Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 1 of 130 UNITED STATES DISTRICT COURT DISTRICT OF MASSACHUSETTS STUDENTS FOR FAIR ADMISSIONS, INC., * * Plaintiff, * * v. * * PRESIDENT AND FELLOWS OF * HARVARD COLLEGE (HARVARD * CORPORATION), * * Defendant. * Civil Action No. 14-cv-14176-ADB FINDINGS OF FACT AND CONCLUSION OF LAW Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 2 of 130 Table of Contents I. INTRODUCTION ....................................................................................................................... 4  II. PROCEDURAL HISTORY ....................................................................................................... 4  III. FINDINGS OF FACT: DIVERSITY, ADMISSIONS PROCESS, AND LITIGATION ........ 6  A. Diversity at Harvard............................................................................................................ 6  1. Harvard’s Interest in Diversity ..................................................................................... 6  2. Admissions Office’s Efforts to Obtain a Diverse Applicant Pool ................................ 8  B. The Admissions Process ................................................................................................... 11  1. The Application .......................................................................................................... 12  2. Alumni and Staff Interviews....................................................................................... 13  3. Application Review Process ....................................................................................... 16  i.  Admissions Office and Personnel .............................................................. 16  ii.  Reading Procedures .................................................................................... 18  iii.  Committee Meetings and Admissions Decisions ....................................... 23  4. Harvard’s Use of Race in Admissions ........................................................................ 27  C. Prelude to this Lawsuit...................................................................................................... 31  1. The Unz Article .......................................................................................................... 31  2. Analysis by Office of Institutional Research.............................................................. 32  i.  Mark Hansen’s Admissions Models........................................................... 32  ii.  Low-Income Admissions Models .............................................................. 35  3. The Ryan Committee .................................................................................................. 38  4. The Khurana Committee ............................................................................................ 39  5. The Smith Committee ................................................................................................. 40  IV. FINDINGS OF FACT: NON-STATISTICAL EVIDENCE OF DISCRIMINATION.......... 41  A. Sparse Country .................................................................................................................. 41  B. The OCR Report ............................................................................................................... 43  C. More Recent Allegations of Stereotyping and Bias .......................................................... 45  V. FINDINGS OF FACT: STATISTICAL ANALYSIS.............................................................. 50  A. Sources of Statistical Evidence ......................................................................................... 50  B. Admission Rates and Ratings by Race ............................................................................. 53  C. Descriptive Statistics ......................................................................................................... 57  1. Professor Card’s Multidimensionality Analysis ......................................................... 57  2. Professor Arcidiacono’s Academic Index Decile Analysis ........................................ 60  2 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 3 of 130 D. Overview of Logistic Regression Models......................................................................... 62  E.  Regression Models of School Support, Profile, and Overall Ratings ............................... 67  1. Relationship Between Race and School Support Ratings .......................................... 67  2. Relationship Between Race and Personal Ratings ..................................................... 68  3. Regression Models of the Academic, Extracurricular, and Overall Ratings .............. 73  F.  Regression Models of Admissions Outcome .................................................................... 74  G. Absence of Statistical Support for Racial Balancing or Quotas ....................................... 80  VI. FINDINGS OF FACT: RACE-NEUTRAL ALTERNATIVES ............................................. 83  A. Eliminating Early Action .................................................................................................. 85  B. ALDC Tips........................................................................................................................ 86  C. Augmenting Recruiting Efforts and Financial Aid ........................................................... 87  D. Increasing Diversity by Admitting More Transfer Students............................................. 88  E.  Eliminating Standardized Testing ..................................................................................... 88  F.  Place-Based Quotas .......................................................................................................... 89  G. SFFA’s Proposed Combinations of Various Race-Neutral Alternatives .......................... 90  VII. CONCLUSIONS OF LAW ................................................................................................... 92  A. Overview ........................................................................................................................... 92  B. SFFA Has Standing........................................................................................................... 93  C. The Supreme Court and Race-Conscious Admissions ..................................................... 94  D. Harvard’s Admission Program and Strict Scrutiny ........................................................ 102  1. Compelling Interest .................................................................................................. 106  2. Narrowly Tailored .................................................................................................... 107  E.  Count II: Harvard Does Not Engage in Racial Balancing ............................................. 112  F.  Count III: Harvard Uses Race as a Non-Mechanical Plus Factor ................................... 116  G. Count V: No Adequate, Workable, and Sufficient Fully Race-Neutral Alternatives Are Available ...................................................................................................................... 119  H. Count I: Harvard Does Not Intentionally Discriminate .................................................. 122  VIII. CONCLUSION .................................................................................................................. 127  3 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 4 of 130 BURROUGHS, D.J. I. INTRODUCTION Plaintiff Students for Fair Admissions, Inc. (“SFFA”) alleges that Defendant President and Fellows of Harvard College (“Harvard”) discriminates against Asian American applicants in the undergraduate admissions process to Harvard College in violation of Title VI of the Civil Rights Act of 1964, 42 U.S.C. §§ 2000d et seq. (“Title VI”). 1 Harvard acknowledges that its undergraduate admissions process considers race as one factor among many, but claims that its use of race is consistent with applicable law. II. PROCEDURAL HISTORY On November 17, 2014, SFFA initiated this lawsuit by filing a complaint that alleged that Harvard violates Title VI by intentionally discriminating against Asian Americans (“Count I”), using racial balancing (“Count II”), failing to use race merely as a “plus” factor in admissions decisions (“Court III”), failing to use race merely to fill the last “few places” in the incoming freshman class (“Count IV”), using race where there are available and workable race-neutral alternatives (“Count V”), and using race as a factor in admissions (“Count VI”). [ECF No. 1 ¶¶ 428–505]. SFFA seeks declaratory judgment, injunctive relief, attorneys’ fees, and costs. Id. at 119. On February 18, 2015, Harvard filed its answer, in which it denied any liability. See [ECF No. 17]. On April 29, 2015, several prospective and then-current Harvard students filed a motion to intervene. [ECF No. 30]. Although the Court denied the motion to intervene, it 1 There is considerable variation in the terminology individuals use to describe their racial and ethnic identities. This opinion uses the terms Hispanic, African American, Asian American, and white to describe the four racial or ethnic identities that account for the majority of applicants to Harvard because those are the terms the parties have used in litigating this case. The term Asian American, as opposed to Asian, is used because SFFA alleges that Harvard discriminates against United States citizens who identify as Asian American. Where “Asian” alone is used, this generally reflects the language used by others in their own analyses which are referred to herein and may include Asian applicants who would not identify as Asian American. 4 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 5 of 130 allowed the students to participate in the action as amici curiae (friends of the court). Students for Fair Admissions, Inc. v. President & Fellows of Harvard Coll., 308 F.R.D. 39, 51–53 (D. Mass.), ECF No. 52, aff’d, 807 F.3d 472 (1st Cir. 2015). On September 23, 2016, Harvard moved (1) to dismiss the lawsuit for lack of standing and (2) for judgment on the pleadings as to Counts IV and VI. [ECF Nos. 185, 187]. On June 2, 2017, the Court found that SFFA had the associational standing required to pursue this litigation, because it was an organization whose membership included Asian Americans who had applied to Harvard, been denied admission, and were prepared to apply to transfer to Harvard. Students for Fair Admissions, Inc. v. President & Fellows of Harvard Coll. (Harvard Corp.), 261 F. Supp. 3d 99, 111 (D. Mass. 2017), ECF No. 324. On the same date, the Court granted Harvard’s motion for judgment on the pleadings and dismissed Counts IV and VI, namely the failure to use race only to fill the last few places in the incoming freshman class and the use of race as a factor in admissions. Students for Fair Admissions, Inc. v. President & Fellows of Harvard Coll. (Harvard Corp.), No. 14-CV-14176-ADB, 2017 WL 2407254, at *1 (D. Mass. June 2, 2017), ECF No. 325. 2 Following the conclusion of discovery, on June 15, 2018, the parties filed cross motions for summary judgment on the four remaining counts, [ECF Nos. 412, 417], which the Court denied on September 28, 2018. Students for Fair Admissions, Inc. v. President & Fellows of Harvard Coll., 346 F. Supp. 3d 174, 180 (D. Mass. 2018), ECF No. 566. The case proceeded to trial on Counts I (intentional discrimination), II (racial balancing), III (failure to use race merely 2 Although discovery ended on May 1, 2018, [ECF Nos. 363, 364], the Court ordered supplemental document productions during trial when it became apparent that Harvard had modified its admissions procedures to provide admissions officers with more explicit guidance on the use of race despite seemingly contradictory testimony by various witnesses. See [ECF No. 645 at 7:20–19:24]. 5 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 6 of 130 as a “plus” factor), and V (race-neutral alternatives), and from October 15 through November 2, 2018, the Court heard testimony from eighteen current and former Harvard employees, four expert witnesses, and eight current or former Harvard College students who testified as amici curiae. On February 13, 2019, following the parties’ submissions of proposed findings of fact and conclusions of law and responses to each other’s respective submissions, see [ECF Nos. 619, 620], the Court heard final closing arguments. The Court now makes the following findings of fact and conclusions of law in accordance with Federal Rule of Civil Procedure 52(a). III. FINDINGS OF FACT: DIVERSITY, ADMISSIONS PROCESS, AND LITIGATION A. Diversity at Harvard 1. Harvard’s Interest in Diversity It is somewhat axiomatic at this point that diversity of all sorts, including racial diversity, is an important aspect of education. See Brown v. Bd. of Educ., 347 U.S. 483 (1954). 3 The 3 On October 30, 2018, the Court heard testimony from Dr. Ruth Simmons, the current President of Prairie View A&M University. President Simmons was born in a sharecropper’s shack on a plantation in Grapeland, Texas. She attended primary and secondary school in a completely segregated environment in Houston, and then Dillard University, an African American institution supported by the Methodist Church in New Orleans. President Simmons was selected to spend her junior year of college at Wellesley, where she studied alongside white students in the United States for the first time. After graduating from Dillard University, President Simmons traveled to France, where she studied as a Fulbright Scholar. She then returned to the United States and earned a Ph.D. from Harvard’s Department of Romance Languages and Literatures. President Simmons held positions at Princeton University, Spelman College, and Smith College before becoming President of Brown University. She retired from Brown University after eleven years and returned to Texas, where she worked on nonprofit projects in the Houston area before being persuaded to come out of retirement to serve as the president of Prairie View A&M. President Simmons offered expert testimony on Harvard’s interest in diversity. Her testimony and her life story, perhaps the most cogent and compelling testimony presented at this trial, demonstrate the extraordinary benefits that diversity in education can achieve, for students and institutions alike. See [Oct. 30 Tr. 6:11–70:23]. 6 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 7 of 130 evidence at trial was clear that a heterogeneous student body promotes a more robust academic environment with a greater depth and breadth of learning, encourages learning outside the classroom, and creates a richer sense of community. See [Oct. 19 Tr. 185:23–187:24; Oct. 23 Tr. 24:13-20, 31:2–34:11, 59:8–14; Oct. 30 Tr. 27:20–28:8]. The benefits of a diverse student body are also likely to be reflected by the accomplishments of graduates and improved faculty scholarship following exposure to varying perspectives. See [Oct. 30 Tr. 28:9–30:11]. Harvard College’s mission, as articulated in its mission statement, is “to educate the citizens and citizen-leaders for our society” and it seeks to accomplish this “through . . . the transformative power of a liberal arts and sciences education.” [DX109 at 1]. 4 In aid of realizing its mission, 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. See [Oct. 17 Tr. 182:17–183:7; Oct. 23 Tr. 24:13–20]. This interest in diversity and the wide-ranging benefits of diversity were echoed by all of the Harvard admissions officers, faculty, students, and alumni that testified at trial. SFFA does not contest the importance of diversity in education, but argues that Harvard’s emphasis on racial diversity is too narrow and that the full benefits of diversity can be better achieved by placing more emphasis on economic diversity. See [ECF No. 620 ¶¶ 216, 231]. Consistent with Harvard’s view of the benefits of diversity in and out of the classroom, Harvard tries to create opportunities for interactions between students from different backgrounds and with different experiences to stimulate both academic and non-academic learning. [Oct. 23 Tr. 39:3–17; Oct. 30 Tr. 25:11–26:6, 27:20–28:8]. As examples, student living assignments, the available extracurricular opportunities, and Harvard’s athletic programs 4 “DX” refers to an exhibit offered by Harvard. 7 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 8 of 130 are all intended to promote a sense of community and encourage exposure to diverse individuals and viewpoints. [Oct. 23 Tr. 39:18–41:23]. Harvard has evaluated and affirmed its interest in diversity on multiple occasions. See [Oct. 17 Tr. 182:4–14]; see, e.g., [PX302; DX26; DX53]. 5 Most recently, in 2015, Harvard established the Committee to Study the Importance of Student Body Diversity, which was chaired by Dean Rakesh Khurana 6 (the “Khurana Committee”). [Oct. 23 Tr. 34:12–22]. The Khurana Committee reached the credible and well-reasoned conclusion that the benefits of diversity at Harvard are “real and profound.” [PX302 at 17]. It endorsed Harvard’s efforts to enroll a diverse student body to “enhance[] the education of [its] students of all races and backgrounds [to] prepare[] them to assume leadership roles in the increasingly pluralistic society into which they will graduate,” achieve the “benefits that flow from [its] students’ exposure to people of different backgrounds, races, and life experiences” by teaching students to engage across differences through immersion in a diverse community, and broaden the perspectives of teachers, to expand the reach of the curriculum and the range of scholarly interests. [PX302 at 1–2, 6]; see also [Oct. 23 Tr. 37:14–38:17]. The Khurana Committee “emphatically embrace[d] and reaffirm[ed] the University’s long-held view that student body diversity – including racial diversity – is essential to [its] pedagogical objectives and institutional mission.” [PX302 at 22]. 2. Admissions Office’s Efforts to Obtain a Diverse Applicant Pool Harvard’s Office of Admissions and Financial Aid (the “Admissions Office”) is tasked with deciding which students to accept to the College and which to reject or waitlist. [Oct. 15 5 “PX” refers to an exhibit offered by SFFA. 6 Dean of Harvard College Rakesh Khurana attended SUNY-Binghamton and Cornell University for his undergraduate studies. He received a Ph.D. in organizational behavior from Harvard University. [Oct. 22 Tr. 192:17–193:11]. 8 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 9 of 130 Tr. 64:1–70:8]. Deciding which applicants to admit is challenging given the overall talent and size of the applicant pool. For example, there were approximately 35,000 applications for admission to the class of 2019. [Oct. 17 Tr. 184:2–4]. Harvard, targeting a class size of roughly 1,600 students, admitted only about 2,000 of those applicants, based on its expectation that approximately 80% of admitted students would matriculate. [Id. at 184:22–185:11]. 7 Among the applicants for that class, approximately 2,700 had a perfect verbal SAT score, 3,400 had a perfect math SAT score, and more than 8,000 had perfect GPAs. [Id. at 184:14–21]. Clearly, given the size and strength of its applicant pool, Harvard cannot admit every applicant with exceptional academic credentials. To admit every applicant with a perfect GPA, Harvard would need to expand its class size by approximately 400% and then reject every applicant with an imperfect GPA without regard to their athletic, extracurricular, and other academic achievements, or their life experiences. Because academic excellence is necessary but not alone sufficient for admission to Harvard College, the Admissions Office seeks to attract applicants who are exceptional across multiple dimensions or who demonstrate a truly unusual potential for scholarship through more than just standardized test scores or high school grades. [Id. at 181:12–183:7]. To help attract exceptionally strong and diverse annual applicant pools, Harvard engages in extensive and multifaceted outreach efforts. Each year, roughly 100,000 students make it onto Harvard’s “search list” through data, including test scores, that the college purchases from ACT 8 which administers the ACT, and the College Board, which administers the PSAT and the SAT. 7 Harvard admitted 5.8% of applicants to its class of 2017 and 5.7% to its class of 2018. [Oct. 15 Tr. 157:21–25]. 8 The American College Testing Company changed its name to ACT in the 1990s. 9 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 10 of 130 [Oct. 15 Tr. 130:2–131:1; Oct. 17 Tr. 146:2–16]. High school students who make the search list receive a letter that encourages them to consider Harvard and may also receive follow-up communications. See [Oct. 15 Tr. 131:5–134:16; Oct. 17 Tr. 146:3–12; PX55]. Harvard also uses the search list to target students as part of its extensive in-person recruiting efforts, which includes Harvard admissions officers travelling to over 100 locations across the United States to speak with potential applicants and encourage them to consider Harvard. [Oct. 15 Tr. 131:13– 20; Oct. 17 Tr. 146:7–12, 179:8–21]. The search list is also sent to Harvard’s “schools committee,” which is comprised of more than 10,000 alumni who help recruit and interview applicants and help persuade admitted students to attend Harvard. [Oct. 15 Tr. 131:21–132:7]. In addition to recruiting students based largely on test scores, Harvard places particular emphasis on communicating with potential low-income and minority applicants whose academic potential might not be fully reflected in their scores. Since the 1970s, Harvard has recruited minority students, including Asian Americans, through its Undergraduate Minority Recruitment Program (“UMRP”). [Oct. 24 Tr. 95:15–21]. The UMRP writes letters, calls, and sends current Harvard undergraduates to their hometowns to speak with prospective applicants. [Id. at 95:12– 102:3]. The program, led by a full-time director and an assistant director, employs between two and ten Harvard students for most of the year, with twenty-five to thirty students working for the program during its peak season. [Id. at 201:1–204:22]. Despite these efforts, African American and Hispanic applicants remain a relatively modest portion of Harvard’s applicant pool, together accounting for only about 20% of domestic applicants to Harvard each year, even though those groups make up slightly more than 30% of the population of the United States. See [PX623; DX713]; U.S. Census Bureau, QuickFacts, Census.gov, https://www.census.gov/quickfacts/fact/table/US/RHI225218. In contrast, Asian 10 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 11 of 130 American high school students have accounted for approximately 22% of total applicants in recent years, although Asian Americans make up less than 6% of the national population. See [DX713]; U.S. Census Bureau, QuickFacts, Census.gov, https://www.census.gov/quickfacts/fact/table/US/RHI225218. Harvard’s recruiting efforts also target low-income and first-generation college students irrespective of racial identity through a recruiting program that operates in conjunction with the Harvard Financial Aid Initiative (“HFAI”). Harvard’s financial aid program guarantees full funding of a Harvard education for students from families earning $65,000 or less per year and also caps contributions at 10% of income for families making up to $150,000 per year. [Oct. 24 Tr. 102:10–104:19; PX316 at 6]. Harvard, through the HFAI recruitment program, employs students who return to their hometowns and visit high schools to talk about the affordability of Harvard and other colleges with need-blind admissions programs. [Oct. 24 Tr. 144:1–22]. Today, more than half of Harvard students receive need-based aid. [Id. at 150:3–6]. B. The Admissions Process Several Harvard admissions officers testified generally about reviewing application files as well as about their review of specific files. The Court credits this testimony. They each described a time-consuming, whole-person review process where every applicant is evaluated as a unique individual. See, e.g., [Oct. 17 Tr. 205:6–223:10; Oct. 24 Tr. 174:19–175:23]; see also [DD1]. 9 Admissions officers attempt to make collective judgments about each applicant’s personality, intellectual curiosity, character, intelligence, perspective, and skillset and to evaluate each applicant’s accomplishments in the context of his or her personal and socioeconomic circumstances, all with the aim of making admissions decisions based on a more complete 9 “DD” refers to demonstrative evidence presented by Harvard. 11 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 12 of 130 understanding of an applicant’s potential than can be achieved by relying solely on objective criteria. [Oct. 16 Tr. 16:15–22; Oct. 17 Tr. 182:17–183:7, 209:16–223:10]; see, e.g., [Oct. 18 Tr. 22:9–48:4; DX293]. 1. The Application Students apply to Harvard either through the early action program or the regular decision program. 10 All applications are reviewed in the same way regardless of whether a student has applied for early action or regular decision. [Oct. 18 Tr. 15:5–10]; see [PX1]. The Admissions Office may accept, reject, or waitlist applicants, or, in the case of early action applicants, defer them into the regular decision applicant pool. [Oct. 18 Tr. 124:14–125:9]. Students who apply for early action are admitted at a higher rate than regular decision applicants. [Oct. 25 Tr. 242:19–243:17]. Students apply to Harvard by submitting the Common Application or the Universal College Application. [Oct. 17 Tr. 186:1–10; Nov. 1 Tr. 27:13–19]. A complete application generally includes standardized test scores, high school transcript(s), information about extracurricular and athletic activities, intended concentration and career, a personal statement, supplemental essays, teacher and guidance counselor recommendations, and other information about the applicant, including high school and personal and family background, such as place of birth, citizenship, disciplinary or criminal history, race, siblings’ names and educations, and 10 Harvard eliminated its early action program for the classes of 2012 through 2015, in part to improve the socioeconomic diversity of its students. [PX316 at 15]; see [DX728]. Eliminating early action, however, did not have the expected effect on class diversity, and Harvard’s peer institutions largely continued with their early action and early decision programs. [PX316 at 15]. Harvard became concerned that it was losing some of the most competitive applicants to other colleges that offered early decision or early action and decided to reverse course and reinstate its early action program for the class of 2016. [Oct. 17 Tr. 163:9–164:1; Oct. 18 Tr. 89:13–91:19; Oct. 22 Tr. 100:6–101:15, 185:2–186:8; Oct. 23 Tr. 158:14–160:19; DX39 at 4]. 12 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 13 of 130 parents’ education, occupation, and marital status. See, e.g., [DX195, DX262, DX276, DX293, DX527, SA1, SA2, SA3, SA4]. 11 Applicants can also supplement their applications with samples of their academic or artistic work, which may be reviewed and evaluated by Harvard faculty. [Oct. 17 Tr. 189:5–14; Oct. 18 Tr. 31:21–32:13]; see, e.g., [DX276 at 41; DX293 at 42]. Applicants may, but are not required to, identify their race in their application by discussing their racial or ethnic identity in their personal statement or essays or by checking the box on the application form for one or more preset racial groups (e.g. American Indian or Alaskan Native, Asian, Black or African American, Native Hawaiian or Other Pacific Islander, or White) and may also select or indicate a subcategory of these groups. See [Oct. 18 Tr. 52:8–14; Oct. 26 Tr. 98:2–6; SA2 at 4; SA3 at 8]. 12 If applicants disclose their racial identities, Harvard may take race into account, regardless of whether applicants write about that aspect of their backgrounds or otherwise indicate that it is an important component of who they are. [Oct. 26 Tr. 91:17– 92:1]. 2. Alumni and Staff Interviews Most applicants interview with a Harvard alumnus. [Oct. 15 Tr. 128:2–6]. Harvard selects alumni to interview candidates based predominantly on geographic considerations. Alumni interviewers are provided with an Interviewer Handbook that describes the admissions process. [Id. at 127:9–128:1]; see [DX5]. Although interviewers have broad discretion in deciding where to conduct the interview, what information to request in advance, and what to 11 “SA” refers to evidence offered by student amici. 12 Harvard could elect not to receive information about applicants’ race for all applicants or some racial subgroups. In fact, Harvard no longer receives information about applicants’ religious affiliation, [Oct. 19 Tr. 186:7–187:18], although it does continue to receive some information about applicants’ religions and beliefs from applicants who choose to write about their religious identities in their essays or their personal statements, [id. at 246:25–247:17]. 13 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 14 of 130 ask, Harvard specifies several questions that alumni interviewers should not ask and also instructs alumni not to advise applicants on their chances of admission, given that “this analysis can only be accomplished with full access to all the material in an applicant’s file and through the extensive discussions shared and comparisons made through the Committee process.” [DX5 at 30–34]. Alumni interviews generally last from 45 minutes to an hour. [Oct. 17 Tr. 218:25– 219:9]. Alumni interviewers do not have all of the information that is available to admissions officers at the time of admissions decisions, but their evaluations can be uniquely helpful to admissions officers, as alumni interviews are often an applicant’s sole in-person interaction with a Harvard representative. [Id. at 219:17–220:10]. 13 Alumni interviewers complete an evaluation form that requests numerical ratings for applicants in academic, personal, and overall categories that align with the rating categories later used by Harvard admissions officers. See [PX88 at 50– 52]. 14 Alumni interviewers also score applicants in a single category that captures extracurricular and athletic activities, community involvement, employment, and family commitments, while admissions officers score applicants in separate extracurricular and athletic categories. See [PX88 at 51; SA1 at 29]; see also infra Section III.B.3. Ratings generally fall between 1 and 4, with 1 being the strongest. The ratings criteria used by alumni (i.e. when to 13 Alumni interviewers may ask students about their standardized test scores, interests, and high schools, but alumni generally do not have access to teacher recommendations, counselor reports, and transcripts, all of which are critical to admissions officers’ evaluation of applicants. [Oct. 17 Tr. 218:25–219:9]. 14 Alumni ratings for applicants’ personal and overall qualities may be reprinted by admissions officers on the summary sheets that sit at the front of application files. See [Oct. 17 Tr. 219:10– 13; DD1 at 15]; e.g. [DX276 at 1]. Some applicants are scored by admissions officers before alumni ratings are available. See [Oct. 24 Tr. 119:7–25]. 14 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 15 of 130 rate applicants 1, 2, 3, 4, or worse for the various rating categories) roughly correspond to the criteria used by the admissions officers. Compare [PX1 at 5–7], with [PX88 at 50–52]. Beyond providing numerical ratings, alumni interviewers write comments explaining their ratings on the interview evaluation form, which is then placed in the applicant’s file. See, e.g., [SA1 at 29]. Although the Interviewer Handbook contains a section on distinguishing excellences including “ethnic . . . factors,” alumni interviewers are not explicitly told to boost the ratings they assign to applicants based on race or ethnicity. [DX5 at 11]. Alumni interviewers are, however, told to “[b]e aware of, and suspect, your own biases” and that awareness of one’s biases is important because “no one can really be ‘objective’ in attempting to evaluate another person . . . .” [Id. at 35]. In addition to alumni interviews, which are offered to most applicants, a small percentage of applicants interview with an Admissions Office staff member. [Oct. 19 Tr. 177:14–19]. Although some staff interviews are offered on a first come, first served basis, many applicants secure staff interviews because they are well-connected or particularly attractive candidates, or because they are from a part of the country where an alumni interview may be unavailable. [Oct. 17 Tr. 219:14–220:12; Oct. 19 Tr. 175:8–181:14]. Students who have staff interviews tend to be among the strongest applicants and are admitted at a comparatively high rate. See [Oct. 19 Tr. 178:24–182:18]. 15 Asian American applicants are less likely to have a staff interview than white, African American, or Hispanic applicants. [PX619]. Among applicants who receive a staff interview, 59% of African Americans, 48% of Hispanics, 53% of whites and 44% of Asian 15 Less than 3% of all applicants, but more than 20% of recruited athletes, legacies, applicants on the dean’s or director’s interest lists, and children of faculty or staff (“ALDCs”) receive a staff interview. [PX619]. Approximately 52% of all applicants and 79% of ALDC applicants who receive staff interviews are admitted. [Id.]. 15 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 16 of 130 Americans are admitted. [Id.]. The lower admission rate for staff-interviewed Asian Americans is driven primarily by the fact that Asian American applicants are less likely than African American and Hispanic applicants, and far less likely than white applicants, to be recruited Athletes, Legacies, on the Dean’s or Director’s interest list, or Children of faculty and staff (“ALDCs”), all of whom are advantaged in Harvard’s admissions process. See [id.]. 16 3. Application Review Process i. Admissions Office and Personnel The Admissions Office is tasked with deciding which applicants to admit and which to reject or waitlist. See [Oct. 19 Tr. 160:1–11]. Dean of Admissions and Financial Aid William Fitzsimmons, 17 Admissions Director Marlyn McGrath, 18 and Financial Aid Director Sally Donahue 19 oversee the Admissions Office, which has approximately seventy employees, including the forty admissions officers who read applicant files and directly participate in the process of deciding which applicants to admit (the “Admissions Committee”). [Oct. 17 Tr. 180:3–13; Oct. 19 Tr. 232:18–20]. Harvard’s admissions staff is a diverse group of individuals that includes Asian Americans. [Oct. 18 Tr. 20:22–21:2]. Several admissions officers testified at trial and forcefully denied the suggestion that racial animus or conscious prejudice against 16 ALDCs are disproportionately white, with 8% of white applicants being ALDCs compared to 2.7% of African American, 2.2% of Hispanic, and 2% of Asian American applicants. [PX619]. 