After Harvard: The Fight Against Race-Based Admissions at the US Naval Academy

An in-depth investigation into race-based admissions at America’s elite military academies—and the landmark case that challenged them (and lost)

Table of Contents

• Introduction and Background

  • About This Report

  • Outline and Summary Overview of Report

• 1. Understanding USNA Admissions

  • 1.1 Initial Application and Eligibility Requirements

  • 1.2 The Whole Person Multiple (WPM) and Candidate Evaluations

  • 1.3 The Nomination Process

    • 1.3.1 Congressional Nominations

    • 1.3.2 Service-Connected Nominations

  • 1.4 Selection and Capacity Constraints

    • 1.4.1 Class Size and Competitive Pool

    • 1.4.2 Special Admission Categories (Athletes and Prep Applicants)

    • 1.4.3 The STEM Degree Mandate and Academic Priorities

  • 1.5 Admission Pathways for Nominated Candidates

    • 1.5.1 Pathways for Congressional Nominees

    • 1.5.2 Pathways for Service-Connected Nominees

  • 1.6 USNA’s Stated Use of Race in Admissions

  • 1.7 Flexibility in the Admissions Process

• 2. Review of Findings from SFFA’s Analysis of USNA Data

  • 2.1 Overall Admission Rates

  • 2.2 Admission Rates to USNA

  • 2.3 Modeling Racial Preferences in USNA Admissions

  • 2.4 Quantifying Racial Preferences in Practical Terms

  • 2.5 Racial Preferences Across Admissions Channels

  • 2.6 Admissions to NAPS

  • 2.7 The Combined Effect of Racial Preferences

  • 2.8 Why Racial Preferences May Be Even Larger Than Estimated

    • 2.8.1 Disparities in USNA Performance Outcomes

    • 2.8.2 Additional Sources of Underestimation

• 3. USNA’s Response

  • 3.1 USNA’s Rebuttal of Arcidiacono’s Report

    • 3.1.1 Omitted Variable Bias

    • 3.1.2 Coding of Race/Ethnicity

    • 3.1.3 The Inappropriateness of Logit Models in the Context of USNA Admissions

  • 3.2 The Diversity Imperative

    • 3.2.1 Claim #1: Racial Diversity Enhances Military Effectiveness

    • 3.2.2 Claim #2: Racial Inequality Presents Risks to Cohesion and Battlefield Performance

    • 3.2.3 Claim #3: Racial Diversity is Critical for Recruitment, Retention, and Maintaining the Military’s Domestic and International Legitimacy

  • 3.3 USNA’s Diversity Dilemma

    • 3.3.1 Performance Disparities at USNA

    • 3.3.2 Impact of Performance on Entry into Elite Warfare Communities

    • 3.3.3 The Leadership Pipeline and Barriers to Advancement

    • 3.3.4 The Role of Promotion and Attrition in Minority Underrepresentation

    • 3.3.5 The Limits of Race-Conscious Admissions for Demographic Engineering

• 4. Judge Bennett’s Ruling

  • 4.1 What Strict Scrutiny Requires

  • 4.2 Is a Racially Diverse Officer Corps a Compelling National Security Interest?

    • 4.2.1 The Supposed ‘Diversity Consensus’

    • 4.2.2 The “Diversity Bonus”

    • 4.2.3 Conclussion of Law: Do Racial Preferences Advance a Compelling National Security Interest?

  • 4.3 Is USNA’s Consideration of Race Narrowly Tailored?

    • 4.3.1 Does USNA Use Race as a Negative?

    • 4.3.2 Has USNA Exhausted Race-Neutral Alternatives?

    • 4.3.3 Is USNA’s Use of Race Time-Limited?

• 5. Policy Options

  • 5.1 Judicial Strategy

  • 5.2 Legislative Action

  • 5.3 Executive Action

    • 5.3.1 Enforcing and Monitoring the Ban on Racial Preferences

    • 5.3.2 Establishing a Commission to Reassess the Diversity Rationale

    • 5.3.3 Strategic Implementation

• Conclusion: The Stakes of SFFA vs. USNA

Introduction and Background

The views expressed in this report are my own and do not reflect the views of any of the parties involved in the case or the institutions referenced. Where this report references materials from Students for Fair Admissions (SFFA) v. U.S. Naval Academy (USNA), it does so solely on the basis of publicly available records from the case—most of which were downloaded via PACER. All cited documents can be accessed and downloaded here.

Race-conscious admissions are illegal in American universities—and, as of February 2025, are now banned at the institutions that train the nation’s military leaders. This dramatic shift followed President Donald Trump’s January 27, 2025, executive order prohibiting race- or sex-based preferences across the U.S. Armed Forces, and a subsequent directive by Defense Secretary Pete Hegseth enforcing those principles throughout the Department of Defense. In response, the Superintendent of the U.S. Naval Academy (USNA), Vice Admiral Yvette Davids, formally revised the Academy’s admissions policy on February 14, 2025. The new guidance prohibits consideration of race, ethnicity, or sex at any stage of the admissions process—a change confirmed in Senate testimony and a Department of Justice court filing.¹

Although USNA has insisted that its use of race was “limited” and “non-determinative,” it also admitted to never having studied the impact of race on the composition of its admitted classes. Moreover, statistical analysis of admissions data tells a different story: a White applicant with a 5% chance of admission would have a 50% chance if evaluated as Black, and more than 70% of Black admits would not have been admitted under a race-neutral system. These findings, among similarly damning others examined in this report, directly contradict USNA’s characterization of its policies.

Nevertheless, Judge Richard Bennett ultimately ruled in favor of USNA on December 6, 2024, allowing race-based admissions to continue. His ruling effectively treated footnote 4 of the Supreme Court’s SFFA v. Harvard/UNC decision—which states that the ruling does not “address the propriety of race-based admissions [in the military academies]”—as a carve-out for service academies, shielding them from the Court’s holding and exempting them from the strict scrutiny standard applied to civilian universities. Accordingly, rather than applying the exacting strict scrutiny standard affirmed in Harvard, Bennett deferred entirely to the government’s assertions, accepting race-conscious admissions as essential to national security without demanding serious empirical proof.

Now, however, the policy defended in that decision has been rescinded. In light of the Academy’s revised guidance, the Department of Justice has moved to suspend appellate briefing while the parties consider whether the case has been rendered moot—and whether the district court’s ruling should be vacated. As of this writing, the case remains in legal limbo.

Yet even if the case is declared moot, its significance is far from extinguished. First, the risk of policy reversal under a future administration is real—if not inevitable. Without stronger institutional safeguards, nothing prevents a future Secretary of Defense—or Academy Superintendent—from reinstating race-based preferences. Second, if allowed to stand, the district court ruling sets a dangerous precedent: that other government agencies in a future administration may invoke vague claims of “national security” to justify racial classifications. The result would be a profound shift in equal protection jurisprudence, granting the executive branch a new and expansive justification for racial classifications in areas far beyond military personnel policy. Finally, the legal, empirical, and policy arguments advanced in SFFA v. USNA offer a critical lens for evaluating the proper role of merit, fairness, and equality in military officer selection.

About This Report

While the Supreme Court's 2023 decision in SFFA v. Harvard/UNC drew national attention to affirmative action in higher education, the application of race-conscious admissions policies at military academies has remained largely opaque—until now.

This report provides the first in-depth analysis of how racial preferences operate at a U.S. service academy. Focusing on SFFA v. USNA, it systematically examines the role of racial preferences in USNA’s admissions process and assesses their impact on the qualifications, performance, and eventual service assignments of admitted students. It also scrutinizes the legal and empirical justifications advanced by the government, including the claim that racial diversity is essential to military effectiveness.

In addition, the report critiques U.S. District Judge Richard Bennett’s December 2024 ruling, highlighting its deference to government assertions and its failure to apply meaningful constitutional scrutiny. Finally, it offers a series of policy recommendations—judicial, legislative, and executive—to ensure that the ban on racial preferences at service academies endures beyond the current Trump administration.

All told, this report not only brings long-overdue transparency to a system that has operated with minimal public scrutiny—it also lays a crucial foundation for preventing the quiet return of race-based preferences under future administrations. It is thus essential reading for policymakers, military leaders, legal scholars, and citizens concerned by the spread of racial equity-focused policies and norms in military institutions, as well as anyone interested in understanding how such policies have operated in practice, how they have been defended and justified, and how similar measures could resurface in the future.

Outline and Summary Overview of Report

This report is divided into five sections, each building upon the previous one to examine the role of racial preferences in USNA admissions, Judge Bennett’s ruling, and the broader legal and policy implications.

Section 1 provides an in-depth overview of USNA’s admissions process, which differs significantly from civilian universities due to its legally mandated nomination system and strict eligibility requirements. Unlike traditional colleges, USNA applicants must first secure a nomination—typically from a member of Congress—before they can be considered for admission. While USNA publicly maintains that race is not considered in most admissions decisions, the academy explicitly acknowledges its “limited” use in a variety of discretionary decisions, including the selection of Congressional slate winners, the awarding of Additional Appointee (AA) admissions, and the granting of Letters of Assurance (LOAs). By outlining the mechanics of the nomination and admissions process in detail, this section establishes the foundation for understanding how USNA exercises discretion in its admissions decisions and how that discretion creates opportunities for racial preferences.

Section 2 empirically assesses the extent to which race influences USNA’s admissions decisions. Using statistical analysis conducted by Duke University economist and SFFA expert witness Peter Arcidiacono, this section demonstrates that racial preferences play a decisive role in admission outcomes, directly contradicting USNA’s claims that racial considerations are limited and non-determinative. Admissions data reveal stark racial disparities, with non-White applicants—particularly Black applicants—admitted at significantly higher rates than White applicants with comparable WPM scores, especially in the middle qualification deciles. Regression models controlling for a broad range of factors confirm that these disparities persist even after accounting for potential omitted variables. Arcidiacono’s analysis also reveals how racial preferences extend beyond direct admissions into indirect pathways, particularly the Naval Academy Preparatory School (NAPS), where Black applicants are admitted at disproportionately high rates and whose matriculants enjoy near-automatic acceptance into USNA the following year. Additional discretionary mechanisms—including LOAs, medical waivers, and adjustments through the Recommendations of the Admissions Board (RAB)—further amplify racial preferences.

Beyond admissions, statistical analysis of matriculant performance data reveals that racial preferences at USNA have far-reaching consequences for academic performance, discipline, and graduation rates. Black midshipmen consistently earn lower grades, are more likely to be placed in remedial courses, and are significantly underrepresented on the Commandant’s List—a merit-based designation awarded to midshipmen with strong academic and leadership evaluations. Even after controlling for observable qualifications, racial disparities in performance persist, suggesting that Black applicants underperform relative to their admissions credentials and that the magnitude of racial preferences may be even larger than initially estimated. Further, internal USNA data reveal that Black midshipmen are overrepresented in conduct and honor offenses, underrepresented among the top Order of Merit (OOM) rankings, and disproportionately concentrated in the bottom OOM deciles. Minority midshipmen—except for Asians—also experience higher attrition and lower graduation rates than their White peers, with Black midshipmen typically graduating at rates 10 to 15 percentage points lower than Whites. These disparities persist despite USNA’s institutional shift toward a “developmental model” intended to boost overall graduation rates, raising serious concerns about whether racial preferences are setting some students up for failure rather than success. Taken together, these findings further undermine USNA’s claims that race is merely a minor, non-determinative factor in its admissions process.

Section 3 evaluates USNA’s defense of its race-conscious admissions practices, analyzing both its statistical rebuttal to Arcidiacono’s findings and its broader justification for racial preferences. The first part critiques USNA’s counteranalysis, which relied on the testimony of Dr. Stuart Gurrea, a legal economist with no peer-reviewed academic publications who was paid more than $500,000 in taxpayer money to challenge Arcidiacono’s findings. Rather than conducting an independent statistical analysis, Gurrea merely attempted to cast doubt on Arcidiacono’s methodology, arguing that the results were distorted by omitted variable bias, that Arcidiacono’s racial classification methods were flawed, and that logistic regression was an inappropriate tool for studying USNA admissions. However, these critiques fail under scrutiny. Not only do Gurrea’s arguments lack empirical support, but his suggested methodological changes, if taken seriously, would render USNA’s admissions process unmeasurable and immune to external scrutiny. That USNA relied on such a weak counteranalysis—rather than producing its own rigorous statistical study—suggests that it either could not refute Arcidiacono’s findings or feared what a full empirical investigation would reveal.

The second part of Section 3 shifts to USNA’s core legal defense, which hinges on the claim that racial diversity is essential for military effectiveness. Judge Bennett’s ruling largely accepts this assertion at face value, yet a closer examination—and even the government’s own testimony—reveals that the military has never conducted a systematic study proving that racial diversity improves battlefield performance, unit cohesion, or morale. A review of military diversity reports and testimony from military officials confirms that these justifications rest entirely on speculation rather than empirical evidence. Likewise, claims that increasing racial diversity in officer ranks is necessary to improve minority recruitment and retention lack empirical support. Historically, Black enlistment has remained strong regardless of the racial composition of the officer corps, and retention rates show no clear relationship to racial preferences. By critically assessing USNA’s diversity rationale, this section argues that it fails to meet the ‘compelling interest’ standard and is based more on untested assumptions than hard data.

Section 4 examines Judge Bennett’s ruling, arguing that it is not a neutral application of strict scrutiny but rather a legal justification designed to protect USNA’s race-conscious admissions policies. Unlike previous affirmative action rulings, which at least engaged with competing evidence, Bennett systematically defers to USNA, shifting the burden of proof onto SFFA and failing to demand clear empirical support for USNA’s claims. The first part critiques Bennett’s handling of the “compelling interest” standard, showing how he accepts USNA’s assertions about military diversity’s benefits without requiring concrete, measurable proof. Instead of engaging with the absence of empirical support, he relies on anecdotal testimony, politically motivated reports, and historical misrepresentations to justify his conclusions.

The second part of Section 4 examines how Bennett misapplies the standard of narrow tailoring. Instead of requiring USNA to exhaustively explore race-neutral alternatives, he dismisses them outright, applying an unconstitutional standard that requires such alternatives to precisely replicate current racial outcomes. His rejection of socioeconomic-based alternatives ignores clear evidence that such policies could maintain substantial racial diversity while expanding opportunities for disadvantaged applicants. Moreover, his justification for racial preferences as “time-limited” is illusory. By tying their continuation to an undefined future demographic balance, he effectively allows them to persist indefinitely, amounting to unconstitutional racial balancing. Beyond its immediate implications, Bennett’s ruling sets a dangerous precedent. If upheld, it would exempt military institutions from constitutional constraints on racial classifications, allowing them to justify racial preferences indefinitely under the guise of national security.

Section 5 outlines a comprehensive, three-pronged strategy—judicial, legislative, and executive—for ensuring that race-based admissions at USNA and other service academies are permanently eliminated and not quietly reinstated under future administrations. While the USNA case is now in legal limbo following the Academy’s policy reversal, judicial options remain crucial: parallel litigation at West Point and the Air Force Academy still offers a path to definitive constitutional resolution. However, SFFA must resist efforts to moot any of these cases absent structural safeguards. If mootness is ultimately declared, it must seek vacatur of the Bennett’s district court ruling to ensure it carries no precedential weight.

On the legislative front, embedding a statutory ban within the National Defense Authorization Act (NDAA) would offer more durable protections than executive orders alone, especially if pursued with active support from the Trump administration. Meanwhile, executive action provides immediate tools for enforcement, transparency, and deeper structural reform. These include auditing and data disclosure requirements to ensure compliance with the new ban, and the creation of an independent commission to reexamine the military’s longstanding but empirically untested “diversity rationale.” Finally, the section recommends strategically linking the admissions ban to the commission’s work by framing the policy as a temporary suspension pending evidence review. This approach would reframe the debate in empirical terms, shift the burden of proof onto proponents of racial preferences, and build institutional and legal barriers to their return. Taken together, these recommendations form a robust blueprint for securing the permanent end of race-conscious admissions at the service academies.

1. Understanding USNA Admissions

For clarity, references in this report to the 2023–2027 admissions cycles correspond to students who matriculated into the USNA Classes of 2023 through 2027. Since candidates apply nearly a year before matriculating, these admission cycles roughly align with applications submitted between 2018 and 2022. For example, an applicant who applied in 2018 and was accepted would have matriculated into the Class of 2023, while an applicant who applied in 2022 would belong to the Class of 2027. This distinction is important when interpreting data throughout the report.

1.1 Initial Application and Eligibility Requirements

Applying to the U.S. Naval Academy (USNA) differs significantly from applying to traditional universities. As early as their junior year of high school, prospective candidates must complete a preliminary application, which serves as an initial screening tool to ensure they meet USNA’s minimum eligibility requirements. These include age (17–22 years), U.S. citizenship (with a valid Social Security number), marital status (unmarried), dependency status (no dependents), and pregnancy status (not pregnant).²

Candidates who pass this initial stage proceed to the formal application process, which includes written essays (e.g., personal statements), letters of recommendation, and evaluations from high school guidance counselors. These evaluations provide information on a candidate’s GPA, class rank, and whether they are classified as “disadvantaged” or of a “minority” background.³

Before the COVID-19 pandemic, USNA required applicants to submit standardized test scores (SAT or ACT). However, for the Classes of 2025–2027, it adopted a “test-flexible” policy, which it has since discontinued.⁴

Beyond academic qualifications, applicants must also pass a medical examination, complete the Candidate Fitness Assessment (CFA), and participate in an interview with a Blue and Gold Officer (BGO)—all of which must be completed by January 31 of their senior year.

1.2 The Whole Person Multiple (WPM) and Candidate Evaluations

Once an applicant’s file is submitted, it is evaluated using the Whole Person Multiple (WPM) score, a composite metric calculated through an algorithm based on USNA’s weighting formula. Before the class of 2025 admissions cycle, the WPM score was determined by an applicant’s highest SAT math and verbal scores, high school class rank, rankings of athletic and non-athletic extracurricular activities, and character-related appraisals from math and English teachers (e.g., “Makes friends easily”). However, after adopting its “test-flexible” policy, USNA removed standardized test scores from the WPM calculation, redistributing their weight to class rank and extracurricular activities.⁵

A more subjective element of the WPM is the Recommendations of the Admissions Board (RAB), which allows the admissions board to adjust an applicant’s raw WPM score by up to ±9,000 points, with any adjustment beyond that requiring approval from the Dean of Admissions.⁶ Factors influencing RAB adjustments include unusual life experiences, hardship, the quality of the Blue and Gold Officer (BGO) interview, participation in Advanced Placement (AP) courses, personal statements, extracurricular involvement, and character assessments. These adjustments are voted on by the admissions board and take effect with majority approval. For the admissions cycles spanning 2023–2027, the average applicant received a net adjustment of +2,450 points (standard deviation = 1,870).⁷

WPM scores generally range from 40,000 to 80,000 points.⁸ Applicants for the class years 2023–2027 averaged raw WPM scores of 66,580 (standard deviation = 6,370).⁹ Scores of 70,000 or higher are considered “highly qualified” and eligible for early appointment offers. While a score of 58,000 is the general minimum for qualification, exceptions may be made for candidates with exceptional qualifications, graduates of schools with high college admission rates, or individuals with unique experiences or accomplishments.¹⁰

The WPM is designed “to assist in predicting first-year success at the Naval Academy as well as comparing candidates when being considered for offers of appointment from official nomination sources.”¹¹ Notably, USNA maintains that an applicant’s racial or ethnic background is neither considered in the calculation of WPM scores nor in RAB adjustments.¹²

1.3 The Nomination Process

1.3.1 Congressional Nominations

One of the most distinctive aspects of the USNA application process is the requirement to secure a nomination from an official source. The most common category is congressional nominations, which account for more than 80% of USNA’s admitted class.¹³ These nominations are granted by the Vice President, Members of Congress, Delegates to Congress representing Washington, D.C., and U.S. territories, as well as the Governor and Resident Commissioner of Puerto Rico.

Each member of Congress is allocated five nomination “charges”—or admission slots—at USNA at any given time.¹⁴ When a nominee is admitted, they are “charged” to that member. Members with fewer than five charges at the end of an academic year are considered to have a “vacancy,” allowing them to nominate new candidates. Typically, members have one vacancy per year and may nominate up to 10 candidates for each vacancy, all of whom must reside in the member’s district or state.

Members of Congress may submit their nominations using one of three methods.¹⁵ The most common approach, used by approximately 65% of members, is the competitive method, in which a member submits a slate of up to 10 nominees without ranking them in order of preference. USNA then evaluates and ranks these nominees based on their Whole Person Multiple (WPM) scores, with the highest-scoring qualified candidate expected to receive the appointment. However, USNA retains some flexibility in this process. If the second-highest candidate is within 4,000 WPM points of the top candidate, the Slate Review Committee (SRC) may choose the lower-scoring nominee instead.¹⁶ Importantly, USNA acknowledges that in such cases, race or ethnicity “may be one of numerous factors” in determining the slate winner.¹⁷

A second method, known as the principal numbered-alternate method, allows members of Congress to designate a principal nominee—effectively their first-choice candidate—while ranking alternates in order of preference. If the principal nominee is fully qualified, USNA is required to admit them regardless of their WPM score. However, if the principal declines the offer or fails to meet qualifications, the appointment moves down the list to the next highest-ranked alternate.

Finally, the principal competitive-alternate method is similar to the previous approach but does not require members to rank alternates. If the principal nominee is ineligible or declines the appointment, USNA evaluates the remaining candidates based on their WPM scores and selects the highest-scoring qualified individual.

1.3.2 Service-Connected Nominations

A second category of nominations is service-connected nominations, which are reserved for the children of servicemembers, active-duty Navy or Marine Corps personnel, and members of the Reserve Officers’ Training Corps (ROTC). In addition to these groups, the USNA Superintendent is authorized to nominate up to 50 candidates annually.¹⁸ While the Superintendent has broad discretion in selecting these nominees and has been described as able to “bring in whoever [he] want[s],” these nominations are most commonly awarded to recruited athletes.¹⁹ Even so, they may also be extended to other qualified candidates who have not yet received an admissions offer. In such cases, USNA acknowledges that a candidate’s race may be considered as a “nondeterminative factor” in the decision-making process.²⁰ At the same time, it asserts that race has not been a factor in these decisions since 2009.²¹ However, that this claim rests on the anecdotal recollections of the Dean of Admissions rather than a documented policy change does not instill much confidence.

1.4 Selection and Capacity Constraints

1.4.1 Class Size and Competitive Pool

On average, 5,088 domestic applicants submitted complete applications annually during the 2023–2027 admission cycles, out of an average total of 14,102 applications submitted each year.²² After excluding foreign applicants, those deemed medically or physically unqualified, and those without a nomination, the effective applicant pool averages 3,098 per year.²³ By law, the total USNA student body (the “Brigade of Midshipmen”) is capped at 4,400 midshipmen across all four years, with each incoming class typically comprising just under 1,200 individuals.²⁴ Thus, in each of the five admission cycles, an average of 3,098 applicants competed to fill approximately 1,200 available admission slots.

1.4.2 Special Admission Categories (Athletes and Prep Applicants)

In practice, however, the “true” pool of competition is even smaller, as certain groups enjoy near-automatic admission. Most notably, Blue Chip Athletes (BCAs) face almost no real competition for admission. Of the 1,268 BCA applicants during the 2023–2027 admission cycles, all but two were admitted—yielding an acceptance rate of 99.8%.²⁵ BCAs typically account for about one-fifth of each incoming class.²⁶

A similar dynamic exists for candidates admitted through the Naval Academy Preparatory School (NAPS) and other USNA-affiliated private prep programs—an admissions pathway explored in more detail later. Nearly all NAPS candidates who meet minimum academic and physical qualifications ultimately receive offers to attend USNA, with admission rates of 94.2% for NAPS graduates and 97% for private prep program graduates during the same period.²⁷ Excluding those who are also BCAs, prep school applicants constituted approximately 13% of all USNA admits for the Classes of 2023–2027.²⁸

1.4.3 The STEM Degree Mandate and Academic Priorities

Beyond selection and capacity constraints, USNA is also required to ensure that at least 65% of each graduating class earns a technical degree in fields such as engineering, mathematics, or the physical sciences.²⁹ While this mandate applies to graduation rather than admissions, it directly influences the selection process by incentivizing the admission of candidates with strong academic performance in STEM disciplines. Applicants with high standardized test scores in math and science, advanced coursework in these subjects, or demonstrated aptitude in technical fields are often prioritized to help ensure the academy meets its STEM quota. As a result, this requirement further narrows the competitive applicant pool by effectively raising the academic bar for non-STEM candidates.

