Students for Fair Admissions v. President & Fellows of Harvard ( 2020 )

United States Court of Appeals
For the First Circuit
No. 19-2005
STUDENTS FOR FAIR ADMISSIONS, INC.,
Plaintiff, Appellant,
v.
PRESIDENT AND FELLOWS OF HARVARD COLLEGE,
Defendant, Appellee.
APPEAL FROM THE UNITED STATES DISTRICT COURT
FOR THE DISTRICT OF MASSACHUSETTS
[Hon. Allison D. Burroughs, U.S. District Judge]
Before
Howard, Chief Judge
and Lynch, Circuit Judge.*
William S. Consovoy, with whom Thomas R. McCarthy, J. Michael
Connolly, Cameron T. Norris, Patrick Strawbridge, Consovoy
McCarthy PLLC, Adam K. Mortara, John M. Hughes, and Bartlit Beck
LLP were on brief, for appellant Students for Fair Admissions,
Inc.
Eric S. Dreiband, Assistant Attorney General, with whom
Elliot M. Davis, Acting Principal Deputy Assistant Attorney
General, Thomas E. Chandler, and Matthew J. Donnelly were on brief,
for the United States, amicus curiae.
Gordon M. Fauth, Jr., Litigation Law Group, Lee C. Cheng, and
* Judge Torruella heard oral argument in this matter and
participated in the semble, but he did not participate in the
issuance of the panel's decision. The remaining two panelists
therefore issued the opinion pursuant to 28 U.S.C. § 46(d).
Asian American Legal Foundation on brief for the Asian American
Coalition for Education and the Asian American Legal Foundation,
amici curiae.
Wencong Fa, Joshua P. Thompson, and Pacific Legal Foundation
on brief for Pacific Legal Foundation, Reason Foundation, Center
for Equal Opportunity, Individual Rights Foundation, and the
Chinese American Citizens Alliance - Greater New York, amici
curiae.
Randall B. Clark, C. Boyden Gray, Adam R.F. Gustafson, James
R. Conde, T. Elliot Gaiser, and Boyden Gray & Associates on brief
for Economists Michael Keane, Hanming Fang, Christopher Flinn,
Stefan Hoderlein, Yingyao Hu, Joseph Kaboski, Glenn Loury, Thomas
Mroz, John Rust, and Matthew Shum, amici curiae.
Jun Xiao, Yong Li, Xiaoying Yu, and Haiying Li, on brief pro
se.
Timothy J. Perry, Perry Krumsiek, LLP, T. Russell Nobile,
Robert D. Popper, and Judicial Watch, Inc. on brief for Judicial
Watch, Inc., amicus curiae.
Cody J. Wisniewski on brief for Mountain States Legal
Foundation, amicus curiae.
Dwight Duncan and Dennis J. Saffran on brief for National
Association of Scholars, amicus curiae.
Seth P. Waxman, with whom Paul R.Q. Wolfson, Danielle Y.
Conley, Brittany Blueitt Amadi, William F. Lee, Felicia H.
Ellsworth, Andrew S. Dulberg, Debo P. Adegbile, Michelle Liszt
Sandals, Greg Schmidt, Emma Simson, Alex Hemmer, Wilmer Cutler
Pickering Hale and Dorr LLP, Ara B. Gershengorn, and Harvard
University, Office of the General Counsel were on brief, for
appellee President and Fellows of Harvard College.
David Hinojosa, with whom Jon M. Greenbaum, Genevieve
Bonadies Torres, Lawyers' Committee for Civil Rights Under Law,
Lawrence E. Culleen, Elisabeth S. Theodore, Nancy L. Perkins,
Janine M. Lopez, Emma Dinan, Camille Heyboer, Arnold & Porter Kaye
Scholer LLP, Niyati Shah, Marita Etcubanez, Eri Andriola, Asian
Americans Advancing Justice, Oren M. Sellstrom, and Lawyers for
Civil Rights were on brief, for Students, Alumni, and Prospective
Students of Harvard College, amici curiae.
Jin Hee Lee, with whom Sherrilyn Ifill, Janai Nelson, Samuel
Spital, Rachel Kleinman, Cara McClellan, Michaele N. Turnage
Young, Jennifer A. Holmes, NAACP Legal Defense and Educational
Fund, Inc., Kate R. Cook, Kenneth N. Thayer, and Sugarman, Rogers,
Barshak & Cohen, P.C. were on brief, for Coalition for a Diverse
Harvard, Association of Black Harvard Women, First Generation
Harvard Alumni, Fuerza Latina of Harvard, Harvard Asian American
Alumni Alliance, Harvard Asian American Brotherhood, Harvard Black
Alumni Society, Harvard Islamic Society, Harvard Japan Society,
Harvard Korean Association, Harvard Latino Alumni Alliance,
Harvard Minority Association of Pre-Medical Students, Harvard
Phillips Brooks House Association, Harvard Progressive Jewish
Alumni, Harvard South Asian Association, Harvard University Muslim
Alumni, Harvard Vietnamese Association, Harvard-Radcliffe Asian
American Association, Harvard-Radcliffe Asian American Women’s
Association, Harvard-Radcliffe Black Students Association,
Harvard-Radcliffe Chinese Students Association, Kuumba Singers of
Harvard College, Native American Alumni of Harvard University,
Native Americans at Harvard College, Task Force for Asian American
Progressive Advocacy and Studies, and 21 Colorful Crimson, amici
curiae.
Jessica L. Ellsworth, Stephanie J. Gold, Jo-Ann Tamila Sagar,
Hogan Lovells US LLP, Peter McDonough, and American Council on
Education on brief for American Council on Education and 40 Other
Higher Education Organizations, amici curiae.
Samuel P. Groner, Harrison D. Polans, Fried, Frank, Harris,
Shriver & Jacobson, LLP, Steven M. Freeman, and Amy Feinman on
brief for Anti-Defamation League, amicus curiae.
Madeleine K. Rodriguez, Dean Richlin, Hemmie Chang, Sarah
Burg, Rachel Hutchinson, Jacqueline Chávez, Foley Hoag LLP,
Kenneth Kimerling, and Asian American Legal Defense and Education
Fund on brief for Asian American Legal Defense and Education Fund
and Other Asian American Education and Youth-Serving Organizations
and Higher Education Faculty.
Jaime A. Santos, Sabrina M. Rose-Smith, William Evans, and
Goodwin Procter LLP, on brief for the National Association of
Basketball Coaches, Women's Basketball Coaches Association, Geno
Auriemma, James A. Boeheim, John Chaney, Tom Izzo, Michael W.
Krzyzewski, Joanne P. McCallie, Nolan Richardson, Bill Self, Sue
Semrau, Orlando Smith, Tara Vanderveer, Roy Williams, Jay Wright,
and 326 Additional Current or Former College Head Coaches, amici
curiae.
Matthew S. Hellman, Ishan K. Bhabha, and Jenner & Block LLP,
on brief for Brown University, Columbia University, Cornell
University, Dartmouth College, Duke University, Emory University,
Johns Hopkins University, Massachusetts Institute of Technology,
Princeton University, Stanford University, University of Chicago,
University of Pennsylvania, Vanderbilt University, Washington
University in St. Louis, and Yale University, amici curiae.
Maura Healey, Attorney General of Massachusetts, Elizabeth N.
Dewar, State Solicitor, Ann E. Lynch, Assistant Attorney General
of Massachusetts, David Ureña, Assistant Attorney General of
Massachusetts, Xavier Becerra, Attorney General of California,
Philip J. Weiser, Attorney General of Colorado, Kathleen Jennings,
Attorney General of Delaware, Karl A. Racine, Attorney General for
the District of Columbia, Clare E. Connors, Attorney General of
Hawai'i, Kwame Raoul, Attorney General of Illinois, Aaron M. Frey,
Attorney General of Maine, Brian E. Frosh, Attorney General of
Maryland, Keith Ellison, Attorney General of Minnesota, Aaron D.
Ford, Attorney General of Nevada, Hector Balderas, Attorney
General of New Mexico, Letitia James, Attorney General of New York,
Josh Shapiro, Attorney General of Pennsylvania, Peter Neronha,
Attorney General of Rhode Island, and Mark R. Herring, Attorney
General of Virginia, on brief for Massachusetts, California,
Colorado, Delaware, the District of Columbia, Hawai'i, Illinois,
Maine, Maryland, Minnesota, Nevada, New Mexico, New York,
Pennsylvania, Rhode Island, and Virginia, amici curiae.
Derek T. Ho, Bradley E. Oppenheimer, Minsuk Han, Joseph L.
Wenner, and Kellogg, Hansen, Todd, Figel & Frederick P.L.L.C. on
brief for Professors of Economics, amicus curiae.
Sarah E. Harrington and Goldstein & Russell, P.C. on brief
for 678 Social Scientists and Scholars on College Access, Asian
American Studies, and Race, amici curiae.
Anton Metlitsky, Patrick D. McKegney, Apalla U. Chopra,
Bradley N. Garcia, Anna O. Mohan, and O'Melveny & Myers LLP on
brief for Walter Dellinger, amicus curiae.
Mark S. Davies, Thomas M. Bondy, Katherine M. Kopp, Sheila A.
Baynes, Sarah H. Sloan, E. Joshua Rosenkranz, Darren S. Teshima,
and Orrick, Herrington & Sutcliffe LLP on brief for Amgen Inc.,
Apple Inc., Applied Materials, Inc., Cisco Systems, Inc., Cummins
Inc., General Electric Company, Gilead Sciences, Inc.,
GlaxoSmithKline LLC, Intel Corporation, Micron Technology, Inc.,
Microsoft Corporation, Twitter, Inc., and ViiV Healthcare Company,
amici curiae.
Keffe B. Clemons on brief for Verizon Services Corp., amicus
curiae.
November 12, 2020
LYNCH, Circuit Judge. Students for Fair Admissions,
Inc. ("SFFA") brought suit on November 17, 2014, against the
President and Fellows of Harvard College and the Board of Overseers
(collectively, "Harvard"). The suit alleged that Harvard
College's admittedly race-conscious undergraduate admissions
process violates Title VI of the Civil Rights Act of 1964, 42
U.S.C. § 2000d et seq. ("Title VI") by discriminating against Asian
American applicants in favor of white applicants.
SFFA asserts that Harvard fails to meet the Supreme
Court's standards for the use of race in admissions which are
asserted to be justified by diversity in these ways: (1) it engages
in racial balancing of its undergraduate class; (2) it
impermissibly uses race as more than a "plus" factor in admissions
decisions; (3) it considers race in its process despite the
existence of workable race-neutral alternatives; and (4) it
intentionally discriminates against Asian American applicants to
Harvard College. SFFA seeks a declaratory judgment, injunctive
relief, attorneys' fees, and costs.
The district court denied Harvard's motion to dismiss
SFFA's suit for lack of Article III standing. See Students for
Fair Admissions, Inc. v. President & Fellows of Harvard Coll.
("SFFA I"), 261 F. Supp. 3d 99, 111 (D. Mass. 2017).
After a fifteen-day bench trial at which thirty
witnesses testified, the district court issued a 130-page opinion
- 5 -
with findings of fact and conclusions of law. See Students for
Fair Admissions, Inc. v. President & Fellows of Harvard Coll.
("SFFA II"), 397 F. Supp. 3d 126, 132 (D. Mass. 2019). It made
numerous factual findings, including as to competing expert
witness testimony and credibility determinations about the
testimony of witnesses. See id. at 158-83. The district court
found that Harvard had met its burden of showing its admissions
process did not violate Title VI. See id. at 197, 199, 201, 204.
It entered judgment for Harvard on all counts. See id.
SFFA appeals from this judgment, and Harvard renews its
argument that SFFA lacks standing.1
After careful review of the record, we hold that SFFA
has associational standing to bring its claims and that under
governing Supreme Court law Harvard's race-conscious admissions
program does not violate Title VI.2
1 The district court dismissed two of SFFA's original
claims before trial: that Harvard uses race to fill more than just
the last few places in its class and that Harvard considers race
in its admissions process generally. SFFA does not challenge the
dismissal of these claims on appeal.
2 We acknowledge and thank all of the amici curiae for
their helpful submissions in this matter. The following amici
submitted briefs in support of SFFA: the United States; the Asian
American Coalition for Education and the Asian American Legal
Foundation; the Pacific Legal Foundation and the four other
institutions on the brief; Michael Keane and the nine other
economists on the brief; Jun Xiao and the three other individuals
on the brief; Judicial Watch, Inc.; the Mountain States Legal
Foundation; and the National Association of Scholars. The
following amici filed briefs in support of Harvard: Students,
Alumni, and Prospective Students of Harvard College; the American
- 6 -
I. Facts
We recount the relevant basic facts before we turn to
our legal analysis.
1) SFFA
SFFA is a validly incorporated 501(c)(3) nonprofit
organization. Its bylaws state that it was formed to "defend human
and civil rights secured by law, including the right of individuals
to equal protection under the law, through litigation and any other
lawful means."
SFFA was incorporated on July 30, 2014. Its original
board of directors had three self-appointed members: Edward Blum,
President, Abigail Fisher, Secretary, and Richard Fisher,
Treasurer. In November 2014, when it filed suit against Harvard,
SFFA's bylaws also provided for "affiliate members" who
Council on Education and the forty other higher education
organizations on the brief; the Anti-Defamation League; the Asian
American Legal Defense Fund and the forty-five other Asian American
education and youth-serving organizations and faculty on the
brief; the National Association of Basketball Coaches, the Women's
Basketball Coaches Association, and the 339 current and former
coaches on the brief; Brown University and the fourteen other
colleges and universities on the brief; Massachusetts and the
fifteen other states on the brief; the eighteen Professors of
Economics; the 678 Social Scientists and Scholars on College
Access, Asian American Studies, and Race; Walter Dellinger; and
the Coalition for a Diverse Harvard and the twenty-five other
organizations on the brief. Amgen, Inc. and thirteen other
corporations submitted a brief in favor of preserving diversity
and inclusion efforts like those previously approved by the Supreme
Court.
- 7 -
"support[ed] the purposes and mission of the Corporation." All
members of SFFA were affiliate members. Membership was free, but
affiliate members could not vote for any officers or directors of
the organization. SFFA had forty-seven affiliate members when it
sued Harvard, including Asian American members who had applied to
and been rejected by Harvard. Several of these members submitted
declarations stating that they voluntarily joined SFFA, supported
its mission, had been in contact with SFFA, received updates about
this litigation, and were able to express their views on it.
In June 2015, SFFA amended its bylaws to eliminate
"affiliate members" and replace them with "general members."
Unlike affiliate members, general members have the right to vote
for a member-elected director and must pay membership dues to SFFA.
As of 2017, SFFA's membership had increased to approximately 20,000
members.
2) Harvard's Admissions Process
Harvard's admissions process is complex and highly
competitive. Each year, Harvard College attempts to admit a class
of roughly 1,600 students. For its class of 2019, Harvard received
around 35,000 applications. Because of the size of the applicant
pool relative to Harvard's available slots, Harvard cannot admit
all applicants who would succeed academically. Harvard has
determined that academic excellence alone is not sufficient for
admission. Rather, Harvard seeks students who are not only
- 8 -
academically excellent but also compelling candidates on many
dimensions.
Harvard's application process3 is accurately described
as having six components: (1) Harvard's pre-application
recruitment efforts; (2) students' submission of applications; (3)
Harvard's "first read" of application materials; (4) admissions
officer and alumni interviews; (5) subcommittee meetings of
admissions officers to recommend applicants to the full admissions
committee; and (6) full admissions committee meetings to make and
communicate final decisions to applicants. Harvard also uses a
system of "tips" for certain applicants. Tips may be considered
during or after the third stage.
Pre-Application Recruitment Efforts by Harvard
Harvard engages in significant recruitment efforts.
Each admissions cycle, it buys the names and contact information
of students who do well academically and on standardized tests
from ACT and the College Board, the two main organizations that
administer standardized tests used in college admissions. Harvard
uses this information to assemble a "search list." That list in
relevant years has had more than 100,000 students. Students on
3 Because SFFA's allegations of discrimination relate only
to United States citizens or permanent residents, the following
discussion is limited to the application process for domestic
applicants. The statistical analyses performed by the parties'
experts exclude data from foreign applicants.
- 9 -
the search list receive communications encouraging them to
consider applying to Harvard. A student's presence on the search
list has no effect on whether Harvard will admit them.4 Harvard
purchases information about high school students of all races,
including Asian Americans. It purchases information for some
groups of students, such as African American or Hispanic students
and for students from states Harvard has labeled "Sparse Country,"5
who have lower standardized test scores than other students.
