Opinion
1:14CV954
2021-10-18
STUDENTS FOR FAIR ADMISSIONS, INC., Plaintiff, v. UNIVERSITY OF NORTH CAROLINA, et al., Defendants.
Andrew Allen Freeman, Daniel Alan M. Ruley, Bell Davis & Pitt, P.A., Winston-Salem, NC, Thomas R. McCarthy, Bryan K. Weir, James F. Hasson, John Michael Connolly, William S. Consovoy, Consovoy McCarthy PLLC, Arlington, VA, Patrick Strawbridge, Consovoy McCarthy Park PLLC, Boston, MA, for Plaintiff. Patrick J. Fitzgerald, Amy L. Van Gelder, Marianne H. Combs, Skadden Arps Slate Meagher & Flom, LLP, Chicago, IL, Stephanie A. Brennan, Tamika Henderson, N.C. Department of Justice, Raleigh, NC, Lara A. Flath, Skadden Arps Slate Meagher & Flom, LLP, New York, NY, for Defendants University of North Carolina, University of North Carolina Board of Governors, John C. Fennebresque, W. Louis Bissette, Jr., Joan Templeton Perry, Roger Aiken, Hannah D. Gage, Ann B. Goodnight, H. Frank Frainger, Peter D. Hans, Thomas J. Harrelson, Henry W. Hinton, James L. Holmes, Jr., Rodney E. Hood, W. Marty Kotis, III, G. Leroy Lail, Scott Lampe, Steven B. Long, Joan G. MacNeill, Mary Ann Maxwell, W. Edwin McMahan, W.G. Champion Mitchell, Hari H. Math, Anna Spangler Nelson, Alex Parker, R. Doyle Parrish, Therence O. Pickett, David M. Powers, Robert S. Rippy, Harry Leo Smith, Jr., J. Craig Souza, George A. Sywassink, Richard F. Taylor, Raiford Trask, III, Phillip D. Walker, Laura I. Wiley, Thomas W. Ross, University of North Carolina at Chapel Hill, Carol L. Folt, James W. Dean, Jr., Stephen M. Farmer.
Andrew Allen Freeman, Daniel Alan M. Ruley, Bell Davis & Pitt, P.A., Winston-Salem, NC, Thomas R. McCarthy, Bryan K. Weir, James F. Hasson, John Michael Connolly, William S. Consovoy, Consovoy McCarthy PLLC, Arlington, VA, Patrick Strawbridge, Consovoy McCarthy Park PLLC, Boston, MA, for Plaintiff.
Patrick J. Fitzgerald, Amy L. Van Gelder, Marianne H. Combs, Skadden Arps Slate Meagher & Flom, LLP, Chicago, IL, Stephanie A. Brennan, Tamika Henderson, N.C. Department of Justice, Raleigh, NC, Lara A. Flath, Skadden Arps Slate Meagher & Flom, LLP, New York, NY, for Defendants University of North Carolina, University of North Carolina Board of Governors, John C. Fennebresque, W. Louis Bissette, Jr., Joan Templeton Perry, Roger Aiken, Hannah D. Gage, Ann B. Goodnight, H. Frank Frainger, Peter D. Hans, Thomas J. Harrelson, Henry W. Hinton, James L. Holmes, Jr., Rodney E. Hood, W. Marty Kotis, III, G. Leroy Lail, Scott Lampe, Steven B. Long, Joan G. MacNeill, Mary Ann Maxwell, W. Edwin McMahan, W.G. Champion Mitchell, Hari H. Math, Anna Spangler Nelson, Alex Parker, R. Doyle Parrish, Therence O. Pickett, David M. Powers, Robert S. Rippy, Harry Leo Smith, Jr., J. Craig Souza, George A. Sywassink, Richard F. Taylor, Raiford Trask, III, Phillip D. Walker, Laura I. Wiley, Thomas W. Ross, University of North Carolina at Chapel Hill, Carol L. Folt, James W. Dean, Jr., Stephen M. Farmer.
TRIAL FINDINGS OF FACT AND CONCLUSIONS OF LAW
LORETTA C. BIGGS, District Judge.
TABLE OF CONTENTS
I. INTRODUCTION...585
II. PARTIES AND PROCEDURAL HISTORY...586
A. Parties...586
B. Procedural History...587
III. FINDINGS OF FACT: UNC'S INTEREST IN THE EDUCATIONAL BENEFITS OF DIVERSITY IS SUBSTANTIAL AND LONG-STANDING...588
A. UNC Has Made the Deliberate Decision to Pursue the Educational Benefits of Diversity...588
B. UNC Has Offered a Principled, Reasoned Explanation for Its Decision...589
C. The Educational Benefits of Diversity that the University Seeks are Sufficiently Measurable to Permit Judicial Scrutiny...591
D. The Educational Benefits that the University seeks are being Experienced...592
E. UNC's Efforts in Pursuing the Educational Benefits of Diversity While Ongoing, Are Unfinished...593
IV. FINDINGS OF FACT: UNC'S ADMISSIONS PROCEDURES MANDATE A HOLISTIC EVALUATION OF CANDIDATES...594
A. Overview of UNC's Admissions Process...595
B. Application Requirements...595 C. Application Readers...596
D. Application Evaluation Process...598
1. Admissions Procedures...598
2. Admissions Criteria...599
V. FINDINGS OF FACT: NON-STATISTICAL EVIDENCE DOES NOT DEMONSTRATE DISCRIMINATION...601
A. There is No Evidence that UNC Conceals the Improper Use of Race Behind Opaque Procedures...602
1. UNC's Intent...602
a. Provisional Decision Stage...602
b. SGR Process...603
2. UNC's Multiple Checks, Balances, and Quality Controls...604
B. There is No Non-Statistical Evidence that Race is a Predominant Factor in Admissions...605
1. UNC Uses Race as One of Multiple Factors in Recruiting Prospective Students...605
2. The Reporting of Race Does Not Suggest Its Improper Use...606
3. The Consideration of Race at all Stages of the Admissions Process is Consistent with How UNC Evaluates All Factors...607
C. There is No Evidence that UNC Has Not Taken a Serious Accounting of Its Use of Race...607
1. There is Evidence of Attention to the Use of Race Throughout the Admissions Process...608
2. There is Evidence that UNC Has Defined and Assessed the Concept of Critical Mass...608
a. UNC Has Defined Critical Mass by Reference to the Educational Benefits of Diversity Critical Mass is Designed to Produce...608
b. UNC Has Conducted Periodic Reviews to Reach the Conclusion that it Has Not Attained Critical Mass...610
VI. FINDINGS OF FACT: STATISTICAL EVIDENCE DOES NOT DEMONSTRATE DISCRIMINATION...612
A. The Parties’ Experts are Highly Qualified...612
B. Overview of Research Questions and Process...613
1. Professor Arcidiacono...613
2. Professor Hoxby...615
C. Descriptive Statistics...615
1. Basic Findings...615
2. Deciles Analysis...617
D. Econometric Modeling...619
1. Professor Arcidiacono's Process...620
2. Accuracy and Fit of Professor Arcidiacono's Preferred Model...621
a. Professor Arcidiacono's Model Predicts Admissions Decisions with a High Degree of Accuracy...621
b. Professor Arcidiacono's Model is Overfit...622
c. There is No Evidence that UNC Has Implemented a Formulaic Process that Precludes Individualized Consideration of Each Candidate...623
3. Professor Arcidiacono Computes Some Variables in Ways that Undermine His Conclusions...623
a. Imputation of GPA and Standardized Test Scores...623
b. Reader Ratings...625
c. Shapley Decomposition...626
4. Professor Arcidiacono's Quantitative Analysis Fails to Show that Race is a Predominant Factor...627
a. Transformational Analysis...627
b. Average Marginal Effect...629
c. Admitted URM Analysis...630
d. Capacity Constraints...631
5. Professor Hoxby's Regression Model Demonstrates that Race is Not a Predominant Factor in UNC's Admissions Process...631
a. Explanatory Power of Professor Hoxby's Model...632
b. Magnitude of Racial Preferences...633
6. Conclusions...634
VII. FINDINGS OF FACT: UNC HAS MADE A SERIOUS, GOOD FAITH EFFORT TO CONSIDER RACE-NEUTRAL ALTERNATIVES...634
A. The University Has Long Considered Race-Neutral Alternatives...635
B. UNC Has Augmented Its Recruitment and Financial Aid Programs...636
1. Student Recruitment...636
2. Financial Aid...637
C. UNC Has Increased Diversity Through Transfer Students...638
D. Eliminating the Early Action Deadline and Legacy Preferences Would Have a De Minimis Effect...639
E. Changing the Admissions Evaluation Process...640
1. Professor Hoxby's Process...640
2. Socioeconomic Simulations...643
3. Top X Percent Simulations...645
4. Geography-Based Simulations and Additional Concepts...647
F. Conclusions...648
VIII. CONCLUSIONS OF LAW...648
A. Overview...648
B. Controlling Supreme Court Precedent Permits Race-Conscious Admissions...649
1. Regents of University of California v. Bakke , 438 U.S. 265, 98 S.Ct. 2733, 57 L.Ed.2d 750 (1978)...649
2. Gratz v. Bollinger , 539 U.S. 244, 123 S.Ct. 2411, 156 L.Ed.2d 257 (2003) (" Gratz "); and Grutter v. Bollinger , 539 U.S. 306, 123 S.Ct. 2325, 156 L.Ed.2d 304 (2003) (" Grutter ")...650
3. Fisher v. Univ. of Tex., 570 U.S. 297, 133 S.Ct. 2411, 186 L.Ed.2d 474 (2013) (" Fisher I "); and Fisher v. Univ. of Tex. , 579 U.S. 365, 136 S. Ct. 2198, 195 L.Ed.2d 511 (2016) (" Fisher II ")...652
C. The University's Undergraduate Admissions Program Withstands Strict Scrutiny...655
1. Conclusions of Law: The University Has a Compelling Interest in Pursuing The Educational Benefits of Diversity...655
2. Conclusions of Law (Count I): UNC Has Demonstrated that Its Use of Race in Admissions Is Narrowly Tailored...658
a. SFFA's Non-Statistical Evidence Does Not Support a Conclusion that UNC Has Engaged in Discrimination...659
b. SFFA's Statistical Evidence Does Not Support a Conclusion that UNC Has Engaged in Discrimination...660
3. Conclusions of Law (Count II): UNC Has Demonstrated that a Nonracial Approach Would Not Promote the Interests of Diversity About as Well as Its Race-Conscious Process...662
CONCLUSION...666
ORDER...667
I. INTRODUCTION
Plaintiff, Students for Fair Admissions, Inc. ("SFFA"), initiated this action on November 7, 2014, against the named University of North Carolina Defendants ("UNC Defendants"), alleging that the use of race in its undergraduate admissions process at the University of North Carolina at Chapel Hill ("the University" or "UNC") violates the Equal Protection Clause of the Fourteenth Amendment to the United States Constitution, 42 U.S.C. §§ 1981, 1983 (" Sections 1981 and 1983"), and Title VI of the Civil Rights Act of 1964, 42 U.S.C. § 2000d et seq. ("Title VI"). (ECF No. 1 at 2 (Compl.).) Specifically, SFFA alleges that the University "has intentionally discriminated against certain of [its] members on the basis of their race, color, or ethnicity in violation of the Fourteenth Amendment and [federal law]" by: (1) "employing an undergraduate admissions policy that does not merely use race as a ‘plus’ factor in admissions decisions in order to achieve student body diversity"; (2) "employing racial preferences in undergraduate admissions where there are available race-neutral alternatives capable of achieving student body diversity"; and (3) "employing an undergraduate admissions policy that uses race as a factor in admissions." (Id. ¶¶ 198, 205, 215.) SFFA seeks declaratory and injunctive relief as well as attorneys’ fees and costs. (Id. at 64.)
The UNC Defendants counter that the admissions process at the University is constitutionally permissible under current law in that it withstands strict scrutiny. First, they argue that "the University's compelling interest in the educational benefits that flow from a diverse student body" is clear and uncontested. (ECF No. 245 at 8.) Second, they argue that "consistent with Supreme Court guidance, the University engages in an individualized, holistic review of each application for admission, considering race flexibly, as only one factor" in the evaluation process. (Id. at 10.) Third and finally, UNC Defendants argue that it has demonstrated a serious, good-faith consideration of race-neutral alternatives but has "found none that would allow it to achieve its compelling interest about as well and at tolerable administrative expense." (Id. at 11.) In this trial UNC Defendants seek judgment in their favor on Counts I and II of the Complaint. (ECF No. 30 at 115 (Answer).) Count III of Plaintiff's Complaint was resolved by prior Order of this Court. (ECF No. 210); see Section II.B.7, infra.
II. PARTIES AND PROCEDURAL HISTORY
A. Parties
Plaintiff Students for Fair Admissions, Inc. is a voluntary membership association with more than 20,000 members, including applicants and prospective applicants to institutions of higher education. (ECF No. 225 ¶ 1 (Joint Statement of Undisputed Facts ("J. Statement")).) It is a non-profit, 501(c)(3) organization whose stated mission is "to defend human and civil rights secured by law, including the right of individuals to equal protection under the law, through litigation and any other lawful means." (PX106; ECF No. 225 ¶ 1 (J. Statement).)
The UNC Defendants are all entities or individuals that are a part of the Defendant University of North Carolina System (the "UNC System"). (See ECF No. 225 (J. Statement).) The University was founded in 1789 as the nation's first public college and is a constituent institution of the UNC System. (Id. ¶¶ 2, 4.) The UNC System's Board of Governors, also a Defendant in this case along with its individual members, has responsibility for the planning, development, and overall governance of the UNC System pursuant to Chapter 116 of the North Carolina General Statutes. (Id. ¶ 6.) Defendant Peter Hans is President of the UNC System and is chief administrative and executive officer. (Id. ¶ 7.) Defendant Kevin Guskiewicz, UNC's Chancellor, is responsible for carrying out the policies of the Board of Governors and the Board of Trustees, and he is vested with executive authority over the University, subject to the direction of Defendant Hans. (Id. ¶ 11.) Robert Blouin, the University's Executive Vice Chancellor and Provost, and Stephen Farmer, UNC's Vice Provost for Enrollment and Undergraduate Admissions, are also Defendants in this case. (Id. ¶¶ 16–17.) These Defendants are referred to collectively as "UNC Defendants."
In addition to the University, the UNC System is composed of sixteen additional constituent institutions. (ECF No. 1 ¶ 25 (Compl.).) Only the University's admissions process is at issue in this case.
Defendant-Intervenors (or "Student-Intervenors") are a racially diverse group of underrepresented students of color who applied, attended, and/or recently graduated from UNC. (See ECF Nos. 79 at 4; 82.) This Court granted Student-Intervenors permissive intervention to present evidence on two issues: (1) the effect of UNC's existing, and SFFA's proposed, admissions processes on the critical mass of underrepresented students at the school; and (2) the history of segregation and discrimination at UNC and in North Carolina. (ECF No. 79 at 14.)
For purposes of this case, "students of color" refers to students who identify with historically underrepresented and marginalized racial and ethnic groups, including Black, Hispanic, and Native American students. Student-Intervenors use the following terms interchangeably: "Black" and "African American"; "Native American" and "Indigenous"; and "Hispanic" and "Latino" or "Latina." Student-Intervenors use the term "white" to mean white/Non-Hispanic. (ECF No. 246 at 6.)
B. Procedural History
1. On March 20, 2015, the parties filed a Joint Stipulation By All Parties to Dismiss Various Parties and Claims to include dismissals of, among other things: (1) all claims pursuant to Section 1981 stated in Counts I, II, and III of Plaintiff's Complaint against all Defendants; (2) all claims pursuant to Section 1983 stated in Counts I, II, and III of Plaintiff's Complaint against the University of North Carolina, the University of North Carolina Board of Governors, and the University of North Carolina-Chapel Hill; (3) all claims pursuant to Title VI against certain named individual Defendants; and (4) all claims against the UNC Board of Trustees and its individual members. (ECF Nos. 29 at 1–5; 30 at 2.)
2. On March 24, 2015, Defendants filed their Answer denying any liability and asserting a number of affirmative defenses. (ECF No. 30.) In addition to seeking judgment in their favor on Counts I and II of the Complaint, UNC Defendants seek attorneys’ fees and costs. (Id. at 115.)
3. On June 30, 2015, a Motion to Intervene in Defense of UNC was filed by those hereinafter identified as Student Intervenors. (ECF No. 39.) The Court, on January 13, 2017, granted the motion insofar as it allowed permissive intervention under Rule 24(b) for the limited presentation of evidence as outlined above. (ECF No. 79); see Section II.A, supra.
4. In the interim between the filing of this motion and the Court's Order, Defendants sought a stay of the proceedings, (ECF No. 46), pending a decision by the United States Supreme Court in Fisher v. University of Texas at Austin, 579 U.S. 365, 136 S. Ct. 2198, 195 L.Ed.2d 511 (2016) (" Fisher II "). Plaintiff initially opposed the stay; however, on September 30, 2015, Plaintiff and Defendants agreed jointly to seek a partial stay of the proceedings pending the resolution of Fisher II on the condition that Defendant make production of certain agreed-upon materials. (ECF No. 64.) The Court, on October 1, 2015, granted the parties’ joint motion for partial stay and further ordered that Defendants produce stipulated documents to Plaintiff in the interim. (ECF No. 65.)
5. On October 25, 2017, the UNC Defendants filed a Motion to Dismiss pursuant to Fed. R. Civ. P. 12(b)(1) alleging a lack of standing. (ECF No. 106.) On September 29, 2018, the Court denied this motion, concluding that SFFA satisfied the requirements necessary for associational standing to sue on behalf of its members. (ECF No. 150.)
6. On January 18, 2019, the UNC Defendants filed its Motion for Summary Judgment, (ECF No. 152), and on the same date SFFA filed a cross Motion for Summary Judgment, (ECF No. 158). On September 30, 2019, the Court denied each parties’ motion. (ECF No. 190.)
7. On May 8, 2020, the UNC Defendants filed an Unopposed Motion for Partial Judgment on the Pleadings as to only Count III of Plaintiff's Complaint. (ECF No. 209.) UNC asked the Court to enter judgment on Count III of Plaintiff's Complaint, (id. at 1), in which Plaintiff contended that "the use of racial preferences" in UNC's admissions process "should be forbidden" outright, (ECF No. 1 ¶ 217 (Compl.)). Plaintiff argued that Supreme Court precedents allowing race-conscious admissions "were wrongly decided at the time they were issued" and should be overruled. (Id. ¶ 216.) On May 28, 2020, the Court found that there was no genuine issue of material fact as to whether UNC used race in its admissions process and, "considering existing Supreme Court precedent related to that claim," granted the unopposed motion in the UNC Defendants’ favor. (ECF No. 210 at 2–3.) This Order resolved Count III while "preserving Plaintiff's right to appeal" the ruling. (Id. at 1.)
8. The trial on the remaining Counts, I and II of Plaintiff's Complaint, began on November 9, 2020, and lasted for eight days. The parties expressed to the court that they were able to streamline the trial by agreeing to narrow the issues on which evidence was to be presented and agreeing to the introduction of numerous documents including a number of expert reports and other evidence. Consequently, the record evidence is far more voluminous than that addressed in the trial transcript and has now been fully reviewed by this Court.
Based on the evidence admitted at trial, the Court makes the following Findings of Fact and Conclusions of Law.
FINDINGS OF FACT
III. FINDINGS OF FACT: UNC'S INTEREST IN THE EDUCATIONAL BENEFITS OF DIVERSITY IS SUBSTANTIAL AND LONG-STANDING
At trial, UNC Defendants produced substantial, credible, and largely uncontested evidence that it has made the deliberate decision to pursue the educational benefits that flow from student body diversity; has offered a principled, reasoned explanation for this decision; and that the benefits the University seeks to achieve are sufficiently measurable to permit judicial scrutiny.
A. UNC Has Made the Deliberate Decision to Pursue the Educational Benefits of Diversity
The University articulates in its Mission Statement that, as "the nation's first public university," it exists "to serve as the center for research, scholarship, and creativity and to teach a diverse community of undergraduate, graduate, and professional students to become the next generation of leaders." (DX001; DX003 at 2.) The University has long recognized that diversity and the educational benefits that flow from it are critical to its mission; and further, to fulfill its mission, the University must admit and enroll a diverse student body. (DX003 at 3; DX006 at 28; DX120 ¶ 20;) The University decided early on to define diversity broadly as "all the ways in which people differ, including primary characteristics, such as age, race, gender, ethnicity, mental and physical abilities, and sexual orientation; and secondary characteristics, such as education, income, religion, work experiences, language skills, geographic location, and family status." (DX088 at 6; see also DX083 at 7 (describing an early decision by the University not to "apply a narrow definition of the term").)
The UNC Board of Governors adopted this mission statement in 2009. (DX001.)
The University expressly established as an institutional goal the admission of underrepresented populations to enhance the educational process. (DX088 at 6.)
The University's commitment to pursue diversity is well documented in numerous institutional documents, a few of which are outlined here. As early as 1998, the University's Faculty Council expressly made clear that the University has an obligation to "create and sustain an environment of educational excellence" and "foster mutually beneficial interactions among students, faculty, staff, and administrators who possess diverse backgrounds and wide varieties of perspectives and life experiences." (DX003 at 2.) The University's 2003 and 2011 Academic Plans—which provide a set of strategies for achieving the University's academic goals—both reflect a commitment to these principles as well. (Nov. 12 Trial Tr. 521:9–523:4; DX006 at 5–6; DX007 at 17.) In 2004, the Office of Minority Affairs conducted the University's first diversity assessment which will be discussed in greater detail below. (DX083 at 6.) In 2010, the University conducted its second diversity assessment which reaffirmed the University's commitment to diversity and its educational benefits. In 2016, the Faculty Council reaffirmed the University's commitment to the values of diversity and inclusion by unanimously passing Resolution 2016-12: On Commitment to Diversity and Inclusion. (DX002 (resolving that diversity "is a critical element of academic excellence and a deeply-held institutional value"); see also Nov. 12 Trial Tr. 523:8–524:13.) Also, the 2017 Provost's Report details the University's long-standing commitment to diversity and inclusion, and identifies the key resolutions, strategic framework, plan of action, and campus programming necessary to capture the benefits that diversity provides. (DX003).
While these documents are not exhaustive of the voluminous evidence introduced by UNC related to its interest in diversity, they reflect, and the Court finds, that over the last two decades the University has recognized and actively pursued the educational benefits of diversity as one of its institutional priorities. Further, these documents illustrate that the University's commitment to diversity and its educational benefits is substantial and well-documented.
B. UNC Has Offered a Principled, Reasoned Explanation for Its Decision
In 2004, then-Chancellor James Moeser charged the Office for Minority Affairs with conducting UNC's first University-wide diversity assessment. (DX083 at 6.) In April 2005, UNC published the results of this assessment in the Report of the Chancellor's Task Force on Diversity. (Id. ) Following a months-long investigation including both quantitative and qualitative research, (id. at 9), the Task Force, a large committee of faculty, students, and staff, ultimately concluded that a diverse and inclusive community is a "critical element for a 21st century educational institution," (id. at 23). Though it identified a number of reasons for this conclusion, it focused on two particular concepts. First, it acknowledged that the "[v]estiges of [prior] discrimination, by [UNC] and society at large, remain today," and argued that the University had an obligation to remedy such effects. (Id. at 5.) Second, it found that diversity "matters ... because the world is a diverse place, and we must all learn to live and work in it." (Id. ) The Task Force articulated, as core values of the University with respect to diversity, the following:
Despite its comparatively recent embrace of diversity, one of Student-Intervenor's experts, Southern historian Dr. David Cecelski, provided the Court with credible evidence that UNC "has been a strong and active promoter of white supremacy and racist exclusion for most of its history." (ECF No. 179-14 at 9.) According to Dr. Cecelski, "[o]ver the centuries, the University's leaders have included the State's largest slaveholders, the leaders of the Ku Klux Klan, the central figures of the white supremacy campaigns of 1898 and 1900, and many of the State's most ardent defenders of Jim Crow and race-based Social Darwinism in the twentieth century." (Id. at 10.) Dr. Cecelski's expert report—which details UNC's reckoning with race over the full course of its history and illuminates the history of racial discrimination in North Carolina's K-12 public schools—is an important contribution to the Court's understanding of the context of this case. Dr. Cecelski provides considerable administrative and legislative findings that illuminate the extent of the educational disparities historically sanctioned by the State, and he concludes that, though recently "the University's faculty, administrators and trustees have made important strides to reform the institution's racial outlook and policies, ... those efforts have fallen short of repairing a deep-seated legacy of racial hostility and disrespect for people of color." (Id. )
• The University believes that it can achieve its educational, research, and service mission only by creating and sustaining an environment in which students, faculty, and its staff represent diversity, for example, of social backgrounds, economic circumstances, personal characteristics, philosophical outlooks, life experiences, perspectives, belief, expectations, and aspirations, to mention some salient factors.
• The University will achieve and maintain diversity on the campus through the admission of students and employment of faculty and staff who broadly reflect the ways in which we differ.
• The University promotes intellectual growth and derives the educational benefits of diversity by creating opportunities for intense dialogue and rigorous analysis and by fostering mutually beneficial interactions among members of the community.
In 2014 the University published a new Diversity Report which reenforced that attracting and retaining underrepresented minority students enriches the educational experience for all the members of the University community. (DX088.)
The University's subsequent work and research in this area has continued to develop and expand on these core values. In recent years, for instance, UNC has described the achieving of diversity as a goal "of paramount importance, as diversity enhances student growth and development in the cognitive, affective, and interpersonal domains." (DX088 at 6.) Another particularly illustrative example is the 2017 Provost's Report on the Educational Benefits of Diversity, which points to specific educational benefits of diversity the University seeks to attain, which include: (1) promoting the robust exchange of ideas; (2) broadening and refining understanding; (3) fostering innovation and problem-solving; (4) preparing engaged and productive citizens and leaders; and (5) enhancing appreciation, respect, and empathy. (DX003 at 5–6). In discussing these benefits, the Report stresses that they are interrelated and together strengthen the educational experience that the University provides to its students. (Id. )
In addition, leaders in the University community likewise provided additional reasons why the University has a significant interest in the educational benefits of diversity that were more specific to the roles they each played on campus. Stephen Farmer, who began working at UNC in 2000 and became the school's Vice Provost for Enrollment and Undergraduate Admissions in 2004, (Nov. 12 Trial Tr. 511:7–512:11), testified that the University is interested in the educational benefits of diversity so that "students will learn how to navigate in a complex multicultural world," (id. at 521:2-8.) Dr. Joseph DeSimone, the Chancellor's Eminent Professor of Chemistry, submitted a declaration averring that diversity of experience—to include differences among ethnic, cultural, socioeconomic, and gender factors—provides a "fertile ground for innovation" in his research lab and prevents the type of "groupthink" that stifles new ideas. (DX127 ¶¶ 20–22, 29.) Other University professors likewise testified that diversity promotes discovery and innovation and expands fields of inquiry. (DX118 ¶ 35; DX140 ¶ 9; DX145 ¶¶ 11–12.) Professors also averred that diverse classrooms expand discussions and offer opportunities for learning by, for example, generating more robust conversations about the history of the South, (DX121 ¶ 9), and providing "powerful and impactful learning moment[s]" in political science classes when discussing policing, (DX118 ¶ 30).
The Court finds that UNC has offered a principled, reasoned explanation for its decision to pursue the educational benefits of diversity. Further, the University has determined, in its academic judgment based on experience and evidence that, in order to prepare the next generation of leaders for North Carolina and the nation, the University must provide students with the experience of learning and living alongside people of different backgrounds. (Nov. 12 Trial Tr. 518:18-24.) In addition, the University has identified specific educational benefits it seeks to achieve from its cultivation of a diverse student body.
C. The Educational Benefits of Diversity that the University Seeks are Sufficiently Measurable to Permit Judicial Scrutiny
Defendants detailed several ways in which UNC measures its progress towards achieving the educational benefits of diversity, both quantitatively and qualitatively. One of the ways it does this is by collecting and reporting basic descriptive data of the student population. For instance, UNC's Office of Diversity and Multicultural Affairs has reported (1) the underrepresented minority ("URM") enrollment rate, disaggregated by race and gender; (2) four-year graduation rates, disaggregated by race, gender, and first-generation status; (3) and first-year male GPAs, disaggregated by race. (DX088 at 8–11.) It has additionally tracked the number of divisions on campus which have highlighted the University's diversity statement on their websites, (id. at 15), implemented diversity education or orientations, (id. at 20), and/or conducted research on the lived experiences of underrepresented groups, (id. at 26), among other things.
The term "underrepresented" in this context refers to any group "whose percentage enrollment within the undergraduate student body is lower than their percentage within the general population in North Carolina." (ECF No. 155-4 at 8.) This framework was established as part of a 1981 consent decree between the UNC System and the U.S. Department of Health, Education, and Welfare. (Id. ) "[T]he University has for more than three decades considered as underrepresented those students identifying themselves as African American or [B]lack; American Indian or Alaska Native; or Hispanic, Latino, or Latina." (Id. at 8–9.)
In addition, UNC has collected feedback on diversity issues from faculty and staff as well, including regular climate studies. (DX003 at 17; DX120 ¶ 48.) These studies likewise include partnerships with third-party groups that allow the University to benchmark its progress with other similarly situated institutions. (DX119 ¶¶ 87–88.) Defendants have provided a series of detailed reports prepared by UNC's Office of Institutional Research and Assessment that summarize student, faculty, and staff feedback at great depth. (See DX084; DX085; DX086; DX087.)
Defendants have also produced an "Inventory of Assessments Related to the Delivery of the Educational Benefits of Diversity and Inclusion" that details dozens of evaluations conducted by academic departments, student affairs, and the Chancellor's and Provost's Offices. (See DX061.) This inventory provides the full scope of an intensive effort to measure the University's progress towards its diversity goals. Moreover, the University has created the Educational Benefits of Diversity Working Group—reporting to the Chancellor and the Diversity and Inclusion Executive Council, (Nov. 16 Trial Tr. 809:14-21)—that oversees UNC's ongoing efforts and ensures that these assessments provide relevant and coordinated information to the highest levels of the University.
The Court finds that these assessments are mapped to the diversity goals that UNC has articulated and provide the University with the data necessary to track their progress. It is apparent that the University takes these efforts seriously and has a process in place to collect, review, and analyze the results. The Court therefore finds that the University's goals related to the educational benefits of diversity are sufficiently measurable to permit judicial scrutiny and are currently being measured by the University.
D. The Educational Benefits that the University Seeks Are Being Experienced
The educational benefits of diversity also are experienced and observed by university faculty, staff, students, and alumni across disciplines and professions. (See Nov. 16 Trial Tr. 780:24–781:11; DX116 ¶ 8; DX118 ¶¶ 20, 29; DX124 ¶ 12; DX131 ¶ 18; DX136 ¶¶ 10–13; DX139 ¶ 15; DX140 ¶ 9; DX144 ¶ 10; DX146 ¶¶ 14, 17.)
Other record evidence likewise reflects that a diverse student body improves students’ capacity to work effectively with others: exposure to diversity breaks down stereotypes, creates common understanding, and encourages empathy. (Nov. 16 Trial Tr. 911:6-10; DX118 ¶¶ 20, 30; DX123 ¶ 19; DX124 ¶ 24; DX126 ¶ 16; DX128 ¶ 13; DX130 ¶¶ 27–28; DX131 ¶ 19; DX132 ¶ 33; DX133 ¶¶ 19, 22; DX135 ¶ 12.)