1.5 Admission Pathways for Nominated Candidates

1.5.1 Pathways for Congressional Nominees

While more than 80% of those admitted to USNA during the 2023–2027 class cycles were Congressional nominees, fewer than half (49.4%) secured their place by winning a Congressional slate.³⁰ For nominees who do not win their slate, there are two primary pathways to admission. The first is through the Qualified Alternates (QA) system, which allows up to 150 (recently extended to 200) Congressional nominees per admissions cycle to receive appointments.³¹ By statute, these QAs are selected based on their Whole Person Multiple (WPM) scores, with appointments ostensibly granted to the top 150 non-slate-winning Congressional nominees. USNA claims that these selections are made solely on the basis of WPM scores, leaving no room for discretion or the consideration of race.³² However, as later sections will demonstrate, USNA appears to have greater latitude in these decisions than it publicly reveals.

If not selected as a QA, and provided that “vacancies in the incoming class of midshipmen remain following all the appointments authorized by law,” a Congressional nominee may still gain admission as an Additional Appointee (AA).³³ During the 2023–2027 admissions cycles, an average of 266 candidates per year were admitted as AAs.³⁴ Unlike QAs, AA appointments are not determined strictly by WPM scores. Instead, USNA describes AA selection as being based on “a holistic assessment of candidates who are expected to make valuable contributions to the brigade environment.”³⁵ While broadly defined, this assessment may include factors such as leadership potential, extracurricular involvement, and other subjective qualities. USNA further acknowledges that race or ethnicity may be one of many nondeterminative factors considered in AA appointments.³⁶

If neither admitted as a QA nor AA, a final, albeit indirect, path to USNA is through an offer of admission to a Naval Academy preparatory program, the largest of which is the Naval Academy Preparatory School (NAPS). NAPS accounted for 80% of all prep admits during the 2023–2027 admission cycles.³⁷ Generally, prep school offers are extended to applicants who, by USNA’s assessment, demonstrate leadership potential and physical fitness but require additional academic preparation before gaining admission.³⁸ Although WPM scores are not formally used in selecting prep school candidates, academic deficiencies often factor into the decision to offer a prep pathway.³⁹

Unlike direct USNA admissions, securing a nomination is not a prerequisite for prep school admission. This gives USNA greater discretion in determining who receives prep school offers, as nominations do not restrict or dictate selections. However, according to USNA, priority is typically given to enlisted sailors and Marines, as well as recruited athletes.⁴⁰ Conditional on meeting minimum academic and fitness performance requirements—including maintaining a cumulative GPA of 2.0 (a threshold lowered from 2.2 prior to the 2019–2020 academic year⁴¹)—just over 94% of those admitted to NAPS are subsequently admitted to USNA in the following admissions cycle, regardless of their initial WPM score.⁴²

1.5.2 Pathways for Service-Connected Nominees

Service-connected nominees accounted for nearly 15% of all USNA admits during the 2023–2027 admission cycles.⁴³ However, over 60% of these candidates did not secure a service-connected slate and were instead admitted through a USNA prep program. Within this subgroup, approximately 30% initially failed to secure any type of nomination when first applying to USNA. The remainder consisted of candidates who held service-connected nominations but were deemed academically unprepared for direct admission.⁴⁴

Importantly, nominations from one USNA admission cycle do not carry over to the next. As a result, prep admits must reapply for a nomination and are encouraged to seek endorsements from all available sources when applying to USNA.⁴⁵ Consequently, a candidate who was admitted to a prep program as a Congressional nominee may ultimately matriculate into USNA as a service-connected nominee in the following cycle.

Overall, approximately 43% of those who entered USNA from a prep program during the 2023–2027 class cycles were admitted as service-connected nominees.⁴⁶ The remainder were admitted through Congressional channels, including a non-trivial subset who secured Congressional slate wins. This further underscores the flexibility USNA exercises in the assignment of admission charges—a topic I turn to shortly.

1.6 USNA’s Stated Use of Race in Admissions

USNA openly acknowledges considering applicants’ race or ethnicity in certain admissions decisions. As noted in previous sections, it explicitly concedes that race may be a factor in (a) deciding Congressional slate winners through the Slate Review Committee (SRC), particularly when candidates’ WPM scores are within a certain threshold and the lower-scoring candidate is selected; (b) awarding Additional Appointee (AA) appointments; and (c) issuing Superintendent nominations—though, according to the Dean of Admissions, race has not been a factor in these since 2009.

In addition, USNA acknowledges considering race when conferring Letters of Assurance (LOAs)—early but conditional offers of admission that are contingent on completing the application process and meeting medical and physical qualifications.⁴⁷ While LOAs are typically extended to candidates with WPM scores exceeding 70,000, they may also be granted to lower-scoring applicants who demonstrate qualities that USNA deems valuable. Such qualities include, but are not limited to, “unusual life experience” and “overcoming significant hardship or adversity.”⁴⁸ USNA explicitly states that race or ethnic background may also be considered, though purportedly as only one factor among others.⁴⁹ Because 84% of admitted candidates during the 2023–2027 admission cycles were recommended for LOAs by the admissions board, these recommendations play a decisive role in shaping the composition of each incoming class.⁵⁰

Trial proceedings revealed an additional avenue through which race may influence admissions decisions: waitlist selections. Although USNA did not publicly disclose this practice before litigation, it admitted in court that race can, at least occasionally, be a factor in deciding whether to take a candidate off the waitlist and extend an offer of admission.⁵¹

While acknowledging that race factors into the foregoing areas of the admissions process, USNA repeatedly insists that this consideration is “limited,” “non-determinative,” and always part of a “holistic assessment that takes into account all aspects of an applicant’s file”.⁵² It maintains that no candidate is admitted “solely because of their race” and contends that, due to the congressional nomination process and WPM scoring, “race is not considered at all for many appointments to the Academy.”⁵³

Yet despite these assertions, USNA simultaneously claims to have no insight into what the racial composition of its classes would look like absent the consideration of race. By its own admission, it has never conducted modeling to explore this question and offers no clear explanation for why such an analysis has not been undertaken.⁵⁴ At the same time, USNA argues that eliminating racial considerations would cause minority enrollment to “drop dramatically,” a claim that stands in tension with its characterization of race as a minor and non-determinative factor in admissions.⁵⁵

This raises an important question: If USNA has never assessed how race affects admissions outcomes, how can it confidently claim that its role is ‘limited’ and ‘non-determinative’? And if race is truly marginal, why does USNA anticipate a drastic demographic shift without it?

To further complicate matters, testimony from USNA’s key trial witnesses regarding the role of race in the admissions process was at times unclear, if not outright contradictory. When questioned by USNA’s defense lawyers, the academy’s director of nominations and appointments stated that the admissions board does not consider race when determining whether candidates are “qualified” or “not qualified.”⁵⁶ Yet under cross-examination by an SFFA attorney, she acknowledged that “ethnic heritage, racial or ethnic diversity” could be among the factors considered in qualification determinations.⁵⁷ This was later corroborated by a member of USNA’s admissions board, who, when asked whether “race can be used to determine whether a candidate is qualified,” answered in the affirmative.⁵⁸

1.7 Flexibility in the Admissions Process

USNA’s admissions process is far more flexible—and discretionary—than the academy publicly acknowledges. One major source of this flexibility is the substantial number of applicants who secure nominations from multiple sources, which allows USNA to shift candidates between different admission channels. In addition, USNA enjoys considerable latitude in selecting replacements for slate winners who decline offers or are ultimately deemed unqualified.

For example, consider a candidate who is the third alternate on a district’s Congressional slate but also the principal nominee on a Senatorial slate. If the top two alternates on the Congressional slate decline admission offers, USNA can declare this candidate the district slate winner. This, in turn, leaves the Senate vacancy open, allowing USNA to fill it with a lower-scoring candidate who would not have qualified for a QA slot.

A similar dynamic occurs with “competitive” slates or when principals on “principal competitive-alternate” slates decline offers or fail to meet qualifications. In theory, a qualified candidate with a WPM score exceeding the second-highest scorer by more than 4,000 points should win these slates.⁵⁹ However, if this top candidate also qualifies for QA admission, and the second-highest scorer does not meet the QA threshold, USNA may assign the top candidate to a QA slot and award the Congressional slate to the lower-scoring nominee.

More broadly, discovery evidence suggest that when a slate winner declines an offer, USNA is not bound by WPM rankings in selecting a replacement. This means a lower-scoring candidate may be chosen over more qualified applicants. Evidence introduced at trial further indicates that race sometimes influences these replacement decisions. For instance, during the discovery process, SFFA obtained emails listing potential replacements for slate winners who had declined admission, where race or ethnicity was prominently featured as a primary identifying characteristic.⁶⁰

There are also numerous cases in which NAPS applicants (i.e., those applying to USNA from NAPS) win Congressional slates over higher-scoring competitors.⁶¹ This is surprising not only given the generally low WPM scores of NAPS applicants but also because it occurs even in admission cycles where WPM scores for NAPS candidates are not reported in the data. This further calls into question the transparency of USNA’s admissions process, particularly the extent to which WPM rankings genuinely drive slate outcomes.

Taken together, these factors cast serious doubt on USNA’s claim that race plays only a minor role in admissions. The academy’s own admissions process—steeped in discretion—creates ample opportunities for racial considerations to influence outcomes at multiple stages. And we don’t have to take USNA at its word. Thanks to trial disclosures and the analysis conducted by SFFA’s expert witness, Peter Arcidiacono, the data tell the real story.

2. Review of Findings from SFFA’s Analysis of USNA Data

As part of the discovery process in this case, USNA was ordered to provide SFFA with applicant files and admissions data for all individuals who applied during the admissions cycles for the Classes of 2023–2027. These data were analyzed by Duke University economist Peter Arcidiacono, who was retained by SFFA and has conducted similar statistical analyses in its (ultimately successful) challenges to affirmative action policies at Harvard and the University of North Carolina.

Arcidiacono’s primary task was to answer the question that USNA was unwilling to quantify: To what extent does race/ethnicity influence admissions decisions at USNA and NAPS? While USNA acknowledges using race in a “limited” and “non-determinative” fashion in certain areas of the admissions process, the actual scope and impact of these considerations remained unclear. Arcidiacono’s analysis systematically evaluates these claims and quantifies the extent to which race affects admissions outcomes.

The following section reviews the key findings from this analysis and their implications for USNA’s admissions policies and practices.

2.1 Overall Admission Rates

Applicants from NAPS and other USNA-affiliated prep programs are virtually guaranteed admission to the next incoming USNA class, provided they meet minimal performance requirements. Thus, with few exceptions, admission to a prep program effectively ensures admission to USNA. Accordingly, Arcidiacono’s analysis begins with descriptive statistics outlining the rates at which applicants are offered admission to either USNA or a prep program, broken down by race and ethnicity. These data are visualized in Figure 2.1A below.

Figure 2.1A. Admission Rates to USNA and USNA-Affiliated Prep Programs by Race/Ethnicity, Regardless of Nomination Status.

Note. Bars represent the percentage of applicants who received an offer of admission to either USNA or a USNA-affiliated prep program. Sample restricted to domestic applicants who submitted complete applications and passed the fitness and medical exams. Data correspond to applicants for the USNA Classes of 2023–2027. Sample sizes are reported in parentheses.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, Table 3.6, at 31.

As shown, White applicants (44.1%) and those who declined to specify their racial background (38.9%) were the only subgroups with overall admission rates below 50%. By contrast, Black applicants had the highest offer rate at 77.9%, followed by Native American/Hawaiian applicants (61.5%), Asian applicants (61.1%), and Hispanic applicants (56.5%).

While these disparities are striking, they do not, in and of themselves, constitute definitive evidence of racial preferences. One reason is that some applicant pools are significantly more advantaged in the admissions process than others. Recall that Blue Chip Athletes (BCAs) are virtually guaranteed admission. Consequently, racial groups with disproportionately high numbers of BCAs may appear to benefit from racial preferences when, in reality, their elevated admission rates stem, at least in part, from athletic recruitment. Similarly, groups with higher proportions of prep school applicants will also exhibit inflated admission rates, as applying from a prep school—though less advantageous than being a BCA—still significantly increases the likelihood of admission.

Figure 2.1B. Distribution of Applicant Pools by Race Among USNA Admits

Note. Sample sizes reported in parentheses. Sample includes only applicants admitted to USNA. “BCA” refers to Blue Chip Athletes, while “Prep” refers to those admitted to USNA from the Naval Academy Preparatory School (NAPS) or another USNA-affiliated private prep program.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, Table 3.8, Panel B, at 35.

Figure 2.1B above helps illustrate this dynamic by showing how admitted students from each of the four major racial groups are distributed across different applicant pools. Black admits provide the clearest example of a group whose admission rates are inflated due to their disproportionate representation among BCAs and prep school applicants. Approximately 60% of Black admits were either BCAs or prep school applicants, compared to just 29% of White admits. However, the share of BCA and prep school applicants among Hispanic (35%) and Native American/Hawaiian admits (35%) was only slightly higher than among White admits, while the share among Asian admits was even lower. Therefore, the elevated admission rates for these three minority groups cannot be solely attributed to BCA or prep school pathways. As we will see, even after accounting for these factors, Black applicants continue to enjoy significantly higher admission rates than similarly qualified applicants of other racial groups.

2.2 Admission Rates to USNA

Given that the presence of BCA and prep school applicants can confound assessments of racial preferences in admissions, a more accurate analysis requires excluding these groups. This is reflected in Figure 2.2A, which presents admission rates for each racial group across three applicant pools: all applicants, applicants excluding BCAs, and applicants excluding both BCAs and prep school candidates. In the latter group, admission rates are relatively consistent across racial groups—with one notable exception: Asian applicants are admitted at a significantly higher rate (54.8%) than all others. Differences between White (36.1%), Black (37.3%), and Hispanic (35.9%) applicants are modest and not statistically significant.

Figure 2.2A. Admission Rates to USNA by Race and Applicant Pool

Note. Bars represent the percentage of applicants who received an offer of admission to USNA, disaggregated by applicant pool and race/ethnicity. The sample is restricted to domestic, complete applications that received a nomination and passed both the fitness and medical exams. Sample sizes for each racial/ethnic group are shown in the legend, while total sample sizes for each applicant pool are listed next to the corresponding labels.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, Table 3.7, at 33.

It would be a mistake, however, to interpret this relative parity as evidence that racial preferences in USNA admissions are minimal or non-existent. Such a conclusion assumes that applicants within each racial group are relatively equal across variables that significantly influence admission decisions. As shown in Figure 2.2B, this assumption does not hold. Among non-BCA/non-Prep applicants, Black admits score significantly lower than their White counterparts across all components of the WPM, including standardized test scores, class rank, and extracurricular evaluations. On several metrics—such as SAT scores—Black admits even score lower than or are comparable to White applicants who were denied admission to USNA. To a lesser extent, Hispanic admits also score lower than White admits on every metric included here, though they tend to receive somewhat higher RAB adjustments (not shown). Asian admits, meanwhile, outperform or match White admits on academic components of the WPM (e.g., SAT scores and class rank), as well as on the overall WPM. However, they tend to score significantly lower in areas such as athletic extracurricular involvement (Athletic Scores) and physical fitness (CFA Scores).

Figure 2.2B. Application Summary Statistics by Race, Removing BCAs and Prep Pool

Note. Sample (N=12,304; White N=8,022; Black N=778; Hispanic N=1,551; Asian N=1,444) restricted to domestic, complete applications that received a nomination and passed the fitness and medical exams. Athletic and Non-Athletic Scores measure extracurricular involvement in athletic and non-athletic activities, respectively. RSO Scores reflect character ratings from applicants’ high school math and English teachers. CFA Scores measure Candidate Fitness Assessment performance. Combined WPM Scores are a weighted composite of all listed factors except CFA Scores. Dashed horizontal lines indicate the average scores for admitted applicants.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, Table 3.16, at 47.

More compelling evidence of the extent of racial preferences emerges when examining admission rates by WPM decile, as shown in the right panel of Figure 2.2C. In general, racial preferences—particularly for Black applicants—are most pronounced in the lower-middle to upper-middle deciles of the WPM distribution, with disparities diminishing at the extremes. For instance, in the first decile, admit rates are similarly low across all racial/ethnic groups: 3.1% for White applicants, 4% for Black applicants, 3% for Hispanic applicants, and 4.7% for Asian applicants. However, by the 5th decile, the admit rates diverge sharply: 20.1% for White applicants, compared to 59.7% for Black applicants, 29.7% for Hispanic applicants, and 38.2% for Asian applicants.

Figure 2.2C. Racial Group Shares (Left) and Admission Rates (Right) by WPM-23 Decile, Excluding BCAs and Prep Applicants

Note. Sample restricted to non-Blue-Chip, non-Prep-Pool, domestic, complete applications that received a nomination and passed the fitness and medical exams, and that have a valid WPM Score. Sample sizes are reported in parentheses. Deciles are computed separately for each class year. WPM-23 refers to the raw WPM score (excluding RAB points), calculated using the 2023–2024 component weights for all admission cycles.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, Tables 3.19–3.20, at 50–51.

The scale of preferences for Black applicants becomes even more evident when comparing admission rates across deciles. For example, Black applicants in the 4th decile (47%) are admitted at a slightly higher rate than White applicants in the 8th decile (46.6%), despite being four deciles lower in the WPM distribution. By the 9th decile, admission rates across groups begin to converge, but significant disparities remain. White applicants in this decile have a 66% chance of admission, compared to 90.3% for Black applicants, 83.6% for Hispanic applicants, and 87% for Asian applicants.

A main reason why preferences tend to be strongest for Black applicants is reflected in the left panel of Figure 2.2C, which plots the racial group shares of each decile. Specifically, nearly two-thirds (65.8%) of non-BCA/non-prep Black applicants are concentrated in the bottom four deciles, compared to 46.7% of Hispanic, 31.6% of White, and 26.2% of Asian applicants. In contrast, just 17% of Black applicants place in the top four deciles, compared to 32.2% of Hispanic, 48.2% of White, and 52.7% of Asian applicants.

Given these distributions, a race-blind admissions system that selected all applicants from the top four WPM deciles would yield a markedly different racial composition among admitted students. Under such a system, the share of non-BCA/non-prep admits would shift to 71.4% White (compared to 61.3% in actuality), 14.3% Asian (16.7%), 8.5% Hispanic (11.8%), and just 1.9% Black (6.1%). If applied to the entire qualified applicant pool—meaning BCAs and prep applicants would no longer be virtual admission shoo-ins—Whites would constitute 70.4% of admits (compared to 58.7% in actuality), Asians 13.8% (14.3%), Hispanics 9.3% (12.5%), and Blacks 2.7% (10.5%).

Ultimately, USNA would be unable to increase Black—and, to a lesser extent, Hispanic—representation to levels even approaching their share of the U.S. population if all applicants were held to the same standards as White applicants. There are simply too few Black applicants in the top deciles for race-neutral policies to achieve this outcome. Even with racial preferences, Black applicants remain underrepresented among admits.

This imbalance—along with the smaller proportion of Hispanic applicants in the top deciles relative to Whites and Asians—may also explain why substantial racial preferences are extended to Asian applicants, despite their overrepresentation among admits relative to their share of the wider population. If USNA cannot achieve representational parity for each minority group individually, it can at least seek parity in the broader “non-White vs. White” sense--a metric it regularly tracks.⁶²

2.3 Modeling Racial Preferences in USNA Admissions

While Arcidiacono’s analysis has demonstrated that admission rates for non-BCA/non-Prep applicants vary significantly by race within the same WPM decile, the extent to which these disparities persist after accounting for additional factors remains an open question. To address this, Arcidiacono employs a series of eight logistic regression models, each progressively incorporating more control variables.

The results of these models are presented in Figure 2.3A, where the bars represent logit coefficients, indicating differences in the log odds of admission relative to White applicants. Each model includes a progressively richer set of controls, with Model 1 adjusting only for sex and class year. As in Figure 2.2A, this initial model shows that Asian applicants are more likely to be admitted than White applicants, while Black and Hispanic applicants are admitted at roughly similar rates. Female applicants (not shown) also receive a modest admissions boost relative to males, though the gap is smaller than that observed between Asian and White applicants.

Figure 2.3A. Logit Estimates of USNA Admissions, Excluding BCAs and Prep Applicants

Note. Bars represent logit coefficients, with error bars indicating 95% confidence intervals. In each panel, error bars that cross the dashed vertical line are not statistically significant at the 95% confidence threshold. Model sample sizes are reported in parentheses. Four observations with missing CFA scores are dropped from Model 3 onward. Model 8 further restricts the sample to observations with no missing values for household income, private high school attendance, or the percentage of the applicant’s high school cohort attending a four-year college.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rebuttal Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 222, Table 4.1R, at A21.

As additional controls are introduced, the coefficients for each racial group increase in magnitude. For instance, Model 3 accounts for first-generation college status, household income, and community-level socioeconomic indicators (e.g., the percentage of an applicant’s zip code eligible for free or reduced-price lunch and the average salary in the zip code). It also controls for individual WPM components (excluding RAB points) and CFA scores. After these adjustments, the coefficient for Black applicants jumps from 0.025 in Model 1 to 1.981, for Hispanic applicants from -0.020 to 0.839, and for Asian applicants from 0.720 to 0.955.

The next major increases in coefficients occur in Model 5, which incorporates interactions between WPM components and class year to account for changes in the WPM weighting formula introduced in the Class of 2025 admissions cycle. It also adjusts for slate characteristics, including the number of available vacancies, the method of nomination, the number of nominees, average WPM scores, and the distance between a nominee’s WPM and the top score on the slate. These controls capture differences in slate types (e.g., principal numbered-alternate vs. competitive) and the degree of competition faced by applicants. After accounting for these factors, the estimated racial preference for Black applicants rises sharply to nearly 3.0 (2.924), while the coefficients for Hispanic and Asian applicants also increase (1.117 and 1.405, respectively).

Adding BGO interview scores, legacy status (e.g., family members who attended USNA or another service academy), and RAB points for AP, IB, and honors coursework in Model 6 has virtually no effect on these coefficients. Further controlling for total RAB points in Model 7 leads to a slight increase in the coefficient for Black applicants (3.077) while leaving estimates for other racial groups largely unchanged.

Finally, Model 8 restricts the sample to applicants with complete data for household and community socioeconomic indicators, reducing the sample size by 38%. This adjustment accounts for missing data that disproportionately affect Black applicants and may otherwise overstate their socioeconomic advantage. As a result, the Black coefficient rises further to 3.7, while coefficients for other racial groups see only modest increases.

These results underscore a critical point: the apparent racial parity in baseline admission rates masks substantial disparities in applicant qualifications. When qualifications are statistically equalized, the magnitude of racial preferences in USNA admissions—particularly for Black applicants—becomes strikingly clear. Moreover, the fact that these estimates increase as additional control variables are introduced suggests that the preferences are even stronger than they initially appear. The models explain a significant share of admissions variability (Pseudo R² values of 0.537 and 0.561 for Models 7 and 8), reinforcing the robustness of these findings.⁶³

This robustness is also directly at odds with USNA’s rebuttal, which claims that Arcidiacono’s estimates are inflated due to omitted variable bias. To the contrary, as later sections will discuss, omitted variables do not appear to drive these results. If anything, these estimates are likely conservative, given that the models account for a broad range of applicant characteristics yet still reveal striking racial disparities in admissions outcomes.

2.4 Quantifying Racial Preferences in Practical Terms

At this point, a casual reader or someone less familiar with statistical analysis might wonder: what do the preceding logit coefficients actually mean in practical terms?

One way to illustrate the real-world impact of these estimates is to assess how a White applicant’s admission probability would change if they were evaluated as a member of a different racial group. Professor Arcidiacono conducts such an analysis using the estimates from Model 6 in Figure 2.3A, which he refers to as his “preferred model”.⁶⁴ To estimate these probabilities, Arcidiacono applies the logistic transformation to the model coefficients, adjusting the applicants’ race while holding all other variables constant.⁶⁵ This analysis is performed for all class years combined (the “pooled” model) and separately for applicants to the 2023–2024 and 2025–2027 admission cycles to changes to the WPM formula.

