA proposed a
preliminary random sampling of application files, which would permit SFFA and its
expert to determine how many total files are needed to draw statistically reliable
conclusions. SFFA thus requested that Harvard produce 400 complete application files
from each of the four largest racial categories (i.e. White, African-American, Asian,
Hispanic-American) from each of the past four admissions cycles, split evenly between
admitted and rejected applicants. That initial sample totals only 6,400 application files—
merely four percent of the total applications Harvard received over the last four years. In
short, SFFA agreed to Harvard’s proposal to use statistically significant sampling and
SFFA has proposed a two-step process that allows Harvard to produce, at the outset, the
smallest sample possible consistent with accepted standards for expert analysis.
Harvard, in a complete about-face, has now reneged on its promise to produce a
“statistically significant sample” of files. Instead, Harvard now agrees to produce only
160 total files—roughly one-tenth of one percent (0.1%) of the same four-year applicant
pool—half of which Harvard wants to hand pick rather than select randomly. 3 This
proposal is statistically worthless and contrary to the principles of sampling that the First
Circuit and other courts routinely apply. SFFA simply cannot perform any meaningful
analysis based on 160 non-randomly selected files. SFFA respectfully asks this Court to
order Harvard to produce the preliminary sample it has requested.

3

Because “statements in [a] 26(f) report constitute judicial admissions,” Carter v. Reiner,
Reiner & Bendett, P.C., Civil No. 3:06CV00988(AWT), 2007 WL 2221432, *1 n.1. (D. Conn.
July 30, 2007), this Court could compel production of the sample solely on the basis of Harvard’s
prior representation.

3

Second, SFFA requested the names of a small subset of the thousands of alumni
interviewers that Harvard uses as a key part of its application process. Alumni interviews
are often the only face-to-face meeting Harvard’s representatives ever have with an
applicant. Accordingly, alumni interviewers are uniquely positioned to discuss how race
affects the admission of the applicants they interview. SFFA limited its request to certain
key geographic areas where there are disproportionate high numbers of

eminently

qualified Asian-American applicants. Harvard, however, refuses to provide the names of
that subset of alumni interviewers for those geographic areas, despite their important role
in the admissions process and the fact that information about alumni interviewers is
readily available and, most importantly, clearly discoverable since they are eyewitnesses.
SFFA respectfully requests that this Court order Harvard to produce documents sufficient
to identify these individuals.
Third, SFFA seeks information about Harvard’s transfer student admissions
process. Harvard has refused to produce any documents about its transfer admissions,
contending that it is irrelevant to this case. But the Complaint alleges that Harvard
improperly uses race in considering transfer students; it also contends that the transfer
program could provide a race-neutral means for Harvard to achieve diversity. And some
of SFFA’s members hope to apply for transfer to Harvard if SFFA prevails. SFFA
therefore respectfully requests that the Court compel Harvard to produce this information.
ARGUMENT
I.

Legal Standard.
Several principles govern the scope of permissible discovery in this case. To

begin, the Rules provide for broad discovery into any matter “reasonably calculated to
lead to the discovery of admissible evidence.” Fed. R. Civ. P. 26(b)(1), 34(a)(1). “[A]

4

request for discovery should be considered relevant if there is any possibility that the
information sought may be relevant to the subject matter of the action.” Schuurman v.
Town of N. Reading, 139 F.R.D. 276, 277 (D. Mass. 1991) (quotation omitted).
Moreover, because this action alleges discrimination against Asian Americans in
violation of Title VI, SFFA is entitled to particularly broad discovery. “The great weight
of the policy in favor of discovery in civil rights actions supplements the normal
presumption in favor of broad discovery.” Grenier v. Jonas, No. 09-121, 2011 WL
1791093, at *2 (D. Vt. May 10, 2011) (quoting Floyd v. City of New York, 739 F. Supp.
2d 376, 381 n.21, 381-82 (S.D.N.Y. 2010)). Moreover, “determining the existence of a
discriminatory purpose demands a sensitive inquiry into such circumstantial and direct
evidence of intent as may be available.” Rogers v. Lodge, 458 U.S. 613, 618 (1982); see
also Raza v. City of New York, 998 F. Supp. 2d 70, 82 (E.D.N.Y. 2013) (“Limiting the
scope of discovery is especially inappropriate when, as here, the central fact at issue,
discriminatory intent, is difficult to establish.”).
Finally, Harvard retains the burden to produce evidence that its use of race can
withstand strict scrutiny. Because this Court (as well as the appellate courts) must give
“close analysis to the evidence of how the process works in practice,” there must be a
complete record with evidence sufficient to understand Harvard’s admissions decisions.
See Fisher v. Univ. of Texas at Austin, 133 S. Ct. 2411, 2421 (2013).