Students have observed and experienced the importance of learning from classmates of different backgrounds and recognize that exposure to these differences is critical to their future success. (Nov. 13 Trial Tr. 680:13-20; Nov. 18 Trial Tr. 1298:8–1299:2; DX133 ¶ 24; DX148 ¶ 17; see also DX123 ¶¶ 4, 20–21; DX150 ¶ 4.) For example, Mary Cooper, a white alumna who served as student body president in 2011, averred that learning in a diverse student body prepared her to work with, coach, and teach others who do not look like her or who have not had the same experiences. (DX123 ¶ 21.) Likewise, Peter Henry, a Black male alumnus from a wealthy Chicago suburb, observed that when he became study partners with a white male from rural North Carolina he learned a "real lesson" about the "implicit assumptions" he made about people. (DX131 ¶ 19.)
Student-Intervenors also testified credibly and compellingly about the importance of racial and ethnic diversity to their educations while at the University. (Nov. 18 Trial Tr. 1297:12–1298:7 (Wingate-Bey) (testifying that racial and ethnic diversity enriched her educational experience); Nov. 16 Trial Tr. 885:25–886:3 (Polanco) ("[T]he ethnic and racial diversity I experienced at UNC helped me have more of an understanding of people from different backgrounds that are different from my own, which makes me better ....").) Further, alumni of the University confirm that exposure to diversity in college is necessary to prepare future leaders for their careers. (Nov. 16 Trial Tr. 911:23–912:3; Nov. 18 Trial Tr. 1269:14–1270:1, 1305:12-25; DX116 ¶ 18; DX117 ¶¶ 16–18; DX129 ¶ 17; DX131 ¶¶ 11–12; DX141 ¶¶ 12–14; DX147 ¶¶ 12–13; DX149 ¶ 17.) For example, student-intervenor Rimel Mwamba, a 2018 graduate, testified that her experiences as a student will enable her to treat and care for a diverse patient population in her career as a doctor. (Nov. 19 Trial Tr. 1371:24–1372:11.)
E. UNC's Efforts in Pursuing the Educational Benefits of Diversity While Ongoing, Are Unfinished
The University strives to foster a campus environment where the educational benefits associated with a diverse student body can be realized inside the classroom as well as outside the classroom. (Deposition of Carol Folt at 31:5–35:19 ("Folt Dep."), 75:11-18, 173:9-22; Deposition of Christopher Faison ("Faison Dep.") at 26:6–29:5; DX002 at 1; DX120 ¶ 42; DX125 ¶ 12.)
The University's expert, Professor Mitchell Chang of the University of California, Los Angeles, assessed the University's programming, including affinity groups, student housing initiatives, campus discussion forums, academic preparedness and mentoring programs, discipline-specific initiatives, course offerings, and classroom assignments. (DX108A at 29–32, 35–41.) Professor Chang determined that the University has implemented a variety of programs and policies that demonstrate that it systematically "engages diversity" to intentionally create the conditions necessary for the achievement of educational benefits. (DX108A at 5–6, 26, 33–34; see also IX001 at 38, 44, 47.)
Dr. Panter testified that the evidence reviewed by the EBD Working Group to date supports that while University students are receiving educational benefits from diversity, the University still has significant work to do in many different spaces in order to fully realize its goals. (Nov. Trial Tr. 805:18–806:7 (Panter).) Specifically, the University continues to face challenges admitting and enrolling underrepresented minorities, particularly African American males, Hispanics, and Native Americans. (Nov. 12 Trial Tr. 570:8-11 (Farmer).) For example, in 2013, the enrollment of African American men in the first-year class fell below 100 students. (Id. at 519:20-21, 570:15-17.)
In addition, Student-intervenors credibly testified that there were far fewer students of color on campus than they expected and that they experienced low minority representation. (Nov. 16 Tr. 879:2-7 (Polanco); Nov. 18 Trial Tr. 1285:19-25 (Ornelas); id. at 1299:3-17 (Wingate-Bey); id. at 1366:4-6 (Mwamba).) This underrepresentation causes minority students to experience loneliness and tokenism. (Nov. 16 Trial Tr. 879:17–880:14; Nov. 18 Trial Tr. 1300:13-18.) Underrepresented minority students also report feelings of isolation and unfair pressure to represent their race or ethnicity—effects that do not create the experience the University wants, or the students deserve. (Nov. 12 Trial Tr. 571:15–572:1 (Farmer); Nov. 18 Trial Tr. 1262:10–1263:5 (Brennen); DX128 ¶¶ 9–11; DX138 ¶ 9; DX142 ¶ 9; DX143 ¶¶ 14–18, 22–23; DX144 ¶¶ 3–6; IX001 at 65.) University professors likewise observe a lack of minorities in certain classes, fields, or areas of campus, this underrepresentation limits opportunities for exposure and learning. (DX126 ¶¶ 19–21; DX145 ¶ 15.)
Student Intervenors’ experiences are not isolated incidents. In 2016, UNC conducted its Undergraduate Diversity and Inclusion Campus Climate Survey, which documented the experiences of minority students on campus. (DX106.) In the survey, approximately half of African American students, a third of Hispanic students, a quarter of Asian American students, and 43% of Native American students agreed or strongly agreed with the statement "I feel pressured in the classroom to represent the views of all people from my racial and ethnic background," as compared to 4% of white students. (Id. at 65.) Approximately 40% of African American students, 30% of Hispanic students, and 25% of Asian American students agreed or strongly agreed with the statement "I feel that I need to minimize aspects of my racial or ethnic culture to fit in here." (Id. at 72.) In addition, 91 percent of students heard insensitive and disparaging racial remarks made by other students, and of students who experienced bias, 70% of Hispanic students, 82% of Asian American students, 95% of African American students, and 100% of American Indian students report they've specifically experienced bias due to their race. (Id. at 78.)
Student Intervenors argue that these experiences are in part due to a lack of "meaningful demographic representation" at the University. (See ECF No. 154-21 at 48.) Their expert, Dr. Uma Jayakumar, reported that in terms of racial identity, 72% of UNC students identify as white, while only 0.5% identify as American Indian/Alaska Native, 8% as African American, 8% as Hispanic, and 12% as Asian/Asian American. (Id. ) They note that these numbers reflect much less diversity than North Carolina as a whole, which a 2016 five-year survey found was 21.5% African American, 8.9% Hispanic, 2.6% Asian American, 1.2% American Indian/Alaska Native, and 69.2% white. (Id. at 49.) The demographic composition of faculty and administrators demonstrated even less diversity. (Id. )
The Court finds that the University's efforts in pursuing the educational benefits of diversity are substantial and ongoing. As highlighted in the testimony of the former Vice Provost, while "the [U]niversity has worked hard" to obtain the benefits of diversity, it is not "where [it] need[s] to be." (Nov. 12 Trial Tr. 569:12-25 (Farmer).)
IV. FINDINGS OF FACT: UNC'S ADMISSIONS PROCEDURES MANDATE A HOLISTIC EVALUATION OF CANDIDATES
Based on credible testimony and other evidence, and for the reasons detailed below, the Court finds that: (a) The University uses a race conscious admissions program to enhance student diversity; (b) UNC's admissions policies mandate that race is taken into consideration only as one of many "plus" factors and that the weight of each factor is determined in the context of all of the information present in a student's application; and, (c) UNC prohibits the use of race as a defining feature of any application, and its policies are flexible enough to allow the Admissions Office to assess all of the qualities of individual candidates. The Court further finds that the University policies do not provide for the evaluation of candidates in separate admissions processes according to their race, nor have they established any procedures that mechanically award points—or any consistent or predetermined weight—to an applicant on the basis of race.
A. Overview of UNC's Admissions Process
UNC's Board of Trustees sets the admissions policy for the University and has delegated responsibility for admitting new students to the Office of Undergraduate Admissions ("Admissions Office"). (ECF No. 225 ¶¶ 13–14 (J. Statement).) This Office consists of approximately 120 employees and is managed by the University's Vice Provost for Enrollment and Undergraduate Admissions. (Id. ¶¶ 15–16.) During the years at issue in this case, Mr. Farmer served in that role. (Id. ¶ 16; see also Nov. 13 Trial Tr. 639:6–640:15 (Farmer) (testifying that he would begin a new position on January 1, 2021).)
The University's admissions process is highly selective. (Nov. 12 Trial Tr. 519:8-14 (Farmer).) In a typical admissions cycle, UNC receives approximately 43,500 applications for its freshman class of 4,200. (Id. at 519:18-21.) Generally, about 47–50% of North Carolina residents who apply are offered admission while only 12–14% of out-of-state students are invited to enroll. (Id. at 519:10-14.) The difference in these rates is largely explained by two interacting factors: though two-thirds of applicants are non-North Carolinians, the UNC System's Board of Governors has capped out-of-state enrollment at 18%. (Id. at 519:24–520:6; ECF No. 225 ¶¶ 38–40 (J. Statement).)
If UNC exceeds this cap, the Board of Governors may reduce the school's operating budget for the following year. (ECF No. 225 ¶ 39 (J. Statement).) For example, when UNC enrolled too many nonresident students in the 2014-15 academic year, the Board of Governors penalized the University one million dollars. (Id. )
B. Application Requirements
All students applying to UNC must complete the Common Application, a streamlined online form now used by over 800 colleges across the country. (ECF No. 225 ¶¶ 24–25 (J. Statement).) As part of this process, applicants submit an essay responding to one of seven Common Application prompts, two "short answers" responding to prompts provided by UNC, their standardized test scores, and a letter of recommendation from at least one of their teachers. (ECF No. 154-7 ¶ 11 (Rosenberg Decl.).) Each student's high school counselor also submits the candidate's official high school transcript as well as a "secondary school statement," which provides UNC's Admissions Office with "information about an applicant's high school and the available curriculum" as well as "how the applicant compares with the rest of the applicant's high school class through comparative statistics on class rank and grade point average ("GPA") distribution." (Id. ¶ 12.)
Students are also able to submit additional letters of recommendation, resumes, artwork, music samples, or disability-related documentation if they so choose. (Id. ¶ 13.) During the years in question, any student wishing to receive financial aid must also complete the Free Application for Federal Student Aid ("FAFSA") form as well as the College Board's College Scholarship Service ("CSS") Profile questionnaire. (ECF No. 225 ¶ 77 (J. Statement).) (But see Nov. 12 Trial Tr. 556:8-14 (Farmer) (testifying that UNC no longer requires the CSS Profile because it imposed a "burden and a barrier for low-income students").) The University additionally requires that candidates pay an application fee but waives this fee for any student who demonstrates that they meet the fee waiver guidelines of the National Association for College Admission Counseling, the ACT, or the College Board. (ECF No. 225 ¶¶ 33–34 (J. Statement).) UNC will also waive the application fee for any candidate who is serving on active duty in the U.S. Armed Forces. (Id. ¶ 34.) Including all documents, applications may be as long as thirty pages. (See Nov. 13 Trial Tr. 711:6-7.)
In filling out the Common Application, candidates are also invited, but not required, to provide demographic information "such as gender, race, and ethnicity." (ECF No. 154-7 ¶ 10 (Rosenberg Decl.); see also Nov. 12 Trial Tr. 546:11-22 (Farmer).) According to uncontradicted testimony from Associate Director of Admissions for Evaluation Jared Rosenberg, who manages the day-to-day process of reviewing UNC's applications, (Nov. 13 Trial Tr. 677:23–678:4 (Rosenberg); ECF No. 154-7 ¶ 1 (Rosenberg Decl.)), the absence of such information is not held against any applicant during the evaluation process, (ECF No. 154-7 ¶ 10 (Rosenberg Decl.)). Though "race, ethnicity, or national origin may be used at any part of the process," at no point during that process are candidates of different racial groups reviewed, considered, or evaluated in separate groups. (Nov. 12 Trial Tr. 542:20–543:6 (Farmer); see Nov. 13 Trial Tr. 696:4-6 (Rosenberg).) As is discussed in more depth below, the University likewise mandates that race, ethnicity, and national origin may "never be used as anything other than one part of the comprehensive, holistic, and individualized review afforded to each candidate." (Nov. 12 Trial Tr. 542:21-23 (Farmer).)
Candidates for admission to UNC have the option of choosing between two different deadlines. The first, October 15, is a "non-binding early action deadline." (ECF No. 154-7 ¶¶ 8–9 (Rosenberg Decl.); Nov. 12 Trial Tr. 557:24-25 (Farmer).) The second, January 15, is referred to as the "regular decision deadline." (ECF No. 154-7 ¶ 9 (Rosenberg Decl.); Nov. 12 Trial Tr. 558:1 (Farmer).) These dual deadlines primarily function to give the Admissions Office a way "to space our reading of 40-some-thousand applications out over six months instead of over a smaller window of time." (Nov. 12 Trial Tr. 558:1-4 (Farmer).) In a typical year, early applicants make up about two-thirds of the candidate pool and are more likely to be North Carolina residents than nonresidents. (Id. at 558:8-11.) According to Farmer, "[t]here is no preference for early action" candidates, and Admissions Office leaders "continuously" talk with staff about evaluating early and regular decision candidates uniformly so that applications from the two windows are treated "as close to the same as we can." (Id. at 557:22, 558:12-21.)
C. Application Readers
After an application is submitted, it is reviewed by one of the approximately forty Admissions Office readers. (ECF No. 154-7 ¶ 17.) This group consists of both full-time admissions staff who balance their reading with other responsibilities and seasonal employees whose only job is to evaluate applications. (Nov. 13 Trial Tr. 681:7-18.) All readers have earned at least an undergraduate degree, and most have advanced degrees and/or professional experience in secondary or postsecondary education. (Id. at 681:21–682:3.) Many have been reading UNC applications for more than a decade, (id. at 682:2-3, 23-25), and all staff read and evaluate applications for approximately thirty hours per week, (id. at 681:11-13, 682:9-11). In hiring readers, the University "look[s] for individuals who are going to advocate for students, who can approach reading an application empathetically, sympathetically, and also with an eye towards understanding the entire student." (Id. at 683:16-21 (Rosenberg).)
Readers are initially trained in three phases. (Id. at 685:4-10.) The first phase, called New Reader Training, introduces first-time readers to the "ABCs of reading," including the logistics of the work. (Id. at 685:10-15; ECF No. 154-7 ¶¶ 20–21 (Rosenberg Decl.).) It also allows new readers to "learn about the University's admissions policies and procedures, admissions goals, and how to evaluate applications and make admissions decisions." (ECF No. 154-7 ¶¶ 20–21 (Rosenberg Decl.).) After the first two days of New Reader Training, all new and returning readers together attend the Reader Kick-Off for the second phase of training. (Id. ¶ 22.) In addition to reviewing the most recent version of the Reading Document which outlines University policies with regards to admissions, (id. ), this training provides context to readers by sharing statistics summarizing past applicant pools and admitted classes, and it gives readers the chance to discuss individual applications in small and large groups, (Nov. 13 Trial Tr. 686:13-24, 687:12-17). In the third and final phase, new readers reconvene to reflect on what they have learned. (Id. at 685:20-25.)
The Admissions Office also trains readers specifically on how to consider race and ethnicity in the evaluation process. During this training, Mr. Rosenberg "remind[s] readers that the University aims to enroll a diverse class across multiple dimensions, including but not limited to diversity of experience; ideas; backgrounds; socioeconomic status; racial and ethnic background; and first-generation college status." (ECF No. 154-7 ¶ 24 (Rosenberg Decl.).) To that end, the Admissions Office "instruct[s] readers to consider each applicant as an individual based on all relevant factors revealed in his or her application in order to understand the candidate holistically and comprehensively." (Id. ) Race and ethnicity, therefore, must be considered "as one factor among many based on a holistic review of all circumstances relevant to an individual applicant." (Id. ¶ 25.) Further, readers are instructed that "there are no quotas, fixed points, or separate admissions processes based on a particular candidate's race or ethnicity." (Id. )
Mr. Farmer described this process as "putting together that jigsaw puzzle of students who come from different places, who have different ideas, who have different backgrounds, who travel different paths." (Nov. 12 Trial Tr. 530:19-22 (Farmer).)
After this initial training is complete, the Admissions Office provides ongoing instruction in multiple ways. For instance, in some years more experienced readers have read behind new readers and provided them with feedback on their evaluations. (Id. ¶ 26.) More recently, all readers—new and returning—have read applications in pairs for the first three weeks of the reading period. (Id. ) While this process affords readers a chance to learn from each other and ensure that they are aligned, it is very rare for a second reader to change the first reader's decision once training is complete. (Nov. 13 Trial Tr. 708:6-15 (Rosenberg) (testifying that "the vast majority of the time"—and possibly as often as 99%—second readers sign off on the first reader's decisions).) Throughout the year, weekly staff meetings provide time for additional training and small group discussions among readers. (ECF No. 154-7 ¶ 27.)
Mr. Rosenberg described the ultimate goal of reader training as introducing staff to UNC's "individual, comprehensive[,] and holistic review." (Nov. 13 Trial Tr. 690:11-14 (Rosenberg).) In other words, the goal "is for them to understand that when they read an applicant, they're reading the entire applicant, not just the test score, not just the GPA, not just an essay. They're a whole person." (Id. at 690:15-18.) To that end, the Admissions Office wants readers to "understand the context" of each applicant's experience, including their home life, school life, and the environment in which they have grown up. (Id. at 690:19-25.) This is particularly important because "success can be defined differently in different environments," and therefore a reader must "understand that some students won't have the curriculum that other students have simply because their high schools don't offer it, [or] that ... some students may have a lot of test prep options, while others may not." (Id. at 691:1-8.) As Mr. Farmer testified,
[n]o student lives in the abstract. Students are real people. They come from real families. They live in real neighborhoods. They go to real schools that are in real communities. We don't feel as though we can understand any student fully unless we try to understand as fully as we can the context within which the student has lived and done his or her work ....
(Nov. 12 Trial Tr. at 533:7-15 (Farmer).)
D. Application Evaluation Process
The aforementioned Reading Document provides readers with high-level guidance in their review of applications, (ECF No. 225 ¶ 45 (J. Statement)), and includes UNC's Mission Statement, statements adopted by the University's faculty, and Admissions Office policies, (see ECF No. 155-4 at 2–6 (Reading Doc.)). In describing how the Admissions Office may use race, ethnicity, or national origin, the document repeatedly cites Supreme Court precedent as guideposts for its policy. (See id. at 9 (citing Grutter v. Bollinger , 539 U.S. 306, 123 S.Ct. 2325, 156 L.Ed.2d 304 (2003) ; Fisher v. Univ. of Tex. , 570 U.S. 297, 133 S.Ct. 2411, 186 L.Ed.2d 474 (2013) (" Fisher I "); and Fisher II , 136 S. Ct. at 2198 ).) It additionally describes the application evaluation process as well the criteria for admission. (Id. at 6–8).
1. Admissions Procedures
When an application is submitted, it is assigned randomly to a reader who then "reads the application, assesses the applicant across specified attributes, formulates an opinion about whether the student should be offered admission based on the totality of information in the applicant's record, and writes a comment defending his or her recommended decision." (Id. at 6.) Comments are designed to be "a summary of what [a reader] ha[s] read," and their purpose is to help a second reader understand why a first reader has made a particular admissions decision. (Nov. 13 Trial Tr. 688:2-6 (Rosenberg).) The time to evaluate an individual application may vary, but readers report that they are able to complete about five applications per hour. (Id. at 721:9-22). At this stage, reader decisions are considered provisional, and, in some cases, applications are automatically evaluated by a second reader. (ECF No. 155-4 at 6 (Reading Doc.); see also Nov. 12 Trial Tr. 534:23–535:5 (Farmer) (explaining the use of second reads for out-of-state students or for in-state applications in which the first reader is unsure of their decision).)
A minor exception to this description of the process—which is not at issue in this case—is that applications from students applying from locations outside of the United States are not assigned randomly. (ECF No. 225 ¶ 44 (J. Statement).)
Once all provisional decisions have been entered, applications then go to a process called "school group review" ("SGR"), where a committee composed of experienced staff members reviews every decision. (Nov. 12 Trial Tr. 535:16–536:5 (Farmer); ECF No. 225 ¶ 61 (J. Statement).). As a part of this stage, and pursuant to SGR Policy, the Admissions Office generates a report for every high school showing all of the high school's students who have applied for admission. (ECF No. 225 ¶ 67 (J. Statement).) These reports display the application deadline each student applied under; their provisional admissions decision; their class rank, GPA, and test scores; the ratings assigned to them by their initial readers; and their status as residents, legacies, or special recruits. (Id. ) At one time, the reports included information about race as well, but this data point has since been removed. (Nov. 13 Trial Tr. 705:19-24.) That said, during the years of 2014–17, a member of an SGR committee would have been able to open an applicant's file during the process and find this information for any student under consideration. (Id. at 705:25–706:5.)
The term "legacy" refers to a child of a UNC alumnus. (Nov. 9 Trial Tr. 14:8-9.)
Broadly speaking, the SGR exists to accomplishes two goals. (Nov. 12 Trial Tr. at 536:13–537:25.) First, it operates as "quality control" by ensuring (1) that data has been entered correctly for each applicant and (2) by allowing the committee to see provisional decisions arranged by high school so that it can identify any "anomalous" outcomes such as admitting several students from a school but denying its valedictorian. (Id. at 536:20–537:12 (Farmer).) While there might be "a really good reason" why an otherwise top-ranked student is denied admission and a lower achieving student at the same high school was invited to enroll, Mr. Farmer explained that the Admissions Office wants "to make sure that before decisions go out the door that we can defend them to counselors and schools ... [and] to families when they call." (Id. )
The SGR's second goal "is to make sure we don't over admit the class." (Id. at 537:13-14.) According to the Reading Document, the Admissions Office uses a statistical model that predicts how many provisionally admitted students will ultimately enroll. (ECF No. 155-4 at 6 (Reading Doc.).) In the event that the predicted number of enrolled students exceeds the total number of seats in the entering class, "the committee may need to fine-tune the number of admitted students and then reevaluate applications." (Id. ) Such reevaluation is conducted by the Director of Admissions, the Deputy Director of Admissions, the Associate Director of Admissions, or a designated subcommittee. (Id. )
Once the SGR is complete, decisions are communicated to all candidates, and any student who is not admitted may appeal to the Vice Provost for Enrollment and Undergraduate Admissions. (Nov. 12 Trial Tr. 538:16-21.) Should the Vice Provost decline the appeal, a candidate may appeal this decision to UNC's Provost. (Id. )
2. Admissions Criteria
The Reading Document states that the goal of each evaluation "is to understand the candidate individually, comprehensively, and holistically." (ECF No. 155-4 at 7 (Reading Doc.).) To that end, "[c]andidates for admissions are evaluated on everything the admissions process reveals about them and not on the basis of formulas or preset scoring requirements." (Id. ) Though the Reading Document describes it as "difficult, if not impossible, to list every criterion that might be used over the course of an admissions season," it nevertheless enumerates approximately forty criteria in eight categories that may be considered at every stage of the evaluation process. (Id. at 7–8.) That said, neither the categories nor individual criteria comprise "a checklist that all candidates must satisfy," nor are they exhaustive descriptions of applicant qualities that a reader may consider in his or her evaluation. (Id. at 8.)
Those eight categories and underlying criteria are described in the Reading Document as follows:
• Academic program criteria : rigor, breadth, and pattern of courses taken, all viewed within the context of the entire applicant pool, and the student's high school and any previously attended post-secondary institutions.
• Academic performance criteria : grade-point average, rank in class, individual grades, trends in grades, and patterns in grades, all viewed within the contexts of the entire applicant pool and the student's high school and any previously attended post-secondary institutions.
• Standardized testing criteria : results from the SAT or ACT, and available SAT Subject, Advanced Placement, and International Baccalaureate exams, as well as occasional results from state-mandated end-of-course exams, all viewed in light of the documented strengths and limitations of these tests, for all first year and sophomore transfer candidates
• Extracurricular activity criteria : engagement outside the classroom; persistence of commitment; demonstrated capacity for leadership; contributions to family, school, and community; work history; unique or unusual interests.
• Special talent criteria : in music, drama, athletics, and in writing[.]
• Essay criteria : idea, organization, voice, vocabulary, sentence structure and grammar; evidence of self-knowledge and reflection; insightfulness; unique or unusual perspectives.
• Background criteria : relative advantage or disadvantage, as indicated by family income level, education history of family members, impact of parents/guardians in the home, or formal education environment; experience of growing up in rural or center-city locations; status as child or step-child of Carolina alumni.
• Personal criteria : curiosity; kindness; creativity; honesty and integrity; motivation; character; impact on community; exceptional achievement in-or-out of the classroom; history of overcoming obstacles or setbacks; openness to new cultures and new or opposing ideas; talent for building bridges across divisions in school or community or among individuals from different backgrounds.
(Id. at 7–8.)
As they read applications, readers are responsible for providing numerical ratings for five of these categories: academic program, academic performance, extracurricular activity, personal, and essay. (Nov. 9 Trial Tr. 71:22–72:15.) The Admissions Office considers these ratings valuable to the extent that they may later offer "a quick way to understand what an application may look like or represent" without having to read the application in its entirely. (Nov. 13 Trial Tr. 697:11-14 (Rosenberg).) There are no minimum ratings required for admission, the scores are not added together at any point to create a composite score, and there is credible and uncontradicted evidence that the "ratings do not equal a decision at all." (Id. at 697:7-19; see also ECF No. 225 ¶¶ 50–51 (J. Statement).)
According to UNC's Associate Director for Research in the Office of Undergraduate Admissions, Jennifer Kretchmar, readers stopped assigning numerical ratings to essays in 2018. (Nov. 9 Trial Tr. 72:9-15 (Kretchmar).)
With regards to race, ethnicity, and national origin, the Reading Document is clear that "[a]t no point in the process are candidates of different racial or ethnic backgrounds reviewed in separate groups." (ECF No. 155-4 at 8 (Reading Doc.).) Further, the University does not have "explicit or implicit quotas for any particular racial or ethnic group, or for underrepresented students as a whole, or for students of color as a whole." (Id. ) That said, the Reading Document also states that "[c]onsistent with the Supreme Court's decision in Grutter , the race or ethnicity of any student may—or may not—receive a ‘plus’ in the evaluation process depending on the individual circumstances revealed in the student's application." (Id. at 9.) Further,
while a "plus" that is awarded may be significant in an individual case and tip the balance towards the admission of the student, it is not automatically awarded, and not considered in terms of numeric points or as the defining feature of an application. Even if awarded, a "plus" does not automatically result in an offer of admission.
(Id. )
The Court finds that at no time in the admissions process are candidates considered in separate groups according to their race nor is a candidate insulated from comparison with all other applicants. Further, there is no evidence that candidates are awarded automatic points on the basis of their race, nor that points are added together at any stage in the process. The only numerical ratings that readers use function merely as internal shorthand descriptions of a reader's impression of a particular candidate.
Moreover, the evidence demonstrates, and this Court finds that race may be used only as a "plus" factor that is sometimes—but not always—provided to URM candidates within the context of their full application. Race is one of more than forty criteria considered in every application, and the evaluation process is flexible enough to consider all of the pertinent elements of diversity that may be present for any particular applicant. Finally, UNC's policies are clear that race may never be used as the defining feature of a candidate's evaluation.
V. FINDINGS OF FACT: NON-STATISTICAL EVIDENCE DOES NOT DEMONSTRATE DISCRIMINATION
In support of their contention that UNC discriminates unlawfully in its admissions process, Plaintiff has provided both statistical and non-statistical evidence. Plaintiff's contentions regarding its non-statistical evidence fall into three distinct categories. First, Plaintiff argues that, despite the University's publicly stated commitment to holistic admissions, the actual process does not live up to its billing. According to Plaintiff, this is because UNC either conceals an improper use of race behind opaque procedures or is unable to ensure that the work of its large Admissions Office is consistent with its stated mission. Second, Plaintiff contends that the consideration of race permeates the admissions process to a degree that suggests race is a predominant factor. In the words of Plaintiff's Counsel, "[a]lthough UNC's witnesses claim that race is just one aspect of many, [the University's] myopic focus on race goes far beyond any attention it gives to the representation of other aspects of diversity ... they claim to value." (See Nov. 9 Trial Tr. 21:18-23 (Strawbridge).) Third, and finally, Plaintiff argues that UNC has not adequately reflected upon its practice to fully reckon with the role race plays in University admissions.
The Court finds, however, that Plaintiff has provided no evidence that UNC conceals the improper use of race behind opaque procedures, has allowed race to become a predominant factor in the admissions process, or has failed to adequately reflect on the role race plays in admissions decisions.
A. There is No Evidence that UNC Conceals the Improper Use of Race Behind Opaque Procedures
In spite of its policies and procedures, Plaintiff challenges the degree to which UNC legitimately aims to have a "holistic" admissions process and whether it has reasonably implemented the checks and balances necessary to ensure that these goals are realized. The Court finds that there is no evidence that UNC conceals the improper use of race behind opaque procedures.
1. UNC's Intent
a. Provisional Decision Stage
Plaintiff focuses on two parts of the admissions process in its attempt to demonstrate that Mr. Farmer's description of the admissions process obscures the true intent of the University. First, it points to the initial application evaluation stage in which readers make provisional decisions. (See ECF No. 247 ¶¶ 73–74.) Plaintiff contends that reader emails regarding candidates show that "Admissions officers frequently highlight the applicant's race as a key feature" of their application. (Id. ¶ 74.) To that end, it identifies eight statements made by UNC readers—out of the hundreds of thousands of application files and materials shared during discovery, (ECF No. 225 ¶¶ 100–02 (J. Statement))—that it argues demonstrate how readers "commonly boil[ ] down [applicants] to their race." (Id. ; see also Nov. 9 Trial Tr. 22:16-21 (Strawbridge); ECF No. 247 ¶ 74.) While it is true that each of these comments refers to the race of a candidate, none of them indicates that race is considered outside of a holistic admissions process, much less as the defining feature of any application.
For instance, two of the comments simply mention the race of a candidate without reference to any weight the factor had in an admissions decision. (See ECF No. 247 ¶ 74.) Two other comments refer to the race of a candidate with other strong attributes tending to show that readers considered several aspects of their candidacies in reaching their provisional decision. (Id. ) Additionally, two comments refer not to admissions but to a candidate's potential for merit scholarship consideration; both comments also refer to at least one other factor the reader considered. (Id. )
One email, in the words of Plaintiff, "express[es] disappointment that an applicant with perfect test scores was Asian and not ‘Brown.’ " (Id. ) However, nothing in this description by Plaintiff refers to the evaluation process completed by the reader. The comment does not suggest, for instance, that the student would not be admitted because they were Asian American, nor even that two such candidates with the same scores would be evaluated differently. Finally, one reader commented that they were conducting additional research on a candidate's residency status and "going through this trouble because [a candidate] is a biracial (black/white) male." (Id. ) This aside tends to suggest that a reader was actively looking for additional factors that might tip the scale in favor of a borderline student who was a URM. It is notable that neither the student's race nor gender were tipping points in the evaluation process for the student at issue. Further, this comment implies that other factors would ultimately play a determinative role in whether the student was admitted.
In sum, while it appears that readers may at times communicate with each other regarding the race of candidates, there is no evidence in any of these statements that race was considered as anything other than a "plus" factor or outside of the type of holistic process UNC describes. Indeed, every comment that invokes a provisional decision additionally describes a non-racial factor with which the candidate's race interacts—such as "solid everything," "[s]tellar academics," standardized test scores, gender, or residency status. The Court therefore finds that these emails provide no evidence that readers have "boiled down" any candidates to their race. Moreover, these eight emails do not establish a routine practice.