The results, shown in Figure 2.4A below, translate abstract statistical coefficients into real differences in admission odds. Beginning with the pooled models (dark blue bars) in the top row, Professor Arcidiacono estimates that a White applicant with a modest 5% probability of admission would have a 50.3% chance of admission if treated as Black, a 14.8% chance if treated as Hispanic, an 18.3% chance if treated as Asian, and a 15.3% chance if treated as Native American/Hawaiian. Notably, the admission ‘bonus’ for being treated as a black applicant increased from 40% in the 2023–2024 cycle to 59.4% in the 2025–2027 cycle. This latter period coincided, perhaps not coincidentally, with heightened public focus on racial equity following the killing of George Floyd. Turning to the right panel, a White applicant with a 25% chance of admission would see their probability rise to 85.7% if treated as Black, 51.3% if treated as Hispanic, and 59.1% if treated as Asian.

Figure 2.4A. Estimated Admission Probability of a White Applicant if Treated as a Different Race

Note. Bars represent estimated admission probabilities for White applicants if treated as a different racial group, based on Model 6 from Figure 2.3A. Estimates are presented for all five class years combined (pooled model) and separately for the 2023–2024 and 2025–2027 admissions cycles. Dashed vertical lines denote the different White admission probabilities.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rebuttal Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 222, at A22.

Another way to quantify racial preferences is to estimate how admission probabilities would change if all applicants were evaluated under the same criteria as White applicants. To this end, and again using estimates from his preferred model, Professor Arcidiacono calculates the average marginal effects for each group. These represent the difference between the average admission probability for a member of a given racial group and the average admission probability for a White applicant, effectively isolating the role of racial preferences in determining admission outcomes.

Figure 2.4B presents these results. Across all five admission cycles, Black applicants would see their average admission probability drop by 24.3 percentage points, from 37.4% to 13.1%, if held to the same standards as White applicants. Similarly, the average Hispanic applicant would experience an 11.1-point reduction (35.9% to 24.8%), and the average Asian applicant would see a 17.6-point reduction (54.8% to 37.2%). Once again, the estimated drop in admission odds for Black applicants is larger in 2025–2027 (-29 points) than in 2023–2024 (-18.6 points). It also more than doubled in size for Hispanic applicants between these periods, from -7 points in 2023-2024 to -14.5 points in 2025–2027.

Figure 2.4B. Estimated Change in Admission Probabilities if Applicants Were Evaluated as White

Note. Bars represent the estimated change in average admission probability for each racial group if evaluated under the same criteria as White applicants, based on Model 6 from Figure 2.3A. Dark blue bars indicate the actual admission rate with racial preferences, green bars represent the estimated admission rate without racial preferences, and light blue bars denote the average marginal effect (AME), or the difference between the two. Estimates are presented for all five admission cycles combined and separately for the 2023–2024 and 2025–2027 cycles.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rebuttal Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 222, at A23.

Still another way of demonstrating the practical effects of racial preferences is to consider how many admits who benefited from racial preferences would still have been admitted in their absence. Professor Arcidiacono tackles this question using a Bayesian framework, which leverages the estimated admissions model to calculate admission probabilities both with and without racial preferences. By comparing these probabilities, the analysis accounts for observed characteristics and adjusts for the unobserved traits that likely played a role in securing admission under the status quo. This approach offers a clearer estimate of how many admits depended on racial preferences to secure admission.

Results for this analysis are graphed in Figure 2.4C. Most striking are the findings for Black admits, only 33.2% of whom, on average across the five admission cycles, would have been admitted in the absence of racial preferences. As shown by the blue bars, this represents a more than 50% reduction in admission odds for a staggering 73% of Black admits—a figure that rises to 78.1% during the 2025–2027 period. In contrast, clear majorities of admits from the other three racial groups would still have gained admission even without the benefit of racial preferences. However, their numbers would decline by approximately a third, with 68% to 70% of Hispanic and Asian admits “surviving” the removal of preferences.

Figure 2.4C. Estimated Probability of Admission for Previously Admitted Applicants if Racial Preferences Were Removed

Note. Bars represent the estimated share of previously admitted applicants who would still have been admitted under a race-neutral model. Estimates are based on Bayesian posterior probabilities derived from Model 6 in Figure 2.3A. Green bars indicate the share of admits who would still gain admission without racial preferences, while dark blue bars represent the share of admits whose admission probability would drop by more than 50% under a race-neutral model. Estimates are presented separately for all five admission cycles combined, as well as for the 2023–2024 and 2025–2027 cycles.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rebuttal Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 222, at A24.

2.5 Racial Preferences Across Admissions Channels

To this point, Professor Arcidiacono’s analyses have quantified racial preferences in USNA admissions as a whole, without examining how these preferences vary across the distinct admission channels discussed in Section 1. His next analysis addresses this gap by employing a nested logit model, which estimates the probability of earning admission through one channel relative to another. This method accounts for the fact that candidates do not apply separately to each pathway but instead are sorted into different categories by USNA. As such, it allows for a more precise examination of how racial preferences might differentially impact decisions across pathways, such as winning a Congressional slate, being selected as a Qualified Alternate (QA) or Additional Appointee (AA), or securing admission as a service-connected nominee.

Figure 2.5A visualizes the logit coefficients for each non-White racial group, derived from four related models adjusted for all control variables in Professor Arcidiacono’s preferred specification. The first model, limited to Congressional nominees with WPM scores above the 150th-ranked QA cutoff, examines the likelihood of being selected as a QA versus winning a Congressional slate. Notably, the only statistically significant coefficient is for Black nominees, who have a substantial coefficient of 0.893. This finding is surprising given that QAs are ostensibly selected from the highest WPM-scoring Congressional nominees who fail to win a slate—yet Black applicants consistently have the lowest average WPM scores across all admission channels.⁶⁶

Figure 2.5A. Nested Logit Estimates by Admission Channel

Note. Bars represent logit coefficients, with error bars indicating 95% confidence intervals. Error bars that cross the dashed vertical line are not statistically significant at the 95% confidence level. All models exclude Blue Chip Athletes and applicants from the Prep Pool. Sample sizes are as follows: 2,008 for the Qualified Alternate model, 941 for the Additional Appointee model, 10,743 for the Congressional Slate Winner model, and 1,536 for the Service-Connected model.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rebuttal Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 222, Table D.86R, at D88.

A second model, consisting of Congressional nominees with WPM scores below the 150th-ranked QA cutoff (rendering them ineligible for QA appointments), estimates the odds of being admitted as an Additional Appointee (AA) versus winning a Congressional slate. Recall that USNA explicitly acknowledges that race and ethnicity may be considered “as part of a holistic consideration of a candidate” when making AA selections. Unsurprisingly, the coefficients for Asian (1.103), Black (2.216), and Hispanic (1.063) nominees in this model are all positive and substantial, reflecting the influence of racial preferences in this admission channel.

A third model includes all Congressional nominees and examines the odds of winning a Congressional slate versus being denied admission altogether. As its results largely mirror those from Model 6 of Figure 2.3A—with substantial coefficients for each group, though slightly smaller, reflecting the stronger preferences observed in the QA and AA channels—I proceed directly to the fourth and final model. This model estimates the odds of earning admission as a service-connected nominee relative to being rejected. Here, too, the coefficients for each racial group are positive, statistically significant, and varyingly large. Notably, the Black coefficient (2.310) closely resembles that observed in the ‘Congressional Slate Winner vs. No Admission’ model, suggesting that racial preferences for Black applicants are comparably strong across admission channels. In contrast, the coefficients for service-connected Asian (0.600 vs. 1.318) and Hispanic (0.908 vs. 1.091) nominees are considerably and slightly lower than for their Congressional counterparts, respectively.

Before concluding this discussion, I revisit the question of racial preferences in the QA channel, one of the few admissions pathways that USNA explicitly claims to be ‘free’ of racial considerations. This claim hinges on the premise that (medically and physically qualified) Congressional nominees selected as QAs are chosen strictly according to their WPM score rankings. However, there is reason to question USNA’s strict adherence to this rule, as WPM score rankings fluctuate considerably across each admissions cycle.

These fluctuations are largely driven by ‘initial’ QAs either declining admission offers or being designated as winners of Congressional slates, thereby creating openings for lower-ranked candidates. For instance, during the most recent admissions cycle in the data (2027), provisional QA lists obtained by SFFA show that the QA cutoff—the 150th highest WPM score—initially reached 74,851. Yet, the actual observed cutoff was 70,478.⁶⁷ Further analysis reveals that 74 of the 150 candidates ultimately selected as QAs had WPM scores below the highest provisional cutoff.

One way to assess whether racial preferences influence QA assignments is to compare the racial composition of QA candidates and QA admits who scored above the initial QA threshold with those who scored below it. Figure 2.5B visualizes this comparison, presenting data for QA-eligible candidates across all five admission cycles (left panel) and for candidates from the 2025–2027 cycles (right panel). Above the initial QA threshold, the racial composition of candidates and QA admits closely aligns. Below the threshold, however, this alignment diverges significantly. White applicants comprise just over 72% of all candidates but less than 55% of QA admits, resulting in a QA rate of 27.5%. By contrast, Black, Hispanic, and Asian applicants make up 2.9%, 7.6%, and 13.4% of all candidates but represent 5.7% (QA rate of 70.4%), 9.6% (45.7%), and 26.6% (71.8%) of QA admits, respectively. These disparities are even more pronounced during the 2025–2027 admissions cycles, where the QA rate for White applicants was 31.4%, compared to 85.6% for Black, 63% for Hispanic, and 81.7% for Asian applicants.

Figure 2.5B. Racial Composition of QA Candidates and Admits by Initial QA Cutoff

Note. Sample restricted to applicants eligible for Qualified Alternate (QA) consideration who were not admitted through another channel. To be included, applicants must have exceeded the final QA threshold. Sample sizes for Panel A outcomes are reported in parentheses. Blue and teal bars represent the share of candidates and admits who placed above the initial QA threshold (i.e., the 150th-ranked WPM score), while red, orange, and green bars represent the share of candidates, admits, and admit rates for those below the initial threshold.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, Table 4.6, at 74.

Together with the results from the nested logit model, these findings demonstrate that strong racial preferences are present across all USNA admissions channels, including those ostensibly ‘off-limits’ to such considerations. At the same time, they reveal that the magnitude of these preferences varies by racial group and admission channel. Preferences for Black applicants are consistently the largest and relatively uniform across channels. In contrast, preferences for Asian and Hispanic applicants, while smaller overall, show more variation, with stronger effects observed in Congressional channels compared to service-connected ones.

As significant as these preferences are in the direct admission channels, they are even more pronounced for most groups in the indirect ‘prep school’ channel—a topic we turn to next.

2.6 Admissions to NAPS

Up to this point, the analyses have excluded applicants applying from NAPS and other USNA-affiliated prep schools, as these candidates are virtually guaranteed admission to USNA. However, such applicants account for nearly 20% of incoming USNA classes. Any comprehensive audit of racial preferences in USNA admissions would be incomplete—and risk underestimating their scope—if these indirect admission channels were not considered.

As NAPS accounts for more than 80% of all prep admits, and because its medical and physical eligibility requirements align closely with those of USNA, Professor Arcidiacono’s analysis focuses specifically on this program. In addition to being medically and physically qualified, NAPS-eligible applicants must meet three primary criteria: a) they must have been denied direct admission to USNA, b) they must have submitted a complete USNA application, and c) they must not have previously attended or currently be attending another prep program. Recall that securing a nomination is not a prerequisite for admission to NAPS. However, nearly 70% of those admitted during the 2023–2027 admission cycles were, in fact, nominees, reflecting the overlap between NAPS admits and the broader nomination process.⁶⁸

Notably, prep school admissions are not among the areas in which USNA acknowledges considering race as a factor, suggesting an implicit denial of its use in these decisions.⁶⁹

The first thing that jumps out in these data is that, regardless of nomination status, most Black applicants who are denied direct admission into USNA are offered admission to NAPS. In fact, Black non-nominees (69.3%) are admitted at significantly higher rates than their nominee counterparts (52.3%). As a result, while Blacks constitute just 8.6% of the total NAPS-eligible sample, they account for 35.2% of all NAPS admits.⁷⁰ By contrast, White applicants, who comprise 66.6% of eligible applicants, represent only 33.6% of admits—an overall admission rate of just 7.3%. Except for applicants who declined to specify their racial/ethnic backgrounds (7.2%), no other group has a lower overall admit rate. Hispanic applicants are admitted at a rate of 22.2%, Asians at 13.1%, and Native American/Hawaiians at 25.9%.

Figure 2.6A. Admission Rates to NAPS by Race and Nomination Status.

Note. Bars represent the percentage of applicants who received an offer of admission to NAPS, disaggregated by race and nomination status. The sample is restricted to domestic, complete applications that passed the fitness and medical exams, were not admitted directly to USNA, and did not come from the prep pool. Sample sizes for each racial/ethnic group are reported in parentheses next to their labels, while total sample sizes for each applicant pool are listed in the legend.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, Table 3.21, at 54.

However, these figures may be misleading for two reasons. First, the sample includes what can be termed ‘latent BCAs’—i.e., athletic recruits who are not yet officially designated as BCAs but who will be assigned that status when they matriculate at USNA the following year. Second, as previously noted, USNA explicitly prioritizes enlisted sailors and Marines for NAPS admissions. Thus, to the extent that black applicants and those from other minority groups are disproportionately represented within these subpopulations, their higher admission rates may partly reflect these institutional priorities rather than exclusively racial preferences.

To address these and other potential sources of bias, Professor Arcidiacono begins by identifying and excluding applicants who are designated as BCAs in the subsequent USNA admissions cycle. Since this adjustment cannot be applied to those admitted to NAPS during the 2027 cycle (due to the absence of 2028 data), the analysis is restricted to the 2023–2026 cycles. Additionally, Professor Arcidiacono employs a series of logistic regression models that control for various factors, including whether applicants received service-connected nominations from the Secretary of the Navy, which are specifically allocated to enlisted Navy and Marine personnel.⁷¹

Results from these models, presented as logit coefficients, are plotted in Figure 2.6B. While some controls slightly moderate the coefficients for racial minority groups, these coefficients remain large and statistically significant across all models. For instance, in Model 1—which controls only for sex, first-generation college status, class year fixed effects, and nomination status—the Black coefficient is 2.907. By Model 4, which incorporates all controls from Arcidiacono’s preferred USNA admissions model along with indicators for service-connected nominations, this coefficient declines only marginally, to 2.874. A similar pattern holds for the smaller but still substantial coefficients for other racial groups. Collectively, these results strongly indicate that the large disparities in admission rates observed in Figure 2.6A are indeed driven by racial preferences rather than race-neutral factors.

Figure 2.6B. Logit Estimates of NAPS Admissions

Note. Bars represent logit coefficients, with error bars indicating 95% confidence intervals. Error bars that cross the dashed vertical line are not statistically significant at the 95% confidence threshold. Model sample sizes are reported in parentheses. A small number of observations with missing WPM components or CFA scores are dropped from Models 3 onward. The sample excludes Future Blue Chip Athletes, applicants from the Class of 2027 admissions cycle, and those with missing BGO interviews.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rebuttal Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 222, Table 4.7R, at A26.

To illustrate the practical effects of these preferences, Professor Arcidiacono employs the same quantification methods used in the earlier USNA admission models. Figure 2.6C, for instance, shows how a White applicant's probability of receiving a NAPS offer would change if treated as a member of another racial group. These estimates are derived from a fifth model, which builds on Model 4 from Figure 2.6B by adding an interaction term between race and an indicator variable for class years 2025 and beyond. This addition is critical, as it reveals that racial preferences in NAPS admissions expanded significantly between the 2023–2024 and 2025–2026 admission cycles. For example, a White applicant with a 5% chance of receiving a NAPS offer would see their probability rise to 37% if treated as Black during the 2023–2024 cycles—and to an astonishing 68% if treated as Black during the 2025–2026 cycles. Similarly, a White applicant with a 25% chance would see their odds rise to 79% during the 2023–2024 cycles and to 93% during the 2025–2026 cycles.

Figure 2.6C. White NAPS Admission Probability (%) if Treated as a Different Race

Note. Results are based on Model 5 from the pooled NAPS logit model. This model follows the same specification as Model 4 in Figure 2.6B but includes an interaction term (Race × 1[ClassYear ≥ 2025]) to capture shifts in the magnitude of racial preferences in the post-2024 admissions cycles. Dashed vertical lines in each panel denote different White admission probabilities.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rebuttal Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 222, Table 4.8R, at A27.

The magnitude of racial preferences in NAPS admissions is particularly striking when examining their impact on Black admit rates, as shown in Figure 2.6D. Absent preferences, the Black admit rate would plummet from 33.2% to 9.7% during the 2023–2024 cycles and from 70.7% to 18.4% during the 2025–2026 cycles—a staggering drop of over 52 percentage points in the latter period. Moreover, Figure 2.6E reveals that only 25% to 26% of the black applicants admitted to NAPS across these two periods would have still been admitted without the benefit of preferences, with 89% of 2023-2024 admits and 85% of 2025-2026 admits seeing a more than 50% drop in their admission probabilities, respectively.

Figure 2.6D. Estimated Change in Admission Probabilities if Applicants Were Evaluated as White

Note. Bars represent the estimated change in average admission probability for each racial group if evaluated under the same criteria as White applicants, based on estimates from Model 5 of the pooled NAPS logit models. Green bars indicate the actual admission rate with racial preferences, light blue bars represent the estimated admission rate without racial preferences, and dark blue bars denote the average marginal effect (AME)—the difference between the two. Estimates are presented separately for the 2023–2024 and 2025–2026 admissions cycles.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rebuttal Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 222, Table 4.9R, at A27.

Figure 2.6E. Estimated Probability of Admission for Previously Admitted Applicants if Racial Preferences Were Removed

Note. Bars represent the estimated share of previously admitted applicants who would still have been admitted under a race-neutral model. Estimates are based on Bayesian posterior probabilities derived from Model 5 of the pooled NAPS logit models. Green bars indicate the share of admits who would still gain admission without racial preferences, while dark blue bars represent the share of admits whose admission probability would decrease by more than 50% under a race-neutral model. Estimates are presented separately for the 2023–2024 and 2025–2026 admissions cycles.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rebuttal Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 222, Table 4.10R, at A28.

While this discussion has largely focused on Black applicants and admits, who benefit from the largest preferences by far, the preferences extended to other racial groups are far from negligible. For example, as shown in Figure 2.6E, only 45.6% of Hispanic admits during the 2025–2026 cycles would have been admitted without preferences, down from 59.2% in the 2023–2024 cycles. Similarly, just 48.8% of Asian admits (compared to 62.3% in the earlier cycles) would have gained admission in the absence of these preferences.

These findings, when combined with the results from the USNA admissions models, demonstrate that racial preferences are deeply embedded in every major pathway to admission. If USNA’s definition of ‘limited’ racial consideration includes preferences of this magnitude, the term has effectively lost all meaning.

2.7 The Combined Effect of Racial Preferences

If racial preferences in preparatory school admissions ultimately contribute to and magnify racial preferences in direct USNA admissions, what then is their total combined impact on the racial composition of USNA admits?

To answer this question, Arcidiacono runs a “capacity constraint” simulation, modeling what USNA’s admitted class would look like under a race-neutral system. This approach ensures a direct apples-to-apples comparison by holding the total number of admitted students constant while redistributing seats based on academic and non-racial qualifications. However, because prep school admits do not matriculate into USNA until the following admission cycle, modeling their contribution requires a steady-state assumption—meaning the number of applicants to NAPS in a given year is expected to be roughly equal to the number of applicants transitioning from NAPS to USNA in that same year. Additionally, due to the small number of non-NAPS prep students, the simulation assumes they have comparable qualifications to NAPS students and treats all prep program admits as a single group. These simplifying assumptions allow for a clearer estimate of how race-neutral admissions policies would alter the overall composition of USNA admits.

The top row of panels in Figure 2.7A visualizes each racial group’s share of direct (i.e., non-prep) USNA admits, USNA admits from the prep school pipeline, and overall USNA admits under both the current race-conscious admissions system and a hypothetical race-neutral alternative. The bottom row presents the corresponding absolute numbers, illustrating the concrete impact of eliminating racial preferences on the total number of admits from each group. For Whites, the results are striking: under a race-neutral system, the White share of overall admits would increase by 9.4 percentage points, rising from 58% to 67.4%, resulting in a net increase of 649 White admits over the 2023–2027 admissions cycles. Given that most White applicants are admitted directly rather than through the prep school pipeline, the majority of this increase comes from direct (non-prep) admits, where the White share grows from 63.2% to 70.9%, adding 434 additional White admits. The prep school pipeline, however, also plays a significant role, with White prep admits increasing from 35.1% to 52.1%, yielding an additional 215 admits.

Figure 2.7A. Simulating the Total Effect of Removing Racial Preferences on the Racial Composition of USNA admits

Note. This simulation applies Arcidiacono’s preferred pre- and post-period Models 5, setting all race coefficients and interactions to zero while adjusting predicted admission probabilities to maintain a constant total number of admits. Blue Chip Athlete admissions are treated as fixed since they are excluded from the estimated models. Sample sizes for each admission pool are reported in parentheses. Each racial group has two bars, with results shown across two rows and three columns. In the top row, green bars represent each group’s actual share of USNA admits, while dark blue bars indicate their share under a race-neutral model. The bottom row presents absolute admit numbers, where green bars reflect the number of admits under the status quo and dark blue bars represent admissions if racial preferences were removed. The three columns distinguish between applicants admitted directly to USNA (USNA Non-Prep Admits), applicants admitted through a prep program (USNA Admits from Prep), and the total of both pathways (Overall Admits).
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Richard D. Kahlenberg, Plaintiff's Exhibit 219, Appendix D, Table 1, at 6 (D. Md. July 15, 2024).

By contrast, Black, Hispanic, and Asian applicants would experience declines in admission shares, with the magnitude varying by group. Black admits, for example, would see their overall share fall from 11.5% to 6.9%, a net loss of 314 admits. This reduction is fairly evenly split between direct admissions (164 fewer Black direct admits) and the prep school pipeline (150 fewer Black admits via prep). Similarly, Hispanic admits would experience a 2.1 percentage point decline, translating to 145 fewer total admits, with 102 of those losses occurring in direct admissions. Asian admits would also decline, dropping from 14% to 11.5%, resulting in 173 fewer Asian admits, with nearly all of that reduction (163 admits) coming from the direct admissions process rather than the prep pipeline.

These figures make clear that racial preferences at USNA have a substantial effect on admissions outcomes, with preferences for Black, Hispanic, and Asian applicants coming at the direct expense of White applicants. The fact that eliminating racial preferences leads to such dramatic shifts in admit numbers underscores the extent to which racial classifications drive USNA’s admissions decisions—far beyond what could be justified as a mere “plus” factor.

2.8 Why Racial Preferences May Be Even Larger Than Estimated

As substantial as the racial preferences estimated in Professor Arcidiacono’s analysis appear to be, there are compelling reasons to believe they may be even greater in practice. One key indicator is the consistent increase in the logit coefficients for each racial group as additional control variables were added to Arcidiacono’s USNA admissions models. This pattern suggests that, far from being overstated, the estimated racial disparities are more likely to grow if additional unobserved factors—currently excluded due to a lack of access to relevant data—were incorporated into the models. In simpler terms, just as Black and other non-White applicants perform worse than their White counterparts on observable admissions metrics, they likely perform worse on unobserved factors as well. If this is the case, failing to account for these unobservables results in an underestimation of the extent to which racial preferences shape USNA admissions outcomes.

Naturally, it is impossible to test this hypothesis directly without access to these unobserved factors. However, its implications can be examined indirectly. If non-White applicants tend to be weaker than White applicants on unobservable characteristics, we would expect to see corresponding performance disparities among matriculated students. More specifically, non-White midshipmen should not only underperform their White peers on average, but some portion of this underperformance should persist even after controlling for all available academic, physical, and extracurricular credentials. The presence of such a residual gap would serve as indirect evidence that important unobserved factors—ones not accounted for in the admissions models—are driving these disparities, further reinforcing the conclusion that the estimated racial preferences in USNA admissions are understated.