Given the

“searching review” that strict scrutiny requires, this Court should be especially skeptical
of Harvard’s attempts to thwart discovery into core aspects of its admissions process.
II.

Harvard Should Produce the Preliminary, Randomly Selected Sample of
Admissions Files SFFA Has Requested.

5

A. Harvard’s admissions process makes a review of application files an
essential part of developing the record in this case.
From the outset, the parties have recognized that any legal evaluation of SFFA’s
claims will require substantial discovery into Harvard’s admissions process. To
understand that process, and to tailor discovery accordingly, SFFA took the deposition of
Marlyn E. McGrath, Harvard’s Director of Admissions. In that deposition, SFFA learned
several important facts. First, Harvard divides the applicant pool into geographic regions,
and applicants compete by region for admission. See Strawbridge Decl, Ex. B at 158:1221. Harvard uses “readers” to conduct the initial evaluation of applications and issues
instructions to guide them in performing that task. See id. at 162:17-163:2.
Second, Harvard reduces each application to a numerical score by rating
applicants on a 1 to 4 scale (with 1 being the best) in four categories: academics,
extracurricular activities, athletics, and “personal.”

See id. at 158:22-160:5.

Each

category includes multiple criteria; for example, an applicant’s academic score reflects
grades, class rank, SAT or ACT scores, and academic achievements. See id. at 161:1424. The applicant’s overall score, however, is not an average or a composite of the
category scoring; rather, it is based on an overall judgment “standing back from the
folder.” See id. at 172:11-173:15. Harvard claims that race is not considered in any
aspect of this process. See id. at 165:7-165:18.
Third, once the scoring process is complete, the committee for each geographic
region decides which applicants will move on to full committee review. See id. at 177:919.

To facilitate that process, the committee is given a “docket” containing key

information about applicants such as their application score, their high school, and other
background information. See id. at 180:10-14. The race of each applicant is included on

6

the docket and race is considered in deciding which applicants will move forward. See
id. at 185:6-23. Once each region selects the applicants who will advance, the process is
repeated by the full committee, which makes the final decisions on which applicants will
be admitted to Harvard. See id. at 190:6-19. Once again, race is listed on the full
committee docket and is a factor in its decisions. See id. at 193:11-15; 194:11-198:7.
This process makes clear why review of application files is necessary. Harvard
claims that an applicant’s race is not a factor in the initial scoring of applications. The
only way for SFFA’s expert to test that claim is to review individual files to determine
whether, “compared with others similarly situated,” certain applicants were “selectively
treated” on the basis of race. See Marrero-Gutierrez v. Molina, 491 F.3d 1, 9 (1st Cir.
2007) (citations and quotations omitted).

Likewise, Harvard claims that while an

applicant’s race is considered during committee review, it is done consistent with Title
VI—viz., not to discriminate against a racial group or to engage in racial balancing, but
only as the Supreme Court has authorized. Here too, the way for SFFA’s expert to test
this claim is to review a significant number of files to see whether race is being used
consistent with the “demanding burden of strict scrutiny articulated in Grutter and …
Bakke.” Fisher, 133 S. Ct. at 2415. In short, there is no other way to determine “how the
process works in practice.” Id. at 2421.
That is why SFFA initially anticipated seeking the approximately 148,000
admission files from the past four cycles. Upon further reflection, however, SFFA never
made a formal request for all 148,000 application files.