Harvard specifically recruits minority students --
including African American, Hispanic, and Asian American students
-- through its Undergraduate Minority Recruitment Program
("UMRP"). Low-income applicants and those who are the first in
their family to go to college are encouraged to apply through the
Harvard Financial Aid Initiative ("HFAI"). Despite these efforts
to expand its applicant pool, the demographics of Harvard's
4 Students on the search list are twice as likely to be
admitted to Harvard as students who are not on the search list.
But correlation does not imply causation. Because the search list
mechanically includes students who do well academically and have
high test scores -- students who would be stronger applicants to
Harvard than those who have less impressive academic and testing
credentials -- it does not follow that being on the search list
causes these students to be admitted at higher rates.
5 Harvard uses "Sparse Country" as a shorthand for areas
of the United States that are sparsely populated. Sparse Country
includes twenty states: Alabama, Alaska, Arizona, Arkansas, Idaho,
Louisiana, Maine, Mississippi, Montana, Nebraska, Nevada, New
Hampshire, New Mexico, North Dakota, Oklahoma, South Dakota, Utah,
Vermont, West Virginia, and Wyoming.
- 10 -
applicants and its admitted students do not mirror those of the
United States on many criteria, including race.
Information Contained in Submitted Applications
Applicants apply to Harvard by Early Action or Regular
Decision. The review process is identical for both. Early Action
students receive admissions decisions more quickly.
Harvard uses the Common Application, a standardized
application that applicants fill out once to apply to different
colleges and universities of their choice. As part of the Common
Application, students submit a great deal of information,
including about their standardized test scores, transcripts,
extracurricular and athletic activities, awards, parents' and
siblings' educational information, parents' occupations and
marital status, teacher and guidance counselor recommendations,
intended field of study, personal statement, and additional
supplemental essays or academic material. Some information --
like racial identity -- can, but need not, be submitted.
Consideration of Submitted Applications: Harvard's "First
Read" of Application Materials
During the admissions cycles directly challenged by
SFFA, Harvard's admissions office was led by William Fitzsimmons,
the Dean of Admissions and Financial Aid, Marlyn McGrath, the
Admissions Director, and Sally Donahue, the Financial Aid
Director. Harvard staffs its admissions office with approximately
- 11 -
seventy people, forty of whom are admissions officers. Admissions
officers read applications and decide which applicants will be
offered admission. New admissions officers are trained and
supervised by more experienced admissions officers. This training
includes instruction on how to consider race.
All admissions officers are given a copy of Harvard's
reading procedures, which explain how to evaluate applications.6
The reading procedures include guidelines for assigning numerical
ratings to applicants in certain categories. The guidelines detail
the factors admissions officers should consider when assigning
these numerical ratings.
Before admissions officers begin reading applications,
they divide the applications based on geography. Admissions
officers are organized into subcommittees dedicated to reviewing
applications from specific regions.7 Within each subcommittee,
admissions officers further specialize by reading all applications
from certain high schools. Each high school has a dedicated "first
reader" for applications from that high school. First readers do
6 The following descriptions reflect the reading
instructions given to admissions officers for the class of 2018
and were in effect during the period analyzed via statistical
analysis in this lawsuit. Harvard has since revised them for later
class years. Any relevant changes are discussed below.
7 To limit inter-geographic discrepancies in application
evaluations, each admissions officer serves on at least two
subcommittees.
- 12 -
not always have an applicant's complete application when reviewing
a file. For example, an application might be missing an alumni
interview report or a teacher recommendation when it is first
reviewed. Every application has a set of numerical ratings
assigned by the first reader. Some applications are also given
additional rounds of ratings by additional readers.8 Faculty
members might also review files.9
There are six types of ratings assigned during the
reading stage: academic ratings, extracurricular ratings,
athletic ratings, school support ratings, personal ratings, and
overall ratings. Each rating is numeric. Higher numbers often,
but not always,10 indicate a worse rating. Each of the six
categories has subjective components and none are formulaically
8 New admissions officers have their first fifty
applications reviewed by second readers. Otherwise, second
readers are rare and used if the case is complex, if the case
raises issues of policy, or if the case would be helped by a second
reader's subject-matter knowledge or geographic expertise. The
chair of the docket may also serve as a third reader. This title
is a misnomer, as an application need not receive a second read to
be given a third read.
9 Faculty members review files if they have specific
expertise relevant to an application. For example, a Visual and
Environmental Studies professor might review an applicant's film.
Some faculty members also volunteer to review files from certain
regions.
10 In some categories, ratings of five or six denote special
circumstances (e.g., home or employment responsibilities that
preclude participation in extracurricular activities) and are not
necessarily worse than lower numbers.
- 13 -
assigned. Admissions officers fine tune their ratings with "+"
and "-" marks. For example, an applicant with a "3+" academic
rating would be stronger than an applicant with a "3", who would
in turn be stronger than an applicant with a "3-" rating. Ratings
of "1" are exceedingly rare and are predictive of admission in
most categories.11 These ratings are preliminary and do not dictate
a student's admission to Harvard. It is not uncommon for students
with worse ratings to be admitted over students with better
ratings. We describe the six categories to which these numeric
ratings are assigned.
Academic Rating
The academic rating is designed to quantify a student's
academic ability and is based largely, though not exclusively, on
grades and standardized test scores. Other factors that go into
the rating include less quantifiable characteristics like a
student's demonstrated love of learning, potential for future
academic improvement, teacher and guidance counselor letters and,
if available, academic or faculty evaluations.
11 During the period analyzed by the parties' experts, .07%
of applicants received an overall rating of 1 and all were
admitted. Of the .45% of applicants who received an academic
rating of 1, 68.9% were admitted. Of the .31% of applicants who
received an extracurricular rating of 1, 49.1% were admitted. Of
the .82% of applicants who received an athletic rating of 1, 86%
were admitted. And of the .03% of students who received a personal
rating of 1, 69.6% were admitted. Schools support ratings of 1
are relatively more frequent and less predictive of admission.
- 14 -
An academic rating of "1" signifies "summa cum laude"
potential. Harvard's reading procedures describe "1" applicants
as genuine scholars with near-perfect standardized test scores and
grades. They may have even demonstrated an ability to produce
original scholarship. "2" applicants have "magna cum laude"
potential and are excellent students with top grades. If they
submit SAT scores, they are typically in the mid- to high-700s in
each testing area. ACT scores for "2" students are typically at
or above 33. "3" students have "cum laude" potential and have SAT
scores in the mid-600s to low-700s or ACT scores between 29 and
32. Finally, "4," "5," and "6" applicants have "adequate
preparation," "marginal potential," and "marginal or worse"
achievement, respectively, with lower grades and standardized test
scores.
Extracurricular Rating
The extracurricular rating measures a student's
commitment to non-academic pursuits, broadly defined. A score of
"1" indicates "truly unusual achievement," a score of "2" reflects
"strong secondary school contribution in one or more areas such as
class president, newspaper editor, etc.," a score of "3" represents
"solid participation but without special distinction," and a score
of "4" corresponds to little or no participation. The reading
procedures instruct admissions officers to give scores of "5" or
- 15 -
"6" if special circumstances precluded the applicant from engaging
in more traditional extracurriculars.
Athletic Rating
The athletic rating measures an applicant's commitment
to athletic pursuits. "1" ratings are reserved for recruited
varsity athletes. A "2" rating represents a "[s]trong secondary
school contribution in one or more areas" of athletics with
"possible leadership role(s)." A "3" is "active participation,"
and "4" indicates little to no participation. Unlike "4" ratings
in other categories, an athletic rating of "4" is not a negative.
"5" or "6" ratings indicate circumstances that hinder or prevent
participation in athletics.
School Support Ratings
Admissions officers also assess the strength of an
applicant's high school support by reading teacher and guidance
counselor recommendations. Each recommendation receives its own
rating. "1" indicates "strikingly unusual support," typically
indicated by phrases in recommendations like "the best ever." A
"2" corresponds to "very strong support," a "3" is "above average
positive support," "4" is "somewhat neutral or slightly negative,"
and a "5" is "negative or worrisome." Admissions officers assign
general school support ratings, but teacher and guidance counselor
recommendations can also affect other rating scores (e.g., the
academic and personal ratings).
- 16 -
Personal Rating
The personal rating features heavily in this litigation.
It attempts to measure the positive effects applicants have had on
the people around them and the contributions they might make to
the Harvard community. Factors considered include an applicant's
perceived leadership, maturity, integrity, reaction to setbacks,
concern for others, self-confidence, likeability, helpfulness,
courage, kindness, and whether the student is a "good person to be
around." Admissions officers generally assess an applicant based
on the applicant's admissions essays, teacher and guidance
counselor recommendations, accomplishments, and alumni interview
report, but almost any information in a student's application can
factor into the personal rating.
According to Harvard's written reading procedures, a
score of "1" is "outstanding," a score of "2" is "very strong,"
and a score of "3" is "generally positive." Applicants who receive
ratings of "4," "5," or "6" are typically described as "bland or
somewhat negative or immature," having "questionable personal
qualities," or having "worrisome personal qualities,"
respectively.12
12 These descriptions were updated in 2018 to be more
descriptive. For example, in the updated instructions, an
applicant with a "1" personal rating is described as follows:
"Truly outstanding qualities of character; student may display
enormous courage in the face of seemingly insurmountable obstacles
in life. Student may demonstrate a singular ability to lead or
- 17 -
Harvard maintains that race itself does not play a role
in a student's numerical personal score. It does admit that
experiences tied to an applicant's race -- for example, experiences
with prejudice or discrimination and how the applicant has overcome
this adversity -- could inform their personal rating. Before 2018,
the reading procedures Harvard distributed to its admissions
officers did not mention whether race should be included in
assigning the personal rating. After SFFA brought this suit,
Harvard modified these instructions to explicitly say that "an
applicant's race or ethnicity should not be considered in assigning
the personal rating." These updated instructions took effect
beginning with the class of 2023.
Overall Rating
Before the admissions officers take a candidate to the
committee stage, they assign an overall rating to each applicant.
This rating takes all available information into account and is
not a formulaic weighting of the other ratings. Harvard's reading
procedures include guidelines for assigning an overall rating: a
"1" signifies an "exceptional" candidate with >90% chance of
admission; a "2" is a strong student with a 50-90% chance of
admission; a "3" is a "solid contender" with a 20-40% chance of
inspire those around them. Student may exhibit extraordinary
concern or compassion for others. Student receives unqualified
and unwavering support from their recommenders."
- 18 -
admission, a "4" is a "neutral" candidate with respectable
credentials, and a "5" is an applicant with below-average
credentials.
Unlike the other ratings, Harvard acknowledges that
admissions officers can and do take an applicant's race into
account when assigning an overall rating. Race can be, but need
not be, a basis for a "tip." So can other factors. We discuss
tips later, after a description of interviews and the committee
deliberations.
Admissions Officer and Alumni Interviews
Concurrent with the admissions office's review of
applications, Harvard uses its alumni and admissions officers to
interview applicants. Harvard provides all interviewers with an
Interview Handbook. The Handbook provides guidance about what
information interviewers should not discuss (e.g., the applicant's
chance of admission) or questions they should not ask (e.g.,
interviewers "should not ask questions that suggest students are
being ethnically screened or go through a 'special' admissions
process"). Alumni interviewers have access to some of the same
information as Harvard's admissions officers. The information
they receive does not include teacher and guidance counselor
recommendations or transcripts. After conducting an interview,
alumni give written comments and assign applicants numerical
scores that are broadly similar to the scores assigned by an
- 19 -
admissions officer. However, alumni only assign academic,
personal, extracurricular, and overall scores. They do not assign
athletic or school support ratings.
Subcommittee Meetings and Recommendations to Full Committee
After reading applications, assigning preliminary
scores, and gathering any additional information from faculty and
alumni, Harvard decides those applicants to whom it will offer
admission. Admissions subcommittees meet to discuss the
applicants in their region over three to five days. First readers
typically serve as advocates for applicants they believe should be
admitted. Not every applicant is discussed, but it is common for
members of the committee to discuss candidates not singled out by
first readers.
The subcommittees make recommendations to the full
admissions committee. Their recommendations are not dispositive.
The final decisions are made only at the full committee meeting.
It is not uncommon for applicants who were not recommended for
admission by a subcommittee to later be admitted (and vice versa).
The subcommittees consider a variety of factors, including race,
when making their recommendations.
Full Committee Meetings and Final Decisions on Offers to Admit
Harvard then holds a full committee meeting -- comprised
of forty members -- to finalize decisions. Any applicant may be
discussed at the full committee meeting, though the focus is
- 20 -
typically on those recommended by the various subcommittees. All
committee members have access to each student's application. This
application frequently includes additional information that was
not available when the first reader assigned ratings to the
applicant. The applicants presented are discussed one by one, and
every member of the full committee votes on admission. An
applicant must secure a majority of the full committee's vote to
be offered admission.
After the full committee votes on applications, it
typically has a pool of more than 2,000 tentative admits, more
than can be admitted. To finalize the class, it conducts a "lop
process" to winnow down the pool even further. Before deciding
which applications will be lopped, members of the admissions
committee are informed of various demographic characteristics of
the admitted applicants, including race. Admissions officers then
compile a "lop list" of applicants who might be lopped. This list
includes information about tentative admits -- race, athletic
rating, legacy status, and socioeconomic status -- relating to
some of Harvard's admissions tips. After enough applicants have
been lopped, Harvard sends decisions to applicants.
"Tips" Given to Benefit ALDC Applicants and Non-ALDC
Applicants
Since at least 1990, Harvard has used a system of "tips"
in its application review process. Tips are plus factors that
- 21 -
might tip an applicant into Harvard's admitted class. The tip
system is an overlay of Harvard's process and tip factors can be
considered at multiple points in Harvard's review. Tips can be
incorporated in Harvard's ratings and during the committee review
stage.
Harvard has memorialized a non-exhaustive list of tip
factors in a handbook it gives to admissions officers and alumni
interviewers. The factors contained in Harvard's written
materials are outstanding and unusual intellectual ability,
unusually appealing personal qualities, outstanding capacity for
leadership, creative ability, athletic ability, legacy status, and
geographic, ethnic, or economic factors. Harvard admits that race
is considered.
Harvard's Consideration of Race
Harvard's use of tips that take race into account is the
focus of many of SFFA's claims. We consider how and when Harvard
claims to consider race. It admits that race can be considered
during Harvard's "first read" of application materials only when
assigning an applicant's overall rating. It also admits that an
applicant's race can be considered in both subcommittee and full
committee meetings. Harvard denies that race is considered in
assigning an applicant's personal rating during the "first read."
We describe the background against which Harvard's tip
taking race into account is used. Admissions officers are
- 22 -
provided, from time to time, with summaries containing demographic
information. These "one-pagers" provide a snapshot of various
demographic characteristics of Harvard's applicant pool and
admitted class and compares them to the previous year. In addition
to race, these sheets summarize the applicant pool on a variety of
other dimensions (e.g., gender, geographic region, intended
concentration, legacy status, whether a student applied for
financial aid, etc.). Information from this sheet is periodically
shared with the full admissions committee, and the committee uses
this information in part to ensure that there is not a dramatic
drop-off in applicants with certain characteristics -- including
race -- from year to year. Harvard keeps abreast of the racial
makeup of its admitted class in part because doing so is necessary
to forecast yield rates. The yield rate is the percent of admitted
applicants who accept an offer of admission.13 Empirically, Asian
American and white students accept offers of admission at higher
rates than African American, Hispanic, Native American, and
multiracial applicants.
13An alternative interpretation of "yield rate by race" is
the number of students of a particular race admitted divided by
the total number of students admitted. We discuss this statistic
when reviewing the district court's findings of fact.
- 23 -
ALDCs: Athletes, Legacy Applicants, Dean's Interest List
Applicants, and Children of Faculty or Staff
Harvard also provides tips to ALDC (recruited athletes,
legacy applicants, applicants on the Dean's Interest List,14 and
children of faculty or staff) applicants. ALDC applicants make up
less than 5% of applicants to Harvard but around 30% of the
applicants admitted each year. These applicants have a
significantly higher chance of being admitted than non-ALDC
applicants. SFFA does not challenge the admission of this large
group of applicants who can and do receive tips.