17 Dean Fitzsimmons began working in the Admissions Office in 1972 as an Assistant Director of Admissions. He later served as Director of Admissions and worked for the Harvard Fund, before becoming Dean of the Admissions Office in 1986. [Oct. 15 Tr. 123:6–124:13]. 18 Director McGrath received a Ph.D. in 1978 and became a Residential Dean at Harvard the same year. She also worked in academic planning for the Faculty of Arts and Sciences at Harvard, before becoming the Director of Admissions in 1987. [Oct. 19 Tr. 156:6–157:8]. 19 Director Donahue recently retired from her leadership role but continues to assist the Admissions Office, including by reading applications. [Oct. 19 Tr. 242:11–17]. Director Donahue did not testify at trial. 16 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 17 of 130 Asian Americans infect Harvard’s admissions process. See, e.g., [Oct. 24 Tr. 175:11–17]. Consistent with this, the Court finds no persuasive documentary evidence of any racial animus or conscious prejudice against Asian Americans. There is significant turnover in the Admissions Office, which frequently hires relatively young admissions officers who leave to pursue other opportunities after a few years. [Oct. 19 Tr. 233:4–240:3]; see [DX25 at 117–20]. New admissions officers go through an orientation process that includes training on evaluating applicants and how to consider race. [Oct. 18 Tr. 187:13–188:18; Oct. 19 Tr. 43:18–44:2; Oct. 24 Tr. 139:7–24, 222:12–223:14]. The training utilizes a casebook that contains lightly edited application files from past years, and new admissions officers are guided on how to evaluate and score applicants based on those files. See [Oct. 19 Tr. 257:2–20]. The first fifty or one hundred application files reviewed by a new admissions officer are also reviewed by a more senior admissions officer who gives feedback to the less-experienced colleague as part of the training process. See [Oct. 16 Tr. 13:16–20; Oct. 24 Tr. 139:18–22]. The Admissions Office holds an annual retreat that sometimes includes professional development sessions on evaluating applicants, and admissions officers receive an annual training from Harvard’s general counsel that covers the permissible use of race in the admissions process. [Oct. 19 Tr. 45:12–47:10]. The Admissions Office has not historically provided new admissions officers with any written guidance on how to consider race in the admissions process, although Harvard amended its admissions reading procedures in 2018 for the class of 2023 to explicitly instruct admissions officers that they “should not take an applicant’s race or ethnicity into account in making any of the ratings other than the Overall rating” and that for the overall rating “[t]he consideration of race or ethnicity may be considered only as one factor among many.” [PX723 at 3 (emphasis omitted)]; see [Oct. 16 Tr. 19:12–17]. 17 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 18 of 130 ii. Reading Procedures Applications are divided into geographic dockets based on high school location. [Oct. 16 Tr. 8:2–20; DX5 at 16]. A subcommittee of the full Admissions Committee is responsible for the initial evaluation of applications within each docket. [DX5 at 16–17]. Docket subcommittees generally include a senior admissions officer who serves as docket chair and three to six additional admissions officers. [Id. at 17]. Each subcommittee member is responsible for reading all applications from a subset of the docket’s high schools. [Oct. 17 Tr. 204:6–205:5]. Because the same reader and subcommittee review all applicants from the same high school, admissions officers develop a familiarity with their respective high schools’ grading practices, academic rigor, and recommendation styles, all of which help them to fairly and consistently evaluate applicants, both from particular high schools and across high schools within their docket. [Id.]; see [Oct. 24 Tr. 110:17–111:17]. Applications are initially reviewed by an admissions officer or “first reader” who assigns the applicant ratings based on reading procedures that are updated on an annual basis. See [PX1; DX5 at 17]. Except for the recent changes to the reading procedures to provide more explicit guidance on the use of race, the substantive guidance on rating applicants has remained largely the same in recent years. [Nov. 1 Tr. 123:19–124:21, 128:19–129:10, 168:16–172:25]; see [PX720; PX721; PX722; PX723; DX742; DX743; DX744]. First readers, and any subsequent readers, assign an overall rating; four profile ratings: (1) academic, (2) extracurricular, (3) athletic, and (4) personal; and at least three school support ratings that reflect the strength of each teacher and guidance counselor recommendation submitted on behalf of an applicant. [Oct. 17 Tr. 206:14–209:8, 217:15–218:3]. Application readers may also rate the strength of any additional recommendations submitted by an applicant. [Id. at 218:4–10]. The ratings generally 18 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 19 of 130 range from 1 to 4, with 1 being the strongest rating. [Oct. 16 Tr. 10:19–11:17; Oct. 17 Tr. 207:13–16]. Ratings of 5 and 6 are also available and indicate either weakness or special circumstances, for example where family responsibilities prevent the applicant from participating in extracurricular activities. [Oct. 16 Tr. 10:21–11:1; PX1 at 5–7]. Admissions officers may also use “+” (stronger) and “–” (weaker) signs to fine tune a rating, with a rating of 2+ being stronger than a rating of 2, which is stronger than a rating of 2–. [Oct. 16 Tr. 11:11–17]; see [Oct. 18 Tr. 31:2–8]. Each of the profile ratings assigned by the first reader and any subsequent readers are preliminary and used as a starting point for any later consideration of the applicant by a docket subcommittee or the full Admissions Committee. [Oct. 17 Tr. 221:6–19]. The academic rating reflects the applicant’s academic strength and potential based on grades, standardized test scores, letters of recommendation, academic prizes, any submitted academic work, and the strength of the applicant’s high school. See [id. at 209:16–210:14; Oct. 19 Tr. 55:4–9; Oct. 24 Tr. 113:5–12]. An academic rating of 1 indicates summa cum laude potential, a genuine scholar, and near-perfect scores and grades (in most cases) combined with unusual creativity and possible evidence of original scholarship; an academic rating of 2 indicates magna cum laude potential, superb grades, and mid- to high-700 SAT scores or a score above 33 on the ACT; an academic rating of 3 indicates cum laude potential, excellent grades, and mid-600 to low-700 SAT scores or an ACT score of 29 to 32; and an academic 4 indicates adequate preparation, respectable grades, and low- to mid-600 SAT scores or an ACT score of 26 to 29. [PX1 at 5–6]. The extracurricular rating is an assessment of an applicant’s involvement in activities during high school and his or her potential to contribute to the extracurricular student life at Harvard. [Oct. 17 Tr. 212:4–213:1]. It may also account for family or personal circumstances 19 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 20 of 130 that have limited the applicant’s participation in extracurricular activities. [Id. at 207:13–23]. An extracurricular rating of 1 indicates national-level, professional or other truly unusual achievement that suggests an applicant may be a major contributor at Harvard; an extracurricular rating of 2 indicates strong contributions to an applicant’s high school in one or more areas, such as being class president or achieving recognition for extracurricular accomplishments on a local or regional level; an extracurricular rating of 3 indicates solid participation but without special distinction; and an extracurricular rating of 4 indicates little or no participation. [PX1 at 6]. An athletic rating of 1 indicates that an applicant is a recruited athlete, an athletic rating of 2 indicates strong high school contribution and possibly leadership roles in athletics, an athletic rating of 3 indicates active participation, and an athletic rating of 4 indicates little or no participation in athletics. [Id.]. The personal rating reflects the admissions officer’s assessment of what kind of contribution the applicant would make to the Harvard community based on their personal qualities. [Oct. 17 Tr. 213:22–216:1; Oct. 18 Tr. 39:1–25]. Although the reading procedures have not historically provided detailed guidance on what qualities should be considered in assigning a personal rating, relevant qualities might include integrity, helpfulness, courage, kindness, fortitude, empathy, self-confidence, leadership ability, maturity, or grit. See [Oct. 17 Tr. 213:22–214:19; Oct. 19 Tr. 227:6–228:2; Oct. 24 Tr. 117:4–24]. For the application cycles that were the subject of the statistical analysis performed in this case, the reading procedures specified that a personal rating of 1 meant “outstanding,” 2 meant “very strong,” 3 meant “generally positive,” and 4 meant “bland or somewhat negative or immature.” [PX1 at 6; PX71 at 6]. The personal rating criteria, perhaps in response to this lawsuit, were overhauled for the class of 2023, and the reading procedures now explicitly state that “an applicant’s race or 20 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 21 of 130 ethnicity should not be considered in assigning the personal rating” and encourage admissions officers to consider “qualities of character” such as “courage in the face of seemingly insurmountable obstacles,” “leadership,” “maturity,” “genuineness, selflessness[,] humility,” “resiliency,” “judgment,” “citizenship,” and “spirit and camaraderie with peers.” [PX723 at 5]. The overall rating reflects the admissions officer’s impression of the strength of the application, taking account of all information available at the time the rating is assigned. [Oct. 18 Tr. 186:12–15; Oct. 19 Tr. 49:3–15; PX1 at 5]. An overall rating of 1 is exceptional and a clear admit, an overall 2 reflects strong credentials, an overall 3 indicates good credentials, and an overall 4 indicates respectable credentials. [PX1 at 5; DX744 at 3]. 20 Admissions officers are permitted to take an applicant’s race into account when assigning the overall rating. [Oct. 17 Tr. 221:3–5]. Applicants are also assigned school support ratings that indicate the strength of their teacher and guidance counselor recommendations. [Oct. 17 Tr. 217:15–218:10; Oct. 18 Tr. 204:3–22]. A school support rating of 1 indicates strikingly unusual support, a 2 indicates very strong support, a 3 indicates above average positive support, and a 4 indicates somewhat neutral or slightly negative support. [PX1 at 7]. Teacher and guidance counsel recommendations may inform the profile ratings, for example if a teacher discusses a student’s academic or extracurricular commitments, but the school support ratings are distinct from the profile ratings and do not impact the profile ratings in a formulaic manner. See [Oct. 31 Tr. 36:10–37:16]. Harvard also considers whether applicants will offer a diverse perspective or are exceptional in ways that do not lend themselves to quantifiable metrics. Harvard may give 20 The summaries here reflect the Class of 2018 reading procedures. Although the ratings guidelines are routinely revised, the guidelines and reading procedures for the classes of 2014 through 2019 do not differ in material respects. 21 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 22 of 130 applicants a “tip” for “distinguishing excellences,” such as capacity for leadership, creative ability, and geographic, economic, and racial or ethnic factors. See [Oct. 17 Tr. 191:8–200:20; DX5 at 9–11]. The Admissions Committee gives some applicants large tips for non-academic reasons where an individual’s talents or background suggests that admitting them will be especially beneficial to the Harvard community. See [DX5 at 11]. ALDCs are the four most notable groups of applicants, other than racial minorities, who receive such tips. [Oct. 17 Tr. 12:10–14:23, 198:22–201:17; Oct. 18 Tr. 48:14–21; Oct. 23 Tr. 204:10–16; PX104; PX106; PX111]. Recruited athletes receive a tip in the admissions process because they are being recruited by one of Harvard’s varsity sports teams and are presumably exceptionally talented, but legacy applicants, those on the dean’s or director’s interest lists, and children of faculty and staff obtain an admissions tip that is primarily or exclusively a product of family circumstances. Harvard’s objective in giving tips to applicants based on criteria other than individual merit, such as to legacies and the children of its faculty and staff, is to promote the institution and is unrelated to the racial composition of those applicant groups. [Oct. 17 Tr. 198:22–200:11]. When reviewing an application, “first readers” generally begin with the application summary sheet, which is a two to three page document that is prepopulated with much of the key information about an applicant, including the applicant’s high school, citizenship, test scores, GPA, class rank, and race. E.g. [DX195 at 2]. The summary sheet also contains blank spaces for ratings and notes, to be filled in by the first reader and a potential second reader. [Oct. 18 Tr. 22:18–23:3]; e.g. [DX195 at 2–4]. After reviewing an application file, the first reader rates the strength of the teacher and guidance counselor letters of recommendation, assigns the academic, extracurricular, athletic, personal, and overall ratings to the applicant, and writes any notes about the applicant. [Oct. 17 Tr. 206:24–207:12]. The reader then sends the application to the docket 22 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 23 of 130 chair if it merits further review, at which point the docket chair will review the file, record his or her own ratings of the applicant based on the same criteria, and add written comments. See [Oct. 19 Tr. 250:12–251:2]; e.g. [DX195 at 2–3]. Even if the first reader does not pass an application on for further review, the application and the first reader’s scoring remain available to all admissions officers and may be discussed later in the admissions process. [Oct. 18 Tr. 12:1–13, 16:7–17:5]. Although docket chairs are frequently the “second reader,” other admissions officers may also serve as a second reader as circumstances require, for example when the first reader is new to the Admissions Office. [Oct. 17 Tr. 206:1–13]. iii. Committee Meetings and Admissions Decisions After the application files for the early action or regular decision cycle have been reviewed by the early readers, the docket subcommittees meet as a group to collectively evaluate the applications in their dockets and come up with a list of recommended admits for the full Admissions Committee. [Id. at 204:10–12; Oct. 18 Tr. 12:14–13:5]. The subcommittees consider early admission applicants in November and meet again to consider regular decision applicants in late January or February. See [DX41]. First readers act as the advocate for the applicants whose applications they initially reviewed. [Oct. 16 Tr. 8:7–9:2; Oct. 17 Tr. 204:10– 12]. Subcommittees generally go through their docket of applications high school by high school, with the first readers for each high school presenting the applicants they view as legitimate contenders for admission. [Oct. 18 Tr. 9:20–10:7]. All applications on a subcommittee’s docket, including those that the first readers view as legitimate contenders and those that they do not intend to present to the subcommittee, are included in a binder which helps the subcommittee members compare and contrast applicants. [Id. at 108:8–11:25]. In some subcommittee meetings, summary information about the applicant under discussion, including 23 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 24 of 130 race, is projected on a screen so that it can be easily viewed by all subcommittee members during the discussion of that applicant. [Oct. 24 Tr. 191:23–192:24]. The subcommittees make recommendations on applicants, including to admit, waitlist, and reject, and may also place applications on hold to await additional information or defer an early decision applicant to the regular admissions pool. [Oct. 18 Tr. 12:14–13:5]. Subcommittees may take race into account in making these initial recommendations. [Oct. 24 Tr. 128:12–25]. The initial recommendations are not final, and the application review process is fluid. It is common for some applicants who are not initially recommended for admission by a subcommittee to be admitted, and for some applicants who are initially recommended for admission to be waitlisted or rejected, especially where more information about an applicant becomes available later in the admissions process. [Oct. 18 Tr. 13:6–15]. As the process progresses and after the subcommittees decide more definitively which applicants to recommend for admission, the full Admissions Committee, comprised of all forty admissions officers who read applications, meets to collectively decide which applicants to admit. [Id. at 13:18–21]. Additionally, there is a standing committee, which includes faculty members, that assists the Admissions Office in its review and evaluation of applications, and those faculty members are also invited to attend the full Admissions Committee meetings. [Id. at 13:19–14:8]. The full committee meets in late November and early December to discuss early action applicants and in March to consider regular decision applicants. [Id. at 14:9–11; DX41]. Almost all applicants who are recommended for admission by the subcommittees are discussed by the full committee. [Oct. 18 Tr. 15:17–19]. Additionally, every admissions officer has access to every application file and may call the full committee’s attention to applicants who have not been recommended by a subcommittee. [Id. at 12:1–13, 16:7–9]. Applications are 24 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 25 of 130 projected on a screen while the full committee discusses the applicant, and the full application file is available to committee members electronically. [Id. at 17:6–11]. At the time of the full committee meeting, there is often more information available to the full committee than was available to the application’s earlier readers and the applicable subcommittee because additional high school grades, alumni interview evaluations, and other information frequently becomes available later in the admissions process. [Id. at 17:12–20]. The full Admissions Committee makes decisions by in-person majority votes. [Id. at 17:21–18:2]. In making admissions decisions, Harvard’s goal is to admit the best freshman class for Harvard College, not merely a class composed of the strongest applicants based solely on academic qualifications. [DX5 at 9–10]. Although the reading procedures reflect the traits that Harvard looks for in applicants, Harvard does not decide which applicants to admit based on any formula. See [Oct. 17 Tr. 221:20–223:6]. As the Interviewer Handbook describes: The Admissions Committee values objective criteria, but holds a more expansive view of excellence. Test scores and grades indicate students’ academic aptitude and achievement. The Committee also scrutinizes applications for extracurricular distinction and personal qualities. Students’ intellectual imagination, strength of character, and their ability to exercise good judgment—these are other, critical factors in the admissions process, and they are revealed not by test scores but by students’ activity outside the classroom, the testimony of teachers and guidance counselors, and by alumni/ae interview reports. Seeking evidence of these three criteria—academic excellence, extracurricular distinction, and personal qualities— the Committee reads with care all the components of each applicant’s file: the high school transcript, standardized test scores, extracurricular activities, personal statement, teacher and secondary school recommendations, and the personal interview report. Attempts to define and to identify precise elements of character, and to determine how much weight they should be given in the admissions process, require discretion and judiciousness. But the Committee believes that the “best” freshman class is more likely to result if we bring evaluation of character and personality into decisions than if we do not. We believe that a diversity of backgrounds, academic interests, extracurricular talents, and career goals among students who live and learn together affects the quality of education as much as a great faculty or vast material resources. 25 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 26 of 130 [DX5 at 10]. The Admissions Office sets a target number of students to admit based on the roughly 1,600 spots available each year and the expected matriculation or yield rate for admitted applicants. See [Oct. 15 Tr. 160:18–161:5]. After the full committee completes its review of all applicants recommended for admission, Harvard often needs to remove some students from the admit list to reach its target number of admitted students. [Oct. 23 Tr. 191:1–4]. When it becomes necessary to reduce the list of prospective “admits”, the Admissions Committee uses a “lop process” in the closing days of the full committee meetings that involves discussing candidates again and then “lopping” some from the admit list. [Oct. 24 Tr. 130:22–131:10; Nov. 1 Tr. 244:3–245:15]. 21 In aid of this, a potential lop list is prepared that may contain the HFAI status, athletic rating, legacy status, gender, and race of the applicants whom the committee is expected to consider lopping. [Oct. 24 Tr. 131:16–24]. Dean Fitzsimmons then informs the Admissions Committee of the characteristics of the admitted class, which may include racial composition, and the committee decides, as a group, which students to lop off the admit list based on many factors, which may include race. See [id. at 196:1–200:16]. After the Admissions Committee concludes the full committee meetings, applicants are notified whether they have been admitted, wait-listed, or rejected, or in the case of early action students, whether they have been deferred into the regular decision process. See [Oct. 18 Tr. 124:16–125:9]. Additionally, some applicants may be offered deferred admission or “z-listed,” meaning they are offered a spot in the class following the class year for which they applied. [Oct. 19 Tr. 167:25–168:23]. 21 Some subcommittees engage in a similar lop process, as they select students to be recommended to the full committee for admission. [Oct. 24 Tr. 130:22–131:6]. 26 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 27 of 130 Application Review Process [DD1 at 4]. 4. Harvard’s Use of Race in Admissions Throughout the admissions process, the Admissions Office leadership tracks the racial composition of the applicant pool, the students recommended for admission to the full committee, and the students admitted by the full committee. The composition of applicants and admitted students helps the Admissions Office see how well its efforts to achieve a diverse class are working by showing, for example, whether Harvard is seeing increases in applications from students with the backgrounds that it has placed a special emphasis on recruiting, and whether minority students have been admitted in numbers that will likely lead to a racially diverse entering class. See [Oct. 18 Tr. 81:20–82:18]. To do this tracking, Dean Fitzsimmons, Director McGrath, and a few other admissions officers receive “one-pagers” that provide a snapshot of the projected class and compare it to the prior year. [Id. at 80:2–5; Oct. 23 Tr. 178:21–179:10]. The one-pagers contain statistics on applications and admission rates by gender, geography, academic interest, legacy status, financial aid circumstances, citizenship status, racial or ethnic group, and on recruited athlete status and applicants flagged as disadvantaged. [Oct. 18 Tr. at 77:5–78:2]; e.g. [PX165 at 2]. 27 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 28 of 130 The racial breakdown shown on the one-pagers is provided based on three methodologies, the “old methodology,” the “new methodology,” and the federal government’s “Integrated Postsecondary Educational Data System” (“IPEDS”), [Oct. 18 Tr. 78:3–13]; e.g. [PX165 at 3], with the Admissions Office preferring the new methodology. 22 [Oct. 18 Tr. 81:6–19, 85:5–7]. Dean Fitzsimmons shares the breakdown of the admitted class as reflected on the onepagers with the full committee from time to time. [Id. at 80:6–18; Oct. 19 Tr. 195:21–196:16]. For example, at the start of the full Admissions Committee meetings, he usually states how many students are being recommended for admission by the subcommittees and how the breakdown of the class compares to the prior year in terms of racial identities and other demographics. [Oct. 24 Tr. 83:7–16; Oct. 26 Tr. 104:22–106:14]. The leadership of the Admissions Office monitors the breakdown of the class as the full committee meetings progress and through the lop process. See [Oct. 23 Tr. 181:4–23]. Although there are no quotas for subcategories of admitted students, if at some point in the admissions process it appears that a group is notably underrepresented or has suffered a dramatic drop off relative to the prior year, the Admissions Committee may decide to give additional attention to applications from students within that group. [Oct. 19 Tr. 198:23– 200:10]. 23 22 The new methodology better reflects the racial diversity that results from students who identify with multiple racial groups than the IPEDS methodology. [Oct. 18 Tr. 83:17–84:9]. Harvard has found the IPEDS methodology less reflective of the actual diversity of its class because, for example, it classifies all applicants who identify as Hispanic as only Hispanic irrespective of other racial groups they may also identify with. [Id. at 84:10–24]. This avoids double counting but results in the underreporting of the representation of minority racial and ethnic groups because many students identify with two or more racial groups. [Id. at 84:10–85:7]. 23 Harvard also shares statistics on admissions by race with the Association of Black Admission and Financial Aid Officers at the Ivy League and Sister Schools to learn about the practices of other schools. [Oct. 24 Tr. 83:17–85:17]. 28 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 29 of 130 In addition to giving the Admissions Office some perspective on whether it is admitting a diverse class, the collective racial composition of applicants and admitted students helps Harvard better forecast its overall yield rate because different racial groups historically accept offers to attend Harvard at differing rates. [Oct. 15 Tr. 160:18–162:7]. As examples, admitted Asian American students usually matriculate at a higher rate than white students, while admitted Hispanic, African American, Native American, and multiracial applicants matriculate at a lower rate. [Oct. 18 Tr. 80:21–81:5]; see [PX324]. Because of these variations in yield rates by racial group, Harvard uses the racial makeup of admitted students to help determine how many students it should admit overall to avoid overfilling or underfilling its class. See [Oct. 15 Tr. 162:1–15]. In addition to monitoring the likely racial makeup of the admitted class, admissions officers use race in evaluating applicants and assigning an overall rating. [Oct. 17 Tr. 221:3–5; Oct. 18 Tr. 49:20–50:3, 186:16–25]. Although race may act as a tip or plus factor when making admissions decisions, it is only ever one factor among many used to evaluate an applicant. [Oct. 18 Tr. 49:10–16, 167:2–169:24]; see [DX5 at 11]. Race is only intentionally considered as a positive attribute. [Oct. 16 Tr. 22:18–23:4; Oct. 18 Tr. 197:5–11]; see [Oct. 30 Tr. 80:1–23]. Admissions officers are not supposed to, and do not intentionally, take a student’s race directly into account when assigning ratings other than the overall rating, but Harvard’s reading procedures did not instruct readers not to consider race in assigning those ratings until 2018, when Harvard amended the reading procedures for the class of 2023 to provide more explicit guidance on the appropriate use and non-use of race. See [Oct. 18 Tr. 49:20–50:3; Oct. 19 Tr. 252:21–253:13; Oct. 24 Tr. 121:21–122:4, 140:6–25; Nov. 1 Tr. 124:3–125:11; PX723 at 1, 3]. Further, some admissions officers may take an applicant’s race into account indirectly, for 29 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 30 of 130 example when an applicant’s race has influenced other personal qualities that the admissions officer believes will add to the Harvard community. [Oct. 19 Tr. 48:11–49:1; Oct. 24 Tr. 138:1– 10]. No admission officer who testified perceived Harvard to be engaged in discrimination against Asian Americans. For example, Senior Admission Officer Charlene Kim 24 was asked what her reaction was to the allegation that Harvard discriminated against Asian Americans. She responded: I think now just concern. It’s not what I know our office to be. It’s not who I am. . . . I would never be part of a process that would discriminate against anybody, let alone people that looked like me, like my family, like my friends, like my daughter. And so I’m actually really grateful to be able to be here to share my little bit of my experience on the admissions committee . . . . I’m not here to say that it’s perfect, but I know that we don’t discriminate against anyone. [Oct. 24 Tr. 175:11–22]. To summarize the use of race in the admissions process, Harvard does not have a quota for students from any racial group, but it tracks how each class is shaping up relative to previous years with an eye towards achieving a level of racial diversity that will provide its students with the richest possible experience. It monitors the racial distribution of admitted students in part to ensure that it is admitting a racially diverse class that will not be overenrolled based on historic matriculation rates which vary by racial group. Although racial identity may be considered by admissions officers when they are assigning an applicant’s overall rating, including when an applicant discloses their race but does not otherwise discuss it in their application, race has no specified value in the admissions process and is never viewed as a negative attribute. 24 Ms. Kim is a senior admissions officer, the assistant director of financial aid, and the director of Harvard’s first-generation program. She graduated from the University of California at Berkeley and received a master’s degree from New York University. She began working in the Admissions Office in 2008. [Oct. 24 Tr. 125:12–25, 141:18–142:1]. 30 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 31 of 130 Admissions officers are not supposed to, and do not intentionally, consider race in assigning ratings other than the overall rating. C. Prelude to this Lawsuit 1. The Unz Article This lawsuit followed magazine and news articles that raised the specter of Asian American students being penalized in college admissions based on their racial identity. Harvard’s response to that controversy demonstrates Harvard’s concern about the perception that its admissions process was racially biased but also the complexity of the statistical evidence upon which the allegations here are based. On or about November 28, 2012, Ron Unz, a Harvard alumnus, published an article titled “The Myth of American Meritocracy” in The American Conservative (the “Unz Article”). [PX218]. Unz asserted that elite universities were biased against Asian Americans and employed “de facto Asian quotas” as evidenced by a gap between Asian American representation among America’s most academically accomplished high school students and their comparatively low representation at elite colleges. [Id. at 9]. The Unz Article, which itself included language that suggested certain unsavory biases, 25 did not attract much attention until approximately one month later when David Brooks of the New York Times published an article that promoted the Unz Article as one of the best magazine articles of the year and argued that stagnant Asian American representation at Harvard between 1995 and 2011 smelled like a quota system. See [Oct. 17 Tr. 24:19–25:17]. The two articles together and their allegations of racial 25 The article relies in part on data based on perceptions about the proportion of national merit scholarship semifinalists from California whose “names seem to be Jewish.” [PX218 at 12]. Although the Court recognizes that this article might have interested some sociologists, it was not unreasonable for some Harvard admissions officials to view the article as “profoundly antiSemitic” and, as a result, to view it as less than serious scholarship. [Oct. 17 Tr. 158:2–159:10]. 31 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 32 of 130 bias sparked concern among Harvard’s leadership and some of its alumni, who encouraged Harvard to respond to the allegations. See [id. at 25:8–37:25; PX227; PX238]. 2. Analysis by Office of Institutional Research i. Mark Hansen’s Admissions Models Following the 2012 Christmas and 2013 New Year’s holidays, Dean Fitzsimmons attempted to develop a response to the Unz Article, including soliciting input from Harvard’s Office of Institutional Research (“OIR”). [Oct. 17 Tr. 37:14–38:16; Oct. 23 Tr. 208:13–209:21; PX230; PX236; PX238]. 26 As part of OIR’s initial evaluation of the statistical evidence, research analyst Mark Hansen 27 prepared four rough logistic regression models, using data on applicants and admission outcomes for the classes of 2007 through 2016, to project Harvard’s admitted classes using a limited set of variables, including applicants’ race. [Oct. 24 Tr. 14:5– 26 OIR is a university-wide office that provides statistical analysis in response to requests from across Harvard University and sometimes on its own initiative when it anticipates a need for such work. During the relevant time period, the office typically had approximately 30 ongoing projects and received numerous additional ad hoc requests each year. [Oct. 19 Tr. 126:5–23]. OIR’s objective was and remains to offer accurate, timely, and digestible research that is tailored to diverse audiences with the goal of promoting informed decision-making and furthering the core missions of the university. [Oct. 18 Tr. 210:9–14; PX465]. 27 Mr. Hansen studied mathematics at Boston University before obtaining a master’s degree from Harvard’s Graduate School of Education. He was hired as a management fellow by OIR in the summer of 2010 and was promoted to research analyst in 2011. He left OIR in the summer of 2013 to work for MIT’s Office of Institutional Research. [Oct. 24 Tr. 10:19–11:25]. 32 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 33 of 130 24]; see [PX12 at 32–35]. 