Rather, since in the Joint

Statement, Harvard proposed “producing a statistically significant sample of redacted
applications, which will be sufficient to evaluate SFFA’s claims,” id. at 13, SFFA agreed

7

to Harvard’s offer. SFFA currently seeks only a preliminary sample of 1,600 randomly
selected application files for each of the last four admissions cycles, split evenly (i.e., 400
each) among white, black, Asian-American, and Hispanic or Latino applicants for each
year. See Ex. C. to Strawbridge Decl. at 25 (Request No. 24). This preliminary sample
will permit a determination as to how many additional files might be needed to constitute
a statistically significant sample size to provide a basis for an expert opinion.
Unfortunately, Harvard has reneged on its prior representations. Instead of the
statistically significant sample it promised to provide, Harvard now refuses to produce
anything more than a total of 160 application files from only the most recent admissions
cycle. Ex. C at 26 (Response No. 24). It insists that 80 of these 160 files must be handselected by Harvard, with the other half to be selected by SFFA. Id. In other words,
Harvard is unwilling to engage in any process of random selection, without which
statistical sampling is impossible, and it refuses to produce any files for any admissions
cycle other than the Class of 2019. Moreover, counsel for Harvard has made clear that
this would not be a preliminary sample, but the entirety of its production of application
files. See Ex. C at 26. Harvard thus apparently believes that a non-random sample of
one-tenth of one percent of the relevant pool is all that SFFA is entitled to in this case
(notwithstanding its prior representation that it would produce a statistically significant
sample). Harvard’s offer is not a serious proposal. SFFA’s motion should be granted.
B. Statistical sampling is an appropriate way to analyze Harvard’s
admissions process while minimizing the burden of discovery.
There is nothing unusual about using statistical sampling to draw conclusions
from a random sample of application files about the total population of files. Statistical
sampling has long been accepted by the courts in a variety of contexts to lessen the

8

burdens of discovery and to help the district courts manage cases without prejudicing
either party. The kinds of files SFFA needs here are routinely produced in civil rights
litigation. Discovery of a statistically significant sample of Harvard’s application files is
therefore entirely appropriate here.
First, the use of a sample to draw statistically valid conclusions about a greater
population has long been accepted by the federal courts. As the First Circuit has noted,
“sampling of similar claims and extrapolation from the sample is a recognized method of
proof.” United States v. Lahey Clinic Hosp., Inc., 399 F.3d 1, 18 n.19 (1st Cir. 2005).
Courts have substantial experience with sampling and extrapolation in a variety of
contexts, including discovery of the class members of large class actions,4 monitoring
compliance with an existing injunctive order,5 review of mortgage loan files in securities
litigation,6 and determination of the number of instances of alleged fraud.7
Nor is there any doubt that Harvard’s application files are properly subject to
discovery. Because of the necessity of such files to test the ordinary defenses made to
allegations of discrimination, see supra at 7, courts routinely require defendants to
produce records containing evaluations or qualifications of individuals who compete with
or are in a comparable position as the alleged victim of discrimination.

See, e.g.,

4

Morangelli v. Chemed Corp., No. 10-00876, 2011 WL 7475, at *1 (E.D.N.Y. 2011);
Seabron v. Am. Family Mut. Ins. Co., 862 F. Supp. 2d 1149, 1166 (D. Colo. 2012), order clarified
on reconsideration (June 26, 2012).
5

Hickenlooper v. Montez, No. 92–cv–00870–CMA, 2014 WL 4413221, at *3 n.4 (D.
Colo. Sept. 8, 2014).
6

Massachusetts Mut. Life Ins. Co. v. Residential Funding Co., LLC, 989 F. Supp. 2d 165
(D. Mass. 2013) (Saris, C.J.); Fed. Hous. Fin. Agency v. JPMorgan Chase & Co., No. 11-6188
DLC, 2012 WL 6000885, at *1 (S.D.N.Y. Dec. 3, 2012).
7

U.S. ex rel. Loughren v. UnumProvident Corp., 604 F. Supp. 2d 259, 261 (D. Mass.

2009).