The racial makeup of ALDC applicants is different than
non-ALDC applicants. Around 67.8% are white, 11.4% are Asian
American, 6.0% are African American, and 5.6% are Hispanic. In
contrast, only 40.3% of non-ALDC applicants are white, 28.3% are
Asian American, 11% are African American, and 12.6% are Hispanic.
3) Prior Analyses of Harvard's Use of Race in Harvard's
Admissions Process
Because SFFA argues, and Harvard denies, that its claims
are supported by prior reviews of Harvard's admissions policies,
we describe those analyses.
14 The Dean's Interest list is a list of applicants the
Dean of Admissions gives special attention to. It primarily
includes the children or relatives of donors, and it includes a
rating of how important the donor is to Harvard.
- 24 -
1990 U.S. Department of Justice Office of Civil Rights ("OCR")
Report
In 1990, OCR investigated potential claims of anti-Asian
bias in Harvard's admissions process but concluded there was no
difference in the numeric ratings given to white and Asian American
applicants.
At the time, Harvard reviewed applications using a
process like the one challenged in this lawsuit in the sense that
it used a rating system and tips for race and ALDC applicants.
OCR's report stated that "descriptions of Asian American
applicants were found that could have implications for the
stereotyping of Asian American applicants." But "[t]hey could not
be shown to have negatively impacted the ratings given to these
applicants." The OCR report concluded that Harvard did not use a
quota system for Asian Americans, that its rating system was not
designed or implemented to harm Asian Americans, and that, "taken
as a whole, there was no significant difference between the
treatment of Asian American applicants and the treatment of white
applicants."
Harvard's Office of Institutional Research Studies
Harvard has conducted its own studies of its admissions
process in the past decade.
The first of these -- commissioned in part due to a
widely publicized November 2012 article by Ron Unz in The American
- 25 -
Conservative titled "The Myth of the American Meritocracy" -- was
conducted by Harvard's Office of Institutional Research (OIR).
This review started in December 2012 and went through several
stages. OIR's initial rough, preliminary logistic regression
models, which omitted many variables actually considered in
Harvard's admission process (like socioeconomic or family
circumstances), showed that Harvard was better able to predict the
racial makeup of its admitted class if it included race as an
explanatory variable in the regression.
After this analysis by OIR, Harvard's Dean of Admissions
requested further analysis on Harvard's use of low-income tips.
OIR's next analysis used application data for the classes of 2009
to 2016 and found that "low income students clearly receive a 'tip'
in the admissions process" but that "the tip for legacies and
athletes is larger" than for low income applicants. It also found
that "there are demographic groups that have negative effects."
The admittedly limited model suggested but did not conclude that
Asian American applicants were statistically significantly less
likely to be admitted to Harvard than white applicants. OIR
explained that the model "has several limitations" and that
expanding the set of explanatory variables in the model "might
result in a better fitting model, one that accounts for more of
the variation in individual applicants and their potentially
unique contributions to the entering class."
- 26 -
Harvard's Dean of Admissions then asked OIR to
investigate whether Harvard's tip for low-income applicants was
consistently applied across applicants of different races.
Specifically, given the concerns raised about Harvard's treatment
of Asian American applicants, he wanted an analysis of the effect
of being both Asian American and low-income. OIR inserted an
interaction term into the logistic regression model used in its
previous report in order to capture the effect of being both Asian
American and low income.15 This term had a statistically
significant positive coefficient, suggesting that, holding all
else constant, being both Asian American and low income results in
an applicant having a higher chance of admission than the sum of
the effect of being Asian American or low-income alone suggests.
This model, however, admittedly suffered from the same problems
and limitations as OIR's earlier models in that it omitted many
variables that Harvard actually considered in its admissions
process.
15 An interaction term between two categorical (i.e., non-
numeric) variables in a logistic regression model captures whether
the effect of having one characteristic changes when another
characteristic is also present. See Reference Manual on Scientific
Evidence [hereinafter, "FJC Reference Manual"] (3d ed. 2011) at
316-17. Here, the interaction term measures whether being both
Asian American and low-income simultaneously results in a greater
or lower change in an applicant's admission chances beyond the
effects of being either Asian American or low-income
independently.
- 27 -
4) Harvard's Articulation of Its Justification for Its Use of
Race to Accomplish Diversity
The Khurana Committee
In 2015, Harvard created the "Committee to Study the
Importance of Student Body Diversity" (the "Khurana Committee").
This committee was chaired by Rakesh Khurana, the dean of Harvard
College. The report it issued (the "Khurana Report") was part of
Harvard's effort to explain the benefits Harvard derives from
diversity, including racial diversity, as required by Supreme
Court precedent. Indeed, the Khurana Committee recognized that
"[t]he question before [it] is one the Supreme Court has asked
public institutions of higher education to answer in connection
with the consideration of an applicant's race" in admissions.
The Khurana Committee sought input from a variety of
sources at Harvard. In reaching its conclusions, it relied on
discussions with students, alumni, faculty, athletic coaches,
extracurricular advisors, the deans of admissions and student
life, residential faculty deans, and other committees on
admissions, financial aid, and educational policy. Based on its
findings, it "emphatically embrace[d] and reaffirm[ed] the
University's long-held . . . view that student body diversity --
including racial diversity -- is essential to [Harvard's]
pedagogical objectives and institutional mission."
- 28 -
The Khurana Report began by reiterating the mission of
Harvard College: "to educate the citizenry and citizen leaders for
our society." It detailed how Harvard has pursued this mission
over time and how student body diversity is essential to achieving
it. At its founding, Harvard's student body was limited by the
socioeconomic, geographic, and political factors of the 17th
century. Harvard educated the small subset of men who had
completed high school, had the means to travel to Harvard, and
wanted to pursue post-secondary education.
Leading up to the Civil War, Harvard's President called
for increased diversity. In a report to Harvard's Board of
Overseers, he recognized the "disastrous condition of public
affairs." He wrote that Harvard must assemble a student body "from
different and distant States" to "remove prejudices, by bringing
them into friendly relations."
Successive presidents of Harvard reaffirmed the
importance of diversity. In the late 19th century, President
Charles William Eliot revised the Harvard curriculum to promote
tolerance, mutual understanding, and camaraderie. He wrote that
diversity "promotes thought on great themes, converts passion into
resolution, cultivates forbearance and mutual respect, and teaches
. . . candor, moral courage, and independence of thought." A
century later, President Neil Rudenstine praised these reforms in
his own 1996 evaluation of diversity at Harvard. He wrote that
- 29 -
President Eliot recognized that a diverse student body was "capable
of shaping lifelong attitudes and habits" and "indispensable to
the healthy functioning of a democratic society." In 2015, then-
President Drew Gilpin Faust said at the beginning of Harvard's
academic year that diversity is "an integral part of everyone's
education." She said she meant not just racial diversity but also
other types, including religious, ethnic, national, political, and
gender diversity. By "everyone," she said she meant not just
students but also the wider Harvard community, including faculty
and staff.
The Khurana Report made Harvard's reasons for
accomplishing diversity by its limited use of race even more
explicit. In our view, at least these specific goals are
articulated in this report: (1) training future leaders in the
public and private sectors as Harvard's mission statement
requires; (2) equipping its graduates and itself to adapt to an
increasingly pluralistic society; (3) better educating its
students through diversity; and (4) producing new knowledge
stemming from diverse outlooks.
The Khurana Report recognized that Harvard, historically
and currently, trains leaders. Signers of the Declaration of
Independence attended Harvard. Today, Harvard College graduates
"are founding and running global companies in technology, retail,
finance, and healthcare, among others." Graduates will become
- 30 -
leaders not just in the United States but throughout the world.
The Khurana Report stated that global companies want to hire and
promote graduates who have been educated in diverse settings.
Businesses have internalized diversity and inclusion as core
values. It concluded that if Harvard does not provide its students
with the opportunity to engage with other students in a diverse
undergraduate environment, "[its] students likely would be
constrained in their pursuit of excellence, and [Harvard] would be
remiss in failing to provide them with the skills they need to
flourish after graduation."
The Khurana Report also stated that providing a diverse
learning environment is necessary not only for Harvard to prepare
its students to become leaders but also for it to train leaders at
all. It quoted the Supreme Court in Grutter v. Bollinger: "In
order to cultivate a set of leaders with legitimacy in the eyes of
the citizenry, it is necessary that the path to leadership be
visibly open to talented and qualified individuals of every race
and ethnicity." 539 U.S. 306, 332 (2003). The Khurana Report
said that Harvard "must demonstrate to the world that individuals
of all races, ethnicities, all backgrounds may thrive, succeed,
and lead" to continue to train leaders. If Harvard cannot produce
a diverse set of graduates, it may no longer produce leaders.
Next, the Khurana Committee said that Harvard's students
graduate into a world that has become increasingly diverse. The
- 31 -
Khurana Report quoted from the amicus brief Harvard filed in Fisher
II: "Whatever their field of endeavor, Harvard's graduates will
have to contend with a society that is increasingly complex and
influenced by developments that may originate far from their homes.
To fulfill their civic and other responsibilities, Harvard's
graduates cannot be blind either to the challenges facing our
increasingly pluralistic country or to the unresolved racial
divisions that stubbornly persist despite decades of substantial
efforts to resolve them." The Khurana Report said student body
diversity encourages students to "examine ways of processing the
world dissimilar to their own" and that through classroom
discussion with others "one learns to negotiate pluralism." The
Khurana Report said that racial diversity in particular is
important because Harvard's graduates enter a society where
"negative life experiences attributable to differences in racial
and ethnic heritage" are still commonplace. It concluded that
Harvard "would fail in a foundational aspect of [its] mission if
[it] disregarded that fact as [it] prepare[s] [its] students for
such a complex and heterogeneous society."
Third, the Khurana Committee detailed how diversity is
important to Harvard's goal of educating students, faculty, and
staff. One of these goals is transformational. In addition to
encouraging students to embrace their own identities, Harvard
wants its students to take on an additional, complementary one.
- 32 -
It wants them to become members of "the community of educated men
and women." Harvard views this identity as "inclusive of but not
bounded by race or ethnicity." It "is sensitive to and
understanding of the rich and diverse range of others' identities"
and "opens empathic windows to imagining how other identities might
feel." To achieve this transformation, Harvard creates
opportunities for students to interact with others who are unlike
themselves.
The Khurana Report cited a book, Making the Most of
College, by Professor Richard J. Light, a professor at Harvard's
Graduate School of Education. The book, based on in-depth
interviews of thousands of Harvard students, provided examples of
how diversity, including racial diversity, can promote learning,
empathy, and understanding. The Khurana Report observed that
certain conversations described in the book "likely could not have
occurred as candidly as [they] did" at Harvard had "any or all of
[the students involved] been white." The Khurana Committee
recognized that certain learning environments are not possible
unless some students "share[] a relationship to the issue that
[others] simply [do] not have." And the educational benefits of
diversity cannot be obtained "[i]f the only contact students had
with others' lived experiences was on the page or on the screen"
because "it would be far too easy to take short cuts in the exercise
of empathy, to keep a safe distance from the ideas, and the people,
- 33 -
that might make one uncomfortable." Personal contact is essential.
It also said that the conversations in the book make clear that
students who have some characteristics in common can and do have
different perspectives from each other. The Khurana Report drew
two conclusions: (1) race "plays an irreplaceable role in
[Harvard's] conception of a diverse student body" and (2) Harvard
cannot and does not treat race monolithically because students of
the same race do not "share the same views, experiences, or other
characteristics."
Finally, the Khurana Committee said that diversity is
important to producing new knowledge. Its report quoted from an
amicus brief filed by Harvard and other colleges and universities
in Regents of the University of California v. Bakke: diversity
"broadens the perspectives of teachers and thus tends to expand
the reach of the curriculum and the range of scholarly interests
of the faculty." The Khurana Report said that entirely new fields
of study, like Women's Studies, Latin American Studies, and Labor
Studies, were created in part because universities like Harvard
admitted more women, Latinx, and low-income students. It also
said that older fields -- like Classics and Philosophy -- were
reshaped by "new perspectives on gender and sexuality, race,
ethnicity, and social class." And it said that the social
sciences, natural sciences, and medicine have all been transformed
by a more diverse student body. A homogeneous research community
- 34 -
leads to "shared blind spots" that different perspectives can
reveal. It concluded that diversity is necessary to create
knowledge.
5) Harvard's Committees to Study Race-Neutral Alternatives
The Ryan Committee
In 2014, in response to the creation of an external,
unaffiliated website accusing its admissions process of being
unfair, Harvard formed a committee to examine race-neutral
alternatives to its race-conscious admissions process (the "Ryan
Committee"). The Ryan Committee was chaired by Professor James
Ryan, then the Dean of Harvard's Graduate School of Education.
This committee met only a few times before disbanding in December
2014.
The Smith Committee
In June 2017, Harvard established the "Committee to
Study Race Neutral Alternatives in Harvard College Admissions"
(the "Smith Committee"). This committee was chaired by Michael
Smith, then the Dean of Harvard's Faculty of Arts and Sciences.
Like the Khurana Committee, the Smith Committee was part of
Harvard's effort to ensure compliance with Supreme Court
precedent. It issued a report (the "Smith Report") examining
whether Harvard could achieve its interest in diversity without
considering race in its admissions process.
- 35 -
The Smith Report began with a summary of the ways Harvard
already recruits diverse applicants. Harvard identifies low-
income students and encourages them to apply with informational
mailings highlighting Harvard's financial aid program. Harvard's
admissions process is need-blind, which means that an applicant's
inability to pay tuition will not harm their chances of being
admitted. It sends representatives -- admissions officers,
current undergraduates, and alumni -- to conduct recruitment
events throughout the United States. These events include
secondary schools and regions that have not historically sent
students to Harvard. It has created a "First Generation" program
to encourage students who would be the first in their family to
attend college to apply. As part of this program, it connects
current first-generation Harvard students with potential
applicants to answer their questions. As discussed earlier,
Harvard encourages racially diverse applicants to apply through
UMRP and low-income applicants to apply through HFAI.
Once it has admitted diverse applicants, Harvard
encourages them to matriculate. It provides admitted students
with a price calculator so that they can determine if Harvard is
affordable. And it has worked to make Harvard affordable. Since
2004, Harvard has steadily reduced the effective tuition that low-
income students pay. Since 2012, Harvard has expected no parental
contribution from families earning less than $65,000 a year and a
- 36 -
parental contribution limited to 10% of family income for students
from families earning less than $150,000 a year. It widely
publicizes this information. Harvard also hosts a weekend-long
on-campus event for admitted students. It provides financial aid
to low-income admitted applicants to ensure that they can attend
the event. Admittees are encouraged to meet each other and current
students, including those with similar backgrounds. To facilitate
this, Harvard hosts an Economic Diversity and First Generation
Students Reception and a multicultural reception.
Expanding Recruitment or Modifying Admissions Criteria
The Smith Committee then considered six of SFFA's race-
neutral proposals to increase diversity by expanding Harvard's
recruitment efforts or modifying its admissions criteria: (1)
increasing efforts to recruit racially and socioeconomically
diverse students; (2) expanding partnerships with schools or
organizations that serve applicants of modest socioeconomic
backgrounds; (3) increasing financial aid; (4) adopting place-
based preferences; (5) increasing transfer admissions; and (6)
increasing the weight for socioeconomic background.
The Smith Report concluded that Harvard already devotes
significant resources to recruitment efforts and that expanding
them further would not increase diversity. It said that a more
racially diverse applicant pool is itself not helpful. Harvard
needs a recruitment process that is directed at students who are
- 37 -
likely to be accepted. Expanding the applicant pool beyond those
students would increase the number of disappointed applicants and
could even discourage younger students from applying in the future.
Harvard also reevaluates these programs at least twice a year to
determine how they can be improved or expanded.
It also concluded that expanding partnerships with
schools or organizations that serve applicants of modest
socioeconomic means would be insufficient for it to meet its
diversity goals. Harvard already engages in significant outreach.
The Report concluded that Harvard's "current efforts are so
substantial that we do not believe that seeking out additional
partnerships of this nature, or deepening current partnerships
could yield more than an incrementally small number of applicants
who would be admitted to Harvard and would not otherwise have
applied."