28 His most expansive model used applicants’ academic index, 29 academic rating, legacy and recruited athlete status, personal rating, extracurricular rating, gender, and race as inputs to predict the admitted class. See [PX12 at 33]. The classes projected by this model had racial demographics that approximated the actual class based on the probability of admission assigned to applicants by the model. See [id. at 34–35]. Mr. Hansen’s less complete models, which did not include variables for racial identities, projected admitted classes with far more Asian students than Harvard’s actual admitted classes, suggesting either that racial tips resulted in fewer Asian students being admitted or that factors correlated with Asian identity that were not included in Mr. Hansen’s models were significantly affecting which applicants Harvard chose to admit. See [id. at 33–34]. 28 At trial, SFFA emphasized a 17-page draft presentation, replete with blank spaces and typographical errors, that Mr. Hansen prepared in February 2013 but did not circulate to others. See [PX9]. In this draft presentation, Mr. Hansen summarized his findings as follows:     Athletes and Legacies explain the difference in raw admit rates for Asian and White applicants. Asian applicants have higher average ratings and test scores (excluding the personal rating). Differences exist in the raw admit rates of Asian and White students with similar test scores and academic indices. Even top scores and ratings don’t guarantee admission. Personal rating is important in models of the admissions process and drive some of the demographic differences we see. [Id. at 2]. Much of the information in the draft presentation, including the above summary, was never shared with the Admissions Office. See [PX12]. Further, it does not appear that anyone affiliated with Harvard other than Mr. Hansen, saw the draft report prior to this litigation. [Oct. 19 Tr. 111:14–22; Oct. 24 Tr. 50:9–14]. 29 The academic index is a metric that provides an indication of overall strength by taking account of standardized test scores and high school grades. [Oct. 16 Tr. 84:9–23]. 33 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 34 of 130 Mr. Hansen’s models could lead a casual observer to conclude that race plays a significantly larger role in Harvard’s admissions process than it actually does. The models incorporate far fewer variables than those prepared by the parties’ economic experts for this litigation and omit many variables that are important to the admissions process. Compare [PX12 at 33], with [PD38 at 26]. 30 Even Mr. Hansen’s most complete model almost certainly suffers from considerable omitted variable bias in light of the likely correlation between race and important variables that Mr. Hansen did not include. Most notably, his models contain no controls for socioeconomic and family circumstances that correlate with race and also affect admissions decisions. See [PX12 at 33]. Given these deficiencies in the models, they are entitled to little weight for the purpose of determining whether Harvard discriminates against Asian American applicants, particularly given the availability of the experts’ far more comprehensive models and the testimony offered by fact witnesses in this case. See [Oct. 19 Tr. 19:19–20:8]. Mr. Hansen’s models do suggest, consistent with other evidence, that Asian Americans applicants excel in academic metrics; that tips for legacies and recruited athletes result in more white students being admitted; that a projection of Harvard’s class based only on the profile ratings, academic metrics, and athlete and legacy statuses is incomplete and results in a projected class that is vastly less racially diverse than the one Harvard achieves; and that, absent any consideration of race, Harvard’s classes would have drastically fewer African American and Hispanic students. See [PX12 at 33–34]. 31 30 “PD” refers to demonstrative evidence presented by SFFA. 31 The Court notes that Mr. Hansen’s models suggest that any increase in Asian American admits would come largely at the expense of African Americans and Hispanics. 34 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 35 of 130 A limited selection of slides depicting Mr. Hansen’s logistic regression models were included in a February 25, 2013, presentation for Dean Fitzsimmons that focused on and included much more information on the reintroduction of Harvard’s early action program and an analysis of issues related to the accessibility and affordability of a Harvard education. [Oct. 17 Tr. 83:24–84:16; PX12 at 32–37]. The slides on Mr. Hansen’s models that were shared with Dean Fitzsimmons included a statement that they were “preliminary and for discussion,” and they were not presented or understood as evidence of discrimination. See [Oct. 17 Tr. 83:24– 84:16; PX12 at 32–36]. Dean Fitzsimmons concluded that Mr. Hansen’s models were incomplete, and he elected not to discuss those slides or the information they contained with Harvard’s leadership at that time. [Oct. 17 Tr. 84:3–85:1]. More than a year later, Mr. Hansen’s models were shared with Dean Khurana, shortly after he became the dean, in advance of a “highlevel meet-and-greet type meeting” that was intended to generally familiarize Dean Khurana with OIR’s work. [Oct. 23 Tr. 44:3–8, 45:6–10, 46:12–17]; see [PX41 at 50]. Dean Khurana also found Mr. Hansen’s models incomplete and viewed them as a puzzling approach to understanding Harvard’s admissions process. [Oct. 23 Tr. 47:4–49:18]. ii. Low-Income Admissions Models Following the February 2013 meeting with OIR, Dean Fitzsimmons requested that Dr. Erin Driver-Linn 32 and Ms. Erica Bever 33 further analyze the effect of low-income status, which Dean Fitzsimmons hoped and expected would confirm that Harvard was providing a tip to lowincome applicants. [Oct. 17 Tr. 172:22–173:21]. This analysis was intended to respond, at least 32 During the relevant period, OIR was led by Dr. Driver-Linn, who holds a Ph.D. in social psychology from Harvard. [Oct. 19 Tr. 69:9–70:7]. 33 Ms. Bever joined OIR in 2007 and transitioned to the Admissions Office where she now serves as a senior admissions officer and the director of research. [Oct. 18 Tr. 200:7–201:1]. 35 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 36 of 130 in part, to criticism that elite colleges, like Harvard, were not doing enough to attract low-income students. See [PX26 at 2]. On May 1, 2013, Ms. Bever, Dr. Driver-Linn, and Mr. Hansen sent Dean Fitzsimmons a summary of their initial findings in a memorandum titled “Harvard College Admissions and Low Income Students.” [Id.]. Their analysis found that Harvard students from lower income backgrounds generally have lower SAT scores but that they are admitted at higher rates when controlling for their SAT scores. [Id. at 2–3, 6–7]. After reviewing the distribution of SAT scores by family income, OIR’s memorandum discussed the need to model the admissions process to better evaluate whether the Admissions Office was providing a tip to low-income students, given that the relationship between income and admission, controlling only for SAT scores, could have been the result of a relationship between income and other factors, such as race. [PX26 at 3–4]; see [PX28 at 4 (indicating that applicants with family incomes of less than $60,000 accounted for 25% of Hispanic, 24% of African American, 18% of Asian American, and 10% of white applicants)]. As OIR’s memo to Dean Fitzsimmons summarized: The differences [in students’ SAT scores by income] could be related to other factors important in the admissions process. In order to control for those potential issues, we implement a logistic regression model to predict the probability of admission, controlling for demographic characteristics and a variety of metrics used to asses qualification for admission. Demographic characteristics include gender and race/ethnicity. Qualifications used in admission include academic index, academic rating, extracurricular rating, personal rating, athletic rating, and legacy status. This approach has several limitations; we picked a small set of variables that would factor in admissions decisions. The selection of a wider set of variables 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. For example, the model does not capture exceptional talent in art or music explicitly (although ratings may capture some aspect of these attributes). In addition, our model is limited to main effects, not examining interactions between variables. Our analysis should not be considered exhaustive. 36 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 37 of 130 [PX26 at 3]. To the extent that OIR’s initial analysis suggested that Harvard was providing an admissions tip to applicants from low-income backgrounds, that tip appeared less significant than tips for legacies and recruited athletes. See [id. at 8–9]. OIR explained that: To get a sense of the size of the admissions advantage conferred to low-income applicants relative to other groups of applicants, the so-called “thumb on the scale,” we include low-income status in a second logistic regression model. . . . The variables with the largest effects on the probability of admission are athletic rating, personal rating, and legacy status. Compared to athletes and legacies, the size of the advantage for low income students is relatively small. [Id. at 3]. The memorandum also noted that “Asian applicants with an academic 1 or 2 are admitted 12% of the time compared against an admit rate of 18% for non-Asian applicants” and provided a chart illustrating this disparity. [Id. at 4, 9]. Further, the memo stated that certain “issues” should be considered before sharing the analysis publicly, including that there are “demographic groups that have negative effects,” although the only demographic group for which OIR’s analysis returned a negative coefficient was “Asian.” [Id. at 4]. Although the model returned a negative coefficient for Asian applicants, neither OIR nor Dean Fitzsimmons viewed the report as indicative of discrimination. [Oct. 17 Tr. 109:15–19; Oct. 19 Tr. 152:22–153:15]. After receiving the May 1, 2013 memorandum, Dean Fitzsimmons asked OIR to examine the effect of Asian racial identity on admissions outcomes to confirm that the low-income tip was being fairly and consistently applied to all groups, but he did not ask OIR to further examine the effect of being Asian on admissions outcomes across the board. [Oct. 17 Tr. 127:22–128:12, 129:13–17]. OIR added an interaction term for Asian and Low Income which allowed the model to return coefficients that accounted for the possibility that the tip for low income varied by race. See [PX28 at 7–8]. On June 3, 2013, OIR shared with Dean Fitzsimmons its additional analysis, [Oct. 17 Tr. 129:13–130:13], which showed a coefficient for the interaction term of “Asian and 37 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 38 of 130 Low Income” that was positive and statistically significant but of a lesser magnitude than the negative coefficient for Asian identity, see [PX28 at 7; PX29]. This updated analysis suggested that although low-income Asian American applicants were provided a tip relative to their higher income Asian American peers, the magnitude of that tip might not overcome the negative relationship between Asian racial identity and admissions outcome, when holding constant some variation in the profile ratings, gender, and applicants’ academic index. [Id. at 7]; see also [DD10 at 27]. Nevertheless, the data reassured Dean Fitzsimmons that the Admissions Office was “treating Asian Americans in an evenhanded manner.” [Oct. 17 Tr. 134:3–11]. As with Mr. Hansen’s February 2013 models, OIR’s May 2013 models suffer from significant omitted variable bias, and the magnitude of the negative coefficient for Asian applicants is relatively modest considering the number and significance of omitted observable and unobservable factors. See [PX28 at 7]. As a result, the OIR analysis is weak evidence of bias against Asian American applicants, particularly relative to the more thorough econometric analysis that has been done by the parties’ economic experts in connection with this litigation. Dean Fitzsimmons’ non-inference of actual discrimination based on the relatively modest negative Asian coefficient was reasonable given the limitations of OIR’s model and his own experience with and confidence in the Admissions Office’s process. Dean Fitzsimmons did not ask for additional analysis based on OIR’s results, nor did he make any changes to Harvard’s admissions process in response to that analysis, because his review of the data did not lead him to believe that the Admissions Office was biased against Asian American applicants. [Oct. 17 Tr. 137:11–17, 138:7–24]. 3. The Ryan Committee In April 2014, Harvard learned of a website that had launched with the url harvardnotfair.com. Harvard’s staff recognized that the website was being promoted by some of 38 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 39 of 130 the same individuals who had financed Fisher v. University of Texas at Austin, 570 U.S. 297 (2013) (“Fisher I”), and 136 S. Ct. 2198 (2016) (“Fisher II”). [Oct. 23 Tr. 211:7–15]; see [PX283]. Apparently in response to the prospect of litigation, Harvard University formed a committee to examine race-neutral alternatives to its race-conscious admissions practices (the “Ryan Committee”). See [Oct. 22 Tr. 13:14–19, 129:13–130:17]. The Ryan Committee, chaired by Jim Ryan, the Dean of the Graduate School of Education, included more than two dozen members from across the university. [Oct. 16 Tr. 69:3–7; Oct. 22 Tr. 13:20–14:2]; see [PX300; PD19]. The committee’s work never really got “off the ground,” owing at least in part to its broad membership and the conflicting scheduling demands of many committee members. [Oct. 16 Tr. 69:10–70:15; Oct. 19 Tr. 76:8–77:10]. After meeting only a few times, it disbanded in December 2014, shortly after this lawsuit was filed. [Oct. 16 Tr. 70:2–6; PX316 at 2]; see [ECF No. 1]. No substantive analysis of any race-neutral alternatives examined by the Ryan Committee was entered into evidence. See [Oct. 19 Tr. 77:14–24 (“I believe the team did some work, under privilege. . . . Under direction of counsel.”)]. 4. The Khurana Committee In 2015, following the filing of this lawsuit and the disbandment of the Ryan Committee, Harvard established the Khurana Committee, officially titled “the Committee to Study the Importance of Student Body Diversity,” which was chaired by Dean Khurana. 34 [Oct. 22 Tr. 210:23–211:21; PX302 at 22]. The Khurana Committee “sought to examine and restate the 34 In addition to Dean Khurana, the members of the Committee to Study the Importance of Student Body Diversity included Mahzarin R. Banaji, the Richard Clarke Cabot Professor of Social Ethics; Emma Dench, the McLean Professor of Ancient and Modern History and of the Classics; Yukio Lippit, the Harris K. Weston Associate Professor of the Humanities; David R. Pilbeam, the Henry Ford II Professor of Human Evolution; and, Jonathan L. Walton, the Plummer Professor of Christian Morals and Pusey Minister of the Memorial Church. [Oct. 23 Tr. 35:14–18; PX302 at 22]. 39 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 40 of 130 benefits that the College derives – as an institution, and for its students and faculty – from student body diversity of all kinds, including racial diversity.” [PX302 at 1]. The Khurana Committee’s report, referenced supra at Part III.A, was prepared with the assistance of counsel and in the face of litigation, but nonetheless reflects an extensive and thoughtful examination of the benefits of diversity to Harvard College. [Oct. 22 Tr. 211:10–212:11]. The committee concluded its report by stating: We emphatically embrace and reaffirm the University’s long-held view that student body diversity – including racial diversity – is essential to our pedagogical objectives and institutional mission. It enhances the education of all of our students, it prepares them to assume leadership roles in the increasingly pluralistic society into which they will graduate, and it is fundamental to the effective education of the men and women of Harvard College. [PX302 at 22]. In February 2016, Harvard’s Faculty of Arts and Sciences voted unanimously to adopt the report. [Nov. 1 Tr. 198:19–24]. Although the Khurana Committee was keenly aware that it was addressing a question that “the Supreme Court has asked public institutions of higher education to answer in connection with the consideration of an applicant’s race in the admissions processes as one factor among many in an individualized review,” its focus was limited to Harvard’s interest in diversity, rather than the viability of race-neutral alternatives. See [PX302 at 1]. 5. The Smith Committee In June 2017, Harvard established the “Committee to Study Race Neutral Alternatives in Harvard College Admissions,” chaired by Michael Smith, the Dean of the Faculty of Arts and Sciences, with Dean Fitzsimmons and Dean Khurana serving as the other committee members (the “Smith Committee”). [PX316 at 1, 3]. The Smith Committee evaluated whether raceneutral means, singly or in combination, would enable Harvard to achieve its diversity-related educational objectives. [Id. at 8–9]. Prior to 2017, Harvard had repeatedly expressed the 40 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 41 of 130 importance of its race-conscious admissions policy and its understanding that diversity across multiple dimensions was critical. See generally [id.]. Harvard had not, however, conducted a detailed empirical analysis of the viability of race-neutral alternatives for at least fifteen years. See [Oct. 16 Tr. 66:21–67:6; Oct. 19 Tr. 194:3–195:3]. The Smith Committee worked with Harvard’s attorneys and had access to the analyses done by the experts in this case. [PX316 at 3]. The committee held seven meetings between August 2017 and April 2018 and then issued a report that was drafted by Harvard’s attorneys. [Oct. 23 Tr. 65:20–66:4; PX316 at 1, 3]. It examined all of the race-neutral alternatives proposed by SFFA, and additionally considered eliminating preferences for athletes and the use of test scores in the admissions process. See [PX316 at 6–18]. The Smith Committee concluded that no workable race-neutral admissions practices could, at that time, promote Harvard’s diversityrelated educational objectives while also maintaining the standards of excellence that Harvard seeks in its student body through its whole-person, race-conscious admissions program, and recommended that Harvard reexamine the issue in five years. [Oct. 22 Tr. 133:21–134:15; Oct. 23 Tr. 126:25–127:6, 134:15–19; PX316 at 18–19]. IV. FINDINGS OF FACT: NON-STATISTICAL EVIDENCE OF DISCRIMINATION As will be more fully discussed, the parties rely heavily on statistical evidence related to the admissions process. Additionally, to corroborate its statistical evidence, SFFA makes several other arguments in support of its contention that Harvard discriminates against Asian American applicants. A. Sparse Country First, as discussed above, Harvard uses a search list, which is primarily compiled based on potential applicants’ ACT, SAT, or PSAT test scores to help Harvard market itself to a diverse array of high school students. The ACT, SAT, or PSAT score that students need to make 41 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 42 of 130 the search list varies by gender, high school GPA, geography, and race. See [Oct. 15 Tr. 136:5– 139:21; PX2]. For example, to make Harvard’s class of 2018 search list, a white male high school student from outside “sparse country” 35 needed an SAT score of 1380, while black, Chicano, Hispanic, Native American, and Puerto Rican students needed only an 1100. See [PX2]; see also [PX50]. As SFFA points out, there are some anomalies in the search list selection criteria that are difficult to explain. As an example, assuming an applicant reported a sufficiently high GPA, for the class of 2018, Harvard lowered the SAT score required to make the search list to 1310 for students from “sparse country” who identified their race as white, other, or unidentified while not simultaneously lowering the required score for Asian American students from the same states to the same level. Consequently, Asian American students from the same states needed to score 1350 or 1380, depending on their gender, to make the search list. See [Oct. 15 Tr. 150:3–9; PX2; PX50]. Some Asian American students therefore did not make the search list, when white students from the same area who had similar grades and SAT scores did. See [Oct. 15 Tr. 151:22–152:2]. SFFA, while recognizing that the list is a marketing tool, would have the Court consider this “sparse country” disparity between the scores required for Asian Americans and whites to make the search list as evidence of Harvard’s intent to impose more selective admissions criteria on Asian Americans for the purpose of artificially suppressing Asian American representation at Harvard. 35 Sparse country for the purposes of the PSAT search includes twenty predominantly rural 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. [Oct. 15 Tr. 144:25–147:20; PX2]. 42 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 43 of 130 Notably, however, in some of the same years that Harvard did not lower the sparse country SAT search list score for Asian Americans commensurate with the lower requirement for whites, it selected Asian Americans for the search list based on lower ACT scores than similarly situated white students from more urban states. See [Oct. 17 Tr. 151:13–152:4; PX2]. Overall, the inconsistencies in the search criteria do not seem to be linked to efforts to advantage or disadvantage any particular racial group, and it was unclear from the testimony at trial whether these variations were accidental or intentional. At root, although being placed on the search list results in recruitment and is correlated with a higher likelihood of admission, the search list is fundamentally a marketing tool that does not affect individual admissions decisions. [Oct. 15 Tr. 129:24–132:25]. B. The OCR Report SFFA next points out that the specter of Harvard discriminating against Asian Americans in its admission process has been raised before. The argument is, at least in part, that repeated instances of smoke should heighten concerns about a fire. In the late 1980s, Harvard faced allegations of bias against Asian American applicants in its admissions process that culminated in a 1990 report by the United State Department of Justice Office of Civil Rights (“OCR”) (“OCR Report”). [PX555 at 2, 15–16]. The OCR Report reached an “overall conclusion that Harvard did not discriminate against Asian American applicants to its undergraduate program in violation of Title VI of the Civil Rights Act,” but its findings indicated that some admissions officers took race into account when assigning the personal rating during the period preceding the 1990 report. See [id. at 45–46]. Further, The OCR Report found recurring characterizations of Asian American applicants that were broadly consistent with stereotypes, noting that: 43 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 44 of 130 In addition to examining the ethnic reader’s comments, OCR’s concern for the potential stereotyping of Asian American applicants prompted a review of reader comments for negative characterizations which could have an impact on the admissions decision and ratings. On its face, reader comments revealed several recurring characterizations attributed to Asian American applicants. Quite often Asian American applicants were described as being quiet/shy, science/math oriented, and hard workers. For example, one reader’s comment embraced all of these in describing an Asian American applicant when she wrote: “[A]pplicant seems like a reserved, hard-working, aspiring woman scientist/doctor.” While such descriptions may not seem damaging, OCR was conscious that problems of “model minority” stereotypes could negatively impact Asian American applicants as a whole. This concern was also raised when OCR's file review came upon comments such as: “He’s quiet and, of course, wants to be a doctor . . .” suggesting that most or all Asian American applicants “want to be a doctor.” Or more pointedly: “[A]pplicant’s scores and application seem so typical of other Asian applications I’ve read: extraordinarily gifted in math with the opposite extreme in English.” OCR noted that in a number of cases, Asian American applicants were described as “quiet, shy, reserved, self-contained, soft spoken” and that these characteristics were underlined for added emphasis by the reader. While white applicants were similarly described, OCR found such descriptions ascribed to Asian American applicants more frequently. In some cases these comments actually originated from the interviews, teacher or counselor recommendations, or self-descriptions given by the applicant. .... OCR recognized that reader comments were also sometimes echoes of other reviewers’ commentaries related to the applicant. OCR also noted a few cases in which the readers referred to an applicant as “a classic V.N. [Vietnamese] bootstrap case” or “a classic BC/NC (blue collar/non-college background) Asian American from the inner-city.” While it was clear from the context of the statement that the readers were not criticizing the applicants, and that there was no negative intention, the comments do suggest a tendency to stereotype by calling the applicants “classic.” [Id. at 24–25]. Following the conclusion that Harvard did not discriminate against Asian American applicants and despite some of the specific findings, Harvard did not hold a meeting or 44 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 45 of 130 otherwise require that its admissions officers modify their evaluation practices to avoid actual stereotyping or the appearance of stereotyping. [Oct. 16 Tr. 59:17–24]. In the instant case, the admissions officers who testified at trial uniformly asserted that they do not and have not directly considered race in assigning ratings, other than the overall rating. 36 The Court credits the admissions officers’ testimony and concludes that Harvard has made clear to its admissions officers in more recent years that they should not use race in assigning the profile ratings. Harvard perhaps should have instituted an explicit written policy stating which ratings could take race into account before 2018, but that error has now been remedied. See [PX723 at 3, 5]. C. More Recent Allegations of Stereotyping and Bias SFFA also points to more recent examples of admissions officers referring to Asian American applicants as “quiet,” “hard worker,” “bright,” but “bland,” “flat,” or “not exciting.” See, e.g., [DX50 at 186, 178, 693, 1040, 1062]. Harvard’s admissions officers are tasked with carrying out a particularly delicate job in that they are instructed to consider race in the admissions process, including for applicants who 36 Senior Admissions Officer Christopher Looby’s deposition testimony is the sole instance in which an admissions officer allegedly admitted that race was directly used in assigning a personal rating between the 1990 OCR Report and the present. SFFA relied on Mr. Looby’s deposition testimony in its opening argument, stating, “he’ll tell the truth that he’s been using race in the personal rating for ten years.” [Oct. 15 Tr. 27:22–24]. Mr. Looby joined Harvard’s financial aid office in 2008 and has been reading admissions files since approximately 2010. See [Oct. 18 Tr. 148:16–25]. When asked during his deposition if he would “take a student’s race into account when assessing his or her personal qualities,” Mr. Looby answered that “just like with the academic rating, it’s one factor of many I consider.” [Id. at 182:8–19]; see also [Looby Dep. 51:12–17, June 30, 2017]. Mr. Looby testified at trial that he misunderstood the deposition question, and that he meant to state that he used race as one factor in assessing an applicant’s overall rating just as he considered the academic rating in assigning the overall rating. [Oct. 18 Tr. 185:19–23]. His response to the deposition question appears to have been a misstatement, and the Court concludes that Mr. Looby meant to indicate at his deposition that he would consider the academic rating and race in assigning the overall rating, not the personal rating. See [id. at 182:18–184:8]. 45 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 46 of 130 have indicated a race or ethnicity but have not elaborated on the importance of that identity, without engaging in unlawful discrimination. This job is especially sensitive due to the lengthy history of discrimination against many racial minorities in the United States, including Asian Americans, as well as Harvard’s own history of discriminating against Jewish applicants beginning in the 1920s. [Oct. 24 Tr. 188:17–25; Oct. 29 Tr. 34:22–35:13, 161:17–162:16]; see Korematsu v. United States, 323 U.S. 214 (1944), abrogated by Trump v. Hawaii, 138 S. Ct. 2392 (2018). It is true that Asian American applicants continue to face both positive and negative stereotypes, such as perceptions that they are timid, hard-working, and are inclined towards medicine and science. See [Oct. 29 Tr. 56:1–56:20]. It is also true that Asian Americans have significantly higher median incomes (perhaps indicative of the strong work ethic in many Asian American communities) 37 and are more likely to hold science, technology, engineering, and mathematics occupations than the United States population more broadly. 38 Therefore, in reviewing applicant files and comments made by admissions officers, the Court is sensitive to the challenge of differentiating among discriminatory comments that evidence actual stereotyping, animus, or racism and comments about a particular applicant that may incidentally 37 Although Asian Americans tend to have higher incomes than Americans with other racial identities, the evidence suggests that Asian American applicants to Harvard are more likely to come from modest socioeconomic backgrounds than white applicants. [PX28 at 2–5]. 38 See Anthony Martínez & Asiah Gayfield, The Intersectionality of Sex, Race, and Hispanic Origin in the STEM Workforce 8 (U.S. Census Bureau, Social, Econ., and Hous. Statistics Div., Working Paper No. 2018-27, Feb. 2019), https://www.census.gov/content/dam/Census/library /working-papers/2019/demo/sehsd-wp2018-27.pdf; Kayla Fontenot, Jessica Semega & Melissa Kollar, Income and Poverty in the United States: 2017 at 2–5, (U.S. Census Bureau Current Population Reports, Sept. 2018), https://www.census.gov/content/dam/Census/library/ publications/2018/demo/p60-263.pdf; Liana C. Landivar, Disparities in STEM Employment by Sex, Race and Hispanic Origin at 2, 7, 12, 16 (U.S. Census Bureau Am. Cmty. Survey Reports, Sept. 2013), https://www2.census.gov/library/publications/2013/acs/acs-24.pdf. 46 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 47 of 130 reference a stereotypical characteristic, like “hard working,” but which may also reflect an actual strength or weakness of that particular applicant. SFFA has not shown 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. The docket binder that contains notes to the effect that several Asian American applicants were “quiet” or “flat” also includes notes for white, African American, and Hispanic applicants who were also described as “quiet,” “shy,” or “understated.” [DX50 at 620, 975, 1054]. In the absence of a pattern or a more pervasive use of stereotypes, the Court accepts that there are Asian American applicants who were “quiet” and that the use of this word with regard to such an applicant would be truthful and accurate rather than reflective of impermissible stereotyping. In addition to SFFA’s concerns about Asian American applicants being referred to as “quiet” and the like, SFFA also points out that there is a statistical relationship between race and the use of the term “standard strong,” which some admissions officers use to indicate a strong applicant who is nonetheless unlikely to be admitted because he or she is not sufficiently distinguished within Harvard’s exceptional applicant pool. [Oct. 25 Tr. 133:20–134:1]. Asian Americans were labeled “standard strong” more frequently than white applicants, and significantly more frequently than African American or Hispanic applicants. See [id. at 135:4– 10]. In a sample of 10% of the applicants to the class of 2018, admissions officers noted that 255 students were “standard strong.” [Id. at 134:6–11]. Not one of the 255 standard strong applicants in the sample was admitted. [Id. at 135:16–18]. The standard strong applicants included 126 white applicants, 114 Asian American applicants, 12 Hispanic applicants, and 3 African American applicants. [Id. at 134:23–135:3]. Approximately 15% of Asian American applicants in the original 10% sample were labeled standard strong, compared to 12% of white 47 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 48 of 130 applicants, 4% of Hispanic applicants, and 1% of African American applicants. [Id. at 135:4–10; PD38 at 41]. Additionally, the Asian American applicants considered standard strong averaged higher academic indexes, math SAT scores and academic ratings than standard strong applicants from other racial groups. See [Oct. 25 Tr. 135:4–136:9]; see also [PD38 at 41]. These statistics on the use of “standard strong” are consistent with the profile ratings Harvard admissions officers assigned to Asian American applicants and white applicants, which show that Asian American applicants excelled, on average, on academic and extracurricular ratings, but were weaker when evaluated on personal and athletic criteria. See [PX621; PX622]. There is not a significant difference, however, between the white and Asian American applicants who were labeled “standard strong” as reflected by the sum of their profile ratings. See [Oct. 31 Tr. 94:16–97:18]; see also [DX709]. Further, the higher proportion of standard strong Asian American applicants is consistent with the fact that Asian American applicants to Harvard’s class are disproportionately unlikely to be among the weakest applicants: less than 21% of Asian American applicants received an overall rating of 4 or worse, compared to 24% of white applicants, 41% of Hispanic applicants, and 52% of African American applicants. [PX621]. As such, it is not surprising that a higher proportion of Asian Americans than white applicants were labeled standard strong. In addition to the use of phrases that align with stereotypes of Asian American applicants and the use of the words “standard strong,” SFFA has identified a few instances in which Harvard’s Admissions Office’s leadership acted in a manner that SFFA argues shows some degree of racial bias. Although the Court concludes that none of these incidents reflects any actual bias against Asian Americans by Harvard’s admissions officers, they do merit brief mention. 