9

Weahkee v. Norton, 621 F.2d 1080, 1082 (10th Cir. 1980) (holding that plaintiff was
entitled to files of employees who were hired and promoted over him because “[t]he
qualifications and job performance of these employees in comparison with the plaintiff’s
qualifications and performance is at the heart of this controversy” and that this is “the sort
of information that EEOC seeks, and we have allowed, in investigating and litigating
discrimination claims”); Diaz-Garcia v. Surillo-Ruiz, 45 F. Supp. 3d 163, 166-67 (D. P.R.
2014) (permitting discovery of other employee files because they may provide evidence
that employees were hired for political reasons or that grounds for plaintiffs’ dismissal
were pretextual). Indeed, Harvard itself previously has been ordered to produce similar
records in a sex discrimination case. Jackson v. Harvard University, 111 F.R.D. 472, 476
(D. Mass. 1986) (ordering production of ten years’ worth of faculty tenure evaluations
because “plaintiff is entitled to an opportunity to show that others with similar
qualifications achieved the rank or position or that women at Harvard Business School
are evaluated by a stricter standard than their male colleagues, or that the evaluations of
plaintiff when compared with those of men who were granted tenure, reveal a covert bias
against women faculty”).
Given the large number of application files at issue here—approximately
148,000—extrapolation from a sample is a reasonable way to limit the burden on Harvard
while permitting a statistically defensible analysis of the extent to which race is playing a
factor in Harvard’s allegedly “holistic” process. Once the sample is produced, this
analysis can be done by assigning values to Harvard’s criteria for scoring applicants and
then determining, through regression analysis, the extent to which race, rather than the

10

other criteria,

appears to be influencing the admissions process.

See Ex. A to

Strawbridge Decl. ¶ 28.
None of this should be a point of controversy for purposes of this motion. After
all, it was Harvard’s position before this Court on April 23 that statistically significant
sampling was an appropriate method of conducting discovery on application files. The
only question for this Court is which of the parties’ sampling proposals hews closer to
basic statistical principles and thus promises to provide worthwhile analytical results.
C. The sample must be of sufficient size such that reliable conclusions can be
drawn regarding Harvard’s admissions decisions.
The sheer difference in the relative size of the proposed samples—6,400 versus
160—weighs heavily in favor of SFFA’s proposal. SFFA’s preliminary sample is crafted
to provide a strong basis for reviewing the application files and determining how many
more files would be necessary for a statistically reliable analysis.

The preliminary

sample must be large enough to analyze the effect of race on various aspects of Harvard’s
admissions process, see Ex. A ¶ 30, and to support inferences about the entire pool with
the requisite degree of statistical confidence, see id. ¶ 31.
There is no dispute that “[s]tatistical evidence drawn from an extremely small
universe has little probative value.” White v. Vathally, 570 F. Supp. 1431, 1435 (D. Mass.
1983), aff’d, 732 F.2d 1037 (1st Cir. 1984) (quotation omitted); Schmid v. Frosch, 680
F.2d 248, 250 (D.C. Cir. 1982) (“[S]mall samples provide a less reliable basis for making
an inference about the treatment of two groups than do larger samples of decisions.”);
United States v. City of New York, 637 F. Supp. 2d 77, 95 (E.D.N.Y. 2009) (“Courts have
sometimes declined to rely on statistical significance analysis when a sample size was too
small.”) (collecting cases).

11

SFFA’s proposal of a preliminary sample totaling four percent of Harvard’s
approximately 148,000 application files during the last four years is well below the
mainstream sample range granted by courts in comparable circumstances. 8 See, e.g.,
Hickenlooper, 2014 WL 4413221, at *3 n.4 (ordering 10% sample for discovery of
compliance with ADA); Seabron, 862 F. Supp. 2d at 1166, order clarified on
reconsideration (June 26, 2012) (ordering 25% sample for discovery of class plaintiffs);
Perez–Benites v. Candy Brand, LLC, No. 07-1048, 2010 U.S. Dist. LEXIS 115590, at *56 (W.D. Ark. Oct. 27, 2010) (limiting discovery of class plaintiffs to 20% sample); Katt
v. New York City Police Dep’t, No. 95-8283, 1997 WL 394593, at *3 (S.D.N.Y. July 14,
1997) (ordering production of random sample of 19% of files regarding sexual
harassment complaints); Skibo v. City of New York, 109 F.R.D. 58 (E.D.N.Y.1985)
(ordering production of a random sample exceeding 10% of files containing complaints
against police officers).
The clear weight of authority favoring a statistically significant sample holds true
even in recent cases involving thousands of files. For example, in mortgage-related
litigation initiated by the Federal Housing Finance Agency, the district court approved a
methodology employing a sample of 50,000 mortgage loan files out of a population of
1.1 million mortgages—an overall percentage of 4.5 percent. Fed. Hous. Fin. Agency v.
JPMorgan Chase & Co., No. 11-6188, 2012 WL 6000885, at *1 (S.D.N.Y. Dec. 3,
2012); Fed. Hous. Fin. Agency v. JPMorgan Chase & Co., 2015 WL 2183875, at *38
n.76 (S.D.N.Y. May 11, 2015); see also Massachusetts Mut. Life Ins. Co., 989 F. Supp.
8