Next, the Committee determined that increasing financial
aid further would not increase diversity. According to Harvard's
financial aid office, 90% of families already pay the same or less
to send their child to Harvard than they would pay to send their
child to a state school. 70% of African American students'
families and 60% of Hispanic students' families qualify for zero
parental contribution under Harvard's financial aid program. The
Committee found no evidence that "members of any racial or ethnic
group are choosing to attend other schools instead of Harvard on
- 38 -
the basis of the need-based financial aid available at those
institutions." It relied on statistical evidence showing that
Harvard's previous expansions of financial aid did not result in
significant increases in the number of African American or Hispanic
applicants or admitted students. It concluded that increasing
financial aid was not a viable race-neutral alternative.
The Smith Committee also rejected place-based
preferences like admitting the top student from each zip code or
high school. It found such preferences "fundamentally
incompatible with the core mission of the Harvard admissions
process, which is to recruit, admit, and enroll the most
extraordinary students in the world, wherever they may be found."
Such preferences were also practically impossible to implement, as
Harvard "does not have room to admit even one student" from every
high school or zip code in the United States. And even if Harvard
did have room, the committee found that it would be impossible to
identify the best student from each area because Harvard does not
rank applicants on a single dimension.
The Smith Committee found that increasing the number of
transfer students would also not increase diversity. Harvard's
ability to admit transfer students is limited by the space it has.
98% of undergraduate students live on campus. Some years, it does
not have space for any transfer students at all. To increase the
transfer students it admits, Harvard would need to admit fewer
- 39 -
freshmen to reserve spots for transfer students. This would only
increase diversity if the transfer applicant pool were more diverse
than Harvard's freshman applicant pool. But the Smith Committee
found that the transfer pool is "less diverse and less impressive
than the pool of freshman applicants."
Finally, the Smith Committee found that increasing the
weight it places on socioeconomic background in admissions would
not further Harvard's diversity goals. Harvard "believes that
excellence can and should be found in all backgrounds" and
determined that "[a] focus on socioeconomic circumstances that
outweighed all other factors could equally reduce the depth and
breadth of the Harvard class as well as its excellence in many
dimensions." In order to reach a level of racial diversity similar
to what it currently achieves, Harvard would need to give
applicants from lower socioeconomic backgrounds such an extreme
tip that it would "overwhelm other considerations in the admissions
process" and result in "significant changes in the composition of
the admitted class." Harvard would admit substantially fewer
students with the highest academic, extracurricular, personal, and
athletic ratings. As academic excellence "remains an
institutional imperative" at Harvard, the Smith Report concluded
that these changes were not feasible. The Smith Committee also
found that using socioeconomic status as a proxy for race would
result in many non-white students in Harvard's class coming from
- 40 -
modest socioeconomic circumstances. Achieving racial diversity in
this way would come "at the cost of other forms of diversity,
undermining rather than advancing Harvard's diversity-related
educational objectives."
Eliminating Admissions Practices
Next, the Smith Committee evaluated whether Harvard
should eliminate three of its core admissions practices: (1) early
action; (2) deferred admission and tips for ALDC applicants; and
(3) consideration of standardized test scores.
Harvard already tried eliminating Early Action. Harvard
thought that eliminating Early Action would "encourage an even
greater number of diverse students to apply and matriculate" and
eliminated Early Action from 2007 to 2011. It found that doing so
"reduced [its] ability to attract a broadly diverse and
academically excellent class." Considering Harvard's "direct
experience and experimentation with Early Action, the committee
[did] not believe that abolishing Early Action again would
contribute to diversity on campus."
The Smith Committee next concluded that eliminating ALDC
tips and deferred admission would not be adequate race-neutral
alternatives. If Harvard were to eliminate its consideration of
race and make no other changes, the Smith Report says that the
share of African American admitted students would be 5.6% and the
share who are Hispanic or Other would be 8.9%. If it were to
- 41 -
eliminate the consideration of race and eliminate deferred
admission and ALDC tips, the report says that the share of African
American admitted students would be 5.3% and the share who are
Hispanic or Other would be 9.3%. The Smith Committee found that
eliminating these admissions practices resulted in negligible, and
sometimes negative, changes to diversity.
The Smith Committee also defended Harvard's use of
deferred admission and ALDC tips. It found them consistent with
Harvard's values. Deferred admission allows Harvard to admit
students who would benefit from a gap year. Giving tips to
athletes allows Harvard to admit students who have demonstrated
"discipline, resilience, and teamwork" and allows Harvard to field
competitive athletic teams, which fosters a sense of community on
campus. Preferencing legacy applicants "helps to cement strong
bonds between the university and its alumni," fosters community-
building, and encourages alumni to donate their time and money to
support Harvard. Harvard's tip for "Dean's List" students is "far
too small for the cessation of any such practice to contribute
meaningfully to campus diversity." Finally, giving tips to the
children of faculty and staff is important to workforce retention.
Eliminating this tip would disadvantage Harvard relative to other
schools in recruiting faculty and staff.
Finally, the Smith Committee found that eliminating
consideration of standardized test scores would come at a
- 42 -
significant cost to Harvard's other educational objectives.
Relative to Harvard's current admitted classes, "the proportion of
students with the highest academic ratings would decline by 17%,
and the proportion of students with the highest extracurricular or
personal ratings would decline by 7%." The Smith Committee also
found that standardized tests are imperfect measures but "provide
useful information that the committee would lose if it excluded
any consideration of them."
After considering all of the race-neutral alternatives
proposed by SFFA, the Smith Committee concluded that "they will
not work at Harvard [] at this time." The Smith Committee
recommended that Harvard revisit race-neutral alternatives in
2023.
II. Procedural History and Descriptions of the District
Court's Rulings
The district court conducted a fifteen-day bench trial
from October 15, 2018 to November 2, 2018. Closing arguments were
held on February 13, 2019. Below, we describe the district court's
rulings. As to all of its conclusions, we provide further
descriptions during our legal analysis. Our description of the
district court's legal conclusions is brief because we review those
conclusions de novo. See, e.g., AcBel Polytech, Inc. v. Fairchild
Semiconductor Int'l, Inc., 928 F.3d 110, 116 (1st Cir. 2019).
- 43 -
1) The District Court Held That SFFA Had Article III Standing
Before trial, the district court held that SFFA had
associational standing to sue Harvard under Hunt v. Washington
State Apple Advertising Commission, 432 U.S. 333 (1977). See SFFA
I, 261 F. Supp. 3d at 111. It held that SFFA satisfied Hunt's
requirements because SFFA included members who had standing to sue
Harvard themselves, because the lawsuit was germane to SFFA's
purpose, and because the injunctive relief sought by SFFA does not
require the participation of its members who have standing. See
id. at 110-11. It rejected Harvard's argument that it had to apply
an additional test from Hunt -- the "indicia of membership" test
-- to determine if SFFA was a traditional voluntary membership
organization capable of having associational standing because it
found that SFFA "adequately represents the interests of its current
members without needing to test this further based on the indicia-
of-membership factors." Id. at 109.
Harvard challenges SFFA's standing in defense of the
outcome reached.
2) The District Court Held That Harvard's Limited Use of Race in
Its Admissions Program Survives Strict Scrutiny
The district court utilized Supreme Court precedent to
apply strict scrutiny to evaluate Harvard's race-conscious
admissions program. SFFA II, 397 F. Supp. 3d at 189.
Specifically, the district court articulated the requirements in
- 44 -
Fisher v. University of Texas at Austin ("Fisher II"), 136 S. Ct.
2198 (2016), Fisher v. University of Texas at Austin ("Fisher I"),
570 U.S. 297 (2013), Parents Involved in Community Schools v.
Seattle School District No. 1, 551 U.S. 701 (2007), Gratz v.
Bollinger, 539 U.S. 244 (2003), Grutter v. Bollinger, 539 U.S. 306
(2003), and Regents of the University of California v. Bakke, 438
U.S. 265 (1978). To survive strict scrutiny, Harvard's use of
race must further a compelling interest and be narrowly tailored
to do so. See Fisher II, 136 S. Ct. at 2208, 2210.
The court held that Harvard had a compelling interest in
student body diversity that was sufficiently precise to permit
judicial scrutiny. SFFA II, 397 F. Supp. 3d at 192. It credited
the Khurana Report and the witness testimony at trial. Id. at
134, 192.
It held that Harvard met its burden under strict scrutiny
and Supreme Court precedent to show its use of race in admissions
was narrowly tailored. It held that Harvard did not engage in
racial balancing, did not use race as a mechanical plus factor,
and did not have workable race-neutral alternatives. SFFA II, 397
F. Supp. 3d at 192-201.
The District Court Held That Harvard Did Not Engage in Racial
Balancing
A race-conscious admissions program cannot be narrowly
tailored if it implements a quota or racial balancing. See Fisher
- 45 -
I, 570 U.S. at 311; Grutter, 539 U.S. at 334. The district court
found that Harvard's admissions program "intends to treat every
applicant as an individual," that "[e]very applicant competes for
every seat," and that Harvard has no target numbers for members of
certain races or permissible levels of fluctuation. SFFA II, 397
F. Supp. 3d at 196. It found that Harvard's admissions office's
use of "one-pagers" did not evidence racial balancing. Id. at
197.
SFFA did not offer expert testimony to support its racial
balancing claim. Id. at 177. The court found that "the racial
composition of Harvard's admitted classes has varied in a manner
inconsistent with the imposition of a racial quota or racial
balancing." Id. at 176. It found that the share of Asian American
applicants admitted to Harvard has increased roughly five-fold
since 1980 and roughly two-fold since 1990. Id. at 177. It also
found that there had been more year-over-year variation in Asian
American applicants admitted to Harvard than there had been in
Asian American applicants to Harvard. Id.
The District Court Held That Harvard Did Not Use Race as a
Mechanical Plus Factor
If a university considers race, it cannot do so in a
mechanical way. See Gratz, 539 U.S. at 270-71. The district court
found that there was no evidence of any mechanical use of tips in
Harvard's individualized, holistic review process. SFFA II, 397
- 46 -
F. Supp. 3d at 198. It found that Harvard considers race
contextually. Id. Harvard's tip was "comparable to the size and
effect" of tips sanctioned by the Supreme Court in Grutter and
Fisher II. See id.
The District Court Held That Harvard Had No Workable Race
Neutral Alternatives
A race-conscious admissions program is not narrowly
tailored if a university uses it despite workable race-neutral
alternatives. See Fisher I, 570 U.S. at 312. The district court
found that eliminating race as a factor in admissions, without
taking any remedial measures, would reduce African American
representation at Harvard from 14% to 6% and Hispanic
representation from 14% to 9%. SFFA II, 397 F. Supp. 3d at 178.
It found that at least 10% of Harvard's class would not be admitted
if Harvard did not consider race and that race is a determinative
tip for approximately 45% of all admitted African American and
Hispanic students. Id.
The court examined six race-neutral alternatives
proposed by SFFA and statistically modeled by the parties' experts:
(1) eliminating Early Action; (2) eliminating ALDC tips; (3)
improving recruiting efforts and financial aid; (4) admitting more
transfer applicants; (5) eliminating standardized testing; and (6)
instituting place-based quotas. See id. at 179. The court also
considered combinations of these alternatives. Id. It held that
- 47 -
none of these race-neutral alternatives were workable. Id. at
183.
3) The District Court Held That Harvard Did Not Intentionally
Discriminate Against Asian Americans
At trial, the parties presented non-statistical and
statistical evidence on whether Harvard intentionally
discriminated against Asian American applicants. See id. at 153,
203.
On the non-statistical evidence, the court found that
Harvard's pre-application search list was primarily a marketing
tool, that the 1990 OCR report did not evidence racial bias, and
that the trial testimony showed that Harvard's admissions officers
were not biased and did not stereotype Asian Americans. Id. at
154-58.
The statistical evidence presented at trial was
extensive. The parties' experts16 reviewed descriptive statistics17
related to Harvard's admissions process, presented logistic
16 Harvard relied on the expert testimony of David Card, an
economics professor at the University of California, Berkeley.
See SFFA II, 397 F. Supp. 3d at 158-59 n.40. SFFA relied on the
expert testimony of Peter Arcidiacono, an economics professor at
Duke University. Id.
17 The term "descriptive statistics" refers generally to
statistics that summarize data (like the mean, median, or standard
deviation). They are often "building blocks" in statistical
analyses. See FJC Reference Manual at 213, 230.
- 48 -
regression models18 of Harvard's academic, extracurricular,
athletic, school support, and personal ratings, and created
logistic regression models of Harvard's admissions process. See
id. at 158-77.
Both Professors Card and Arcidiacono used applicant-
level admissions data19 for applicants to Harvard's classes of 2014
through 2019 to build their models. Id. at 159, 162. The dataset
included "hundreds of variables relating to each applicant's
demographic characteristics, personal background, geographic
information, test scores, high school grades, ratings assigned by
Harvard's admissions officers, and Harvard's admissions decision."
Id. at 159. The experts had access to aggregate information for
the classes of 2000 to 2017 and a sample of actual application
files for applicants from the classes of 2018 and 2019. Id.
Harvard also presented statistics on the racial make-up of its
admitted classes from 1980 to 2019. Id. at 160.
18 Logistic regression models can be used to measure the
effect of one explanatory variable (here, race) on a response
variable (here, depending on the model, Harvard's numerical
ratings or admission to Harvard) while controlling for other
explanatory variables (e.g., standardized tests scores, grades,
parental income, etc.). See generally Alan Agresti, Foundations
of Linear and Generalized Linear Models 165, 202 (2015).
Professors Card and Arcidiacono agreed that logistic regressions
are the most useful tool to understand the effect of race on
Harvard's admissions process. SFFA II, 397 F. Supp. 3d at 165.
19 Their dataset contained only domestic applicants and did
not consider transfer applicants or incomplete applications. Id.
at 159 n.41.
- 49 -
The district court first made factual findings related
to the descriptive statistics. It found that Asian Americans were
admitted to Harvard at a lower rate (between 5% and 6%) than white
applicants (between 7% and 8%) to the classes of 2014 through 2017.
Id. It found that Asian Americans tended to score better on
Harvard's academic and extracurricular ratings than white
applicants but had worse personal ratings than non-Asian American
applicants. Id. at 161-62. Both experts presented non-regression
explanations for these statistics. Id. at 162-65. The court found
problems with both and determined that logistic regression models
were the most useful tools to determine whether Harvard
discriminated against Asian Americans because these models can
isolate the effects of race by controlling for other variables
affecting the modeled outcome. Id. at 165-66.
The court analyzed two sets of logistic regression
models: the first set measured the effect of race on Harvard's
numerical ratings, while the second set measured the effect of
race on the probability of admission to Harvard. Id. at 167-72.
One of the purposes of the first set of models was to determine
whether Harvard's numerical ratings -- like the personal rating,
which SFFA argues was influenced by race -- should be included in
the second set of models. Id. at 169. The court recognized that
variables that are influenced by race should be excluded from the
second set of models. Id. at 166. But variables that are not
- 50 -
influenced by race and include information that the second set of
models would not otherwise capture should be included. Id.
To determine whether the personal rating was influenced
by race, the court considered a model of it.20 Id. at 169-71.
This logistic regression model showed that there was a negative
correlation between an applicant's personal rating and Asian
American identity even when controlling for various factors
related to admission. Id. at 169. The court found that
correlation does not imply causation. It found that the
correlation between race and the personal rating did not mean that
race influences the personal rating. Id. at 170.
The district court then found that Harvard's expert's
model including the personal rating "results in a more
comprehensive analysis." Id. at 173. Nonetheless, it considered
a model including the personal rating and a model excluding it.
Id. at 175. The model including the personal rating showed that
Asian American identity has no statistically significant effect on
an applicant's chance of admission to Harvard. Id. The model
20The court also considered logistic regression models for
Harvard's other ratings and found that the academic,
extracurricular, athletic, and school support ratings should be
included in the models because they were not influenced by race.
Id. at 167-72.
The court made other findings related to the differences in
the experts' modeling assumptions, but the court's decision on the
personal rating is the only modeling choice SFFA challenges on
appeal.
- 51 -
excluding the personal rating showed that Asian American identity
had a slightly negative effect on an applicant's chance of
admission to Harvard. Id. The district court found the
statistical evidence "inconclusive" and held that it did "not
demonstrate any intent by admissions officers to discriminate
based on racial identity." Id. Based on the non-statistical and
statistical evidence, it held that there was no intentional
discrimination. Id. at 203-04.
III. Legal Analysis
When, as here, the district court conducts a bench trial,
we review its findings of fact for clear error. See, e.g., Sawyer
Brothers, Inc. v. Island Transporter, LLC, 887 F.3d 23, 29 (1st
Cir. 2018). Under this standard, "we will set aside a trial
court's factual findings only if 'after careful evaluation of the
evidence, we are left with an abiding conviction that those
determinations and findings are simply wrong.'" Id. (quoting N.