48 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 49 of 130 In April 2012, Director McGrath was asked to respond to a letter to President Drew Gilpin Faust from an elderly alumnus. See [PX461 at 3–6]. The alumnus’ letter argued that Harvard should be admitting more students from Massachusetts, proposed admissions quotas based on religious affiliation and skin color, and stated that Harvard has “a large number of oriental students.” [Id. at 5]. Director McGrath wrote a polite response, stating that the alumnus’ “comments on the importance of attracting a strong representation of students from Massachusetts resonates well in the Admissions Committee,” but also that Harvard “has become more representative of the ethnic and economic diversity of the country and, the University believes, better positioned to make significant contributions to the country.” [Id. at 1]. Director McGrath’s carefully crafted response rejected a proposal that was inconsistent with Harvard’s values and did not endorse the suggestions made in the letter, while seemingly trying not to alienate its author. In a January 2014 exchange, Director McGrath sent her daughter, who served as an alumni interviewer for Harvard, a list of the top applicants from Utah prepared by the alumni interviewers for that state, noting that she was “sending this along for your amusement. Pure Utah.” [PX265]. In responding, Director McGrath’s daughter wrote back, “Hahaha. Very Thorough!! I also love that the top-tier list is, as you’ve told me before, all Asians except for a couple . . . .” [Id.]. The email, which reflects amusement at the unusual degree of thoroughness of the Utah alumni interviewers, does not reflect a negative view of Asian Americans. 39 39 Director McGrath testified that she thought it notable that the Utah schools committee put all Asian Americans at the top of their list because it “confounds the stereotype that many people have of the population of Utah.” [Oct. 19 Tr. 221:3–11]. 49 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 50 of 130 In sum, comments on application files and Admissions Office correspondence do not suggest any pervasive bias against Asian Americans among Harvard’s admissions officers or its admissions leadership, nor has the Court identified any individual applicant whom it can determine was discriminated against or intentionally stereotyped by an admissions officer, including by the use of the words “standard strong.” V. FINDINGS OF FACT: STATISTICAL ANALYSIS A. Sources of Statistical Evidence In addition to testimony based on the lived experiences of witnesses, the parties introduced statistics and econometric models through expert witnesses. This statistical evidence is perhaps the most important evidence in reaching a resolution of this case, given SFFA’s heavy reliance on the data to make out its claims. Harvard presented its statistical evidence primarily through Professor David Card, and SFFA presented its statistical evidence primarily through Professor Peter Arcidiacono. 40 Both Professors Card and Arcidiacono are very well-qualified experts, but they fundamentally disagree about whether the statistics show that Asian Americans are discriminated against in the Harvard admissions process. Their disagreement results from differences in their respective statistical models of admissions outcomes, based on their inclusion of different applicants and use of different control variables. Therefore, decisions by the Court as to which applicants and control variables belong in the admission outcome model are pivotal. 40 Harvard’s expert, Professor David Card, and SFFA’s economist, Professor Peter Arcidiacono, are both highly respected economists. Professor Card is an economics professor at the University of California at Berkeley, where he teaches undergraduate and graduate level economic courses. He has published numerous articles and books and is a winner of the John Bates Clark Prize. [Oct. 30 Tr. 73:7–76:2; DX133]. Professor Arcidiacono is a professor of labor economics at Duke University. He teaches undergraduate and graduate-level economic courses and has published numerous peer-reviewed articles. His research is focused on labor economics, and more narrowly, higher education. [Oct. 25 Tr. 14:7–17:14]. 50 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 51 of 130 In sum, as discussed more fully below, Professor Arcidiacono excludes ALDCs from his model despite the fact that they make up about 30% of each admitted class, analyzes the data in aggregate rather than independently modeling each admissions cycle, excludes certain variables that he contends are unreliable and have unexpected effects on the model, selectively interacts certain variables, omits the personal rating based on his finding that it is influenced by race, and then, based on that data and approach, concludes that Asian Americans are discriminated against in the admissions process. See [Oct. 26 Tr. 62:9–63:25; Oct. 30 Tr. 145:15–148:11; DX695; PD38 at 45]. Professor Card creates an independent model for each admissions cycle, includes the personal rating because he concludes that it does not reflect race and, in any event, includes information that is important to the admissions process such that omitting it skews the outcome, includes the other variables that Professor Arcidiacono omits, and does not interact variables. Using this approach, he comes out with a very slight, and not statistically significant, negative coefficient for Asian American identity and concludes, based on that data and approach, that Asian Americans are not discriminated against in Harvard’s admissions process. See [Oct. 31 Tr. 172:19–173:15; DX695; DD10 at 34–35]. The statistics and econometric models used by Professors Arcidiacono and Card were generated using primarily data produced by Harvard in this litigation. Consistent with this Court’s orders, Harvard provided applicant-by-applicant admissions data for more than 150,000 domestic applicants to Harvard’s classes of 2014 through 2019, 41 as well as aggregate 41 Because this lawsuit concerns only allegations of discrimination against United States citizens or permanent residents, foreign applicants were removed from the data set. Further, transfer applicants and those who submitted incomplete applications or for whom Harvard’s database was for some other reason incomplete were also removed. [Oct. 25 Tr. 25:3–26:24; Nov. 1 Tr. 99:12–100:10]. Statistics on “applicants” referred to by these findings of fact are therefore based on data for the approximately 150,000 domestic applicants to Harvard’s 2014 to 2019 classes for whom Harvard’s database contained a single, complete record. See [PD38 at 1]. 51 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 52 of 130 information for the classes of 2000 through 2017, and a sample of actual application files and summary sheets from the classes of 2018 and 2019. [Oct. 25 Tr. 23:8–26:13; PD38 at 1]. For each applicant to the classes of 2014 through 2019, Harvard’s database includes 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. [Oct. 25 Tr. 23:16–24:8; Oct. 26 Tr. 73:22–74:2]. On behalf of SFFA, Professor Arcidiacono supplemented this data by merging it with College Board data on applicants’ high schools and neighborhoods. [Oct. 25 Tr. 24:9–12]. The parties dispute whether ALDC applicants should be included when computing admissions statistics and modeling Harvard’s admissions process. ALDC applicants are admitted at higher rates than the applicant pool more broadly. SFFA argues that because ALDC applicants are granted significant tips that are not available to most applicants, they are not typical. [Id. at 27:2–25, 29:4–30:7]. SFFA therefore presented numerous statistics based on non-ALDC applicants which it identifies as the “Baseline Dataset.” [Id. at Tr. 27:2–25]. The Baseline Dataset excludes approximately 7,400 ALDCs, leaving a total of 142,728 applicants in the dataset. [Oct. 25 Tr. 30:8–31:3; PD38 at 1–2]. SFFA has also presented data based on a subgroup of dataset applicants that include legacies, dean’s and director’s list applicants, and children of faculty and staff (“LDCs”), but not recruited athletes, which SFFA refers to as the “Expanded Dataset.” [Oct. 25 Tr. 40:17–41:7]. Although ALDCs represent only a small portion of applicants and are admitted or rejected through the same admissions process that applies to other applicants, they account for approximately 30% of Harvard’s admitted class. [Oct. 30 Tr. 153:6–154:8, DX706; DD10 at 38, 40]. For reasons discussed more fully infra at Section V.F, the Court agrees with Professor Card 52 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 53 of 130 that including ALDCs in the statistics and econometric models leads to more probative evidence of the alleged discrimination or lack thereof. Nevertheless, the Court has referenced numerous statistics based on data that excludes some or all ALDCs because SFFA used those metrics at trial. In addition to statistics based on Harvard’s admissions database, Harvard presented statistics on the racial make-up of its admitted classes from 1980 to 2019, [Oct. 31 Tr. 119:23– 124:6; DX711; DX713; DD10 at 100–04], and SFFA used statistics based on an analysis of 480 sample application files, two-thirds of which were selected by SFFA and one-third by Harvard, [Oct. 25 Tr. 24:21–24]. Both Harvard and SFFA also relied on statistics and models that were prepared by OIR before this lawsuit was filed. See, e.g., [PX9; PX12; PX21]. B. Admission Rates and Ratings by Race Asian Americans were admitted to Harvard at slightly lower rates than white applicants in the years leading up to this lawsuit, with between 5% and 6% of Asian American and between 7% and 8% of white applicants being admitted to the classes of 2014 through 2017. See [PX319 at 15–16]; see also [PD38 at 20]. 42 The admissions rates differ more significantly among certain subgroups, but the admissions rates for Asian American ALDCs are generally similar to or higher than those for white ALDCs. 88.6% of Asian American recruited athletes, 48.1% of Asian American children of faculty or staff, and 47.7% of Asian Americans on the dean’s or director’s interest lists are admitted, compared to 88.1%, 47.9%, and 43.1% of white applicants in those groups, respectively. [PX634]. Asian American legacies are admitted at a rate of 42 Overall admission rates for Asian American applicants are lowered slightly because they are underrepresented among ALDCs, who are admitted at a rate of 43.6% or nearly eight times the 5.5% admissions rate for non-ALDC applicants. [Oct. 30 Tr. 154:17–155:19; DX679; DD10 at 39]. 53 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 54 of 130 35.2%, as are white legacies. [Oct. 25 Tr. 121:7–122:4; PX634]. SFFA’s economic expert, Professor Arcidiacono, acknowledges that Asian American ALDCs were likely not discriminated against. 43 [Oct. 25 Tr. 122:16–123:17, 126:1–8]. Non-ALDC Asian American applicants have admission rates that are similar to white applicants, although the admission rates relative to whites varies by year from between 0.2 percentage points lower to 0.9 percentage points higher. [Id. at 68:2–70:2; PD38 at 20]. 44 With the exception of 2019 where the admission rates favored Asian American applicants, the differences in admission rates for non-ALDC white and Asian American applicants was not statistically significant. See [PD38 at 20]. The gist of SFFA’s argument, however, is not that Asian Americans were excluded altogether, but rather that the non-ALDC Asian American applicants were stronger than the non-ALDC white applicants and should have been admitted at a higher rate. 45 43 Although its expert agrees that Asian American ALDCs were not discriminated against, SFFA continues to argue that they were, but that the strength of their applications overcame the bias. The Court ultimately finds that excluding ALDCs distorts the analysis. 44 The highest annual admissions rate for Asian American applicants relative to white applicants, and the only year for which the admission rates for Asian American and white applicants differed to a statistically significant degree, was the class of 2019, which was selected after the allegations of discrimination that led to this lawsuit emerged. [Oct. 25 Tr. 68:2–22; PD38 at 20]. 45 As reflected by the data, Harvard does not systematically exclude Asian Americans, nor does SFFA claim that it does. As of 2016, the United States population was approximately 60% white and 5.9% Asian. U.S. Census Bureau, Quick Facts, Census.gov, https://www.census.gov/ quickfacts/fact/table/US/RHI225218. Among applicants to Harvard’s class of 2019, 21.2% were Asian American and 57.6% were white. [DX713]. Among those domestic applicants who Harvard admitted, 40% of the class identified as white and 24% identified as Asian American. It is entirely possible, and not without historical precedent, that an admissions process could discriminate against Asian Americans (or Jews) despite their over-representation in a class as compared to the general population. The Court nonetheless includes these numbers to give some context to the overall admissions data. 54 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 55 of 130 Asian Americans would likely be admitted at a higher rate than white applicants if admissions decisions were made based solely on the academic and extracurricular ratings. Among Expanded Dataset applicants, more than 60% of Asian American applicants received academic ratings of 1 or 2, compared to 46% of white applicants, 9% of African American applicants, and 17% of Hispanic applicants. [Oct. 25 Tr. 49:17–50:5; PX623]. Overall, strong academic applicants are particularly abundant, with a higher percentage of applicants (42%) scoring a 1 or 2 on the academic rating as compared to the percent that score a 1 or 2 on any other rating. [DD10 at 4]. 46 Asian American applicants’ stronger academic ratings broadly align with their stronger performance across a range of qualitative indicators of academic strength. [Oct. 25 Tr. 41:18–46:9; PD38 at 4–7]. Asian American applicants also average relatively high extracurricular ratings. More than 28% of Expanded Dataset Asian American applicants receive an extracurricular rating of 1 or 2, compared to 25% of white applicants, 16% of African American applicants, and 17% of Hispanic applicants. [Oct. 25 Tr. 52:12–22; PX623]. Although Harvard admissions officers do not believe that Asian American applicants, as a group, have worse personal qualities than other applicants and Harvard alumni interviewers assign personal ratings of 1 or 2 to Expanded Dataset Asian American and white applicants with a similar frequency, [Oct. 23 Tr. 204:1–9; Oct. 24 Tr. 138:11–16; Oct. 25 Tr. 55:7–12], Harvard admissions officers assign Asian American applicants personal ratings that are, on average, slightly weaker than those assigned to applicants from other racial groups, [PX623]. Among Expanded Dataset applicants, 22.6% of white applicants receive a personal rating of 1 or 2, compared to 18% of Asian Americans, 19.4% of African Americans, and 19.1% of Hispanics. 46 24% of applicants receive an extracurricular rating of 1 or 2, 21% of applicants receive a personal rating of 1 or 2, and 10% of applicants receive an athletic rating of 1 or 2. [Oct. 30 Tr. 86:25–88:2; DD10 at 4]. 55 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 56 of 130 [Id.]. The statistics are similar for Baseline Dataset applicants, with 17.6% of Asian Americans receiving a personal rating of 1 or 2, compared to 18.7% of Hispanics, 19% of African Americans, and 21.3% of whites. [Oct. 25 Tr. 55:13–22; PX621]. At least a partial cause of the disparity in personal ratings between Asian American and white applicants appears to be teacher and guidance counselor recommendations, with white applicants tending to score slightly stronger than Asian Americans on the school support ratings. [PX621; PX623; PD38 at 4–5, 8–10]. Among Expanded Dataset applicants, 31.9% of white applicants received a “teacher 1” rating (the rating for the first of two teacher recommendations submitted) of 1 or 2 compared to 31.6% of Asian American applicants. [PX 623]. For the “teacher 2” rating (the rating for the second teacher recommendation), 33.6% of white applicants received a rating of 1 or 2 compared to 32.3% of Asian American applicants. [Id.]. In the Expanded Dataset, 27.4% of white applicants and 26.4% of Asian American applicants receive a guidance counselor rating of 1 or 2. [Id.]. Although these differences may appear slight, they are significant in that the stronger high school academic and extracurricular performance of Asian American applicants on average would lead one to expect that those applicants would receive stronger teacher and guidance counselor recommendations than white applicants. On average, Asian American applicants are also assigned lower athletic ratings, particularly compared to white applicants, who average especially strong athletic ratings. See [PX621; PX623; DX692 at 2]. Among non-recruited athlete applicants, only 5% of Asian Americans received an athletic rating of 2, compared to 14% of whites, 7% of African Americans, and 8% of Hispanics. [PX623]. When recruited athletes are included in the calculation, the disparity between white and Asian American applicants receiving strong athletic 56 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 57 of 130 ratings increases, with white applicants receiving athletic ratings of 1 or 2 at roughly three times the rate of Asian American applicants. [Oct. 30 Tr. 96:25–97:19; DX692 at 2; DD10 at 10]. C. Descriptive Statistics In addition to the regression analyses used in this case, Professors Card and Arcidiacono also offered descriptive statistics that support their respective arguments on the question of discrimination. In constructing these statistics, both experts used the same dataset consisting of applicants to the classes of 2014 through 2019 (except that Professor Arcidiacono prefers to remove ALDCs). Professor Card uses the dataset to compare admission rates by racial group for applicants who scored 1s and 2s across similar numbers of profile ratings. He, on behalf of Harvard, uses this multidimensionality analysis to argue that the statistical evidence does not support a conclusion that Harvard discriminates against Asian Americans relative to whites. Meanwhile, Professor Arcidiacono uses an academic decile analysis in which he divides applicants into deciles based on applicant academic index score and then shows that Asian Americans in the top academic deciles are receiving strong personal and overall ratings at lower rates than applicants from other racial groups with similar academic qualifications. He, on behalf of SFFA, argues that the lower average overall and personal ratings for Asian American applicants who have similar levels of academic strength to non-Asian American applicants suggest that Harvard is engaged in a discriminatory admissions process. 1. Professor Card’s Multidimensionality Analysis Professor Card’s statistical analysis shows that the students most likely to be admitted to Harvard are those that do well across the profile and school support ratings, rather than merely excelling on just one rating. In coming to this conclusion, Professor Card analyzed the relationship between race and applicant strength across multiple profile ratings, which he terms an analysis of “multidimensional accomplishments.” [Oct. 30 Tr. 89:3]. Only 7,000 applicants 57 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 58 of 130 per year, or roughly 27%, receive a rating of 1 or 2 in at least two profile ratings, and only 7% of applicants receive ratings of 1 or 2 in three or all four profile ratings. [Id. at 89:19–90:17; DX672; DD10 at 5]. The 7% of applicants who score highly in three or four of the four profile ratings are seemingly the most multidimensional under Harvard’s scoring system; 70% of those applicants are admitted and make up 46% of all admitted applicants. [Oct. 30 Tr. 93:15–94:12; DX672; DD10 at 8]. The 20% of applicants who receive two profile ratings of 1 or 2 account for 38% of admitted students. [Oct. 30 Tr. 93:15–94:12; DX672; DD10 at 8]. Meanwhile, applicants with one or no ratings of 1 or 2 account for 73% of applicants but only 15% of admitted students. [DX672; DD10 at 8]. White applicants are slightly more likely than Asian American students to receive three profile ratings of 1 or 2, with approximately 900 or 9% of all white applicants receiving three such scores relative to 500 or 8% of all Asian American applicants. [Oct. 30 Tr. 95:18–96:10; DX692 at 2; DD10 at 9]. Professor Card has also offered support for his conclusion that white applicants are disproportionately strong in non-academic traits by removing all academic inputs from his model of admissions probability to rank applicants to Harvard. See [Oct. 31 Tr. 69:20–71:5; DD10 at 77]. By doing so, he creates a “non-academic index,” and his analysis shows that white students do disproportionately well in this metric, with 12% of white applicants ranking in the top decile compared to only 7.8% of Asian American applicants. See [Oct. 31 Tr. 70:17–19; DD10 at 77]. Professor Card’s multidimensionality analysis thus suggests that a partial cause of the racerelated disparities in admission rates, when controlling for academic performance, is that Asian American applicants’ disproportionate strength in academics comes at the expense of other skills and traits that Harvard values. See [DX692 at 2–4]. 58 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 59 of 130 The Court notes, however, that the profile ratings are not equally distributed in terms of the number of 1s, 2s, 3s, or 4s assigned, nor are they equally correlated with an applicant’s chances of admission. For example, being a recruited athlete (and therefore receiving an athletic rating of 1) vastly improves an applicant’s odds of admission, with 86% of recruited athletes typically admitted and Asian Americans especially underrepresented in that group. [Oct. 25 Tr. 31:11–23; PD38 at 2]. Although Harvard highly values applicants who will contribute to its varsity sports, it also admits a significant number of applicants who do not participate in high school athletics, and who therefore receive an athletic rating of 4 or lower. [Oct. 25 Tr. 28:21– 29:8, 31:11–23; PD38 at 10]. Academic, extracurricular, or personal ratings of 4 or lower are relatively rare and more likely to result in rejection than an athletic rating of 4 or lower. [Oct. 25 Tr. 52:6–54:12; PD38 at 8–10]. 39% of admitted non-ALDC applicants are scored as athletic 4s or lower, while less than 1% of admitted Baseline Dataset applicants are scored as academic, extracurricular, or personal 4s. [Oct. 25 Tr. 52:23–53:13; PD38 at 10]. Further, personal ratings of 1 are exceptionally rare and are awarded to fewer than 10 applicants in a typical year, whereas athletic, extracurricular, and academic ratings of 1 are more common, though they are still each awarded to less than 1% of applicants. See [PX623; PD38 at 2]. Although the profile ratings are not of equal importance, are not assigned on a set curve, and do not have any assigned mechanical weight, receiving multiple ratings of 1 or 2 is strongly correlated with admission. [Oct. 30 Tr. 88:12–89:3, 90:18–92:12; DX672; DD10 at 6, 8]. Because the number of 1s and 2s awarded in each of the four profile ratings every year vastly exceeds the number of students Harvard can admit, Harvard tends to admit applicants with multiple profile ratings of 1 or 2 who are also significantly distinguished in some other way— which, as discussed supra at Part III.B.3, may include accomplishments or characteristics that are 59 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 60 of 130 remarkable even when measured against a very accomplished applicant pool or that are likely to be underrepresented in Harvard’s class. To summarize, Professor Card uses his multidimensionality analysis to show that the Harvard admissions process favors applicants who score well across the profile and school support ratings and to counter the argument that Harvard’s admissions process is biased based on a comparison of admission rates for students who are similarly-situated academically. Professor Card is correct that an analysis predicated on an applicant’s academic profile ignores statistical disparities between racial groups across other dimensions that favor non-Asian American applicants. Most notably, white applicants are significantly more likely to have made strong high school contributions to athletics, and this disparity counteracts the effect that Asian American applicants’ relative academic and extracurricular strength would otherwise have on their admission rate. Professor Card’s analysis shows that strength across multiple dimensions is highly correlated with admission to Harvard and results in fewer admitted Asian American applicants. That being said, because Professor Card’s multidimensional analysis gives equal weight to each profile rating, it overvalues the athletic rating which favors white applicants, despite the fact that it is seemingly less important than the academic, personal, and extracurricular ratings for obtaining admission to Harvard, at least for applicants who are not recruited athletes. See [DX692 at 2]. Further, because the multidimensionality analysis uses all the profile ratings, any bias in the ratings, including in the personal rating, is baked into his analysis. 2. Professor Arcidiacono’s Academic Index Decile Analysis In contrast, for his descriptive statistics analysis, Professor Arcidiacono compared applicants by analyzing the relationship between race and various ratings, including school support, academic, extracurricular, personal, and overall ratings for applicants who are 60 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 61 of 130 academically similarly-situated—that is who fall into the same academic index deciles. For this analysis, he splits Baseline Dataset applicants into ten equally sized groups based on their academic index, which reflects the strength of an applicant’s standardized test scores and high school grades, with “Decile 10” containing the 10% of applicants to Harvard who had the strongest academic index scores and “Decile 1” containing the applicants with the weakest scores. See [PD38 at 6]. The deciles reflect only numerical academic metrics in contrast to the academic ratings assigned by Harvard, which incorporate an assessment of academic potential and other non-numerical factors. [Oct. 22 Tr. 137:12–24]. Professor Arcidiacono’s deciles show that Asian American applicants are disproportionately represented in the top academic deciles. See [Oct. 25 Tr. 44:12–48:8; PX624; PX626; PD38 at 6–8]. More than a third of Baseline Dataset Asian American applicants fall in the strongest two academic index deciles, while African American and Hispanic applicants are underrepresented among applicants with the highest academic indexes. [Oct. 25 Tr. 47:22–48:5; PD38 at 8]. Professor Arcidiacono’s academic decile analysis shows a racial disparity in applicants’ personal and overall ratings that appears to favor white applicants based on a comparison of applicants within the same academic decile. See [PD38 at 16, 18–19]. For example, among Baseline Dataset applicants, 22.2% of Asian Americans in the top decile of applicants by academic index (i.e. those with the strongest high school GPA and standardized test scores) receive personal ratings of 1 or 2, compared to 29.6% of whites, 34.2% of Hispanics, and 47% of African Americans; similar variances by race are also present for students in the second and third strongest academic deciles. [Id. at 19]. Similarly, only 12.9% of Baseline Dataset Asian Americans in the top academic index decile receive an overall rating of 1 or 2, compared to 15.6% of whites, 27.4% of Hispanics, and 47% of African Americans. [Id.]. 61 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 62 of 130 Professor Arcidiacono’s decile analysis also shows that the disparities between Asian American and white applicants’ school support ratings are magnified when comparing applicants within the same academic index deciles. Among non-recruited athletes, white applicants are only approximately 1 percentage point more likely to receive teacher or guidance counselor ratings of 1 or 2 than Asian American applicants. [PX623]. White applicants in the top academic deciles, however, receive school support ratings of 1 or 2 at a rate that is 4 to 6 percentage points higher than Asian Americans in the same academic deciles. [Oct. 26 Tr. 37:25–40:17]. In sum, Professor Arcidiacono bases his decile analysis on the academic index, which only accounts for test scores and grades—criteria in which Asian American applicants are, on average, especially strong. He argues that the personal rating is compromised, that the athletic rating is not that important, and that Asian American applicants do well on the limited measures that remain and should therefore be admitted at a higher rate than they are. This approach 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. D. Overview of Logistic Regression Models Professors Card and Arcidiacono both believe that the descriptive statistics discussed above help to provide context, but also agree that logistic regressions are the most useful econometric tool in evaluating the probable effect of race and other traits on Harvard’s admissions process. See [Oct. 25 Tr. 79:11–83:24; Oct. 30 Tr. 101:5–17]. Their respective logistic regression models seek to isolate the effects of race through models that include, and thereby control for, other variables that affect the modeled outcome. [Oct. 25 Tr. 215:12–19; Oct. 30 Tr. 176:18–179:3]. 62 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 63 of 130 The Court notes at the outset that although logistic regression models are seemingly the best available econometric tool, they cannot capture all of the factors that Harvard considers and can therefore account for only part of the variation in admissions decisions, or other modeled outcomes. See [Oct. 25 Tr. 80:13–24; Oct. 30 Tr. 114:10–23]. Further, no model is perfect, and models can be affected by biases that are inherent in the control variables that they use. See [Oct. 25 Tr. 91:17–92:11, 215:16–19]. To limit the impact of variables affected by bias, variables that are themselves impacted by the independent variable of primary interest, which is race, should generally be excluded from regression models. Including such variables dilutes the implied effect of race by allowing that effect to be partially captured by the race-influenced variable itself. See [id. at 215:16–19; Nov. 1 Tr. 77:22–78:4]. Excluding variables for this reason may, however, make a model less accurate because it also results in the removal of information relevant to the modeled outcomes. Here, although Professors Arcidiacono and Card both endorse the use of regression models, they disagree on whether the personal rating should be included as a control variable. Professor Arcidiacono contends that personal ratings are themselves affected by race and that they should therefore not be used in the admissions model. [Oct. 25 Tr. 99:11–18]. Professor Card argues that the personal rating variable should be included, and thereby implicitly contends that race correlates with personal qualities that affect personal ratings, but that race does not itself affect the personal ratings assigned by admissions officers, or at least that any causal effect of race on the personal rating is insignificant relative to the value of the variable in controlling for a race-correlated, but not directly race-caused, relationship. [Nov. 1 Tr. 79:9–14]. 47 Further, 47 Directly race-caused means a cause internal to the Harvard admissions process, as distinct from the potential for an effect of race on inputs to that process. 