Fortunately, and perhaps inadvertently, SFFA was able to obtain at least some individual-level performance data for midshipmen in the USNA Classes of 2025 and 2026, including course grades.⁷² Additionally, publicly available records on midshipmen who made the Commandant’s List each semester provide further insight, spanning all five class years under study.

2.8.1 Disparities in USNA Performance Outcomes

The available individual-level course grade data spans the Fall semesters for the USNA Classes of 2025 and 2026, covering final grades in Math, Science, and English courses. Figure 2.8A compares mean final grades across these subjects, disaggregated by race and matriculant pool. Overall, Black plebes consistently earn the lowest grades across all pools, followed by Hispanic plebes, whose grades are slightly lower than those of White plebes. Asian plebes perform on par with White plebes, while White plebes generally achieve the highest grades. Notably, Black-White disparities are most pronounced in Science courses (chemistry, physics), while overall semester QPA (Quality Point Average) differences are largest within the non-prep/non-BCA pool—the very subpopulation for which racial preferences play the most significant role in direct admissions.

Figure 2.8A. Average Fall Semester Course Grades by Matriculant Pool and Race

Note. Bars in each panel represent the average Fall Semester course grades (1=D, 2=C, 3=B, 4=A) for the USNA Classes of 2025–2026, except for the bottom-right panel, which reports credit-weighted average grades (i.e., Quality Point Average/QPA) across the three courses. “BCA” refers to Blue Chip Athletes, and “Prep” includes those who matriculated to USNA from a NAPS, Foundation, or Civilian Prep program. QPA estimates for BCA/Prep matriculants are not available. Dashed black vertical lines represent sample means. The sample is restricted to domestic USNA matriculants with complete WPM information. The full sample size is 2,257 (White N=1,312, Black N=252, Hispanic N=305, Asian N=314). Three observations were excluded due to missing WPM information.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, at 90, D124–D127 (Table 5.1, Tables D.114, D.116, D.118, and D.121)

These disparities extend beyond final grades. As shown in Figure 2.8B, non-prep/non-BCA Black (and, to a lesser extent, Hispanic) plebes are significantly more likely than their White counterparts to be enrolled in remedial Math (16.3% vs. 5%) and English (7.6% vs. 0.8%) courses. This suggests that observed racial disparities in course grades would be even larger if adjusted to account for differences in remedial enrollments, as remedial courses cover less advanced and may be graded on a more lenient scale.

Figure 2.8B. Fall Semester Remedial Course Rates

Note. Sample includes domestic, non-international students from the USNA Classes of 2025 (N=1,127) and 2026 (N=1,133). “BCA” refers to Blue Chip Athletes, and “Prep” includes those who matriculated to USNA from a NAPS, Foundation, or Civilian Prep program. The left panel reports the percentage of students in each racial and matriculant group enrolled in the remedial English course (HE101: Practical Writing), while the right panel reports the percentage enrolled in the remedial Math course (SM005: Pre-Calculus). Sample sizes for each racial/ethnic group are reported in parentheses next to their labels, while total sample sizes for each applicant pool are listed in the legend. No BCA/Prep Asian matriculants were enrolled in a remedial course, resulting in the absence of a corresponding bar for this group.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, Table D.29, at D30.

Even more telling are the racial disparities in rates of appearing on the Commandant’s List—a merit designation requiring a minimum semester QPA of 2.9 (equivalent to a B to B-) and at least ‘B’ grades in non-academic areas such as Aptitude for Commission, Conduct, and Physical Readiness. Given these relatively modest requirements, nearly half (47.6%) of all 2023–2027 matriculants—including 57.2% of those in the non-BCA/non-prep pool—appeared on the list at least once. Yet, as shown in Figure 2.8C, only 20% of Black matriculants (35% of Black non-BCA/non-prep plebes) met this threshold, compared to 53% of White matriculants (60.6% in the non-BCA/non-prep pool), 50.8% of Asians (55.5% in the non-BCA/non-prep pool), and 43% of Hispanics (52.7% in the non-BCA/non-prep pool).

Figure 2.8C. Percent appearing on Commandant’s List at least once by Matriculant Pool and Race

Note. Bars represent the percentage of matriculants who appeared on at least one Commandant’s List during any semester from the 2020–2024 academic years. Sample includes domestic USNA matriculants with complete WPM information from the Classes of 2023–2027. “BCA” refers to Blue Chip Athletes, while “Prep” includes those who matriculated to USNA from a NAPS, Foundation, or Civilian Prep program. Sample sizes for each matriculant group are reported in parentheses.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, Table 5.1, at 90.

While these performance disparities are unsurprising given the substantial racial preferences afforded to Black applicants, the key question is whether they persist after controlling for all observable admissions criteria, such as high school class rank, SAT scores, and extracurricular achievements. If they do, it would indicate that Black applicants—and perhaps other groups—systematically underperform relative to their observed qualifications. This, in turn, would suggest that these groups also perform worse on unobserved factors not captured in Arcidiacono’s admissions models, further implying that the actual magnitude of racial preferences at USNA is even greater than his estimates suggest.

With respect to Fall semester course grades, Model 4 in Figure 2.8D shows that controlling for all variables in Arcidiacono’s preferred specification nearly eliminates Black-White disparities in Calculus. Disparities in Science and English also shrink but remain statistically significant. In contrast, grade differences between Whites and other racial groups are relatively minor and, in most cases, not statistically significant, even in the minimally adjusted Model 1.

Figure 2.8D. OLS Estimates of USNA Grades by Course Type

Bars represent OLS regression coefficients estimating differences in course grades relative to White matriculants (reference group), with grades measured on a 4-point scale (1=D, 2=C, 3=B, 4=A). Error bars denote 95% confidence intervals. Coefficients with error bars crossing the dashed vertical line are not statistically significant at the 95% confidence threshold. Model sample sizes are reported in parentheses. Two observations with missing CFA scores are excluded from Model 4.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rebuttal Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 222, Table 5.3R, at A33.

Controlled models of Commandant’s List appearances arguably provide an even stronger test of the persistence of performance disparities, as this outcome spans the length of a midshipman’s tenure at USNA and encompasses both academic and non-academic performance indicators. Results from these logit models are presented in Figure 2.8E, with Model 5 corresponding to Arcidiacono’s preferred specification. Unlike the earlier Fall course grades model, racial disparities in Commandant’s List appearances remain significant across all minority groups relative to Whites, even after controlling for all observable predictors of academic success. The baseline Black-White disparity is the largest at -1.529 in Model 1, though it declines to -0.537 in Model 5. The Asian-White disparity, by contrast, is the smallest at baseline (-0.074) but actually increases to -0.216 when observables are equalized. The Hispanic-White disparity falls between the two extremes, declining from -0.391 in Model 1 to -0.092 in Model 5.

Figure 2.8E Logit Estimates of Appearing on a Commandant’s List

Note. Bars represent logit regression coefficients estimating the likelihood of appearing on a Commandant’s List at least once, relative to White matriculants (reference group). Error bars denote 95% confidence intervals. Coefficients with error bars crossing the dashed vertical line are not statistically significant at the 95% confidence threshold. Model sample sizes are reported in parentheses. Twelve observations with missing CFA scores are excluded from Models 3 and beyond. Model 6 further restricts the sample to applicants with no missing values for household income, private high school attendance, the percentage of their high school attending four-year colleges, or the percentage of their high school eligible for free and reduced-price lunch.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rebuttal Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 222, Table D.112R, at D118–D120.

In sum, results from the performance models indicate that non-White matriculants—especially Black midshipmen—perform worse, and White matriculants better, than their observable qualifications predict. These findings strongly suggest that critical competence-related variables were omitted from Arcidiacono’s admissions models. If included, these variables would likely yield even larger estimates of racial preferences at USNA.

2.8.2 Additional Sources of Underestimation

‘Early Notify’ Recommendations

Professor Arcidiacono concludes his report by identifying several additional areas where racial preferences likely influence admissions outcomes but are not fully captured in his models. The first concerns the issuance of Letters of Assurance (LOAs)—conditional admission offers for which USNA concedes that race can be a “non-determinative” factor. LOAs play a critical role in shaping the final pool of admitted students, with 84% of those admitted to the Classes of 2025–2027 having been recommended by the admissions board to receive one.⁷³

For many applicants, receiving an LOA significantly reduces the uncertainty of waiting for a final decision, making them more likely to commit to USNA rather than accept offers from competing institutions. Conversely, applicants who do not receive an LOA must either wait and risk rejection or accept another offer. In this way, LOAs not only act as an early selection mechanism but also functionally influence yield rates, disproportionately benefiting those who receive them. If White applicants are less likely to receive LOAs, this would create a systemic disadvantage not fully accounted for in traditional admissions models.

Since USNA did not provide direct data on LOA recipients, Professor Arcidiacono approximated this outcome using the “Most Recent Board Result” field. Available only for the 2025–2027 admission cycles, this field includes an “Early Notify” category, which is used by the admissions board to recommend LOA recipients. While not a perfect substitute for direct LOA data, this approximation allows for a reasonable assessment of how LOAs are distributed and whether race plays a role. Among applicants in the 2025–2027 cycles, those designated as “Early Notify” had an admit rate of 79.5%, compared to just 16.5% for those classified only as “Qualified”.⁷⁴

The top row of panels in Figure 2.8F displays the percentage of applicants classified as ‘Qualified’ and, among that group, their likelihood of being recommended for ‘Early Notify,’ disaggregated by race. Notably—focusing on Panel B—non-BCA/non-prep White applicants have one of the highest ‘Qualified’ rates (73.7%), second only to Asians (75.3%). Yet, paradoxically, they have the lowest ‘Early Notify’ rate (39.7%) of any group. In stark contrast, Black applicants have the lowest ‘Qualified’ rate (37.1%) but the highest ‘Early Notify’ rate (67%).

This pattern is particularly striking given that LOAs are generally issued to candidates who score above 70,000 on the WPM. The discrepancy becomes even more apparent in the bottom row of Figure 2.8F: the average WPM of Black ‘Early Notify’ designees (67,964) is actually lower than that of ‘Qualified’ White applicants (68,092) who were not recommended for an LOA.

Figure 2.8F. Qualified Share, Early Notify Share of Qualified, and Average WPM Scores by Race

Note. Sample restricted to domestic, complete applications that had a Most Recent Board Result (excluding “USNA Deferred”) and a WPM Score. “Qualified” refers to applicants with a Most Recent Board Result of “Early Notify”, “Qualified”, or “Qualified Prep Pool”.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rebuttal Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 222, Table 6.2, at 98.

Further reinforcing these disparities, a series of logit models regressing ‘Early Notify’ designations on race—while controlling for the same variables used in Arcidiacono’s admissions models—show that racial preferences become even more pronounced when additional applicant characteristics are accounted for.⁷⁵

Initially, when controlling only for gender and socioeconomic indicators, Black applicants appear somewhat less likely than Whites to be recommended for ‘Early Notify’ (logit coefficient = -0.349). However, once WPM components are included, this coefficient reverses direction and surges to 2.292, rising further to 2.909 in the final model. A similar pattern emerges for Asian and Hispanic applicants, whose initial coefficients of 0.621 and -0.082 grow to 1.364 and 1.067, respectively.

These findings strongly suggest that race is a significant factor in LOA recommendations, contradicting USNA’s claim that race plays only a “non-determinative” role in admissions. Given the heavy weighting of LOAs in shaping the admitted class, this further indicates that racial preferences systematically disadvantage White applicants—not only in direct admission decisions but also in their likelihood of committing to USNA.

Beyond LOAs, Professor Arcidiacono identifies three additional areas where race likely influences admissions outcomes—either directly or indirectly—yet remains unaccounted for in his models. These include (1) the granting of medical waivers, (2) the rating of Blue and Gold Officer (BGO) interviews, and (3) RAB adjustments.

Medical Waivers

Applicants who fail USNA’s medical exam have the option to appeal for a waiver, but not all appeals are resolved. Because passing the medical exam or obtaining a waiver is a prerequisite for admission, disparities in waiver approvals directly influence which applicants remain eligible. While USNA denies considering race/ethnicity when deciding whether to grant a waiver, the data reveal significant racial differences in waiver approval rates.⁷⁶ Black applicants receive medical waivers at nearly twice the rate of White applicants (21% vs. 11%), while White applicants are the most likely to have unresolved waivers.⁷⁷ These discrepancies persist even after excluding Blue Chip Athletes and prep program applicants, suggesting that the differential treatment favoring minority applicants extends beyond athletics and preparatory pathways.

BGO Interview Ratings

The Blue and Gold Officer (BGO) interview is a key qualitative component of an applicant’s evaluation, and Arcidiacono’s analysis uncovers some racial disparities in its scoring.⁷⁸ Black and Hispanic applicants initially receive slightly lower ratings than their White and Asian counterparts. However, once other applicant characteristics are accounted for, Black applicants receive a modest boost in their predicted BGO scores, while Asian applicants experience a small penalty.

Since Arcidiacono’s admissions models include BGO scores as a control, this suggests that his estimates may slightly understate racial preferences for Black applicants while slightly overstating them for Asians. However, given the relatively small differences in coefficients, the overall impact of this bias is likely minimal.

RAB Points

RAB adjustments to applicants’ WPM scores provide further evidence of racial preferences.⁷⁹Accounting for other applicant characteristics, Black, Hispanic, and Asian applicants receive significantly higher RAB points than White applicants. Since Arcidiacono’s preferred admissions models do not include RAB points as a control, this suggests that racial preferences in USNA admissions could be even larger than estimated.

Interestingly, these RAB models also show that first-generation college students and applicants from lower-income families receive positive RAB adjustments. However, neither group exhibits a significant positive association with admission in Arcidiacono’s models, suggesting that these adjustments have little to no practical impact on their likelihood of admission. This raises broader questions as to whether USNA has truly exhausted race-neutral alternatives in its pursuit of diversity, as required under strict scrutiny.

3. USNA’s Response

Professor Arcidiacono’s findings demonstrate that USNA has been far from forthcoming about the extent to which race is considered throughout its admissions process and the broad range of areas where these considerations take place. These revelations cast serious doubt on USNA’s repeated assurances that its use of race is limited, non-determinative, and narrowly tailored to achieve its stated goals.

How did USNA respond to these findings? Interestingly—and in stark contrast to how institutions like UNC and Harvard approached previous affirmative action cases—USNA’s initial response did not involve conducting its own exculpatory analysis of admissions outcomes. In fact, USNA neither provided any quantitative analysis nor directly addressed Arcidiacono’s findings. Instead, its opening salvo comprised a series of expert reports emphasizing the purported importance of racial diversity in enhancing military readiness, mission effectiveness, the military’s domestic and international legitimacy, and the recruitment and retention of personnel. As I will document in a later section, these reports were poorly argued and largely devoid of quantitative evidence to substantiate their sweeping claims about the benefits of diversity.

However, while USNA initially avoided directly addressing Arcidiacono’s findings, it eventually commissioned a report to counter his analysis. Authored by Dr. Stuart Gurrea, the managing director of the consultancy group Secretariat, this report marked USNA’s delayed attempt at rebuttal. Gurrea holds a PhD in economics but has never published a single academic journal article. Instead, his career has been spent almost entirely outside academia, primarily as a “litigation economist” or “hired gun”, frequently representing federal government interests in litigation cases.

In this case, Gurrea’s compensation appears to have exceeded half a million dollars in taxpayer funds, including not just his hourly rate of $750 for hundreds of hours of billed time, but also a share of the additional fees charged by his supporting team at Secretariat.⁸⁰ While expert witnesses on both sides of litigation are routinely compensated handsomely for their time, the use of public money—especially in defense of a contested government policy—demands a high standard of analytical rigor and impartiality. Gurrea’s work, however, fell well short. Unlike Arcidiacono, who produced three detailed reports supported by replicable empirical models, Gurrea offered no independent analysis of admissions patterns. Instead, as the following section will illustrate, his work primarily sought to muddy the waters—raising doubts about Arcidiacono’s findings without presenting evidence that meaningfully challenged them.

3.1 USNA’s Rebuttal of Arcidiacono’s Report

The central argument of Dr. Gurrea’s rebuttal report, reinforced in his trial testimony, is that the results of Professor Arcidiacono’s analyses are biased, unreliable, and therefore fail to demonstrate that USNA’s consideration of race is more substantial than it claims. Gurrea offers three central reasons to support this conclusion, each of which is examined below.

3.1.1 Omitted Variable Bias

Dr. Gurrea’s first critique is that Professor Arcidiacono’s models fail to account for critical race-neutral factors that USNA considers in its admissions decisions. He highlights qualitative attributes—such as those reflected in letters of recommendation, interview notes, and personal statements—that are not readily quantifiable and are therefore excluded from Arcidiacono’s analysis.⁸¹ For example, Gurreau references a hypothetical Black applicant whose personal statement discusses overcoming racism, arguing that such an experience “may be viewed as an indication of strength of character and contributes positively to a candidate’s selection.”⁸² USNA’s defense attorneys echoed this point, citing a similar hypothetical case of a Black applicant with “relatively weak grades but a strong essay about having to overcome racism growing up”.⁸³ According to Gurrea, the exclusion of these qualitative factors likely inflates Arcidiacono’s estimates of racial preferences.

In addition to omitting qualitative race-neutral factors, Gurrea contends that Arcidiacono’s models fail to adequately control for applicants’ socioeconomic status and whether they hail from Congressional districts underrepresented at USNA—an “observable relevant factor for admission.”⁸⁴ Regarding socioeconomic status, Gurrea objects to Arcidiacono’s use of a binary indicator for whether an applicant’s household income is above or below $80,000, arguing that such data is often misreported and that the predictive significance of this monetary threshold varies by location. Although Arcidiacono supplemented this control with IRS-reported zip-code-level income data, Gurrea argues that these supplemental measures are “problematic,” do not “reliably control for applicant socioeconomic status,” and fail to “fully reflect how the Academy assesses socioeconomic status.”⁸⁵

In light of these limitations, Gurrea concludes that Arcidiacono’s results are incomplete and potentially misleading, as they fail to account for important qualitative and quantitative factors that, he argues, could explain the observed disparities without attributing them to racial preferences.

In truth, Gurreau’s arguments are designed to mislead. As Arcidiacono rightly points out, it is telling that Gurreau failed to run his own admission models to test the plausibility of his claims.⁸⁶ Such an exercise does not require perfectly accounting for every qualitative or quantitative variable that might influence admissions outcomes. When it comes to the qualitative factors Gurreau emphasizes—like personal statements or letters of recommendation—it suffices to control for observed quantitative variables that logically correlate with these unmeasured constructs and thus serve as reasonable proxies.

Critically, these proxies need not perfectly capture the unobserved factors they represent. What matters is how the race coefficients change—especially in terms of direction—when these proxies are included in the model. If controlling for these proxies leaves the racial disparities largely intact, it strongly suggests that the unmeasured qualitative factors Gurreau references are unlikely to explain away the observed disparities.

The same principle applies to the quantitative factors Gurreau claims were insufficiently controlled. Even if we accept Gurreau’s critique of the socioeconomic indicators in Arcidiacono’s models, these indicators (i.e., household income, first-generation college, zip-code level free or reduced-price school lunch eligibility rates, zip-code level income) are highly correlated with socioeconomic status. Similarly, community-level socioeconomic indicators and the slate-related characteristics Arcidiacono controls are logically correlated with what USNA considers ‘underrepresented’ districts. Including these variables provides a meaningful indication of the direction—and extent—to which racial differences might shift if the underlying constructs were perfectly captured. And, as Figure 2.3A earlier demonstrated, the inclusion of such controls, far from diminishing the observed racial disparities, actually resulted in even larger differences in admissions outcomes.

Gurreau’s failure to engage in this straightforward empirical exercise reveals the speculative nature of his critique. By neglecting to test whether his proposed factors could plausibly alter Arcidiacono’s findings, Gurreau essentially asks us to accept his claims on faith rather than evidence.

Fortunately, Arcidiacono addresses this gap by comparing the results from his preferred admissions model to those from alternative models that incorporate some of the variables Gurrea claims inflate the observed disparities when excluded. This comparison, shown in Figure 3.1A, plots the estimated logit coefficients from both models. The results perhaps explain why Gurrea avoided empirically testing his own assertions: incorporating these additional variables actually amplifies the coefficients for all racial minority groups. Specifically, controlling for indicators of unusual life experiences, hardship, or adversity increases the Black coefficient from 2.958 to 3.152, the Hispanic coefficient from 1.195 to 1.238, and the Asian coefficient from 1.449 to 1.520. Adding RAB adjustments further elevates the Black coefficient to 3.307, while slightly reducing the coefficients for Hispanics (1.213) and Asians (1.297).

Figure 3.1A. Logit estimates of USNA Admissions: Arcidiacono’s Preferred Model vs. Gurrea’s Suggested Model

Note. Bars represent logit coefficients, with error bars denoting 95% confidence intervals. Model sample sizes are reported in parentheses. Gurrea’s preferred analytical sample excludes service-connected nominees, arguing that their applications are evaluated under different criteria.⁸⁷ This exclusion not only reduces his model’s sample size compared to Arcidiacono’s but also slightly inflates the estimated coefficients for non-White racial groups.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Supplemental Expert Report of Peter Arcidiacono, Plaintiff's Exhibit 518, Table 3.1F, at 11.

All told, these patterns of results make Gurrea’s claim that unobserved qualitative factors are driving Arcidiacono’s estimates of racial preferences highly implausible. There is no evidence to suggest that properly accounting for these factors would meaningfully diminish the estimates. If this were the case, we would expect at least some moderation in the magnitude of the coefficients when adjusting for proxies of these constructs. Instead, the observed pattern shows the opposite—the coefficients grow larger. These findings also lend additional support to Arcidiacono’s assessment that his estimates of racial preferences are conservative rather than overstated. While it’s not impossible that other unobserved race-neutral variables drive the differences, they would all (emphasis) have to be a) either uncorrelated or strongly negatively correlated with the model observables, and b) strongly positively correlated with both race and admission. But given the diverse array of observables in the model, and given how much variation in admissions is already explained by Arcidiacono’s models, this prospect is, again, highly implausible.

3.1.2 Coding of Race/Ethnicity

Dr. Gurrea’s second critique is that Professor Arcidiacono’s categorization of race and ethnicity is “ad hoc and unjustified.”⁸⁸ Specifically, he objects to Arcidiacono’s integration of USNA’s separate race and ethnicity variables and his use of a hierarchical coding scheme to assign applicants to a single racial/ethnic category.

Arcidiacono’s approach follows a structured “one-drop rule” framework, classifying applicants based on a cascading hierarchy.⁸⁹ Applicants are categorized as “Black” if they identify as any part Black or African American, regardless of ethnicity. If not categorized as Black but identifying as Hispanic or Latino, they are classified as Hispanic, regardless of any additional racial identifications. Those who are neither Black nor Hispanic but report being any part “American Indian/Alaskan Native” or “Native Hawaiian or Other Pacific Islander” are coded as “Native American/Hawaiian.” Applicants who identify as Asian and do not fall into any of the prior categories are categorized as “Asian.” Finally, applicants are categorized as “White” only if they report “White” as their sole racial identity and do not identify as Hispanic or Latino.

Gurrea raises two main objections to this approach.⁹⁰ First, he argues that USNA considers race and ethnicity as separate variables in its admissions process, and therefore Arcidiacono’s integration of the two fails to reflect how USNA evaluates applicants. He claims that Arcidiacono’s framework erases the independent influence of ethnicity and arbitrarily prioritizes certain racial identities over others. For example, Hispanic applicants may identify with multiple races, yet Arcidiacono exclusively classifies them as Hispanic unless they also identify as Black. According to Gurrea, this “waterfall” approach inconsistently treats multiracial applicants based on their specific racial identifications, rather than allowing for separate effects of race and ethnicity.⁹¹

Second, Gurrea contends that Arcidiacono’s hierarchical coding scheme lacks justification.⁹² He argues that the decision to assign applicants identifying as any part Black exclusively to the Black category—while treating Hispanic as a catch-all for non-Black Hispanics—unnecessarily collapses multiracial identities. This, he claims, introduces inconsistencies, particularly since most Black applicants identify as a single race while many Hispanic applicants identify as White.

Gurrea proposes an alternative “data-driven” categorization scheme, in which race and ethnicity are treated as separate variables.⁹³ Under this framework, Hispanic applicants are classified as an ethnic group regardless of racial identity, while multiracial applicants who do not clearly fit into one of the major racial categories are placed in an “Other or Multiracial” category to preserve additional information. He argues that this approach better reflects how USNA processes applicant data and avoids the perceived arbitrariness of Arcidiacono’s framework.