Ins. Co. of N.Y. v. Point Judith Marina, LLC, 579 F.3d 61, 67 (1st
Cir. 2009)). We also "remain mindful that the trial court 'sees
and hears the witnesses at first hand and comes to appreciate the
nuances of the litigation in a way which appellate courts cannot
hope to replicate.'" Paraflon Investments, Ltd. v. Fullbridge,
Inc., 960 F.3d 17, 24 (1st Cir. 2020) (quoting Calandro v. Sedgwick
Claims Mgmt. Servs., Inc., 919 F.3d 26, 33 (1st Cir. 2019)). We
- 52 -
review the district court's legal conclusions de novo. See, e.g.,
AcBel Polytech, 928 F.3d at 116.
1) SFFA Has Standing to Bring This Suit, Contrary to Harvard's
Assertion
A federal court's power to decide a suit is limited by
Article III of the U.S. Constitution, which restricts federal
courts' jurisdiction to cases or controversies. See, e.g., Warth
v. Seldin, 422 U.S. 490, 498 (1975). Federal courts have "an
independent obligation to assure that standing exists, regardless
of whether it is challenged by any of the parties." Summers v.
Earth Island Inst., 555 U.S. 488, 499 (2009).
The "prerequisites for associational standing ensure
that Article III's case or controversy requirement is satisfied."
United States v. AVX Corp., 962 F.2d 108, 116 (1st Cir. 1992).
"An association has standing to sue on behalf of its members when
three requisites have been fulfilled: (1) at least one of the
members possesses standing to sue in his or her own right; (2) the
interests that the suit seeks to vindicate are pertinent to the
objectives for which the organization was formed; and (3) neither
the claim asserted nor the relief demanded necessitates the
personal participation of affected individuals." Id. (citing
Hunt, 432 U.S. at 343). Harvard does not dispute that SFFA
satisfies these requirements, and, for the reasons stated in the
- 53 -
district court's opinion, we agree that it does. See SFFA I, 261
F. Supp. 3d. at 109-11.
Instead, Harvard argues that SFFA lacks standing
because it is not a "genuine" membership organization as required
by Hunt and in doing so misreads the Supreme Court's decision in
Hunt. Its argument hinges on the resolution of two issues: (1)
whether this Court must apply the "indicia of membership" test to
determine whether SFFA is a traditional voluntary membership
organization and (2) if so, whether SFFA satisfies this test.
Because we hold that the indicia of membership test does not apply
to SFFA, we do not address whether SFFA satisfies it.
In Hunt, the Court addressed whether the Washington
State Apple Advertising Commission -- "a state agency, rather than
a traditional voluntary membership organization" -- could claim
associational standing. 432 U.S. at 344. To resolve the issue,
it introduced a threshold hurdle that certain organizations must
clear before invoking associational standing. Id. Dubbed the
"indicia of membership" test by later courts, see, e.g., Friends
of the Earth, Inc. v. Chevron Chem. Co., 129 F.3d 826, 829 (5th
Cir. 1997), this test requires courts to determine if organizations
that are not voluntary membership organizations, "for all
practical purposes, perform[] the functions of a traditional trade
association." 432 U.S. at 344. The Court identified a number of
factors to consider: whether the organization's purpose is to
- 54 -
protect and promote the interests of its non-members, whether these
non-members are "the primary beneficiar[ies] of its activities,"
and whether non-members elect its members, are the only people who
may be members, or finance the organizations' activities,
including litigation costs, through assessments levied upon them.
Id. at 344–45. If these indicia of membership are present, non-
membership organizations are deemed sufficiently similar to
traditional voluntary membership organizations and can claim
associational standing (provided that they also satisfy the
traditional three-part test). Id.
Harvard argues that we must apply the indicia of
membership test to SFFA to determine whether it is, in fact, a
traditional voluntary membership organization. SFFA argues that
because it is, on its face, a traditional voluntary membership
organization, the indicia of membership test is inapplicable. We
agree with SFFA.
Harvard cites a number of cases applying the indicia of
membership test. However, none support the proposition that the
indicia of membership test must be met here. These cases hold the
indicia of membership test is used in certain factual circumstances
which are not present in this case. See Sorenson Commc'ns, LLC v.
Fed. Commc'ns Comm'n, 897 F.3d 214, 225 (D.C. Cir. 2018) (applying
the indicia of membership test when it was "unclear if [the
plaintiff was] the sort of organization that would qualify as a
- 55 -
'membership association,'" id. at 225, and concluding that it was
not because the plaintiff was an unincorporated online
"information forum" where users could sign up for e-mail updates,
id. 223); Heap v. Carter, 112 F. Supp. 3d 402, 418 (E.D. Va. 2015)
(noting that the indicia of membership test is used to determine
whether an organization is the functional equivalent of a
membership organization and determining that the plaintiff lacked
associational standing because it "provided no details about who
the membership is or whether [it] truly can be considered a
voluntary membership organization or a functional equivalent");
Package Shop, Inc. v. Anheuser-Busch, Inc., CIV. A. No. 83-513,
1984 WL 6618, at *8 (D.N.J. Sept. 25, 1984) (applying the indicia
of membership test when the plaintiff was a self-described trade
association but its membership list was outdated and handwritten
and it had refused to follow corporate formalities). Further,
this Court has routinely applied the three-part associational
standing test without also applying the indicia of membership test.
See, e.g., Merit Const. All. v. City of Quincy, 759 F.3d 122, 126–
27 (1st Cir. 2014).
Harvard's reading of Hunt is at odds with decades of
decisions21 since Hunt that have not applied the indicia of
21 Other courts have explicitly refused to hold that every
organization claiming associational standing must pass the indicia
of membership test. See Brady Campaign to Prevent Gun Violence v.
Salazar, 612 F. Supp. 2d 1, 29 (D.D.C. 2009) (finding that "[t]he
- 56 -
membership test to organizations which, on their face, are
voluntary membership organizations.
When suit was filed in November 2014,22 SFFA was a validly
incorporated 501(c)(3) nonprofit with forty-seven members who
joined voluntarily to support its mission of "defend[ing] human
and civil rights secured by law, including the right of individuals
to equal protection under the law." We have already described its
bylaws and membership structure and do not repeat this description.
These facts are sufficient to conclude that SFFA is a valid
membership organization and applying an indicia of membership test
to SFFA is unwarranted. SFFA has associational standing to pursue
its claims.
[defendant's] standing argument is based upon a flawed reading
of Hunt" because "[t]he inquiry into the 'indicia of membership'
. . . is necessary only when an organization is not a 'traditional
membership organization'"); California Sportfishing Prot. All. v.
Diablo Grande, Inc., 209 F. Supp. 2d 1059, 1066 (E.D. Cal. 2002)
("[T]he 'indicia of membership' requirement in Hunt applies only
to situations in which an organization is attempting to bring suit
on behalf of individuals who are not members.").
22 Harvard argues that SFFA amended its bylaws after filing
suit to make itself appear more like a traditional voluntary
membership organization. Because standing must be established
when a suit is filed, we consider whether SFFA was a traditional
voluntary membership organization as of November 2014. See Davis
v. Fed. Election Comm'n, 554 U.S. 724, 734 (2008) ("[T]he standing
inquiry remains focused on whether the party invoking jurisdiction
had the requisite stake in the outcome when the suit was filed.").
- 57 -
2) Harvard's Limited Use of Race in its Admissions Program
Survives Strict Scrutiny
Because Harvard accepts federal funds, it is subject to
Title VI. 42 U.S.C. § 2000d ("No person in the United States
shall, on the ground of race, color, or national origin, be
excluded from participation in, be denied the benefits of, or be
subjected to discrimination under any program or activity
receiving Federal financial assistance."). Title VI's protections
are coextensive with the Equal Protection Clause of the Fourteenth
Amendment. See Bakke, 438 U.S. at 287 ("Title VI must be held to
proscribe only those racial classifications that would violate the
Equal Protection Clause or the Fifth Amendment."); Alexander v.
Sandoval, 532 U.S. 275, 280 (2001) (describing the preceding
language in Bakke as "[e]ssential to the Court's holding").
Harvard is subject to the same limitations on its use of race in
admissions as state-run institutions.
When Title VI applies, a university is prohibited from
considering race in its admission process "unless the admissions
process can withstand strict scrutiny." Fisher I, 570 U.S. at
309. Harvard admits that it considers race in its admissions
process and at times provides tips to applicants based on their
race. Strict scrutiny applies regardless of racial animus. See
id. Strict scrutiny requires that the university's use of race
must further a compelling interest. See Grutter, 539 U.S. at 326.
- 58 -
The Supreme Court has held that attaining student body
diversity may be a compelling interest. Fisher I, 570 U.S. at
310. Whether an asserted interest in diversity is a
constitutionally acceptable compelling interest requires that a
university have made certain showings. And even once those
showings are made, the university must also show that the means
utilized to further that interest are narrowly tailored, which
means that the university must make additional showings. SFFA
argues that Harvard fails strict scrutiny because it engages in
racial balancing, uses race as a mechanical plus factor, and has
race-neutral alternatives.23
Compelling Interest
When assessing a university's interest in student body
diversity, it has long been the law that a school "bears the burden
to prove 'that the reasons for any [racial] classification [are]
clearly identified and unquestionably legitimate.'" Id. (quoting
City of Richmond v. J.A. Croson Co., 488 U.S. 469, 505 (1989)).
"[T]he decision to pursue 'the educational benefits that flow from
student body diversity' . . . is, in substantial measure, an
23 It is not entirely clear how SFFA's arguments about
Harvard's use of race to benefit African American and Hispanic
applicants relate to SFFA's central allegation that Harvard
discriminates against Asian American applicants in favor of white
applicants. We understand SFFA's arguments as attacking the use
of race to admit African American and Hispanic candidates, to the
detriment of Asian American and white applicants.
- 59 -
academic judgment to which some, but not complete, judicial
deference is proper." Id. (quoting Grutter, 539 U.S. at 330).
But "asserting an interest in the educational benefits of diversity
writ large is insufficient." Fisher II, 136 S. Ct. at 2211. A
university's goals "must be sufficiently measurable to permit
judicial scrutiny of the policies adopted to reach them." Id.
SFFA's appellate challenge to us does not contest that Harvard has
a compelling interest in diversity. Nonetheless, Supreme Court
precedent compels us to assess whether Harvard's interest in
diversity is both clearly identified, definite, and precise.
The Fisher II majority held that that the University of
Texas's interest in diversity was definite and precise when it
identified the educational values it sought to realize through its
race-conscious admissions program. Id. These values included
promoting cross-racial understanding, breaking down racial
stereotypes, fostering a robust exchange of ideas, cultivating "a
set of leaders with legitimacy in the eyes of the citizenry,"
exposing them to different cultures, and preparing them for the
challenges of an increasingly diverse workforce. Id. The
University of Texas commissioned a year-long study and issued a
thirty-nine-page proposal giving a reasoned explanation for its
decision to pursue its diversity goals. Id. The Fisher II
majority also held that record evidence from admissions officers
reiterating this same explanation for the university's use of race
- 60 -
further supported the precision and particularity of the program.
Id.
Harvard has identified specific, measurable goals it
seeks to achieve by considering race in admissions. These goals
are more precise and open to judicial scrutiny than the ones
articulated by the University of Texas and approved by the Fisher
II majority. Harvard's interest in diversity is established both
by the Khurana Report and other statements and testimony at trial.
We rely on the Khurana Report described before and do
not repeat its description. The articulated purpose of the Khurana
Report was to enable courts to assess whether Harvard's interest
was sufficiently compelling to comply with strict scrutiny and
Supreme Court precedent. The Khurana Committee produced the report
after a thoughtful, rigorous study of the importance of diversity
to Harvard.
As for process, the Khurana Committee relied on input
and data from students, alumni, faculty and staff, and other
stakeholders in Harvard's admissions process. It considered how
a diverse environment prepares Harvard's graduates to enter the
public and private sectors and how those sectors prefer graduates
who have been exposed to a wide range of ideas and people. In our
view, at least these specific goals were articulated in the Khurana
Report: (1) training future leaders in the public and private
sectors as Harvard's mission statement requires; (2) equipping
- 61 -
Harvard's graduates and Harvard itself to adapt to an increasingly
pluralistic society; (3) better educating Harvard's students
through diversity; and (4) producing new knowledge stemming from
diverse outlooks. These goals make clear that Harvard's interest
in diversity "is not an interest in simple ethnic diversity, in
which a specified percentage of the student body is in effect
guaranteed to be members of selected ethnic groups," but "a far
broader array of qualifications and characteristics of which
racial or ethnic origin is but a single though important element."
Parents Involved, 551 U.S. at 722 (quoting Grutter, 539 U.S. at
324-25). Race is one piece of Harvard's interest in diversity.
It is "considered as part of a broader effort to achieve 'exposure
to widely diverse people, cultures, ideas, and viewpoints.'" Id.
at 723 (quoting Grutter, 539 U.S. at 330).
Testimony at trial also supported Harvard's interest in
diversity. The district court made a factual finding that "Harvard
values and pursues many kinds of diversity within its classes,
including different academic interests, belief systems, political
views, geographic origins, family circumstances, and racial
identities." SFFA II, 397 F. Supp. 3d at 133. It found that
"Harvard tries to create opportunities for interactions between
students from different backgrounds and with different experiences
to stimulate both academic and non-academic learning." Id. at
134. It based these findings on the testimony of "all of the
- 62 -
Harvard admissions officers, faculty, students, and alumni that
testified at trial." Id. at 133. Harvard's interest in student
body diversity and its consideration of race to attain it is also
not unique. Many other colleges and universities consider an
applicant's race, in addition to many other factors, in
admissions.24 And the business community has communicated its
interest in having a well-educated, diverse hiring pool both in
this case and in the prior governing Supreme Court cases.25
Harvard has sufficiently met the requirements of Fisher
I, Fisher II, and earlier cases to show the specific goals it
achieves from diversity and that its interest is compelling.
24 According to the College Board's profiles of colleges
and universities, the schools that consider race are diverse on
numerous dimensions, including in terms of religious affiliation,
location, size, and courses of study offered. For example, the
University of Notre Dame, Georgetown University, Boston College,
Villanova University, Catholic University, Lafayette College,
Gonzaga University, Marquette University, the College of the Holy
Cross, and Fordham University all consider race in their admissions
process. See College Board: BigFuture,
https://bigfuture.collegeboard.org/. The following colleges and
universities submitted an amicus brief in support of Harvard and
also consider race in their admissions process: Brown University,
Columbia University, Cornell University, Dartmouth College, Duke
University, Emory University, Johns Hopkins University, the
Massachusetts Institute of Technology, Princeton University,
Stanford University, the University of Chicago, the University of
Pennsylvania, Vanderbilt University, Washington University in St.
Louis, and Yale University.
25 An amicus brief filed by 14 leading American companies
says that "policies like those approved by the Supreme Court in
Grutter are essential to [their] ongoing efforts to attract and
benefit from the best possible people."
- 63 -
We turn to the heart of SFFA's challenge. To survive
strict scrutiny, Harvard's use of race must also be narrowly
tailored and consistent with Supreme Court precedent.
Narrow Tailoring
Narrow tailoring requires that "[t]he means chosen to
accomplish the [university's] asserted purpose must be
specifically and narrowly framed to accomplish that purpose.”
Fisher I, 570 U.S. at 311 (quoting Grutter, 539 U.S. at 333).
"[N]o deference is owed when determining whether the use of race
is narrowly tailored to achieve the university's permissible
goals." Fisher II, 136 S. Ct. at 2208 (citing Fisher I, 570 U.S.
at 311).
Accordingly, we give no deference to Harvard. Under the
Supreme Court's precedent, a university's admissions program
cannot be narrowly tailored if it (1) involves racial balancing or
quotas, see Fisher I, 570 U.S. at 311,26 (2) uses race as a
mechanical plus factor, see id. at 312, or (3) is used despite
workable race-neutral alternatives, see Fisher II, 136 S. Ct. at
2208.27 "[I]t remains at all times the University's obligation to
26 Grutter says that "[t]o be narrowly tailored, a race-
conscious admissions program cannot use a quota system" and that
racial balancing is an "unlawful interest." 539 U.S. at 323, 334.
Because the substance of these claims is similar, we review them
together under the narrow tailoring prong of strict scrutiny.