63 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 64 of 130 the personal rating captures other relevant characteristics unrelated to race that will not be taken into account at all by the modeling if the personal rating is excluded, such as the extent to which an applicant demonstrates character, leadership ability, self-confidence, grit, or other distinctive qualities that might benefit the Harvard community. Logistic regressions result in two metrics that are relevant here: “coefficient” and “marginal effect.” Coefficients indicate how much weight the model suggests each variable has in determining the modeled variable (here, admissions outcome or an assigned rating), holding the other included variables constant. [Oct. 25 Tr. 76:22–78:8]. To generate a coefficient for a discrete variable such as race (e.g. where an applicant is white, Asian American, African American, or Hispanic), a model omits one of those characteristics to create a baseline group (e.g. white applicants) and the coefficients that the model generates for the included groups (e.g. Asian American, African American, and Hispanic) then indicate the implied effect of each of those characteristics on the dependent variable (e.g. admissions outcome) relative to the baseline group, holding constant the control variables that are included in the model (e.g. academic rating, disadvantaged status, parental occupation, etc.). [Id. at 77:24–78:8]. In the experts’ models, a positive coefficient is associated with a higher probability of admission or a stronger rating, and a negative coefficient is associated with a lower probability of admission or a weaker rating. [Id. at 77:3–78:23]. The Court has and will continue to note when race appears “statistically significant” in an analysis, which indicates that the coefficient for some racial group is of a magnitude that would occur infrequently due to random variation if race and the modeled variable were not related when controlling for the other variables included in the model. It is critical to understand that a statistically significant variable in an econometric model is not proof 64 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 65 of 130 of a causal relationship. A statistically significant coefficient may be the result of random variation, omitted variables, or other flaws in the model. A marginal effect is a measurement of the change in the projected outcome of the model (e.g. odds of admission to Harvard) that is associated with changing a given variable, while holding other variables constant. The magnitude of the change in probability will depend on the other variables. For example, a model might not suggest a large effect on an applicant’s probability of being admitted based on being African American, as opposed to being white, for a student who is academically unprepared (i.e. race won’t make a difference for an unprepared student), but might imply a significantly increased probability of admission associated with being African American rather than white for an applicant who is well-prepared. See [id. at 78:24–80:6; PD38 at 25]. An “average marginal effect” is the average of the marginal effect associated with the variable of interest for that group. For example, one could calculate the average marginal effect of African American racial identity relative to white identity on the odds of admission or of achieving a given rating by averaging the probability changes attributable to the coefficient for African American identity in a relevant model. See [Oct. 25 Tr. 21:18–22:17, 80:8–12, 96:19–97:24; PD38 at 24–25, 31]. Again, it must be understood that the average marginal effect reflects an average statistical relationship between a variable of interest (such as race) and a modeled variable (such as admissions outcome), and that relationship may or may not be causal in nature. Professors Card and Arcidiacono each prepared models of the admissions process in which the dependent variable is the admissions decision (admitted or rejected). [Oct. 25 Tr. 216:22–217:7; Oct. 30 Tr. 101:15–17]. Their admissions models are broadly similar and predict the probability of admission for domestic non-transfer applicants by accounting for a wide range 65 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 66 of 130 of observable variables including gender, disadvantaged status, first generation college applicant status, fee waiver, whether the applicant applied for financial aid, academic index, intended major, secondary school type, indicators of parental education, whether parents attended an Ivy League school, whether parents are alive, geographic indicators, and standardized test results. See [Oct. 30 Tr. 143:16–25]; see also [PD38 at 26]. There are, however, several critical differences in the structure and control variables utilized by Professor Card’s and Professor Arcidiacono’s respective models. As a result of these differences, Professors Card’s model returns a coefficient for Asian American identity that is negative but not statistically significant, meaning that the model does not strongly suggest that Asian American as opposed to white racial identity affects an applicant’s chances of admission, whereas Professor Arcidiacono’s model returns a negative coefficient for Asian American identity that is statistically significant, meaning that his model suggests that Asian American identity is associated with a lower chance of admission, when controlling for the other variables he includes. [Oct. 25 Tr. 115:1–11; Oct. 30 Tr. 129:9–16, 132:21–134:11]. The modeling differences that result in these disparate conclusions are discussed infra at Section V.F. Professor Arcidiacono also prepared a series of ordered logit estimates that SFFA contends show how well the ratings assigned for the academic, extracurricular, personal, overall, and school support ratings can be predicted. [Oct. 25 Tr. 90:10–91:23, 150:2–6; Oct. 30 Tr. 101:15–17; Oct. 31 Tr. 188:22–189:1; Nov. 1 Tr. 76:3–11, 79:15–19; PD38 at 28]. These models are similar to Professor Arcidiacono’s model of admissions decisions, except that they are intended to be probative of the effect of race on the ratings assigned by admissions officers rather than the applicants’ probability of admission. [Oct. 25 Tr. 90:16–91:10]. Professor Arcidiacono’s preferred ordered logit models control for application year, application docket, 66 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 67 of 130 academic index, SAT scores, SAT II scores, high school GPA, extremely low-grade applicants, parental education level, including whether a parent attended an Ivy League school, whether the applicant’s parents are alive, expected college major, gender, high school type, neighborhood, disadvantaged status, receipt of an application fee waiver, first generation college applicant status, whether the applicant applied for financial aid, profile ratings other than the dependent variable, teacher recommendation ratings, guidance counselor rating, alumni ratings, and certain interactions among those variables. See [id. at 82:13–85:3; PD38]. To the extent that Professor Arcidiacono’s models imply that Asian American identity is associated with the ratings assigned by admissions officers, his models suggest that the magnitude and direction of the relationship (bonus or penalty) varies by rating and depends on whether an Asian American applicant is male or female and whether or not they are economically disadvantaged. See [PD38 at 28–35]. E. Regression Models of School Support, Profile, and Overall Ratings Although the experts’ models of admissions outcomes are most probative of whether Harvard has engaged in discrimination against Asian Americans relative to white applicants, there are also related statistical relationships between race and the profile and school support ratings. Because those ratings serve as inputs for the proposed admissions outcome models, the Court will briefly address the extent to which race might appear to impact the ratings assigned by admissions officers before turning to the admissions outcome models themselves. 1. Relationship Between Race and School Support Ratings As discussed supra at Section V.C, Asian American applicants have lower average school support ratings than white applicants. There are several conceivable explanations for the disparity including actual differences in non-academic strengths, a correlation between the quality of the guidance counselor or teacher recommenders and the racial makeup of high schools, biased teachers and guidance counselors, or biased Harvard admissions officers. 67 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 68 of 130 Considering the testimony of Harvard’s admissions officers and the admissions process itself, the Court views Harvard admissions officer bias as an unlikely explanation for the disparity in school support ratings and concludes that race-related variance in the school support ratings result from some combination of the other potential causes, all of which are beyond Harvard’s control. Further, when considering regression analyses, because school support ratings can be included in admissions outcome models, any racial effect that impacts admissions decisions through the school support ratings can be controlled for. Importantly, however, although the school support ratings themselves provide only an overall numeric evaluation of recommendations, the school support materials are in fact more nuanced and the substance of them informs perceptions about applicants across numerous dimensions. [Oct. 31 Tr. 36:16–37:16]. Considering the stereotypes and biases that favor and disfavor Asian American applicants in different evaluation dimensions, the impact of race on the school support ratings could be understood to suggest that the overall numeric score masks more subjective disparities in how applicants from different racial groups are presented by their recommenders. See [id.]. Therefore, to the extent Asian Americans are presented by guidance counselors and high school teachers as weaker in personal characteristics that Harvard values and those presentations inform the personal rating, omission of the personal rating results in an omitted variable bias that cannot be fully captured by including a school support rating control variable. 2. Relationship Between Race and Personal Ratings Professor Arcidiacono’s preferred model suggests that Asian American identity reduced a Baseline Dataset applicant’s probability of receiving a personal rating of 2 or higher. The model implies that when holding constant nearly all of the available observable variables, Asian American identity is associated with a lower probability of being assigned a strong personal 68 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 69 of 130 rating by an admission officer. More precisely, his model suggests that an average Baseline Dataset Asian American applicant has a 17.8% probability of receiving a 2 or higher on the personal rating, which is lower than the 21.6% chance that the model suggests the applicant would have in the absence of any racial preference. [Oct. 25 Tr. 96:24–97:24; PD38 at 31]. Harvard did not offer a competing regression model to show that no statistically significant relationship between Asian American identity and the personal rating exists, and the Court therefore concludes that the data demonstrates a statistically significant and negative relationship between Asian American identity and the personal rating assigned by Harvard admissions officers, holding constant any reasonable set of observable characteristics. The Court finds, however, that Professor Arcidiacono’s preferred model likely overstates the direct effect of Asian American identity on the personal rating. First, as discussed supra at Section III.B.4, Harvard’s witnesses credibly testified that they did not use race in assigning personal ratings (or any of the profile ratings) and did not observe any improper discrimination in the admissions process. [Oct. 18 Tr. 49:17–19; Oct. 19 Tr. 48:24–49:19, 253:4–17; Oct. 23 Tr. 50:24–51:4, 219:21–24; Oct. 24 Tr. 122:5–8; Nov. 1 Tr. 246:18–247:4, 253:18–25]. The uniformity of these observations is persuasive given the collective manner in which admissions decisions are made, with all members of the Admissions Committee participating in all decisions and having real-time visibility into the process for each applicant. Any causal relationship between Asian American identity and the personal rating must therefore have been sufficiently subtle to go unnoticed by numerous considerate, diligent, and intelligent admissions officers who were immersed in the admissions process. Second, Professor Arcidiacono’s model explains only a portion of the variation in personal ratings and likely suffers from considerable omitted variable bias. The model does not 69 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 70 of 130 include variables for several factors that influence personal ratings and may correlate with race, such as the extent to which applicants’ essays and personal statements demonstrated their abilities to overcome obstacles or personal achievements that might reasonably be perceived as an indication of leadership ability or other personal strengths. [Oct. 31 Tr. 35:15–36:9]. 48 Third, as discussed supra at Section V.C, E, teacher and guidance counselor recommendations seemingly presented Asian Americans as having less favorable personal characteristics than similarly situated non-Asian American applicants, and the school support ratings do not fully reflect more subtle racial disparities. As the experts’ analyses demonstrate, some race-correlated variation in teacher and guidance counselor recommendations is likely a cause of at least part of the disparity in the personal ratings. See supra at Sections V.C, E. Professor Card’s analysis shows that the school support ratings for Asian American applicants are generally weaker than the ratings for white students when comparing white and Asian American students who receive the same academic rating. [DX692 at 4]; see [DD10 at 68]. For example, approximately 43% of white students who receive an academic score of 2 have school support ratings (from their two teacher and one guidance counselor recommendations) that sum 48 Speculating on how unobserved variables may be influencing the model’s implied effect of race on the personal ratings is fraught with difficulty. Although the Court has not received statistical evidence on the effect of race on specific high school achievements, it is likely that some high school achievements are themselves effected by racial biases. One might question the effect, positive or negative, of being Asian American on the probability of being selected to a leadership position such as class president, captain of a sports, math, or debate team, or the likelihood of being identified as an outspoken advocate, a natural leader, or an intellectual superstar. Professor Arcidiacono’s models account for some of these considerations, to some degree, through inclusion of the school support ratings, but much of the variation in applicants’ qualities cannot easily be boiled down to econometrically digestible variables. [Oct. 31 Tr. 35:15–36:9]. It is possible that Asian American applicants to Harvard are being disadvantaged by biases in their high schools that affect their college applications. Admissions officers have no easy mechanism to measure or correct for these biases, except to carefully review individual applicants in a holistic way and to recognize and consider applicants’ accounts of how their racial identities have shaped their pre-college experiences. 70 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 71 of 130 to 7 or less (indicating very strong recommendations), while only about 37% of Asian American applicants with an academic score of 2 receive similarly strong school support ratings. [Oct. 31 Tr. 55:11–56:2]. Because teacher and guidance counselor recommendation letters are among the most significant inputs for the personal rating, the apparent race-related or race-correlated difference in the strength of guidance counselor and teacher recommendations is significant. See [id. at 54:6–56:2; DD10 at 67–68]. The Court reiterates that to the extent that disparities in the personal ratings are explained by teacher and guidance counselor recommendation letters, Harvard’s admissions officers are not responsible for any race-related or race-correlated impact that those letters may have. Additionally, correlation between race and the personal and school support ratings does not clearly demonstrate a causal relationship, given the correlation between race and numerous factors that likely influence teacher and guidance counselor recommendations and admissions officers’ evaluation of personal and overall strength. For example, a privileged student and a disadvantaged student with the same academic performance may well not receive similar teacher and guidance counselor recommendations. Similarly, 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, but if working outside of school correlates to race and informs teacher, guidance counselor, and admissions officers’ evaluation of applicants, the school support and personal ratings may correlate with race, although race might not be the cause of the differential. In other words, race-correlated disparities in personal ratings for applicants who have similar academic qualifications may reflect underlying differences in the backgrounds and experiences of applicants that happen to correlate with race but are not racially motivated. That being said, it is not clear that these sorts of considerations adequately explain the difference 71 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 72 of 130 in personal ratings between white and Asian American applicants in Professor Arcidiacono’s decile analysis or the similar analysis Professor Card has offered. Overall, the disparity between white and Asian American applicants’ personal ratings has not been fully and satisfactorily explained. Because some of the disparity in personal ratings is due to teacher and guidance counselor recommendations, the issue becomes whether the remaining disparity reflects discrimination. The disparity in personal ratings between Asian American and other minority groups is considerably larger than between Asian American and white applicants and suggests that at least some admissions officers might have subconsciously provided tips in the personal rating, particularly to African American and Hispanic applicants, to create an alignment between the profile ratings and the race-conscious overall ratings that they were assigning. See [PD38 at 33]. It is also possible, although unsupported by any direct evidence before the Court, that part of the statistical disparity resulted from admissions officers’ implicit biases that disadvantaged Asian American applicants in the personal rating relative to white applicants, but advantaged Asian Americans over whites in the academic rating. Further, the Court cannot accurately estimate what portion of the difference in the personal ratings resulted from the strength of the personal qualities that Harvard seeks to measure or from differences in how Asian American applicants are presented to Harvard by high schools relative to other applicants, as opposed to being the effect of implicit biases. Taking account of all the available evidence, it is possible that implicit biases had a slight negative effect on average Asian American personal ratings, but the Court concludes 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 (e.g. recommendations or high school 72 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 73 of 130 accomplishments) or underlying differences in the attributes that may have resulted in stronger personal ratings. 3. Regression Models of the Academic, Extracurricular, and Overall Ratings Unlike the personal ratings, the experts agree that the academic and extracurricular variables should be included in the admissions outcome model and that the overall rating should not be included because Harvard acknowledges that it is directly affected by racial identity. See [PD38 at 26; DD10 at 46–47]. Nevertheless, because the profile ratings may all be impacted by race in a very marginal manner, the Court will briefly discuss the econometric models of these variables. Professor Arcidiacono’s logistic regression models for the academic, extracurricular, and overall ratings suggest a non-uniform effect of Asian American identity on those ratings. [Oct. 25 Tr. 91:11–92:20, 109:23–110:13; PD38 at 28–33]. The academic and extracurricular ratings models return positive coefficients for Asian American identity, while the overall rating model returns a negative coefficient for Asian Americans (with the exception of disadvantaged Asian American females). See [Oct. 25 Tr. 92:24–94:10, 107:8–13, PD38 at 29, 32–33]. A comparison between the strength of Asian American applicants on the observable characteristics included in Professor Arcidiacono’s academic and extracurricular rating models and the coefficients for Asian American groups suggests that Asian Americans have traits, other than their racial identity, that make them likely to score well in academic and extracurricular ratings. [Oct. 25 Tr. 107:8–110:17; PD38 at 32–33]. This implies that the positive coefficients for Asian American identity in the academic and extracurricular ratings models are likely at least partially the result of unobservable characteristics that correlate with race, and Professor Arcidiacono has posited that is indeed likely the cause. [Oct. 25 Tr. 108:24–109:8, 110:14–17]. The Court finds, however, that although omitted variables are likely partially responsible for the positive coefficients for Asian American identity in Professor Arcidiacono’s models for the 73 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 74 of 130 academic and extracurricular ratings, those coefficients could also partially be the result of slight implicit bias that favors Asian Americans in these areas. Professor Arcidiacono’s model of the overall rating yields negative coefficients for Asian American males and non-disadvantaged Asian Americans females. [PD38 at 29]. This suggests that Asian Americans who are not also disadvantaged females might receive lower overall ratings because of their racial identity relative to similarly-situated white applicants, see [Oct. 25 Tr. 92:20–94:10; PD38 at 29], but the result is subject to substantially the same criticism that Harvard lodges against Professor Arcidiacono’s admissions outcome model, namely that Professor Arcidiacono’s modeling choices do not fully reflect the actual admissions process and that his decision to exclude ALDC applicants was results-driven. Regardless, the Court finds it unnecessary to delve further into the overall rating disparity because it is the odds of admission, not an apparent disparity in the odds of receiving a high overall rating, that is primarily at issue, and Harvard acknowledges and intends that race may be factored into the overall rating. See [Oct. 18 Tr. 167:17–168:6]. F. Regression Models of Admissions Outcome As noted supra at Section V.D, both Professors Arcidiacono and Card prepared models of domestic non-transfer applicants’ probability of admission to Harvard based on a wide array of variables. [Oct. 25 Tr. 21:18–23:23, 215:12–15; Oct. 30 Tr. 176:18–177:7]. Professor Card’s preferred model returns a negative coefficient for Asian American identity (suggesting a lower likelihood of admission), but the relationship is slight, not statistically significant, and is positive (suggesting an increased likelihood of admission) for some class years. [Oct. 30 Tr. 129:9–16, 74 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 75 of 130 132:21–134:11; DX685; DD10 at 30]. 49 Professor Arcidiacono’s preferred model returns a statistically significant negative coefficient for non-ALDC Asian American applicants, which implies a penalty for non-ALDC Asian American applicants relative to non-ALDC white applicants. [Oct. 25 Tr. 115:1–118:10; PD38 at 34]. Professors Card’s and Professor Arcidiacono’s preferred models differ in the following significant respects: (1) Professor Arcidiacono interacts race and disadvantaged status; (2) Professor Arcidiacono excludes the personal rating from the model; (3) Professor Arcidiacono excludes ALDC applicants; (4) Professor Arcidiacono pooled the 2014–2019 applicant data into a single model with effects for class years, whereas Professor Card used separate year-by-year models and thereby allowed the effect of variables to vary by admissions cycle; and (5) Professor Arcidiacono excludes parental occupation, intended career, and an indicator for whether applicants interviewed with a staff member. See [Oct. 31 Tr. 88:21–91:23; DD10 at 84]. For the reasons discussed below, the Court finds both experts’ approaches to be econometrically defensible, but prefers Professor Arcidiacono’s approach with respect to interacting race and disadvantaged status and prefers Professor Card’s inclusion of ALDC applicants, use of year-byyear models, and inclusion of parental occupation, intended career, and staff interview variables, and finds models with and without the personal rating to be worthy of consideration. Professor Arcidiacono reasonably interacted race and disadvantaged status. [Oct. 25 Tr. 150:11–19]. This approach is consistent with the approach taken by OIR in response to Dean 49 Professor Card also modeled the admissions outcomes for two subgroups of Asian Americans: females and applicants from California. He found that Asian American identity within these subgroups returned positive coefficients that were not statistically significant. [Oct. 25 Tr. 154:7–155:3; Oct. 30 Tr. 136:8–137:8]. These models show that to the extent biases influenced the admissions process, those biases were not uniform across the Asian American applicant population. 75 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 76 of 130 Fitzsimmons’ request and reflects the possibility of some interaction between race and disadvantaged status. See [Oct. 17 Tr. 127:22–129:17; Oct. 25 Tr. 150:11–151:1; PX26]. It was not unreasonable, however, for Professor Card not to interact the selected variables. The inclusion of these interaction terms has only a modest impact on the average marginal effects of Asian American identity generated by the admissions models, and their inclusion alone does not result in Asian American identity having a statistically significant effect when the terms are added to Professor Card’s model. [Oct. 31 Tr. 89:11–18; DD10 at 84]. There is a reasonable econometric basis for removing the personal ratings from the admissions models given the possibility that the personal ratings are affected by race. See [Oct. 25 Tr. 91:17–92:1]. Removing the personal rating, however, expands the omitted variable bias because the relationship between race and the personal rating is likely partially reflective of biases external to the Admissions Office, characteristics that are correlated with race, and life experiences that are impacted by race. See supra at Section V.C. Therefore, although the Court believes that including the personal rating results in a more comprehensive analysis, models with and without the personal rating are econometrically reasonable and provide evidence that is probative of the effect of race on the admissions process. Professor Card’s inclusion of ALDCs in the admissions model is preferred by the Court. Although ALDCs benefit from sizable tips owing to their respective statuses as recruited athletes, legacies, dean’s or director’s list members, or children of faculty or staff, they are evaluated through the same basic admissions process as other applicants and their admission outcomes provide data that is probative of whether Harvard is discriminating against Asian Americans. [Oct. 17 Tr. 203:19–22; Oct. 25 Tr. 30:13–31:3, 233:25–234:3]. Including ALDCs in the model is particularly warranted where they account for approximately 30% of Harvard’s 76 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 77 of 130 admitted students and therefore provide a significant portion of the datapoints for admitted students. [DX706, DD10 at 38]. Professor Arcidiacono acknowledges that Asian American ALDCs are not discriminated against. See [Oct. 25 Tr. 120:23–126:8]. His corresponding suggestion that only non-ALDC Asian Americans face discrimination in the admissions process is inadequately supported by non-statistical evidence. Further, it does not seem likely that Harvard would discriminate against non-ALDC Asian Americans, but not discriminate against ALDC Asian American applicants or that there would be a race-related explanation for treating the two groups differently, especially given the Court’s conclusion based on the testimony of Harvard’s admission officers that any race-related discrimination against Asian American applicants relative to white applicants is unintentional. Additionally, the tips that only ALDCs receive, for example for being recruited athletes, can be adequately accounted for through the inclusion of variables for those characteristics. See [Oct. 30 Tr. 157:24–158:14]. Overall, including ALDCs leads to a model that more accurately reflects how the admissions process works and takes into account a larger percentage of the admitted class. In the view of the Court, looking at only a portion of a class or carving out the segments where there is less likely to be discrimination distorts the analysis just as carving out the segments where there is mostly likely to be discrimination would do the same but to the benefit of the other party. [Id. at 166:21–167:20]. For similar reasons, Professor Card’s modelling of each individual admissions cycle is preferable to Professor Arcidiacono’s pooling of applicants from the six admissions cycles of available data. Professor Card’s year-by-year approach conforms to the reality that the effect of various characteristics in the admissions process may change slightly between years, as Harvard’s institutional interests or admissions policies shift or when the composition of the 77 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 78 of 130 applicant pool changes. [Id. at 167:25–170:15]; see, e.g., [DX703; DX704]. Further, modeling each annual admission cycle independently recognizes that having a class that is 30% African American one year and 0% the next is not the same as having 15% each of those years. Professor Arcidiacono pooled the admissions cycles to achieve a more precise estimate of the effect of Asian American racial identity, but Professor Card’s model achieves a lower standard error, which is an indication of the precision of the model. [Oct. 30 Tr. 172:21–175:18; DD10 at 45]. Professor Arcidiacono omitted intended career, staff interview indicator, and parental occupation from his model. [Oct. 25 Tr. 145:21–148:12]. The Court prefers a model that includes these variables because they play a role in the admissions process. [Oct. 26 Tr. 8:25– 9:21, 10:17–11:6; Oct. 31 Tr. 9:3–7]. Further, these variables correlate with race and therefore create a significant potential for omitted variable bias if excluded. [Oct. 31 Tr. 10:16–18, 11:15– 12:21, 21:19–22:14; DX677; DX681; DD10 at 54]. Professor Arcidiacono excluded these variables primarily because of data issues, including unexplained year-to-year fluctuations in the distribution of parental occupation and intended career categorizations. [Oct. 25 Tr. 145:21– 148:12; DD10 at 50–52, 56]. As examples, numerous parents who were categorized as having low-skill jobs for the class of 2014 would likely have been categorized as being self-employed for the class of 2015, and there is a substantial decrease in the number of parents categorized as unemployed among applicants to the class of 2017 versus the class of 2018. [Oct. 25 Tr. 146:4– 147:9; DD10 at 51–52]. Although the data for parental education and intended career are not as consistent year-to-year as would be ideal, including the variables is preferable because their exclusion results in omitted variable bias that exaggerates the effect of race that is implied by the models. [Oct. 30 Tr. 146:18–147:6; DX695; DD10 at 35]. Professor Card’s model deals 78 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 79 of 130 effectively with data categorization inconsistencies by treating each admission cycle separately, and SFFA has not shown that the data is unreliable within any admissions cycle. [Oct. 30 Tr. 169:12–24]. This data might well need to be excluded if using one data pool for all admission years, but there is no need to exclude it when modeling admissions decisions year-by-year. Professor Card included a staff interview indicator variable, while Professor Arcidiacono excluded the indicator based on his conclusion that the score from an interview should matter, not whether an applicant was interviewed. [Oct. 25 Tr. 148:13–18]. Interviewing with an Admissions Office staff member seemingly affects an applicant’s probability of admission to Harvard, perhaps because it provides an applicant with a potential advocate in the Admissions Office irrespective of how well the applicant performs in that interview, and the Court concludes that including the indicator variable is preferable. See [Oct. 31 Tr. 25:7–27:8]. The Court finds that Professors Card and Arcidiacono each presented a viable econometric model but will rely on Professor Card’s model with the interaction terms utilized by Professor Arcidiacono and then consider results both with and without the personal rating variable included. This model would return a slightly negative coefficient and average marginal effect for Asian American identity, but that coefficient is only statistically significant in the version of the model where the personal rating variable is excluded. See [Oct. 30 Tr. 146:6–17; DD10 at 35]. In fact, without any modifications, Professor Card’s model returns a slight positive average marginal effect for Asian American identity in three of the six admission cycles that the experts analyzed. [DD10 at 30]. Whether the personal rating variable is included or not, the lower probability of admission to Harvard that appears associated with Asian American identity is slight, with an average marginal effect of Asian American racial identity on admissions probability that is well below minus one percentage point (i.e. closer to zero). The model does 79 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 80 of 130 not demonstrate any intent by admissions officers to discriminate based on racial identity, and the implied effect is so slight that it is possible that the coefficient would be positive, at least with the personal rating included, if the model was better able to account for unobserved factors. It 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. If so, the effect was so slight that it went unnoticed by careful and conscientious observers within the Admissions Office. The implied effect varies by admissions cycle and, with the personal rating variable included, results in a positive, statistically insignificant effect for the 2019 class year, which suggests, even though the change is not statistically significant, that any implicit biases against Asian Americans dissipated or were eliminated after the Admissions Office was confronted with the allegations at issue here. See [Oct. 30 Tr. 163:22–164:22; DD10 at 41]. G. Absence of Statistical Support for Racial Balancing or Quotas Harvard does not have any racial quotas and has not attempted to achieve classes with any specified racial composition. [Oct. 18 Tr. 112:1–21, 197:16–19; Oct. 19 Tr. 65:13–25, 197:14–20; Oct. 24 Tr. 123:15–18, 174:10–18, 210:2–9; Nov. 1 Tr. 249:24–250:6]. As discussed supra at Section III.B.4, Harvard evaluates the likely racial composition of its class and provides tips to applicants to help it achieve a diverse class. Those tips are necessary to achieve a diverse class given the relative paucity of minority applicants that would be admitted without such a tip. In trying to assure a diverse class, when reviewing an individual applicant, the admissions officers consider various qualitative and numerical indicators of diversity, including the racial composition of the group of students who are expected to be admitted. Although Harvard tracks and considers various indicators of diversity in the admissions process, including race, the racial composition of Harvard’s admitted classes has varied in a 80 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 81 of 130 manner inconsistent with the imposition of a racial quota or racial balancing. See [Oct. 31 Tr. 119:10–121:10; DX711]. As Figures 1 and 2 show, there has been considerable year-to-year variation in the portion of Harvard’s class that identifies as Asian American since at least 1980. See [DX711 at 2; DD10 at 100–101]. Figure 1: Percent Change in Year-to-Year Admittance of Students by Race. [DD10 at 100; DX711]. 81 Case Document 672 Filed 09/30/19 Page 82 of 130 Figure 2: Percent of Applicants and Admitted Students by Race 1980 through 2019. at 1001980 1985 1990 1995 2000 2005 2010 2015 Asian American Applicants African American Applicants Hispanic Applicants ?Asian American Admitted Students ?African American Admitted Students?Hispanic Admitted Students The demographic makeup of Harvard?s classes from 1980 through 2019 show signi?cant changes to the composition of each class, and there has been more year-to-year variation in admitted Asian American applicants than year-to-year variation in the number of applicants. at 104]. From 1980 to 2019, Asian Americans went from 4.1% of applicants and 3.4% of admitted students to 21.2% of applicants and 20.6% of admitted students. Since 1980, the Asian American proportion of the admitted class has increased roughly ?ve-fold, and since 1990 the Asian American proportion of the admitted class has increased roughly two- fold. SFFA did not offer expert testimony on racial balancing and instead asserts that the claim can be resolved without any expert analysis. [Oct. 25 Tr. 202:6?203: 1; ECF No. 627 1] 84]. 82 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 83 of 130 The Court finds that the statistical evidence shows that Harvard has not imposed racial quotas or otherwise engaged in impermissible racial balancing. VI. FINDINGS OF FACT: RACE-NEUTRAL ALTERNATIVES Under the strict scrutiny rubric established by the Supreme Court, Harvard may consider race to achieve diversity only if there is no workable race-neutral alternative to the consideration of race to ensure a sufficiently diverse class. SFFA introduced models on race-neutral alternatives through an expert, Richard Kahlenberg. 