In response, Professor Arcidiacono acknowledges that while Gurrea’s categorization scheme is “unobjectionable in the abstract,” it does not align with how college admissions typically operate at any of the universities he has studied.⁹⁴ More critically, Gurrea’s characterization of his framework as “data-driven” is difficult to square with the fact that it produces a less accurate model of admissions outcomes for minority applicants. This suggests that USNA employs a hierarchical approach similar to the one Arcidiacono has observed at other institutions.

This is evident in Figure 3.1B, which compares the accuracy of predicted admission rates for different race/ethnicity combinations using Arcidiacono’s versus Gurrea’s coding schemes. For example, Arcidiacono’s treatment of multiracial Black applicants as ‘Black’ yields far more accurate predictions of this group’s admit rate than Gurrea’s approach. Arcidiacono’s estimate (43.39%) is only 4 points removed from the actual multiracial Black admission rate (47.48%), whereas Gurrea’s estimate (29.95%) is off by more than 17 points. When using Gurrea’s preferred sample—excluding service-connected nominees—the error in Arcidiacono’s prediction narrows to just 3.3 points, while the error in Gurrea’s prediction expands to 18.1 points. Similarly, because Gurrea underestimates admit rates for multiracial Black applicants, he overestimates them for multiracial non-Black applicants, with a prediction error of 5.1 points compared to just 2.9 points for Arcidiacono’s estimate. Similar trends emerge for White and non-White Hispanic applicants, where Gurrea’s coding scheme consistently produces larger prediction errors than Arcidiacono’s.

Figure 3.1B. Actual and Predicted Admit Rates by Race/Ethnicity Combinations

Note. Green bars represent observed admission rates. Blue and gray bars depict predicted admission rates using Arcidiacono’s and Gurrea’s racial/ethnic categorization schemes, respectively. Predictions are derived from Arcidiacono’s preferred model, which alternates between using his own categorization scheme and Gurrea’s for demonstration purposes. Gurrea’s sample excludes service-connected nominees.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Supplemental Expert Report of Peter Arcidiacono, Plaintiff's Exhibit 518, Table 3.2F, at 16.

Interestingly, Gurrea concedes that his categorization scheme results in a poorer fit to the data compared to Arcidiacono’s. However, he attempts to downplay this by claiming that both models fit the data “extremely closely.”⁹⁵ As Arcidiacono points out, Gurrea can only make this claim by focusing on overall dataset fit statistics, which are dominated by the large sample of non-Hispanic White applicants.⁹⁶ Because racial preferences primarily affect minority admissions, aggregate fit statistics obscure significant disparities in predictive accuracy for minority groups. In other words, the “closeness” of the fit statistics masks meaningful differences where it matters most—among applicants subject to racial preferences.

Despite the weaker predictive fit of Gurrea’s scheme, he argues that Arcidiacono’s estimates “are not robust” under its adoption, claiming that the coefficient on Black “drops significantly”.⁹⁷

Figure 3.1C. Comparing the Impact of Race/Ethnicity Categorization on Estimated Marginal Effects in Admissions

Note. The top row presents predicted admit rates and average marginal effects (AMEs) using Arcidiacono’s sample, with the left panel applying his race categorization scheme and the right panel applying Gurrea’s. The bottom row repeats this structure using Gurrea’s sample. Because Gurrea treats race and ethnicity as separate variables, the number and identity of racial/ethnic groups differ across panels. All predictions are derived from Arcidiacono’s preferred admissions model (Model 6), with only the race/ethnicity categorization scheme varied for comparison.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Supplemental Expert Report of Peter Arcidiacono, Plaintiff's Exhibit 518, Table 3.3F, at 17.

Notably, Gurrea’s rebuttal report omitted baseline admit rates for different racial groups—a key omission. As shown in the bottom-right panel of Figure 3.1C, even under Gurrea’s categorization scheme and preferred sample, Black applicants are predicted to be admitted at a rate more than 2.5 times higher than if they were treated as White. According to Arcidiacono, this actually understates the true effect, as Gurrea removes race and ethnicity coefficients separately rather than simultaneously. When both are turned off together, the predicted Black admit rate rises to 2.7 times higher, with the average marginal effect increasing from +18.9 to +19.6 percentage points.

This is hardly a major departure from Arcidiacono’s original estimates, which predict a Black admit rate 2.96 times higher under race-conscious admissions, corresponding to an average marginal effect of +26.6 percentage points. Furthermore, the estimated effects for Asian applicants remain nearly identical across categorization schemes—a fact Gurrea fails to acknowledge.

Finally, Gurrea attempts to downplay the role of race and ethnicity in admissions by arguing that removing these variables from the model reduces overall fit by “less than three percentage points.”⁹⁸ However, this framing is misleading, as overall fit statistics are heavily driven by non-Hispanic White applicants, whose outcomes dominate the dataset and are less affected by racial preferences. In reality, for minority applicants, removing race and ethnicity leads to significantly larger reductions in predictive accuracy, underscoring their decisive role in admissions decisions.

3.1.3 The Inappropriateness of Logit Models in the Context of USNA Admissions

A third strategy Gurrea employs to discredit Arcidiacono’s findings is to argue that logistic regression models are fundamentally inappropriate for analyzing USNA admissions. Specifically, he claims that USNA’s rolling admissions, fixed class sizes, and preference for variety in the incoming class create interdependencies among admissions decisions, violating the assumption of independence among observations—a key requirement for logit models.⁹⁹

To support this claim, Gurrea cites economist Kenneth Train’s warnings about using logit models in scenarios where unobserved factors are correlated over time. He argues that because USNA admits students throughout the cycle, a decision made for one applicant could affect decisions for others later on. For instance, if a candidate from an underrepresented district is admitted early, it might reduce the chances of another applicant from that district being admitted later.¹⁰⁰ Likewise, he contends that because class sizes are fixed, admitting one candidate necessarily limits the available slots for others, further undermining the independence of admissions decisions.

However, as Arcidiacono points out, Gurrea’s critique misapplies Train’s concerns.¹⁰¹ Train’s warnings primarily address cases where the same individual or decision-maker makes repeated choices—such as a consumer purchasing the same product over time—where unobserved preferences persist across decisions. But at USNA, admissions decisions are made independently for each applicant, based on individual qualifications. The unobserved factors in Arcidiacono’s models—such as high school grades, extracurricular achievements, and leadership potential—are unique to each candidate and not systematically dependent on decisions made for others.

Moreover, the structure of USNA’s admissions process does not support Gurrea’s claim of interdependence. While some admissions decisions are deferred until later in the cycle, this does not mean that early decisions drive later ones in a way that fundamentally alters outcomes. Arcidiacono demonstrates this by analyzing the timing of admissions decisions. Gurrea cites a histogram showing a spike in review dates toward the end of the cycle, claiming this proves that admissions are interdependent.¹⁰² But, as Arcidiacono notes, this pattern is common at competitive universities, where marginal candidates are often held for final review.¹⁰³ More importantly, Arcidiacono shows that while rejection decisions do cluster toward the end of the cycle, admissions decisions are distributed more evenly throughout. This undermines Gurrea’s assertion that earlier acceptances systematically constrain later ones.

Arcidiacono also addresses variability across admissions cycles by including class-year fixed effects in his models. These controls ensure that differences in applicant pools and admissions standards across years are accounted for, rendering Gurrea’s claims about interdependence irrelevant.

Notably, Gurrea’s argument is inconsistent with how similar admissions processes have been analyzed in other affirmative action cases. Institutions like Harvard and UNC also use rolling admissions, maintain fixed class sizes, and seek “variety” in their incoming classes. Yet in the Harvard case, David Card—a Nobel Prize-winning economist hired to rebut Arcidiacono’s findings—did not argue that logit models were inappropriate for admissions analysis. In fact, Card himself relied on logit models to study Harvard’s admissions. Despite claiming to have read Card’s report, Gurrea admitted under cross-examination that he could not recall whether Card had criticized the use of logistic regression—a striking lapse, given that Card used the very framework Gurrea seeks to discredit.¹⁰⁴

Perhaps the most troubling implication of Gurrea’s argument is that it suggests USNA’s admissions process cannot be modeled at all. If accepted, this claim would render empirical scrutiny of USNA’s use of race impossible, effectively shielding the institution from accountability. Public institutions like USNA are subject to transparency standards, and dismissing statistical analysis as inherently inadequate undermines those standards. Gurrea’s position amounts to a request that we simply take USNA at its word—a deeply unsatisfactory stance given the stakes involved.

Ultimately, Gurrea’s arguments not only fail to undermine Arcidiacono’s findings but also highlight the broader credibility gap between the two experts. While Arcidiacono, as an academic, is held to the scrutiny of his peers and the risk of professional reputational damage, Gurrea operates in the world of litigation consulting, where client satisfaction—rather than intellectual rigor—is the primary incentive. Given that USNA had the resources to hire a more renowned expert, its decision to rely on Gurrea—who lacks peer-reviewed academic publications and, more importantly, whose arguments lack empirical support—raises further questions about the strength of its defense. In previous affirmative action cases, institutions like Harvard secured the testimony of world-class economists like David Card to bolster their arguments. The government’s failure to find an equivalent expert here suggests either poor strategic judgment or difficulty in securing a credible scholar willing to defend USNA’s admissions policies.

3.2 The Diversity Imperative

If Gurrea’s testimony was weak, the ‘diversity experts’ USNA enlisted to defend its use of race in admissions were no stronger. These experts share two glaring shortcomings. First, they repeat the same sweeping, unqualified claims about the supposed benefits of racial and ethnic diversity in the military, often portraying diversity as indispensable to military readiness, effectiveness, and cohesion. Second, the literature they cite in support of these claims—frequently drawing from overlapping sources—fails to provide direct or empirical evidence linking racial diversity (let alone racial diversity artificially created through racial preferences) to the outcomes they champion. Much of it is tangential at best, addressing non-racial forms of diversity or contexts with little relevance to the unique realities of military service.

And yet, this testimony is pivotal to USNA’s defense. Even if USNA’s consideration of race were genuinely “limited”—a notion belied by Arcidiacono’s findings—strict scrutiny requires more. To pass constitutional muster, USNA must demonstrate that its consideration of race is narrowly tailored and the only feasible means to further compelling national security imperatives. These expert reports, therefore, are central to USNA’s attempt to frame its admissions policies as indispensable to achieving military readiness and effectiveness.

A prime example of this reasoning is the expert report and testimony of David Lyall, a political scientist and director of the Political Violence FieldLab at Dartmouth College. His testimony, alongside that of other USNA witnesses, forms the foundation of the Academy’s central arguments: (1) that racial diversity enhances military effectiveness, (2) that racial inequality presents risks to cohesion and battlefield performance, and (3) that racial diversity is critical for recruitment, retention, and maintaining the military’s domestic and international legitimacy. These claims—frequently repeated across multiple expert reports—are examined in detail below.

3.2.1 Claim #1: Racial Diversity Enhances Military Effectiveness

In support of this first claim, Lyall contends that research demonstrates that diversity confers several advantages (a “diversity bonus”) on modern battlefields.¹⁰⁵ First, he notes that diverse teams outperform homogeneous ones in problem-solving and decision-making, particularly for complex tasks. Second, he contends that cognitive heterogeneity—while not reducible to race or ethnicity alone—fosters innovation by introducing a broader range of ideas, debates, and communication styles. Lastly, he asserts that diverse teams may be more resilient and adaptive in uncertain or challenging environments.

The Diversity Claim-Evidence Gap

The first thing that jumps out in Lyall’s testimony is how he subtly shifts from discussing ‘racial and ethnic diversity’—the specific focus of his claim about combat performance—to making broader, more abstract arguments about ‘diversity’ in general. This rhetorical maneuver is not incidental—it allows him to cite studies that support diversity in an abstract sense while avoiding the inconvenient fact that no direct evidence links racial diversity to improved military performance.

Indeed, while each of Lyall’s three points is accompanied by citations—15 in total—none provide empirical evidence directly linking racial or ethnic diversity to improved military performance. In fact, most do not even address racial/ethnic diversity, let alone in the context of the U.S. (or any) military. Some, like the literature review The Science of Teams in the Military: Contributions From Over 60 Years of Research, do not mention diversity at all.¹⁰⁶ Of the five references containing empirical analysis, none examine the effects of racial/ethnic diversity in military contexts—or any context.¹⁰⁷ For instance, a 2004 study simulated fictitious agents to assess whether teams with diverse skills outperform more able but homogeneously skilled groups.¹⁰⁸ Another, using Amazon Mechanical Turk respondents, found that groups solve complex problems better and faster than individuals working alone.¹⁰⁹

To be sure, certain forms of diversity may enhance performance in specific contexts. Yet none of Lyall’s references demonstrate that racial diversity—let alone racial diversity engineered through preferences—improves performance on any battlefield, whether on land, at sea, or in the air.

But this evidentiary void is hardly unique to Lyall’s testimony. The same lack of direct, empirical support is evident across the reports of all the experts USNA hired to validate its diversity rationale. For example, in attempting to substantiate many of the same claims as Lyall, Jeannette Haynie—a research scientist at CNA Corporation and former advisor to the Department of Defense (DoD)—cites a non-peer-reviewed Harvard Business Review article titled Why Diverse Teams Are Smarter.¹¹⁰ Yet the studies referenced in this article either fail to examine the relationship between racial/ethnic diversity and team performance, have no connection to the military context, or—in the case of the widely criticized 2015 McKinsey study—have since been discredited.¹¹¹

If that doesn’t sufficiently convey the lack of empirical support in Haynie’s report, consider her reliance on small-sample qualitative studies, including a 2015 Master’s Thesis with just five participants and five citations.¹¹² Needless to say, such sources fall far short of the rigorous evidence needed to justify the bold claims about the indispensable role of racial diversity in military readiness and effectiveness.

And yet, such sources are perhaps the best Haynie, Lyall, and USNA’s other ‘diversity experts’ have to work with. As a 2014 article from the U.S. Air Force’s flagship academic journal—which Haynie herself cited but conspicuously omitted this key point—observed:

“The military has no wide-range studies that examine whether diverse teams resolve complex problem sets better than nondiverse teams.”¹¹³

A decade later, nothing has changed.

According to a comprehensive 2024 literature review cited by one of SFFA’s attorneys during Haynie’s cross-examination:

“Despite recent focus by the Navy and other military branches on promoting healthy cultures to enable performance among an increasingly diverse workforce, there exists little research on the relationship between diversity and military performance.”¹¹⁴

The same assessment was given by Admiral Rick Cheeseman, the Chief of Naval Personnel, during a 2023 congressional hearing. When asked whether the Navy had any data demonstrating that diversity efforts improved combat effectiveness, his response was unequivocal:

“We do not have any specific data right now that talks about any diversity, equity, and inclusion efforts and how it relates to combat effectiveness.”¹¹⁵

Even Haynie herself had to admit under cross-examination that, to her knowledge, the Department of Defense “has never done any kind of quantitative analysis, large or small, that examines whether diverse teams resolve complex problems better than nondiverse teams.”¹¹⁶

This absence of hard data was further highlighted during cross-examination when Haynie was asked how she could conclude that diversity and inclusion were vital to military team performance given that the DoD has never conducted a wide-ranging study on the subject. Instead of providing a direct answer, she pivoted to discussing her “roots in the Marine Corps,” the need to “build a comprehensive picture,” and the importance of “continuing to learn and grow.”¹¹⁷

When pressed to cite concrete studies demonstrating a performance advantage for diverse teams in military contexts, she could only recall a study that was “currently underway” and vaguely referenced reports from RAND and NATO without elaboration.¹¹⁸ Likewise, when asked to explain why she states that DoD has consistently viewed diversity as critical to mission success, she simply listed historical steps the military has taken to promote diversity—without ever explaining on what empirical basis the DoD reached this conclusion in the first place.¹¹⁹

This evasiveness was emblematic of USNA’s entire case—asserting benefits of racial diversity while failing to produce evidence that these benefits exist in real-world military operations.

When you reflect on it, the dearth of quantitative analysis on these questions is truly shocking, calling into question the integrity of the military’s judgment that “a racially and ethnically diverse officer corps is a national security imperative.”¹²⁰ The Department of Defense has had decades to conduct such research. If it genuinely prioritized building a body of empirical evidence to support its diversity goals, it could have utilized this time to field natural and/or controlled experiments, leveraging random variation in the racial composition of military units to examine the effects on objective measures of performance.

Instead, the DoD appears to have taken these claims as self-evident truths, relying on untested assumptions and cherry-picked anecdotes rather than robust, replicable research. This lack of evidence is not just an oversight—it represents a failure to apply the same rigor to its diversity rationale as it presumably applies to other critical military objectives. For an institution tasked with defending the nation, this casual approach to such a consequential issue raises serious concerns about its commitment to evidence-based decision-making.

In fairness, both Haynie and Lyall contend that ethical and logistical considerations make such experimental research difficult, if not impossible, to conduct.¹²¹ But that claim doesn’t hold up to scrutiny. Studying the effects of natural variation in unit racial composition—which, by the testimony of one witness, “is constantly in flux”—on relevant performance outcomes is neither unethical nor logistically prohibitive.¹²² In fact, one of the few exceptions to the general void of empirical analysis in this context—a 2024 Naval Postgraduate School study—did exactly that, examining the effects of racial diversity in destroyer ship crews on several performance metrics.¹²³ Remarkably, despite claiming to have conducted a thorough literature review, Haynie was unaware of this study—arguably the most, if only, relevant piece of research for the DoD and Navy’s diversity argument.¹²⁴ While this could have been an innocent oversight, the fact that the study found no significant relationship between crew racial diversity and performance metrics raises questions that cannot be ignored.

Even if evidence existed to suggest that racial diversity enhances performance in certain contexts, it is both unrealistic and intellectually dishonest to assume that such effects are universally positive. Social science research has long recognized the potential downsides of diversity—including increased conflict, communication barriers, and challenges to cohesion—with researchers characterizing its impact on team performance as “a double-edged sword”.¹²⁵

Racial diversity engineered through affirmative action introduces its own unique complications. For one, the perception—or reality—that individuals have been selected based on criteria other than merit can undermine trust and cohesion within the ranks. Officers and leaders presumed to have benefited from racial preferences may face unwarranted skepticism about their competence, eroding their authority and effectiveness. Conversely, those who feel excluded from such preferences may harbor resentment, viewing the system as unfair or biased against them. In the military, where trust, cohesion, and shared purpose are paramount, these dynamics could carry severe consequences.

A case study Lyall cites on post-Apartheid reforms to South Africa’s military provides a striking example of how affirmative action policies—intended to foster diversity—can have the opposite effect, fueling resentment and degrading force readiness.¹²⁶ While Lyall omits this evidence, it illustrates the unintended consequences of racial engineering, a theme explored further in the next section.

In truth, it’s not even clear that the DoD or Navy genuinely cares whether its diversity claims and goals are grounded in empirical analysis. This apparent indifference was starkly illustrated during Professor Arcidiacono’s pre-trial deposition, when the leading USNA attorney asked him:

“So it sounds like you do believe that the military requires a statistical analysis in order to reach a judgment about the value of diversity in the officer corps?”¹²⁷

When “Diversity” Means Race

Notably, USNA’s own admissions policies are difficult to reconcile with its stated diversity rationale. The Academy asserts that it values a wide range of diversity factors—including gender, socioeconomic background, geographic variety, first-generation status, and demonstrated interest in STEM disciplines—yet its own admissions records reveal that racial diversity receives by far the most institutional attention.¹²⁸ The dean’s brief, which is updated throughout the admissions cycle and distributed to USNA’s top admissions officials, meticulously tracks racial categories but omits any comparable tracking of socioeconomic status, rural/urban backgrounds, first-generation college applicants, or candidates who have faced hardship or adversity.¹²⁹ Similarly, USNA generates specialized reports identifying qualified minority applicants who have not yet received offers, yet produces no equivalent reports for applicants from disadvantaged socioeconomic backgrounds.¹³⁰ If all forms of diversity matter equally, why does USNA’s internal decision-making prioritize race above all else?

This inconsistency exposes a deeper flaw in USNA’s logic: if certain qualities—such as resilience, cognitive diversity, and adaptability—enhance military effectiveness, then it would be far more efficient to measure and select for those qualities directly rather than using race as a crude proxy. By prioritizing racial diversity while giving significantly less weight to other diversity factors, USNA implicitly acknowledges that its policies are not designed to maximize battlefield effectiveness but rather to achieve a predetermined racial composition in its incoming classes. This raises an unavoidable question: is the military’s push for racial diversity truly about enhancing operational performance, or is it about advancing a broader ideological commitment to racial representation for its own sake?

As the next section explores, USNA’s diversity rationale is built on more than just unproven claims about effectiveness—it also relies on untested assumptions about cohesion, morale, and the risk of racial unrest.

3.2.2 Claim #2: Racial Inequality Presents Risks to Cohesion and Battlefield Performance

Building on his previous claim, Lyall argues that racial inequality within the military threatens cohesion, unit effectiveness, and overall battlefield performance.¹³¹ This argument relies primarily on his analysis of the Project Mars dataset, which he compiled over seven years with the assistance of 134 undergraduate and graduate students. The dataset encompasses 825 ground armies across 250 conventional wars between 1800 and 2011 and serves as the basis for Lyall’s Military Inequality Coefficient (MIC)—a measure designed to quantify racial and ethnic inequality within armies on the eve of war. The MIC ranges from 0 (perfect equality) to 1 (perfect inequality), accounting for both the size and composition of ethnic groups within the military as well as how these groups are treated by the state—whether through formal policies or informal practices.

Lyall’s analysis finds that militaries with higher MIC values—indicating greater levels of marginalization and inequality—experience significantly worse combat outcomes, including higher casualty rates, increased desertions, and diminished overall effectiveness.¹³² He attributes these patterns to the corrosive effects of inequality on trust, cohesion, and unit morale, which he claims ultimately undermine battlefield performance.

Breaking: Mistreated Soldiers Perform Poorly in Combat

Whatever the merits of Lyall’s research, its relevance to the modern U.S. military—and, by extension, the USNA case—is tenuous at best. For the central thrust of Lyall’s findings is merely that subjugated, enslaved, or otherwise oppressed peoples tend not to make loyal or effective soldiers. While this conclusion may be grounded in rigorous analysis—albeit one that is purely cross-sectional, limited to land wars, and omits critical control variables such as technological superiority—it is neither groundbreaking nor particularly relevant to a case concerning racial diversity achieved through affirmative action in the modern U.S. military.

Indeed, Lyall’s Project Mars dataset is severely limited in what it can reveal about the modern U.S. military. As Arcidiacono notes, it includes only 21 U.S. military observations, a mere six of which are from after the Vietnam War, and just 15 from the post-2000 era.¹³³ Moreover, since 1900, all U.S. observations fall within the lowest band of the MIC index, meaning there is no meaningful variation from which to infer anything about modern racial dynamics in the armed forces.

Still, Lyall could have tested whether racial diversity adds value once inequality is held constant—for example, by examining whether highly diverse but inclusive armies outperform less diverse yet equally inclusive ones. Tellingly, he does not. In fact, he acknowledges that the number of ethnic groups in an army—his proxy for diversity—has a small negative effect on performance, which he attributes to the “transaction costs” of diversity without inclusion: linguistic barriers, coordination difficulties, and interpersonal friction.¹³⁴ Fair enough. But if the benefits of diversity are conditional on inclusion—defined here as the absence of mistreatment or oppression of minority ethnic groups—then the relevant question is not whether inclusion improves performance (which Lyall’s findings already support), but whether diversity itself contributes anything once inclusion is present. That is the issue most relevant to the rationale behind race-conscious admissions at USNA—and it remains unaddressed in Lyall’s analysis, either because his data are not equipped to examine it or because they provide no evidence to support it.