27 There is some tension in the Supreme Court's precedent
about how extensively Harvard can consider race. If Harvard's use
of race is too extensive, it could be impermissibly mechanical
- 64 -
demonstrate, and the Judiciary's obligation to determine, that
admissions processes" are narrowly tailored. See Fisher I, 570
U.S. at 311-12.
There Was No Error in Holding That Harvard Did Not Engage in
Racial Balancing
SFFA first argues that Harvard's admissions policy is
not narrowly tailored because Harvard engages in racial balancing.
The United States, as an amicus in support of SFFA, makes a similar
argument. Under the Supreme Court's precedent, racial balancing
is impermissible. See, e.g., id. at 311. A university "is not
permitted to define diversity as 'some specified percentage of a
particular group merely because of its race or ethnic origin.'"
Id. (quoting Bakke, 438 U.S. at 307). "Racial balancing is not
transformed from 'patently unconstitutional' to a compelling state
interest simply by relabeling it 'racial diversity.'" Parents
under Gratz or used beyond the "factor of a factor of a factor" in
the holistic review process approved in Fisher II. See Gratz, 539
U.S. at 272; Fisher II, 136 S. Ct. at 2207. But if Harvard's use
of race is not extensive enough, it cannot reach its diversity
goals, thus undercutting the rationale for using race at all. See
Fisher II, 136 S. Ct. at 2212 (addressing petitioner's argument
that "considering race was not necessary because such
consideration has had only a '"minimal impact" in advancing the
[University's] compelling interest'") (citation omitted). To
address this tension, the Fisher II majority held that "it is not
a failure of narrow tailoring for the impact of racial
consideration to be minor" and that "[t]he fact that race
consciousness played a role in only a small portion of admissions
decisions should be a hallmark of narrow tailoring, not evidence
of unconstitutionality." Id.
- 65 -
Involved, 551 U.S. at 732. However, universities may pay "some
attention to numbers" without "transform[ing] a flexible
admissions system into a rigid quota." Grutter, 539 U.S. at 336
(quoting Bakke, 438 U.S. at 323).
SFFA argues that, focusing on the classes of 2009 to
2018,28 "the percentage of [Harvard's] class by race always fell
within a narrow range." For these classes, the share of Asian
Americans admitted ranged from a low of 17.5% in 2013 to a high of
20.3% in 2016 with various percentages in between.
SFFA also argues that Harvard uses "one-pagers"
displaying the racial makeup of the admitted class to ensure racial
balancing and to "closely monitor the racial makeup of [its]
class." In support of this argument, it cites one Harvard
admissions officer's testimony that Harvard uses its one-pagers to
prevent "a dramatic drop-off in some group [from] last year."
First considering the broader context, the share of
admitted Asian American applicants for the classes of 1980 to 2019
has increased from a low of 3.4% in 1980 to a high of 20.6% in
2019. The share of Asian American applicants has ranged from a
low of 4.1% in 1980 and a high of 22.5% in 2014 over the same
period. The level of variation in the share of admitted Asian
28SFFA chooses to start its analysis with Harvard's class
of 2009 and end with the class of 2018 because Grutter was decided
in 2003 and it argues this is the "ten-year period between Grutter
and this suit."
- 66 -
American applicants is inconsistent with a quota, as is the fact
that the share of admitted Asian Americans co-varies almost
perfectly with the share of Asian American applicants.
Even if we were to restrict the analysis to the period
SFFA favors, the same pattern holds. The amount by which the share
of admitted Asian American applicants fluctuates is greater than
the amount by which the share of Asian American applicants
fluctuates. This is also true for Hispanic and African American
applicants. It is the opposite of what one would expect if Harvard
imposed a quota. The fact that Harvard's admitted share of
applicants by race varies relatively little in absolute terms for
the classes of 2009 to 2018 is unsurprising and reflects the fact
that the racial makeup of Harvard's applicant pool also varies
very little over this period. The district court properly
concluded that Harvard does not utilize quotas and does not engage
in racial balancing.
Next, SFFA's argument on the impermissibility of one-
pagers is foreclosed by Grutter and Fisher II. The Grutter
majority held that the "consultation of the 'daily reports,' which
keep track of the racial and ethnic composition of the class" does
not "'sugges[t] there was no further attempt at individual review
save for race itself' during the final stages of the admissions
process." Id. (quoting id. at 392 (Kennedy, J., dissenting)).
- 67 -
The Grutter Court also pointed to two additional factors
-- present here -- supporting its conclusion that the university's
admissions program did not function as an unconstitutional quota:
(1) the variation in the percent of minority applicants admitted
each year and (2) the uncontradicted testimony by admissions
officers "that they never gave race any more or less weight based
on the information contained in these reports." Id.
Harvard's witnesses testified that they used one-pagers
for three main reasons: (1) to assess how well its diversity
recruitment efforts (e.g., via UMRP and HFAI) were working; (2) to
manage its yield rates; and (3) to avoid drop offs in students
with particular characteristics due to inadvertence or lack of
care.
Harvard may permissibly use one-pagers to assess the
effectiveness of its pre-application recruitment efforts. Given
the Fisher II majority's command that universities must "continue
to use . . . data to scrutinize the fairness of its admissions
program [and] to assess whether changing demographics have
undermined the need for a race-conscious policy," Harvard's use of
one-pagers for this purpose evidences narrow tailoring because it
allows Harvard to assess whether its race-conscious admissions
policy is still necessary. 136 S. Ct. at 2214.
Managing yield rates is also permissible. Harvard is a
residential college with a limited number of beds. It needs to
- 68 -
carefully monitor the number of applicants it admits to avoid
becoming overcrowded. Applicants with different demographics
accept offers of admission at different rates. For example,
applicants from "Sparse Country" accept offers of admission at
lower rates than other applicants. Engineering admittees yield at
lower rates. And applicants of different races also enroll at
differing rates. To help manage its class size, Harvard includes
geographic data, intended concentration, and race -- in addition
to many other factors, like gender, ALDC status, and economic
status -- on its one-pagers. This is permissible.
One-pagers also avoid drop-offs in admitted students
with certain characteristics, including race, due to inadvertence
or lack of care. An admissions officer testified that if Harvard
did observe a dramatic decrease in representation of members of a
particular race in its admitted class that was not due to
inadvertence or lack of care, "[s]ome things can't be avoided."
Harvard's use of race in this way has been approved by the Supreme
Court. See Grutter, 539 U.S. at 335 (finding Harvard's "flexible
use of race" instructive and refusing to characterize its use of
race as a quota when "Harvard certainly had minimum goals for
minority enrollment, even if it had no specific number firmly in
mind").
- 69 -
There Was No Error in Holding That Harvard Did Not Use Race
as a Mechanical Plus Factor
Next, SFFA argues that Harvard's admissions program is
not narrowly tailored because, in its view, Harvard's
consideration of race is mechanical. The Supreme Court has found
race-conscious admissions policies unconstitutional as mechanical
when they give pre-defined boosts to applicants solely because of
race, when they preclude individualized consideration of
applicants, and when race becomes the decisive factor in admission.
See Gratz, 539 U.S. at 271-72; see also Fisher II, 136 S. Ct. at
2207 (holding that consideration of race "does not operate as a
mechanical plus factor for underrepresented minorities" when it is
contextual).
In Gratz, the University of Michigan's undergraduate
program graded applicants using a point system. Id. at 255.
Applicants scoring over 100 points were guaranteed admission. Id.
The school automatically awarded "every applicant from an
underrepresented racial or ethnic minority group . . . 20 points."
Id. at 256. The Court found this system impermissible because it
did not allow for individualized consideration of applicants, as
required by Bakke, and because automatically distributing a fifth
of the points required for admission "mak[es] 'the factor of race
. . . decisive' for virtually every minimally qualified
underrepresented minority applicant." Id. at 272 (quoting Bakke,
- 70 -
438 U.S. at 317). In contrast, the Court has found race to be
sufficiently non-mechanical when its consideration is
individualized and can benefit any applicant. See Fisher II, 136
S. Ct. at 2207 (explaining that "the consideration of race, within
the full context of the entire application, may be beneficial to
any UT Austin applicant -- including whites and Asian–Americans"
(quoting Fisher v. Univ. of Texas at Austin, 645 F. Supp. 2d 587,
606 (W.D. Tex. 2009))).
SFFA argues that Harvard's race-conscious admissions
process does not pursue student-body diversity, places too much
weight on race, and that Harvard's use of race has no defined end
point.
SFFA's contention that Harvard elevates racial diversity
above other types of diversity is not supported by the evidence.
Harvard has demonstrated that it values all types of diversity,
not just racial diversity. Harvard's use of race in admissions is
contextual and it does not consider race exclusively.
Next, Harvard's process does not weigh race so heavily
that it becomes mechanical and decisive in practice. Harvard's
undergraduate admissions program considers race as part of a
holistic review process. This use was previously praised by the
Supreme Court as a way of considering race in a non-mechanical
way. Unlike the program in Gratz, Harvard does not award a fixed
amount of points to applicants because of their race.
- 71 -
The district court made a number of pertinent factual
findings against SFFA's arguments, all of which are supported by
the evidence. SFFA counters by pointing to evidence that Asian
American applicants with high grades and test scores are admitted
at lower rates than applicants of other races. The district court
considered this evidence and found that it "likely over emphasizes
grades and test scores and undervalues other less quantifiable
qualities and characteristics that are valued by Harvard and
important to the admissions process." SFFA II, 397 F. Supp. 3d at
165. It credited other evidence showing that the effect of race
on a student's chance of admission is "not disproportionate to the
magnitude of other tips applicants may receive." Id. at 199.
The Supreme Court has also approved admissions programs
where race has a larger effect on a student's chances of admission
than Harvard's use of race. In Grutter, the Supreme Court upheld
the University of Michigan Law School's race-conscious admission
program that, if eliminated, would have reduced the
underrepresented minority population of the admitted class from
14.5% to 4%, a 72.4% decrease. See 539 U.S. at 320. Here, without
considering race, the share of African American and Hispanic or
Other students enrolled at Harvard would decrease by 45%. See
SFFA II, 397 F. Supp. at 198. The impact of Harvard's use of race
- 72 -
on the makeup of its class is less than the one at issue in
Grutter.29
In Harvard's holistic admissions process, tips are used
for athletic ability, legacy status, geographic and economic
factors, race at times, and perhaps other reasons. But the
outcomes of Harvard's admissions process do not indicate that race
is impermissibly "'decisive' for virtually every minimally
qualified underrepresented minority applicant" within it. Gratz,
539 U.S. at 272 (quoting Bakke, 438 U.S. at 317). According to
SFFA's own expert's analysis, Harvard rejects more than two-thirds
of Hispanic applicants and slightly less than half of all African
American applicants who are among the top 10% most academically
promising applicants to Harvard in terms of standardized test
scores and GPA. Gratz precludes programs where race is decisive
for minimally qualified candidates. Harvard's admissions process
29 The United States attempts to make the impact of
Harvard's use of race appear more significant than it is. It
argues that Harvard "inflicts an 11.1% penalty" on Asian Americans
because, absent the consideration of race, their representation
would increase from 24% to 27%. It then claims that Harvard
provides a 133% bonus to African Americans because their
representation increases from 6% to 14%. While these calculations
are correct, similar calculations show that race was used about as
extensively in the program approved in Grutter. That program,
using the government's language and calculations, inflicted a
penalty of 10.9% on applicants who were not underrepresented
minorities (because their representation would increase from 85.5%
to 96% absent the consideration of race) while simultaneously
giving a 263% bonus to underrepresented minority applicants
(because their representation increased from 4% to 14.5% with the
consideration of race). See Grutter, 539 U.S. at 320.
- 73 -
is so competitive that race is not decisive for highly qualified
candidates. The district court also found that some Asian
American applicants are advantaged by Harvard's use of race. See
SFFA II, 397 F. Supp. 3d at 178 ("The policy of considering
applicants’ race may improve the admission chances of some Asian
Americans who connect their racial identities with particularly
compelling narratives.").
Relatedly, the United States argues that Harvard
"considers race at virtually every step of its admission process."
It reads Fisher II as mandating that race only be considered at
one step in a university's admissions process because race was
considered at only one point in the University of Texas at Austin's
process. This argument is not persuasive for several reasons.
Its premise is questionable. The Fisher II majority does say that
"race enters the admissions process . . . at one stage and one
stage only." 136 S. Ct. at 2207. But the pervasiveness of the
University of Texas's consideration of race was one of Justice
Alito's chief criticisms in his dissent, which was joined by Chief
Justice Roberts and Justice Thomas. Justice Alito cited the
district court's finding that "[b]ecause an applicant's race is
identified at the front of the admissions file, reviewers are aware
of it throughout the evaluation," Fisher v. Univ. of Texas at
Austin, 645 F. Supp. at 597, before writing that "[c]onsideration
of race therefore pervades every aspect of UT's admissions
- 74 -
process," Fisher II, 136 S. Ct. at 2220 (Alito, J., dissenting).
It is difficult to imagine how a school could both consider an
applicant's race and holistically review their application, as
required by Supreme Court precedent, at only a single point in the
admissions process. This is true because the applications
themselves frequently contain racially identifiable information,
as we have described earlier.
Regardless, there is nothing in Fisher II suggesting
that a university can only consider race once or that only a single
use of race is a necessary component of a narrowly tailored policy.
The Court made clear that as long as race is "considered in
conjunction with other aspects of an applicant's background" and
is "but a 'factor of a factor of a factor' in the holistic-review
calculus," it will not be considered impermissibly mechanical.
Fisher II, 136 S. Ct. at 2207. Harvard has shown that its holistic
consideration of race is not impermissibly extensive.
Finally, SFFA and the United States argue that Harvard's
use of race has no end point because Harvard has not identified a
stopping point for its use of race. It derives this argument from
Grutter's statement that the "use of race must have a logical end
point," 539 U.S. at 342, and its hope that "25 years from now, the
use of racial preferences will no longer be necessary to further
the interest approved today," id. at 343. This argument is also
not persuasive and is insensitive to the achievement of the
- 75 -
university's legitimate goals once it has met the requirements
established by the Supreme Court. Indeed, the Supreme Court never
mentioned Grutter's 25-year timeline in Fisher I or Fisher II.
Harvard's failure to identify a specific level of
diversity it would need to achieve before it stopped using any
consideration of race is not fatal to its admissions program. The
Fisher II majority held that because "the University is prohibited
from seeking a particular number or quota of minority students, it
cannot be faulted for failing to specify the particular level of
minority enrollment at which it believes the educational benefits
of diversity will be obtained." 136 S. Ct. at 2210. Importantly,
the evidence is that Harvard has periodically reviewed its use of
race in the past, has periodically and recently considered race-
neutral alternatives, and has made it clear that it will continue
to do so in the future. See Grutter, 539 U.S. at 342 (requiring
that universities conduct “periodic reviews to determine whether
racial preferences are still necessary”). The same committee that
reaffirmed Harvard's need to pursue racial diversity in admissions
emphasized that "it will be important to reassess, periodically,
the necessity of considering race and ethnicity in the admissions
process." No Supreme Court precedent requires Harvard to identify
a specific end point for its use of race.
- 76 -
There Was No Error in Holding That Harvard Considered Race-
Neutral Alternatives and Legitimately Concluded That the
Alternatives Were Not Workable
SFFA argues that Harvard's use of race fails the narrow
tailoring prong of strict scrutiny because Harvard has disregarded
race-neutral ways to achieve its diversity goals. See Fisher I,
570 U.S. at 312 ("Narrow tailoring . . . . involves a careful
judicial inquiry into whether a university could achieve
sufficient diversity without using racial classifications.").
Courts must not defer to a university's consideration of workable
race-neutral alternatives. Id. They must "be satisfied that no
workable race-neutral alternatives would produce the educational
benefits of diversity." Id. Universities have an "ongoing
obligation to engage in constant deliberation and continued
reflection" on whether workable race-neutral alternatives exist.
Fisher II, 136 S. Ct. at 2215.
The Supreme Court has distinguished "conceivable"
alternatives from "workable" alternatives, requiring only the use
of workable alternatives. Id. at 2208. Harvard has met its
burden, including through the Smith Report, to show that it has
carefully considered all alternatives. It has concluded that they
are not workable and would undercut its educational objectives.
"Narrow tailoring does not . . . require a university to choose
between maintaining a reputation for excellence or fulfilling a
- 77 -
commitment to provide educational opportunities to members of all
racial groups." Grutter, 539 U.S. at 339 (citations omitted).