50 The Smith Committee’s conclusions and the analysis performed by Professor Card and Mr. Kahlenberg all convincingly establish that no workable race-neutral alternatives will currently permit Harvard to achieve the level of racial diversity it has credibly found necessary for its educational mission. Harvard’s race-conscious admissions policy has a significant impact on the racial diversity of its class, with African American and Hispanic applicants being the primary beneficiaries in terms of their admissions probabilities. The policy of considering applicants’ race may improve the admission chances of some Asian Americans who connect their racial identities with particularly compelling narratives, but overall results in fewer Asian American and white students being admitted. See [Oct. 31 Tr. 127:22–128:15]. Any race-neutral alternative will be deemed workable only if it would allow Harvard to achieve the benefits that it derives from its current degree of diversity within a given class year, while also being practicable, affordable, and not requiring a material decline in academic quality or any of the other measures of excellence valued by Harvard. 50 Mr. Kahlenberg is a senior fellow at The Century Foundation, where he has worked for the last twenty years. He graduated from Harvard College in 1985 and received a juris doctor from Harvard Law School in 1989. Mr. Kahlenberg has published works on numerous socioeconomic subjects, including the use of race-neutral alternatives in college admissions. [Oct. 22 Tr. 7:15– 12:10]. 83 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 84 of 130 Currently, although always considered in conjunction with other factors and metrics, race is a determinative tip for approximately 45% of all admitted African American and Hispanic applicants. See [DX721 at 1]. At least 10% of Harvard’s admitted class, including more than one third of the admitted Hispanics and more than half of the admitted African Americans, would most likely not be admitted in the absence of Harvard’s race-conscious admissions process. See [Oct. 31 Tr. 127:22–129:2; DX721; DD10 at 107]. 51 In the absence of any other adjustments to Harvard’s admissions policy, eliminating consideration of race would cause the African American representation at Harvard to decline from approximately 14% to 6% of the student population and Hispanic representation to decline from 14% to 9%. [Oct. 31 Tr. 126:21– 129:2]. Over the course of four years, the number of African American and Hispanic students at Harvard would fall by nearly 1,000 students. See [Oct. 25 Tr. 167:20–168:4; PD38 at 39]. The Court notes that Harvard’s current admissions policy does not result in underqualified students being admitted in the name of diversity—rather, the tip given for race impacts who among the highly-qualified students in the applicant pool will be selected for admission to a class that is too small to accommodate more than a small percentage of those qualified for admission. 52 Therefore, removing attention to race, without a workable race-neutral alternative, 51 The econometric models fail to fully reflect the number of students for whom race is determinative. Among other factors, the increased Asian American representation that the models project would likely not include all Asian American students who are admitted under the current race-conscious approach. In the total absence of a race-conscious policy, some Asian American applicants who excelled on academic, athletic, or other metrics of success would likely replace some number of Asian American students from disproportionately less advantaged backgrounds who tell compelling stories about their personal identities that require an understanding of their race. See, e.g., [Oct. 18 Tr. 52:19–56:21; Oct. 29 Tr. 147:6–152:12]. 52 Moreover, other tips in the admissions process, like so many facets of modern-day American life, disproportionately benefit individuals in the majority and more affluent group. 84 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 85 of 130 would cause a sharp decline in the percentage of African American and Hispanic students at Harvard without resulting in a particularly significant increase in the overall academic strength of the class. 53 The parties’ experts, as well at the Smith Committee, examined numerous race-neutral alternatives to determine if they, alone or in combination, could conceivably limit the decline in racial diversity in Harvard’s class in the absence of a race-conscious admissions policy. See [Oct. 22 Tr. 18:1–11; Oct. 31 Tr. 129:3–130:4; PX316 at 6–18]. These alternatives included eliminating early action, tips for ALDC applicants, the practice of offering deferred admissions or z-listing applicants, and consideration of standardized test scores, as well as expanding recruiting and partnership efforts, admitting more transfer students, utilizing a place-based quota system, and expanding preferences for economically disadvantaged applicants. [Oct. 22 Tr. 33:15–49:8; Oct. 31 Tr. 130:5–130:23, 133:10–20; PX316 at 6–18; DD10 at 109]. As more fully set forth below, Harvard has demonstrated that none of these approaches, individually or in combination, would allow it to reach the level of racial diversity that it believes necessary to achieve its educational mission without significant consequences to the strength of its admitted class. A. Eliminating Early Action In an earlier effort to both increase diversity and level the admissions playing field for less advantaged applicants, Harvard eliminated its early action program for the classes of 2012 through 2015, believing that early action disproportionately benefitted affluent applicants and hoping that other elite colleges would follow its lead, which they largely did not. [PX316 at 15]. 53 Similarly, removing the tips for recruited athletes would result in a sharp decline in admitted athletes, removing the tips for children of faculty or staff would reduce their representation, and eliminating the tip for legacies would decrease their numbers as well. In other words, removing any tips changes the make-up of the admitted class, but not necessarily its overall quality. 85 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 86 of 130 This actually had the unintended consequence of decreasing matriculation rates among some categories of African American and Hispanic applicants, apparently because the most qualified of those prospective applicants were choosing to attend other colleges that offered early admission or early decision. [Oct. 23 Tr. 156:17–157:22; DX39 at 2–4]. As a result, Harvard reinstituted an early action program for the class of 2016. [PX316 at 15; DX39 at 4]. Harvard’s actual experience is more probative of the probable result of such a change than econometric prognostications and shows that the likely effect of removing early action on African American and Hispanic enrollment is negative or near zero. [Oct. 31 Tr. 133:20–135:24; DX728 at 3]. As such, eliminating early action does not present a viable race-neutral option for achieving student body diversity. B. ALDC Tips Preferences or tips for ALDC applicants and related deferred admissions also disproportionately benefit socioeconomically advantaged applicants. See [PX316 at 16–17]. Although removing tips for these applicants would improve socioeconomic diversity at Harvard and increase the number of Asian American students, it would not significantly increase the number of African American and Hispanic students if implemented alone. [Oct. 31 Tr. 131:8– 133:8; DX720; DD10 at 112]. Professor Card reasonably estimated that eliminating tips for race and ALDC status, along with eliminating deferred admissions, would cause African American enrollment to decline from 14% to 5% and Hispanic enrollment to decline from 14% to 9%. [Oct. 31 Tr. 132:15–133:19; DX720; DD10 at 112]. Eliminating tips for ALDC applicants would have the effect of opening spots in Harvard’s class that could then be filled through an admissions policy more favorable to non-white students, but Harvard would be far less competitive in Ivy League intercollegiate sports, which would adversely impact Harvard and the student experience. [Oct. 30 Tr. 40:12–41:21]. Eliminating tips for legacies, applicants on the 86 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 87 of 130 dean’s and director’s interest lists, and children of faculty or staff would also come at considerable costs, and 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. [Oct. 23 Tr. 164:19–167:2; Oct. 30 Tr. 20:17– 21:8, 35:25–43:13; PX316 at 16–17]. Therefore, eliminating tips for ALDC applicants and related deferred admissions practices is not alone an adequate race-neutral alternative given the limited probable impact on racial diversity and the likely adverse consequences for Harvard and student life. The Court notes that reasonable minds can differ on the importance of college athletics, alumni relations, and admitting the children of faculty and staff, but takes no position on these issues other than to note that these are topics best left to schools to figure out for themselves. As relevant here, eliminating these tips or preferences is largely unrelated to the goal of diversity or the issue of race, and in any event, is not a race-neutral alternative that would obviate the need for considering race in admissions. C. Augmenting Recruiting Efforts and Financial Aid Harvard looked at expanding recruiting and partnership efforts and providing more financial aid as a way to increase diversity without having to consider race in the application process. The college already makes significant outreach efforts and provides exceptionally generous financial aid. [PX316 at 9–11]. In addition to the HFAI and UMRP programs discussed supra at Section III.A.2, the Smith Committee’s report describes additional community-based outreach efforts and considered but rejected the potential for pipeline programs that are inconsistent with Harvard’s recruitment goals. [PX316 at 10]. 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 87 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 88 of 130 reducing the cost of a Harvard education. See [Oct. 31 Tr. 158:15–161:2; PX316 at 10–11; DD10 at 131–133]. D. Increasing Diversity by Admitting More Transfer Students Harvard might also increase diversity by admitting, as transfers, students who might not have applied or been accepted to Harvard at the outset. For example, it is conceivable that if Harvard expanded its efforts to attract and admit transfer students, it might be able to admit some transfer applicants who did not have the perspective to see attending Harvard as an option or who excelled during two years at another college, thereby demonstrating an academic prowess that might not have been evident right out of high school. Despite the facial appeal of these scenarios, Harvard has demonstrated that accepting an increased number of transfer applicants is also not a viable race-neutral alternative because these applicants are, on average, less diverse and less qualified than applicants to its freshman class. [Oct. 31 Tr. 146:24–149:21; DX730; DD10 at 124–125]. Further, Harvard operates as a four-year residential college and the number of transfer students that it can admit is constrained by the number of available beds, meaning that there is not room for transfer students unless other class members drop out. [PX316 at 12–13]. E. Eliminating Standardized Testing Eliminating consideration of standardized testing is likewise not an adequate race-neutral alternative to considering race in the admissions process. Harvard considers standardized tests to be reflective of academic or intellectual strength and uses SAT and ACT test scores in assigning academic ratings. [PX721 at 4]. Harvard has demonstrated that eliminating consideration of standardized test scores in the admissions process would lead to a reduction in the academic qualifications of its admitted class, at least as measured by the criteria Harvard presently uses. [Oct. 31 Tr. 143:23–146:11; DX722 at 3; DD10 at 121]. As the Smith Committee found, standardized tests are “imperfect measures,” but they can be a useful metric when considered in 88 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 89 of 130 tandem with an applicant’s background. [PX316 at 18]. Although eliminating consideration of standardized test scores might improve diversity slightly, the effects on the academic strength of Harvard’s admitted class makes eliminating the consideration of standardized test scores an unviable race-neutral alternative. See [Oct. 31 Tr. 153:4–154:17; DX723 at 3]. F. Place-Based Quotas The Smith Committee considered place-based quotas, such as admitting the top student from each high school class or from each zip code. [PX316 at 11–12]. Harvard’s evaluation and rejection of these ideas reflects the reality that Harvard is far too selective and high schools and zip codes in the United States too numerous for such an admissions policy to be even close to workable. [Oct. 22 Tr. 107:6–108:2]. Harvard could achieve somewhat improved racial diversity in the absence of a raceconscious admissions policy by increasing the tips for students from disadvantaged economic backgrounds and areas. Under any reasonable implementation, however, this race-neutral approach would result in fewer African Americans than are admitted under the current system and would also come at the expense of traditional measures of academic strength, such as SAT scores. See [Oct. 22 Tr. 125:6–10, 126:17–127:23; PD27; PD29; PD31; PD33]. Mr. Kahlenberg proposes a quota system where Harvard commits to enrolling students from broad neighborhood clusters constructed to generate more representation from racially diverse and disproportionately economically disadvantaged areas, [Oct. 22 Tr. 35:23–36:16], but given the logistical challenges of such an arrangement coupled with the questionable legality of any sort of quota system, as discussed infra at Section VII.G, place-based quotas are not an available and workable race-neutral alternative. 89 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 90 of 130 G. SFFA’s Proposed Combinations of Various Race-Neutral Alternatives Mr. Kahlenberg presented four simulations, labeled A, B, C, and D, that model the combined effect of various allegedly race-neutral alternatives on Harvard’s class. [Oct. 22 Tr. 16:7–14, 29:20–47:6]. The simulations, using the admissions models developed by Professors Card and Arcidiacono with the models’ implied racial tips removed, project the diversity of Harvard’s class with various modifications to the models that are aimed at increasing racial diversity by increasing the tip given to economically disadvantaged applicants, further preferencing applicants from disadvantaged geographic areas, and by removing preferences currently used in Harvard’s admissions process for ALDC students or LDC students that disproportionately benefit white applicants. [Oct 22 Tr. 27:11–27:7]. 54 These simulations show that Harvard could achieve a significant increase in socioeconomic diversity and an increase in the total representation of African American, Hispanic and other (i.e. non-white and non-Asian American) students in its classes but only if it abandoned all preferences for legacies, applicants on the dean’s or director’s interest lists, and children of faculty or staff, and implemented a sizable tip based on economic and geographic indicators of disadvantage. See [PD27; PD29; 54 In all of the simulations, the implied effects of tips given to LDCs are removed. [Oct. 22 Tr. 34:17–35:9; PD32]. Simulation B, which utilizes Professor Card’s model and simulation, projects the effect of removing preferences for recruited athletes as well. [Id. at 41:3–42:9]. The simulations all impose some form of a socioeconomic and/or geographic status boost. [PD32]. Model A expands the boost associated with disadvantaged status such that it is half the magnitude of the tip that the model suggests is currently granted to recruited athletes and forces equal selection of applicants from 33 neighborhood clusters, [Oct. 22 Tr. 35:23–36:16]; Model B boosts for socioeconomically disadvantaged students based on census tract income, [id. at 41:20– 42:1]; and Simulations C and D modify the socioeconomic and census tract boost used in Simulation B and consider whether an applicant attended a disadvantaged high school, [id. at 43:7–44:16]. Models A and C also remove the admissions models’ implied preference for early action applicants, while models B and D include that preference. [Id. at 42:2–3, 46:10–12; PD32]. 90 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 91 of 130 PD31; PD33]. For example, Simulation D projects that 49% of Harvard’s class would be from an economically disadvantaged background, relative to the 12% in the class of 2019. [PD33]. Mr. Kahlenberg’s changes to the admissions policy would come at significant costs. In addition to the loss of benefits provided by tips for ALDCs or LDCs, the simulations show a 53 to 71-point decline in average SAT scores. [PD27; PD29; PD31; PD33]. These declines in average SAT score would be associated with more significant declines in the expected strength of Harvard’s class across the profile ratings, with the amount of the expected decline varying depending on the simulation selected. For example, under Simulation C, the portion of the admitted class achieving a 1 or 2 in each profile rating falls by between 13% and 26%. [DX729 at 11; DD10 at 141]. The simulations also imply substantial changes to the academic interests of Harvard’s admitted classes that would pose administrative and staffing challenges. [DX729]. For example, Mr. Kahlenberg’s models would likely lead to more students being admitted who indicated an intended concentration in engineering and fewer admitted students who intend to concentrate in the humanities, which would likely require Harvard to expand and contract its academic programs accordingly. Finally, and perhaps most significantly for present purposes, Mr. Kahlenberg’s simulations uniformly suggest that African American representation in Harvard’s incoming class would fall nearly one-third to approximately 10% of the class. [Oct. 22 Tr. 127:16–23]. In order to achieve, without race-conscious policies, comparable numbers of African American students in its admitted classes to those Harvard currently achieves, Harvard would likely need to eliminate all ALDC preferences, eliminate consideration of standardized tests, significantly expand the tip for disadvantaged applicants, and find a way to increase the number of disadvantaged applicants so that more of those disproportionately minority applicants could be 91 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 92 of 130 admitted. [Oct. 31 Tr. 153:4–154:3; DX723 at 1]. These changes, even assuming they could be achieved, would result in a significant decline in the strength of Harvard’s admitted classes across multiple dimensions, including its potential for academic and scholarly achievement. See [Oct. 31 Tr. 154:2–24; DX723 at 3; DD10 at 127]. Harvard plausibly concludes that reshaping its incoming classes in this way would have negative effects on Harvard’s attractiveness to potential students, adversely affect the educational experience at Harvard generally, and that the resulting decrease in the number of African American students would exacerbate “feelings of isolation and alienation among racial minorities in Harvard’s community.” See supra at Section III.A.1; [PX316 at 8]. The Court therefore concludes that Harvard has demonstrated that there are no workable and available race-neutral alternatives, singly or taken in combination, that would allow it to achieve an adequately diverse student body while still perpetuating its standards for academic and other measures of excellence. This conclusion is corroborated by the work of the experts retained by both sides, none of whom have proposed alternatives that would allow Harvard to meet its diversity goals while not unduly compromising on its other legitimate institutional objectives. VII. CONCLUSIONS OF LAW A. Overview The Court first affirms its previously expressed view that SFFA has standing and then turns to SFFA’s four pending Title VI claims: impermissible racial balancing (Count II), failure to use race merely as a “plus factor” (Count III) the availability of race-neutral alternatives (Count V), and intentional discrimination (Count I). Ultimately, the Court finds that Harvard has met its burden of showing that its admissions process complies with the principles articulated by 92 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 93 of 130 the Supreme Court in Fisher II, 136 S. Ct. at 2208, and concludes that judgment must issue for Harvard on each of the remaining claims. B. SFFA Has Standing The constitutional extent of federal court jurisdiction is limited by Article III, which provides that “judicial power” extends to “Cases” and “Controversies” that, inter alia, arise “under this Constitution [or] the Laws of the United States.” U.S. Const. Art. III § 2, cl. 1. “Over the years, [Supreme Court] cases have established that the irreducible constitutional minimum of standing contains three elements:” (1) “an injury in fact—an invasion of a legally protected interest which is (a) concrete and particularized and (b) actual or imminent, not conjectural or hypothetical;” (2) “a causal connection between the injury and the conduct complained of—the injury has to be fairly traceable to the challenged action of the defendant, and not the result of the independent action of some third party not before the court;” and (3) “it must be likely, as opposed to merely speculative, that the injury will be redressed by a favorable decision.” Lujan v. Defenders of Wildlife, 504 U.S. 555, 561 (1992) (citations and modifying punctuation omitted). “The party invoking federal jurisdiction bears the burden of establishing these elements.” Id. Under the doctrine of associational standing, “an association may have standing solely as the representative of its members even in the absence of injury to itself, in certain circumstances.” Camel Hair & Cashmere Inst. of Am., Inc. v. Associated Dry Goods Corp., 799 F.2d 6, 10 (1st Cir. 1986). As the Supreme Court has held: [A]n association has standing to bring suit on behalf of its members when: (a) its members would otherwise have standing to sue in their own right; (b) the interests it seeks to protect are germane to the organization’s purpose; and (c) neither the claim asserted nor the relief requested requires the participation of individual members in the lawsuit. Hunt v. Wash. State Apple Advert. Comm’n, 432 U.S. 333, 343 (1977). 93 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 94 of 130 During this litigation, SFFA demonstrated that its members included individuals who had standing to pursue this litigation on their own, that this litigation was germane to SFFA’s purpose, and that the injunctive relief SFFA seeks does not require the participation of those members in this lawsuit. See Students for Fair Admissions, 261 F. Supp. 3d at 110–11. Harvard argued at the summary judgment stage that the case had become moot because the SFFA members who the Court found had individual standing were no longer participating in the college admissions process or seriously interested in transferring. “Mootness usually results when a plaintiff has standing at the beginning of a case, but, due to intervening events, loses one of the elements of standing during litigation . . . .” Wild Earth Guardians v. Pub. Serv. Co. of Colo., 690 F.3d 1174, 1182 (10th Cir. 2012) (citing Arizonans for Official English v. Arizona, 520 U.S. 45, 68 n.22 (1997)). At summary judgment, the Court found that “Harvard ha[d] not established that the case ha[d] become moot based on the [members’] alleged disinterest in transferring.” Students for Fair Admissions, 346 F. Supp. 3d at 191 (D. Mass. 2018). Harvard now asserts that the Court should have applied a more stringent standard, including requiring SFFA to show that its members control its conduct and possess certain “indicia of membership.” [ECF No. 619 ¶¶ 326–30]. Harvard’s standing arguments are preserved for appeal. C. The Supreme Court and Race-Conscious Admissions Although this Court, as it must, relies principally on the Supreme Court’s most recent guidance as set forth in Fisher II, a brief synopsis of the case law which culminated in Fisher II follows. The Supreme Court directly confronted the issue of affirmative action or race-conscious admissions in the context of higher education for the first time in Regents of University of California v. Bakke, 438 U.S. 265 (1978) (plurality opinion). In that case, the Supreme Court struck down an admissions policy at the University of California at Davis Medical School 94 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 95 of 130 pursuant to Title VI and the Equal Protection Clause. Bakke, 438 U.S. at 271 (1978). At that time, the Medical School admitted most of its minority students through a “special admissions program” that filled 16 of the class’ 100 spots with economically or educationally disadvantaged applicants who were members of a minority group. Id. at 272–75. White applicants could compete for 84 of the seats in the Medical School’s class, while all 100 seats were potentially open to minority students. Id. at 289. Justices Brennan, White, Marshall, and Blackmun would have found Title VI coextensive with the Equal Protection Clause and upheld the medical school’s policy on the basis that the government may use race to remedy disadvantages to minorities caused by past racial prejudice. Id. at 355, 324–79 (concurring in part and dissenting in part). Chief Justice Burger and Justices Stevens, Stewart, and Rehnquist would have found the special admissions program in violation of Title VI, irrespective of the Equal Protection Clause. Id. at 408–21. Justice Powell, who announced the judgment of the Supreme Court, agreed with Justices Brennan, White, Marshall, and Blackmun that Title VI proscribes only those racial classifications that would violate the Equal Protection Clause, but unlike his fellow justices, concluded that diversity was an asserted state interest that could withstand strict scrutiny and that to satisfy strict scrutiny, the medical school’s approach to diversity had to “encompass[ a broad] array of qualifications and characteristics of which racial or ethnic origin is but a single though important element.” Id. at 315. Although no majority agreed on a particular rationale, the Supreme Court determined that the medical school’s special admissions program was unconstitutional because it involved “the use of an explicit racial classification” that told “applicants who are not Negro, Asian, or Chicano that they [were] totally excluded from a specific percentage of the seats in an entering class.” Id. at 319. 95 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 96 of 130 Nevertheless, a majority of the Supreme Court believed that race could be used in higher education admissions, and it was understood that Justice Powell’s opinion in Bakke permitted the use of race or ethnic background as a “plus” factor to further the goal of diversity in education. Justice Powell attached the Harvard College Admissions Program as an appendix to his opinion in Bakke and used it as a basis to conclude that the “assignment of a fixed number of places to a minority group is not a necessary means toward” diversity. Id. at 316, 321–24. In contrast with Harvard’s admissions process, which purported to treat “each applicant as an individual in the admissions process” and did not foreclose applicants from competing for the last available seat “simply because he was not the right color or had the wrong surname,” id. at 318, the “fatal flaw” in the medical school’s “preferential program” was its “disregard of individual rights,” id. at 320. Twenty-five years later, the Supreme Court revisited the subject of racial preferences in higher education admissions in a pair of cases concerning the University of Michigan’s Law School and its College of Literature, Science, and the Arts. In Grutter v. Bollinger, 539 U.S. 306 (2003), the Supreme Court concluded that the admissions process at the University of Michigan Law School was constitutionally permissible. 539 U.S. at 325. The law school considered applicants with a focus on academic ability coupled with a flexible assessment of applicants’ talents, experiences, and potential to contribute to the learning of those around them. Id. at 315. Admissions officials were required to consider all the information available in an applicant’s file, including a personal statement, letters of recommendation, undergraduate grades, admissions test scores, and an essay describing the ways the applicant would contribute to the life and diversity of the law school. Id. at 315. While not restricting the types of diversity eligible for consideration or defining diversity solely in terms of racial or ethnic status, the law school was 96 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 97 of 130 committed to “racial and ethnic diversity with special reference to the inclusion of students from groups which have been historically discriminated against.” Id. at 316. In deciding Grutter, the Supreme Court clarified that strict scrutiny applies to the use of race in college admissions, agreed that the law school had a compelling interest in obtaining the educational benefits that flow from a diverse student body, and concluded that the law school’s race-conscious admissions process was sufficiently narrowly tailored. Id. at 333–34. The Supreme Court found that the law school’s goal of “enroll[ing] a critical-mass of minority students,” did not run afoul of the requirement that a school not attempt to attain “some specified percentage of a particular group merely because of its race or ethnic origin,” which would “amount to outright racial balancing” and be “patently unconstitutional.” Id. at 329–30 (quoting Bakke, 438 U.S. at 307). Instead, as distinct from a quota, the concept of “critical mass [was] defined by reference to the educational benefits that diversity is designed to produce,” including racial understanding, breaking down stereotypes, advancing learning outcomes, and preparing students for a diverse workforce and society. Id. at 330. The Supreme Court noted that the law school’s admissions program bore the hallmarks of a narrowly tailored plan: truly individualized consideration including the use of race in a “flexible, nonmechanical way,” no quotas or separate admissions tracks for members of certain racial groups, and no insulating “applicants who belong to certain racial or ethnic groups from the competition for admission.” Id. at 334. In upholding the law school’s admissions process in Grutter, the Supreme Court again approved of “the Harvard plan,” as described by Justice Powell in Bakke. See id. at 335. Like Harvard, the University of Michigan Law School did not have a “quota,” meaning “a program in which a certain fixed number or proportion of opportunities are ‘reserved exclusively for certain minority groups.’” Id. (quoting Richmond v. J.A. Croson Co., 488 U.S. 469, 496 (1989) 97 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 98 of 130 (plurality opinion)). Rather, the law school pursued a “permissible goal” that “require[d] only a good-faith effort to come within a range demarcated by the goal itself,” and “permit[ed] consideration of race as a ‘plus’ factor in any given case while still ensuring that each candidate ‘competes with all other qualified applicants.’” Id. (punctuation omitted) (first quoting Sheet Metal Workers v. EEOC, 478 U.S. 421, 495 (1986) and then quoting Johnson v. Transp. Agency, Santa Clara Cty., 480 U.S. 616, 638 (1987)). The Court noted that the Harvard plan, previously endorsed by Justice Powell in Bakke, “certainly had minimum goals for minority enrollment, even if it had no specific number firmly in mind,” but it reiterated that Justice Powell had “flatly rejected the argument that Harvard’s program was ‘the functional equivalent of a quota’ merely because it had some ‘plus’ for race, or gave greater ‘weight’ to race than to some other factors, in order to achieve student body diversity.” Id. (quoting Bakke, 438 U.S. 317–18, 323). Further, like the Harvard plan, Michigan Law’s admissions process was “flexible enough to consider all pertinent elements of diversity in light of the particular qualifications of each applicant, and to place them on the same footing for consideration, although not necessarily according them the same weight.” Id. (quoting Bakke, 438 U.S. at 317). Although race was given substantial weight in the admissions process, the law school also considered “the broad range of qualities and experiences that may be considered valuable contributions to student body diversity,” including fluency in several languages, a history of overcoming personal adversity and family hardship, exceptional records of extensive community service, and successful careers in other fields, and “actually [gave] substantial weight to diversity factors besides race.” Id. at 338. While race may have been “‘outcome determinative for many members of minority groups[]’ who do not fall within the upper range of LSAT scores and grades,” that possibility 98 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 99 of 130 was not dispositive given that “the same could be said of the Harvard plan discussed approvingly by Justice Powell in Bakke.” Id. at 338 (quoting Grutter, 539 U.S. at 389 (Kennedy, J., dissenting)). The Supreme Court noted in Grutter that “all underrepresented minority students admitted by the Law School [had] been deemed qualified,” although minority applicants were “less likely to be admitted in meaningful numbers on criteria that ignore[d]” race and experiences with racial inequality, which were of “particular importance to the Law School’s mission.” Id. On the same day the Supreme Court decided Grutter, it held in Gratz v. Bollinger, 539 U.S. 244 (2003), that the admissions process at the University of Michigan College of Literature, Science, and the Arts violated Title VI and the Equal Protection Clause. Gratz, 539 U.S. at 275. The University of Michigan admitted or rejected applicants to the College of Literature, Science, and the Arts based on the number of points that an applicant scored under a rubric that offered points for high school GPA, standardized test scores, the academic strength of the applicant’s high school, the applicant’s high school curriculum, in-state residency, alumni relationship, personal essay, and other achievements. Id. at 255. Underrepresented minority applicants received an additional 20 points scored in a “miscellaneous” category which provided a significant bump