Plaintiff additionally takes issue with the "swiftness with which UNC's readers process applications," arguing that "a mere ten to twelve minutes to review, evaluate, score, and comment on each 30-page plus application suggests a formulaic review process." (Id. ¶ 311.) This claim, without additional context, is difficult to credit. For one, Plaintiff provides no comparison data with any other admissions process to set forth what amount of time would be necessary for a professional reader—who evaluates over a thousand applications per cycle—to assess each candidate holistically. Nor do they address how familiarity with the essay prompts, secondary school statements, or admissions rubric might allow UNC's readers to identify the key parts of an application more quickly than one unfamiliar with the process. In the same way a first semester law student is impressed by the speed at which their slightly more experienced upper-class peers can identify the holding in a case, it would seem entirely consistent with general experience that a well-educated and well-trained reader would be able to conduct a holistic analysis of an application in a time that may seem "swift" to the untrained eye.
Moreover, it is uncontested that the majority of applications are evaluated by a second reader, and all candidates are reviewed again in the subsequent SGR process. Accordingly, the Court finds that there is no basis to find that the speed at which readers conduct their first read of applications is a shortcoming of UNC's admissions process or suggests that the University is not able to provide an individualized approach to its candidates.
b. SGR Process
The final stage of the admissions process that Plaintiff suggests acts as a veil to disguise an unlawful use of race is the SGR process, in which a committee of veteran readers looks at provisional admissions decisions for all candidates from a single high school. (See ECF No. 247 ¶ 86 (contending that UNC uses the SGR process to "shape[ ] and fine-tune[ ] the class, including from a racial/ethnic standpoint) (citing PX070); see also ECF No. 1 ¶¶ 48–50 (Compl.).)
Yet Plaintiff provides no basis for this assertion, and none of Plaintiff's witnesses testified to that end. (See generally Nov. 10 Trial Tr. 301:25–302:5 (Arcidiacono) (agreeing that he is "offering no opinion here as to whether or not school group review was the mechanism used by UNC to allegedly insert racial preferences into admissions"); Nov. 12 Trial Tr. 465:17-21 (Kahlenberg) (agreeing that he is "not offering any opinion in this case on whether or not the school group review process was used to manipulate the racial composition of the admitted class").) Further, Plaintiff admitted during closing arguments that this is "a claim that was not developed at trial." (Nov. 19 Trial Tr. 1378:18–1379:2 (Strawbridge).)
On the other hand, Defendants have provided uncontradicted evidence demonstrating that the SGR process changes the racial composition of the class very little and, to the extent that it does, has only reduced the number of admitted URM students. Defendants’ expert witness, Dr. Caroline Hoxby, tracked the number of students by race who entered the SGR process as provisionally accepted from 2013-14 to 2015-16 as well as those who were ultimately accepted after the process was complete. (Nov. 17 Trial Tr. 978:5–980:24.) In 2013-14, there were 1.0% fewer admitted Hispanic students in the class and no change in African American students after the SGR process than had been provisionally accepted. (Id. ) The following year, African American representation dropped by 0.3% and Hispanic representation remained flat. (Id. ) In the third and final year of Dr. Hoxby's study, both African American and Hispanic representation dropped, by 0.1% and 0.2%, respectively. (Id. ) Over the course of three years, only one time did any single underrepresented race increase its representation in the class following SGR, and this was in 2015-16 when Pacific Islander students went from 0.0% of the admitted class prior to SGR to 0.1% afterwards. (Id. ) At the same time—in direct contrast to the URM trend—there were more white and Asian American students admitted as part of UNC's classes following the completion of the SGR every year. (Id. ) Accordingly, the Court finds that the SGR process has not been used to "fine-tune" the racial composition of UNC's admitted classes in favor of URM candidates.
2. UNC's Multiple Checks, Balances, and Quality Controls
Plaintiff counters, however, that even an admissions process intended to be holistic by university leadership "could be abused," and Mr. Farmer acknowledged that this is possible. (Nov. 11 Trial Tr. 652:7-9 (Farmer).) Therefore, in order to prevent abuse, Mr. Farmer agreed with Plaintiff's Counsel that the Admissions Office "needs to guarantee that its holistic admissions process has effective checks, balances, and quality control." (Id. at 652:3-6.) The Court finds that there is uncontested evidence that such checks, balances, and quality controls exist throughout the UNC admissions process.
For example, checks and balances are in place from the moment applications are submitted. As described above, the reading protocol pairs staff members together at the beginning of every admissions cycle to evaluate each other's provisional decisions. (ECF No. 154-7 ¶ 26.) Further, in some preset circumstances, the admissions policy calls for an automatic second read of applications that fall into certain categories. (ECF No. 155-4 at 6 (Reading Doc.); see also Nov. 10 Trial Tr. 534:23–535:5.) New readers are also provided with ongoing training where more experienced staff members evaluate their work and give them regular feedback. (ECF No. 154-7 ¶ 26.)
Additionally, senior members of the Admissions leadership team testified credibly and without contradiction that they read hundreds of admission files every year to ensure that readers are engaging in holistic decisions. Mr. Rosenberg, for instance, testified that he "read[s] behind [every reader] multiple times, many, many times throughout the year." (Nov. 13 Trial Tr. 703:14-20 (Rosenberg) (testifying that he is "constantly seeing or reviewing files," "reading their comments," and "understanding how they got to their decisions.").) Mr. Farmer also testified that he is confident that UNC's readers are engaging in holistic evaluations because he "had seen the hard work that our staff had done in training, and I, myself, had read behind people in the office on thousands of folders." (Nov. 13 Trial Tr. 656:14-18 (Farmer).) Finally, quality control is also one of the primary goals of the SGR process where every provisional decision is reviewed by "a committee comprised of experienced members of the Admissions office." (ECF No. 225 ¶¶ 59, 61 (J. Statement).)
Though his testimony is discussed in much more depth below, Plaintiff's expert witness Dr. Peter Arcidiacono stated that, based on his statistical analysis, he agreed that the UNC admissions process is holistic. (See, e.g. , Nov. 9 Trial Tr. 163:23–164:5 (Arcidiacono).)
In sum, while there is strong evidence that University leadership has adopted checks, balances, and quality control measures to ensure that its policy of incorporating a holistic admissions policy is followed at all stages of the process by all participants, Plaintiff has provided no evidence to show that this process has been used inappropriately.
B. There is No Non-Statistical Evidence that Race is a Predominant Factor in Admissions
Plaintiff contends that the consideration of race permeates the admissions process to such an extent that it must be a predominant factor as the University makes its decisions. According to Plaintiff, this racial focus eclipses other aspects of diversity that the University "claim[s] to value." (Nov. 9 Trial Tr. 21:18-23 (Strawbridge).) Plaintiff fails, however, to provide any meaningful evidence to sustain such an allegation.
1. UNC Uses Race as One of Multiple Factors in Recruiting Prospective Students
Plaintiff first argues that UNC's student recruitment efforts demonstrate the use of race as a predominant factor in its admissions process. As background, the University begins its recruitment process each year by identifying its prospect pool. (Nov. 13 Trial Tr. 728:20-23.) It first adds to this pool students who have independently expressed interest in attending UNC. (Id. at 728:24–729:2.) Then the University adds other individuals based on a combination of their standardized test scores, GPA, residency status, and demographic data. (Id. at 729:2-4.) To identify this latter group, the Admissions Office works with the College Board and the ACT and purchases contact information of high school students who meet the preset criteria. (Id. at 729:2-23.)
In recruiting North Carolina residents, UNC purchases the information of any student who self-reports at least a B average in high school and who scores a 3 or higher on an AP exam or an 1100 or higher on the SAT. (Id. at 730:1-7.) In a separate batch, the University also purchases information for any North Carolina student who scores between 900 and 1100. (Id. at 730:8-15.) Instead of mailing these students recruitment materials, however, UNC sends them information from the Carolina College Advising Corps about ways to prepare for college. (Id. )
The University also purchases information for students with equivalent ACT scores. (Nov. 13 Trial Tr. 730:1-7.) Though not noted subsequently, this practice holds true for all data purchases described in this section.
In recruiting out-of-state students, UNC purchases contact information for those who score 1400 or higher on the SAT and have maintained a B grade point average. (Id. at 731:1-4.) The University also sends recruitment materials to out-of-state students with a B average who score between 1250 and 1390 if they identify as a first-generation college student, low-income, or a URM. (Id. at 731:4-9.) According to UNC's Associate Director for Recruitment Michael Davis, this final band of students is a part of the University's "priority populations for recruitment" and is included so that the school may "cast[ ] as wide a net as possible to talk to those students and let them know about the opportunities at Carolina." (Id. at 731:12-16.)
Plaintiff argues that this practice "requires some racial groups to perform better on standardized tests than others in order to be deemed admissible (and therefore recruited)." (ECF No. 247 ¶ 47.) As an example, Plaintiff observes that, in some years, "URMs in North Carolina could score as low as a 26 on the ACT and be recruited, whereas white and Asian-American students needed at least a 29." (Id. ) Plaintiff notes that recruitment efforts based in part on race also extended to University recruitment events as well. (Id. ¶ 48 (citing Nov. 13 Trial Tr. 750:7–751:7).)
Despite its objections here, it does not escape the Court's attention that there is tension between this argument and Plaintiff's allegation that, as a race-neutral alternative, UNC could achieve its diversity goals "by improving its recruitment of socioeconomically disadvantaged, high-achieving minorities." (ECF No. 1 ¶ 132 (Compl.).)
While this practice does implicate race, at most it only expands the pool of applicants to the University and has no bearing on the actual decision-making criteria the Admissions Office uses to make its decisions. There is no evidence, for instance, that an increase in the number of URM applicants in any way changes the nature of a reader's holistic evaluation. Nor is there evidence that the University considers students in these pools to be admissible based on these criteria alone. According to testimony, it appears that UNC merely believes that students in priority populations are less likely to have considered the University or college in general. Notably, every student whose contact information is purchased must also achieve a certain standardized test score and GPA in order to be added to the list, and no student is added to the pool—much less admitted at this stage—based on race alone.
2. The Reporting of Race Does Not Suggest Its Improper Use
Plaintiff also takes issue with the use of what the parties refer to as "core reports." These reports "identified the racial/ethnic composition of the admitted class ... with comparisons to the prior year" and were sent in various forms to senior admissions staff during some of the years at issue in this case. (ECF No. 247 ¶ 77 (citing Nov. 9 Trial Tr. 87:5–92:2).) In addition to identifying the racial makeup of the class, at different times these reports listed information about gender, residency, citizenship, and legacy status. (Nov. 9 Trial Tr. 88:22–89:11.) Today these reports also include information about first-generation college students as well as those who qualify for fee waivers. (Id. at 89:14-16.)
There is no evidence, however, that these reports have affected admissions decisions in any way. Indeed, it is unclear how they might have such an impact given that they were only viewed by senior staff members not directly involved in provisional decision making. While Mr. Farmer acknowledged that "some people in the office" had access to core reports disaggregated by race prior to 2015, (Nov. 12 Trial Tr. 546:23–547:6 (Farmer)), Mr. Rosenberg stated in uncontradicted testimony that he had no knowledge of any seasonal or admissions office staff readers having access to "any reports showing the racial and ethnic breakdown of the provisionally admitted class." (Nov. 13 Trial Tr. 702:1-5 (Rosenberg).)
Nevertheless, as an added precaution, admissions policy was changed in 2015 so that any person who had a reason to see such data must thereafter recuse themselves from reading any additional applications. (Nov. 12 Trial Tr. at 547:7–548:2.) Although according to Mr. Farmer "readers had never made decisions in light of information that they knew" about the racial composition of the class, this change was enacted so that "there could be no confusion and there could be no misinterpretation of [the University's] practices." (Id. at 548:3-9 (Farmer).) Today, readers only learn the racial makeup of the incoming first-year class in the summer after reading is complete, the waitlist is disbanded, and the class is fully enrolled. (Nov. 13 Trial Tr. 702:23–703:1.)
Precautions notwithstanding, the Court finds that—even if readers themselves had access to data showing the real-time projected composition of an incoming class—these core reports do not undermine the University's position that race was used as one factor among many in a holistic admissions process. Indeed, there is no evidence before the Court that the reports—which included a variety of factors in addition to race—were ever used in this way.
3. The Consideration of Race at all Stages of the Admissions Process is Consistent with How UNC Evaluates All Factors
The Court briefly addresses Plaintiff's contention that UNC "seeks to take race independent of any other factor into account at all stages of its admissions process." (Nov. 19 Trial Tr. 1377:22-24 (Strawbridge).) This language presumably comes from the University's Reading Document, which states that "race, ethnicity, or national origin may be used at any stage in the admissions process." (ECF No. 155-4 at 8.) This phrasing, however, echoes earlier language in the Reading Document which, referring to other factors, states that "more than forty criteria ... are used at every stage of the admissions process." (Id. at 7.) Read in this context, the consideration of race throughout the evaluation of an application is unexceptional and does not distinguish the consideration of race from the rigor of a student's high school schedule, work history, special talent in music, or creativity. (See id. at 7–8.) Moreover, the Reading Document provides more context for this statement by making it clear that race "is never used as anything other than one part of the comprehensive, holistic, and individualized review afforded to each candidate." (Id. at 8.) The Court therefore finds that the availability of race as an optional factor in considering an application at "every stage" of the admissions process is consistent with all other factors and not evidence of a non-holistic evaluation process.
C. There is No Evidence that UNC Has Not Taken a Serious Accounting of Its Use of Race
Third, and finally, Plaintiff argues that UNC has not taken a serious accounting of the role that race plays in its admissions process. This contention is expressed in two ways. First, Plaintiff alleges that UNC has not taken advantage of all of its resources to study the impact of race in admissions. Second, Plaintiff contends that the University's lack of definition for the term "critical mass" shows that UNC "has no idea whether its current level of racial diversity achieves its diversity goals." (ECF No. 247 ¶ 38.) The Court, however, finds that there is no basis for either of these conclusions.
1. There is Evidence of Attention to the Use of Race Throughout the Admissions Process
Plaintiff first alleges that the Admissions Office's internal research history shows a lack of urgency to the use of race in the admissions process. According to Plaintiff, although Jennifer Kretchmar, UNC's Associate Director for Research in the Office of Undergraduate Admissions, (Nov. 9 Trial Tr. 62:2-4), has completed studies on the effect that gender, legacy status, admissions deadline, and standardized test scores have played on admission decisions, she has never conducted a study that would analyze the extent to which race plays a role, (ECF No. 247 ¶ 93 (citing Nov. 9 Trial Tr. 83:3–84:6 (Kretchmar) and Nov. 13 Trial Tr. 653:11–656:25 (Farmer))). Mr. Farmer, who had the authority to commission such a study, testified that he believed such an analysis was not necessary because he "felt confident in the work that our staff was doing." (Nov. 13 Trial Tr. 654:23–655:11 (Farmer).)
One University group, the Committee on Race Neutral Strategies, subsequently completed a study in 2016 to measure whether race was a predominant factor in admissions. (Nov. 16 Trial Tr. 842:11–845:7 (Panter).) This study concluded that "underrepresented minority status does not meaningfully drive the prediction accuracy of the final multivariate model." (Id. at 844:23-25)
For its part, Plaintiff argues that not asking Ms. Kretchmar to complete a study demonstrates that "UNC willfully blinded itself to the true impact race has in its admissions decisions." (ECF No. 247 ¶ 92.) There are, however, other ways to assess the impact of race in admissions outside of a statistical analysis. To that end, it would have been reasonable for Mr. Farmer to rely on his firsthand monitoring of the process in concert with the Admissions Office's intense focus on race in its ongoing training. In sum, the Court finds that, while a study by Ms. Kretchmar would have provided yet another way for UNC to examine its practice, the presence of other assessments rendered this particular type of analysis ultimately unnecessary to demonstrate the Admissions Office's engagement with the role race plays in its admission process.
2. There is Evidence that UNC Has Defined and Assessed the Concept of Critical Mass
In 2005, UNC issued its first diversity plan following a "formal, year-long assessment" that found " ‘widespread agreement’ among students, faculty, and staff that ‘they had learned and benefited’ from interactions with colleagues from different backgrounds." (ECF No. 155-4 at 4 (Reading Doc.).) Among other things, this plan called for the admission of students who could contribute to the diversity of the University as well as the "achievement of critical masses of underrepresented populations." (Id. ) The absence of critical mass, the University found, "impedes the educational process," places "undue pressure on underrepresented students," and limits the degree to which all students can experience the "educational benefits of a diverse learning environment." (Id. )
a. UNC has Defined Critical Mass by Reference to the Educational Benefits of Diversity Critical Mass is Designed to Produce
Plaintiff, however, contends that UNC does not discuss the concept of "critical mass" in its Admissions Office, has not determined if it has achieved a critical mass of underrepresented students, and has not defined the term. (ECF No. 247 ¶¶ 37–39.) There is evidence on the record to support these claims, at least in part. For instance, former University Provost James Dean testified in a deposition that, "[i]n all of [his] conversations with Steve Farmer, that phrase [critical mass] has never come up." (Id. ¶ 37 (quoting Dean Dep. 144:22–145:5).) Ms. Kretchmar also testified that "critical mass isn't language that we use in our everyday work." (Nov. 9 Trial Tr. 108:14-15 (Kretchmar).) Several witnesses for UNC were likewise unable to point to a written definition of the term. (See, e.g. , Nov. 13 Trial Tr. 660:21-25 (Farmer) (agreeing that no definition of the term appears in the final report of the Committee for Race-Neutral Alternatives).).
In Grutter v. Bollinger , the University of Michigan's law school sought to enroll a " ‘critical mass’ of minority students" through its holistic admissions process. 539 U.S. at 329, 123 S.Ct. 2325. The Court held that an interest in "critical mass" was "not simply ‘to assure within its student body some specified percentage of a particular group merely because of its race or ethnic origin’ "; such a specific number "would amount to outright racial balancing which is patently unconstitutional." Id. at 329–30, 123 S.Ct. 2325 (quoting Regents of Univ. of Cal. v. Bakke , 438 U.S. 265, 307, 98 S.Ct. 2733, 57 L.Ed.2d 750 (1978) (opinion of Powell, J.)). Rather, the concept "is defined by reference to the educational benefits that diversity is designed to produce." Id. at 330, 123 S.Ct. 2325. Enumerated among those benefits in Grutter , and discussed at some length above, were cross-racial understanding, breaking down stereotypes, improved classroom discussions, the promotion of learning outcomes, student preparation for postgraduate life, and the cultivation of strong leaders. Id. Though there is not substantial evidence that UNC's Admissions Office used the phrase "critical mass" regularly, there is significant evidence that it defined, discussed, and measured the concept the phrase describes by reference to the educational benefits that diversity is designed to produce. In testimony, Mr. Farmer discussed critical mass as follows:
I think that critical mass, as I said earlier, is complicated, and I think that critical mass has to be assessed not exclusively in terms of numbers but really in the lived experience of our students: What they're learning, how they're thriving, what they're contributing to the learning and the thriving of others .... [M]eaningful representation is important in critical mass for the reasons that I've just described. It's important that students not feel isolated. It's important that students feel free to be all they are instead of feeling trapped into being one-dimensional, and it's important that students feel free from stereotype. All those things are important, and meaningful representation is important for those reasons.
(Nov. 12 Trial Tr. 567:23–568:11 (Farmer).)
Further, Ms. Kretchmar summarized when acknowledging that the term "critical mass" was not often used in the Admissions Office, "to the extent that critical mass is aiming towards certain ends, achieving educational benefits of diversity, those are things that we talk about frequently." (Nov. 9 Trial Tr. 108:14-17 (Kretchmar).) Not only are these benefits discussed, but Mr. Farmer testified that the university has "codified them" and is "assessing [itself] against those specific benefits." (Nov. 12 Trial Tr. 520:24–521:1 (Farmer).)
With regards to the benefits themselves, multiple witnesses spoke credibly, consistently, and without contradiction about what that they sought to achieve by enrolling a critical mass of students, demonstrating to the Court that the Admissions Office is well-versed in this concept. Among those benefits expressly described at trial—either in the words of university staff or University documents—were:
• cross-racial understanding through living and learning alongside one another, (Nov. 12 Trial Tr. 518:18-24 (Farmer); id. at 525:22 (quoting the Provost's report on the educational benefits of diversity));
• breaking down stereotypes, (id. at 544:13-21);
• improved classroom discussion through different perspectives, (id. at 522:24–523:4 (quoting from the 2011 Academic Plan); 524:7-13 (quoting a faculty resolution));
• academic excellence, (id. at 523:24–524:4 (quoting a 2016 faculty resolution discussing the enhancement of the learning environment when students, faculty, and staff from diverse backgrounds interact); Nov. 16 Trial Tr. 781:2-11 (Panter) (describing classrooms where students "com[e] together around a meaningfully conceived intellectual pursuit" and with different members contributing in different ways);
• promotion of innovation, new ideas, and problem-solving, (Nov. 12 Trial Tr. 520:13-16, 521:4-5, 524:7-13 (Farmer) (quoting a faculty resolution); id. at 525:20-23 (quoting the Provost's report on the educational benefits of diversity));
• teaching students how to navigate the world, (id. at 521:2-8; 522:6-7 (quoting one of the University's academic priorities); id. at 525:24-25 (quoting a provost's report on the educational benefits of diversity); id. at 527:4-14));
• the cultivation of leaders, (id. at 517:22-25 (citing the University mission); id. at 525:24-25 (quoting the Provost's report on the educational benefits of diversity) id. at 526:14-23 (describing student leadership in University ROTC programs));
• enhancing appreciation, respect, and empathy for others, (id. at 525:25–526:1 (quoting the Provost's report on the educational benefits of diversity));
• improving the experience of underrepresented groups so that they were not isolated or having to act as spokespeople for their race, (id. at 544:5-21, 568:4-12).
Accordingly, the Court finds that UNC has defined the term "critical mass" not by numbers nor unconstitutional quotas, but by reference to the educational benefits of diversity this concept is designed to produce.
b. UNC Has Conducted Periodic Reviews to Reach the Conclusion that it Has Not Attained Critical Mass
The Court in Grutter also held that "race-conscious admissions policies must be limited in time." Grutter , 539 U.S. at 342, 123 S.Ct. 2325. Universities are able to meet these durational requirements through "sunset provisions in race-conscious admissions policies and periodic reviews to determine whether racial preferences are still necessary to achieve student body diversity." Id. The Court therefore turns to examine to what extent the University has endeavored to examine its progress towards meeting its diversity goals and end race-conscious admissions at the school.
According to Mr. Farmer, UNC "has been assessing the health of our campus, the success of our students, and the success of diversity initiatives for a long time." (Nov. 12 Trial Tr. 572:6-9 (Farmer) (testifying that this work had occurred over decades).) One person who has been directly involved in these efforts is Abigail Panter, UNC's Senior Associate Dean for Undergraduate Education and a Professor of Psychology and Neuroscience at the University. (Nov. 16 Trial Tr. 767:5-8 (Panter).) Dr. Panter has chaired UNC's Advisory Committee for Undergraduate Admissions as well as the Committee for Race-Neutral Strategies, and she has been a member of the Educational Benefits of Diversity and Inclusion Working Group. (Id. at 776:17–777:12.)
Dr. Panter described three components in the University's assessment of the educational benefits of diversity: (1) longitudinal data that measures the experiences of individual students and is tied to a set of local and national assessments; (2) reports that evaluate how students work through the general education curriculum, including department-level tools that assess student engagement; and (3) measures designed to assess faculty training and inclusive teaching. (Id. at 782:11–785:10.) In partnership with UNC's Institutional Research and Assessment department, Dr. Panter has additionally led cohort studies, climate surveys, and focus groups, and she has conducted intensive research on course evaluations in her efforts to produce a full picture of the University's progress on its diversity initiatives. (Id. at 785:15–790:8.) More recently, in 2017, Dr. Panter has joined the Educational Benefits of Diversity Working Group, established by Provost Robert Blouin, whose charge is "to coordinate and enhance the assessment of the University's ongoing efforts to realize the educational benefits of the diversity and inclusion for its undergraduate students." (Id. at 794:12-17 (quoting DX005 at 2).)
With regards to whether UNC has already achieved its goals, this working group has reported that many UNC students are indeed benefitting from the University's efforts in this area. (Id. at 804:15–805:22.) That said, Dr. Panter insisted that there remains "a lot of work to do in many different spaces." (Id. at 806:3-7 ("We are not done. We are not done.").) Mr. Farmer echoed this understanding and testified that his experience in speaking with students and faculty, studying University surveys, and reading The Daily Tar Heel student newspaper has led him to the conclusion that "students wish there were more diversity, including more racial and ethnic diversity, on our campus, and the feeling is particularly pronounced among underrepresented students." (Nov. 12 Trial Tr. 568:25–569:25 (Farmer) ("I've honestly never heard a person say at the university that we're where we need to be.").) He testified that, for example, "the low enrollment of African American men [has] caused a lot of harm. It caused a lot of hurt." (Id. at 570:15-21.) For instance, Mr. Farmer alluded to a time in 2013 when the enrollment of first-year African American men fell below a hundred, and three students filmed a video in front of the Admissions Office where they described their experiences at UNC. (Id. at 570:22–571:2.) Shortly thereafter, a group of American Indian students spoke out in the same way. (Id. at 571:2-4.)
These stories echo the experiences described by the UNC students and alumni who testified at trial on behalf of Intervenors. One alumna, Cecilia Polanco, testified, for instance, that it was "pretty apparent once I arrived on campus that there were much, much less students of color than I thought." (Nov. 16 Trial Tr. 879:2-7 (Polanco).) As a result, she "often felt alone and a bit invisible in some spaces because I was ... the only Latina in some of those spaces." (Id. at 879:23-25.) She said that, despite her hope that she would not be the person "called on to speak on Latino issues or immigrant issues," she often was. (Id. at 880:1-4.) "And it was uncomfortable," she testified, "because I didn't want to be a speaker for my whole community just based on my experience. It felt like tokenization a lot." (Id. at 880:4-6.)
This idea was reiterated by multiple other alumni. Star Wingate-Bey, a 2016 UNC graduate, testified that she was the only Black student in many of her classrooms during her time at the University. (Nov. 18 Trial Tr. 1299:12-17 (Wingate-Bey).) In her words, those experiences "can feel isolating." (Id. at 1300:13-15.) She testified that she often felt as though she were "the token or the sole representative for [her] race[,] or the fact checker for [her] race, which can be a bit of a burden, in class." (Id. at 1300:15-18.)
Andrew Brennen, a 2019 UNC graduate, likewise testified to the discomfort he felt in class as the only African American student when issues of race were discussed. (Id. at 1262:10–1263:5 (Brennen).) Mr. Brennen also described hearing racial slurs directed at him multiple times as he walked on UNC's campus. (Id. at 1265:17–1266:10.) As a Black student, he testified that, "in the moment, you smile, you deescalate," but he "was always careful" and "always did what [he] needed to do" to avoid making the situations even more difficult. (Id. at 1265:23, 1267:3-10.)
The Court finds that the University's ongoing review of its diversity goals is robust and able to assess UNC's progress on a consistent basis. That said, while some portions of the student body appear to be reaping the benefits of diversity, there is both quantitative and qualitative evidence that UNC has not yet achieved its aims. To that end, while UNC has not set forth a proposed time period in which it believes it can end all race-conscious admissions practices, the evidence unmistakably demonstrates that such a time has not yet been achieved.
VI. FINDINGS OF FACT: STATISTICAL EVIDENCE DOES NOT DEMONSTRATE DISCRIMINATION
The parties additionally provided the Court with detailed statistical evidence to support their claims. While SFFA relied almost exclusively on statistical evidence to support its case, UNC counters SFFA's evidence with its own experts. Each engaged an expert witness to create and describe econometric models that purported to demonstrate the role that race plays in the University's admissions decisions. Both experts’ models—while, at times, employing different statistical methods—contribute to the overall understanding of UNC's process despite the experts’ starkly divergent conclusions.
A. The Parties’ Experts are Highly Qualified
The credentials of the parties’ experts are very impressive. Plaintiff's expert Professor Peter Arcidiacono has a Ph.D. in economics from the University of Wisconsin and has been a professor of economics at Duke University for more than two decades, earning tenure in 2006. (Nov 9. Trial Tr. 115:10-24.) His research focuses on labor economics and applied microeconomics, and he has also completed studies on the economics of higher education. (Id. at 116:7-18.) Professor Arcidiacono specializes in using empirical models which, in his words, involve econometric modeling to develop an "estimator" and then applying that estimator to real data. (Id. at 116:19–117:3 (Arcidiacono).) Professor Arcidiacono testified that he has published about forty papers, almost all of which have involved econometric modeling, and almost a quarter of which "relate to affirmative action." (Id. at 118:21–119:6.) Additionally, he has served as a coeditor or associate editor for multiple journals and has been appointed as a research associate for the National Bureau of Economic Research ("NBER"). (Id. at 123:20–124:2.) In 2018, Professor Arcidiacono was elected as a Fellow of the Econometric Society and in 2020 was elected as a Fellow of the International Association for Applied Econometrics. (Id. at 124:6-9.) In a similar action brought by Plaintiff against another university in which he was also an expert witness, a federal district court found Professor Arcidiacono to be "highly respected" and "well-qualified." Students for Fair Admissions, Inc. v. President & Fellows of Harvard Coll. , 397 F. Supp. 3d 126, 158 n.40, 159 (D. Mass. 2019).
The credentials and experience of Defendant's expert Professor Caroline Hoxby are equally impressive. Professor Hoxby earned a Ph.D. in economics from the Massachusetts Institute of Technology in 1994 and immediately began teaching as a professor of economics at her undergraduate alma mater, Harvard University. (Nov. 17 Trial Tr. 929:16–930:3 (Hoxby).) In 2007, she moved to Stanford University where she is currently a named professor in economics specializing in the economics of education. (Id. at 931:1-9.) She was appointed by President George W. Bush to serve as the Program Director of the Economics of Education Program for the NBER, is a member of the American Academy of Arts and Sciences and has been awarded the Smithsonian's award for ingenuity for her original and thought-provoking research. (Id. at 933:10-14, 935:4-13.) She has also received the Fordham Award as the top educational researcher in the United States. (Id. at 933:14-21.)
There is little need to differentiate between the qualifications of these two experts. The Court finds that both are entirely qualified to create the types of statistical models they offer as evidence in this case. The experts differ with regards to the specific questions they sought to answer and the methods they used to reach their conclusions. However, the Court is well-positioned to balance the trade-offs in these different approaches and consider how they have impacted the experts’ ultimate findings. The Court begins this analysis first by providing a brief description of the questions that each expert testified that he or she attempted to answer and an overview of how each conducted their work.
B. Overview of Research Questions and Process
1. Professor Arcidiacono
Professor Arcidiacono testified that he sought to understand (1) "how formulaic UNC's admissions decisions are" and (2) "the role that race plays in those admissions decisions." (Nov. 9 Trial Tr. 114:24-25 (Arcidiacono).) To answer these questions, he completed what he described as a four-part process. (Id. at 125:13-18.) The first part of the process was to create a data set. (Id. at 125:18-21.) UNC provided application data to Plaintiff for every candidate who applied between the 2011-12 admissions cycle and the 2016-17 cycle. (Nov. 18 Trial Tr. 1122:4-9 (Hoxby).) This included over 200,000 applications, with about 65,000 from in-state applicants and 135,000 from nonresidents. (ECF No. 225 ¶ 100 (J. Statement).) For each applicant, the University provided the following data points:
The parties’ stipulated that UNC "produced data relating to all applicants from the 2012-13, 2013-14, 2014-15, 2015-16, and 2016-17 admissions cycles (that is, for the incoming freshman classes that would make up the graduating classes of 2016 to 2021)." (ECF No. 225 ¶ 99 (J. Statement).) This statement lists five individual classes before parenthetically noting a graduating range of six classes. Based on the experts’ testimony citing to data from six different admissions cycles, the Court concludes that the class of 2011-12 was unintentionally omitted in the Joint Stipulation.