Harvard has implemented many of the policies SFFA
proposes, like eliminating Early Action and increasing financial
aid and outreach, but found those policies insufficient.30 This
accords with Fisher II. 136 S. Ct. at 2212-13 (finding race-
neutral alternatives unworkable when the university had introduced
scholarship programs and "submitted extensive evidence of the many
ways in which it already had intensified its outreach efforts").
Harvard has increased the financial aid it offers to
low-income applicants since it instituted its HFAI program in 2004.
It made multiple changes over the years to increase the
affordability of a Harvard education. When this case went to
trial, Harvard's budget for undergraduate financial aid was around
$200 million dollars and more than half of Harvard undergraduates
received financial aid. About 90% of Harvard students pay the
same or less in tuition as they would at a state school. The
initial introduction of HFAI and Harvard's first increase of the
zero parental contribution limit to $60,000 did have positive
effects on Harvard's ability to attract and admit racially diverse
30 Some policy changes backfired. Harvard abandoned its
Early Action program for the classes of 2012 through 2015 but
reinstated it after finding that "many highly talented students,
including some of the best-prepared low-income and
underrepresented minority students, were choosing programs [at
other schools] with an early-action option."
- 78 -
applicants. But later changes to the program had little to no
effect on the racial makeup of Harvard's applicants or admitted
class. As the district court found, Harvard "has already reached,
or at least very nearly reached, the maximum returns in increased
socioeconomic and racial diversity that can reasonably be achieved
through outreach and reducing the cost of a Harvard education."
SFFA II, 397 F. Supp. 3d at 180. Further, the Smith Report, while
not entitled to any deference, does show that Harvard has carefully
considered and rejected race-neutral alternatives. The Smith
Report was supported by the testimony of Harvard officials. It
was also supported by the testimony of Harvard's economic expert.
SFFA focuses its argument on "Simulation D,"31 one of its
proposed race-neutral alternatives. Under this scenario, Harvard
would eliminate its consideration of race, eliminate LDC tips, and
increase the tip for low-income applicants.
If Harvard were to adopt Simulation D, its analysis shows
that the admitted share of white and African American applicants
would decrease (from 40% to 33% and 14% to 10%, respectively) and
the share of Asian American and "Hispanic and Other" applicants
would increase (from 24% to 31% and 14% to 19%, respectively).
The average student's high school grade point average would remain
31 Simulation D is also called Simulation 7 elsewhere in
the record.
- 79 -
unchanged, but the average SAT score would decrease from 2244 to
2180.
Harvard proved that Simulation D was not a workable
alternative. Not only would SAT scores drop, but the fraction of
applicants with academic, extracurricular, personal, and athletic
ratings of 1 or 2 would decrease by more than 10% (ranging from an
11% decrease for the personal rating to a 22% decrease for the
athletic rating). If Harvard were to increase its tip based on
socioeconomic status, it would make sacrifices on almost every
dimension important to its admissions process, including one
designed to measure a student's academic excellence. As the Fisher
II majority held when it found the plaintiff's alternative of
"altering the weight given to academic and socioeconomic factors
in the University's admissions calculus" unworkable, "the Equal
Protection Clause does not force universities to choose between a
diverse student body and a reputation for academic excellence."
Fisher II, 136 S. Ct. at 2213.
Next, the district court made a factual finding that
removing LDC tips "would adversely affect Harvard’s ability to
attract top quality faculty and staff and to achieve desired
benefits from relationships with its alumni and other individuals
who have made significant contributions to Harvard." SFFA II, 397
F. Supp. 3d at 180. The loss of top faculty would negatively
affect the educational experience of students at Harvard and its
- 80 -
reputation for excellence. See Grutter, 539 U.S. at 339. The
loss of the ability to cultivate relationships with donors and
alumni through LDC tips would harm Harvard's ability to raise
funds, and the staffing changes needed to accommodate a drastic
shift in student concentrations would present Harvard with
sizeable administrative expenses. SFFA II, 397 F. Supp. 3d at
179-80, 182. Under Simulation D, the number of humanities
applicants admitted would drop by 14%. See id. at 182. The number
of engineering applicants admitted would rise by 8%. See id.
Finally, African American representation in Harvard's
admitted class would decrease by about 32% under Simulation D.
Harvard's consideration of race is not impermissibly extensive,
but considering race is meaningful to Harvard's admissions process
because it prevents diversity from plummeting. Harvard's race-
conscious admissions program ensures that Harvard can retain the
benefits of diversity it has already achieved.32 The district
32Dean Smith testified about how his committee evaluated
race-neutral alternatives and what it considered to be an
acceptable level of racial and ethnic diversity. He said that,
having seen the progress Harvard has already made in achieving
racial diversity and the benefits it has had, one consideration
was whether Harvard would be "moving backwards from where [it is]
today." Regarding any decline in African American representation
specifically, he testified that alienation and isolation is
already a problem among African American students at Harvard and
that Harvard is "not looking to make that worse." He also
testified that Harvard does not view underrepresented minorities
interchangeably. The increase in representation among Hispanic
and Other students under Simulation D would not cure the decrease
in African American representation.
- 81 -
court found that the dramatic decline in diversity under Simulation
D could adversely affect the educational experience at Harvard and
increase feelings of isolation and alienation among Harvard's
students. Id. at 183. Similarly, the Fisher II majority found
that the University of Texas's compelling interest could not be
met with race-neutral alternatives when it had introduced
"evidence that minority students admitted [under a race-neutral
regime] experienced feelings of loneliness and isolation." 136 S.
Ct. at 2212.
Ample testimony in the record, including from Harvard
students and alumni, supported this finding.33 As these witnesses'
testimony makes clear, a meaningful reduction in
representation -- and a 32% reduction in African American
representation is clearly meaningful -- would make Harvard less
33 As an example, one student testified that, when choosing
which school to attend, she "wanted to make sure that there would
be other students who were people of color like myself . . . so
that I could have a more safe environment, a more welcoming
environment, and a better . . . learning environment." After
arriving at Harvard, she found that "students of color were a huge
minority in almost every space" and that walking into a class with
few or no people of color made her nervous and reticent; she would
hold back in discussions to avoid being "seen or stereotyped as
someone who . . . is just talking about communities of color
because that's where I came from." And she described how a large
reduction in the number of Black or Latinx students would be
"catastrophic" because "there are so few students of color and
under-represented minority groups at Harvard as it is" and that
"any sort of reduction in any of those groups would be really
detrimental to the community at Harvard, both for students of color
[and] students in general." At trial, other students and recent
alumni expressed similar viewpoints.
- 82 -
attractive and hospitable to minority applicants while limiting
all students' opportunities to engage with and learn from students
with different backgrounds from their own. Enabling students to
understand, relate to, and learn from people of different
backgrounds is one of the main goals of Harvard's race-conscious
admissions program. It is a compelling interest, but under
Simulation D Harvard cannot achieve it. Harvard has carried its
burden of showing that no workable race-neutral alternatives
exist.
3) The District Court Did Not Err in Finding Harvard Does Not
Intentionally Discriminate Against Asian American Applicants
SFFA's final claim is that Harvard's admissions policy
intentionally discriminates against Asian Americans. It argues
that the district court "could not rule out that Asian Americans
are penalized in Harvard’s admissions process" and says both the
non-statistical and statistical evidence showed that Harvard
discriminates against Asian Americans. From this premise, it
argues that Harvard cannot carry its burden of disproving
intentional discrimination under strict scrutiny.34
34 SFFA's intentional discrimination claim does not fit
neatly into the strict scrutiny framework. Harvard disputes
whether strict scrutiny applies to this claim. It admits to using
race in its admissions process. But it has never admitted to
discriminating against Asian American applicants and denies doing
so.
SFFA argues that strict scrutiny applies to any challenge
related to a university's race-conscious admission policy and
- 83 -
To make out a prima facie case of intentional
discrimination on the basis of race under Title VI, a plaintiff
must show that the defendant treated members of one race
differently and less favorably than members of another race and
that the defendant did so with a racially discriminatory
purpose. See Washington v. Davis, 426 U.S. 229, 239-40 (1976)
(describing the intentional discrimination standard applicable
under the Equal Protection Clause of the Fourteenth Amendment);
Sandoval, 532 U.S. at 280 (finding the protections of Title VI
coextensive with those of the Equal Protection Clause); accord
Goodman v. Bowdoin Coll., 380 F.3d 33, 43 (1st Cir. 2004) (citing
Tolbert v. Queens Coll., 242 F.3d 58, 69 (2d Cir. 2001)).
Assuming arguendo that SFFA is correct that strict
scrutiny operates on its intentional discrimination claim by
shifting the burden to Harvard to disprove intentional
discrimination, Harvard can succeed only if it disproves any of
that, under strict scrutiny, Harvard bears the burden of disproving
SFFA's intentional discrimination claim. Harvard argues that
SFFA's intentional discrimination claim does not get the benefit
of strict scrutiny until SFFA has established that Harvard has
discriminated against Asian Americans and acted with racial animus
against them. See Vill. of Arlington Heights v. Metro. Hous. Dev.
Corp., 429 U.S. 252, 265 (1977) ("Proof of racially discriminatory
intent or purpose is required to show a violation of the Equal
Protection Clause."). The district court adopted SFFA's view.
See SFFA II, 397 F. Supp. 3d at 190. We need not decide this issue
because we hold that Harvard prevails even applying the more
demanding standard advanced by SFFA.
- 84 -
these elements. Harvard must show that it did not discriminate
on the basis of race or that its discrimination was not intentional
(i.e., it did not act with animus or "stereotyped thinking or other
forms of less conscious bias"). Thomas v. Eastman Kodak Co., 183
F.3d 38, 42 (1st Cir. 1999). We hold, as did the district court,
that Harvard has carried this burden.
The District Court Did Not Err in Concluding That the Non-
Statistical Evidence Did Not Show Discrimination Against
Asian American Applicants
SFFA argues that three pieces of non-statistical
evidence show intentional discrimination. First, it says that
Harvard's admissions process as a whole -- and its use of a
personal rating in particular -- is highly subjective, which makes
it susceptible to stereotyping and bias. Next, it argues that
Harvard ignored warnings that its process might be racially biased,
citing the 1990 OCR Report that, as the district court
acknowledged, "found recurring characterizations of Asian American
applicants that were broadly consistent with stereotypes." SFFA
II, 397 F. Supp. 3d at 154. SFFA says that because Harvard took
no steps to remedy these characterizations, its process is biased
against Asian Americans. Finally, SFFA says that Harvard's "post-
filing conduct" evidences past discrimination because, in response
to this lawsuit, Harvard amended its written handbook on reading
procedures to explicitly instruct admissions officers that they
- 85 -
should not consider race when assigning a personal rating and
increased the number of Asian Americans it admitted.
Supreme Court precedent makes clear that the fact that
Harvard's application process is subjective is insufficient to
overcome other evidence in the record that Harvard is not biased
against Asian Americans and does not stereotype them. First, there
is no requirement that universities use entirely objective
criteria when considering race to admit applicants.35 Cf. Bakke,
438 U.S. at 317-18 (describing, and approving of, Harvard's
subjective, "flexible" admissions system where "the weight
attributed to a particular quality may vary from year to year
depending upon the 'mix' both of the student body and the
applicants for the incoming class"). Harvard presented testimony
from multiple admissions officers that its admissions process,
though subjective, did not facilitate bias or stereotyping. The
district court found that the "testimony of the admissions officers
that there was no discrimination against Asian American applicants
with respect to the admissions process as a whole and the personal
35 Indeed, it is unclear whether Harvard could even adopt
a more objective system and still comply with the Supreme Court's
precedent. A more objective system could be viewed as mechanically
taking race into account. Cf. Gratz, 539 U.S. at 270 ("[T]he
University's policy, which automatically distributes 20 points, or
one-fifth of the points needed to guarantee admission, to every
single 'underrepresented minority' applicant solely because of
race, is not narrowly tailored to achieve the interest in
educational diversity.").
- 86 -
ratings in particular was consistent, unambiguous, and
convincing." SFFA II, 397 F. Supp. 3d at 203. No witness who
testified "had seen or heard anything disparaging about an Asian
American applicant." Id. The district court was permitted to
assess the credibility of witnesses and credit the testimony of
Harvard's admissions officers. See Cooper v. Harris, 137 S. Ct.
1455, 1478 (2017) (upholding a district court's credibility
determination that a witness "skirted the truth . . . when he
claimed to have followed only race-blind criteria in drawing
district lines" because a reviewing court "cannot disrespect such
credibility judgments"); Cumpiano v. Banco Santander Puerto Rico,
902 F.2d 148, 158-59 (1st Cir. 1990) (finding, in a discrimination
case, that "whether or not to credit [defendant's employee's]
testimony as to his knowledge and intent was predominantly a
credibility question" and that "[a]ppellate courts should always
be reluctant to erase the trial judge's answer to such a query").36
The nature of Harvard's admissions process, as the
district court recognized, offset any risk of bias. SFFA II, 397
F. Supp. 3d at 203. An applicant must secure a majority of votes
at a full-body admissions committee meeting with forty admissions
officers to be admitted to Harvard, which mitigates the risk that
36SFFA did not present any evidence to the contrary or
"present a single Asian American applicant who was overtly
discriminated against or who was better qualified than an admitted
white applicant." SFFA II, 397 F. Supp. 3d at 203.
- 87 -
any individual officer's bias or stereotyping would affect
Harvard's admissions process.
Nor was there error in the district court's fact finding
that earlier reports did not show that Harvard discriminated
against or stereotyped Asian American applicants. The district
court acknowledged that the 1990 OCR Report "found recurring
characterizations of Asian American applicants that were broadly
consistent with stereotypes." Id. at 154. This same report also
concluded that "Harvard did not discriminate against Asian
American applicants." Id. The court also considered SFFA's
contention that Harvard referred to Asian American applicants as
"quiet," "flat," "shy," and "understated," and that this showed
stereotyping. Id. at 157. The district court found that using
these words "with regard to such an applicant would be truthful
and accurate rather than reflective of impermissible
stereotyping." Id. It found that Harvard considers applicants
holistically and that the evidence did not show "that any applicant
was referred to by these types of descriptors because of their
race or that there was any sort of systemic reliance on racial
stereotypes." Id. at 157, 193. To the contrary, the evidence
showed that admissions officers referred to applicants of all races
using similar language, not just Asian Americans. Id. at 157.
SFFA's last two arguments on this point are that it was
only after it filed its lawsuit that Harvard both updated its
- 88 -
written reading procedures to say that race should not be
considered when assigning the personal rating and increased the
number of Asian American applicants it admitted. As to the first
argument, Harvard updates its reading procedures annually. The
district court found that "Harvard has made clear to its admissions
officers in more recent years that they should not use race in
assigning the profile ratings." Id. at 156.
As to SFFA's second argument, while the share of Asian
American applicants admitted to Harvard did increase from 19.1%
for the class of 2018 to 20.6% for the class of 2019 after this
lawsuit was filed, the argument made ignores the broader context.
Asian Americans constituted 20.3% of the admitted class of 2016,
showing that the 20.6% share of Asian American admittees for the
class of 2019 was not anomalous. Neither was the post-filing
increase from 19.1% for the class of 2018 to 20.6% for the class
of 2019. The share of Asian Americans similarly increased from
17.6% for the class of 2010 to 19.5% for the class of 2011 and
from 19.3% for the class of 2015 to 20.3% for the class of 2016
before this lawsuit was filed. Indeed, the number of Asian
Americans admitted to Harvard has been steadily increasing for
decades.
Because statistical evidence can "generate an inference
of intentional discrimination," Haidak v. Univ. of Massachusetts-
- 89 -
Amherst, 933 F.3d 56, 75 (1st Cir. 2019), we now turn to the
parties' statistical evidence.
The District Court Did Not Err in Concluding That the
Statistical Evidence Did Not Show Discrimination Against
Asian American Applicants
SFFA's main argument with respect to the statistical
evidence revolves around whether to include an applicant's
personal rating in the model of Harvard's admissions process. If
the personal rating is included, as done by Harvard's expert, being
Asian American has a statistically insignificant effect on an
applicant's chance of admission. If the personal rating is
excluded, as done by SFFA's expert, it shows that being Asian
American has a statistically significant37 negative effect on an
applicant's chance of admission to Harvard. SFFA argues that it
was clear error for the district court to consider a model
including the personal rating. It also says the district court
37 An explanatory variable has a "statistically
significant" effect if, assuming the effect of the explanatory
variable on the response variable were actually zero, the effect
observed in the model is different enough from zero that it is
unlikely to be due to chance. "Different enough" is quantified
using a significance level. The experts in this case used a
significance level of 5%. For example, the experts called the
effect of Asian American identity on the chance of admission to
Harvard statistically significant if, assuming being Asian
American actually had no effect on admission chances, the
probability of observing an effect of Asian American identity at
least as extreme as the effect observed in the model was less than
5%. We use the same terminology.