Although Harvard claims that less than four years is an appropriate time frame, and
proposes only producing application files from one admissions cycle, Ex. C at 26, SFFA’s
Complaint specifically alleges that Harvard engages in racial balancing over a four-year period,
adjusting the numbers by race each year to maintain its preferred racial distribution. Compl. ¶¶
300-304.

12

2d at 167-68 (approving methodology based upon sample of discovery into 9,900
mortgage loan files for population of 278,609 loans, or 3.6 percent). These cases support
the reasonableness and scientific basis for SFFA’s proposal here.

The preliminary

sample SFFA has requested will yield an ultimate request for files that falls within this
accepted range. Indeed, it is possible that the sample may be all the files SFFA requires
to analyze Harvard’s admissions process. In all events, however, there is no legal or
statistical basis for Harvard’s proposal to provide 0.1 percent of application files.
D. The sample must be random to avoid selection bias and stratified to
ensure that SFFA’s expert can competently examine the effect of race on
the admissions process.
An equally glaring flaw in Harvard’s proposal is its refusal to generate the sample
using random selection. Instead, Harvard proposes to choose (and, no doubt, to cherrypick) 80 files and to let SFFA pick 80 additional files. Given a pool of approximately
37,000 files for each admission year, Harvard’s process is deliberately designed to yield
unrepresentative results. It also is the antithesis of sound statistical practice. As Dr.
Arcidiacono explains, “a biased sample is not representative of the entire population and
the inferences drawn from such a sample cannot reliably characterize for the entire
population.” Ex. A at ¶ 22. Random sampling “eliminates these problems by ensuring
that each member of the population has an equal chance of being selected for the sample;
this ensures that the sample is representative of the whole population.” Id. at ¶ 23.
Common sense dictates that a selectively chosen sample is unlikely to reflect the
larger population. “Probability sampling methods, in contrast, ideally are suited to avoid
selection bias. . . . [and] leaves no room for selection bias.” Reference Manual on
Scientific Evidence, Federal Judicial Center at 100 (2d ed. 2000); see also United States
v. Skodnek, 933 F. Supp. 1108, 1118 (D. Mass. 1996) (crediting affidavit from a Harvard
13

statistics professor describing “the rationale for random sampling as a method of
objectively choosing a sample that is representative of the population”). Indeed, courts
often reject non-random sampling because of concerns about selection bias. See, e.g., In
re Countrywide Fin. Corp. Mortgage-Backed Sec. Litig., 984 F. Supp. 2d 1021, 1040
(C.D. Cal. 2013) (rejecting admissibility of expert report because “[w]ithout a random
sample . . . there is no reliable way to draw conclusions about the relationship between
Plaintiffs’ certificates and Countrywide RMBS as a whole”); In re Chevron U.S.A., Inc.,
109 F.3d 1016, 1019-20 (5th Cir. 1997) (rejecting selective sampling and holding that
“the sample must be a randomly selected one of sufficient size so as to achieve statistical
significance to the desired level of confidence in the result obtained”).
In addition to random selection, SFFA proposes a sample stratified by the race of
the applicant, meaning that the sample includes equal numbers of applicants from each of
the four largest racial categories. Ex. C at 25. Harvard’s proposal, by contrast, is not
stratified in any way. Id. at 26. Under its proposal, Harvard could hand-pick 80 Hispanic
students, and therefore massively distort the representative nature of the 160-file sample
Stratified random sampling, in contrast, would ensure that the sample includes “[a]
sufficient number of both rejected and accepted applicants is necessary to draw any
meaningful conclusions.” Ex. A. at ¶ 25. As a result, it ensures fewer files are needed
than would be the case with simple random sampling. Id. at ¶ 35. The stratified sample
will then permit SFFA’s expert “to discern differences or relationships between
subgroups within” the applicant pool. Id. at ¶ 26. Stratified sampling permits analysis of
the effect of practices or process across characteristics such as. Id.; see also Reference
Manual on Scientific Evidence, at 100 n.52 (discussing the advantage of “[m]ore