[1] admissions and enrollment decisions, [2] whether the applicant applied for early action, [3] academic background measures such as test scores and grades, [4] the five ratings that the University's admissions officers assigned to each file, and [5] demographic information such as residency status, race/ethnicity, gender, and whether the applicant is a first-generation college student.
(Id. ¶ 101.) Professor Arcidiacono testified that, before beginning his analysis, he first decided which of these applications were "relevant" to the questions he was answering. (Nov. 9 Trial Tr. 125:19-21 (Arcidiacono).) As a result of this review, he removed the applications of any student who withdrew or otherwise submitted an incomplete application, any application listing a reader rating of zero, any application that was flagged in UNC's system as "special," and any application for a student who was applying from outside of the United States. (Id. at 128:18–130:17.) These exclusions removed 37,555 students from the data set, or about 19% of all applications provided. (See ECF No. 247-1 at 3 (Arcidiacono demonstratives).)
Professor Arcidiacono removed the applications of students who withdrew or submitted incomplete applications because these students were "not really competing for admissions in the same way." (Nov. 9 Trial Tr. 128:24–129:3 (Arcidiacono).) For instance, students who withdrew their applications from consideration before receiving a decision might have extremely strong credentials but nevertheless would still not be admitted to the University.
Professor Arcidiacono removed students with admissions ratings of zero because such a rating would indicate "bad data" in the sense that zero was not one of the rating options available to UNC readers and therefore must signal a mistake. (Nov. 9 Trial Tr. 129:6-9 (Arcidiacono).)
Though Professor Arcidiacono was not able to discern the meaning of all of the "special" flags in the UNC system, he testified that this group included candidates, for instance, who were recruited athletes, stating that "the admissions process operates very differently for them." (Nov. 9 Trial Tr. 129:10-21 (Arcidiacono) (observing that applicants with "special" flags were admitted 97% of the time).)
Professor Arcidiacono excluded international applicants given his understanding that "this case is about domestic applicants; and at least at a lot of universities, for example, need-blind admissions will not apply to foreign applicants." (Nov. 9 Trial Tr. 130:12-17 (Arcidiacono).)
The second step in Professor Arcidiacono's analysis was to complete a descriptive analysis of the applications that remained. (Nov. 9 Trial Tr. 125:22-23 (Arcidiacono).) In this stage, he looked more closely at "basic patterns" in the data such as the admission rates of students disaggregated in multiple ways. (Id. )
Third, Professor Arcidiacono built several iterations of an econometric model with the goal of simulating UNC's admissions process as closely as possible using only the observable variables available to him. (Id. at 125:24–126:3.) As part of this phase, he also attempted to approximate how important the different variables were in UNC's decision-making, and he assigned mathematical weights for each variable—such as GPA, race, or first-generation status—that he considered. (Id. at 163:12–164:5.) Using these weights, he then assigned each student a probability of admission. (Id. ) Fourth, and finally, Professor Arcidiacono attempted to quantify the effect of race in his model by modifying the race of individual applicants to see how such a change would alter the model's projection for a student's probability of admission. (Id. at 126:4-7.) In Professor Arcidiacono words, this amounted to "turning on or off racial preferences" within the model itself. (Id. )
2. Professor Hoxby
Professor Hoxby testified that she sought to answer whether Plaintiff's allegations that race and ethnicity were dominant factors in the admissions process were accurate. (Nov. 17 Trial Tr. at 936:10-14 (Hoxby).) In answering this question, Professor Hoxby's process mirrored Professor Arcidiacono's in many ways, and she constructed nine increasingly complex regression models attempting to simulate the University's process. (Id. at 939:9-19.) She diverged from Professor Arcidiacono's analysis in two primary ways: by using a slightly different set of variables in her preferred model and by performing what is known as a Shapley decomposition to separate those variables into "[b]uckets or bins" to determine the effect that any single factor had on the overall admissions process. (Id. at 948:16–950:19.) The Court next looks at the results of these experts’ analyses.
As discussed in more detail below, Professor Hoxby also sought to answer, "whether there were potential race-neutral or race-blind alternatives to the current admission process that would allow UNC to attain its current levels of racial and ethnic diversity and academic preparation." (Nov. 17 Trial Tr. 936:16-20 (Hoxby).) She was additionally charged with responding to the opinions proffered by Plaintiff's experts, Professor Arcidiacono and Richard Kahlenberg. (Id. at 936:21-23.)
C. Descriptive Statistics
1. Basic Findings
The first part of Professor Arcidiacono's analysis produced a set of descriptive statistics. (Nov. 9 Trial Tr. 125:22-23 (Arcidiacono).) As a baseline, Professor Arcidiacono testified that 48% of in-state applicants were accepted compared to only 14% of out-of-state applicants. (ECF No. 247-1 at 4 (Arcidiacono demonstratives).) He next disaggregated this information by race and found that 51% of white applicants and 54% of Asian American applicants who were North Carolina residents were admitted while only 31% of in-state African American applicants and 41% of in-state Hispanic applicants were invited to enroll. (Id. at 5.) On the other hand, out-of-state URMs had a higher acceptance rate than their white and Asian American counterparts: 17% of out-of-state African American candidates and 20% of Hispanic candidates gained admission compared to 11% of white candidates and 17% of Asian American candidates. (Id. ) Looking at all students regardless of residency and race, 26% of UNC applicants were admitted. (Id. at 4.) Overall, this included 31,740 white and Asian American students, who were admitted 25.8% of the time, and 7,270 African American and Hispanic students, who were admitted 24.3% of the time. (See id. at 6–7.)
Professor Arcidiacono did not include any other racial groups in this part of his analysis. (See ECF No. 247-1 at 5–19 (Arcidiacono demonstratives).)
Professor Arcidiacono also testified that, "in part because of the history of our country, [some trends in applicant data] are pretty different across racial groups." (Nov. 9 Trial Tr. 137:1-5 (Arcidiacono).) For instance, he observed that in-state African American candidates "have substantially lower test scores, lower rank in their class, and lower grades than other applicant groups." (Id. at 137:5-7.) Further, he found that the "academic patterns are very similar" for out-of-state students as well. (Id. at 139:15-20.)
Professor Arcidiacono further testified that he found patterns along racial lines with regards to the internal ratings that UNC readers gave to applicants. As detailed above, readers during this time gave candidates a numerical rating in five different categories: academic program, academic performance, extracurricular activity, personal qualities, and essay. (Id. at 71:22–72:8 (Kretchmar).) Professor Arcidiacono testified that "whites and Asian Americans scored better on UNC's program rating, performance rating, extracurricular rating, and essay rating than African Americans and Hispanics." (Id. at 137:15-20 (Arcidiacono); see also id. at 140:7 (noting that this holds true for both in-state and out-of-state candidates).) However, he testified that "[w]hen we get to the personal quality rating, it flips, and there African Americans and Hispanics rate higher on the personal quality rating than whites and Asian Americans." (Id. at 137:20-23.)
The Court observes that readers rate applicants at least in part on the opportunities available to them. (See ECF No. 155-4 at 7 (Reading Document) (describing the "academic program criteria" as "the rigor, breadth, and pattern of courses taken, all viewed within the context of the entire applicant pool " (emphasis added)).) As recently as 2020, a North Carolina court found that many of the state's Black and Hispanic students continue to lack equal access to college preparatory resources. See Hoke Cty. Bd. Of Ed. v. State , 95-CVS-1158, at *2, 16, 27 (N.C. Super. Ct., Jan. 21, 2020). The court in Hoke concluded that the state failed in the last twenty-four years to provide resources to all students, "especially students of color," sufficient to prepare them for post-secondary education or vocational training as required by the North Carolina Constitution, and racial disparities in public schooling have actually "increased" in recent years. Id. at *2, 16. Those students who nevertheless succeed in earning admission to institutions like UNC must do so without the benefit of well-funded college preparatory programs.
That said, the Court observes that this is not an apples-to-apples comparison because Professor Arcidiacono did not use the same calculations for each of the ratings. For the first three ratings—academic program, academic performance, and extracurricular activities—Professor Arcidiacono averages the scores of applicants. (Id. at 136:11-18.) For the essay and personal ratings, he instead calculates the percentage of students who scored above a five in this area because "five is by far the most common" rating that readers assign. (Id. at 136:19-21; see also ECF No. 247-1 at 6–7 (Arcidiacono demonstratives).) It is unclear to the Court why this would be a sufficient reason to alter the method of calculation between two variables one is trying to compare. The Court therefore has no evidence before it that the average personal rating is higher or lower for any particular racial group.
This evidence instead shows that URM students were, on the whole, more likely to score above a 5 on their personal ratings than their white and Asian American peers. Even this conclusion, however, does not always hold true when disaggregated by race. For instance, in-state Asian American applicants and in-state African American applicants were both rated above a 5 on the personal rating 20% of the time. (See ECF No. 247-1 at 6 (Arcidiacono demonstratives).) By comparison, 18% of white in-state candidates scored above a 5 and 23% of in-state Hispanic students did the same. (Id. ) These numbers are slightly higher for all groups in the out-of-state pool, with Asian American and African American applicants again having the same percentage of students (24%) who earned higher-than-average ratings. (Id. at 7.)
In sum, this evidence demonstrates that URM students were admitted to UNC at a lower rate than their white and Asian American counterparts during these six admissions cycles. On average, URM students were rated lower by UNC readers on their academic program, academic performance, and extracurricular activities. A lower percentage of URM students scored above a 5 on UNC's essay rating than did white and Asian American candidates, and a higher percentage scored above a 5 in the personal ratings. UNC raters were not asked to provide scores for candidates in three categories—standardized testing, special talent, or background—and therefore there are no descriptive statistics for any students along those criteria.
2. Deciles Analysis
Professor Arcidiacono next conducted a deciles analysis where he looked at applicant outcomes across a range of academic data. To complete this analysis, he created an "academic index" ("AI") that "equally combine[d] [a candidate's] SAT score and their high school grades." (Nov. 9 Trial Tr. 141:11-25 (Arcidiacono).) This AI was designed to be a summary statistic explaining a student's academic qualifications in a single number. (Id. ) Though Professor Arcidiacono testified that schools in the Ivy League use an AI with "slightly different weights," (id. at 142:1-6), he acknowledged that UNC does not make any such calculation as a part of its admissions process, (id. at 143:24-25; Nov. 10 Trial Tr. 316:22-24). In her testimony, Professor Hoxby underscored this point by noting that the AI is "an admissions index that Professor Arcidiacono invented." (Nov. 17 Trial Tr. 958:13-17 (Hoxby) (testifying further that she does not "like to call it the academic index because it's not actually used by UNC, or any other college or university of which I am aware").)
As described below, the experts in this case consistently used SAT scores as a rough proxy for academic ability, preparedness, and potential. That said, witnesses for both parties acknowledged the limitation of this data point when considered in isolation. (See, e.g. , Nov. 12 Trial Tr. 456:20–457:7 (Kahlenberg) (citing research that showed differences in SAT scores as a result of socioeconomic status and not academic ability); id. at 598:7-12 (Farmer) ("[W]hen you compare two people one to another, a difference of 60 points on the SAT probably isn't a material difference, especially if you're able to assess each of those two candidates on many dimensions.").) That said, the use of standardized test scores here appears to be the best option available to approximate, at the very least, how UNC assesses such qualities in its candidates. As Professor Hoxby summarized, there is no reason to "consider the SAT or the ACT to be perfect measures of academic preparedness, not at all. These tests are imperfect. They have issues. There are issues around bias in these tests. There are issues regarding test retaking. So, they are not perfect academic indicators, but they are standardized across high schools, and so they are the way that most economists and statisticians do try to judge academic preparedness. As imperfect as they are, they are better than the alternative—they are better than other types of indicators." (Nov. 17 Trial Tr. 985:10-19 (Hoxby).)
After creating this index, Professor Arcidiacono calculated an individual AI for each candidate whose application showed a GPA on a four-point scale and an SAT score. (Nov. 9 Trial Tr. 142:16–143:1 (Arcidiacono).) This included about 95% of in-state applicants and about 70% of out-of-state applicants. (Id. at 142:18-20.) Despite using a smaller sample, Professor Arcidiacono testified that the subset of students he included in this analysis "seem to be pretty representative" of the entire pool given its size and the similarity between the admissions rates between the subset and the full sample, including when disaggregated by race. (Id. at 143:2-23; see also ECF No. 247-1 at 9 (Arcidiacono demonstratives).)
Professor Arcidiacono then ranked the remaining students by AI score and divided them into ten deciles so that the top decile, decile ten, would include the candidates whose scores were in the top 10% of the candidate pool. (Nov. 9 Trial Tr. 144:15-21.) The next decile down, decile nine, included candidates whose AI scores were between the eightieth and ninetieth percentiles. (Id. ) This pattern was continued down to the lowest decile of those candidates scoring in the bottom 10% on this metric. (Id. )
Professor Arcidiacono then sought to calculate to what extent his AI calculation matched UNC's reader ratings. Professor Arcidiacono found that "people who do better on the academic index also tend to do better on the program rating." (Id. at 146:19-21.) He found the same result with respect to academic performance as well. (Id. at 146:22-25.) Though the differences between deciles were much less pronounced for extracurricular ratings, essay scores, and personal qualities, (id. at 147:5-12; see also ECF No. 247-1 at 10 (Arcidiacono demonstratives)), these patterns held true across all of the ratings and for both in-state and out-of-state students, (Nov. 9 Trial Tr. 147:5–148:19). Professor Arcidiacono's "overall takeaway" was that "higher scores on test scores and grades are also associated with UNC ratings." (Id. at 148:21-23.)
Professor Arcidiacono tested this question by simply calculating the percentage of students in each decile whose reader ratings were above the median in each category and comparing the results in a chart. (Nov. 9 Trial Tr. at 145:10-11 (Arcidiacono).)
Professor Arcidiacono next analyzed how different racial groups were distributed across his AI deciles. Looking first at in-state candidates, he found that white applicants were "disproportionately in the higher academic index deciles." (Id. at 149:2-23.) He found the same to be true for Asian American candidates, observing that nearly 20% of Asian American applicants were in the top decile. (Id. at 150:2-8, 155:7-10; see also ECF No. 247-1 at 13 (Arcidiacono demonstratives).) On the other hand, he found that African Americans "disproportionately have lower test scores and grades," and that Hispanic candidates do as well. (Nov. 9 Trial Tr. 150:9–151:1, 152:15-19 (observing that 17% of Hispanic candidates were in the lowest decile); id. at 155:7-11 (observing that more than half of African American applicants were in the lowest two deciles).) Professor Arcidiacono testified that he conducted research for out-of-state applicants as well, and that he found "broadly similar patterns" in the data. (Id. at 156:5-9.)
Finally, in concluding this part of his analysis, Professor Arcidiacono looked at admissions rates for each decile disaggregated by race and found that "when you're in the very top decile, pretty much everybody is getting in." (Id. at 154:3-5.) In the middle deciles of his AI calculation, however, there are meaningful differences between races. For instance, in decile five, "whites and Asian Americans have admit rates that are below 30%, but the African American admit rate is over 40 points higher, at 71%, and the Hispanic admit rate is almost 54%." (Id. at 154:13-18.) Finally, Professor Arcidiacono attempted to "fill the class" with students based on their AI scores alone and found that the share of URMs would be cut substantially, especially out-of-state, if UNC selected its students solely on the basis of his AI calculation. (Id. at 159:21–161:16.)
Professor Hoxby, for her part, takes issue with this analysis. She testified that Professor Arcidiacono "assumes that ... all deciles are equally relevant to the UNC admissions process, which they are not." (ECF No. 251-1 at 16 (Hoxby demonstratives).) As she observes in the data, "[m]ost UNC admitted students come from the top few deciles," and therefore to emphasize applicants who are "on-the-bubble"—where any single factor may tip the scales for a student on the cusp of admission—exaggerates the impact that race has on the process writ large. (Id. ; see also Nov. 17 Trial Tr. 958:25–959:20 (Hoxby).)
Ultimately, Professor Arcidiacono acknowledged that "the decile analysis by itself is insufficient to show that the differences of admission rates are the result of racial preferences." (Nov. 10 Trial Tr. 317:8-11 (Arcidiacono); see also Nov. 9 Trial Tr. 162:16-19 (Arcidiacono) (testifying that these descriptive statistics on their own are "just suggestive").) This is likely due to several factors. For one, and most glaringly, all of the data that comes from this analysis is predicated on the use of an admissions index that UNC has never applied to candidates.
Further, it is uncontested that UNC's admissions decisions are based on more than the GPA and standardized test scores accounted for in this index. As Professor Arcidiacono acknowledged in explaining his analysis, "UNC takes into account lots of characteristics" in evaluating candidates. (Nov. 9 Trial Tr. 159:24.) Therefore, using only these two variables, in Professor Hoxby's words, "tends to treat [all other factors] as though they didn't exist, as though it were just random which students got admitted from Decile 5." (Nov. 17 Trial Tr. 962:15-18 (Hoxby).) As she testified, students of all races are both admitted and denied from any given decile, "so there must be other factors that are unobservable to the statisticians but can be seen by the admissions officers." (Id. at 962:19-23.)
Though this was not the focus of Professor Arcidiacono's analysis, the Court also observes that 3% of in-state, top decile African American candidates were denied admissions by UNC, more than double the percentage of their white counterparts with AIs in this decile. (See ECF No. 247-1 at 13 (Arcidiacono demonstratives) (showing 1.2% of white students and 1.8% of Asian American students being denied in the top decile); see also Nov. 10 Trial Tr. 371:23–372:3 (Arcidiacono).) If URM candidates were largely defined in the admissions process by their race, one would expect to find that every URM demonstrating academic excellence at the highest level would be admitted or, at the very least, be admitted at a higher rate than their non-URM peers. This deciles analysis suggests instead that race does not even act as a tipping point for some students with otherwise exceptional qualifications. On the contrary, it strongly implies that a holistic admissions process is taking into account a number of factors in addition to these three criteria.
The Court finds that this deciles analysis is insufficient to demonstrate that UNC has departed in any way from the admissions process it has described and fails to provide evidence that race is a predominant factor.
D. Econometric Modeling
After creating a data set in the first phase of his work and compiling descriptive statistics in his second phase, Professor Arcidiacono then began to respond more directly to the two questions he sought to answer: how formulaic is UNC's admissions process, and what role does race play in its decisions? To that end, he next constructed an econometric model designed to "approximate what actually happens in UNC's admissions process as best we can." (Id. at 164:8-9.) Professor Hoxby also constructed an econometric model designed to further illuminate the role race plays in the University's evaluation of candidates.
For this work, he used a logit model commonly used in situations where the outcome is binary. (Nov. 9 Trial Tr. 163:1-8 (Arcidiacono).)
Before more closely examining the two experts’ models and for the reasons described in more detail below, the Court first finds Professor Hoxby's model to be more probative on the issue of whether race is a dominant factor in UNC's admissions than Professor Arcidiacono's. This is primarily because the data that Professor Hoxby uses to build her model more accurately reflect the University's actual process and is free of some of the assumptions that meaningfully undermine Professor Arcidiacono's conclusions. That said, even if the Court were to credit Professor Arcidiacono's model fully, each expert's calculations demonstrate that race plays only a minor role in UNC's admissions process.
The Court therefore finds that there is no evidence that race was used by the University as a predominant factor in evaluating candidates nor that it was the defining feature of any individual application. Instead, the evidence tends to show that the Admissions Office afforded each candidate a comprehensive, holistic, and individualized review consistent with the process outlined in the Reading Document. The Court now turns to the models themselves.
1. Professor Arcidiacono's Process
Professor Arcidiacono started by choosing several variables to add as "inputs" into his model. (Id. at 163:12-22; see also ECF No. 247-1 at 24 (Arcidiacono demonstratives).) In his preferred model, which forms the basis of his analysis, Professor Arcidiacono used the following factors:
Professor Arcidiacono identified his fourth iteration, Model 4, as his preferred model. (Nov. 9 Trial Tr. 185:4-11 (Arcidiacono).)
• race/ethnicity;
• gender;
• application deadline used (i.e., early or regular decision);
• legacy status;
• first-generation college status;
• use of a fee waiver;
• application year;
• SAT math and verbal section scores;
• GPA;
• class rank;
• UNC reader ratings; and
• intended college major.
(See ECF No. 247-1 at 25 (Arcidiacono demonstratives).) He additionally added several controls that accounted for missing data and also included variables that controlled for interactions between different factors such as race and gender. (Id. )
He next tried to determine how accurately a mathematical formula incorporating these variables could predict "the probability that an applicant was admitted or rejected." (Nov. 9 Trial Tr. at 163:23–164:1 (Arcidiacono).) Though UNC does not calculate such probabilities, Professor Arcidiacono nevertheless testified that this model output was necessary as opposed to a binary admit/reject decision. This is because, regardless of how many variables one includes, there will always be unobservable factors that are unaccounted for in a holistic process, and therefore any model could not predict an admissions decision with full certainty. (Id. at 164:1-5 (testifying that, even though "we see an extraordinary amount of variables in this case, ... you don't get to see everything").) Professor Arcidiacono acknowledged as well that some of the data that was unobservable to him "was, in fact, observable to the UNC admissions officers" and that "there will always be some" unobservables that are "incredibly strong." (Nov. 10 Trial Tr. at 309:9-19 (Arcidiacono).)
An example of an unobservable factor in this case would be a letter of recommendation that is considered as part of a candidate's file by an admissions reader but is not directly accounted for in Professor Arcidiacono's model because there is no concrete data point associated with it. (See Nov. 17 Trial Tr. 947:23–948:8 (Hoxby).) It is also unobservable whether a student with a high rating in personal qualities earned that rating by showing a strong sense of curiosity or by overcoming personal obstacles. (See ECF No. 155-4 at 8 (Reading Doc.).)
In addition to estimating the probability that an individual student would be admitted, the model also approximated the relative importance of the variables in the admissions process by providing a coefficient for each of them. (Nov. 9 Trial Tr. at 164:10–165:7 (Arcidiacono).) Coefficients with higher values indicated a higher level of positive correlation with admission than lower values, suggesting that they are more important in the admissions process. (See id. at 176:6-14.) While Professor Arcidiacono acknowledged that "[o]bviously, correlation does not imply causation," (Nov. 10 Trial Tr. 367:19-20 (Arcidiacono))—meaning that there might be patterns in the data that did not explain how an Admissions officer made his or her decisions—these coefficients nevertheless provided the basis for his subsequent quantitative evaluations.
Because the accuracy and fit of Professor Arcidiacono's model as well as the precision of his model's coefficients form the basis of his conclusions, the Court next looks more closely at these two issues.
2. Accuracy and Fit of Professor Arcidiacono's Preferred Model
a. Professor Arcidiacono's Model Predicts Admissions Decisions with a High Degree of Accuracy
One way to measure the accuracy of Professor Arcidiacono's model would be to see how often it correctly predicted the admissions outcome for each student in the applicant pool. (See Nov. 9 Trial Tr. 196:16–197:3.) To make this calculation, Professor Arcidiacono ran each candidate's data through the model to create a summary statistic he called an "admissions index." (Id. at 197:3-13.) After compiling these scores for every student in his data set, Professor Arcidiacono then sorted students from the top admissions index score to the lowest and filled each class in this way. (Id. at 198:10-20.) Then, he compared those admitted in this hypothetical scenario with the actual admissions decisions to see how often this method mirrored UNC's outcomes. (Id. at 199:5-9.) Using this method, Professor Arcidiacono's model predicted in-state decisions correctly about 92% of the time and out-of-state admissions about 93% of the time. (Id. at 200:1-5, 206:5-7.) According to Professor Arcidiacono, these numbers are "incredibly high" and "much higher that what [he is] used to seeing." (Id. at 208:24-25.) b. Professor Arcidiacono's Model is Overfit
Professor Hoxby found this method to be "nonstandard," "novel," and "ad hoc." (Nov. 18 Trial Tr. 1143:25–1144:3 (Hoxby).) She argues that a more typical way to make this calculation would be to use a "count statistic" where admitted students would be those with an admissions probability above 50%. (Id. at 1142:24–1143:17.)
Professor Arcidiacono also testified that his model fit the admissions data particularly well because it had a relatively high pseudo R-squared value. (See, e.g. , Nov. 9 Trial Tr. 191:14–192:3 (Arcidiacono).) In the end, however, Plaintiff appears to walk back any use of pseudo R-squared to measure the accuracy of a model. (See ECF No. 247 ¶¶ 183–84 (arguing that this metric would not "represent the predictive accuracy of the model" (citing Nov. 10 Trial Tr. 250:8–251:2 (Arcidiacono) and Nov. 18 Trial Tr. 1140:18–1141:16 (Hoxby)).) Even Professor Arcidiacono, when asked to explain the pseudo R-squared value of his model, pointed to how often his model "predicted" the correct admissions decision to define the meaning of the metric in this context. (Nov. 10 Trial Tr. at 250:17–251:2 (Arcidiacono) (testifying that a pseudo R-squared value of .727 was equal to 92% accuracy.) Accordingly, the Court places little weight in the pseudo R-squared values as a measure of accuracy.
Defendants argue, however, that Professor Arcidiacono was able to achieve such a high degree of accuracy because his model is overfit. An overfit model is one that includes so many variables in its calculations that it begins to pick up spurious relationships between those variables and admissions decisions that do not explain the process being modeled. (See, e.g. , Nov. 10 Trial Tr. 252:23–253:11 (Arcidiacono).) Such a model is therefore able to predict outcomes within the data sample it is based upon, but it would be less accurate when applied to data outside of that sample. In a hypothetical example, a model that was overfit might find that applicants whose mothers were born on a Monday have a much higher chance of being admitted to UNC than those whose mothers were born on a Tuesday. While such a pattern could be present in the sample data that is used to construct the model, it would not be expected to be indicative of how readers make their decisions, nor would this pattern be expected to continue to hold true when tested on out-of-sample data. In other words, this connection would only appear to make the model more accurate when in reality it would be responding to the idiosyncrasies of a particular dataset.
Though both experts agree that the degree of overfit is a factor they must take into account, the parties disagree on how to measure it. Professor Arcidiacono contends that an economist trying to create a model that is an accurate predictor of a process would want to have the lowest out-of-sample error possible. (Id. at 257:3-9 ("That's not the whole purpose of the models, but for the purpose of prediction, that's what you want.").) In assessing his models using this metric, he testified that his model more accurately predicted out-of-sample admissions decisions than did Professor Hoxby's. (Id. at 270:7–271:1; see also ECF No. 247-1 at 56 (Arcidiacono demonstratives).)
Professor Hoxby, on the other hand, argued that "the correct test[s] for a nonlinear model" are the Bilger and Manning tests and observed that Professor Arcidiacono did not complete this calculation. (Nov. 17 Trial Tr. at 971:7-14 (Hoxby).) In her opinion, one measures the accuracy of a model by "how well it performs out of sample compared to how well it performs in sample." (Id. at 971:24-25 (emphasis added).) In other words, it is a comparative measure that must take into account more than simply the accuracy of a model for any given sample.
She further contended that Professor Arcidiacono's method of comparing the percentage error for his model with the error in hers "makes no sense" because "all measures of whether a model is overfit have to do with this relative comparison for the same model in sample versus out of sample." (Id. at 973:2-23.) To focus purely on the low out-of-sample error would, in her words, be "just another way of saying he was maximizing [in-sample predictive power] even at the expense of showing a very, very inaccurate model." (Id. at 974:5-15.) Professor Hoxby ultimately testified that Professor Arcidiacono's preferred model is "somewhat overfit." (Id. at 975:25–976:3 ("It's certainly overfit, but it is not grossly overfit.").)
Though both experts provide helpful information on this point, they do appear to be answering slightly different questions. Whereas Professor Arcidiacono stresses the accuracy of his model, Professor Hoxby points out the tension between accuracy and overfit by emphasizing the uncontested point that an increase in in-sample predictive power may come at the cost of accuracy in describing the process writ large.
The Court accepts Professor Hoxby's conclusion that Professor Arcidiacono's model is somewhat overfit and therefore more susceptible to picking up spurious correlations in the data. Professor Hoxby's model, on the other hand, is more cautious in its approach as detailed below. This allows the Court to feel more confident that she has largely eliminated spurious relationships in her calculations.
c. There is No Evidence that UNC Has Implemented a Formulaic Process that Precludes Individualized Consideration of Each Candidate
With regards to whether UNC's admission process is formulaic, Professor Arcidiacono agreed in cross examination that he "could predict decisions very well," but not "that anyone told any admissions officer to try to follow that formula." (Nov. 10 Trial Tr. 319:3-6 (Arcidiacono).) Moreover, though he testified that he could "approximate [UNC's readers’] decisions very well," he acknowledged that any formula they were following is merely an implicit one. (Id. at 318:19–319:2.)
There are indeed patterns in the University's admissions decisions, but it appears that UNC's readers are, in Professor Arcidiacono's words, simply "paying attention to the components" of each individual application—such as GPA, extracurricular activities, and first-generation status—rather than using a formula to make decisions. (See id. at 318:24-25.) This is perhaps unsurprising given that UNC has exerted significant and continuous effort in communicating to its admissions readers what the University values and how the readers should evaluate the tens of thousands of candidates who apply annually.
In sum, the Court finds that while there are patterns in UNC's admissions decisions that map onto the values expressed by the University, there is no evidence that admissions officers are following an explicit formula or that they are automatically providing mechanical points to URMs nor applying consistent weights on the basis of a candidate's race. Further, though rote consideration of applications might be a necessary condition to demonstrate that UNC did not take into account all of the qualities of its candidates, mere calibration among the University's readers provides no basis to find that UNC neglected to provide the individualized approach to any admissions decision.
3. Professor Arcidiacono Computes Some Variables in Ways that Undermine His Conclusions
Having considered the degree to which Professor Arcidiacono's model accurately predicts which students are most likely to be admitted to UNC, the Court next turns to consider his contentions regarding the role of race within the larger admissions process. Before looking at the four calculations Professor Arcidiacono performed in this analysis, it is first useful to consider a few of the criticisms that Defendants have about this part of his work.
a. Imputation of GPA and Standardized Test Scores
Though the applicant data that UNC provided is immense, there are some cases where data is missing from individual applications. (See id. at 280:3-14.) In such cases, the experts had to determine how they would handle these issues in their modeling. For instance, as mentioned above, Professor Arcidiacono removed about 700 students from his sample whose applications listed at least one of their reader ratings as zero. (See ECF No. 247-1 at 3 (Arcidiacono demonstratives).) At other times, applications did not include GPAs or SAT scores. (Nov. 10 Trial Tr. 280:8-14.) In these cases, Professor Arcidiacono did not remove such students from his candidate pool but instead developed a method for imputing such scores. (Id. at 280:15–281:4.) The Court finds his method problematic because, in some cases, it resulted in data that is different from what the Admissions Office saw when evaluating these candidates and exacted what amounted to a penalty that ultimately exaggerated the impact of race in the decision-making process by mechanically lowering the scores of URM applicants.
Professor Arcidiacono described his method for filling in GPAs and SAT scores as "letting the model decide." (Id. at 281:3-4 (Arcidiacono); see also Nov. 9 Trial Tr. 177:24–178:23 (Arcidiacono).) For GPA, the way this worked in practice was that Professor Arcidiacono's model first calculated the average GPA for each race of admitted students; when this metric is missing from a student's application, the candidate is assigned the average for their race. (Nov. 10 Trial Tr. at 281:5-13.) According to Professor Arcidiacono, "that number can be different for different races." (Id. at 281:13-14 (Arcidiacono).)