- 90 -
erred when it concluded that Harvard had disproved intentional
discrimination after it credited one model including the personal
rating and another excluding it. We find no error.
The District Court Did Not Err in Finding the Statistical
Evidence Did Not Show the Personal Rating Was Influenced by
Race
As a matter of basic principles of statistical analysis,
whether the personal rating used in the admissions process should
be included in the regression model hinges on whether the personal
rating is influenced by race and whether that rating includes
information that is not otherwise controlled for in the experts'
admissions models. See FJC Reference Manual at 313-16, 322-24
(explaining that experts typically assume that "changes in
explanatory variables affect the dependent variable, but changes
in the dependent variable do not affect the explanatory variables"
and the concept of omitted variable bias)
SFFA argues that race influences the personal rating and
should be excluded, while Harvard argues to the contrary that race
is only correlated with the personal rating and excluding the
personal rating would introduce omitted variable bias into the
model.
The district court acknowledged that the model without
the personal rating was "econometrically reasonable" and
"provide[d] evidence that is probative of the effect of race on
- 91 -
the admissions process" but made a factual finding that "including
the personal rating results in a more comprehensive analysis" and
so was more accurate. SFFA II, 397 F. Supp. 3d at 173. Considering
all of the evidence, it concluded that "the majority of the
disparity in the personal rating between white and Asian American
applicants was more likely caused by race-affected inputs to the
admissions process . . . or underlying differences in the
attributes that may have resulted in stronger personal ratings"
than Harvard admissions officers' biases. See id. at 171.
There is a clear and important distinction between race
being correlated with the personal rating and race influencing the
personal rating. Race correlating with the personal rating means
that there is a statistical relationship between race and the
personal rating. Race influencing the personal rating means that
this statistical relationship is causal. It means that Harvard
assigns applicants higher or lower personal scores because of their
race. The distinction between correlation and influence is very
important.38 See Samaan v. St. Joseph Hosp., 670 F.3d 21, 33 (1st
38 SFFA implies throughout its brief that to see a negative
correlation between Harvard's personal rating and Asian American
ethnicity and not attribute it to racial bias is to conclude that
"maybe the stereotypes about Asian Americans are true." This is
a false dichotomy. SFFA ignores the possibility that factors
external to Harvard's admissions process but correlated with race
could account for the racial disparity in personal ratings.
Racial differences in Harvard's other ratings help illustrate
SFFA's logical error. For example, Asian Americans do better than
any other racial group on Harvard's academic rating. No one looks
- 92 -
Cir. 2012) ("Correlation is not causation."); Wessmann v. Gittens,
160 F.3d 790, 804 (1st Cir. 1998) ("Even strong
statistical correlation between variables does not automatically
establish causation."). If race is only correlated with the
personal rating, excluding it from regression models could make it
appear as if Harvard discriminates when it does not. If race
influences the personal rating, including it in the experts'
regression models could make it appear as if Harvard does not
discriminate when it does.
The district court found that even when controlling for
a number of other factors, race is correlated with the personal
rating. SFFA II, 397 F. Supp. 3d at 169-70. Based on this fact,
SFFA argues that the court was required to find that race
influences the personal rating. The district court disagreed, and
at this correlation and concludes that there are only two
possibilities: either Asian American students are smarter than
everyone else or Harvard thinks they are and assigns them higher
ratings accordingly. Instead, it is more likely that factors
external to Harvard's admissions process but correlated with race
account for Asian American's higher academic ratings. For this
reason, everyone agreed that Harvard's academic rating should
remain in the logistic regression model even though Professor
Arcidiacono showed that -- even when controlling for obvious
variables like grades and test scores and other factors like
demographics, parental education, geographic region, and
characteristics of applicants' high schools -- there is a positive
correlation between Asian American identity and Harvard's academic
rating. Indeed, when asked what could account for the statistical
relationship between the academic rating and race even when
controlling for other factors, Professor Arcidiacono replied: "We
don't think it's because of race. We think it's because of these
unobservable factors."
- 93 -
we hold that the district court did not clearly err in concluding
that the more accurate statistical model includes the personal
rating.
The district court gave three reasons why the personal
rating should be included in the model. First, it credited
Harvard's witnesses' testimony that they did not consider race in
assigning the personal rating as evidence that race did not
influence the personal rating. Id.
Second, the district court found that Professor
Arcidiacono's analysis establishing a statistical correlation
between race and the personal rating explains only a portion of
the variation in personal ratings. Id. Professor Card testified
at trial that Professor Arcidiacono's model had a low pseudo-R2
value, indicating a poor fit.39 A low pseudo-R2 value alone is not
sufficient to reject Professor Arcidiacono's model of the personal
rating. But it is evidence that important explanatory variables
may have been omitted from it.
Third, this evidence of poor fit properly led the
district court to inquire into what these omitted variables might
39 In linear regression analyses, the R2 value "is a
statistic that measures the percentage of variation in the
dependent variable that is accounted for by all the explanatory
variables." FJC Reference Manual at 345. For technical reasons
not relevant for our purposes, statisticians cannot calculate R2
statistics for logistic regression models and instead typically
report an analogous quantity -- the pseudo-R2 value -- as one
estimate of a model's goodness of fit.
- 94 -
be. See id.; FJC Reference Manual at 314 n. 31 ("A very low R-
squared (R2) is one indication of an unexplained portion of the
multiple regression model that is unacceptably high. However, the
inference that one makes from a particular value of R2 will depend,
of necessity, on the context of the particular issues under study
and the particular dataset that is being analyzed.") It concluded
that Harvard had not discriminated on the basis of race. See SFFA
II, 397 F. Supp. 3d at 171.
Professor Card testified that non-quantifiable aspects
of an applicant's personal essay could contribute to the
correlation between race and the personal rating identified by
Professor Arcidiacono. He also presented statistical evidence
showing that Asian American applicants receive lower teacher and
guidance counselor recommendation ratings than white applicants.
Both personal essays and teacher and guidance counselor
recommendations factor heavily into applicants' personal ratings,
justifying the court's inquiry into what could account for any
correlation between these factors and race. If factors external
to Harvard -- like personal essays and recommendations -- correlate
with race, affect the personal rating, but are not controlled for
in Professor Arcidiacono's model of the personal rating, this calls
into question SFFA's contention that race influences (and is not
just correlated with) the personal rating.
- 95 -
The district court then analyzed whether these external
factors -- personal essays and teacher and guidance counselor
recommendations -- correlate with race. Id. at 169. It found
that non-quantifiable aspects of applicants' personal statements
might be correlated with race but are not controlled for in
Professor Arcidiacono's model. Id.
For instance, in Harvard's experience, applicants do
choose to write about how "their racial identities have shaped
their pre-college experiences" and admissions officers might read
these essays40 as evidence of an applicant's "abilit[y] to overcome
obstacles" and therefore infer their "leadership ability or other
personal strengths." Id. at 169-70 & n.48.
Turning to the teacher and guidance counselor
recommendations, the district court opined that they sometimes
"seemingly presented Asian Americans as having less favorable
personal characteristics than similarly situated non-Asian
American applicants" and detailed why teacher and guidance
40 Record evidence supported this conclusion. One
admissions officer testified that the ability to overcome
obstacles -- including racial ones -- is a factor in assessing
submitted essays and assigning personal ratings. In addition, a
student who had been admitted to Harvard testified about her
application essay discussing how her "ethnoracial identity had
impacted every decision [she] had made, every experience that [she]
had had" and that she "wanted to write about it because [she] felt
like it was something important and something of value that [she]
could bring to a school like Harvard."
- 96 -
counselor recommendation letters might correlate with race for
reasons unrelated to Harvard. Id. at 170-71.
Whether or not the applicant came from a privileged
background was one likely factor. Privileged students likely have
better access to schools with low student-to-teacher ratios and
teachers and guidance counselors with more time to write strong,
individualized recommendations. See id. at 170. Privileged
students likely receive better recommendations than less
privileged students. See id. Privilege is correlated with race.
One of Harvard's amici expands on this point, citing research that
Asian American students are more likely than white students to
attend public high schools where overloaded teachers and guidance
counselors may provide more perfunctory recommendations.
Likewise, the district court reasoned that a "student that works
part time and a student that does not may receive different
recommendations even with the same academic performance and
without reference to race." Id. at 170-71. Because working part
time likely correlates with race, it is plausible that race is
also correlated with the quality of recommendations.41 The finding
41 The district court's reasoning does not itself imply
that teachers and guidance counselors are racially biased and
should not be so understood. Nor does it imply that Harvard must
accept such a characterization. Because race likely correlates
with privilege (resulting in applicants in some racial groups
having teachers that are stretched thinner than applicants in
others) or participation in school-sponsored extracurricular
activities (resulting in applicants in some racial groups having
- 97 -
that teacher and guidance counselor recommendations correlate with
race but were not accounted for in Professor Arcidiacono's model
was not erroneous.42
SFFA also argues that the district court engaged in
speculation with "zero evidentiary basis" by considering how these
omitted variables could be correlated with race and affect the
validity of Professor Arcidiacono's model. The argument is based
on a misperception of the requirements for statistical analysis.
The district court credited Professor Card's testimony and
analysis that the omitted variables discussed above -- the
qualitative components of teacher and guidance counselor
recommendations and personal essays -- called Professor
Arcidiacono's analysis of the personal rating into question. Id.
It was not erroneous for the district court to find that the
personal rating was not influenced by race in a way that precluded
its use in a logistic regression model.
less teacher interaction than applicants in others), teachers or
guidance counselors do not have to be biased to write more generic
or less enthusiastic letters for applicants based on race. An
alternative explanation is that race correlates with how well
teachers or guidance counselors know their students.
42 Contrary to SFFA's argument, the school support ratings
that Professor Arcidiacono included as control variables in his
model do not account for these biases because they reflect
assessments of teacher and guidance counselor recommendations as
a whole. As Professor Card testified, Professor Arcidiacono cannot
control for the specific portions of recommendation letters that
inform the personal rating specifically and might be correlated
with race.
- 98 -
The District Court Did Not Err by Crediting Harvard's Expert's
Logistic Regression Model Including the Personal Rating
Because the court did not err in concluding that the
personal rating was not influenced by race, it was not erroneous
to consider a logistic regression model of Harvard's admissions
process including it. Because all parties agree that Harvard
considers the personal rating in its admission decisions, it should
be included in any model of Harvard's admissions process unless
there is a strong reason to exclude it. See FJC Reference Manual
at 313-14. As Professor Card and Harvard's admissions officers
testified, the personal rating includes information that is not
accounted for elsewhere in the experts' models but is important to
Harvard's admissions process. And since race is correlated with
the personal rating but not influenced by it, excluding it would
increase the risk of misleading regression results. See FJC Manual
at 314 ("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."). Without the
personal rating, the model would suffer from omitted variable bias.
The court properly credited a logistic regression model including
the personal rating.
- 99 -
The District Court Did Not Err by Concluding That the
Statistical Evidence Did Not Show That Harvard Intentionally
Discriminated Against Asian Americans
We repeat that the statistical model using the personal
rating showed no discrimination against Asian Americans. Rather,
it shows that Asian American identity has a statistically
insignificant overall average marginal effect43 on admissions
probability of -.08%. This means that, on average, the model shows
that an Asian American student has a .08% lower chance of admission
to Harvard than a similarly situated white student and that this
effect is statistically insignificantly different from zero.
SFFA relies on one aspect of the statistical analysis
presented. SFFA's preferred model without the personal rating
shows a statistically significant overall average marginal effect
of -0.34%. This means that, on average, the model shows that an
Asian American student has a .34% lower chance of admission to
43 The "average marginal effect" of Asian American identity
is computed by first taking every applicant in the sample and
calculating two probabilities using the logistic regression model:
their probability of admission if they were Asian American and
their probability of admission if they were not Asian American.
The difference in the two probabilities is called marginal effect
of being Asian American for that applicant. The marginal effect
of being Asian American differs for every applicant depending on
other aspects of their application. The average marginal effect
is the average of the marginal effects across all applicants.
The overall average marginal effect uses data from all six
admissions cycles analyzed by the experts. The experts also
analyzed the average marginal effect in each admissions cycle.
- 100 -
Harvard than a similarly situated white student and that this
effect is statistically significantly different from zero. But
because the average marginal effect is calculated using all
applicants to Harvard, including many applicants whom Harvard is
unlikely to admit, this number does not establish that being Asian
American matters for the small subset of applicants who have a
realistic chance of being admitted to Harvard.
The statistically significant negative overall average
marginal effect of Asian American identity in SFFA' preferred model
is also not robust.44 In addition to disappearing entirely when
the personal rating is included in the model, it is almost
undetectable on a year-by-year basis even within SFFA's preferred
model. The average marginal effect of Asian American identity in
the model excluding the personal rating is only statistically
significantly negative in one of the six years analyzed.45 In five
of the six admissions cycles, the effect is statistically
44 We use the word "robust" in the technical sense. A
statistic or procedure is robust when it "does not change much
when data or assumptions are modified slightly." FJC Reference
Manual at 295; see also id. at 295, 322 (defining "robust" and
explaining that "[t]he issue of robustness . . . is of vital
importance").
45 For the classes of 2014, 2015, and 2016, Asian American
identity had a negative but statistically insignificant average
marginal effect on admissions chances. For the classes of 2017
and 2019, Asian American identity had a positive but statistically
insignificant average marginal effect. It is only in 2018 that
the average marginal effect is statistically significantly
negative.
- 101 -
indistinguishable from zero. Indeed, in two years, the model shows
that Asian American identity actually has a positive effect on an
applicant's chance of admission to Harvard.
Based on this evidence, the district court found that
regardless of whether the personal rating is included or not, the
average marginal effect on Asian American identity is close to
zero. Id. at 175. It found that the effect of Asian American
identity varies by admissions cycle and is not always negative in
each admission cycle. Id. Indeed, it found that Asian American
identity could be a positive factor if the model "was better able
to account for unobserved factors." Id. It reasoned that "[i]t
is also possible that the negative coefficient and average marginal
effect reflect a very slight implicit bias that could have played
a modest role in lowering Asian Americans’ admissions probability
in some of the 2014–2019 admissions cycles" but that, if there
were any implicit bias, "the effect was so slight that it went
unnoticed by careful and conscientious observers within the
Admissions Office." Id.
Finally, SFFA argues that "the district court recognized
that one likely explanation for why Asian Americans are penalized
in the admissions process is Harvard’s 'implicit bias'" and that
"calling the bias 'implicit' does not make it legal." First, the
district court called this effect "possible," not likely. Id.
Next, as the court recognized, this possibility was "unsupported
- 102 -
by any direct evidence" before it. Id. at 171. Indeed, there was
ample non-statistical evidence suggesting that Harvard admissions
officers did not engage in any racial stereotyping, and the court
determined that "no credible evidence . . . corroborates the
improper discrimination suggested" by SFFA's preferred model. Id.
at 203. The district court's speculation about what might have
caused a statistically significant effect in one of the two models
it considered does not transform its finding that there was no
"intent by admissions officers to discriminate based on racial
identity" into clear error. Id. at 175.
The district court did not clearly err in finding that
Harvard did not intentionally discriminate against Asian
Americans. See Torres-Lazarini v. United States, 523 F.3d 69, 72
(1st Cir. 2008) ("When the evidence presented at a bench trial
supports plausible but competing inferences, the court's decision
to favor one inference is not clearly erroneous."); see Cape Fear,
Inc. v. Martin, 312 F.3d 496, 500 (1st Cir. 2002).
IV.
Harvard has an "ongoing obligation to engage in constant
deliberation and continued reflection regarding its admissions
policies." Fisher II, 136 S. Ct. at 2215. The issue before us is
whether Harvard's limited use of race in its admissions process in
order to achieve diversity in the period in question is consistent
- 103 -
with the requirements of Supreme Court precedent. There was no
error.
Affirmed. No costs are awarded.
- 104 -