14

complicated methods, such as stratified sampling and cluster sampling”); Massachusetts
Mut. Life Ins. Co., 989 F. Supp. 2d at 175 (“Stratification increases the precision.”).
E. Requiring the production of this preliminary sample does not impose an
undue burden on Harvard.
SFFA’s request for a preliminary sample imposes no undue burden on Harvard.
To the contrary, SFFA’s proposal limits the number of files that Harvard will need to
produce, both at the outset and by the end of discovery. Ex. A at ¶ 35. Although Harvard
would prefer to produce fewer documents, it cannot meet its burden by “‘rely[ing] on a
mere assertion that compliance would be burdensome and onerous without showing the
manner and extent of the burden and the injurious consequences of insisting upon
compliance.’” In re New England Compounding Pharmacy, Inc. Products Liab. Litig.,
No. 13-2419, 2013 WL 6058483, at *6 (D. Mass. Nov. 13, 2013) (quotation omitted)).
Harvard also has invoked its obligations under the Family Educational Rights and
Privacy Act (“FERPA”) and may argue that those obligations somehow increase its
burden to justify denying a statistically sound sample to SFFA. This is a red herring. As
a threshold matter, a large percentage of the produced files will be for students who never
enrolled at Harvard and as to whom FERPA imposes little or no obligations on Harvard.
See 20 U.S.C. § 1232g(a)(6) (excluding from the definition of “student” persons who do
not attend the institution). Moreover, SFFA already has agreed to permit Harvard to
redact a substantial amount of personally identifiable information from the applications it
does produce. See Protective Order (Dkt. No. 55) ¶¶ 2(a), 14.
Furthermore, FERPA permits the disclosure of such information with proper
notice. There is no substantial burden to Harvard’s providing notice to a few thousand
individuals (at most); its admissions office alone sends more than 100,000 letters at the

15

beginning of every recruiting cycle. See Ex. B at 16:7-18. Although Harvard may prefer
to redact files instead of giving notice, that voluntary choice cannot create a burden
justifying resistance to a reasonable, statistically sound sampling approach.
F. Harvard cannot rely on individualized review of applications to defend
against conclusions SFFA’s expert may draw from statistics if it does not
produce a statistically significant sample of those files.
Finally, it is important to understand the ramifications for this case if SFFA is not
permitted to examine a statistically significant sample of files from each admissions cycle
during the relevant period. As the Court explained at the initial scheduling conference, it
is vital that the parties build as “good and accurate a record” as possible. Transcript of
April 30, 2015, Status Conference at 4. Allowing Harvard to thwart discovery on these
files would be tantamount to an order denying SFFA the right to conduct discovery on
issues that go to the heart of this case—issues that will be essential to this Court’s
resolution of SFFA’s claims and the multiple levels of appellate review likely to follow.
Yet the ramifications for Harvard will be equally severe and potentially fatal to its
ability to defend its admissions program. SFFA has requested—and Harvard had agreed
to produce—statistical data concerning admissions decisions during the relevant period.
It is quite likely that this data will establish that Harvard’s admissions policies have a
disproportionate racial effect sufficient to establish an inference of intentional
discrimination, see Vill. of Arlington Heights v. Metro. Hous. Dev. Corp., 429 U.S. 252,
266 (1977), and that Harvard is impermissibly using race in its admissions process in
other ways that cannot be defended under strict scrutiny, see Fisher, 133 S. Ct. at 2415.
Harvard will surely respond by arguing that these statistical disparities can be explained
on a case-by-case basis. In particular, Harvard will likely defend its program as one that