Thus, the central takeaway from Lyall’s findings is simply that armies composed of different ethnic groups perform poorly when those groups are mistreated. But this pattern is not obviously unique to diverse armies. Conceivably, ethnically homogeneous forces whose troops are mistreated (e.g., North Korea) would be similarly susceptible to breakdowns in cohesion and effectiveness as diverse armies lacking inclusion. By conflating “diversity” with “treatment,” Lyall overstates the relevance of his findings to the case at hand—particularly the claim that increasing racial diversity in today’s U.S. officer corps enhances military effectiveness.¹³⁵

The ‘Hedge Against Racial Unrest’ Argument

To make his research relevant to the present case, Lyall must argue not simply that mistreatment harms military performance—a point his data already support—but that racial and ethnic minorities in today’s U.S. military continue to experience mistreatment or marginalization to a degree that compromises cohesion, morale, or readiness.¹³⁶

In making this case, Lyall first suggests that if the contemporary U.S. Army were assessed using the Ethnic Power Relations (EPR) dataset—a framework that absurdly classifies all non-White Americans as “politically powerless”—its Military Inequality Coefficient (MIC) score would be even higher.¹³⁷ Next, he refers to the 2017 Workplace and Equal Opportunity Survey of Active Duty Members (WEOA), which found that non-White personnel were more likely than their White counterparts to report experiencing at least one instance of racial or ethnic harassment or discrimination.¹³⁸ Such experiences, he argues, are linked to declines in key indicators of military readiness, including lower satisfaction with military life, reduced willingness to remain in service, and decreased confidence in personal and unit preparedness for wartime missions.

Lyall, along with other USNA witnesses, including Haynie, further cites the Department of Defense’s Office of People Analytics (OPA) analysis of the WEOA data, which links service members’ perceptions of an unhealthy unit “diversity and inclusion” (D&I) climate to negative outcomes—lower satisfaction with military life, weaker morale, and increased attrition.¹³⁹ The defense even referenced these findings in both its opening and closing arguments, underscoring their centrality to USNA’s case.¹⁴⁰

Lastly, Lyall highlights racial disparities in the military justice system, citing data indicating that Black soldiers and sailors are twice as likely as their White counterparts to be investigated and referred to court-martial.¹⁴¹

From these findings, Lyall and other USNA witnesses—including Haynie—advance a broader implication: that increasing minority representation in officer ranks is necessary to mitigate racial bias and grievances, reduce internal tensions, and maintain a stable force. In essence, this is a modern adaptation of an older idea—that ensuring sufficient racial representation in leadership positions acts as a “hedge” against potential unrest, particularly among Black service members.¹⁴²

Yet this reasoning is deeply flawed—both empirically and conceptually. Even if racial discrimination and harassment persist in the U.S. military, USNA’s witnesses provide no evidence that these issues are widespread enough to meaningfully threaten readiness, cohesion, or retention. Nor do they demonstrate that racial preferences in officer admissions offer an effective remedy. The assumption that increasing minority representation will alleviate racial grievances lacks empirical support and, in some cases, may exacerbate tensions rather than resolve them.

First, the WEOA data Lyall and other USNA witnesses cite suffer from a fundamental limitation: they are cross-sectional, not causal--meaning they capture associations at a single point in time and cannot establish causality. While the USNA witnesses assume that experiencing discrimination causes declines in readiness indicators, it is equally plausible that individuals who are already dissatisfied with military life or struggling in their roles are more likely to perceive or report discriminatory treatment.

Moreover, perceptions of discrimination are highly subjective and susceptible to external influences, including media narratives and political discourse. Indeed, my own research has shown that rates of self-reported discrimination among Black Americans increased during the Black Lives Matter movement and the Trump presidency—suggesting that social and political context can shape whether individuals perceive and report discrimination, regardless of whether actual treatment has changed.¹⁴³

This pattern of subjective interpretation is also evident in the military. During trial testimony, SFFA witness Brigadier General (Ret.) Christopher S. Walker, a Black 40-year veteran of the U.S. Air Force and former commander of the West Virginia Air National Guard with combat experience in the Balkans, Afghanistan, and Iraq, stated that he had never personally witnessed racial prejudice during his service.¹⁴⁴ When confronted by a USNA attorney with survey results indicating that some service members reported discrimination, Walker responded that, in many cases, these claims did not withstand closer examination. He described personally questioning airmen who alleged racial bias and discovering that their complaints often stemmed not from actual discrimination but from frustration with difficult or unsympathetic commanders. As he put it, “When I really, really dig deep and ask them questions, it turns out that, no, it wasn’t really racial discrimination; it’s just that their commander was a butthole.”

Walker’s testimony highlights a key issue with relying on subjective survey data to diagnose systemic bias. Just as perceptions of discrimination can be shaped by broader political discourse, they can also be influenced by personal grievances, leadership styles, and misattributions of intent—making them an unreliable basis for sweeping policy conclusions.

More critically, USNA’s witnesses fail to address a crucial question: do racial minorities actually exhibit lower morale, retention, and confidence in readiness—the very outcomes they claim are undermined by discrimination? If their argument were valid, we would expect to see substantial racial disparities in these indicators. Yet no such disparities are reported.

A closer examination of the WEOA data, which are visualized in Figure 3.2A below, makes clear why. In reality, there are no meaningful racial differences in the shares of active-duty military personnel reporting high personal and unit morale, satisfaction with military life, a willingness to stay on active duty, or confidence that they and their units are prepared for war. If anything, the data cut in the opposite direction: Black personnel actually report higher personal morale (46% vs. 43%), greater satisfaction with military life (63% vs. 61%), and greater willingness to remain on active duty (68% vs. 60%) than their White counterparts. Similarly, while Asians (23.3%) and Hispanics (21%) were also more likely than Whites (12.7%) to report an experience of racial discrimination or harassment, their overall morale and retention responses were indistinguishable from those of Whites. If such experiences had the strong negative effects that USNA’s witnesses suggest, this pattern would be difficult to explain.

Figure 3.2A. 2017 Workplace and Equal Opportunity Survey: Responses to Select Items Among Active-Duty Military by Race

Note. Sample sizes for each racial/ethnic group are reported in parentheses. Error bars represent 95% confidence intervals. Response estimates for each item are derived from tables in the 2017 Workplace and Equal Opportunity Survey of Active-Duty Members: Tabulation of Responses document. Items include: (1) Current level of morale (Q21a, p. 128), (2) Current level of unit morale (Q21b, p. 130), (3) Satisfaction with military way of life (Q7, p. 30), (4) Personal wartime mission preparedness (Q20A, p. 124), (5) Perceived unit wartime mission preparedness (Q20B, p. 126), and (6) Active-duty retention intentions (Q6, p. 28).

USNA witnesses’ reliance on OPA’s analysis of diversity and inclusion (D&I) climate perceptions fares no better. While the OPA study found that service members who perceived an unhealthy D&I climate were more likely to leave the military, it also found that minorities were more likely than Whites to remain. In fact, one of the leading factors driving higher White attrition was frustration with excessive emphasis on diversity initiatives.

What is more, among all respondents, more service members believed that there was too much attention paid to D&I (18%) than reported being in an “unhealthy” unit climate (12%). And among those who did label their unit’s climate as unhealthy, 22% attributed this at least in part to excessive attention paid to racial and ethnic harassment training.¹⁴⁵ In other words, for a significant share of service members, dissatisfaction with unit climate stemmed not from discrimination—but from the belief that the military’s attention to race is excessive.

Differential Standards Produce Differential Behavior

Finally, Lyall’s insinuation that racial disparities in military justice—namely, that Black service members are more likely than Whites to be investigated and referred to court-martial—reflect systemic bias ignores a more plausible and empirically supported explanation: differences in conduct. While Lyall’s claim concerns enlisted personnel, internal USNA data reveal conduct disparities even among midshipmen—an elite, highly selected group. These data come from a study of performance outcome disparities among midshipmen from the USNA classes of 2014–2020, conducted by a team of USNA faculty members, diversity officials, and external consultants, including former and current DEI administrators at the academy.

Finally, Lyall’s insinuation that racial disparities in military justice—namely, that Black service members are more likely than Whites to be investigated and referred to court-martial—reflect systemic bias ignores a more plausible and empirically supported explanation: differences in conduct. While Lyall’s claim concerns enlisted personnel, internal USNA data reveal conduct disparities even among midshipmen—an elite, highly selected group. These data come from a study of performance outcome disparities among midshipmen from the USNA classes of 2014–2020, conducted by a team of USNA faculty members, diversity officials, and external consultants, including former and current DEI administrators at the academy.

Among other findings, the study revealed striking racial differences in midshipmen commission of conduct (e.g., insubordination, unauthorized absences, alcohol/sex-related misconduct) and honor (e.g., cheating, plagiarism, property theft) offenses. As shown in Figure 3.2B, Black midshipmen were far more likely than their peers to commit both types of offenses. Fifty-five percent of non-varsity Black midshipmen had at least one conduct offense, compared to 31% of White, 30% of Hispanic, and 37% of Asian midshipmen. Among varsity athletes, the pattern was similar: 47% of Black varsity athletes committed conduct offenses, compared to 30% of White, 41% of Hispanic, and 25% of Asian varsity midshipmen.

While less common, significant racial disparities were also observed for honor offenses. More than 5% of Black non-varsity and 11% of Black varsity midshipmen committed one—rates that are roughly triple those of their White counterparts and higher than those of all other racial/ethnic minority groups.

Figure 3.2B. Conduct and Honor Offense Rates Among Midshipmen by Race and Varsity Status (Classes of 2014–2020)

Note. Blue bars represent the percentage of non-varsity midshipmen in each racial group who committed at least one conduct (left panel) or honor (right panel) offense. Green bars represent the percentage of varsity midshipmen in each racial group who committed at least one conduct (left panel) or honor (right panel) offense. Dashed horizontal blue and green lines indicate the average offense rates for non-varsity and varsity midshipmen, respectively. The sample consists of 5,384 White, 599 Black, 998 Hispanic, and 535 Asian midshipmen from the USNA classes of 2014–2020.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Plaintiff's Exhibit 148, Email from Steve Vahsen to David Forman re: “Any available time?” with attachment, Figures 22–23, at 29 (D. Md. Sept. 20, 2024).

Consistent with their higher conduct and honor offense rates, the data also show that Black midshipmen are more than three times as likely as White midshipmen to be separated from the academy for conduct-related reasons (5.6% vs. 1.7%). Given that conduct-related separations stem from serious disciplinary infractions—and that USNA prioritizes retaining Black midshipmen to meet diversity objectives—these disparities are more plausibly driven by genuine behavioral differences than by systemic bias in disciplinary enforcement.

Beyond disparities in offense commission and conduct-related separation rates, differences in conduct are also reflected in peer evaluations. Each semester, midshipmen rank their fellow company members and assign attributes from a standardized list of 89 positive and negative descriptors to the highest- and lowest-rated individuals. The authors analyzed four frequently used positive attributes (Logical, Organized, Thorough, Self-Disciplined) and four frequently used negative attributes (Inept, Lackadaisical, Sloppy, Unproductive) to identify racial disparities in peer assessments.

As shown in Figure 3.2C, Black midshipmen were consistently less likely than their peers to receive positive attributes and more likely to receive negative ones. Only 27.3% of Black midshipmen received at least one vote for "Logical," compared to 40.1% of White midshipmen; 30.2% were labeled "Organized" versus 46.1% of Whites; 25.3% were described as "Thorough" compared to 40.4% of Whites; and 36.2% were considered "Self-Disciplined," while 46.8% of Whites received that designation. Conversely, Black midshipmen were far more likely to be rated with negative attributes, such as "Inept" (76.1% vs. 55.5% of Whites), "Lackadaisical" (61.9% vs. 37.3%), "Sloppy" (54% vs. 37.9%), and "Unproductive" (53.2% vs. 26.1%).

Figure 3.2C. Racial Disparities in Peer Evaluations of Midshipmen (Classes of 2014–2020)

Note. Bars represent the percentage of midshipmen in each racial group who received at least one peer nomination for a given attribute during semesterly evaluations. The top panel displays the share of midshipmen receiving positive attributes (Logical, Organized, Thorough, Self-Disciplined), while the bottom panel displays the share receiving negative attributes (Inept, Lackadaisical, Sloppy, Unproductive). The sample consists of 5,384 White, 599 Black, 998 Hispanic, and 535 Asian midshipmen from the USNA classes of 2014–2020.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Plaintiff's Exhibit 148, Email from Steve Vahsen to David Forman re: “Any available time?” with attachment, Figures 13–14, at 18–19 (D. Md. Sept. 20, 2024).

While peer evaluations are inherently subjective and could potentially reflect bias, their strong alignment with objective indicators—such as documented conduct/honor offenses and conduct-related separations—suggests they capture real behavioral patterns rather than arbitrary prejudice.

If such disparities emerge even among the most rigorously screened officer candidates, it stands to reason that they may be even more pronounced in the broader enlisted force, where selection standards are lower, and a much wider range of academic and behavioral backgrounds are present. This pattern suggests that racial preferences in admissions and differential standards may contribute to disparities in performance and discipline rates—not systemic bias in military justice.

But perhaps most damning, the very report Lyall cites to support his claim explicitly tested the bias hypothesis—and found no evidence for it.¹⁴⁶ The report’s authors examined whether racial disparities in military discipline were larger or smaller for low-discretion offenses—cases in which officials had little leeway in whether or how to prosecute. If bias were a major factor, one would expect disparities to be smaller for these offenses. Instead, the disparities were actually larger, suggesting that Black service members were more likely to commit offenses in categories where discretion was limited. In short, the evidence Lyall himself relies on directly contradicts his implication that racial bias is the primary driver of these disparities.

Ultimately, the idea that racial preferences in service academy admissions will help prevent racial unrest within the military is not only unsupported but dangerously shortsighted. Engineering diversity through affirmative action can introduce its own tensions, particularly if meritocratic standards are compromised.

The racial unrest and violence that erupted in the Navy during the early 1970s offers a cautionary tale—but perhaps not for the reasons USNA’s witnesses would have us believe. While historians Beth Bailey and John Sherwood argue that these disturbances were primarily driven by a lack of Black representation in naval leadership, the evidence—including portions of their own testimony—suggests they may be overlooking or understating a critical factor: the rapid decline in the Navy’s recruitment standards following the end of the draft.

Prior to the early 1970s, the Navy practiced what Sherwood described as “qualitative recruitment,” selecting only candidates with relatively high scores on the Armed Forces Qualification Test (AFQT) from the draft pool.¹⁴⁷ As a result, the enlisted force was overwhelmingly White, as Black recruits disproportionately failed to meet the service’s higher entrance requirements. However, with the draft’s phaseout under the Nixon administration, the Navy faced a sudden manpower shortage and was forced to actively recruit enlistees rather than rely on selective induction. To do so, it dramatically lowered its standards, admitting a far higher proportion of recruits who placed in the bottom categories of the AFQT—many of whom lacked even a basic high school education or struggled with literacy.

The effects were immediate. As the proportion of low-quality recruits increased, so too did disciplinary problems, including acts of interracial violence. A 1973 congressional report by the Special Subcommittee on Disciplinary Problems in the U.S. Navy found that many of the sailors involved in racial unrest aboard ships like the Kitty Hawk were of below-average mental capacity and would likely not have been accepted into the Navy under prior recruitment standards.¹⁴⁸ “The Navy’s new recruitment guidelines,” the report concluded, “have created many of the problems the Navy is experiencing today.”

While the Subcommittee found no evidence that racial discrimination was a root cause of the violence, it acknowledged that the decline in recruitment standards may have indirectly contributed to perceptions of racial bias, particularly in job assignments. Many Black sailors, unable to qualify for specialized ratings due to lower test scores, were disproportionately placed in menial roles. Although this outcome reflected aptitude and skill requirements rather than racial exclusion, it was nevertheless perceived by many as evidence of systemic discrimination. This perception—accurate or not—fueled grievances that, in turn, exacerbated tensions within the ranks.

None of this is to say that racial discrimination and Black underrepresentation in the Navy’s officer corps played no role in the unrest of the 1970s. Rather, the point is that it is hardly obvious—and USNA’s witnesses fail to demonstrate—that these factors were more causally important than the lowering of the Navy’s recruitment standards.

Instead of acknowledging the clear connection between relaxed recruitment standards and increased disciplinary problems, the military and political leadership of the time—and now USNA’s witnesses—reframed the issue entirely as a problem of racial discrimination and a lack of Black leadership. This assumption—that increasing minority representation in leadership will mitigate racial grievances, improve cohesion, or enhance military effectiveness—rests on untested assumptions rather than historical or contemporary evidence.

If the Navy’s past mistakes teach us anything, it is that unit cohesion and military effectiveness depend far more on maintaining high standards than on achieving demographic proportionality.

3.2.3 Claim #3: Racial diversity is critical for recruitment, retention, and maintaining the military’s domestic and international legitimacy

A final argument advanced by USNA’s witnesses, including Lyall and Haynie, is that racial diversity is essential for recruitment, retention, and maintaining the military’s legitimacy—both domestically and internationally. They claim that increasing racial diversity in officer ranks enhances minority enlistment and retention rates, strengthens trust between the military and civilian populations, and improves U.S. military legitimacy in foreign operations, particularly in peacekeeping and counterinsurgency efforts.

At first glance, this argument might seem reasonable. If minority communities see more officers who “look like them,” they might be more inclined to enlist or remain in service. Likewise, in international missions, forces that appear racially diverse might be perceived as more neutral by local populations, enhancing cooperation. Yet, as with the previous claims, a closer examination reveals flawed reasoning, unsupported assumptions, and a striking lack of empirical evidence.

Claim #3A. Racial Diversity Improves Retention

USNA’s witnesses contend that racial representation in leadership roles improves retention among minority personnel. The primary quantitative evidence presented to support this claim consists of two sources: (1) a 2020 Naval Postgraduate School Master’s thesis, which Lyall cites in his expert report and was asked about during his trial testimony, and (2) time-series data on officer retention, introduced by the defense during the trial, showing increases in Black and Hispanic officer retention over the past two decades.

2020 Naval Postgraduate School Master’s thesis

According to Lyall, the Naval Postgraduate School study found that “an increase in coethnic peers, immediate supervisors, and senior leadership on medium ships such as destroyers, along with submarines, increases retention among Black first-term sailors and first-term Hispanic officers,” and that an “increase in same-minority senior leadership increases retention of first-term non-Hispanic and Black officers.”¹⁴⁹

However, even a cursory examination of the study reveals that its findings are weak, inconsistent, and, in some cases, directly contradict Lyall’s claims.¹⁵⁰First, while the study assumes that personnel assignments to ships are random due to the Navy’s centralized system, it does not actually test or validate this assumption. The authors fail to conduct balance tests or assess potential assignment biases, making it impossible to determine whether the observed differences in retention rates are caused by greater coethnic representation or by other unmeasured factors. Without such validation, the study cannot establish causality and remains susceptible to confounding variables.

More fundamentally, the study’s results, which are plotted below in Figure 3.2D, fail to provide meaningful support for Lyall’s argument. The supposed benefits of coethnic representation were small—with only one effect reaching significance at the conventional 95% confidence level, which is unimpressive given the model’s large sample sizes (e.g., N=23,944 Black sailors, N=22,527 Hispanic sailors). Moreover, the results were inconsistent and often at odds with Lyall’s underlying premise.

For instance, a 10-percentage point increase in the share of Black enlisted personnel stationed on medium-sized ships and submarines was associated with only a 2.1-percentage point increase (p=0.051) in the retention rate of first-term Black sailors—an effect so small that it borders on statistical insignificance. Moreover, this increase coincided with a 1.0-percentage point increase (p=0.055) in non-Black retention, contradicting the notion that coethnic representation provides a unique benefit to Black personnel.

Figure 3.2D. Impact of Racial Composition on First-Term Retention: NPS Study Coefficient Estimates

Note. Markers denote estimated changes (in percentage points) in first-term retention for Black versus non-Black groups (top row) and Hispanic versus non-Hispanic groups (bottom row). Results are presented separately for three samples: first-term officers on medium-sized vessels (top tick), first-term sailors on medium-sized vessels (middle tick), and first-term sailors on large-sized vessels (bottom tick). Estimates reflect the effect of a 10 pp increase in the respective group's share among first-term peers (column 1), mid-level supervisors (O3–O4 / E5–E6; column 2), and senior supervisors (O5–O6 / E7–E9; column 3). Error bars represent 95% confidence intervals.
Source: Hernandez Rodriguez, J. M., & Serna, C. (2020). The effects of diversity among peers and role models on U.S. Navy retention [Master's thesis, Naval Postgraduate School]. Naval Postgraduate School.

Similarly, a 10-point increase in Black share of senior enlisted leadership (E-7 to E-9) on medium-sized ships and submarines predicted just a 1.3-percentage point increase (p=0.029) in Black enlisted retention while simultaneously increasing non-Black retention by 0.6 percentage points (p=0.035)—again raising doubts about whether the supposed “diversity effect” is anything more than statistical noise.

For first-term Black officers, the results were even more problematic: a 10-percentage point increase in Black senior leadership actually reduced Black junior officer retention by 2.8 percentage points (p=0.096) on medium ships—the very opposite of what Lyall claimed.

Finally, and again contrary to Lyall’s assertions, increases to Hispanic representation showed no meaningful effects on Hispanic retention. Across all three personnel groups, a 10-percentage point increase in the Hispanic share had effects on first-term Hispanic officer and sailor retention that were all statistically indistinguishable from zero at the 95% confidence threshold. Notably, increasing Hispanic representation among senior leadership had a larger effect on non-Hispanic officers (+0.82 points, p=0.011) than on Hispanic officers (+0.56 points, p=0.683)—a result that once again runs directly counter to the claim that coethnic representation uniquely benefits minority retention.

Tellingly, Lyall omitted these inconvenient findings from his report. When confronted with the actual study during cross-examination, he admitted that he had failed to disclose its conflicting results, including the negative effect of Black senior leadership on Black junior officer retention.¹⁵¹ He further acknowledged that he had inaccurately characterized the study’s conclusions, having claimed it showed that increasing minority leadership improves minority retention when, in fact, the study found no such effect for Hispanic personnel and even a negative effect for Black officers in certain contexts.

Time-Series Data on Officer Retention

Beyond citing this Master’s thesis, USNA’s witnesses introduced time-series data showing that minority officer retention among USNA graduates has improved over the past two decades, implying that the gains in representation from racial preferences in USNA admissions contributed to this trend.¹⁵² Yet here, too, the evidence is far from conclusive, as it fails to directly attribute retention increases to racial preferences.

While USNA’s witnesses assert that retention rates have improved, these claims rely on incomplete and selectively presented data. USNA witness Stephanie Miller (Deputy Assistant Secretary of Defense for Military Personnel Policy) presented charts that aggregated all non-White officers into a single “minority” category and applied trend lines “for the ease of being able to observe the trends.”¹⁵³ This phrasing itself suggests that the underlying data exhibited substantial variation and that aggregation was used to obscure year-to-year fluctuations.

Although the data were sealed and not entered into the public record, portions of Miller’s testimony confirm this suspected pattern. Even where retention increases were observed, they were described in testimony as “slight” or “relatively flat”.¹⁵⁴ At the 5-year and 10-year marks, general improvements were noted, but they were not dramatic, and at 15 years, retention for non-White officers was characterized as showing only “some small growth over time.”¹⁵⁵ Moreover, these improvements were generally only observed when all minority groups were collapsed into a single category. When retention was examined separately by racial group, the trends were far more erratic, with some minority groups experiencing stagnation or even declines at key career milestones. For instance, testimony specifically noted that Hispanic officer retention had declined at both the 10-year and 15-year marks, which indicates that any improvements in retention are not uniform across minority groups.¹⁵⁶ What is more, Miller acknowledged that retention improvements were not exclusive to minority officers; White officer retention had also improved in some cases, further complicating any attempt to attribute overall trends to racial preferences in admissions.¹⁵⁷

Indeed, it is hardly obvious that any modest improvements in minority retention necessarily stemmed from USNA’s racial preferences in admissions. Other factors are equally, if not more, plausible. One such factor—which emerged during trial testimony—is the lowering of retention and promotion standards over time. During cross-examination, Miller acknowledged that the military had abandoned its traditional “up or out” policy, which previously required officers to be promoted within two selection cycles or face administrative separation.¹⁵⁸ This change, implemented within the past five years, allowed more officers to remain in service despite not advancing in rank.

In addition to changes in personnel policies, demographic and economic trends also plausibly contribute to overtime variation in retention rates. Without a controlled analysis that isolates the effects of racial preferences from these and other variables, one cannot even begin to make the case that preferences causally influence retention rates.