As detailed above, Professor Arcidiacono found that URM applicants "disproportionately have lower test scores and grades" than their white and Asian American counterparts. (Nov. 9 Trial Tr. 150:12–151:1 (Arcidiacono).) These patterns were particularly pronounced for African American students: Professor Arcidiacono testified, for instance, that more than half of African American applicants were in the lowest two deciles of his AI. (Id. at 155:7-10.) Therefore, it is straightforward to conclude that these GPA imputations mechanically assign lower scores to URM students than they do to white or Asian American students when data is missing.
The Court finds Professor Arcidiacono's treatment of students missing SAT scores to be even more troubling. As background, when students apply to UNC with only an ACT score, the Admissions Office uses a concordance table published by the College Board to convert the ACT score to an SAT scale. (Nov. 13 Trial Tr. 694:18–695:11.) Researchers at the ACT and SAT published this concordance to provide stakeholders with a way of understanding how the two scores compare. (Id. at 695:7-11.)
Instead of using this concordance, Professor Arcidiacono chose a different method, letting the model decide what the appropriate conversion should be based on the race and gender of the candidates. (Nov. 10 Trial Tr. 351:2–352:12, 360:6-14.) In other words, instead of simply translating an ACT score to an SAT equivalent and putting into the model the score that a UNC admissions reader would evaluate, Professor Arcidiacono adopted a more complicated method that takes into account an average score based on the candidate's race and gender. While Professor Arcidiacono testified that "[i]t is not clear that [this method] disfavors" URM candidates, (id. at 352:11-12 (Arcidiacono)), it appears that, given the model's regression to the mean, the only time it would favor such candidates would be when an applicant had particularly low scores such that they were lower than the average URM student when this conversion would increase a score which was below the mean. With the low rate of URM admissions in mind, however, it is not a stretch to recognize that any gain such students might take from the process occurs in score ranges where candidates are typically not admitted and—more to the point—would still operate as much less of a bump than equivalent scores would for non-URM students.
Therefore, an Asian American student taking only the ACT would get a positive "bump" of about 9 points on his converted SAT score while a Hawaiian/Pacific Islander candidate would receive a 50-point deduction in their converted score. (Id. at 359:13-17, 395:22–396:6.) In short, as Professor Arcidiacono testified, if a student scored a 32 on the ACT, they "would be better off being white" in his model given that it would automatically impute a higher score for a non-URM student despite the admissions office seeing the exact same converted score for both candidates. (Id. at 355:19-25.)
The evidence is clear that higher standardized test scores are positively correlated with admission, and therefore this method again exaggerates the ground that URM students must make up to be competitive in UNC's admissions process. The scope of this issue is not particularly clear, and Professor Arcidiacono was not familiar enough with the data he used in his model to further clarify the magnitude of this issue. Defendants’ counsel asked him whether he was aware that "in the course of your process over the six years that the folks who only had ACT scores numbered more than 50,000?" (Id. at 361:4-6 (Fitzpatrick).) Professor Arcidiacono replied that he was "not sure." (Id. at 361:7 (Arcidiacono).) Defendants’ counsel next asked if he was aware that, for the year 2018, he converted ACT scores for about 42% of African American applicants and 45% of Hispanic applicants. (Id. at 361:8-10, 13-14 (Fitzpatrick).) Again, Professor Arcidiacono testified that he did not "have any evidence of that on either account," (id. at 361:11-12, 15-17 (Arcidiacono)), despite the importance of this factor in his model.
Professor Arcidiacono's decision to convert ACT scores to SAT scores by using a regression to the racial mean is very concerning to the Court. While the missing GPAs are unknown, how UNC considers known ACT scores is not. In sum, Professor Arcidiacono eschewed a simple calculation that is actually used by UNC's Admissions Office and replaced it with one that exaggerates the difference in academic credentials between URMs and non-URMs. The Court finds that this decision undermines Professor Arcidiacono's ultimate conclusions because it put a thumb on the scale and mechanically departed from the process it attempted to model.
Professor Arcidiacono testified that he "[broke] out the SAT score by math and verbal" in his model and "you can't use the concordance table for that." (Nov. 10 Trial Tr. 377:4-6 (Arcidiacono).) This statement, however, is unavailing. For one, the Court takes judicial notice that the concordance table published by the College Board—which Mr. Rosenberg testified that the University uses—does indeed break out scores by section. (See College Board, Guide to the 2018 ACT/SAT Concordance , https://www.act.org/content/dam/act/unsecured/documents/ACT-SAT-Concordance-Information.pdf ). Moreover, even if Professor Arcidiacono had difficulty locating or using the subject-area concordance, it does not address the problematic fact that he still opted to create his own system for converting scores rather than mirroring the simple system used in practice by UNC's Admissions Office. Prioritizing the impact of subject-area breakdowns—which are not at issue in this case and were not found to be meaningful by either party—over replicating the University's process is not a justifiable choice.
b. Reader Ratings
The next criticism that Professor Hoxby had for Professor Arcidiacono's model is that it includes as variables each of UNC's reader ratings. In her own model discussed below, Professor Hoxby excluded these ratings because they are "evaluative variables[,] or what an economist would call ‘endogenous variables.’ " (Nov. 17 Trial Tr. 942:8-10 (Hoxby).) This means that "the variables are determined within the process itself." (Id. at 942:10-12.) Those ratings, in other words, do not come from a candidate's application but rather show a subjective, intermediate assessment that readers use as a way of summarizing their impressions. Professor Hoxby testified that these are not "verifiable objective factors," and, in her words, "statisticians are trained not to put that kind of variable into a multiple regression." (Id. at 944:1-3.)
Though at trial Professor Arcidiacono did not generally address the criticism of including intermediate evaluative variables in his model, he did testify that economists use subjective factors in their modeling "[a]ll the time." (Nov. 9 Trial Tr. 174:16-19 (Arcidiacono).) He stated that whether it is "correct" to include this rating "depends on what you're trying to do." (Nov. 10 Trial Tr. 291:10-12.) "[I]f you're trying to figure out how formulaic UNC's admissions process is," he testified, "then you need to include all the components of the formula, and the personal rating is one of those components." (Id. at 291:12-15.) On the other hand, "[f]or the purposes of figuring out how big racial preferences are, you would not want to include it because part of the effect of racial preferences would then be operating through" the rating system. (Id. at 291:17-20, 292:1-2 (contending that including the personal rating will "make the preferences appear smaller than they really are").)
This statement seems to imply that there is some tension between answering Professor Arcidiacono's first question of whether or not the process is formulaic and his second question regarding what role race plays in it. That said, the Court finds that the data Professor Arcidiacono provides tends to show that the inclusion of the personal rating, in his words, "doesn't matter that much," (id. at 292:19-24), and there is no evidence that the inclusion of this metric made a significant impact on Professor Arcidiacono's ultimate conclusions regardless of whether or not endogenous variables are generally appropriate in econometric modeling.
c. Shapley Decomposition
While Professor Hoxby's first two criticisms address inputs into Professor Arcidiacono's model, her third and final criticism speaks to how he calculates the "shares" of each variable or, put another way, how he describes the degree to which certain factors matter in the admissions process. More specifically, Professor Hoxby testified that Professor Arcidiacono should have used what is known as the Shapley decomposition to consider the impact of the consideration of race.
According to Professor Hoxby, the Shapley decomposition is a statistical method that allows an economist to "demonstrate which factors are playing an important role in a model" and is "really the only decomposition method that is accepted." (Nov. 17 Trial Tr. 948:19-21 (Hoxby).) It "decomposes" pseudo R-squared values into buckets or shares for various factors. (Id. at 948:21–949:3.) For instance, Professor Hoxby testified that "race and ethnicity might be in one bucket, and test scores might be in another bucket, high school GPA, class rank could be in yet another bucket." (Id. at 949:2-6.) If a variable plays a dominant role within UNC's admissions process, Professor Hoxby testified that the Shapely decomposition would "[a]bsolutely" reveal its weight because it is "designed to show the marginal effect of any factor reliably." (Id. at 949:15-20.) She further stated that, if the factor were important "even for a subset of applicants," the Shapley decomposition is designed to show this effect. (Id. at 949:20-23.) When Professor Hoxby performed a Shapley decomposition on Professor Arcidiacono's preferred model, she found that, for in-state students, "2.7% of admissions decision was accounted for by race and ethnicity" and, for out-of-state students, 6.7% are accounted for by this factor. (Id. at 955:14-20; see also ECF No. 251-1 at 14 (Hoxby demonstratives).)
Though Professor Arcidiacono disagrees with the use of the Shapley decomposition in this case, Professor Hoxby's actual calculations using this method are uncontested. (See Nov. 10 Trial Tr. 251:3–252:12 (Arcidiacono).)
Professor Arcidiacono, however, testified that using the Shapley decomposition was "conceptually wrong" in this instance. (Nov. 10 Trial Tr. 251:3-7 (Arcidiacono).) He testified that Professor Hoxby is "thinking about how race affects the entire admissions process, and I don't think that's the right way to look at it." (Id. at 251:8-10.) Instead, Professor Arcidiacono stated that he believes one needs to "be thinking about how race matters for the groups that are affected by that." (Id. at 251:10-12.) He summarized this position by agreeing that Professor Hoxby's Shapley decomposition "spread[s] the effect of UNC's racial preferences across those tens of thousands of applicants each year." (Id. at 252:9-12.)
In sum, both experts therefore agree that the Shapley decomposition is a useful calculation in certain instances, but just disagree as to whether it should be applied in the instant case. Based on this testimony, the Court finds that applying the decomposition allows one to see how race impacts the full admissions process. Accordingly, the Court finds that, across the admissions process, race explains only 2.7% of in-state admissions decisions and 6.7% of out-of-state admissions in Professor Arcidiacono's preferred model.
4. Professor Arcidiacono's Quantitative Analysis Fails to Show that Race is a Predominant Factor
Having considered Professor Arcidiacono methods in creating and calculating his preferred model, the Court next looks at the four ways in which he sought to use the model to quantify the effect of race. Those four calculations were (1) a transformational analysis, (2) the average marginal effect of race, (3) an admitted URM analysis, and (4) a calculation that considered each class's capacity constraints. Even setting aside for now the fact that Professor Arcidiacono's calculations are undermined by its imputation of missing standardized test scores, the Court finds that none of these analyses demonstrate that race is used as anything other than a "plus" factor in UNC's admission process. On the contrary, the Court finds that the evidence tends to show that race is not a predominant factor in the University's candidate evaluations.
a. Transformational Analysis
The first type of calculation that Professor Arcidiacono completed is one he termed a "transformational analysis." (Nov. 9 Trial Tr. 210:12-21 (Arcidiacono).) In this calculation, he began by creating a hypothetical non-URM student with a specific probability of admission in his preferred model. (See id. at 213:12-17.) From there, he "flipped on the racial preference" for the candidate—theoretically turning them into an African American or Hispanic student—and then considered how the calculated probability of admission changed. (Id. at 214:25–215:11.) He completed this computation for eight hypothetical white students, disaggregated by gender, first-generation status, and state residency. (See ECF No. 247-1 at 35–36 (Arcidiacono demonstratives).) In general, this calculation resulted in what Professor Arcidiacono termed a "substantial bump" for African American applicants and a slightly smaller one for Hispanic candidates. (Nov. 9 Trial Tr. at 190:11-13 (Arcidiacono); see also ECF No. 247-1 at 26–27 (Arcidiacono demonstratives).)
Though Professor Arcidiacono presented to the Court only examples of students with a 10% or 25% chance of admission initially, (see ECF No. 247-1 at 35–36 (Arcidiacono demonstratives)), he testified that this calculation would work for any percentage of probability, (Nov. 9 Trial Tr. 214:14-15 (Arcidiacono)).
While the expected changes in probabilities do show some meaningful effects, there are significant limitations in this analysis. For one, Professor Arcidiacono does not produce any analysis that "flips off or on" any other variable to allow the Court to compare the impact of race with other factors in the admissions process. For instance, Professor Arcidiacono does not demonstrate whether raising or lowering a student's GPA has a greater or lesser effect in order to contextualize how sensitive his model might be to changes in variables. Additionally, this analysis also relies on the premise that Professor Arcidiacono's model can accurately predict not simply admissions decisions but is also precise enough to accurately predict the probability of an individual candidate's admission. Because a candidate's probability of admissions is never calculated throughout UNC's process and Professor Arcidiacono only has access to the binary results of the admit/reject decisions, whether or not he can calculate probabilities with precision is not only unproven, but untestable given the current data set.
A more obvious limitation, perhaps, is the telling phrase Professor Arcidiacono's uses to describe his process. He stated that in calculating a hypothetical student's probability of admission, that student is "treated as" an URM. (Nov. 9 Trial Tr. 210:12-21 (Arcidiacono).) By this, he means that every other factor in the person's admissions file would be the result of the student having the lived experience of growing up as a URM, but that this context would go entirely unrecognized in a holistic admissions process. In other words, this calculation renders unobservable one of the most contextual factors in a student's lived experience. As Professor Arcidiacono testified, to complete this analysis one must be "operating more in a color-blind world" despite acknowledging that "there's no question" that the "experiences of African Americans are different from the experiences of white Americans and of other groups." (Nov. 10 Trial Tr. 365:25–366:9, 366:25–367:5 (Arcidiacono).)
Professor Hoxby points out that there are a number of statistical problems with this analysis as well and, to illustrate, she described how it differed from a randomized test within a medical context that would be, in her estimation, the correct use of a transformational analysis. (See Nov. 17 Trial Tr. 963:5–965:24 (Hoxby).) In a randomized drug trial, 10,000 people might be given the drug being tested and 10,000 people might be given a placebo. (Id. at 963:9-13.) "It's important in this context," Professor Hoxby testified, "that people be randomly assigned either to the real drug or to the placebo" because, under the law of large numbers, it would be fair to assume that those getting the real drug and those getting the placebo "are the same on all other characteristics on average." (Id. at 963:14-22.)
However, as Professor Hoxby observed, UNC's admissions process "is not a randomized control trial." (Id. at 964:8-12.) The data are not generated in an experiment, but instead by "real behavior of real people" and therefore "all other things are not held constant." (Id. at 964:12-17.) Further, given that there is significant evidence before the Court that the two groups being compared—URMs and non-URMs—do not have all of the same characteristics on average, (see ECF No. 247-1 at 6–7 (Arcidiacono demonstratives) (describing URMs scoring lower than white and Asian Americans candidates on UNC's program rating, being more likely to be a first-generation college student, and being much more likely to qualify for a fee waiver)), Professor Hoxby contends that a transformational analysis is ill-equipped to demonstrate a meaningful effect.
The Court accordingly places little weight on this analysis. Given the testimony of the parties’ expert witnesses acknowledging the differences between the lived experiences of white applicants and URM applicants, changing a single variable without accounting for that variable's impact makes this analysis little more than a theoretical exercise that fails to accurately represent the admissions process as described in uncontested testimony.
b. Average Marginal Effect
Professor Arcidiacono next looked at the average marginal effect of race in the admissions process. In this analysis, he considered "the full set of minority applicants" and considered "what would happen to their probabilities of admission" if they were instead "treated as white applicants." (Nov. 9 Trial Tr. 211:4-10 (Arcidiacono).) In practice, this means that he looked at the admissions probability produced by his model for each African American and Hispanic candidate and averaged those probabilities together to produce a starting point. (See ECF No. 247-1 at 38 (Arcidiacono demonstratives).) He then "turned off" the coefficient for these races and produced and averaged a new set of admissions probabilities for the same set of students. (Id. ) The difference between those two numbers produced what he refers to as the average marginal effect of race. (Id. at 38–39.)
For African American students, Professor Arcidiacono testified that removing his coefficient for race would reduce the average probability of admissions for in-state candidates from 31% to 18% and from 17% to 2% for out-of-state candidates. (Id. at 39.) For Hispanic students, the probability of admissions would be reduced from 41% to 31% in state and from 20% to 6% out-of-state. (Id. ) Professor Arcidiacono then compared the before-and-after percentages for each group and extrapolated that the "bump" his model predicts for race is responsible for 42% of in-state African American admissions and 24% of Hispanic admissions. (Id. at 40.) For out-of-state students, Professor Arcidiacono testified that it accounts for 91% and 70% of African American and Hispanic admissions, respectively. (Id. at 41.)
An example of this calculation may be useful here. On average, according to Professor Arcidiacono, in-state, African American students have a 30.5% chance of being admitted under UNC's current system. (ECF No. 247-1 at 40 (Arcidiacono demonstratives).) If one "turn[s] off" the racial coefficient in his model, the probability of admissions drops 12.7% to a new average probability of 17.8%. (Id. ) That difference of 12.7% is equal to about 42% of the original 30.5% admissions probability. In other words, Professor Arcidiacono testified that, of the actual 30.5% average admission probability, 17.8% is due to non-racial factors and 12.7% is due to a racial bump. (See id. )
Professor Hoxby, however, testified that such numbers are misleading. Among other things, she stated that median marginal effect is a more appropriate way to describe how race is used in the admissions process than average marginal effect. Using the average, as Professor Arcidiacono does, "grossly exaggerates the role of outliers" in the data. (Nov. 17 Trial Tr. 968:1-7 (Hoxby).) As an example, Professor Hoxby describes a hypothetical situation where Professor Arcidiacono's model alters a student's race from African American to white and where the model shows the student's probability of admissions going from 10% to 90%. (Id. at 968:9-15.) In such a case, one student would offset eighty other students whose probability of admission dropped by just 1% in the model's calculation. (Id. at 968:15-17.) In this hypothetical scenario, the median marginal effect would be 1%, and one could surmise that the average is "dominated by outliers." (Id. 968:22–969:3.)
Professor Hoxby testified that this is exactly what one sees in Professor Arcidiacono's analysis. Her calculations showed that the median marginal effect for removing race from his preferred model ranged from 0.6% to 2.0% depending on the group being assessed. (ECF No. 251-1 at 18.) This suggests that, while there might be a set of students who would see a sizable "bump" in their admissions probability in the model because of their race, a typical student would see a very minimal difference in their odds of being admitted.
In this case, the Court need not choose between the average effect and the median effect. Both measures describe different properties of the data which, together, allow the Court to gain a fuller understanding of how the racial preferences operate in Professor Arcidiacono's model. To that end—and leaving aside for now the Court's concern regarding the strength of the model itself and crediting Professor Arcidiacono's untestable admissions probabilities—the Court finds that the consideration of race does appear to be a tipping point for some applicants according to Professor Arcidiacono's model given the relatively high average marginal effect. However, the Court also finds that those affected in the model must be contained to a small group of outliers given the much smaller median marginal effect. Accordingly, it ultimately finds that these measures appear to confirm UNC's descriptions of its admissions process in that race may operate as a tipping point for a small number of students and does not provide evidence that race is a predominant factor in its evaluation of candidates.
c. Admitted URM Analysis
The third analysis that Professor Arcidiacono conducted was what he called an admitted URM analysis. (Nov. 9 Trial Tr. 211:18-20 (Arcidiacono).) In this calculation, he started with all of the URM students that UNC actually admitted. (Id. at 211:20-25.) Then Professor Arcidiacono worked backwards and calculated each candidate's probability for being admitted under his model if they were treated as a white student. (Nov. 10 Trial Tr. 238:7-17 (Arcidiacono).) In this analysis, Professor Arcidiacono testified that the average admitted African American candidate would have, in his model, a 58% chance of being admitted without racial preferences in-state, and only a 9% chance of being admitted if they were a nonresident. (ECF No. 247-1 at 45 (Arcidiacono demonstratives).) An average admitted Hispanic student under this analysis would have a 76% chance of being admitted in-state and a 29% chance of being admitted out-of-state. (Id. )
Professor Arcidiacono further testified that 43% of admitted, in-state African American candidates and 95% of admitted, out-of-state candidates would have had, under his model, less than a 50% chance of being admitted without racial preferences. (Id. ) For Hispanic applicants, 22% of admitted, in-state applicants and 78% of admitted, out-of-state applicants would be projected as having a less than 50% chance at being admitted to UNC. (Id. )
The Court finds that this calculation again relies entirely on the precision of Professor Arcidiacono's admissions index calculating the percentage chance that a student would be admitted to the University. As discussed above, this index is not only unproven but untestable and ultimately does not provide a basis for the Court to find that fluctuations in the average URM admissions rate are meaningful or suggest a process at odds with the one credibly and consistently described by the UNC Admissions Office.
d. Capacity Constraints
The fourth and final way that Professor Arcidiacono attempted to quantify the effect of racial preferences in UNC's admissions process was to conduct a capacity constraints analysis. Here, Professor Arcidiacono "turn[ed] off" all racial preferences and calculated new probabilities for all candidates. (Nov. 10 Trial Tr. 242:22–243:14 (Arcidiacono).) He then reordered the applicant pool according to those numbers and built six new incoming classes with students the model estimated would have the highest chance of admissions. (Id. ) Professor Arcidiacono testified that, over the course of the six admissions cycles he studied under this model, UNC would have admitted 2,948 more white students and 709 more Asian American students. (ECF No. 247-1 at 50 (Arcidiacono demonstratives).) On the other hand, it would have admitted 2,239 fewer African American students and 1,341 fewer Hispanic students. (Id. )
To put those numbers in perspective, the number of students whose admissions decisions Professor Arcidiacono suggests were affected by race add up to about 4.4% of UNC's applicants during this time period. Leaving aside for now the Court's finding that the underlying data Professor Arcidiacono used in his model mechanically overemphasizes race, the Court finds that this low percentage of affected applications appears to confirm the idea that the use of race cannot serve as a predominant factor in UNC's admissions process.
5. Professor Hoxby's Regression Model Demonstrates that Race is Not a Predominant Factor in UNC's Admissions Process
The Court next considers Professor Hoxby's regression model. As discussed above, Professor Hoxby sought to answer whether Plaintiff's allegations that race and ethnicity were dominant factors in UNC's admission process were true. (Nov. 17 Trial Tr. 936:10-14 (Hoxby).) After "careful empirical analysis," she concluded that "UNC's admissions decisions appeared to be fully consistent with the holistic admissions process" and that "UNC's process could not be explained by a formula based on verifiable variables." (Id. at 937:3-7.) "In other words," she testified, "UNC's process does not appear to be formulaic." (Id. at 937:11-12.) She additionally found that "race and ethnicity are not dominant factors in the UNC admissions process." (Id. at 937:15-17.)
Professor Hoxby reached this conclusion by building nine different models of her own. (Id. at 939:9-14.) She testified that, if race were a dominant factor, the data would show at least two results. (Id. at 938:20-22.) The first would be that the multiple regression model could "explain most of the decision between admissions and rejection." (Id. at 938:22–939:1.) The second result would be that, once she "decomposed the explanatory power of the regression model" using the Shapley decomposition, "a lot of the variation would be explained by race and ethnicity." (Id. at 939:4-7.) Professor Hoxby testified that she selected the variables in her model with an eye towards attempting to "replicate or understand the behavior of actual human beings" which she testified is "the goal" of all of her models. (Id. at 941:3-6.) To that end, she asked herself what an admission officer sees or considers in their evaluation of candidates, (id. at 941:6-8), and her final preferred model incorporated nearly all of the same variables that Professor Arcidiacono's did, (see ECF No. 251-1 at 7 (Hoxby demonstratives)). The only variables she did not include were UNC's reader ratings, a control for the year in which a student applied, and an applicant's intended college major. (Compare id., with ECF No. 247-1 at 24 (Arcidiacono demonstratives).) Professor Hoxby also did not use variables in which multiple inputs interacted with each other such as gender and race. (See ECF No. 251-1 at 7 (Hoxby demonstratives).)
In creating her econometric models, Professor Hoxby's used only the four most recent years of applicant data instead of all applications in the six-year period. (Nov. 18 Trial Tr. 1122:4-19 (Hoxby).) Though Plaintiff contends it was "error" not to incorporate all of the possible data into her model, (ECF No. 247 ¶ 174), it provides no basis for this assertion. Professor Arcidiacono, for his part, only points out this difference between the models when he is contending that "more data lessens the risk of you being overfit." (Nov. 10 Trial Tr. 257:19-25 (Arcidiacono).) At no time did he suggest that this decision had any impact on Professor Hoxby's findings. Accordingly, the Court finds that there is no evidence suggesting that there is any distinction in the value between these two datasets based solely on their size.
a. Explanatory Power of Professor Hoxby's Model
While the nature of the pseudo R-squared metric is not a significant part of Professor Arcidiacono's analysis, it plays a greater role in Professor Hoxby's conclusions. This metric generated a lot of discussion in trial and in subsequent briefing primarily because the experts fundamentally disagreed on how to interpret it.
As background, the experts testified that there is a foundational measure referred to as R-squared and a separate metric—or set of metrics—called pseudo R-squared. (Nov. 10 Trial Tr. at 249:2-6 (Arcidiacono).) According to Professor Arcidiacono, the metric R-squared is used in "a normal regression ... where the outcome is continuous," such as earnings. (Id. at 249:8-11.) One can use R-squared to assess "what share of the variation [in the data] is explained" by the variables one has used to build their model. (Id. at 249:8-15.) For example, one might be able to calculate the effect of a college degree on one's earnings. There are, according to Professor Arcidiacono, "many different kinds of pseudo R-squareds," and each is "trying to figure out some way of relating what we see with the R-squared to models ... that are not linear , where the outcome is discrete. " (Id. at 249:16-20 (emphasis added).) Professor Hoxby described the metrics in largely the same way. (See Nov. 17 Trial Tr. 944:17-21, 945:2-5 (Hoxby).)
The parties disagreed, however, on the degree to which these two measures are actually similar. Professor Hoxby testified that that "pseudo R-squared is designed to be analogous as possible to R-squared," with the only difference being the type of model to which it should be applied. (Id. ) Professor Arcidiacono testified, on the other hand, that "what's true about the R-squared and the pseudo R-squared is [that] higher values mean that you're fitting the data better." (Nov. 10 Trial Tr. at 249:21-23 (Arcidiacono).) That said, according to him, "you can't say anything more than that in terms of ... how much of the variation you're actually explaining by using the pseudo R-squared." (Id. at 249:23–250:1 (emphasis added).) In other words, Professor Arcidiacono contended that while a calculation with an R-squared value of 0.5 would suggest that a factor explains about 50% of the variation in the data, a model with a pseudo R-squared value of 0.5 would have no such meaning, nor even a natural interpretation. (Id. at 250:2-12.)
Professor Hoxby testified that the pseudo R-squared of her preferred model is 0.428, "which suggests," in her words, "that the model explained about 42.8%" of the admissions decisions. (Nov. 17 Trial Tr. 946:24–947:7 (Hoxby).) She emphasized that this calculation also shows that the remaining 57.2% of admissions decisions are therefore explained by unobservables that a statistician or econometrician would not be able to explain. (Id. at 947:15–948:10.) Under cross-examination, Professor Hoxby acknowledged that this type of description is an oversimplification to some degree, and "meant to be a kind of intuitive definition for a reader who isn't technical." (Nov. 18 Trial Tr. at 1138:13-20 (Hoxby); see also id. at 1136:3-6 (referring to the description as a "rough interpretation" in "colloquial English").) She testified, however, that "both concepts are trying to see how much more you can explain by fitting a model with explanatory factors rather than using the null model." (Id. at 1135:25–1136:3.)
By comparison, Professor Arcidiacono's model had a pseudo R-squared of 0.727. (Nov. 10 Trial Tr. 250:17-20 (Arcidiacono).)
In attempting to draw a conclusion from these conflicting expert opinions, the Court finds, at a high level, that Professor Hoxby's model reinforces the idea that there are significant parts of the decision-making process that are unobservable to any economist attempting to simulate the process through an econometric model. Though it seems that the Court should take Professor Hoxby's testimony of the exact percentage that is unobservable as more of a rough estimate than a precise conclusion, it is uncontested by both experts that their models demonstrate that there are meaningful parts of UNC's holistic admissions process that are not able to be explained simply through the observable, discrete factors available to them.
b. Magnitude of Racial Preferences
Professor Hoxby next considered the role that race plays in UNC's decisions. After completing a Shapley decomposition, she concluded that race and ethnicity account for only 1.2% "of the total admissions decision." (Nov. 17 Trial Tr. at 951:7-11 (Hoxby).) By comparison, she found that ACT and SAT test scores explain 9.8% of the admissions decisions and are, in comparison to race and ethnicity, "a more important factor ... by several times." (Id. at 952:1-9.)
Professor Hoxby then disaggregated these findings. For instance, she separated students by their state of residency and found that race explained about 1.2% of in-state admissions decisions and 5.1% of out-of-state decisions. (Id. at 953:2-15.) She found, at the same time, that 15% of in-state admissions were explained by test scores while 18.9% of out-of-state admissions were the result of this factor. (Id. ) Using the Shapley decomposition on other variations and on Professor Arcidiacono's model, she found that race never explained more than 6.7% of the admissions decision for any subgroup she tested. (See ECF No. 251-1 at 13–15 (Hoxby demonstratives).) She concluded that "[r]ace simply cannot be a dominant factor in the UNC admissions process because it plays a minor role regardless of which model one uses." (Nov. 17 Trial Tr. at 957:5-14 (Hoxby).)
6. Conclusions
Before summarizing its findings, the Court first reiterates that there are factors considered in UNC's admissions process that are not included in any of the experts’ modeling. These unobservable factors are in some cases as straightforward as letters of recommendation or a special talent in music, drama, athletics, or writing. (See ECF No. 155-4 at 2, 7 (Reading Doc.).)
There are also a number of factors that are accounted for in only a very imprecise way, such as the forty individual criteria described in UNC's Reading Document. This imprecision is most impactful when a criterion is distinct from race, but highly correlated with it, such as the obstacles that a candidate has overcome as a result of racial prejudice. In such a case, an econometric model that takes race into account but measures only obliquely those obstacles might easily attribute to the former what in reality is a consequence of the latter. Given the data available to the experts, both models in this case necessarily fit such a description. This does not take away from the meaningful information that these statistical models provide. That said, identifying a way in which the calculations before the Court might overemphasize the use of race as a discrete factor merely clarifies the limits of these highly complex formulas’ explanatory powers.
As discussed throughout this section, the Court finds that there is no evidence before it that the University's admission process departs in any way from the procedures described in the Reading Document. In sum, it appears that race—or unobservable factors highly correlated with race—may play a determinative role for a small number of URM students. To be sure, "[t]he fact that race consciousness played a role in only a small portion of admissions decisions should be a hallmark of narrow tailoring, not evidence of unconstitutionality." Fisher II , 136 S. Ct. at 2212. Any such students appear to be outliers, and there is no statistical evidence before the Court that there are any individual students for whom race has been the defining feature of their application. When considered as part of the admissions process writ large, the Court credits Professor Hoxby's finding that race plays a role in a very small percentage of decisions: 1.2% for in-state students and 5.1% for out-of-state students. Even assuming some of Professor Arcidiacono's imputations accurately reflected UNC's process, his analysis showed that fewer than 5% of admission decisions would be affected by the use of race over the course of the six-year period the experts analyzed. Further, race appears to be less than or equally important to several other data points considered within a holistic process.
VII. FINDINGS OF FACT: UNC HAS MADE A SERIOUS, GOOD FAITH EFFORT TO CONSIDER RACE-NEUTRAL ALTERNATIVES
The Court next considers whether UNC has engaged in serious, good faith consideration of potential, workable race-neutral alternatives ("RNAs") that would allow it to achieve its goal of diversity about as well as a race-conscious admissions process. In addition to contending that UNC has not put forth sufficient effort to meet this standard, Plaintiff argues that RNAs are available and has identified four distinct practices that UNC could undertake to reach this aim: (1) improving its recruitment process to reach additional highly qualified URMs; (2) recruiting more community college students; (3) eliminating early action admissions and legacy preferences; and (4) changing how it evaluates and selects candidates for admission.