towards the 75 to 100 points that were, depending on the year and the applicant’s in-state residency status, generally required for admission. Id. at 255–56, & n.8. The Supreme Court in Gratz concluded that the admissions policy was impermissible under Justice Powell’s opinion in Bakke because giving every underrepresented minority applicant 20 points did not provide the necessary “individualized consideration” and instead had “the effect of making ‘the factor of race . . . decisive’ for virtually every minimally qualified underrepresented minority applicant.” Id. at 271–72 (quoting Bakke, 438 U.S. at 317). The university’s use of 99 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 100 of 130 race was therefore not narrowly tailored to achieve the asserted compelling interest in diversity and thus violated the Equal Protection Clause and Title VI. Id. at 275–76. More recently, in the Fisher cases, the Supreme Court reviewed the undergraduate admissions program at the University of Texas at Austin (“UT Austin”), which considered race as one factor among many in assigning a personal achievement index which, together with an academic index, determined whether applicants who were not in the top 10% of their Texas high school class would be admitted or rejected. Fisher I, 570 U.S. at 304–07. In 2013 in Fisher I, the Supreme Court vacated the Fifth Circuit’s decision upholding UT Austin’s admissions program because the appeals court had not properly conducted the strict scrutiny analysis. Id. at 303. The Fifth Circuit had undertaken the narrow tailoring analysis with a degree of deference to the university, presuming that the school had made a good-faith decision to use race and then imposing the burden of rebutting that presumption on the plaintiff. Id. at 311–15. The Supreme Court concluded that no such deference to a university was permitted in undertaking the narrow tailoring analysis. Id. Following remand, the Fifth Circuit found that UT Austin had demonstrated that the use of race in its admissions program was narrowly tailored to achieve the rich diversity that contributed to UT Austin’s academic mission and once again affirmed the district court’s judgment that UT Austin’s admissions program did not violate the Equal Protection Clause. Fisher v. Univ. of Tex. at Austin, 758 F.3d 633, 657, 659–61 (5th Cir. 2014). The Supreme Court granted certiorari once more, and in 2016 it affirmed the Fifth Circuit’s ruling. Fisher II, 136 S. Ct. at 2214–15. In Fisher II, the Supreme Court stated the following three controlling principles: First, because racial characteristics so seldom provide a relevant basis for disparate treatment, race may not be considered . . . unless the admissions process 100 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 101 of 130 can withstand strict scrutiny. Strict scrutiny requires the university to demonstrate with clarity that its purpose or interest is both constitutionally permissible and substantial, and that its use of the classification is necessary to the accomplishment of its purpose. Second, . . . the decision to pursue the educational benefits that flow from student body diversity is, in substantial measure, an academic judgment to which some, but not complete, judicial deference is proper. A university cannot impose a fixed quota or otherwise define diversity as some specified percentage of a particular group merely because of its race or ethnic origin. Once, however, a university gives a reasoned, principled explanation for its decision, deference must be given to the University’s conclusion, based on its experience and expertise, that a diverse student body would serve its educational goals. Third, . . . no deference is owed when determining whether the use of race is narrowly tailored to achieve the university’s permissible goals. A university . . . bears the burden of proving a nonracial approach would not promote its interest in the educational benefits of diversity about as well and at tolerable administrative expense. Though narrow tailoring does not require exhaustion of every conceivable race-neutral alternative or require a university to choose between maintaining a reputation for excellence and fulfilling a commitment to provide educational opportunities to members of all racial groups, it does impose on the university the ultimate burden of demonstrating that race-neutral alternatives that are both available and workable do not suffice. Id. at 2208 (citations and modifying punctuation omitted). In applying these principles in Fisher II, the Supreme Court determined that UT Austin had provided a reasoned and principled articulation of concrete and precise goals for its raceconscious admissions program, including destroying racial stereotypes, promoting cross-racial understanding, preparing the student body for an increasingly diverse workforce and society, cultivating leaders with legitimacy in the eyes of the citizenry, providing an educational environment that fosters the robust exchange of ideas, exposure to different cultures, and the acquisition of the competencies required of future leaders. Id. at 2211. The Supreme Court noted “that a university bears a heavy burden in showing that it had not obtained the educational benefits of diversity before it turned to a race-conscious plan,” but found that UT Austin had engaged in good faith studies from which it reasonably “concluded that ‘[t]he use of race-neutral 101 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 102 of 130 policies and programs ha[d] not been successful in achieving’ sufficient racial diversity at the University,” and that this position was supported by both statistical and anecdotal evidence. Id. at 2211–12 (quoting the record). Lastly, none of the plaintiff’s proposed race-neutral alternatives, or any of the other proposals discussed in the course of the litigation, was shown to have been an “‘available’ and ‘workable’ means through which the University could have met its educational goals, as it then understood and defined them” without considering race, because “the Equal Protection Clause does not force universities to choose between a diverse student body and a reputation for academic excellence.” Id. at 2213–14 (quoting Fisher I, 570 U.S. at 312). Most significantly, the controlling principles articulated by the Supreme Court in Fisher II reflect the sum of its holdings in cases concerning higher education admissions over the last forty years and now guide the application of Title VI in this case. D. Harvard’s Admission Program and Strict Scrutiny Title VI provides, “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.” 42 U.S.C. § 2000d. In the higher education admissions context, the contours of Title VI claims are largely shaped by the Equal Protection Clause of the Fourteenth Amendment. The “intentional discrimination proscribed by Title VI is discrimination that violates the Equal Protection Clause of the Fourteenth Amendment.” Weser v. Glen, 190 F. Supp. 2d 384, 396 (E.D.N.Y.), aff’d, 41 F. App’x 521 (2d Cir. 2002); see Bakke, 438 U.S. at 284, 286 (noting that Title VI reflects a “congressional intent to halt federal funding of entities that violate a prohibition of racial discrimination similar to that of the Constitution,” but “proscribe[s] only those racial classifications that would violate the Equal Protection Clause or the Fifth Amendment”); see also 102 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 103 of 130 Grutter, 539 U.S. at 343 (adopting reasoning in Bakke); Gratz, 539 U.S. at 276 n.23 (“We have explained that discrimination that violates the Equal Protection Clause of the Fourteenth Amendment committed by an institution that accepts federal funds also constitutes a violation of Title VI.” (first citing Alexander v. Sandoval, 532 U.S. 275, 281 (2001), then citing United States v. Fordice, 505 U.S. 717, 732, n.7 (1992), and then citing Alexander v. Choate, 469 U.S. 287, 293 (1985))). Although Harvard is not a state actor, Harvard College is a component of Harvard University which receives federal funds and intentionally provides tips in its admissions process based on students’ race. See [ECF No. 570 at 9–10]. Harvard College is therefore subject to the same standards that the Equal Protection Clause imposes upon state actors for the purposes of a Title VI claim. See Students for Fair Admissions, 346 F. Supp. 3d at 192 n.16 (“Harvard does not identify any specific reasons for distinguishing public universities from federally-funded private universities, or explain how the analytical framework would differ for private versus public litigants . . . .”). Under Grutter, “strict scrutiny must be applied to any admissions program using racial categories or classifications.” Fisher I, 570 U.S. at 310; see also Grutter 539 U.S. at 326. Because Harvard both accepts federal funds and uses race in making admissions decisions, its admissions program is subject to strict scrutiny. Harvard argues that the test for a “facially neutral policy” should be applied, 55 but Harvard’s admissions process is not facially neutral. Fisher I, 570 U.S. at 307 (“It is . . . 55 The analysis of a facially neutral policy that has a disparate impact by race is different from the analysis of a policy that admittedly considers race. “In reviewing a uniformly applied facially neutral policy, ‘[d]etermining whether invidious discriminatory purpose was a motivating factor [in its adoption] demands a sensitive inquiry into such circumstantial and direct evidence of intent as may be available.’” Students for Fair Admissions, Inc. v. President & Fellows of Harvard Coll., 346 F. Supp. 3d 174, 193 (D. Mass. 2018) (quoting Anderson ex rel. 103 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 104 of 130 irrelevant that a system of racial preferences in admissions may seem benign. Any racial classification must meet strict scrutiny.”). Although Harvard’s reading procedures do not explicitly preference particular racial groups, Harvard pursues its interest in diversity in part by considering the race of applicants, and its admissions officers may take an applicant’s race into account when making an admissions decision even when the applicant has not discussed their racial or ethnic identity in their application. [Oct. 18 Tr. 52:15–53:13; 167:10–168:11]. Harvard’s acknowledged consideration of race is unlike a facially neutral policy which requires plaintiffs to prove racial discrimination. Cf. Vill. of Arlington Heights v. Metro. Hous. Dev. Corp., 429 U.S. 252, 270–71 (1977) (plaintiffs “failed to carry their burden of proving that discriminatory purpose was a motivating factor” for a rezoning decision that did not explicitly rely on race). Here, the use of race in and of itself is admitted, and the issue becomes whether it is permissible given the justification and the means used to achieve the sought-after diversity—in other words, whether Harvard’s use of race survives strict scrutiny. Notably, the Supreme Court has consistently used strict scrutiny when reviewing school admissions programs that consider race. 56 Dowd v. City of Boston, 375 F.3d 71, 83 (1st Cir. 2004)). Policies that do not explicitly consider race are facially neutral and violate the Equal Protection Clause based on statistical evidence only where they form a clear pattern, unexplainable on grounds other than race. See Yick Wo v. Hopkins, 118 U.S. 356, 374 (1886) (finding unconstitutional the administration of a facially neutral policy for licensing laundries where permits had been denied to 200 Chinese applicants but granted to all but one of 80-odd others permit applicants who were not Chinese); see also Gomillion v. Lightfoot, 364 U.S. 339 (1960) (finding unconstitutional an alteration to the shape of Tuskegee, Alabama “to remove from the city all save four or five of its 400 Negro voters while not removing a single white voter or resident”). A policy that relies on race at least in part is subject to strict scrutiny regardless of its impact. Therefore, cases like Gomillion v. Lightfoot, 364 U.S. 339, 340–41 (1960) and Yick Wo v. Hopkins, 118 U.S. 356 (1886) are inapposite. 56 Where a school admissions program is subject to strict scrutiny, the Court understands this to mean that the admissions program in its entirety is subject to strict scrutiny and not just the admissions decisions that involve the students that it seeks to advantage. Here, Harvard presses 104 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 105 of 130 Strict scrutiny requires that classifications used by Harvard in its admissions program be narrowly tailored to further a compelling interest. 57 See id. (“Strict scrutiny requires the the idea that its admissions program is facially neutral and should be evaluated by a less demanding standard than strict scrutiny. Harvard’s admissions program is facially neutral in that it does not explicitly prioritize any particular racial group over any other and permits its admissions officers to evaluate the racial and ethnic identity of every student in the context of his or her background and circumstances. The policy cannot, however, be considered facially neutral from a Title VI perspective given that admissions officers provide tips to African American and Hispanic applicants, while white and Asian American applicants are unlikely to receive a meaningful race-based tip. In this circumstance, the standard for facially neutral policies could arguably be applied in evaluating any disparate outcomes as between whites and Asian Americans, keeping in mind that the purpose of strict scrutiny is to ferret out inappropriate racial classifications, and given that there is no suggestion of a racially motivated classification involving whites and Asian Americans. See Richmond v. J. A. Croson Co., 488 U.S. 469, 493 (1989) (plurality opinion) (noting that the purpose of subjecting a racial classification to strict scrutiny is to determine “what classifications are ‘benign’ or ‘remedial’ and what classifications are in fact motivated by illegitimate notions of racial inferiority or simple racial politics”); Grutter, 539 U.S. at 326 (“We apply strict scrutiny to all racial classifications to smoke out illegitimate uses of race by assuring that government is pursuing a goal important enough to warrant use of a highly suspect tool.” (quotation marks omitted and modifying punctuation omitted)). In the case of a facially neutral policy, “[p]roof of racially discriminatory intent or purpose is required to show a violation of the Equal Protection Clause.” Vill. of Arlington Heights v. Metro. Hous. Dev. Corp., 429 U.S. 252, 265 (1977). Were that standard to be applied here, the Court would easily find in favor of Harvard on SFFA’s claim of intentional discrimination as there has been no showing of discriminatory intent or purpose. 57 SFFA contends that it may also succeed on its intentional discrimination claim by showing a “pattern or practice” of discrimination through statistically significant evidence of discrimination that then shifts to Harvard the burden of disproving the alleged pattern or practice. [ECF No. 620 ¶¶ 167–76]. This burden shifting framework, which is rooted in the statutory provisions of Title VII, see 42 U.S.C. § 2000e-6, is inapplicable to a non-class, private plaintiff such as SFFA, even assuming that it could apply in a Title VI case. See Chin v. Port Auth. of N.Y. & N.J., 685 F.3d 135, 149–50 (2d Cir. 2012) (holding “that the pattern-or-practice method of proof is not available to nonclass, private plaintiffs in cases such as the one before us” and noting that “all of our sister circuits to consider the question have held that the pattern-or-practice method of proof is not available to private, nonclass plaintiffs”); see also Semsroth v. City of Wichita, 304 Fed. Appx. 707, 715 (10th Cir. 2008); Davis v. Coca-Cola Bottling Co. Consol., 516 F.3d 955, 967– 69 (11th Cir. 2008); Bacon v. Honda of Am. Mfg., 370 F.3d 565, 575 (6th Cir. 2004); Celestine v. Petroleos de Venezuella SA, 266 F.3d 343, 355–56 (5th Cir. 2001), abrogated on other grounds by, Health v. Bd. of Supervisors for the S. Univ. of Agric. & Mech. Coll., 850 F.3d 731 (5th Cir. 2017); Gilty v. Vill. of Oak Park, 919 F.2d 1247, 1252 (7th Cir. 1990); Lowery v. Circuit City Stores, Inc., 158 F.3d 742, 761 (4th Cir. 1998), vacated on other grounds, 527 U.S. 1031 (1999). 105 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 106 of 130 university to demonstrate with clarity that its ‘purpose or interest is both constitutionally permissible and substantial, and that its use of the classification is necessary . . . to the accomplishment of its purpose.’” (quoting Bakke, 438 U.S. at 305)). 1. Compelling Interest In Bakke, Justice Powell found that student body diversity and the educational benefits that flow from a diverse student body was a compelling interest that could justify the consideration of race. 438 U.S. at 315 (“As the interest of diversity is compelling in the context of a university’s admissions program,” the remaining question is “whether the program’s racial classification is necessary to promote this interest.”). Importantly, he went on to explain that “[t]he diversity that furthers a compelling state interest encompasses a far broader array of qualifications and characteristics of which racial or ethnic origin is but a single though important element.” Bakke, 438 U.S. at 315. Twenty-five years later, the Supreme Court, in Grutter, reaffirmed that “student body diversity is a compelling state interest that can justify the use of race in university admissions.” Grutter, 539 U.S. at 325; see also Fisher I, 570 U.S. at 308–09 (reiterating that prior cases had found that “obtaining the educational benefits of student body diversity is a compelling state interest” (citation and internal quotation marks omitted)). Here, for the reasons discussed supra at Section III.A.1, Harvard’s interest in student body diversity is substantial and compelling. Its goals are not “elusory or amorphous,” and are instead “sufficiently measurable to permit judicial scrutiny of the policies adopted to reach them.” Fisher II, 136 S. Ct. at 2211. These goals include “enhance[ing] the education of [its] students of all races and background [to] prepare them to assume leadership roles in the increasingly pluralistic society into which they will graduate,” achieving the “benefits that flow from [its] students’ exposure to people of different background, races, and life experiences” by teaching them to engage across differences through immersion in a diverse community, and 106 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 107 of 130 “broaden[ing] the perspectives of teachers[, and] thus tend[ing] to expand the reach of the curriculum and the range of scholarly interests of [its] faculty.” [PX302 at 1–2, 9]. Harvard’s goals are similar in specificity to goals the Supreme Court found “concrete and precise” in Fisher II. See 136 S. Ct. 2211. Racial categorizations are necessary to achieve those goals. In the absence of such categorizations, racial diversity at Harvard would likely decline so precipitously that Harvard would be unable to offer students the diverse environment that it reasonably finds necessary to its mission. See infra at Section VII.G. 2. Narrowly Tailored Even in the limited circumstance when drawing racial distinctions is permissible to further a compelling state interest, a university is still “constrained in how it may pursue that end: ‘The means chosen to accomplish the [university’s] asserted purpose must be specifically and narrowly framed to accomplish that purpose.’” Shaw v. Hunt, 517 U.S. 899, 908 (1996) (quoting Wygant v. Jackson Bd. of Ed., 476 U.S. 267, 280 (1986)). Therefore, to satisfy strict scrutiny, “a university must make a showing that its plan is narrowly tailored to achieve the only interest that this Court has approved in this context: the benefits of a student body diversity that ‘encompasses a . . . broa[d] array of qualifications and characteristics of which racial or ethnic origin is but a single though important element.’” Fisher I, 570 U.S. at 308 (quoting Bakke, 438 U.S. at 315). “When race-based action is necessary to further a compelling governmental interest, such action does not violate the constitutional guarantee of equal protection so long as the narrow-tailoring requirement is also satisfied.” Grutter, 539 U.S. at 327; see also J.A. Croson Co., 488 U.S. at 493 (plurality opinion) (“The purpose of strict scrutiny is to ensure that “the means chosen ‘fit’ . . . th[e] compelling goal so closely that there is little or no possibility that the motive for the classification was illegitimate racial prejudice or stereotype.”). 107 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 108 of 130 “To be narrowly tailored, a race-conscious admissions program cannot use a quota system,” Grutter, 539 U.S. at 334, but instead must “remain flexible enough to ensure that each applicant is evaluated as an individual and not in a way that makes an applicant’s race or ethnicity the defining feature of his or her application,” Fisher I, 570 U.S. at 309 (quoting Gratz, 539 U.S. at 337). “In other words, an admissions program must be ‘flexible enough to consider all pertinent elements of diversity in light of the particular qualifications of each applicant.’” Grutter, 539 U.S. at 334 (quoting Bakke, 438 U.S. at 317). Thus, individualized consideration in the context of a race-conscious admissions program is paramount. See id.; Bakke, 438 U.S. at 318 n.52 (identifying the “denial . . . of th[e] right to individualized consideration” as the “principal evil” of the medical school’s admissions program). The Court finds that Harvard’s admissions program “bears the hallmarks of a narrowly tailored plan” in that “race [is] used in a flexible, nonmechanical way” and considered “as a ‘plus’ factor in the context of individualized consideration of each and every applicant.” Grutter, 539 U.S. at 334. Like the University of Michigan Law School in Grutter, Harvard “engages in a highly individualized, holistic review of each applicant’s file, giving serious consideration to all the ways an applicant might contribute to a diverse educational environment,” “this individualized consideration [is afforded] to applicants of all races,” and its “race-conscious admissions program adequately ensures that all factors that may contribute to student body diversity are meaningfully considered alongside race in admissions decisions.” Id. at 337–38. The nature of the allegations in this case however, requires that the analysis go further. 58 Given the “serious problems of justice connected with the idea of preference itself,” Bakke, 438 58 Even though Harvard has shown that its admissions policy must consider race to serve its substantial and compelling interests, the application of strict scrutiny requires a “a further 108 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 109 of 130 U.S. at 298, narrow tailoring further requires “that a race-conscious admissions program not unduly harm members of any racial group,” Grutter, 539 U.S. at 341; see also Metro Broad., Inc. v. FCC, 497 U.S. 547, 630 (1990) (O’Connor, J., dissenting) (a race-conscious admissions program must not “unduly burden individuals who are not members of the favored racial and ethnic groups”). The remaining issue is whether Harvard’s admissions program unduly burdens Asian American applicants. Based on Professor Card’s model and the Court’s preferred model with the personal rating included, there is not a statistically significant difference between the chances of admission for similarly situated Asian American and white applicants. Under this rubric, the lack of a statistically significant penalty against Asian American applicants relative to white applicants suggests that the burden Harvard’s race-conscious admissions policy places on Asian American applicants is not undue. However, Professor Arcidiacono’s analysis, and the Court’s preferred model with the personal rating excluded, imply that Asian American applicants are disadvantaged relative to white applicants. The questions in the context of this case then become: why do Asian American applicants score lower on the personal rating, does it unfairly affect their chances of admission, and if so, is this an undue burden on them when measured against Harvard’s compelling interest in diversity? It is possible that the self-selected group of Asian Americans that applied to Harvard during the years included in the data set used in this case did not possess the personal qualities that Harvard is looking for at the same rate as white applicants, just as it is possible that the selfselected white applicants tend to have somewhat weaker academic qualifications than Asian judicial determination that the admissions process meets strict scrutiny in its implementation.” Fisher I, 570 U.S. at 311. Strict scrutiny affords a plaintiff “close analysis to the evidence of how the [admission] process works in practice.” Id. at 312–13. 109 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 110 of 130 American applicants. In other words, assuming Asian American and white applicants have the same academic and extracurricular potential and the same quality of personal attributes as demographic groups, it could be that asymmetric portions of each of these groups apply to Harvard. This would explain why Asian American applicants to Harvard did better than white applicants on the academic and extracurricular ratings and why white applicants to Harvard did better on the personal rating despite the likelihood that Asian Americans are not inherently more intelligent and white applicants are not inherently more personable. This scenario has little evidentiary support, but it, like Professor Card’s model and the Court’s preferred model including the personal ratings, would result in a finding of no undue burden and a narrowly tailored process that satisfied strict scrutiny. 59 Alternatively, it may be that there is overt discrimination or implicit bias at work to the disadvantage of Asian American applicants. To begin at the end, the Court sees no evidence of discrimination in the personal ratings save for the slight numerical disparity itself. The statistical disparity is relatively minor and can be at least partially explained by a variety of factors including race-correlated inputs to the rating such as teacher and guidance counselor recommendations. Just as the Court cannot explain the variations in the academic and extracurricular ratings, it cannot definitively explain the difference in the personal ratings, but it 59 There may be little evidentiary support for this hypothesis because it was not in the interest of either party to develop this scenario. SFFA was wedded to the idea that the Asian American applicants were superior in two profiles and discriminated against on a third, while Harvard was unwilling to overtly argue that Asian American applicants were actually weaker in personal criteria, notwithstanding their stronger average academic performance and Harvard’s acknowledgment that Asian American applicants tend to be stronger in their extracurricular pursuits. The Court does not think, however, that demonstrable, disproportionate strength of a racial group in one area necessarily implies that the same racial group should be strong in all areas. If one assumes that raw talent and race are unrelated, it would be unsurprising to find that applicants that excel in one area, tend to be somewhat weaker in other areas. 110 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 111 of 130 finds that the disparity is small and reflects neither intentional discrimination against Asian American applicants nor a process that was insufficiently tailored to avoid the potential for unintended discrimination. Even if there is an unwarranted disparity in the personal ratings, the Court is unable to identify any individual applicant whose admissions decision was affected and finds that the disparity in the personal ratings did not burden Asian American applicants significantly more than Harvard’s race-conscious policies burdened white applicants. Further, there is no evidence of any discriminatory animus or conscious prejudice. To the contrary, certain statistics can be interpreted to suggest that Harvard’s admissions process unintentionally favored some subsets of Asian Americans, including the ALDCs and certain other discrete demographic groups like disadvantaged Asian females. The most likely causes of these statistical findings, however, is random variation in the admissions process or omitted variable biases, not selective discrimination that favored some Asian Americans and disfavored others. In terms of burden, it is likely that eliminating consideration of race would significantly disadvantage at least some Asian American applicants, as evidenced by the testimony of the amici at trial, all of whom viewed their race or ethnicity as a critical aspect of their life experiences and applications to Harvard. Further, it is vital that Asian Americans and other racial minorities be able to discuss their racial identities in their applications. As the Court has seen and heard, race can profoundly influence applicants’ sense of self and outward perspective. See, e.g., [Oct. 29 Tr. 30:23–33:17, 81:16–82:14, 85:24–90:3. 113:23–117:6, 140:9–148:3, 166:19–172:18, 199:18–204:9]. Removing considerations of race and ethnicity from Harvard’s admissions process entirely would deprive applicants, including Asian American applicants, of their right to advocate the value of their unique background, heritage, and perspective and would 111 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 112 of 130 likely also deprive Harvard of exceptional students who would be less likely to be admitted without a comprehensive understanding of their background. Further, throughout this trial, SFFA did not present a single admissions file that reflected any discriminatory animus, or even an application of an Asian American who it contended should have or would have been admitted absent an unfairly deflated personal rating. There thus remains the distinct possibility that a review of the available applications did not turn up a rejected Asian American applicant who was clearly more qualified than the white applicants who were admitted, or an applicant who received an obviously unjustified personal rating. This would strongly suggest that Asian American applicants were not discriminated against relative to white applicants and were therefore not unduly burdened by Harvard’s admissions program. Although the Court evaluates each of SFFA’s four counts separately below, it concludes that Harvard’s admissions program has been designed and implemented in a manner that allows every application to be reviewed in a holistic manner consistent with the guidance set forth by the Supreme Court. Further, the Court concludes that while the admissions process may be imperfect, the statistical disparities between applicants from different racial groups on which SFFA’s case rests are not the result of any racial animus or conscious prejudice and finds that Harvard’s admissions program is narrowly tailored to achieve a diverse class and the benefits that flow therefrom. E. Count II: Harvard Does Not Engage in Racial Balancing Count II alleges that Harvard engaged in impermissible racial balancing, that is, racial balancing that does not adhere to the parameters established by the Supreme Court. To maintain a permissible race-conscious admissions policy, Harvard may not “impose a fixed quota,” Fisher II, 136 S. Ct. at 2208, or otherwise “‘assure within its student body some specified percentage of a particular group merely because of its race or ethnic origin,’” as such a practice “would amount 112 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 113 of 130 to outright racial balancing, which is patently unconstitutional” under the Equal Protection Clause and therefore prohibited by Title VI. Grutter, 539 U.S. at 329–30 (quoting Bakke, 438 U.S. at 307). The requirement that colleges and universities that accept federal funds abstain from such quota systems stems from the “simple command that the Government must treat citizens as individuals, not as simply components of a racial, religious, sexual or national class.” Parents Involved in Cmty. Sch. v. Seattle Sch. Dist. No. 1, 551 U.S. 701, 730 (2007) (quoting Miller v. Johnson, 515 U.S. 900, 911 (1995)). Quota systems are impermissible because they insulate some “category of applicants with certain desired qualifications from competition with all other applicants.” Grutter, 539 U.S. at 334 (quoting Bakke, 438 U.S. at 315); see Wessmann v. Gittens, 160 F.3d 790, 798 (1st Cir. 1998) (“A single-minded focus on ethnic diversity ‘hinder[s] rather than further[s] attainment of genuine diversity.’” (quoting Bakke, 438 U.S. at 315)). Harvard’s admissions program intends to treat every applicant as an individual. Harvard does not employ a race-based quota, set aside seats for minority students, or otherwise “define diversity as ‘some specified percentage of a particular group merely because of its race or ethnic origin.’” Fisher I, 570 U.S. at 311 (quoting Bakke, 438 U.S. at 307). Every applicant competes for every seat. See [Oct. 18 Tr. 112:1–21]. Although a university could run afoul of Title VI’s prohibition on quotas even where it stopped short of defining a specific percentage and instead allowed some fluctuation around a particular number, see Parents Involved, 551 U.S. at 712 (striking down school district student allocation plan that allowed for 10% variation from the district’s overall white/nonwhite racial balance), Harvard’s admissions policy has no such target number or specified level of permissible fluctuation. As Justice Powell recognized in Bakke and as was affirmed in Grutter, “minimum goals for minority enrollment . . . [without a] specific 113 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 114 of 130 number firmly in mind” did not make Harvard’s program “the functional equivalent of a quota merely because it had some ‘plus’ for race, or gave greater ‘weight’ to race than to some other factors, in order to achieve student body diversity.” Grutter, 539 U.S. at 335 (quoting Bakke, 438 U.S. 317–318, 323). As the Court also held in Grutter: The . . . goal of attaining a critical mass of underrepresented minority students does not transform its program into a quota. As the Harvard plan described by Justice Powell recognized, there is of course “some relationship between numbers and achieving the benefits to be derived from a diverse student body, and between numbers and providing a reasonable environment for those students admitted.” Id. at 335–36 (quoting Bakke, 438 U.S. at 323). SFFA argues that its racial balancing claim is supported by non-statistical evidence, principally that Harvard’s admissions leadership too frequently looked at the “one-pagers” that showed the racial composition of admitted applicants or applicants whom Harvard was likely to admit and that Harvard placed students on its “search list” and sent recruitment letters to applicants based on criteria that disfavored Asian American applicants. The recruitment letters, however, did not affect admissions decisions, and SFFA cannot maintain a viable claim for intentional discrimination based merely on the allegation that some limited number of Asian American applicants did not receive certain pieces of marketing mail. See Weser, 190 F. Supp. 2d at 399 (holding that race-conscious recruiting efforts do “not constitute discrimination”); see also Allen v. Ala. State Bd. of Educ., 164 F.3d 1347, 1352 (11th Cir. 1999), vacated per stipulation, 216 F.3d 1263 (11th Cir. 2000) (“[W]here the government does not exclude persons from benefits based on race, but chooses to undertake outreach efforts to persons of one race, broadening the pool of applicants, but disadvantaging no one, strict scrutiny is generally inapplicable.”); Lutheran Church-Mo. Synod v. FCC, 154 F.3d 487, 492 (D.C. Cir. 1998) (noting that “broad outreach to, as opposed to the actual hiring of, a particular race” would not necessarily trigger strict scrutiny); Honadle v. Univ. of Vt. and State Agric. Coll., 56 F. Supp. 2d 114 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 115 of 130 419, 428 (D. Vt. 1999) (distinguishing “‘inclusive’ forms of affirmative action, such as recruitment and other forms of outreach” from “‘exclusive’ forms of affirmative action, such as quotas, set asides and layoffs” and holding that monitoring racial composition and encouraging recruitment of diverse candidates were not discriminatory practices subject to strict scrutiny). Even if non-receipt of an invitation to apply to Harvard could constitute discrimination, there was no evidence presented at trial that any SFFA member fell into the group of Asian American applicants who did not receive such an invitation because of their race, nor is there any evidence that they suffered an injury as a result. Further, as in Grutter, consulting the one-pagers “which keep track of the racial and ethnic composition of the class” (among other statistics) does not “sugges[t] there was no further attempt at individual review save for race itself during the final stages of the admissions process.” 