As it happens, neither could Miller. When directly asked during cross-examination whether she could at all quantify the extent to which racial preferences contributed to the observed retention trends, she admitted that she could not.¹⁵⁹ Pressed further, she acknowledged that she had no precise measurement but simply “appreciated that having the ability to consider [race] may be a positive factor.” In other words, the claim that racial preferences improved retention was based on assumption rather than measurable evidence.

Beyond its failure to isolate the effects of racial preferences, the presented DoD data is strikingly incomplete in two key ways. First, the data begin in 2001, despite USNA’s practice of racial preferences likely beginning much earlier.¹⁶⁰ This omission raises an obvious question: If racial preferences are essential to increasing minority officer retention, why not examine trends before and after their implementation? Doing so would allow for a more meaningful assessment of whether racial preferences had a measurable impact. When asked why the data only start in 2001, Miller’s response was hardly satisfying—she claimed that this date was chosen because it provided a “20-plus-year career perspective” and was where they had “the most consistent data.”¹⁶¹ But this rationale does nothing to explain why earlier retention data—data that could have provided a more comprehensive picture—were excluded.

Another glaring omission is that all of the retention and career progression data Miller presented were limited to USNA graduates alone, despite the fact that the vast majority of naval officers (roughly 80%) are commissioned through ROTC or Officer Candidate School (OCS). Under cross-examination, she admitted that none of the 24 charts introduced at trial included ROTC or OCS officers, even though these officers make up the overwhelming majority of the Navy’s leadership pipeline.¹⁶² This exclusion is particularly striking given the government’s claim that racial preferences at USNA are necessary not just for the Academy, but for the officer corps as a whole.

If racial preferences uniquely improve retention, one would expect USNA graduates—who benefit from these preferences—to exhibit stronger retention rates than their ROTC and OCS counterparts, who do not. A direct comparison across commissioning sources would not establish causality outright, but it would provide valuable evidence about the plausibility of the government’s claim. If retention gains were significantly greater among USNA graduates than among ROTC and OCS officers, it would lend some credibility to the idea that racial preferences play a role. Conversely, if retention rates were similar across commissioning sources—or if ROTC and OCS officers exhibited equal or better retention—this would cast serious doubt on the claim that racial preferences are driving retention improvements.

Yet, inexplicably, the government provided no such analysis. Excluding the officer majority from the data not only weakens the evidentiary basis for its claims but also raises an obvious question: If racial preferences are truly critical to retention, why avoid testing this claim with the most comprehensive dataset available?

In the end, if racial diversity were genuinely a decisive factor in officer retention, one would expect USNA or the Department of Defense to have conducted rigorous studies isolating its effects. Instead, Miller’s admission that she could not quantify the impact of racial preferences, combined with the selective omission of key data, underscores a larger issue: USNA’s claims are asserted, not demonstrated.

Claim #3B. Racial Diversity Increases Recruitment and Domestic Legitimacy

USNA’s witnesses argue that racial diversity in the officer corps is critical for both recruitment and domestic legitimacy. The claim is twofold: first, that minority representation in leadership roles attracts more minority recruits, and second, that a more racially diverse officer corps bolsters public confidence in the military. However, as the data show, neither of these assertions is supported by empirical evidence.

No Evidence That Minority Representation Increases Enlistment or Domestic Legitimacy

USNA’s witnesses, including Jeannette Haynie, Admiral John Fuller, Stephanie Miller, and Ashish Vazirani (Deputy Under Secretary of Defense for Personnel and Readiness), contend that diversity in the officer corps enhances recruitment because prospective minority recruits are more likely to join the military if they “see themselves” in leadership. Haynie cites a RAND report that suggests minority recruiters may have more success engaging with minority recruits due to homophily—the sociological principle that people tend to associate with those similar to themselves.¹⁶³ However, this claim, drawn from a general sociology paper rather than a military study, does not establish that racial representation in leadership influences minority enlistment rates. Meanwhile, Fuller and Vazirani rely primarily on anecdotal evidence rather than empirical research to support their assertions.

If minority representation in the military—and in the officer corps in particular—were a key driver of recruitment, we would expect Black youth to exhibit lower enlistment intentions and career expectations when their group’s representation in leadership is lower. However, as Figure 3.2E demonstrates, this is not the case. Across nearly five decades of Monitoring the Future (MTF) survey responses, Black (and, since 2006, Hispanic) 12th-grade students have consistently expressed equal or greater interest in military service than White students, even during periods when the officer corps was overwhelmingly White. Similarly, Black students have historically been just as likely as White students to expect to pursue a military career or become an officer. These trends provide no support for the claim that racial representation in leadership plays a meaningful role in recruitment decisions.

Figure 3.2E. Military Service Intentions/Expectations and Confidence in/Feelings Toward the Military by Race

Note. Data are weighted. For the top row of panels and the bottom-left panel, lines represent the share of White, Black, and Hispanic 12th-grade students who selected the response shown in each panel’s title. Hispanic data are available only from 2006 onward, as the survey did not previously include a Hispanic category in its racial classification. For the bottom-middle panel, lines represent the share of general population White, Black, and Hispanic respondents who reported having “a great deal” of confidence in the U.S. military (as opposed to “only some” or “hardly any”). For the bottom-right panel, lines represent the average levels of warmth toward the military among White, Black, and Hispanic general population respondents.
Source: Monitoring The Future: A Continuing Study of American Youth (12^th-Grade Survey, 1976–2023); General Social Survey (1972–2022); American National Election Studies (1964–2012).

To further illustrate this point, Figure 3.2F juxtaposes the military service intentions and expectations of Black 12th-graders (left y-axis) with the Black share of all active military officers (right y-axis). If representation in the officer corps meaningfully influenced Black recruitment, we would expect a clear positive relationship between rising Black officer shares and increasing enlistment propensity.

Yet, statistical analysis confirms that no such relationship exists. Correlation tests reveal that the Black share of active-duty officers is negatively associated with all four measures of military service interest (r = -0.675 to -0.417, p < 0.01 in all cases), directly contradicting the argument that increasing representation enhances Black enlistment intentions. However, these negative correlations disappear entirely when the data are detrended (rΔ = -0.011 to 0.058, p > 0.1 in all cases), indicating that even this inverse relationship is likely spurious and driven by broader structural and geopolitical factors.

Figure 3.2F. Military Service Intentions and Expectations Among Black 12th-Graders vs. Black Share of Active Military Officers

Note. Monitoring the Future (MTF) data are weighted. Black lines (left y-axes) represent the share of Black 12th-graders who selected the response shown in each panel’s title. Blue lines (right y-axes) represent the Black share of active military officers.
Source: Monitoring the Future: A Continuing Study of American Youth (12^th-Grade Survey, 1976–2023); Population Representation in the Military Services: Fiscal Year 2021 Summary Report (Appendix D. Historical Tables, Tables D-33 and D-36).

One might argue that representation among enlisted personnel, rather than the officer corps, is the more relevant factor influencing enlistment decisions. However, additional tests provide no support for this hypothesis. While the raw correlation between the Black share of active enlisted personnel and Black military service intentions appears modestly to moderately positive (r = 0.295 to 0.457), these relationships vanish once detrended (rΔ = -0.025 to 0.062). As before, this suggests that any apparent association is an artifact of broader structural and geopolitical forces rather than a genuine effect of representation.

Similarly, Figure 3.2G calls into question the claim that military legitimacy depends on whether its racial composition reflects that of wider American society. If Black Americans’ confidence in the military were tied to representation, we would expect their trust and warmth toward the institution to rise as Black representation in the enlisted ranks increased. Yet, statistical analysis provides no support for this hypothesis.

Correlation tests reveal no meaningful association between the Black share of active enlisted personnel and Black confidence in the military (r = -0.053, p = 0.780), and this weak relationship becomes slightly more negative when detrended (rΔ = -0.278, p = 0.140). Similarly, correlations with Black respondents’ general “warmth” toward the military, as measured by feeling thermometers, are also negative and insignificant (r = -0.464, p = 0.129; rΔ = -0.366, p = 0.269), though this series consists of just 12 observations, limiting its reliability.

Figure 3.2G. Black Confidence in and Warmth Toward the Military vs. Black Share of Active Enlisted Military Members

Note. General Social Survey (GSS) and American National Election Studies (ANES) data are weighted. In the left panel, black lines (left y-axis) represent the share of general population Black respondents reporting “a great deal” of confidence in the U.S. military (as opposed to “only some” or “hardly any”). In the right panel, black lines (left y-axis) represent average warmth toward the military among general population Black respondents. In both panels, blue lines (right y-axis) represent the Black share of active enlisted military members.
Source: General Social Survey (1972–2022); American National Election Studies (1964–2012); Population Representation in the Military Services: Fiscal Year 2021 Summary Report (Appendix D. Historical Tables, Tables D-23 and D-26).

In closing, while USNA’s witnesses repeatedly assert that increasing officer diversity is essential for recruitment and legitimacy, they offer no empirical support for this claim--and the data analyzed above provide no support for it either. In fact, when asked about the existence of any study that examines the level of racial representation in the Navy’s officer corps that would allow minority recruits “to see themselves as future leaders in the military,” USNA witness Ashish Vazirani indicated he was “not aware” of any such study.¹⁶⁴ Meanwhile, long-term data spanning nearly five decades show that minority enlistment propensity and public confidence in the military have remained largely steady regardless of racial representation in leadership. Instead, enlistment decisions appear to be shaped by broader structural and geopolitical factors, such as economic opportunities, family military traditions, wartime dynamics, and educational benefits—none of which are meaningfully affected by the racial composition of the officer corps.

Claim #3C: Racial Diversity Enhances International Legitimacy

Finally, USNA’s witnesses—including Jason Lyall, Jeannette Haynie, Admiral John Fuller, and Ashish Vazirani—argue that racial diversity in the officer corps enhances the U.S. military’s legitimacy abroad. They claim that a diverse force fosters cultural competence, improves relationships with foreign populations, and strengthens alliances. However, these assertions rely on misrepresented research, anecdotal speculation, and an unsupported assumption that racial identity, rather than training and professionalism, determines military effectiveness in international operations.

No Evidence That Racial Diversity Enhances International Legitimacy

At first glance, this argument may seem plausible—linguistic skills, cultural awareness, and regional expertise are indeed valuable in peacekeeping, counterinsurgency, and military diplomacy. However, these attributes are neither inherently racial nor exclusive to non-White personnel. More importantly, the studies cited by USNA’s witnesses fail to demonstrate a causal link between racial diversity and improved military legitimacy abroad.

In fact, many of the sources Lyall and Haynie rely on do not measure racial diversity at all. Lyall, for example, cites Bove and Ruggeri (2016) to argue that diverse UN peacekeeping forces perform better, yet the study does not examine racial or ethnic diversity; instead, it measures diversity in terms of linguistic, national, and geographic differences.¹⁶⁵ Even within these parameters, the study’s results were highly sensitive to model specification and the diversity measures used—raising questions about their robustness and generalizability.

Similarly, Lyall invokes the Nomikos (2022, 2023) studies on UN peacekeeping in Mali to argue that diverse military forces gain greater trust from local populations.¹⁶⁶ However, the study’s own appendix explicitly states that the perceived neutrality of peacekeepers—not their racial identity—was the key driver of local trust and cooperation. When pressed on this point during cross-examination, Lyall was forced to concede that racial identity was not the determining factor in these findings.¹⁶⁷

For her part, Haynie cites a RAND report in arguing that a diverse and inclusive military “builds trust between the population it serves, provides legitimacy with our international allies, and promotes international engagement.”¹⁶⁸ Yet the RAND report offers no substantive support for this claim. It asserts that military representation may improve legitimacy and engagement but provides no empirical evidence that racial diversity, specifically, drives these outcomes.¹⁶⁹ Instead, the report conflates general societal representation with racial diversity and fails to explain how racial composition—rather than competence, professionalism, or institutional neutrality—determines legitimacy in the eyes of foreign allies or local civilian populations. If legitimacy were primarily a function of racial diversity, we would expect to see measurable improvements in operational outcomes tied to racial composition—something neither RAND nor Haynie demonstrates.

Other USNA witnesses, such as Admiral John Fuller and Ashish Vazirani, rely on broad assertions and anecdotal observations rather than empirical evidence. Fuller testified that racial diversity has “a great impact” on the military’s international credibility, citing personal experiences where diverse crews visiting foreign ports were seen as a reflection of American identity.¹⁷⁰ However, under cross-examination, he conceded that his testimony—which he previously described as “fact-based”—was derived entirely from personal observation rather than systematic analysis.¹⁷¹ Similarly, Vazirani claimed that a diverse officer corps enhances legitimacy with allies and foreign populations by demonstrating American values and improving cultural competence.¹⁷² Yet, he provided no studies or data to substantiate this claim, instead relying on generalized assertions about the benefits of diversity. His argument ultimately rests on the assumption that foreign partners and local populations perceive racial diversity as a marker of legitimacy—an assumption for which no evidence was offered.

Beyond the lack of empirical support, the concept of “legitimacy” as a justification for racial preferences is nebulous and resistant to measurement. As SFFA pointed out during closing arguments, it is unclear how legitimacy—whether domestic or international—could be meaningfully quantified, let alone tied to the racial composition of the military.¹⁷³ Moreover, even if racial diversity were a factor in international legitimacy, it would be just one among many and not necessarily the most important. USNA’s witnesses provide no evidence that the racial composition of the military is a key determinant of how the U.S. military is perceived abroad, nor do they demonstrate that increasing officer diversity has had any measurable effect on mission outcomes in international engagements.

3.3 USNA’s Diversity Dilemma

Even if one accepts the DoD’s diversity goals—which I do not—the reality is that these goals cannot be achieved without lowering accession standards in specific naval communities. The naval communities with the lowest levels of minority representation are precisely those with the most demanding selection criteria—communities that also produce a disproportionate share of senior leadership. Unless the Navy is willing to lower these standards, it cannot achieve its diversity objectives through racial preferences in USNA admissions alone. The impact of such preferences on the composition of senior leadership will be negligible, making them largely symbolic with little practical effect.

3.3.1 Performance Disparities at USNA

USNA performance disparities are evident across multiple measures of academic, military, and physical aptitude, all of which shape midshipmen’s career trajectories. The analysis in Section 2.8.1 offered a preliminary glimpse into these disparities, but only a partial one. That analysis was limited to three fall semester course grades for the Classes of 2025–2026 and commandant’s list appearances for the Classes of 2023–2027—primarily reflecting academic performance in a single semester for two class years. By contrast, internal USNA data spanning the full tenures of multiple class years (2014–2020) offer a far more comprehensive assessment of midshipmen performance. These data capture not only academic standing across all coursework but also military aptitude and physical readiness—key dimensions of midshipmen success that directly impact service selection and long-term career progression.

First, as shown in Figure 3.3A, the internal data reveal significant racial differences in midshipmen aptitude grades, which measure leadership potential and suitability for commissioned service. These grades, assigned on a 4.0 scale, reflect evaluations from midshipmen’s chain of command, including senior raters and company officers. White midshipmen earned the highest aptitude grades (3.18), followed by Asians (3.05) and Hispanics (2.98), while Black midshipmen had the lowest scores (2.70)—a deficit of nearly half a full grade category. These gaps, which persist across both varsity and non-varsity midshipmen, underscore broad and systemic differences in assessed leadership potential.

Figure 3.3A Racial Disparities in Midshipmen Aptitude Grades (USNA Classes 2014–2020)

Note. Bars represent mean aptitude grades by race and varsity status. The dashed horizontal line indicates the overall sample mean. Group sample sizes are reported in parentheses. Aptitude grades are assigned on a 4.0 scale and reflect evaluations from midshipmen’s chain of command, including senior raters and company officers.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Plaintiff's Exhibit 148, Email from Steve Vahsen to David Forman re: “Any available time?” with attachment, Summary Assessment of Midshipmen Equity by Race/Ethnic Group, Figures 11 and 29, at 16 and 32 (D. Md. Sept. 20, 2024).

A similar racial pattern emerges in Physical Readiness Test (PRT) scores, plotted in Figure 3.3B. The PRT is a standardized assessment of midshipmen’s physical fitness and military readiness, measuring endurance, muscular strength, and core stability. Designed to ensure combat readiness and reduce injury risk, the test is scored from “Outstanding (High)” (100 points) to “Probationary” (45 points), though standards differ by sex, making direct male-to-female comparisons inappropriate.¹⁷⁴

As with aptitude grades, White male and female midshipmen score higher than their counterparts in all other racial groups. The largest gaps exist between White and Black midshipmen. Whereas the former average scores (87, 86) place them between the ‘Excellent/High’ and ‘Outstanding/Low’ performance categories, Black men (79) and Black women (77) score nearly a full category lower, placing them at the ‘Low’ to ‘Medium’ level of the ‘Excellent’ category. Notably, this pattern mirrors observed differences in Candidate Fitness Assessment (CFA) scores among admitted applicants, suggesting that, as with academics, racial disparities in physical fitness performance are already present at the point of admission.¹⁷⁵

Figure 3.3B Racial Disparities in Midshipmen Physical Readiness Test Scores (USNA Classes 2014–2020)

Note. Bars represent mean Physical Readiness Test (PRT) scores by race and sex. The PRT assesses endurance, muscular strength, and core stability, with scores ranging from "Outstanding (High)" (100 points) to "Probationary" (45 points). Standards differ by sex, making direct male-to-female comparisons inappropriate. Dashed horizontal line represents the overall sample mean.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Plaintiff's Exhibit 148, Email from Steve Vahsen to David Forman re: “Any available time?” with attachment, Summary Assessment of Midshipmen Equity by Race/Ethnic Group, Figure 15, at 20 (D. Md. Sept. 20, 2024).

Given the disparities observed in aptitude grades and PRT scores, it is unsurprising that similar patterns emerge in midshipmen’s Order of Merit indexes, namely Military Order of Merit (MOM) and Academic Order of Merit (AOM) rankings.

MOM rankings (Figure 3.4C, left panel), a broad measure of non-academic performance that incorporates aptitude grades and PRT scores alongside conduct grades and military coursework, generally follow the same racial hierarchy. Among non-varsity midshipmen, White midshipmen rank 485th on average, followed by Asians (618th), Hispanics (645th), and Blacks (795th), who rank more than 300 places behind their White peers and nearly 200 places below Asians and Hispanics. This pattern holds among varsity midshipmen as well, though the gaps between Whites (550th) and both Asians (623rd) and Hispanics (642nd) are somewhat narrower—whereas Black varsity midshipmen (870th) continue to rank substantially lower than all other groups.

Figure 3.4C Racial Disparities in Midshipmen Order of Merit Rankings (USNA Classes 2014–2020)

Note. Bars represent mean rankings for Military Order of Merit (MOM), Academic Order of Merit (AOM), and Overall Order of Merit (OOM) by race and varsity status. MOM incorporates non-academic performance measures, including aptitude grades, PRT scores, conduct grades, and military coursework. AOM reflects midshipmen’s academic standing across all coursework. OOM is a composite measure incorporating both academic (AOM) and non-academic (MOM) performance but is not a direct function of either. Lower rankings indicate better performance. Dashed horizontal lines represent the sample mean for each index. Group sample sizes are reported in parentheses.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Plaintiff's Exhibit 148, Email from Steve Vahsen to David Forman re: “Any available time?” with attachment, Summary Assessment of Midshipmen Equity by Race/Ethnic Group, Figures 26–28, at 31–32 (D. Md. Sept. 20, 2024).

AOM rankings (Figure 3.4C, middle panel), a comprehensive indicator of academic standing, largely mirror those observed in MOM, with White midshipmen ranking highest on average, followed by Asians, Hispanics, and Black midshipmen, who consistently rank the lowest. Thus, across both non-academic and academic dimensions, minority midshipmen tend to perform worse--and Black midshipmen significantly worse--than their White peers.

Naturally, this reality is broadly reflected in Overall Order of Merit (OOM) scores (Figure 3.4C, right panel), a composite measure that incorporates (but is not a direct function of) academic (AOM) and non-academic (MOM) performance. OOM rankings follow the same racial hierarchy, with White midshipmen ranking highest on average, followed by Asians, Hispanics, and Black midshipmen at the bottom.¹⁷⁶ Notably, while Black midshipmen comprised 7% of all midshipmen across the Classes of 2014–2020, they made up just 0.1% of those ranking in the top 100, yet accounted for 35% of those in the bottom 100 and 26% of those ranked between 1001st and 1100th.¹⁷⁷

3.3.2 Impact of Performance on Entry into Elite Warfare Communities

These disparities in overall performance have significant consequences for midshipmen’s career trajectories, particularly in service assignments to the Navy’s most selective (and least racially diverse) warfare communities, which produce an outsized share of senior leadership.¹⁷⁸ Academic and non-academic performance records, including as captured in Order of Merit rankings, are key determinants of assignments to such communities.¹⁷⁹ For instance, Aviation assignments are based primarily on OOM scores (50%), with additional weight given to Aviation Points (35%) and scores on the Aviation Selection Test Battery (15%).¹⁸⁰ Submarine and Nuclear Surface Warfare (SWO-N) assignments require a minimum CQPR (cumulative GPA) of 2.5, with Naval Reactors staff conducting additional academic and technical screenings before approving candidates for nuclear programs.¹⁸¹ Naval Special Warfare (SEAL) selection involves a screening panel that evaluates academic, athletic, and leadership records, alongside completion of the SEAL Officer Assessment and Selection (SOAS) program.¹⁸²

Given these stringent selection criteria, it is unsurprising that racial disparities in USNA performance translate into racial disparities in the community assignments of USNA graduates. Figure 3.3D visualizes these differences using representation quotients, which compare each racial group’s share of a given warfare community to its share of the overall population of USNA assignees. A ratio of 1.0 indicates proportional representation, values above 1.0 indicate overrepresentation, and values below 1.0 indicate underrepresentation. Black graduates, for instance, are substantially underrepresented among those assigned to Nuclear (0.54), Aviation (0.57), and Special Warfare/Operations (0.19) communities—each of which imposes distinct selection hurdles. By contrast, Black graduates are overrepresented in Non-Nuclear Surface Warfare (2.2), Information Warfare/Cyber (1.6), and Restricted Line/Support (1.3), none of which impose the same academic or physical barriers.

Figure 3.3D Racial Representation in USNA Graduate Service Assignments Relative to Overall Class Composition

Note. Bars represent representation quotients for each racial/ethnic group across different naval and Marine Corps service assignments. Quotients are calculated by dividing a group’s share within a given community by its share among all USNA assignees. A value of 1.0, denoted by the dashed horizontal line, indicates proportional representation, while values above 1.0 indicate overrepresentation and values below 1.0 indicate underrepresentation. Data consist of USNA graduates from the Classes of 2014–2024. Racial/ethnic group sample sizes (i.e., the number of USNA graduates assigned to a community within each racial/ethnic category) are reported in parentheses next to each group label. Each community's share of total assignees across all racial/ethnic groups is reported in parentheses next to the community label. The dashed horizontal line at 1.0 on the y-axis represents proportional representation. For simplicity in presentation, service categories combine related communities while accounting for their varying academic, technical, and physical demands.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, DOJ_USNA Class Years 2014–2024_Service Assignment by Race-Ethnic Group by Class Year, Plaintiff's Exhibit 610 (D. Md. Sept. 16, 2024).

Hispanic and Asian graduates also show distinct patterns of representation. Hispanics are least represented in Special Warfare/Operations (0.81) and most overrepresented in Non-Nuclear Surface Warfare (1.25). Asians, meanwhile, are most concentrated in the academically demanding Nuclear community (1.4) but are least represented in the physically demanding Special Warfare/Operations community (0.19). Notably, this distribution aligns with broader patterns observed among Asian admits to USNA, who tend to score high on academic indicators but lower on physical fitness measures and athletic extracurricular involvement (see Figure 2.2B).

Although just 20% of all naval officers are commissioned through USNA, similar patterns emerge when examining assignments across all commissioning sources.¹⁸³ Data from the 2020 Task Force One Navy report, visualized in Figure 3.3E, confirm that minority representation is especially low in the Navy’s most selective warfare communities. For instance, despite comprising 13.4% of the U.S. population and 19% of all enlisted active naval personnel, Black officers account for just 7.3% of all naval officers—and their representation drops even further in elite communities: 1.2% of Special Warfare officers, 2.0% of submarine officers, and 2.3% of aviation officers.