16

is “holistic and complex” and which considers “many fine applicants” but “can choose
only a few.” See Ex. B. at 118:5-9. It will be Harvard’s position that the statistics do not
tell the whole story because “each applicant is really considered as an individual” based
on a complete review of his or her application file. Id. at 231. Indeed, Harvard has
emphasized the “holistic nature” of its admissions process and the many, many factors
that might influence its decisions. See Answer ¶¶ 124, 147; Joint Statement, at 8;
Strawbridge Decl., Ex. D, at 12-13 (Response to Interrogatory No. 4).
But if Harvard does not produce the statistically significant sample of admissions
files in order to allow SFFA to examine and rebut those claims, it will not be permitted to
make these arguments. Harvard’s ability to carry its burden will rise or fall on the
statistics it produces. Harvard cannot defend its program based on the individualized,
holistic review of the entire application file and then conceal in discovery the very
materials it claims to rely upon to advance that defense. See Rule 37(c)(1) (“If a party
fails to provide information . . . as required by Rule 26(a) or (e), the party is not allowed
to use that information or witness to supply evidence on a motion, at a hearing, or at a
trial, unless the failure was substantially justified or is harmless.”); Cosme-Montalvo v.
Trafon Grp., Inc., No. 11-2197, 2013 WL 1728577, at *3 (D.P.R. Apr. 22, 2013), appeal
dismissed (Nov. 6, 2013) (prohibiting defendants’ use at summary judgment of
documents withheld in discovery). Either SFFA must be provided the opportunity to
review those materials and test Harvard’s claims to see if Harvard’s assertions regarding
the use of race truly match up to its actual practices or Harvard may not rely on the
allegedly individualized, holistic nature of its program in this case.

17

III.

Harvard Should Produce the Requested Names and Contact Information for
a Subset of Its Alumni Interviewers.
Harvard uses a network of its alumni to interview applicants to Harvard. See,

e.g., Complaint ¶ 175. There is no dispute that these interviews are an important part of
the admissions process.

Ms. McGrath testified that: Harvard’s admissions officers

coordinate alumni interviews; alumni interviewers evaluate applicants and submit an
assessment of the interview to the admissions office, where it becomes part of the
admissions file; and that this information is used as part of the admissions decisions. See
Ex. B at 56:8-57:5; 60:2-61:15; Ex. D at 13 (noting that “reports from alumni or
admissions officer interviews” are a factor in admissions). Based on this undisputed
evidence alone, alumni interviewers are certain to have relevant information.
In light of their substantial role in the admissions process, SFFA propounded a
request to Harvard for documents sufficient to show the names of alumni interviewers it
uses in specific geographic areas most likely to have a significant number of competitive
Asian applicants. See Ex. C at 21-22 (Request No. 19). Although Harvard has not
disputed the importance of the alumni interview for applicants, Harvard nonetheless
refuses to provide to SFFA documents responsive to its request on the grounds of alleged
burden, invasion of privacy, and lack of relevance. See Ex C at 22 (Response No. 19).
Harvard’s blanket refusal to provide documents sufficient to show the names of
the certain alumni interviewers that SFFA has requested is indefensible. Many of those
individuals are likely to have relevant information about the admissions process,
including the extent to which race drives admissions decision. Rule 26 specifically
permits discovery into “the identity and location of persons who know of any
discoverable matter.” Fed. R. Civ. P. 26(b)(1); Gerber v. Down E. Cmty. Hosp., 266

18

F.R.D. 29, 35 (D. Me. 2010) (noting that Rule 26(b)(1)’s “language clearly encompasses
the identity of potential witnesses”). Harvard’s effort to shield the names of these thirdparty witnesses on relevancy or privacy grounds is baseless and not made in good faith.
Finally, there is little burden on Harvard to respond to this request; Ms. McGrath
testified that Harvard assigns one admissions officer oversight responsibilities for the
alumni interviewers, which presumably includes access to their basic contact information.
Ex. B. at 50:1-7; 52:16-18. Counsel for Harvard has instead asserted an alleged burden
on the interviewers themselves based on the scope of future third-party discovery that
SFFA may serve. But the rules do not permit prospective, speculative, and premature
objections concerning third-party discovery as a basis to withhold readily available
information of eyewitnesses within the party’s possession. SFFA therefore respectfully
requests that the Court order Harvard to comply with SFFA’s Document Request No. 19.
IV.