539 U.S. at 336 (quotation marks omitted). Throughout the process, Harvard remains committed to its holistic evaluation and its whole person review. Harvard’s use of the onepagers as part of its admissions process and to evaluate whether it would be able to achieve its “goals for minority enrollment” is permissible and does not establish the existence of a quota or impermissible racial balancing. Id. at 335 (emphasis in original). As the Supreme Court has held, “‘[s]ome attention to numbers,’ without more, does not transform a flexible admissions system into a rigid quota.” Id. at 336 (quoting Bakke 438 U.S. at 323). 60 60 In fact, the law requires a “reasoned, principled explanation” for a decision to use race in admissions, and courts examine numerical evidence when evaluating whether race-conscious plans are narrowly tailored to serve a compelling interest. See, e.g., Fisher II, 136 S. Ct. at 2211–12 (considering “anecdotal evidence” including racial representation in enrolled classes and “more nuanced quantitative data” reflecting African American and Hispanic representation in undergraduate classes). 115 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 116 of 130 Further, it may well be necessary to give attention to numerical indicators of racial diversity when an institution elects to adopt a race-conscious admissions program so as to remain compliant with the dictates of strict scrutiny, including monitoring the ongoing need for a raceconscious admissions process and the availability of race-neutral alternatives. See Fisher II, 136 S. Ct. at 2214–15 (requiring UT Austin to “continue to use [] data to scrutinize the fairness of its admissions program; to assess whether changing demographics have undermined the need for a race-conscious policy; and to identify the effects, both positive and negative, of the affirmativeaction measures it deems necessary”). Harvard’s awareness and consideration of the number of minority students likely to enroll throughout its annual admissions cycle is consistent with the fact that there is “some relationship between numbers and achieving the benefits to be derived from a diverse student body, and between numbers and providing a reasonable environment for those students admitted.” Grutter, 539 U.S. at 336 (quoting Bakke 438 U.S. at 323). Additionally, Harvard also considers the racial distribution of its admitted students to assist it in predicting its yield rate and thereby avoid overenrolling its freshman class because students from some racial groups historically matriculate at higher rates than others. These practices do not violate Title VI. As Justice Powell did in 1978, the Court “flatly reject[s] the argument that Harvard’s program [is] ‘the functional equivalent of a quota’” system or an otherwise impermissible means of racial balancing. Id. at 335 (quoting Bakke 38 U.S. at 317–18). Accordingly, judgment for Harvard shall enter on Count II, racial balancing. F. Count III: Harvard Uses Race as a Non-Mechanical Plus Factor Count III alleges that Harvard fails to use race merely as a “plus” factor in admissions decisions. Consistent with what is required by Supreme Court precedent, Harvard has demonstrated that it uses race as a factor that can act as a “plus” or a “tip” in making admissions 116 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 117 of 130 decisions, and that its admissions program is “flexible enough to consider all pertinent elements of diversity in light of the particular qualifications of each applicant, and to place them on the same footing for consideration, although not necessarily according them the same weight.” Grutter, 539 U.S. at 334 (quoting Bakke, 438 U.S. at 317). Although race is an important consideration in deciding to admit many African American and Hispanic applicants, it remains an “individualized consideration in the context of [Harvard’s] race-conscious admissions program” and never becomes “the defining feature” of applications. Id. at 337 (citing Bakke, 438 U.S. at 318 n.52). Admissions policies that fail to use race only as a plus factor typically either employ a quota system or assign some specified value to applicants’ racial identity, and thereby use race in a rigid and mechanical manner that deprives applicants of the truly individualized consideration required by the Supreme Court. See Gratz, 539 U.S. at 270 (finding unconstitutional “the University [of Michigan]’s . . . policy, which automatically distribute[d] 20 points, or one-fifth of the points needed to guarantee admission, to every single ‘underrepresented minority’ applicant solely because of race”); Bakke, 438 U.S. at 272 (striking down quota system); Johnson v. Bd. of Regents of Univ. of Ga., 263 F.3d 1234, 1254 (11th Cir. 2001) (finding University of Georgia’s admissions policy not narrowly tailored where it employed a rigid, mechanical approach that awarded “every non-white applicant [] a 0.5 point bonus, regardless of his or her background and regardless of whether a white applicant with a far more ‘diverse’ background” was available). Although the parties’ experts have estimated the average magnitude of Harvard’s race-related tips based on past admissions decisions and the effect those tips have on the diversity of its classes, the magnitude of the tip for an individual applicant cannot be precisely determined because race is considered in a contextual manner as part of Harvard’s holistic evaluation of each 117 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 118 of 130 applicant. The estimated average magnitude of the tips and the impact of the race-related tips or plus factors on the racial composition of Harvard’s classes, however, are comparable to the size and effect of tips that have been upheld by the Supreme Court. For example, in Fisher II, the Supreme Court noted that the proportion of Hispanic and African American applicants admitted through UT Austin’s holistic review process in 2007, when race was considered, had increased 54% and 94%, respectively, relative to 2003, when the holistic review process had been race-neutral. Fisher II, 136 S. Ct. at 2212. Those figures showed that “race has had a meaningful, if still limited, effect on the diversity of the University’s freshman class.” Id. The impact of UT Austin’s holistic process is comparable to the decline in combined African American and Hispanic enrollment that Harvard would likely experience in the absence of the consideration of race, which is estimated to be approximately 45%, absent alternative measures. In addition, the Supreme Court upheld the University of Michigan Law School’s admissions program where “underrepresented minority students would have constituted 4 percent of the entering class in 2000 instead of the actual figure of 14.5 percent,” and African American applicants to the law school were “nearly guaranteed admissions if they score above 155 on the LSAT,” while “[w]hites scoring [below] 167 . . . on the LSAT [were] routinely rejected.” Grutter, 539 U.S. at 320, 377 (Thomas, J., dissenting). The plus-factor or tips that Harvard employs to achieve racial diversity for its educational mission are not nearly as large. Additionally, the magnitude of race-based tips is not disproportionate to the magnitude of other tips applicants may receive. The effect of African American and Hispanic racial identity on an applicant’s probability of admission has been estimated at a significantly lower magnitude than tips offered to recruited athletes, and is comparable to tips for legacies, applicants on the dean’s 118 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 119 of 130 or director’s interest lists, children of faculty or staff, and strengths that are reflected by Harvard’s profile ratings. Finally, the magnitude of race-based tips as indicated by the relative academic qualifications of admitted minority students at Harvard is modest. Every student Harvard admits is academically prepared for the educational challenges offered at Harvard, and a majority of admitted applicants from every major racial group scores in the 2 range on Harvard’s academic ratings. [PX623]. 61 In other words, most Harvard students from every racial group have a roughly similar level of academic potential, although the average SAT scores and high school grades of admitted applicants from each racial group differ significantly. Accordingly, judgment for Harvard shall enter on Count III, using race as a nonmechanical plus factor. G. Count V: No Adequate, Workable, and Sufficient Fully Race-Neutral Alternatives Are Available Count V alleges that Harvard, in constructing an admissions process that considers race to ensure a diverse class, failed to consider and adopt race-neutral alternatives that would allow it to achieve diversity. Strict scrutiny requires that the Court “verify that it is ‘necessary’ for a university to use race to achieve the educational benefits of diversity.” Fisher I, 570 U.S. at 312 (quoting Bakke, 438 U.S. at 305). “This involves a careful judicial inquiry into whether a university could achieve sufficient diversity without using racial classifications. Although ‘[n]arrow tailoring does not require exhaustion of every conceivable race-neutral alternative,’” id. (quoting Grutter, 539 U.S. at 339–40), or choosing “between maintaining a reputation for excellence or fulfilling a commitment to provide educational opportunities to members of all 61 An academic rating of 2 indicates magna cum laude potential, superb grades, and mid- to high700 SAT scores or a score above 33 on the ACT. See supra at Section III.B.3.ii. The “2 range” includes applicants who were assigned a “2+” or “2-.” 119 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 120 of 130 racial groups,” Grutter, 539 U.S. at 339, “strict scrutiny does require a court to examine with care, and not defer to, a university’s ‘serious, good faith consideration of workable race-neutral alternatives,’” Fisher I, 570 U.S. at 312 (quoting Grutter, 539 U.S. at 339–340). “Consideration by the university is of course necessary, but it is not sufficient to satisfy strict scrutiny: The reviewing court must ultimately be satisfied that no workable race-neutral alternatives would produce the educational benefits of diversity.” Id. If “‘a nonracial approach . . . could promote the substantial interest about as well and at tolerable administrative expense,’ . . . then the university may not consider race.” Id. (citation omitted). In considering the proffered raceneutral alternatives, the Court is mindful of Justice Ginsburg’s astute observation that “only an ostrich could regard the supposedly neutral alternatives as race unconscious.” Id. at 335 (Ginsburg, J., dissenting). Here, as more fully discussed in Section VI, Harvard has demonstrated “that ‘race-neutral alternatives’ that are both ‘available’ and ‘workable’ ‘do not suffice.’” Fisher II, 136 S. Ct. at 2208. In sum, eliminating early action and tips for ALDCs, increasing outreach and community partnerships, offering more financial aid, or admitting more transfer students are all “available” and “workable” in some form and at varying costs, but they would likely have no meaningful impact on racial diversity. Further, any minimal effect that these alternative admissions practices might have on racial diversity, if implemented individually or in combination, would be offset by the decline in African American and Hispanic students that would result if race-conscious admissions practices were eliminated. Several other conceivable alternatives, such as admitting only students who rank at top of their high school class after their junior year or admitting the top student from each zip code, are not workable for Harvard because such programs would vastly over enroll its class. See supra at Section III.A.2; see also Fisher II, 136 S. Ct. at 2213 (“Class 120 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 121 of 130 rank is a single metric, and like any single metric, it will capture certain types of people and miss others. . . . [P]rivileging one characteristic above all others does not lead to a diverse student body.”). SFFA’s expert, Mr. Kahlenberg, proposes a geographic-based quota system using “neighborhood clusters” that is seemingly designed to achieve racial diversity based on socioeconomics rather than attention to race. This proposal has some of the earmarks of impermissible racial balancing, albeit without an explicit, articulated reliance on race. Further, it poses significant logistical challenges, such as how to form the clusters, and how to account for wealthy households in a generally lower income cluster, as well as difficult institutional and philosophical questions such as whether economics can fairly be considered a proxy for race. These issues aside, although Harvard could theoretically impose some form of geographic, place-based quota system, it could not achieve comparable racial diversity through such a program without a significant decline in the academic strength of its class. Further, the legality of the proposed place-based quota system is uncertain. In Fisher II, the Supreme Court upheld the constitutionality of UT Austin’s holistic review program but did not speak to the overall permissibility of place-based admissions policies. 136 S. Ct. at 2213–15. Unlike Harvard’s holistic process which considers every applicant individually, UT Austin admitted most of its class by automatically admitting applicants who graduated in the top 10% of their Texas high school class pursuant to a state law requiring it to admit those students. Id. at 2209. The plaintiff advocated the expansion of the automatic admission percentage, claiming it to be a race-neutral way of increasing diversity. Id. at 2213. The Supreme Court refused to require the expansion of the program, stating, “‘It is race consciousness, not blindness to race, that drives such plans.’ Consequently, petitioner cannot assert simply that increasing the University’s 121 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 122 of 130 reliance on a percentage plan would make its admissions policy more race neutral.” Id. at 2213 (citation omitted) (quoting Fisher I, 570 U.S. at 335–36 (Ginsburg, J., dissenting)). Here, just as in Fisher II, the Court is not persuaded that such a plan would actually be “more race neutral,” id. at 2213. Place-based plans therefore do not suffice, pose complex challenges, and may not even qualify as available race-neutral alternatives. Harvard could adopt a more significant tip for economically disadvantaged students, but every such proposal presented to the Court would result in a significant decline in African American representation. Achieving even roughly comparable levels of combined African American and Hispanic representation to those Harvard presently achieves would require Harvard to sacrifice the academic strength of its class and forgo other admissions policies from which it derives financial, reputational, and academic benefits. See supra at Section III.B.3. As such, Harvard would compromise some degree of its reputation for academic excellence and still be less diverse than it is currently. Title VI does not require such an outcome. See Fisher II, 136 S. Ct. at 2213 (explaining that the Supreme “Court’s precedent [makes] clear that the Equal Protection Clause does not force universities to choose between a diverse student body and a reputation for academic excellence”). Harvard has demonstrated that no workable and available race-neutral alternatives would allow it to achieve a diverse student body while still maintaining its standards for academic excellence. Judgment shall therefore enter in Harvard’s favor on Count IV, race-neutral alternatives. H. Count I: Harvard Does Not Intentionally Discriminate SFFA’s intentional discrimination claim, Count I, requires the Court to determine whether Harvard’s admissions program violates Title VI through intentional discrimination against Asian Americans notwithstanding the Court’s conclusion that Harvard has shown that its 122 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 123 of 130 admissions program serves its compelling interest in diversity, that some racial categorizations are necessary to serve that interest, that it does not engage in proscribed racial balancing, and that no workable and available, fully race-neutral alternatives would suffice to meet Harvard’s goals. SFFA is not claiming that Harvard excludes Asian Americans and in fact, Asian Americans are admitted at virtually the same rate as white applicants. What it does claim is that, based solely on the quantifiable aspects of admissions, Asian Americans should be admitted at an even higher rate and that, if the personal ratings were not depressed, there would be more Asian Americans admitted. In undertaking its analysis, the Court begins with certain fundamentals. First, “given the important purposes of public education and the expansive freedoms of speech and thought associated with the university environment, universities occupy a special niche in our constitutional tradition.” Grutter 539 U.S. at 328–29. Second, a university is free to “make its own judgments as to . . . the selection of its student body.” Id. at 329 (quoting Bakke, 438 U.S. at 312). And third, although deference is owed to a university’s decision to pursue the educational benefits that flow from diversity, the university must show that its use of race is narrowly tailored to achieve its permissible goals. Fisher II, 136 S. Ct. at 2208. To these, the Court reiterates the following findings specific to this case: 1. Throughout this trial and after a careful review of all exhibits and written submissions, there is no evidence of any racial animus whatsoever or intentional discrimination on the part of Harvard beyond its use of a race conscious admissions policy, nor is there 123 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 124 of 130 evidence that any particular admissions decision was negatively affected by Asian American identity. 62 2. A race-conscious admissions program allows Harvard to achieve a level of robust diversity that would not otherwise be possible, at least at this time. 3. The Court firmly believes that Asian Americans are not inherently less personable than any other demographic group, just as it believes that Asian Americans are not more intelligent or more gifted in extracurricular pursuits than any other group. 4. There is a statistical difference in the personal ratings with white applicants faring better that Asian American applicants. Asian American applicants, however, do better on the extracurricular and academic ratings than their white counterparts. All three ratings incorporate subjective and objective elements, and while implicit biases may be affecting 62 The Court notes that under the Title VI standard applicable outside the higher education admissions context, SFFA’s intentional discrimination claim would fail because SFFA has not shown, by a preponderance of the evidence, that (1) Harvard discriminated on the basis of race, (2) that the discrimination was intentional, and (3) that the discrimination was a substantial or motivating factor for admissions decisions. See 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)). The requirement for a “substantial or motivating factor” requires “evidence of racial animus,” id. at 43, and no racial animus was present here. Further, under the standard articulated in Goodman v. Bowdoin College, 380 F.3d 33 (1st Cir. 2004), the Court would enter judgment for Harvard because it has shown that its admissions program was employed to promote diversity, which is not an invidious discriminatory purpose. See supra at Section III.D. Admissions decisions are made only after a careful process that considers and appreciates the diversity that applicants from diverse racial backgrounds, including Asian Americans, provide at Harvard. Harvard’s only intentional consideration of race views increased racial diversity as a positive attribute of its admitted class, which it achieves by considering an individual’s race through an individualized, holistic evaluation of every applicant in the manner envisioned by the Supreme Court. Further, the Court feels confident stating that the statistical disparities in personal ratings and admissions probabilities that have been identified are the result of some external race-correlated factors and perhaps some slight implicit biases among some admissions officers that, while regrettable, cannot be completely eliminated in a process that must rely on judgments about individuals. 124 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 125 of 130 Harvard’s ratings at the margins, to the extent that the disparities are the result of race, they are unintentional and would not be cured by a judicial dictate that Harvard abandon considerations of race in its admission process. 5. Harvard’s admissions program is conceptually narrowly tailored to meet its interest in diversity. In practice, as more fully discussed above, it does not seem to unduly burden Asian Americans despite the fact that some percentage of Asian American applicants have received lower personal ratings than white applicants who seem similarly situated. The reason for these lower scores is unclear, but they are not the result of intentional discrimination. They might be the result of qualitative factors that are harder to quantify, such as teacher and guidance counselor recommendations, or they may reflect some implicit biases. Race conscious admissions will always penalize to some extent the groups that are not being advantaged by the process, but this is justified by the compelling interest in diversity and all the benefits that flow from a diverse college population. Here, any relative burden on Asian Americans (and it is not clear that there is a disproportionate burden) is not enough to warrant a finding that Harvard’s admissions process fails to survive strict scrutiny or to require it to move to an admissions model that foregoes diversity in favor of parity based solely on quantifiable metrics. 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 ratings in particular was consistent, unambiguous, and convincing. Not one of them had seen or heard anything disparaging about an Asian American applicant despite the fact that decisions were made collectively and after open discussion about each applicant in the docket and full committee meetings. Similarly, there is no credible evidence that corroborates the improper 125 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 126 of 130 discrimination suggested by Professor Arcidiacono’s statistical model. Asian American applicants are accepted at the same rate as other applicants and now make up more than 20% of Harvard’s admitted classes, up from 3.4% in 1980. Although Asian Americans can and do bring important and diverse perspectives to Harvard, because only about 6% of the United States population is Asian American compared to nearly a quarter of Harvard’s class, it is reasonable for Harvard to determine that students from other minority backgrounds are more likely to offer perspectives that are less abundant in its classes and to therefore primarily offer race-based tips to those students. Finally, SFFA did not present a single Asian American applicant who was overtly discriminated against or who was better qualified than an admitted white applicant when considering the full range of factors that Harvard values in its admissions process. The statistics themselves are alone not enough to cause the Court to conclude that Harvard has engaged in improper intentional discrimination where Harvard has shown that its admissions policy uses race only in a permissible and narrowly tailored way. Further, although Professor Arcidiacono’s statistics suggest discrimination against certain subsets of Asian American applicants, Professor Card’s analysis of this same data suggests the opposite, thereby leaving the statistical analyses inconclusive. Even assuming that there is a statistically significant difference between how Asian American and white applicants score on the personal rating, the data does not clearly say what accounts for that difference. In other words, although the statistics perhaps tell “what,” they do not tell “why,” and here the “why” is critically important. Further, by its very nature, the personal score includes, and should include, aspects of an applicant and his or her application that are not easily quantifiable and therefore cannot be fully captured by the statistical data. 126 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 127 of 130 Harvard’s admissions process survives strict scrutiny. It serves a compelling, permissible and substantial interest, and it is necessary and narrowly tailored to achieve diversity and the academic benefits that flow from diversity. Consistent with the hallmarks of a narrowly tailored program, applicants are afforded a holistic, individualized review, diversity is understood to embrace a broad range of qualities and experiences, and race is used as a plus factor, in a flexible, non-mechanical way. See Fisher, 136 S. Ct. at 2214; Grutter, 539 U.S. at 337–38. The Admissions program also satisfies the other principles articulated in Fisher II in that it does not have a quota or use a fixed percentage and all applicants compete for all available seats. Further, Harvard has met its burden of showing that there are not currently any available or workable race-neutral alternatives. Finally, there is nothing about Harvard’s admissions process that is at odds with the reason for subjecting racial classifications to strict scrutiny—to ensure “little or no possibility that the motive for the classification was illegitimate racial prejudice or stereotype.” J.A. Croson Co., 488 U.S. at 493. The use of race benefits certain racial and ethnic groups that would otherwise be underrepresented at Harvard and is therefore neither an illegitimate use of race or reflective of racial prejudice. Accordingly, judgment for Harvard shall enter on Count I, intentional discrimination. VIII. CONCLUSION Notwithstanding the fact that Harvard’s admissions program survives strict scrutiny, it is not perfect. The process would likely benefit from conducting implicit bias trainings for admissions officers, maintaining clear guidelines on the use of race in the admissions process, which were developed during this litigation, and monitoring and making admissions officers aware of any significant race-related statistical disparities in the rating process. That being said, the Court will not dismantle a very fine admissions program that passes constitutional muster, solely because it could do better. There is always the specter of perfection, but strict scrutiny 127 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 128 of 130 does not require it and a few identified imperfections, after years of litigating and sifting through applications and metrics, do not alone require a finding that Harvard’s admissions program is not narrowly tailored. Further, the Court emphatically repeats what the Supreme Court said in Fisher II: The University now has at its disposal valuable data about the manner in which different approaches to admissions may foster diversity or instead dilute it. The University must continue to use this data to scrutinize the fairness of its admissions program; to assess whether changing demographics have undermined the need for a race-conscious policy; and to identify the effects, both positive and negative, of the affirmative-action measures it deems necessary. The Court’s affirmance of the University’s admissions policy today does not necessarily mean the University may rely on that same policy without refinement. It is the University’s ongoing obligation to engage in constant deliberation and continued reflection regarding its admissions policies. 136 S. Ct. at 2213–15. The Court here stops well short of requiring an admissions process that is overly data driven. Using statistics to ensure that the distribution of profile ratings or any other measure is exactly even among various groups would potentially run afoul of the prohibition on quotas and, more importantly, defeat the purpose of a comprehensive, holistic review process that allows the admission of applicants with virtues that are not always quantifiable. But now that Harvard and other schools can see how statistical analyses can reveal perhaps otherwise imperceptible statistical anomalies, these sorts of statistics should be used as a check on the process and as a way to recognize when implicit bias might be affecting outcomes. It was always intended that affirmative action programs be limited in duration. In 2003, the Supreme Court articulated its expectation that in twenty-five years, it would not be necessary to use racial preferences to achieve a diverse student body. Grutter, 539 U.S. at 343. As time marches on and the effects of entrenched racism and unequal opportunity remain obvious, this 128 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 129 of 130 goal might be optimistic and may need to change, but it remains imperative that Harvard and other schools that make use of racial preferences to achieve a diverse learning environment ensure, through data and experience, that “race plays no greater role than is necessary to meet its compelling interest” in diversity and to keep in mind that “racial classifications may sometimes fail to capture diversity in all of its dimensions.” Fisher II, 136 S. Ct. at 2210. The wise and esteemed author Toni Morrison observed, “Race is the least reliable information you can have about someone. It’s real information, but it tells you next to nothing.” Emily Langer, From heart of black America, a voice for the voiceless, Boston Globe, Aug. 7, 2019, at C11 (quoting Paul Gray, Books: Paradise Found, Time (Jan. 19, 1998), http://content.time.com/time/subscriber/article/0,33009,987690-5,00.html). Although this has been said, it must become accepted and understood before we close the curtain on race conscious admissions policies. The rich diversity at Harvard and other colleges and universities and the benefits that flow from that diversity will foster the tolerance, acceptance and understanding that will ultimately make race conscious admissions obsolete. As President Ruth Simmons said from the witness stand in this case when asked about the importance of diversity: It’s very hard for me to overstate my conviction about the benefits that flow to all of these areas from a diverse undergraduate student body. I know something about the lack of diversity in one’s education. . . . My father was a janitor, my mother was a maid. They had been sharecroppers, they had few opportunities. I lived through that. I remember it. So to me, the benefits that flow to students is they get a better education, a deeper education, a truer education to deal with what they're going to have to deal with in life. To the institution, it makes for not just an enhanced learning environment but for the opportunity to be unparalleled in their standing because they offer something that is so indispensable for society. And for society, my goodness, I've spoken about the conflicts in society, how deeply they run, how they resurface from time to time. How can we imagine a world in which we are not creating leaders and citizens who have the capacity to 129 Case 1:14-cv-14176-ADB Document 672 Filed 09/30/19 Page 130 of 130 mediate those differences? I cannot imagine it. And so it’s with great conviction that I say that we must continue to offer diverse undergraduate education to our young people to save our nation. [Oct. 30 Tr. 54:11–55:15]. That eloquent testimony captures what is important about diversity in education. For purposes of this case, at least for now, ensuring diversity at Harvard relies, in part, on race conscious admissions. Harvard’s admission program passes constitutional muster in that it satisfies the dictates of strict scrutiny. The students who are admitted to Harvard and choose to attend will live and learn surrounded by all sorts of people, with all sorts of experiences, beliefs and talents. They will have the opportunity to know and understand one another beyond race, as whole individuals with unique histories and experiences. It is this, at Harvard and elsewhere that will move us, one day, to the point where we see that race is a fact, but not the defining fact and not the fact that tells us what is important, but we are not there yet. Until we are, race conscious admissions programs that survive strict scrutiny will have an important place in society and help ensure that colleges and universities can offer a diverse atmosphere that fosters learning, improves scholarship, and encourages mutual respect and understanding. SO ORDERED. September 30, 2019 /s/ Allison D. Burroughs ALLISON D. BURROUGHS U.S. DISTRICT JUDGE 130