For the reasons described below, the Court finds that UNC has engaged in serious, good faith consideration of potential, workable RNAs and has already implemented several of the most promising ideas. Moreover, the University engages in ongoing periodic reviews to assess both the validity of emerging strategies and the extent to which UNC's efforts are achieving the educational benefits of diversity that are consistent with the critical mass it seeks.
Through an exhaustive exploration of RNAs that have been implemented and others that are purely theoretical, the University has shown that there are not any available, workable, or sufficient RNAs that would allow it to achieve its diversity goals. Accordingly, the Court finds that UNC has satisfied its burden of demonstrating that there is no non-racial approach that would promote such benefits about as well as its race-conscious approach at tolerable expense.
A. The University Has Long Considered Race-Neutral Alternatives
The Court first considers the extent to which UNC has already engaged in "serious, good faith consideration of workable race-neutral alternatives." (ECF No. 247 ¶ 230 (citing Grutter , 539 U.S. at 339, 123 S.Ct. 2325 ).) Though the Court has addressed this question above in many ways, it now outlines the University's examination of RNAs over the last several years.
Within the first month of becoming UNC's Admissions Director in 2004, Mr. Farmer began working with the College Board's Access & Diversity Collaborative. (Nov. 12 Trial Tr. 589:24–590:3.) This group brings together policymakers, legal experts, and college administrators to "talk about how [they] could do a better job," discuss the legal landscape, and "hear what other schools [are] doing and how well their tactics [are] working." (Id. at 590:4-9 (Farmer).) UNC administrators have continued to remain involved in those meetings in the seventeen years since, and the University "pay[s] attention to their documents about race-neutral alternative and other practices that other schools were putting into place." (Id. at 590:10-14.)
In 2007, the Admissions Office completed a study that evaluated whether indicators of socioeconomic disadvantage could be used in lieu of race in the admissions process to yield a class with academic credentials and racial diversity similar to those of the actual admitted class. (Id. at 590:16-25.) Mr. Farmer testified that, after reviewing the results of this study, he did not believe that this approach was viable as a RNA because it led to a "substantial decline in the population of underrepresented students." (Nov. 12 Trial Tr. 595:8-12 (Farmer); Nov. 13 Trial Tr. 663:10-16 (Farmer).)
In 2009, at the direction of Mr. Farmer, Ms. Kretchmar conducted a literature review on how banning racial considerations in the admissions process affected other institutions. (Nov. 12 Trial Tr. 595:15-17 (Farmer).) Mr. Farmer testified that "the literature review strongly demonstrated that schools like [UNC] had not found race-neutral alternatives that worked well." (Id. at 596:14-17.)
In 2012, Mr. Farmer and Ms. Kretchmar evaluated what impact a race-neutral admissions strategy designed to admit the top students from every N.C. high school would have if it were implemented at UNC. (Id. at 596:20-23.) For in-state students, Mr. Farmer found that there would be an increase in URM students from 15% to 16%, but that "every academic indicator, other than the share of class ranking in the top 10%, would have declined." (Id. at 597:4-12 (observing that the average SAT score of the class dropped from 1317 to 1262).) Mr. Farmer testified that "although a difference of 60 points between two people might not be a significant difference, a difference of 60 points across a population of 30-some hundred is actually a very significant difference." (Id. at 598:7-20.)
This particular admissions policy is described and discussed in much more depth below.
In September 2013, Mr. Farmer convened the Race-Neutral Alternative Working Group chaired by Barbara Polk, the then-Deputy Director of Admissions. (Id. at 618:17–619:17.) This group consisted of members of faculty and staff from across the University, (Id. at 620:23–621:2), and it met and discussed multiple options, reviewed relevant literature and commentary, and sought to understand other institutions’ experiences with implementing such programs, (Nov. 9 Trial Tr. 96:23-24, 96:17-22, 97:2-6; Nov. 13 Trial Tr. 621:2-11).
The working group modeled the impact of five types of RNAs: (1) an approach that used the strength of a student's high school curriculum and a standardized testing threshold as criteria for admission; (2) a top 10% plan; (3) a top 4.5% plan; (4) a percentage plan that guaranteed admission to top-ranked students while also admitting more students from low-income high schools; and (5) an approach that would guarantee admission to students with standardized test scores above a certain threshold. (Nov. 13 Trial Tr. 630:7-12.) Each alternative the working group analyzed resulted in a decline in racial diversity or a decline in academic quality, or both. (Nov. 13 Trial Tr. 631:1-8; DX045 at 17–20.)
In February 2016, UNC's Advisory Committee convened a Committee on Race-Neutral Strategies ("CRNS"). (Nov. 9 Trial Tr. 105:12-14, 106:1-6.) The committee, which continues to operate, is comprised of faculty and administrators who have expertise in various fields relevant to exploring RNAs, including diversity and inclusion, data integration and analysis, machine learning, modeling, causal inference, student affairs, and undergraduate admissions. (Nov. 16 Trial Tr. 823:22–825:7.) CRNS was charged with: (1) considering whether there are workable race-neutral strategies and practices that the Admissions Office could employ in evaluating applications for undergraduate admission; (2) advising the Admissions Office about these strategies and practices; and (3) reporting to the Advisory Committee on the CRNS's consideration of specific race-neutral strategies approximately every two years. (Id. at 822:9-23.) The CRNS considered the effect of race-neutral practices on the University's diversity objectives and academic goals, as well as the potential administrative expense associated with implementing such practices. (Nov. 16 Trial Tr. 828:6-11.) The CRNS convened fifteen times between Spring 2016 and April 2018. (Id. at 826:25–827:1.)
In sum, the Court finds that UNC has engaged in ongoing, serious, good faith consideration of workable RNAs that would allow it to achieve its goal of diversity about as well as a race-conscious admissions process.
B. UNC Has Augmented Its Recruitment and Financial Aid Programs
1. Student Recruitment
Despite this work, Plaintiff identifies four race-neutral strategies it believes UNC must implement in order to demonstrate its good faith. Its first proposal is that the University invest in programs that would bring in "more highly qualified, socioeconomically disadvantaged minorities into its applicant pool." (ECF No. 1 ¶ 125 (Compl.).) Plaintiff alleges that UNC could accomplish this by implementing programs "designed to recruit students from underrepresented regions," (id. ¶ 127), or by "mailing a well-designed, targeted brochure" that seeks to engage students who might not otherwise consider applying to college, (id. ¶ 128).
The Court finds that UNC already engages in these types of activities well beyond the suggestions proffered by Plaintiff. In Mr. Farmer's words, the University "want[s] to reach out in every way that we know how ... [s]o we start engaging with students really early in their high school careers." (Nov. 12 Trial Tr. 574:19-24 (Farmer).) UNC invites students to campus, hosts special sessions targeted towards affinity groups, reaches out to prospective applicants by telephone, and visits every county in North Carolina each recruiting cycle, often conducting these communications in Spanish for students and their families. (Id. at 574:25–575:10.) The Admissions Office also partners with the Office of Diversity and Inclusion to operate Project Uplift, described by Mr. Farmer as "a big pipeline program that happens in the summer, and it is available to rural, low-income, underrepresented first-generation college and other students." (Id. at 575:11-15.) Additionally, the Admissions Office works closely with the American Indian Center to host "Carolina Horizons," an event that invites Native American students to campus, introduces them to the University's Native American community, and encourages college attendance. (Nov. 13 Trial Tr. 737:8-15 (Davis).) Once students are admitted, the University continues to recruit them through partnerships with student organizations such as the Black Student Movement and similar groups run by Hispanic and American Indian students. (Nov. 12 Trial Tr. 575:16-21.) Additionally, UNC also offers travel grants that allow any applicant eligible for a fee waiver or Pell Grant to take a trip to campus with one parent and experience the University firsthand. (Nov. 13 Trial Tr. 743:7-15.)
Finally, UNC participates in the Carolina College Advising Corps program ("CCAC"). The CCAC seeks to increase the number of low-income, first-generation, and underrepresented students enrolling in college by placing recent UNC graduates as college advisors in select public high schools throughout the state to assist students and their families with college admissions, scholarship applications, and the financial aid process. (Id. at 576:2-9; ECF No. 225 ¶¶ 92–93 (J. Statement).) This program serves about 20% of all Black students in public high schools in North Carolina, 50% of all American Indian students, and a large share of Latino students. (Nov. 12 Trial Tr. 577:19-23.)
2. Financial Aid
Plaintiff's expert Richard Kahlenberg nevertheless contended at trial that UNC could offer scholarships to students, and further argued that the University should use more of its three-billion-dollar endowment to make college more affordable. (Nov. 12 Trial Tr. 498:5-6, 459:8-12).
Mr. Kahlenberg is a senior fellow at the Century Foundation, where he has worked for the last twenty-two years. (Nov. 12 Trial Tr. 391:2; 393:17-18.) He graduated from Harvard College in 1985 and earned his juris doctor from Harvard Law School in 1989. (Id. at 392:3-6.) Mr. Kahlenberg has worked in several professional occupations that include being a legislative assistant, serving as a visiting associate professor of constitutional law, and working for the think tank Center for National Policy. (Id. at 392:10-11, 392:20–393:4.) Mr. Kahlenberg has also published or edited several works that explore various subjects related to socioeconomic status which include their consideration as RNAs in college admissions. (Id. at 393:9–395:5.)
There is, however, strong evidence that the University already has several policies in place to address these concerns. First, the University has a need-blind admissions process, meaning that an applicant's ability, or inability, to pay tuition is not taken into account when making an admissions decision. (Nov. 12 Trial Tr. 554:14-23.) UNC is also currently one of the only two public universities in the country that meets the "full demonstrated need" of every undergraduate student who is eligible for federal aid. (Id. at 581:24–582:4 (Farmer); ECF No. 225 ¶¶ 77–81 (J. Statement).) In total, UNC awards 93% of its available scholarship and grant funds based solely on financial need, (Nov. 12 Trial Tr. 582:17-19), and provided approximately $159 million in scholarships and grants to undergraduate students in the 2016–17 school year, (ECF No. 225 ¶ 82 (J. Statement)).
"Full demonstrated need" is defined as the difference between a university's estimated cost of attendance and the amount the FAFSA and CSS determine the applicant's family can afford to pay, according to a set formula. (ECF No. 225 ¶¶ 77–80 (J. Statement).)
A foundational part of this financial support is the Carolina Covenant program, which offers admitted students whose family income is at or below 200% of the federal poverty guidelines a financial aid package that meets 100% of their cost of attendance. (Id. ¶¶ 84–85.) Carolina Covenant Scholars account for approximately 12–14% percent of each incoming class, (Id. ¶ 86), and the University does not set a cap on the number of students who can participate, (Nov. 12 Trial Tr. 584:22-24.) Mr. Farmer testified that this program entails an outsized financial commitment from the University, and it has continued even amid "serious financial challenges." (Id. at 585:1-3 (Farmer).)
Mr. Kahlenberg's contention that UNC should use more money from its endowment likewise oversimplifies the University's position. According to the evidence provided, the University may spend only the income generated by the endowment, and 90% of those funds are subject to restrictions agreed to by their donors. (Id. at 586:10-22.) The University may access about $6 million per year in unrestricted endowment-generated funds, much of which already supports need-based student aid. (Id. at 586:23–587:2.)
Given these efforts, it is not surprising that the University has been repeatedly recognized for its commitment to low-income students. (Id. at 587:12–588:7.) The Court finds that UNC has been strongly committed to helping underrepresented students afford the costs of college. Accordingly, the Court finds that Plaintiff has failed to show how increasing funds available for need based scholarships could serve as a RNA.
C. UNC Has Increased Diversity Through Transfer Students
Plaintiff next contends that universities "have achieved student body diversity by aggressively recruiting high-achieving community college students." (ECF No. 1 ¶ 129 (Compl.).) Again, this is an area where UNC has undertaken significant efforts. In 2006, UNC established the Carolina Student Transfer Excellence Program ("C-STEP") which offers guaranteed admission to low- and moderate-income students who attend a partner community college, complete required coursework, and earn an associate degree. (Nov. 12 Trial Tr. 578:21–579:10; ECF No. 225 ¶¶ 88, 90 (J. Statement).) Students whose household incomes fall at or below 300% of federal poverty guidelines are eligible for C-STEP, and the University meets 100% of their financial need. (Nov. 12 Trial Tr. 579:12-14; ECF No. 225 ¶¶ 89, 91 (J. Statement).) This program began with three partner colleges and eight students in 2006 and has now grown to include fourteen partner colleges and 400 participating students each year. (Nov. 12 Trial Tr. 580:11-13.)
While Mr. Kahlenberg acknowledged the strength of C-STEP, he nevertheless contended that "programs like that could be increased, enhanced, extended to try to increase ... racial and economic diversity." (Id. at 454:11-22 (Kahlenberg) (observing that the University of California at Berkeley had a higher number of students who were community college transfers).) Defendants’ expert, Professor Hoxby, tested this suggestion empirically to see if an expansion of the program would likely yield such results. (See Nov. 17 Trial Tr. 1031:12–1032:1 (Hoxby).) In her analysis, she assumed that (1) the University would be able to identify all North Carolina public school students who expressed an intention to attend a community college and (2) convince those with the highest test scores to transfer to UNC, thereby ensuring the simulation would maximize academic results. (Id. at 1031:14-23.) Even with these assumptions, however, Professor Hoxby found that an expansion of the C-STEP program would yield substantially lower average test scores for UNC's admitted class. (Id. at 1031:14-1032:1).)
Nevertheless, UNC continues to expand this program every year: not only has the number of C-STEP participants consistently increased, but in the last five years community college students have gone from 28% of the incoming transfer class to 45% in the most recent cycle. (Nov. 12 Trial Tr. 581:3-11.) The Court finds that the University has in good faith adopted this strategy and continues to test it as a potential long-term RNA.
D. Eliminating the Early Action Deadline and Legacy Preferences Would Have a De Minimis Effect
Plaintiff next argues that UNC "employs admission practices and policies that make it more difficult for socioeconomically disadvantaged minorities to gain admission," namely its early admissions deadline and its policy of favoring legacy applicants. (ECF No. 1 ¶¶ 133–145 (Compl.).)
With regards to UNC's early admissions deadline, Mr. Kahlenberg argues that this practice amounts to an "unfair preference[ ] that tend[s] to benefit wealthy and white students." (Nov. 12 Trial Tr. 415:23-25, 4:16:18-21 (Kahlenberg).) Before considering this contention, it is useful to observe that UNC's "early action" program is different from a typical "early admissions" program in that students who submit their applications by UNC's early deadline are not bound to enroll at UNC and may apply to any other institution. (Id. at 501:3-8; id. at 558:22-24 (Farmer).) Nevertheless, Mr. Kahlenberg argues that this policy "still slants toward the wealthy and well-connected and white students who ... realize it's an advantage to apply early, whereas other students who might not have intensive counseling won't have that knowledge." (Id. at 501:12-17 (Kahlenberg).)
There is disagreement between the parties as to whether applying by the early action deadline confers an advantage. (Compare id. at 557:22, 558:12-21 (Farmer), with Nov. 9 Trial Tr. 177:1-3 (Arcidiacono).) The evidence that early action acts as an advantage, however, was not developed at trial, and neither of Plaintiff's experts made any contention regarding any magnitude such an effect might have. Though Mr. Kahlenberg may be correct that such a preference—if it were to exist—could work to the advantage of students with more resources, the Court finds that it has no basis to find that UNC's early action program has such an impact. Accordingly, there is no basis to find that eliminating such a program would help the University achieve the racial diversity it seeks.
The same may be said for Plaintiff's contention that the University should eliminate preferences for legacy applicants. While intuitively, a preference for the children of former students could disproportionately disadvantage URM students at a university that, until only 70 years ago, was exclusively open to white students only, Professor Arcidiacono testified that the number of minorities admitted to the University is "minimally affected" by legacy preference. (Nov. 9 Trial Tr. 317:20-24 (Arcidiacono).) Therefore, there is no basis for the Court to find that eliminating legacy preferences would serve as a viable RNA.
E. Changing the Admissions Evaluation Process
Finally, Plaintiff contends that UNC could achieve the educational benefits of diversity by altering its admissions process. First, it argues that the University could do this "by placing greater emphasis on socioeconomic factors." (ECF No. 1 ¶¶ 66–68 (Compl.) (identifying such factors as single-parent status, parents’ education level, family income, the number of dependents in the family, whether the applicant attended a rural high school, the percentage of students from the applicant's high school eligible for free or reduced-price lunch, or school-wide student-to-teacher ratio).). Second, Plaintiff contends that a "community-based preference" could accomplish the same goal. (Id. ¶ 73.) These models include "percentage plans" that guarantee admission to a certain percentage of the top graduates at each high school or within a certain zip code. (Id. ¶¶ 74–75.)
Experts for both Plaintiff and Defendants conducted analyses to test whether any of the suggested models could provide a RNA. Mr. Kahlenberg testified that he provided Professor Arcidiacono with "high-level ideas," such as changing the weights of some variables in Professor Arcidiacono's preferred model to admit a hypothetical class of students. (Nov. 9 Trial Tr. 466:20–467:2 (Kahlenberg).) Mr. Kahlenberg would then look at the results of those simulations to suggest changes to UNC's admissions practices.
On the other hand, Professor Hoxby conducted what she described as "exhaustive simulations" that attempted to "test the limit" of Plaintiff's proposals. (Nov. 17 Trial Tr. 937:19–938:4 (Hoxby).) She testified that she completed 109 simulations but "was never able to achieve UNC's actual levels of racial and ethnic [diversity] and its level of academic preparation" despite making "very generous assumptions" in her models. (Id. at 1032:25–1033:7; ECF No. 251-1 at 28 (Hoxby demonstratives).)
The Court finds that none of the models proffered by Plaintiff nor Defendant would be viable RNAs that would allow UNC to reproduce the educational benefits of diversity about as well as its current approach.
1. Professor Hoxby's Process
Because Professor Hoxby's research in this area is distinct from her econometric model, the Court first considers her process before looking at the simulations provided by the parties. In conducting her simulations, Professor Hoxby drew on data from the North Carolina Education Research Data Center ("NCERDC") which compiles information directly from the administrative records of North Carolina public schools, (Nov. 17 Trial Tr. 989:20-23 (Hoxby)), and includes information on all North Carolina public school students, (id. at 996:4-5). Using this database allowed Professor Hoxby to increase the size of UNC's potential applicant pool to students who had not applied and access data not otherwise available in the applications the University provided.
The process Professor Hoxby used to test the viability of any individual RNA consisted of several steps. The first step was simply identifying the alternative that she would be assessing. (Id. at 981:5-7.) Professor Hoxby testified that she attempted to "test any alternative that Plaintiff had considered," such as those in the Complaint or otherwise set forth by Mr. Kahlenberg, as well as others that appeared in papers, journal articles, or books that had been referenced in Plaintiff's briefing. (Id. at 981:7-25.) These plans may largely be broken down into four categories: (1) socioeconomic plans, (2) "top X percent" plans, (3) geography-based plans, and (4) additional concepts proposed by Plaintiff's experts. (ECF No. 251-1 at 28 (Hoxby demonstratives).)
The second step for Professor Hoxby in testing the viability of a RNA was deciding who would apply under the proposed alternative model. (Nov. 17 Trial Tr. at 982:3-7.) This was necessary, she testified, because a change in admissions policy would necessarily increase the chances of some students being admitted and accordingly, likely increase their chances of applying. (See id. at 982:8-15 (Hoxby) (noting that, when the University of Texas moved to an admissions plan that automatically admitted students in the top 10% of their high school class, more students who matched that description applied).)
Defendants’ expert Professor Bridget Terry Long confirmed this dynamic, observing that "when you change a policy, you may also be changing incentives and student behavior or family behavior, which is going to influence who ends up in the applicant pool." (Nov. 18 Trial Tr. 1191:22–1192:6.) While it is easy to envision how admitting every student in the top 10% of the class might alter incentives, cases—including this one—are commonly predicated on the idea that a plaintiff would change their behavior if an admissions process were changed in more nebulous ways. (See, e.g., Gratz v. Bollinger , 539 U.S. 244, 251 n.1, 123 S.Ct. 2411, 156 L.Ed.2d 257 (2003) (explaining that one plaintiff "indicated that he intended to apply to transfer if the [university's] discriminatory admissions system is eliminated" (internal citation and quotation omitted)); see also ECF No. 1 ¶ 20 (Compl.) ("Applicant is ready, able, and intends to seek to transfer to [UNC] when it stops discriminating against applicants on the basis of race and ethnicity.").)
Professor Long is the Dean of the Harvard Graduate School of Education and has earned three degrees in Economics, including a Ph.D. from Harvard University. (Nov. 18 Trial Tr. 1185:12–1186:13.) She was appointed by President Obama and confirmed by the U.S. Senate to sit on the National Board of Education Sciences and has served as its chair for two years. (Id. at 1186:15-23.) Her research focuses on college access and success, (id. at 1187:1-3), and she has testified as an expert before Congress four times, most recently on issues related to the FAFSA, (id. at 1188:21-25).
In deciding who would be in the applicant pool, Professor Hoxby testified that she made assumptions that were generous to the RNAs, in essence providing a "ceiling" for their effectiveness that assumed the "highest possible level [of applications] that is realistic." (Nov. 17 Trial Tr. 987:4-12 (Hoxby).) For instance, one assumption Professor Hoxby made is that all students in the current applicant pool would remain despite evidence that this had not happened when other schools had altered their admissions policies. (Id. at 988:1-12.) She additionally used NCERDC data to identify students who did not apply but would now be considered highly qualified under the RNA model being tested. (Id. at 989:1-19.) Professor Hoxby then added 75% of these students to the application pool. (Id. )
The way that this assumption is "generous" to a RNA is that it ensures that any simulated model would have at its disposal as many top students to choose from as possible. In such a scenario, even if only a small number of current applicants were chosen under a different admissions scheme, each would be guaranteed to be ones that maximized the qualifications the new model preferred.
Another assumption that Professor Hoxby made was that students in the NCERDC data who had taken the ACT only one time—when they were required to do so during their junior year of high school—would have taken the test a second time now that they were more likely to be admitted under a new admissions process at UNC. (Id. at 990:18–991:13.) To adjust for this issue, she added the equivalent of 40 points on the SAT to a student's lone ACT score to boost their academic performance. (Id. ) This boost, according to Dr. Hoxby, is based on "serious research conducted by [the] ACT about the effect of retaking the exam." (Id. at 991:14-16.) It likewise would tend to boost the "ceiling" of any RNA by increasing the academic credentials of students in the candidate pool who fit this description.
The final two steps in Professor Hoxby's RNA analysis were deciding who would be admitted into the class and who would enroll. (Id. at 982:16–983:4.) In deciding who would be admitted, Professor Hoxby used the same type of statistical models discussed above that she and Professor Arcidiacono employed to simulate UNC's current process. (Id. at 982:16-24.) In Professor Hoxby's words, these "are not perfect admissions models because, of course, we did not know everything about a student that an admissions officer can see," but they nevertheless operate as a reasonable proxy for the alternative in question. (Id. )
When completing the final step analyzing who would enroll at UNC, Professor Hoxby testified that "[s]tudents who are more likely to have good alternative opportunities are a little less likely to actually matriculate at UNC." (Id. at 992:14-16.) In other words, admissions data shows that students with the highest qualifications and multiple college offers are the most likely to decline admission. Professor Hoxby accounts for this through a yield model she created that assumes the "current enrollment probabilities would continue to hold." (Id. at 991:22-24.)
Professor Hoxby testified that her simulations were exhaustive in three ways. First, she considered "every race-neutral alternative plan that was proposed or suggested, even hinted at in any way by the Plaintiff in the Complaint or in any other expert report." (Id. at 992:21-25.) Second, she states that she "tried very hard under each one of those plans to allow for a wide range of possibilities about how a plan would actually be implemented." (Id. at 993:6-11.) Third, and finally, she stated that she "chose assumptions that try to get me to something that was like a ceiling for each plan" and, for some simulations, "created a way of doing the race-neutral alternative which was purely designed to maximize the power or the ability of the race-neutral alternative to attain the actual." (Id. at 993:12-18.)
2. Socioeconomic Simulations
The first group of simulations the Court considers are those based on socioeconomic status. Before considering these simulations at greater length, the Court first observes that SES and race are not interchangeable. Student-Intervenors offered testimony from their personal experiences that shed light on this issue. For instance,
• Mr. Ward, who attended UNC in the early 1980s, recounted instances of being called a racial epithet and testified that "socioeconomics had nothing to do with that," (Nov. 16 Trial Tr. 920:4–921:4 (Ward));
• Mr. Brennen testified that he does not see much of a link between SES and racial identity, and shared that "people don't see me as someone that is relatively affluent; they see me as a black man," (Nov. 18 Trial Tr. 1258:6-10 (Brennen));
• Ms. Ornelas explained that her racial identity, SES, and first-generation college status, are "important parts to getting a full picture" of who she is, and "the way they uniquely intersect ... is important," (id. at 1287:21–1288:2 (Ornelas) (emphasis added));
• Ms. Mwamba explained that while she has achieved some level of upward mobility in her SES status, her racial identity has given her unique perspective and unique life experiences that would not be the same as someone just because they shared the same SES, (Nov. 19 Trial Tr. 1362:9–1263:8 (Mwamba)).
Therefore, while the Court accepts that an increase in socioeconomic diversity may be valuable in its own right to a university seeking to attain the benefits of educational diversity, achievement of this goal does not obviate a school's interest in racial diversity as well.
When completing the socioeconomic simulations, Professor Hoxby took four specific steps. First, she constructed an SES index for each applicant using the variables available to her and assigned a value to each application. (Nov. 17 Trial Tr. 998:19-25.) Second, in order to "test the full range of every" RNA, she built several different iterations of her model with different SES "emphases" and "thresholds." (Id. at 999:1-23 (Hoxby).) The "emphasis" of a model referred to the number of seats she would reserve for low SES candidates. (Id. ) The "threshold," on the other hand, would define what SES index score would be classified as "low." (Id. ) For instance, one example might include 750 seats in the admissions process reserved for students in the bottom 20% of the SES index being tested. The 750 seats would be the "emphasis" of that particular model and the 20% would be the "threshold" at which a student would be considered "low SES." (Id. ) Professor Hoxby analyzed eighty-two different socioeconomic plans using a wide range of these factors. (Id. at 993:1-2.)
In summarizing her approach, Professor Hoxby emphasized that she made three key assumptions that were favorable to the RNAs. The first is that UNC would be able to identify all of the low SES students and that these students would then apply to the University at a high rate. (Id. at 1003:21–1004:23.) Second, Professor Hoxby assumed that UNC would, in every situation, choose the highest-scoring, low-SES student. (Id. at 1005:1-12.) This, of course, is not how UNC's current process works and would require that the University abandon its holistic approach for this group of candidates. (Id. ) Third, she assumed that the current admitted applicants would continue to enroll at the same rate. (Id. at 1005:13-22.)
In the end, Professor Hoxby testified that she "was never able to find a race-neutral alternative using this index that achieved UNC's actuals," even with the favorable assumptions that she made. (Id. at 1012:1-4.) In some cases, Professor Hoxby showed models where URM applicants were admitted with approximately the same test scores as URMs in UNC's actual admitted class, but their enrollment numbers were reduced by more than half. (ECF No. 251-1 at 34 (Hoxby demonstratives).) In other cases, Professor Hoxby's model was able to increase the size of the URM population, but at a cost of at least 200 points on average on the SAT. (Id. ) The vast majority of the simulations, she found, resulted in both fewer URMs admitted and meaningfully lower test scores among those students. (Id. ; Nov. 17 Trial Tr. 1011:12-21.) Professor Hoxby further testified that "it's simply not possible to proxy very well for race and ethnicity." (Nov. 17 Trial Tr. 1013:16-20 (Hoxby).)
Mr. Kahlenberg argued that at least one of the simulations that Professor Hoxby completed could be "[v]iable." (Nov. 12 Trial Tr. 443:21-24 (Kahlenberg).) In order to reach this conclusion, however, Mr. Kahlenberg abandons holistic admissions entirely by completing the class with "the most academically qualified students remaining using a measure of high school GPA and SAT, equally weighted, for the in-state public high school students." (Id. at 444:21–445:4.) Such a system would prevent UNC from pursuing any other types of diversity and, even with his optimistic assumptions, would also lead to a drop in URM admissions and SAT scores. (See ECF No. 247-2 at 24 (Kahlenberg demonstratives).)
Mr. Kahlenberg presented his own set of three simulations at trial that sought to test whether using SES indicators could achieve the educational benefits of racial diversity. In two of the three examples, URM representation and GPA decreased and, SAT scores dropped in all three. (See ECF No. 247-2 at 14, 16, 18 (Kahlenberg demonstratives).) Professor Hoxby testified that these models were also "terribly unrealistic." (Nov. 17 Trial Tr. 1014:1-15 (Hoxby).) For one, she argued that, while the SES boosts Mr. Kahlenberg included in his models would have a "very, very large effect" on the applicant pool, Mr. Kahlenberg did not take into account how that pool would be altered by such changes. (Id. at 1014:1-6, 1014:19–1015:3.)
Professor Hoxby next pointed out that the boosts Mr. Kahlenberg provided for students with low SES were "extremely large, unrealistically large, so large that it would essentially remove the ability of UNC to practice holistic admissions at all." (Id. at 1015:6-11.) In some cases, Professor Hoxby describes the boost to be as large as a student adding 400 or 500 points to their SAT score. (Id. at 1015:11-16 (observing that "it depends on the simulation, but [the boost] could be as high as 800 points added to the student's SAT score").) The Court finds that Professor Hoxby's exhaustive models demonstrate that an admissions process that heavily favored low-SES candidates would not achieve the educational benefits of diversity about as well as the University's race-conscious policies. Her calculations, which are based on very generous assumptions, show that adopting such policies would compromise UNC's tenuous momentum towards achieving a critical mass of underrepresented students even in the best-case scenarios. The Court further credits Professor Hoxby's uncontested conclusions that Professor Kahlenberg's simulations rely on both unrealistic assumptions and extreme changes to UNC's admissions process and would severely undermine the University's ability to pursue any other type of diversity.
In addition to the models he presented, Mr. Kahlenberg also testified that a SES-based RNA would be more likely to succeed if the University would broaden its definition of socioeconomic disadvantage to account for an applicant's "wealth" rather than income. (Nov. 12 Trial Tr. 450:12-18 (Kahlenberg).) While Mr. Kahlenberg may be correct that such data could allow the University to more easily identify students who have overcome socioeconomic obstacles, (see id. at 451:15–452:1), implementing such a policy, particularly if it meant that a university would no longer be "need-blind," would be a "fundamental change to how admissions committees do their work or how they strive to do their work to not disadvantage low-income students." (Nov. 18 Trial Tr. 1215:2-7 (Long); see also Nov. 12 Trial Tr. 555:23–556:14 (Farmer) (testifying that UNC no longer requires completion of forms that would provide information on a candidate's family wealth because "we realize the profile can be a burden and a barrier for low-income students, and we don't want to impose another barrier for students who might already be struggling to get to us").) The Court credits the testimony of those experienced in the Admissions process and finds that obtaining such information would most likely come at the cost of providing additional barriers to low-SES applicants.
Even outside of the proposed alternatives provided in this case, there is strong evidence that any models based on socioeconomic status are ill-equipped to serve as RNAs. Though using SES "was very attractive" to Professor Long and others in her field when the idea first surfaced two decades ago, "very, very quickly ... researchers from across the board, economists, sociologists[,] using lots of different data sets kept coming to the same conclusion, that you couldn't get racial and ethnic diversity from an SES-based plan." (Nov. 18 Trial Tr. 1204:13–1205:16 (Long).) This is because the majority of low-income students are white, and therefore "if you were going to have a policy that gives preferences according to SES, you're still going to be choosing more white students" from the low-SES pool than you would URMs. (Id. at 1206:10-17.) This is why "no university has actually implemented an SES-based plan that has replaced a holistic, race-conscious admissions approach." (Id. at 1206:23–1207:3.)