Harvard Should Produce Information and Documents Relevant to the Use of
Race in Its Transfer Student Admissions Process.
Harvard has responded to numerous SFFA discovery requests by limiting its

production of documents to the freshman admissions process. See Ex. C at 12 (Response
No. 1), 14 (Response No. 4), 23 (Response No. 21), 29 (Response Nos. 29 and 30), 32
(Response No. 34), 33 (Response No. 36). Harvard bases its categorical refusal to
provide any discovery into its transfer admissions process on the bare assertion that
“documents or information concerning individual transfer applicants . . . are irrelevant to
this Action.” See, e.g., Ex. C at 7 (Objection to Instruction No. 3). In doing so, Harvard
seeks to exclude any discovery into the use of race in the admission of transfer students.
But Harvard admitted in its Answer that its “admissions office considers
applicants for transfer admission in a holistic manner and that, as with applicants for

19

early or regular admission, it considers each applicant’s background and personal
characteristics, including—where relevant—the applicant’s race or ethnicity as it bears
on the holistic review.” Answer ¶ 186. Harvard’s website likewise touts that its transfer
applicants undergo the same “holistic” admissions process that applies to freshman
applicants. See Ex. E. And counsel for Harvard has not articulated any difference in the
ways in which race is used or considered for transfer students.
Moreover, SFFA’s Complaint clearly establishes the specific relevance of the
transfer admissions process to this action:
•

SFFA has members who were denied admission to Harvard as freshmen but retain
interest in applying for transfer if race will not be a factor in considering their
application. See Complaint ¶ 24.

•

The recruitment and admissions of transfer students, particularly from community
colleges, may be a race-neutral alternative to Harvard’s race-based admissions
process. Id. ¶¶ 334-38.

•

SFFA contends, among other things, that Harvard is engaged in ongoing racial
balancing to preserve a specific composition of racial groups on campus. See,
e.g., id. ¶ 300. The transfer admissions process is one way by which Harvard
could achieve this goal, particularly as issues of mismatch, see id. ¶¶ 383-99,
result in attrition among the students admitted to Harvard as freshmen.
Because the transfer admissions process is directly relevant to SFFA’s claims, the

Court should reject Harvard’s categorical objection to producing materials regarding this
process and order it to produce documents and information about transfer admissions.
CONCLUSION
For the reasons given, SFFA respectfully requests that the Court: (a) order
Harvard to comply with SFFA Request for Production Nos. 19 and 24; (b) order Harvard
to produce documents regarding its transfer admissions process on the same terms it
produces information regarding its freshman admission process in response to SFFA’s
discovery requests; and (c) grant such other and further relief as the Court sees fit.
20

Respectfully submitted,
By: /s/ William S. Consovoy
Paul M. Sanford BBO #566318
Benjamin C. Caldwell BBO #67506
BURNS & LEVINSON LLP
One Citizens Plaza, Suite 1100
Providence, RI 02903
Tel: [phone]
Fax: [phone]
[e-mail]
[e-mail]

Dated: July 16, 2015

William S. Consovoy
Thomas R. McCarthy
J. Michael Connolly
CONSOVOY MCCARTHY PARK PLLC
3033 Wilson Boulevard, Suite 700
Arlington, Virginia 22201
Tel: [phone]
Fax: [phone]
[e-mail]
[e-mail]
[e-mail]
Patrick Strawbridge
BBO #678274
CONSOVOY MCCARTHY PARK PLLC
Ten Post Office Square
8th Floor South PMB #706
Boston, MA 02109
Tel: [phone]
[e-mail]
Counsel for Plaintiff Students for Fair
Admissions, Inc.

4820-7134-5189.1

21