To the extent that Mr. Kahlenberg contends that some of his models could nevertheless function as RNAs, Professor Long found that "he overstates how effective race-neutral alternatives have been or would be because he's not paying attention to the details." (Id. at 1210:14-18.) In particular, she points out, his literature review "fails to account for the quality or the relevance of the research or the particular data used" as well as the context of the university at issue. (Id. at 1210:18-23.) As a result, Professor Kahlenberg relies heavily on "thought experiments" which "provide limited insights about what might actually be feasible for colleges and universities to implement," depend upon "data that an admissions committee wouldn't have," or make "assumptions that are not reasonable for the real world." (Id. at 1207:4-18.)
Accordingly, the Court finds that none of the socioeconomic models before it is a workable RNA. Further, the premise on which these models are based has yet to provide even a feasible prototype on which to reasonably attempt such a drastic change to its policy.
3. Top X Percent Simulations
The next type of plan that the parties considered is what may be referred to as a "top X percent" plan. The most well-known examples of this have been implemented at the University of Texas and Texas A&M where students ranked in the top 10% of their high school class are automatically admitted each year. (Nov. 17 Trial Tr. 1016:19–1017:1.) Such plans are predicated on the idea that high schools continue to be largely segregated by race. Professor Hoxby testified that "the more desegregated North Carolina's high schools become in the future, the worse a plan like this would work in terms of achieving racial and ethnic diversity." (Id. at 1019:24–1020:6 (Hoxby) ("These plans really do depend not only on having a high level of segregation currently, but also maintaining that high level of segregation into the future."); see also Nov. 18 Trial Tr. 1241:15-20 (Long) (observing that the need for segregation in order for "a percentage plan or anything place-based" to work is "something that comes up in ... the research literature again and again and again." ).)
Professor Long also pointed out that, when Texas implemented its plan, families with students just outside the top 10% of their class had an incentive to transfer high schools in order to take advantage of the new admissions preference. (Nov. 18 Trial Tr. 1201:8-16 (Long) (observing that there are "several studies that actually document this kind of behavior").) Researchers have found that these transfers ultimately "mitigated the intended effects of the percentage plan policy." (Id. at 1201:17-21.)
There are a greater number of students in North Carolina as compared to Texas relative to the slots available in UNC's freshman classes. (See Nov. 17 Trial Tr. 1017:20-24.) Therefore, a top 10% plan in the Tar Heel State would include more students than UNC admits each year. (Id. ) Accordingly, Professor Hoxby first modeled a top 7.95% plan that would fill all of the University's available slots. (Id. ) She found that there were mixed results with regards to racial diversity. Overall, URM numbers increased as compared to the status quo, including a 19% increase in African American students. (See ECF No. 251-1 at 39 (Hoxby demonstratives).) At the same time, this change would nearly erase the Native American incoming class which experienced a 73% drop from seventy students to nineteen. (Id. ) The number of admitted Hispanic students also dropped slightly. (Id. ) On average, SAT scores dropped by 77 points and, among URMs, average scores dropped by more than 100 points. (Id. ) Professor Hoxby then considered a top 7.29% plan that would include the number of students it would take to fill the actual number of in-state seats in the class. (Nov. 17 Trial Tr. 1020:11-17.) The results were largely similar. (Id. at 1020:20–1021:14.)
While these simulations filled every seat at UNC, Professor Kahlenberg's top X percent models were similar to the process implemented by the University of Texas in that he allocated a certain percentage of the seats in the incoming class to students based on class rank and then filled the remaining seats through a more holistic admissions process. (Nov. 12 Trial Tr. 439:19–440:2 (Kahlenberg).) Using the applications of actual candidates who applied to UNC, (id. at 440:8-11), Mr. Kahlenberg first modeled a 4.5% plan, (id. at 438:17-19). This simulation resulted in the same percentage of in-state URMs (16.0%), including an increase in African American students from 9% to 10% of the class. (ECF No. 247-2 at 20 (Kahlenberg demonstratives).) SAT scores in this simulation dropped for the same group from an average of 1311 to 1280, and GPA dropped marginally. (Id. )
Kahlenberg's ninth simulation was also a percentage plan that took the top 4% of students from all public high schools in North Carolina using the NCERDC data. (Nov. 12 Trial Tr. 441:12-20, 442:13-15.) This simulation was therefore different from the previous iteration in that it included students who did not apply to UNC. (Id. at 441:12-20.) Admitting the top 4% filled three-quarters of UNC's class, and so the remaining 25% of students were selected through a holistic model that did not include UNC's ratings. (Id. at 441:21–442:1.) In this model, the percentage of white students rose from 68% to 75% while the percentage of Asian American and Hispanic students dropped, and African American students remained the same. (See ECF No. 247-2 at 22 (Kahlenberg demonstratives).) Overall, URM representation fell from 16.3% to 14.6% and SAT scores dropped as well, while GPA increased. (Id. )
Professor Hoxby testified, however, that Mr. Kahlenberg's percentage plans were not, in fact, based on high school rank but rather on the admissions index calculated by Professor Arcidiacono's preferred model. (Nov. 17 Trial Tr. 1021:19–1022:1 (Hoxby).) As Professor Hoxby observed, there is an inherent practical challenge to such a plan. Whereas students in Texas, for instance, would know their class rank upon applying, North Carolina students would only know their admissions index after it had been computed by the University. (Id. at 1022:18–1023:3.) For its part, UNC would not know the rank of any of its candidates on this measure unless it had access to real-time data for all high school students—including non-applicants—so that it could set them in the proper order. (Id. ) The Court observes that even if a system could be implemented to allow high schools to enter scores that would provide this admissions index to all stakeholders, the index itself would change every time any student took a standardized test. UNC would either have to cut off the date at which students could submit their SAT or ACT scores or implement a system that dynamically responded to students rising into the top 4.5% or falling out of it on an ongoing basis throughout each selection cycle. Both solutions appear to be largely impractical—not to mention unprecedented—in higher education.
Professor Hoxby also contended that Mr. Kahlenberg again failed to take into account the change in UNC's applicant pool that would be a result of the percentage plans he proffered. (ECF No. 251-1 at 41 (Hoxby demonstratives).) He unrealistically assumed, she testified, that all of the top students in the candidate pools he uses would apply, be admitted, and enroll. (Nov. 17 Trial Tr. 1023:21–1024:18 (Hoxby).)
The Court finds that none of the top X percent simulations offered a workable RNA. Leaving aside the viability of such plans in general, every model that was presented meaningfully changed the composition of the incoming class, was based on unrealistic assumptions, presented practical challenges that would burden all stakeholders with significant complications, and/or severely undermined the University's ability to achieve diversity in non-racial ways.
4. Geography-Based Simulations and Additional Concepts
Professor Hoxby next conducted a set of geography-based simulations. To her knowledge, no other university has implemented such a plan. (Id. at 1024:21-25 (Hoxby) (calling the exercise "entirely theoretical").) She testified that these plans "try in some ways to mimic" the top X percentage plan model by taking a "small level of geography," like a census tract or a zip code, and ranking students within them. (Id. at 1025:13-22.)
In one of her models, Professor Hoxby identified the historical admissions rates for each of the census tracts in the state. (Id. at 1027:4-14.) Using a formula that takes into account only test scores and GPA, she then ranked the students within each tract and chose the top student from each starting with the area that had the lowest historical admissions rate. (Id. at 14-21.) She continued this pattern until all of the seats in UNC's class were filled. (Id. at 1027:15-23.) She testified that this strategy resulted in a "significant decrease in the racial diversity of UNC's admitted group of students and enrolled group of students." (Id. at 1028:1-3.)
She next tested two different types of alternatives proposed by Mr. Kahlenberg where UNC would partner with disadvantaged high schools or community colleges to increase the number of candidates from these areas. (ECF No. 251-1 at 46 (Hoxby demonstratives).) Professor Hoxby found that each scenario she analyzed resulted in "substantially lower average test scores" and at least one also provided "less racial diversity." (Id. )
Accordingly, the Court finds that none of these models provided a workable RNA.
F. Conclusions
When considering the testimony of the two experts, it is clear to the Court that Professor Hoxby's firsthand familiarity with the nuances of the data she is using allows her to provide a deeper level of analysis for each plan that has been presented in evidence. Indeed, the Court finds her research to be, for all practical purposes, exhaustive in considering the race-neutral admissions policies that have been proffered by either party.
When taking into account the assumptions that must be made to attain even the most optimistic outcomes, the Court finds that none of the models before it from either party would be viable in reproducing the educational benefits of diversity about as well as a race-conscious admissions policy. It is possible that such a set of policies exists, and it is incumbent upon UNC to continue to study emerging ideas. However, it would be counterproductive at this time to abandon the current admissions process in favor of untested proposals that, even in the best-case scenarios and under dubious assumptions, exact significant consequences on the University's ability to recruit and enroll an academically prepared student body that is diverse along the several dimensions it values. Such changes are more likely to have a deleterious effect on all of the metrics at issue than to improve any single one of them.
When considering all of the evidence on RNAs before it, the Court finds that UNC has conducted good faith, serious consideration of ways in which its admissions policies can become race-neutral and continues to implement and research the most promising strategies available to it in ongoing, periodic reviews. It further finds that the University has demonstrated that there are not workable or viable RNAs, singly or in conjunction, that would allow it to achieve the educational benefits of diversity about as well as its current race-conscious policies.
VIII. CONCLUSIONS OF LAW
A. Overview
As earlier stated, while SFFA brought three claims for relief in its Complaint, only two were remaining at trial. See ECF No. 210. The two remaining claims, Counts I and II, each allege that UNC has intentionally discriminated against certain of Plaintiff's members on the basis of race, color, or ethnicity in violation of the Equal Protection Clause of the Fourteenth Amendment and Title VI through its undergraduate admissions policies and program at the University. In Count I, Plaintiff alleges that the University does not merely use race as a "plus" factor in admissions decisions to achieve student body diversity. In Count II, Plaintiff alleges that the University employs racial preference in undergraduate admissions when there are available race neutral alternatives capable of achieving student body diversity.
No separate analysis is necessary in this context regarding violations under the Equal Protection Clause of the Fourteenth Amendment and Title VI. See Gratz , 539 U.S. at 276 n.23, 123 S.Ct. 2411 ("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.")
This Court, having made the foregoing Findings of Facts ("Findings") related to the University's undergraduate admissions program, reaches its ultimate conclusion that UNC has met its burden of demonstrating that the University's undergraduate admissions program withstands strict scrutiny and is therefore constitutionally permissible, which is explained in more detail below. The Court will next summarize controlling United States Supreme Court precedent relied upon to reach its ultimate conclusion.
B. Controlling Supreme Court Precedent Permits Race-Conscious Admissions
1. Regents of University of California v. Bakke, 438 U.S. 265, 98 S.Ct. 2733, 57 L.Ed.2d 750 (1978) (" Bakke ")
In 1968, the Medical School of the University of California at Davis ("UC Davis") opened with a class of fifty students. Bakke , 438 U.S. at 272, 98 S.Ct. 2733. Only three of those students in the inaugural class were non-white. Id. In response, the school created a "special admissions program," where applicants who self-identified as "economically and/or educationally disadvantaged"—or, later, as a member of a "minority group"—were evaluated and admitted in a separate process. Id. at 274, 98 S.Ct. 2733. Over the next few years, UC Davis increased its entering class size to 100 and annually reserved sixteen of its seats for students admitted under its special program. Id. at 279, 98 S.Ct. 2733. After being rejected for admission twice, a white student who had been considered under the regular admissions process brought suit against the university under Title VI of the Civil Rights Act of 1964 and the Equal Protection Clause of the Fourteenth Amendment. Id. at 277–78, 98 S.Ct. 2733.
The case eventually reached the Supreme Court where the Justices "produced six separate opinions, none of which commanded a majority." Grutter, 539 U.S. at 322, 123 S.Ct. 2325 (describing the Bakke opinions). Justice Powell, whose opinion set forth the Court's judgment on the narrowest grounds, see id. at 325, 123 S.Ct. 2325, concluded that the use of race in admissions was subject to strict scrutiny, and that UC Davis bore the burden of showing that its use of race was "precisely tailored to serve a compelling government interest," Bakke , 438 U.S. at 299, 98 S.Ct. 2733. In reviewing the interests that UC Davis proposed were substantial enough to meet that burden, Justice Powell accepted only one of the four that the university set forth. First, he rejected a quota system that "assure[d] within its student body some specified percentage of a particular group," writing that it would be "facially invalid." Bakke , 438 U.S. at 307, 98 S.Ct. 2733. Next, though acknowledging that the "State certainly has a legitimate and substantial interest in ameliorating, or eliminating where feasible, the disabling effects of identified discrimination," he nevertheless found that the medical school had failed to provide "judicial, legislative, or administrative findings of constitutional or statutory violations" that could provide a basis for such a conclusion. Id. Third, he found that there was scant evidence in the record that the medical school's interest in admitting underrepresented candidates so that they could improve "the delivery of health-care services to communities currently underserved" would produce such results. Id. at 310, 98 S.Ct. 2733. Justice Powell did find, however, that "the attainment of a diverse student body ... clearly is a constitutionally permissible goal." Id. at 311–12, 98 S.Ct. 2733. "[I]t is not too much to say," he wrote, "that the nation's future depends upon leaders trained through wide exposure to the ideas and mores of students as diverse as this Nation of many peoples." Id. at 313, 98 S.Ct. 2733 (internal quotation marks and citation omitted) (observing that it "may be argued that there is greater force to these views at the undergraduate level").
That said, Justice Powell found that UC Davis's admissions process was not narrowly tailored to such a goal. A constitutionally permissible admissions system, he wrote, would take race into account as a "plus" factor in a particular applicant's file, while not "insulat[ing] the individual from comparison with all other candidates for the available seats." Id. at 317, 98 S.Ct. 2733. According to Justice Powell, a university's policies must be "flexible enough to consider all the pertinent elements of diversity in light of the particular qualifications of each applicant." Id. A school that followed such guidelines would be presumed to be acting in good faith "in the absence of a showing to the contrary." Id. at 318–19, 98 S.Ct. 2733.
Ultimately, the Court held only that a "State has a substantial interest that legitimately may be served by a properly devised admissions program involving the competitive consideration of race and ethnic origin." Grutter , 539 U.S. at 322–23, 123 S.Ct. 2325 (quoting Bakke , 438 U.S. at 320, 98 S.Ct. 2733 ). UC Davis's two-tiered program that considered candidates separately on the basis of their race, however, did not meet this bar, and the Court affirmed the lower court's judgment striking it down. Bakke , 438 U.S. at 271, 98 S.Ct. 2733.
2. Gratz v. Bollinger, 539 U.S. 244, 123 S.Ct. 2411, 156 L.Ed.2d 257 (2003) (" Gratz "); and Grutter v. Bollinger , 539 U.S. 306, 123 S.Ct. 2325, 156 L.Ed.2d 304 (2003) (" Grutter ")
Twenty-five years later, the Supreme Court issued two concurrent opinions that further developed its jurisprudence in this area. The plaintiff in the first of these decisions, Gratz v. Bollinger , challenged the use of race in admissions at the University of Michigan's College of Literature, Science, and the Arts ("LSA"). 539 U.S. at 251–53, 123 S.Ct. 2411. In LSA's process, university officials used what they referred to as a "selection index" where they awarded up to 150 points to candidates for a number of different factors. Id. at 255, 123 S.Ct. 2411. The scores in each of these categories were then added together to create a summary score which led directly to an admissions decision based on the total points awarded. Id. Notably, URM students were automatically assigned twenty points in the "miscellaneous" category by virtue of their membership in an underrepresented racial group. Id. The Court found that this mechanical addition of points based on race, however, was "not narrowly tailored to achieve the interest in educational diversity" that LSA claimed to be seeking. Id. at 270, 123 S.Ct. 2411.
"Justice Powell's opinion in Bakke ," the Court held, "emphasized the importance of considering each particular applicant as an individual, assessing all of the qualities that individual possesses, and in turn, evaluating the individual's ability to contribute to the unique setting of higher education." Id. at 271, 123 S.Ct. 2411. The Court found that LSA's admissions policy did not provide for individual consideration of candidates and, moreover, found that the automatic awarding of twenty points made the factor of race "decisive for virtually every minimally qualified underrepresented minority candidate." Id. at 271–72, 123 S.Ct. 2411 (internal quotation marks and citation omitted).
In the second case, the Court found that Michigan Law School's admissions process passed constitutional muster in its broad, holistic review of each candidate's credentials. Grutter , 539 U.S. at 334, 123 S.Ct. 2325. To be narrowly tailored, the Court reiterated, a race-conscious admission process may not use a quota system that would "insulate each category of applicants with certain desired qualifications from competition with all other applicants"; may consider race as only a "plus" factor; and must be "flexible enough to consider all pertinent elements of diversity." Id. (quoting Bakke , 438 U.S. at 315, 317, 98 S.Ct. 2733 ). The Court further held that each applicant must be "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." Id. at 337, 123 S.Ct. 2325 ("The importance of this individualized consideration in the context of a race-conscious admissions program is paramount."). This does not mean, however, that race could not provide the tipping point for students whose applications were in the gray zone between being admitted and being denied. For instance, the Court found no fault with the University's acknowledgement that, in some cases, an applicant's race "may be a ‘determinative’ factor." Id. at 319, 123 S.Ct. 2325. It additionally reaffirmed the goal of diversity as a substantial interest, holding that the law school's "educational judgment that such diversity is essential to its educational mission is one to which we defer." Id. at 328–29, 123 S.Ct. 2325 ("[A]ttaining a diverse student body is at the heart of the Law School's proper institutional mission.").
The Court in Grutter also recognized the historic backdrop that provides the context for URM candidates—and the impossible task of attempting to separate the race of an applicant from the effect that race has had on his or her life experience. "By virtue of our Nation's struggles with racial inequality," the Court held, URM students "are both likely to have experiences of particular importance to the Law School's mission, and less likely to be admitted in meaningful numbers on criteria that ignore those experiences." Id. at 338, 123 S.Ct. 2325 ; see also id. at 333, 123 S.Ct. 2325 ("Just as growing up in a particular region or having particular professional experiences is likely to affect an individual's views, so too is one's own, unique experience of being a racial minority in a society, like our own, in which race unfortunately still matters.").
The Court also considered testimony regarding the Law School's desire to enroll a "critical mass" of underrepresented students. The Director of Admissions, for instance, testified that this phrase meant "meaningful numbers" or "meaningful representation," such that it "encourage[d] underrepresented minority students to participate in the classroom and not feel isolated." Id. at 318, 123 S.Ct. 2325. That said, the phrase "critical mass" did not imply to the Court that there was an effort on the part of the Law School to implement a quota system, nor did they represent magic words that a school must define outside of describing the substantial "educational benefits that diversity is designed to produce." Id. at 329–30, 123 S.Ct. 2325.
The plaintiff in Grutter also argued that the Law School's plan was not narrowly tailored because race-neutral alternatives existed that could achieve the benefits of diversity. Id. at 339, 123 S.Ct. 2325. The Court disagreed, however, and held that narrow tailoring "does not require exhaustion of every conceivable race-neutral alternative." Id. Nor does it force a school to "choose between maintaining a reputation for excellence or fulfilling a commitment to provide educational opportunities to members of all racial groups." Id. At the same time, narrow tailoring does require "serious, good faith consideration" of potential workable options. Id. at 339–40, 123 S.Ct. 2325 (rejecting two RNAs because they undercut the Law School's ability to make "the individualized assessments necessary to assemble a student body that is not just racially diverse, but diverse along all the qualities valued by the university"). The Court also held that the use of race in admissions "must have a logical end point" and identified two ways in which a university might accomplish this goal: by creating sunset provisions and conducting periodic reviews on the extent to which such policies remain necessary. Id. at 342, 123 S.Ct. 2325.
Though this holding could potentially be read as requiring both of these actions, the Court has approved subsequent admissions practices where a university engaged in a "periodic reassessment" of its policies without mentioning the use of sunset provisions. See Fisher II , 136 S. Ct. at 2209–10.
3. Fisher v. University of Texas , 570 U.S. 297, 133 S.Ct. 2411, 186 L.Ed.2d 474 (2013) (" Fisher I "); and Fisher v. University of Texas , 579 U.S. 365, 136 S. Ct. 2198, 195 L.Ed.2d 511 (2016) (" Fisher II ")
In 2013, the Court considered the admissions process at the University of Texas where race was one of several factors the university used in evaluating undergraduate applicants. Fisher I , 570 U.S. at 300–01, 133 S.Ct. 2411. On its first review of the case in Fisher I , the Court held that the lower court, in affirming a summary judgment for the university, had not applied strict scrutiny consistent with the burden articulated in Grutter and Bakke . Id. at 303, 133 S.Ct. 2411. Accordingly, the Court vacated and remanded the lower court's decision. Id. Subsequently, the lower court again affirmed the summary judgment, and the Supreme Court granted certiorari a second time, 136 S. Ct. at 2207, leading to Fisher II .
As background, the admissions process under consideration required that the University of Texas fill up to 75% of its freshman seats by admitting any student in the state whose GPA ranked in the top portion of their high school class. Id. at 2206. Texas admitted the remaining quarter of its class through a more holistic review where it combined the scores that its admissions officers assigned to each candidate on (1) an academic index and (2) a personal achievement index. Id. Race was considered as a factor within the latter score. Id. The district court, in reviewing the university's holistic admissions process, held that it "is contextual and does not operate as a mechanical plus factor for underrepresented minorities." Id. at 2207 (quoting Fisher v. Univ. of Tex. , 645 F. Supp. 2d 587, 606, 608 (W.D. Tex. 2009) ) (internal quotation marks omitted).
The Supreme Court, in reaching its decision in Fisher II that this admissions process was constitutionally sound, held that " Fisher I set forth three controlling features relevant to assessing the constitutionality of a public university's affirmative-action program." Id. at 2207–08. "First, because racial characteristics so seldom provide a relevant basis for disparate treatment, race may not be considered by a university unless the admissions process can withstand strict scrutiny." Id. at 2208 (internal quotations and citations omitted). To satisfy a strict scrutiny analysis, a university must "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." Id. (internal quotations omitted).
The second principle of Fisher I , expounded by the Court, was that "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." Id. (internal quotations omitted). There is no deference given to a university which might "impose a fixed quota or otherwise define diversity as some specified percentage of a particular group merely because of its race or ethnic origin." Id. (internal quotations omitted). However, when a university provides "a reasoned, principled explanation for its decision" to pursue the educational benefits of diversity, "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." Id. (internal quotations omitted).
The third and final principle of Fisher I , according to the Court, is that "no deference is owed" to a university when a court is evaluating whether the use of race is narrowly tailored to achieve its goals. Id. To that end, a university not only must create a process that considers each candidate individually but also "bears the burden of proving [that] a nonracial approach would not promote its interest in the educational benefits of diversity about as well and at tolerable administrative expense." Id. (internal quotations omitted). The Court reiterated that narrow tailoring, however, "does not require exhaustion of every conceivable race-neutral alternative" or even "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." Id. (internal quotations omitted). Instead, it imposes on the university "the ultimate burden of demonstrating that race-neutral alternatives that are both available and workable do not suffice." Id. (internal quotations omitted).
The Court next addressed several arguments that the plaintiff made in the case, two of which are particularly applicable here. First, the plaintiff contended that the university "must set forth more precisely the level of minority enrollment that would constitute a ‘critical mass.’ " Id. at 2210. "Without a clearer sense of what the University's ultimate goal is," she argued, "a reviewing court cannot assess whether the University's admissions program is narrowly tailored to that goal." Id. The Court, however, rejected this contention. Id. It held that the compelling interest of a diverse student body is not achieved by enrolling a certain percentage of students across racial groups but rather by obtaining "the educational benefits that flow" from such an outcome. Id. ("Increasing minority enrollment may be instrumental to these educational benefits, but it is not ... a goal that can or should be reduced to numbers."). Moreover, the Court held that, "since the University is prohibited from seeking a particular number or quota of minority students, it cannot be faulted for failing to specify the particular level of minority enrollment at which it believes the educational benefits of diversity will be obtained." Id.
However, simply "asserting an interest in the educational benefits of diversity writ large is insufficient." Id. at 2211. Rather, according to the Court, a university's goals "must be sufficiently measurable to permit judicial scrutiny of the policies adopted to reach them." Id. (holding that such goals may not be "elusory or amorphous"). The Court held that the university met this bar through a number of "concrete and precise goals," such as (1) the destruction of stereotypes; (2) cross-racial understanding; (3) preparation of the student body to thrive in a diverse society; (4) the cultivation of leaders; (5) the robust exchange of ideas; and (6) exposure to different cultures. Id. The Court further found that Texas had provided a reasoned and principled explanation for its decision to pursue these goals after finding through an internal study that its race-neutral policies had been unsuccessful in several of these areas. Id. The Court also recognized that these findings were further supported by the depositions and affidavits of several admissions officers. Id.
Next, the plaintiff argued that "there are numerous other available race-neutral means of achieving the University's compelling interest" in the educational benefits of diversity. Id. at 2212 (internal quotations omitted). She suggested three alternatives, none of which the Court accepted. First, she argued that the University "could intensify its outreach efforts" to underrepresented groups. Id. at 2212–13. The Court, however, observed that Texas had already created new scholarship programs, opened regional admissions centers, and increased its recruitment budget. Id. at 2213. She next suggested that the school should alter the weight it gives to academic and socioeconomic factors. Id. However, the Court found that this proposal ignored the University's ongoing efforts to use SES as part of its admissions process and "further ignores this Court's precedent that the Equal Protection Clause does not force universities to choose between a diverse student body and a reputation for academic excellence." Id. (citing Grutter , 539 U.S. at 339, 123 S.Ct. 2325 ).
Finally, petitioner suggested that Texas "uncap the Top Ten Percent Plan and admit more—if not all—the University's students through a percentage plan." Id. The Court found this proposal to be problematic in at least three ways. First, it held that percentage plans are "adopted with racially segregated neighborhoods and schools front and center stage." Id. (citing Fisher I , 570 U.S. at 335, 133 S.Ct. 2411 (Ginsburg, J., dissenting)). Therefore, it is "race consciousness, not blindness to race, that drives such plans," and such a strategy increases the university's reliance on race instead of eliminating it. Id. Second, the Court found that such an approach would "sacrifice all other aspects of diversity" in the admissions process and prioritize the single metric of class rank over any holistic review. Id. The Court held that "privileging one characteristic above all others does not lead to a diverse student body." Id. Third and finally, according to the Court, such an alteration in the process would create untenable "perverse incentives" for applicants and their families, encouraging parents to keep children in less competitive schools or discouraging students from taking challenging classes that might best prepare them for college and their lives thereafter. Id. at 2214 (quoting Gratz , 539 U.S. at 304 n.10, 123 S.Ct. 2411 (Ginsburg, J., dissenting)).
Justice Souter made a similar observation in dissent in Gratz , pointing out that "[w]hile there is nothing unconstitutional" about percentage plans, they "nonetheless suffer[ ] from a serious disadvantage. It is the disadvantage of deliberate obfuscation." 539 U.S. at 297–98, 123 S.Ct. 2411 (2003) (Souter, J., dissenting) (footnote omitted). In other words, the percentage plans "are just as race conscious" as admissions processes that automatically award points to candidates because of their race, as was the case in Gratz , "but they get their racially diverse results without saying directly what they are doings or why they are doing it." Id. at 298, 123 S.Ct. 2411. "Equal protection," Justice Souter wrote, "cannot become an exercise in which the winners are the ones who hide the ball." Id.
Applying the above-described Supreme Court precedent to the University's admissions program in this case, the Court draws the following Conclusions of Law.
C. Conclusions of Law: The University's Undergraduate Admissions Program Withstands Strict Scrutiny
1. Conclusions of Law: The University Has a Compelling Interest in Pursuing the Educational Benefits of Diversity
To satisfy strict scrutiny, the University must first demonstrate that it has a compelling interest which makes its use of a race conscious admissions policy necessary in furthering its interest. The Supreme Court has recognized that pursuing the educational benefits that flow from student body diversity is a compelling interest that may support the use of race in admissions. Fisher II , 136 S. Ct. at 2210 ; Grutter , 539 U.S. at 328, 123 S.Ct. 2325. Further, a university's decision to pursue the educational benefits of diversity is, in substantial measure, an academic judgment that must be reasoned and principled to be given judicial deference. Fisher II , 136 at 2208. Absent a showing to the contrary, a court may presume a university's good faith in choosing to pursue the educational benefits of the diversity. Grutter , 539 U.S. at 329, 123 S.Ct. 2325 ; Bakke , 438 U.S. at 318–19, 98 S.Ct. 2733.
UNC has met its burden of demonstrating that it has a compelling interest in pursuing the educational benefits of diversity, including racial diversity. (See Findings, Part III, supra. ) The University has produced substantial, credible, and largely uncontested evidence that beginning around 1998, it determined that, as a public university responsible to the people of North Carolina, it has an obligation to "create and sustain an environment of educational excellence" and "foster mutually beneficial interactions among students, faculty, staff, and administrators who possess diverse backgrounds and wide varieties of perspectives and life experiences." (Id. § III.A.) As articulated in its Mission Statement, the University exists "to serve as the center of research, scholarship, and creativity and to teach a diverse community of undergraduate, graduate, and professional students to become the next generation of leaders." Id. The University has determined that, to fulfill this mission, it must enroll and admit a diverse student body. Id. The University's interest in and commitment to its pursuit of the educational benefits of diversity are substantial, long-standing, and well documented as is reflected in, and articulated through, its core institutional documents to include its Mission Statement, Faculty Council statements, Academic Plans, and numerous other critical University documents which detail the educational benefits of diversity that the University seeks. Id.
In addition, the University has offered a principled and reasoned explanation for this decision borne out of quantitative and qualitative research. (Id. § III.B.) In 2004, the University engaged a task force comprised of faculty, students, and staff to conduct its first University-wide diversity assessment. The report of the task force was issued in 2005. As a result of this undertaking, the University determined in its academic judgment, based on evidence and experience, that it can achieve its education, research, and service mission only by creating and sustaining an environment in which students, faculty, and staff represent diversity to include age, race, gender, ethnicity, social backgrounds, economic circumstance, personal characteristics, philosophical outlooks, life experiences, and all other ways people differ. According to the University, it is through this broad diversity and the educational benefits that flow from it that students will learn how to navigate in a complex diverse and inclusive world. Further, the University determined that it would achieve and maintain diversity on the campus through the admission of students who broadly reflect the ways in which we differ, including a critical mass of underrepresented populations. (Id. § V.C.2.) The absence of critical mass, the University found, "impedes the educational process," places "undue pressure on underrepresented students," and limits the degree to which all students can experience the "educational benefits of a diverse learning environment." Id.
The University has also articulated specific educational benefits it seeks to achieve that are both concrete and sufficiently measurable to permit judicial scrutiny, to include (1) promoting the robust exchange of ideas; (2) broadening and refining understanding; (3) fostering innovation and problem-solving; (4) preparing engaged and productive citizens and leaders; (5) enhancing appreciation, respect, and empathy, cross-racial understanding, and breaking down stereotypes. (Id. § III.B.) These benefits are not "elusory or amorphous," and are "sufficiently measurable to permit judicial scrutiny." See Fisher II , 136 S. Ct. at 2211. The Supreme Court has recognized many of the same educational benefits of diversity or goals to be achieved and found such benefits to be concrete and sufficiently measurable to permit judicial scrutiny. Id.
Further, the University consistently assesses its progress towards achieving these educational benefits of diversity through both quantitative and qualitative measures. (Findings, Section III.C, supra. ) The University regularly reports much of the data it collects, collects feedback from faculty and staff, and performs regular climate studies. These studies include partnerships with third party groups that allow the University to evaluate its progress with other similarly situated institutions. In addition, the University has a standing Educational Benefits of Diversity Working Group to coordinate a litany of assessments at every level of the University. While assessment of progress in attaining the educational benefits sought by the University is rigorous and ongoing, there continues to be much more that needs to be done to achieve these benefits as reflected in the testimony of student intervenors, faculty, and administrators alike. (Id. § III.E.)
SFFA in its Complaint and during opening argument at trial challenged whether the University's pursuit of the educational benefits of diversity was genuine and authentic; however, the organization appears to have abandoned those claims in that they presented no evidence at trial to support such allegations and SFFA's counsel conceded such in his closing argument. Nevertheless, the University has presented substantial evidence demonstrating its good faith in pursuing the educational benefits that flow from diversity.
SFFA does argue that because of the University's inability to define and measure its goal of attaining a critical mass of underrepresented students, such goal is elusory and amorphous, and further that there is no definite and measurable endpoint. According to SFFA, the University does not discuss the concept of "critical mass" in its Admissions Office, has not determined if it has achieved a critical mass of underrepresented students, and has not defined the term. For this reason, SFFA contends the University should not be accorded judicial deference. (Id. § V.C.2.a.)
This argument was expressly rejected in both Grutter and Fisher II . Grutter , 539 U.S. at 329-330, 123 S.Ct. 2325 ; Fisher II , 136 S.Ct. at 2210-2211. As in this case, in Grutter , the university sought to enroll a "critical mass" of underrepresented minority students as a means of achieving its’ compelling interest in the educational benefits that flow from such diversity. Id. The university there defined "critical mass" of minority students as "meaningful numbers" or "meaningful representation" necessary to foster and promote the educational benefits it was designed to produce. Id. The Court in concluding that the university's program in Grutter bore the hallmarks of a narrowly tailored plan, explained that the university's compelling interest was not in achieving "some specified percentage of a particular group," but rather to achieve "the educational benefits that diversity is designed to produce." Id. Subsequently, in Fisher II , the petitioner argued, as SFFA argues here, that universities must "set forth more precisely the level of minority enrollment that would constitute ‘critical mass.’ " Fisher II , 136 S.Ct. at 2210. However, the Court reiterated that "the compelling interest that justifies consideration of race in college admissions is not an interest in enrolling a certain number or percentage of minority students." Id. As the Court recognized such interest would be little more than an unconstitutional quota to achieve racial balancing. Id. While increasing minority enrollment may be instrumental to furthering the university's compelling interest in the educational benefits that flow from such diversity, according to the Court, "it is not, as petitioners seem to suggest a goal that should be reduced to pure numbers." Id. Rather the concept of critical mass should be defined by reference to the educational benefits that diversity is designed to produce. Id. The Court in Fisher II also noted that "[s]ince the University is prohibited from seeking a particular number or quota of [underrepresented minority] students, it cannot be faulted for failing to specify the particular level of minority enrollment at which it believes the educational benefits will be obtained." Id.
While the University in this case readily acknowledges that it does not regularly use the phrase critical mass, nor point to a specific number of underrepresented minority students it seeks to enroll, it has produced significant evidence that it defines, discusses, and measures critical mass by reference to the educational benefits that it seeks to achieve and that such diversity is designed to produce. (Id. § V.C.2.a-b.) Thus, contrary to SFFA's contention and consistent with Grutter and Fisher II , the Court concludes that the University has defined critical mass, not by numbers or unconstitutional quotas, but rather by reference to the educational benefits of diversity this concept is designed to produce. As earlier concluded by this Court, the benefits of diversity sought by the University here are neither elusory nor amorphous.
In sum, the Court concludes that UNC has met its burden in demonstrating that it has a genuine and compelling interest in achieving the educational benefits of diversity; that the University has offered a reasoned explanation for its decision to pursue these benefits; that the educational benefits sought by the University are concrete and measurable and are not elusory and amorphous; that these benefits are being regularly assessed; and further that the University's decision to pursue such benefits by, among other things, enrolling students that are both academically gifted and broadly diverse is entitled to "judicial deference." Fisher II , 136 S. Ct. at 2208.
Lastly, in terms of whether the University has defined an end point for its use of race in the admissions, the Supreme Court has approved admissions programs, such as the one here, that have engaged in periodic reassessment of its policies without any articulation of a sunset provision. Id. at 2209–10. Here, the University has an ongoing process through which it has and will continue to make such assessments. (Findings, Section V.C.2.b, supra. ) Further, the University has determined that, though it is rigorously pursuing the educational benefits described and there is some realization of progress, there is more work to be done. (Id. ) Thus, the Court also concludes that the durational requirement has been satisfied here.
2. Conclusions of Law (Count 1): UNC Has Demonstrated that It's Use of Race in Admissions Is Narrowly Tailored
Count I of SFFA's Complaint alleges that UNC has "intentionally discriminated against certain of Plaintiff's members on the basis of their race, color, or ethnicity ... by employing an undergraduate admissions policy that does not merely use race as a ‘plus’ factor in admissions decisions in order to achieve student body diversity." (ECF No. 1 ¶ 198.) The Court concludes that the evidence produced at trial does not support SFFA's claim. (See Findings, Parts IV, V, VI, supra. )
The Supreme Court has found that admissions programs which consider race are narrowly tailored when they flexibly are a part of an individualized holistic review of each applicant which gives serious consideration to all the ways an applicant contributes to a diverse education. Grutter, 539 U.S. at 337, 123 S.Ct. 2325. There must be a close "fit" between the university's consideration of race and its "compelling goal" of diversity, so that there is "little or no possibility ... [for] illegitimate racial prejudice or stereotype." Id. at 333, 123 S.Ct. 2325. Thus, to be constitutionally permissible, such a program may not use a quota system that would "insulate each category of applicants with certain desired qualifications from competition with all other applicants"; may consider race as only a "plus" factor; and must be "flexible enough to consider all pertinent elements of diversity." Id. at 334, 123 S.Ct. 2325 (quoting Bakke , 438 U.S. at 315, 317, 98 S.Ct. 2733 ). The Court has further held that each applicant must be "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." Id. at 337, 123 S.Ct. 2325 ("The importance of this individualized consideration in the context of a race-conscious admissions program is paramount."). This does not mean, however, that race could not provide the tipping point for students whose applications were in the gray zone between being admitted and being denied. The Court found no fault with a university's acknowledgement that, in some cases, an applicant's race "may be a ‘determinative’ factor." Id. at 319, 123 S.Ct. 2325.
Consistent with Supreme Court precedent, the University's admissions program here engages in a highly individualized, holistic review of each applicant's file, giving serious consideration to all of the ways an applicant may contribute to a diverse educational environment. (See Findings, Part IV, supra. ) Further, the University's use of race in the admissions process is narrowly tailored in that the University considers race flexibly as a plus factor as one among many factors in its individualized consideration of each and every applicant. (See Parts V, VI, supra. ).
Admissions officers testified credibly, consistently, and without contradiction that they are guided by the University's policies and procedures which mandate that all parts of a candidate's application, rewards different kinds of diversity, and evaluates a candidate within the context of their lived experience. (See generally Part IV, supra. ) All students applying to the University must complete a common application, submit an essay responding to one of seven common prompts, two short answers to prompt provided by UNC, their standardized test scores and a letter of recommendation from at least one of their teachers. (Id. § IV.B) After an application is submitted, it is reviewed by one of approximately forty admissions readers, many of whom have been reading the University's applications for more than a decade. (Id. § IV.C.) The parties stipulated, and the evidence shows, that readers evaluate applicants by taking into consideration dozens of criteria across broad categories. (See id. § IV.D.2.) The admissions office trains readers to consider each applicant as an individual based on all relevant factors revealed in his or her application in order to understand the applicant holistically and comprehensively. The readers also go through extensive training and are specifically trained on how to consider race and ethnicity in the evaluation process. (Id. § IV.C.) Race and ethnicity must be considered "as one factor among many in a holistic review of all circumstances relevant to an individual applicant". Id. Further readers are instructed that there are no quotas, fixed points, or separate admissions process based on a particular candidate's race or ethnicity. Id.
It is further uncontested that applicants are not required to disclose their race or ethnicity and are not penalized if they elect not to do so. (Id. § IV.B.) If race is disclosed voluntarily, it is only one of forty factors across a broad array of categories considered. (Id. ) It is also uncontested that the University's admission process considers all candidates as part of a single evaluation process that does not differentiate students on the basis of race, nor insulate them from comparison with all other applicants. (Id. ) There is no evidence or claim that the University uses a quota system to racially balance its incoming class, nor is there any evidence that students are awarded points automatically or mechanically due to their race or ethnicity. It is likewise uncontested by the experts of all parties that the University's admissions process is a holistic system that takes into account a large number of variables. (Id. Parts V, VI.) According to UNC's expert, Dr. Hoxby, the University's "admission decisions appear to be fully consistent with the holistic admissions process." (Id. § VI.D.5.) Likewise, SFFA's expert, Professor Arcidiacono, concedes that the University's admissions process is individualized and holistic, though he contends that the admissions process is formulaic. (Id. § VI.D.2.c.)
To support its claim that the University's use of race in its’ admissions process is improper, SFFA offers what it characterizes as non-statistical and statistical evidence of discrimination.
a. SFFA's Non-Statistical Evidence Does Not Support a Conclusion that UNC Has Engaged in Discrimination
According to SFFA, its non-statistical evidence demonstrates: (1) that the University conceals its improper use of race behind opaque procedures; (2) that the consideration of race so permeates the University's admissions process that it is a predominant factor in admissions decisions; and (3) that the University has failed to adequately reflect on the role that race plays in its admissions decisions. (Findings, Part V, supra. ) The Court concludes based on its Findings that none of these arguments are supported by credible evidence. (See id. ) The Court will likewise discuss these Findings briefly here.
With respect to their first contention, SFFA focuses on two parts of the admissions process. They first argue that the readers in the provisional decision phase "frequently highlight the applicant's race" as a key feature in the application. (Id. § V.A.1.a.) To support this contention, SFFA offers eight partial email statements out of the hundreds of thousands of application files provided by the University during discovery in which there is some reference to color or race by a reader. (Id. ) None of these emails demonstrate that race was considered outside of the holistic review given to each applicant or that it was a defining feature of the application.
Secondly, they take issue with the length of time the readers spend on each application. (Id. ) Like their first piece of evidence, there is no context given for this assertion, i.e., no comparison data related to other holistic review processes, no discussion of the experience of the reader reviewing the application, nor any evidence that some other review occurred other than the holistic review mandated by the University. The University receives approximately 43,000 applications each year to fill 4,200 available spaces. In addition, not only is there a second reader for many applications, but there is testimony by a number of the administrators that they also personally review thousands of application decisions. (See id. § V.A.2.)
SFFA next takes issue with the SGR process, which is a final step in the admissions process in which a group of veteran readers look at the provisional admissions decisions from a single high school. (Id. § V.A.1.b.) Specifically, they allege that the SGR process is used to fine tune the racial composition of UNC's admitted classes in favor of URMs. (Id. ) Plaintiff concedes that this argument was not developed at trial. (Id. (citing Nov. 19 Trial Tr. 1378:18–1379:2 (Strawbridge)).)
Despite this concession, the University provided uncontradicted testimony demonstrating that the SGR process changes the racial composition very little and, to the extent that it does, has only reduced the number of admitted underrepresented minority students. (§ V.A.1.b.) Defendants’ expert witness tracked the number of students by race who entered the SGR process as provisionally accepted from 2013-14 to 2015-16, as well as those who were ultimately accepted after the process was complete. He found that representation of Hispanic and African American applicants either remained the same or was reduced by 0.1%–0.3% by the SGR process. Over the course of three years, only one time did any single underrepresented race increase its representation in the class following SGR, and this was in 2015-16 when Pacific Islander students went from 0.0% of the admitted class prior to SGR to 0.1% afterwards. At the same time—in direct contrast to the trend of the underrepresented minority students—there were more white and Asian American students admitted as part of UNC's classes following the completion of the SGR every year. (Id. )
b. SFFA's Statistical Evidence Does Not Support a Conclusion that UNC Has Engaged in Discrimination
In addition, the statistical evidence presented in this case demonstrates that race is not a predominant factor in UNC's admission process. (See Findings, Part VI, supra. ) Both parties hired highly qualified experts to address whether race was used as more than a mere "plus" factor in the University's admissions process. (Id. § VI.A.) SFFA's expert, Professor Arcidiacono, testified that he sought to answer (1) how formulaic UNC's decisions are, and (2) the role race plays in those admissions decisions. (Id. § VI.B.1.) UNC's expert, Professor Hoxby, testified that she sought to answer whether Plaintiff's allegation that race and ethnicity were dominant factors in UNC's admissions process were accurate. (Id. § VI.B.2.) Both experts constructed numerous models as detailed in this Court's Findings to answer their identified questions. (Id. § VI.B.1, 2.) To begin, the Court concludes that Professor Hoxby's model was more probative on the issue of whether race is a dominant factor in the University's admissions process primarily because the data used to build her model more accurately reflects the University's actual process and were free of a number of assumptions that undermined Professor Arcidiacono's models. Moreover, Professor Hoxby analyzed the role of race and ethnicity in admissions decisions across the University's entire applicant pool as Supreme Court precedent requires. See Grutter , 539 U.S. at 334, 123 S.Ct. 2325 (requiring admissions programs to be "flexible enough to consider all pertinent elements of diversity in light of the particular qualifications of each applicant ," and consider race flexibly as only a "plus factor" in "individualized consideration of each and every applicant").
As detailed in this Court's Findings, the Court notes a significant number of non-standard choices in the models offered by Professor Arcidiacono that made it difficult to credit his conclusions. (Findings, Section VI.D.3, supra. ) In sum, in designing his statistical models, Professor Arcidiacono made choices that did not in many cases correspond to the University's admissions process, to include: (1) his decision to calculate and apply disparate converted SAT and ACT scores for applicants of different racial and gender groups rather than using the actual data provided by the University, which was voluminous; (2) his use of an academic index created using only test scores and grades, though the University does not use such an index; and (3) his mathematical transformation models that sought to hypothetically transform an applicant from one race to another merely by changing their race and considering no other context. (See id. § VI.D.1, 3, 4.) These anomalies made his testimony far less credible than that of Professor Hoxby. Moreover, as argued by UNC Defendants, Professor Arcidiacono appears to base his opinion that race is a dominant factor on his conclusion that race and ethnicity are determinative—at most—in a subset of applicants and not across the University's admissions process. (See id. § VI.D.4.d.) He concedes that race is not a dominant factor in the University's program as a whole.
Despite the relative weights given to each of the experts’ testimony, neither expert's models demonstrated that race and ethnicity were dominant factors in the University's admission's process as a whole. (Id. § VI.D.6.) According to Professor Hoxby, race plays a role in a very small percentage of decisions: 1.2 percent for in-state students. (Id. § VI.D.5.b.) Even assuming some of Professor Arcidiacono's imputation accurately reflected the University's process, his analysis showed fewer than 5 percent of admission decisions would be affected by the use of race over the course of the six-year period the experts analyzed. (Id. § VI.D.3.a.) In addition, race appears to be less than or equally as important as several other datapoints considered with a holistic process. (Id. § VI.D.5.)
Thus, there is no evidence that race was used by the University as a predominant factor in evaluating a candidate's admission, nor that it was a defining feature of any individual application. While there may be some evidence that race may tip the scale for some small percentage of applicant decisions in a given year, such conclusion does not transform UNC's admittedly holistic process into a constitutionally impermissible one. See Fisher II , 136 S. Ct. at 2212, ("The fact that race consciousness played a role in only a small portion of admissions decisions should be a hallmark of narrow tailoring, not evidence of unconstitutionality."). In addition, even though Professor Arcidiacono did demonstrate that he was able to predict the University's prior admissions decisions with a high degree of accuracy, there was simply no evidence that the University uses a formula, explicit or otherwise. (Findings, Section VI.D.2.c, supra. ) Even assuming there may be some patterns in the University's admissions decisions as articulated by Professor Arcidiacono, as he put it, the readers are simply "paying attention to the components" of each individual application, such as GPA, extracurricular activities, and first-generation status, rather than using a formula to make decisions. (Id. (quoting Nov. 10 Trial Tr. 318:24-25 (Arcidiacono)).) This is not surprising given that the University's readers have significant training in what the University values and how they should approach their evaluation of the tens of thousands of students who apply annually. Professor Hoxby concluded that "UNC's admissions decisions appear to be fully consistent with the holistic admissions process" and that "UNC's process could not be explained by a formula based on verifiable variables." (Id. § VI.D.5 (quoting Nov. 17 Trial Tr. 937:3-7 (Hoxby)).) "In other words, the University's admission's process does not appear to be formulaic." (Id. (quoting Nov. 17 Trial Tr. 937:11-12 (Hoxby)).) This Court concludes that there is no evidence that the University makes its admissions decisions based on a formula, nor does the University mechanically add points to applications of underrepresented minorities nor consistent weights on the basis of a candidate's race.
Based on the voluminous evidence presented and the Court Findings related to the same, this Court concludes that the University's admissions program bears all the hallmarks of a narrowly tailored race conscious admissions program: it engages a highly individualized, holistic review of each applicant's file, which considers race flexibly as a "plus factor" as one among many factors in its individualized consideration of each and every applicant, without insulating the individual from competition with all other candidates for available seats. Nor does the fact that an applicant's race may tip the scale in some small number of admissions alter this Court's conclusion. Thus, race is not a predominant factor in the University's admissions program through policy or practice, nor is it a defining feature of any individual application. Nor does the University use a quota system, formula, or add mechanical points for race in its admissions decisions. Finally, this Court concludes that the consideration of race in the University's admission's program is both necessary to achieve its mission and compelling interest, and a close fit as a means of achieving the benefits of diversity that the University seeks.
The Court will therefore enter judgment in favor of Defendants on Count 1 of Plaintiff's Complaint.
3. Conclusions of Law (Count II): UNC Has Demonstrated that a Nonracial Approach Would Not Promote the Interests of Diversity About as Well as Its Race-Conscious Process
Count II of Plaintiff's Complaint alleges that UNC has intentionally discriminated against Plaintiff's members "by employing racial preferences in undergraduate admissions when there are available race-neutral alternatives capable of achieving student body diversity." (ECF No. 1 ¶ 205.) The Court concludes that the evidence produced at trial does not support SFFA's claim. Rather, this Court concludes that UNC has engaged in ongoing, serious, and good faith consideration of workable race neutral alternatives over a several year period, and further, that UNC Defendants have carried their burden of showing that there are no adequate, workable, or sufficient race neutral alternatives available. (Findings, Part VII, supra. )
Even in a case such as the one now before the Court where the University has demonstrated that its admissions program engages an individualized, holistic review of each applicant's file, and which uses race as only a constitutionally permissible "plus" factor among many factors, the University still "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." Fisher II , 136 S. Ct. at 2208 (internal quotation omitted). That is, to satisfy narrow tailoring, a university must have considered race neutral alternatives that are workable and available and conclude that they do not promote the institution's goals as well as race conscious strategies and at tolerable expense. Id. However, it is not required that every conceivable race-neutral alternative has been tried. Id.
The University has long considered race-neutral alternatives and engages in ongoing periodic reviews to assess the validity of emerging strategies and the extent to which UNC's efforts are achieving the educational benefits of diversity that it seeks. (Findings, Section VII.A, supra. ) In 2004, Mr. Farmer began working with the College Board's Access & Diversity Collaborative, a group of policymakers, legal experts, and college administrators, to learn what other schools are doing and how well their strategies are working related to RNAs, as well as other issues. The University remains involved with this collaborative. In 2007, the admissions office completed a study that evaluated whether indicators of socioeconomic disadvantage could be used in lieu of race in the admissions process to yield a class with academic credentials and racial diversity similar to those of the admitted class. However, the study demonstrated a substantial decline in the population of underrepresented students. In 2009 Ms. Kretchmar conducted a literature review on how banning racial considerations in the admissions’ process affected other institutions. The review strongly demonstrated that schools like UNC had not found race neutral alternatives that work well. In 2012, Mr. Farmer and Ms. Kretchmar evaluated what impact a race-neutral admissions strategy similar to the one employed at the University of Texas would have at UNC. While it demonstrated that there would be an increase in URM of approximately 1%, every other academic indicator, other than share of class ranking in the 10%, would have declined. Also in 2012, the U.S. Department of Education's Office on Civil Rights reviewed the University's admissions’ process and concluded that the University had given serious, good faith consideration to race-neutral alternatives. In 2013, Mr. Farmer convened a Race-Neutral Alternative Working Group, comprised of members of faculty and staff across the University, which discussed and explored options, reviewed literature, and modeled five RNAs, each resulting in a decline in racial diversity, academic quality, or both. In February 2016, UNC's Advisory Committee on Race Neutral Strategies ("CRNS") was first convened and continues to operate today, which is comprised of faculty and administrators who have expertise in fields relevant to exploring RNAs, including diversity and inclusion, data integration, modeling, student affairs, and undergraduate admission. The CRNS was charged with: (1) considering whether there are workable race-neutral strategies and practices that the Admissions Office could employ in evaluating applications for undergraduate admission; (2) advising the Admissions Office about these strategies and practices; and (3) reporting to the Advisory Committee on the CRNS's consideration of specific race-neutral strategies approximately every two years. (Id. ) The CRNS considered the effect of race-neutral practices on the University's diversity objectives and academic goals, as well as the potential administrative expense associated with implementing such practices. (Id. (citing Nov. 16 Trial Tr. 828:6-11 (Panter)).)
This Court concludes that UNC Defendants have demonstrated that the University has engaged in ongoing, serious, good faith considerations of workable race neutral alternatives in an effort to find options to its race conscious process in admissions.
Plaintiff nevertheless contends that there are race neutral alternatives available to UNC that include (1) investing in programs that would bring in "more highly qualified, socioeconomically disadvantaged minorities into its applicant pool" (2) increasing financial aid, scholarship and recruitment efforts, (3) "aggressively recruiting high-achieving community college students," (4) elimination of "admissions policies and practices that make it more difficult for socioeconomically disadvantaged minorities to gain admission," namely its early admissions deadline and its policy of favoring legacy applicants, and (5) changing UNC's admissions process altogether. (Findings, Section VII.B–E (quoting ECF No. 1 ¶¶ 125, 129, 133 (Compl.)).)
Here again, both parties used experts to support their positions. Plaintiff's expert, Mr. Kahlenberg, offered the above strategies. To test his suggested alternatives, according to Mr. Kahlenberg he provided Professor Arcidiacono with what he called "high-level ideas," such as changing the weight of some variables in Professor Arcidiacono's preferred model to admit a hypothetical class of students (Id. § VII.E (quoting Nov. 9 Trial Tr. 466:20–467:2 (Kahlenberg).) He then would look at the results of those simulations to suggest changes to UNC's admissions practices. In addition to this unusual manner of offering expert testimony, Mr. Kahlenberg's testimony demonstrated that he lacked an intimate knowledge of the simulations prepared by Professor Arcidiacono from which he was testifying. UNC Defendants seriously questioned Mr. Kahlenberg's independence. On the other hand, the University provided credible expert evidence that supports and strengthens its assessment that no available race-neutral alternative would allow the University to achieve its compelling interest nearly as well as race conscious strategies at tolerable expense. Professor Hoxby ran more than 100 race-neutral simulations, including socioeconomic status plans, high school top X percent plans, geography-based simulations, and the other suggestions posited by Mr. Kahlenberg. Professor Hoxby testified that her simulations were exhaustive in three ways. First, she considered "every race-neutral alternative plan that was proposed or suggested, even hinted at in any way by the Plaintiff in the Complaint or in any other expert report." (Id. § VII.E.1 (quoting Nov. 17 Trial Tr. 992:21-25 (Hoxby).) Second, she states that she "tried very hard under each one of those plans to allow for a wide range of possibilities about how a plan would actually be implemented." (Id. (quoting Nov. 17 Trial Tr. 993:6-11 (Hoxby)).) Third, and finally, she stated that she "chose assumptions that try to get me to something that was like a ceiling for each plan" and, for some simulations, "created a way of doing the race-neutral alternative which was purely designed to maximize the power or the ability of the race-neutral alternative to attain the actual." (Id. (quoting Nov. 17 Trial Tr. 993:12-18 (Hoxby).) Thus, when considering the testimony of the two experts, it is clear to the Court that Professor Hoxby's first-hand familiarity with the nuances of the data she used allowed her to provide a deeper level of analysis for each plan that was presented in evidence, even those offered by Mr. Kahlenberg.
With respect to Plaintiff's suggestions that UNC increase its financial aid, scholarship, and recruitment efforts, the Court concludes, that the University already engages in these activities well beyond the suggestions offered by Plaintiff. (§ VII.B.1-2.) There was uncontested evidence presented in this case that the University has already taken these steps, including providing exceptional levels of financial aid that cover the full cost of tuition for qualifying students, employing need-blind admissions, and increased and targeted recruiting. (Id. ) The University has also partnered with an organization that increases awareness and access to college through an advising corps designed to connect with students in high schools with fewer resources to guide potential applicants through the complex world of college admissions. (Id. ) Additionally, the University has created partnerships with community colleges across North Carolina to target students who might not otherwise be considering a four-year degree. (Id. § VII.C.)
Further, the Court concludes that the elimination of the admissions policies that Plaintiff complains of—the early action deadline and a preference for out-of-state legacy students—have a de minimis effect on the University's admission process based on expert testimony provided by both sides. (Id. § VII.D.) There is no evidence that incorporating either or both in conjunction with even the most promising race-neutral strategies would lead to another outcome.
Moreover, while UNC's expert Professor Hoxby conducted over 100 simulations to gauge the impact of a broad range of potential changes to the University's admissions process, none of these simulations, even when using very generous assumptions that strongly favored the Plaintiff's proposed plans, achieved diversity about as well as UNC's race-conscious admissions policies. (Id. § VII.D.1-4.) The models proffered by SFFA, or its expert would often force the University to choose between maintaining a reputation for excellence and providing educational opportunities to all racial groups. Supreme Court precedent is clear that it does not require a University to make such choices. See Grutter, 539 U.S. at 339, 123 S.Ct. 2325. Others would dramatically undercut the University's efforts to achieve additional types of diversity that it values. Professor Hoxby's efforts to simulate these race neutral alternatives were, if not exhaustive, as close to exhaustive as would be practicable, and testimony from Professor Long confirmed that there were additionally no actual examples of race neutral alternatives in the real world which the University could follow.
When taking into account the assumptions that must be made to attain even the most optimistic outcomes, the Court finds that none of the models before it from either party would be viable in reproducing the educational benefits of diversity about as well as a race-conscious admissions policy. It is possible that such a set of policies exists, and it is incumbent upon UNC to continue to study emerging ideas. However, it would be counterproductive at this time to abandon the current admissions process in favor of untested proposals that, even in the best-case scenarios and under dubious assumptions, exact significant consequences on the University's ability to recruit and enroll an academically prepared student body that is diverse along the several dimensions it values. Such changes are more likely to have a deleterious effect on all of the metrics at issue than improve any single one of them.
When considering all of the evidence on race neutral alternatives before it and this Court's Findings regarding the same, the Court concludes that the University has conducted good faith, serious consideration of ways in which its admissions policies can become race-neutral and continues to implement and research the most promising strategies available to it in ongoing, periodic reviews. The Court further concludes that the University has met its burden in demonstrating that there are not workable or viable race neutral alternatives, singularly or in conjunction, that would allow it to achieve the educational benefits of diversity about as well as its current race-conscious policies and practices.
Accordingly, the Court will likewise enter judgment on Count II of Plaintiff's Complaint in favor of Defendants.
CONCLUSION
UNC has met its burden of demonstrating with clarity that its undergraduate admissions program withstands strict scrutiny and is therefore constitutionally permissible. In sum, the University has demonstrated that: (1) it has a compelling and substantial interest in pursuing and attaining the educational benefits of diversity and has offered a reasoned decision for doing so, entitling its decision to judicial deference; (2) to accomplish its interest it is necessary that the University admit and enroll a diverse student body to include racial diversity; (3) the University engages in a highly individualized, holistic admissions program that is narrowly tailored in that it considers race flexibly as only a "plus factor" among many factors for each and every applicant and race is not a defining feature in any of its admissions decisions; and (4) the University has conducted and continues to conduct good faith serious consideration of race neutral strategies and has not found any alternatives that would promote its compelling interest about as well and at tolerable administrative expense as its current admissions program.
However, while the University's recognition and pursuit of student body diversity and the educational goals that flow from it are not only constitutionally permissible, but welcomed, the University is far from creating the diverse environment described in its Mission Statement and other foundational documents submitted into evidence in this case. Admittedly, the efforts that the University has undertaken in recent years related to creating a diverse student body demonstrate a marked contrast to the discriminatory and obstructionist policies that defined the University's approach to race for the vast majority of its existence. (See Id. § III.B n.4.) Nevertheless, nearly seventy years after the first black students were admitted to UNC, the minority students at the University still report being confronted with racial epithets, as well as feeling isolated, ostracized, stereotyped and viewed as tokens in a number of University spaces. In addition, the evidence shows that, as a whole, underrepresented minorities are admitted at lower rates than their white and Asian American counterparts, and those with the highest grades and SAT scores are denied twice as often as their white and Asian American peers.
As legal scholar Donna Nixon writes in her recent essay outlining UNC's path to integration, UNC was founded in 1789 but only welcomed its first Black students in 1951 when Harvey E. Beech, James L. Lassiter, J. Kenneth Lee, Floyd B. McKissick, and James R. Walker enrolled in the University's Law School. Donna L. Nixon, The Integration of UNC—Law School First , 97 N.C. L. Rev. 1741, 1742, 1764 n.188 (2019). Even upon their admission, "the men could not fully enjoy the privileges and benefits of enrollment at the state's flagship university," and the students faced "ostracism, threats, hostility, and further discrimination." Id. at 1743.
Ensuring that our public institutions of higher learning are open and available to all segments of our citizenry is not a gift to be sparingly given to only select populations, but rather is an institutional obligation to be broadly and equitably administered. While no student can or should be admitted to this University, or any other, based solely on race, because race is so interwoven in every aspect of the lived experience of minority students, to ignore it, reduce its importance and measure it only by statistical models as SFFA has done, misses important context to include obscuring racial barriers and obstacles that have been faced, overcome and are yet to be overcome. As the Court in Grutter explained, by virtue of our Nation's struggle with racial inequality, such minority students, as in this case, are both likely to have experiences of particular importance to an institution's mission and less likely to be admitted in meaningful numbers on criteria which ignore those experiences. As articulated by one of UNC's experts, "the work associated with diversity and inclusion is complicated and challenging and is an ongoing iterative process." (DX108A at 8 (Expert Report of Mitchell J, Chang).) While the University's current admissions program has captured the context described in Grutter , UNC continues to have much work to do.
Based on the foregoing Findings of Fact and Conclusions of Law, the Court enters the following:
ORDER
IT IS THEREFORE ORDERED AND ADJUDGED that Judgment shall be entered on Counts I and II of Plaintiff's Complaint, (ECF No. 1), in favor of Defendants and against Plaintiff SFFA.
IT IS FURTHER ORDERED AND ADJUDGED that attorneys fees and costs will be addressed under separate order.