Figure 3.3E. Racial Representation in Navy Warfare Communities Across All Commissioning Sources

Note. Bars represent the racial composition of various Navy officer communities, enlisted personnel, and the general U.S. population. The black, purple, green, orange, and blue bars indicate each racial group’s share of officers (circa FY 2019) within specific naval warfare communities (Special Warfare, Submarine Warfare, Surface Warfare, and Aviation, respectively). The red bar represents each group’s share of all Navy officers. The light blue bar indicates each group’s share of active enlisted personnel, while the gray bar represents the group’s share of the U.S. population based on 2018 Census estimates.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Task Force One Navy: Our Navy Team – Navigating A Course to True North, Plaintiff's Exhibit 600, at 15 and 101–102 (D. Md. Sept. 23, 2024).

Hispanics, who make up 18.5% of the U.S. population and 18% of enlisted naval personnel, constitute 8.6% of all officers, but are similarly underrepresented in key warfare communities—comprising just 5.9% of Special Warfare officers, 6.6% of submarine officers, and 7.0% of aviation officers. Asian representation, while at parity among enlisted personnel (5.6%) and officers (5.6%) relative to their U.S. population share, is significantly lower in elite warfare roles. Asians constitute just 1.8% of Special Warfare officers, 3.1% of aviation officers, and 4.3% of submarine officers.

3.3.3 The Leadership Pipeline and Barriers to Advancement

The downstream impact of these racial disparities in service assignments becomes clear when examining the Navy’s senior leadership pipeline. Officers in Special Warfare, Aviation, Submarine Warfare make up roughly one-third (33.9%) of all naval officers, yet account for nearly half (48.2%) of flag officers (O-7 and above) and an even greater share—60%—of the Navy’s most senior leaders (O-10).¹⁸⁴

Consistent with minority underrepresentation in these communities, Figure 3.3F shows that minority representation declines sharply as rank increases. Black officers, who comprise 7.5% of O-1s, shrink to just 2.8% at O-7 and effectively disappear at O-8 and above. Hispanic officers follow a similar trajectory, dropping from 11.4% at O-1 to 2.8% at O-7 and vanishing completely at O-10. Meanwhile, White officers increase from 74.8% at O-1 to 100% at O-10.

Figure 3.3F Racial Composition of Navy Officers by Commissioning Rank

Note: Each bar represents a different naval officer rank, with color-shaded regions indicating the racial composition of officers at each rank level (circa FY 2019). Sample sizes for each rank are provided in parentheses. Data illustrate the decline in minority representation as rank increases, particularly at the flag officer level (O-7 and above).
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Task Force One Navy: Our Navy Team – Navigating A Course to True North, Plaintiff's Exhibit 600, at 102–103.

3.3.4 The Role of Promotion and Attrition in Minority Underrepresentation

But while minority underrepresentation at senior ranks is partly a function of their underrepresentation in the Navy’s most selective warfare communities, it is also a product of their lower promotion rates after entering the officer pipeline. Both of these bottlenecks—at the point of service assignment and at the promotion stage—are ultimately tied to disparities in academic and non-academic performance.

First, as discussed, the most selective warfare communities disproportionately recruit from the top of the academic and physical ability distribution. Given their emphasis on Order of Merit (OOM) rankings, academic performance, and aptitude and PRT scores, midshipmen with lower academic and military records—disproportionately minorities—are far less likely to enter these elite pathways, narrowing the leadership pipeline at its earliest stage.

Second, USNA performance is strongly predictive of promotion outcomes, even though it is not directly considered by promotion boards. Indeed, a recent study of USNA midshipmen found academic performance—not athletic participation, leadership roles, or other extracurricular achievements—to be the strongest predictor of post-USNA career outcomes, such as promotion to senior ranks.¹⁸⁵

This aligns with an earlier study of the U.S. Air Force Academy (USAFA), which was cited during trial testimony.¹⁸⁶ The USAFA study found that Order of Merit rankings—namely AOM and MOM—are among the strongest predictors of eventual promotion to higher officer ranks. Although promotion boards do not consider these rankings directly, they are nonetheless correlated with officer quality traits (e.g. cognitive aptitude) that shape long-term career trajectories. The study concluded that if the Air Force sought to mitigate racial disparities in promotions, the most effective intervention would be to recruit minority cadets who rank higher in USAFA’s order of merit distribution, rather than implementing race-based selection policies later in an officer’s career.

In addition to producing minority officers who are less likely to enter selective naval communities and ascend to higher ranks, USNA’s race-conscious admissions policies also result in higher separation rates among minority midshipmen, further reducing the number who ultimately commission as officers. As shown in Figure 3.3G, Black midshipmen have the highest overall separation rate (20%), followed by Hispanics (14.5%), Asians (10.4%), and Whites (10.1%). While some separations are voluntary, a significant portion result from academic failure, misconduct, or physical/medical disqualification. Among those who separate, 32% of Whites do so involuntarily, compared to 58% of Black midshipmen, 50% of Asians, and 46% of Hispanics. Alternatively, if viewed in terms of group rates, Black midshipmen (11.7%) are 3.6 times more likely than their White peers (3.3%) to be involuntarily separated, while Hispanics (6.7%) are twice as likely, and Asians (5.2%) 1.6 times as likely.

Figure 3.3G. Racial Disparities in USNA Midshipmen Separations (USNA Classes 2014–2020)

Note. Horizontal bars in the top row represent overall, academic, physical, and conduct-related separation rates by race/ethnicity. Vertical bars in the bottom row display representation quotients, calculated by dividing a group’s share of separations by its share among all USNA assignees. A quotient of 1.0 (denoted by the dashed horizontal line) indicates proportional representation, while values above 1.0 signify overrepresentation and values below 1.0 indicate underrepresentation. Data include USNA graduates from the Classes of 2014–2024. Racial/ethnic group sample sizes are reported in parentheses next to each group label.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Plaintiff's Exhibit 148, Email from Steve Vahsen to David Forman re: “Any available time?” with attachment, Summary Assessment of Midshipmen Equity by Race/Ethnic Group, Tables 1–5, at 4–8.

The upshot of these numbers is that a portion of the very minority candidates—including 1 in 5 Black midshipmen—USNA admits to enhance diversity never even commission as officers, compounding the effects of underrepresentation in elite warfare communities and lower promotion rates after commissioning. Thus, while racial preferences in admissions may increase the number of minority midshipmen at the outset, higher attrition rates ultimately undermine these efforts, further limiting the policy’s intended impact on long-term officer representation.

3.3.5 The Limits of Race-Conscious Admissions for Demographic Engineering

All told, race-conscious admissions are inherently limited in their ability to single-handedly diversify the least diverse naval communities and senior ranks. USNA produces only 20% of all naval officers, meaning its admissions policies affect just a fraction of the officer pipeline. Higher attrition rates further shrink the pool of minority graduates available for service assignments. Among those who do graduate, lower academic and military performance records disproportionately filter them out of the most selective warfare communities—the very communities that serve as the primary pathway to senior leadership. This cascading effect ensures that the impact of racial preferences on long-term leadership diversity is, at best, modest.

Yet despite these inherent constraints, the Navy justifies racial preferences in USNA admissions on the assumption that they will lead to a more diverse senior leadership. But the data suggest this assumption is fundamentally misguided—the true bottleneck in shaping flag officer composition is not who is admitted to USNA, but who gains entry into the most competitive warfare communities.

Perhaps the only way to meaningfully alter this trajectory would be to consider race as a factor in service assignments, explicitly steering more minority graduates into elite naval communities. However, as USNA witness Ashish Vazirani conceded during trial testimony, such an approach would violate DoD policy and require changes to qualification standards.¹⁸⁷ In other words, the Navy’s diversity goals cannot be met without either fundamentally altering its selection process or reducing the performance standards of its most selective communities—both of which would carry significant operational risks.

Unless the Navy is willing to openly advocate for lowering the standards of these communities, its diversity goals remain unrealistic. At some point, it must decide whether it prioritizes its meritocratic selection process—or racial balance in leadership. It cannot have both.

4. Judge Bennett’s Ruling

The weakness of USNA’s case is matched only by the shallowness of Judge Richard Bennett’s ruling, which, without exception, uncritically adopts every one of USNA’s talking points. His decision is so one-sided that it blurs the line between adjudication and advocacy.

To be clear, the outcome itself was not surprising. Both Harvard and UNC initially prevailed at the district court level before their policies were ultimately struck down by the Supreme Court. But even by the low bar set in those cases, Bennett’s ruling stands out for its complete lack of skepticism toward USNA’s claims. Rather than applying a rigorous legal standard, he treats the Academy’s justifications as self-evident, adopting them wholesale with no serious scrutiny. His ruling reads less like an impartial legal analysis and more like an advocacy brief on USNA’s behalf. At no point does he subject the Academy’s claims to meaningful examination, demand empirical justification, or seriously engage with counterarguments. Instead, he accepts USNA’s assertions at face value, turning what should have been an exacting legal review into little more than a judicial rubber stamp.

4.1 What Strict Scrutiny Requires

Strict scrutiny is the highest standard of judicial review. And because racial classifications are inherently suspect under the Equal Protection Clause, the Court applies an especially exacting version of strict scrutiny in such cases. To pass this test, the government must prove that its use of race advances a compelling interest—one that is specific, concrete, and backed by substantial evidence. Mere assertions are insufficient; the interest must be demonstrably urgent and essential. Second, the policy must be narrowly tailored to that interest, meaning that no workable race-neutral alternatives exist that could achieve the same goal with comparable effectiveness. If a race-neutral approach could accomplish the objective “about as well,” the government’s use of race fails the test.¹⁸⁸

Even if a policy meets these two requirements, additional constitutional constraints apply. First, race cannot be used as a negative or as a stereotype—meaning no racial group can be placed at a disadvantage, nor can admissions rely on crude generalizations about individuals based on race.¹⁸⁹ The Supreme Court has been unequivocal on this point: any admissions policy that imposes a measurable penalty on applicants of a certain race fails strict scrutiny. Similarly, Grutter made clear that racial preferences cannot be justified by presumed racial viewpoints or characteristics.¹⁹⁰

Second, racial classifications must have a logical endpoint. Racial preferences must be temporary and subject to clear limits, rather than serving as an open-ended mechanism for demographic engineering. If a policy lacks a defined stopping point, it amounts to racial balancing, which is explicitly unconstitutional.¹⁹¹ These constraints ensure that racial classifications remain an exceptional, last-resort measure—not a permanent feature of government policy.

Yet rather than rigorously applying these principles, Bennett treats strict scrutiny as a procedural formality—one that USNA clears with little effort. The most striking failure of his ruling is how he shifts the burden of proof. Instead of requiring USNA to supply empirical evidence demonstrating that racial diversity in the officer corps serves a compelling interest, he assumes its validity. Instead of seriously considering whether race-neutral alternatives could achieve similar goals, he dismisses them outright. Instead of requiring clear limits on racial preferences, he permits them to persist indefinitely under vague assurances of future review.

This is not strict scrutiny—it is judicial abdication. What follows is a systematic examination of how Bennett’s ruling fails at every stage, exposing a review that defers to USNA rather than scrutinizing its claims.

4.2 Is a racially diverse officer corps a compelling national security interest?

4.2.1 The Supposed ‘Diversity Consensus’

Judge Bennett’s ruling assumes, without scrutiny, that racial diversity is indispensable to national security. Rather than treating this claim as an unproven hypothesis requiring rigorous validation, he frames it as settled military doctrine. He begins his ruling by declaring that “a racially diverse officer corps is a national security interest,” citing “military history” and a 2009 diversity commission mandated by Congress as evidence.¹⁹² Yet neither source withstands scrutiny. The claim’s origins are recent, overtly political, and lack empirical validation.

First, the Military Leadership Diversity Commission (MLDC), which Judge Bennett cites as proof of a longstanding military consensus, was neither neutral nor solely focused on military readiness. It was created through an amendment to the 2009 National Defense Authorization Act, spearheaded by then-Representative Ben Cardin with backing from the Congressional Black Caucus—a clear indication that its establishment was driven by political, rather than strictly military, considerations.¹⁹³ Moreover, the commission itself was not composed primarily of military strategists concerned with combat effectiveness.¹⁹⁴ Instead, it was stacked with DEI consultants, bureaucrats, and activists, with only a token military presence. Its recommendations—now being cited as justification for racial preferences—were explicitly geared toward demographic parity, not operational effectiveness.¹⁹⁵ Nowhere in its report did it provide rigorous empirical evidence that racial diversity improves combat effectiveness, unit cohesion, or military performance.

Despite this, the MLDC’s report laid the groundwork for the Department of Defense’s Diversity and Inclusion Board, which Secretary of Defense Mark Esper ordered into existence on June 18, 2020—just weeks after the death of George Floyd and amid nationwide “racial justice” protests.¹⁹⁶ The timing strongly suggests that the Board’s creation was driven by political expediency rather than military necessity. Like the MLDC, this Board was not composed primarily of warfighters or operational experts. Instead, it was dominated by DEI bureaucrats, human resources personnel, and corporate diversity consultants.¹⁹⁷ Among its key participants were the Defense Diversity Working Group and the Executive Defense Diversity Working Group, both of which were explicitly tasked with reshaping military personnel policies under the guise of national security.

Unsurprisingly, the Diversity Board’s final report—which Judge Bennett also cites to bolster the supposed “diversity-national security” consensus—was transparently political.¹⁹⁸ It recycled the same narrative, invoked historical grievances, and framed diversity as a “strategic imperative,” yet failed to provide any empirical evidence that racial diversity enhances military readiness. Instead, it served as a vehicle for further embedding DEI initiatives into military policy without ever demonstrating their necessity or effectiveness in combat environments.

While Judge Bennett further insists that the “diversity-national security” rationale is grounded in “American military history”, in doing so, he simply parrots the narrative constructed by USNA and the DoD’s Diversity and Inclusion Board. This tendentious account selectively interprets past racial tensions in the military as proof that racial preferences are necessary to prevent the recurrence of Vietnam-era racial unrest. But this interpretation is not just misleading—it is internally contradictory.

Bennett’s own ruling acknowledges that significant racial unrest in the military followed the lowering of the Navy’s recruitment standards at the end of the draft—a point I previously highlighted in Section 3.2.2. And yet, he never connects the dots. He describes how, after the draft ended, the Navy lowered recruitment standards to meet manpower needs, which coincided with heightened racial tensions and violence.¹⁹⁹ Yet, rather than recognizing the connection between declining standards and unrest, he fixates solely on the racial composition of the officer corps—both as the cause of unrest and the supposed solution.²⁰⁰

This omission is telling. If the racial unrest of the 1970s was largely a consequence of lowered recruitment standards, then the logical response would have been to restore and uphold those standards—not to implement racial preferences in officer selection. Yet Judge Bennett ignores this entirely, accepting without question USNA’s claim that a more racially diverse officer corps would have prevented unrest. At no point does he demand actual evidence that racial diversity would have mitigated these tensions, nor does he question why racial representation—rather than leadership competence, discipline, or military professionalism—is assumed to be the determining factor in military cohesion.

Judge Bennett’s failure to recognize this contradiction is not merely an oversight—it is symptomatic of his broader refusal to subject USNA’s claims to scrutiny. A legitimate court would have examined the MLDC’s credibility, interrogated the historical claims underpinning USNA’s rationale, and demanded empirical proof that diversity enhances military effectiveness. Instead, Bennett uncritically adopts USNA’s assumptions, treating them as legal justifications rather than subjecting them to the rigorous constitutional scrutiny strict scrutiny demands.

4.2.2. The “Diversity Bonus”

Judge Bennett uncritically defers to USNA’s sweeping, unsubstantiated claims that diversity enhances military cohesion, lethality, recruitment, retention, and strategic legitimacy. Rather than requiring rigorous proof, he treats these assertions as self-evident truths. To justify this deference, he begins by citing several Supreme Court cases, which he claims establish judicial precedent for deferring to military judgment on personnel policies.²⁰¹ Yet none of these cases involve racial classifications or affirmative action in military admissions. Instead, they concern religious dress accommodations (Goldman v. Weinberger; Singh v. Berger) and military policies on sexual orientation (Cook v. Gates). The military has historically received some deference in personnel administration, but racial classifications—being inherently suspect—trigger a far more demanding standard of review, a distinction Bennett entirely ignores.

Even setting aside this heightened scrutiny, these rulings do not support the use of racial preferences to achieve cohesion or effectiveness. If anything, they underscore that cohesion in the military has historically been achieved through uniformity, shared mission, and discipline—not through demographic diversity for its own sake. By misapplying these cases, Judge Bennett manufactures precedent where none exists, substituting deference to legitimate operational needs with deference to an unproven, politically driven personnel policy.

But none of this matters to Judge Bennett, who declares that USNA has “presented significant evidence” proving a compelling national security interest in racial diversity.²⁰² In fact, this supposed “evidence” consists of the same 2020 DoD Diversity and Inclusion report, a handful of tangential studies, and the unverified assertions of USNA’s own witnesses—all of which were previously debunked in Section 3.2.

Misrepresenting the Evidence

Consider the three studies Judge Bennett cites as supposedly supporting “DoD’s military judgment regarding the importance of diversity to unit cohesion and thus to national security.”²⁰³ The first is the OPA ‘climate study,’ discussed in Section 3.2.2. Relying on the testimony of Dr. Jeannette Haynie, Bennett claims this analysis demonstrated that “diverse leadership increases the likelihood that leaders recognize a climate’s impact on their team members, mitigate obstacles, and improve morale and cohesion” while also showing “the impact of improved diversity on the military’s climate and performance of its units.”²⁰⁴

Had Bennett subjected the study to even minimal scrutiny, he would have seen that none of these claims are supported. The study did not demonstrate that diversity improves cohesion. Instead, the study and its underlying data revealed: (a) little to no racial differences in self-reported morale, readiness, and cohesion; (b) that more personnel felt D&I efforts were excessive (18%) than those who reported serving in “unhealthy” unit climates (12%); and (c) that minorities in unhealthy units were actually less likely to leave the military than Whites. Most critically, the study failed to establish any causal link—let alone one between racial diversity and improved military performance.

But it gets even worse. The second “study” cited by Bennett as evidence of diversity’s impact on unit cohesion and lethality is a 2023 RAND report titled “Women, Peace, and Security in Action: Including Gender Perspectives in Department of Defense Operations, Activities, and Investments.” The title alone should raise immediate red flags: this report is about gender inclusion, not racial or ethnic diversity, and it provides no empirical analysis of diversity’s effect on military cohesion or performance. Instead, the authors present a series of anecdotal “vignettes” highlighting instances in which gender perspectives have been incorporated into DoD operations. While this may be of interest to those studying gender policy in international security, it has zero relevance to the question at hand—whether racial preferences in USNA admissions improve military performance. Its inclusion in Bennett’s ruling is not just misleading; it is entirely irrelevant.

A third and final “study” Bennett cites is a 2022 report titled “Leveraging Diversity for Military Effectiveness: Diversity, Inclusion, and Belonging in the UK and US Armed Forces”. Unlike the Women, Peace, and Security report, this one at least mentions racial and ethnic diversity. However, it suffers from the same fundamental flaw: it presents no empirical evidence whatsoever. Instead, it offers speculative assertions, theoretical discussions, and ideological commitments to “inclusion” as an inherent good. Its language is littered with qualifiers—diversity may improve X, inclusion could strengthen Y—but at no point does it provide concrete data demonstrating that increased racial diversity enhances unit cohesion, lethality, or operational effectiveness.

Bennett’s Circular Reasoning and Dismissal of Conflicting Evidence

Despite the absence of empirical validation, Judge Bennett treats these reports as definitive proof that racial diversity enhances unit cohesion and lethality. His reasoning is circular: because DoD officials and USNA witnesses claim that diversity strengthens military effectiveness, and because these reports merely echo those claims, he deems the argument settled. But this is not how strict scrutiny operates. Under constitutional analysis, the burden is on the government to demonstrate—not merely assert—that its use of race serves a compelling interest. The government cannot manufacture its own “proof” by citing reports that do nothing more than restate its own unverified claims.

Yet that is precisely what Bennett allows here. Worse still, he dismisses conflicting evidence outright. When SFFA’s military witnesses—Brigadier General Christopher Walker and Colonel Dakota Wood—testified that, in their experience, racial diversity had no relation to unit cohesion or performance, Bennett waved their testimony away as mere “considered personal opinions” while treating DoD’s politically motivated assertions as “research-backed conclusions.”²⁰⁵ In doing so, he abandoned any pretense of impartial evaluation, effectively shifting the burden of proof onto SFFA rather than holding the government to the high evidentiary standard that strict scrutiny demands.

Redefining Unit Cohesion and Lethality to Dodge Scrutiny

The problems with Bennett’s reasoning go beyond his reliance on unsubstantiated reports. Even if these studies did contain empirical analysis—which they do not—his ruling still fails to establish the necessary causal connection between racial diversity and combat effectiveness. In fact, Bennett’s own decision acknowledges that USNA’s witnesses eventually conceded that “the racial composition of a unit alone has no effect, for example, on that unit’s ability to complete tasks such as firing missiles.”²⁰⁶

Rather than grappling with the implications of this admission, Bennett dismisses it by accusing SFFA of adopting an unduly narrow definition of unit cohesion and lethality--“as a given unit’s ability to complete the tasks necessary to its mission”—one he claims “contradicts” the definition used by DoD and military leaders²⁰⁷ Yet Bennett never actually articulates what this alternative definition entails, nor does he explain why a broader definition would be more appropriate. More fundamentally, his argument betrays a profound misunderstanding of military priorities.

Mission effectiveness is the military’s ultimate goal. Unit cohesion, morale, and leadership are valuable only to the extent that they enhance battlefield performance. The military does not pursue cohesion for its own sake; it prioritizes cohesion because, when properly cultivated, it improves operational readiness and warfighting capability. If racial diversity does not contribute to that performance, then it cannot be classified as a compelling government interest under strict scrutiny.

Bennett sidesteps this reality by treating cohesion, morale, and leadership as independent objectives rather than as conditions that must ultimately enhance mission effectiveness. In doing so, he lowers the government’s evidentiary burden and shields racial preferences from meaningful constitutional review.

Recruitment and Retention

Judge Bennett next turns to the claim that racial diversity in the officer corps improves recruitment and retention—one he proclaims that USNA has provided “substantial data” for and has ultimately “proven”. Yet, as with his analysis of unit cohesion and lethality, he offers no real scrutiny of the evidence, instead accepting the testimony of military officials as conclusive.

A core premise of this claim is that increasing racial diversity in leadership roles encourages more minority recruits to enlist and remain in service. In support of this, Bennett relies on vague references to DoD reports, testimony from military officials, and a handful of studies—none of which establish a causal link between officer diversity and recruitment or retention rates. In fact, some of the very data cited in his ruling contradict the notion that minority representation is a decisive factor in enlistment or career persistence.

For instance, Bennett cites the DoD Officer Retention and Promotion Barrier Analysis Study, which states that “a lack of diversity in recruiters and advertisements may contribute to a lack of diversity in the officer corps.”²⁰⁸ The careful wording here is telling. The use of “may” signals the absence of empirical proof. Instead of demonstrating a causal connection, the report merely speculates that diversity in recruitment materials and leadership could influence minority enlistment. That is not proof—it is a hypothesis. Yet, Bennett presents it as definitive evidence.

A similar issue arises in his discussion of minority retention. Bennett references the DoD Office of People Analytics (OPA) study, which indicates that service members who perceive a “healthy diversity and inclusion climate” report higher intent to stay in the military.²⁰⁹ However, as discussed in Section 3.2.2, this same study also found that minorities already exhibit higher retention rates than their White counterparts, despite Whites being more likely to identify their climates as “healthy.” This raises obvious questions about whether the perceived climate-retention relationship is truly causal. In any case, it bears repeating: the study’s design does not allow for causal inference.

This pattern of superficial analysis extends to Bennett's discussion of the Naval Postgraduate School study that Lyall relied on to support the claim that increasing racial diversity improves minority retention rates. Regarding this study, Bennett writes:

“Plaintiff also disputes the data and testimony Defendants presented, citing one competing study—'The Effects of Diversity Among Peers and Role Models on U.S. Navy Retention’—that showed a negative correlation between racial diversity in the officer corps and minority officer retention.”²¹⁰

Bennett’s framing is misleading. By claiming that SFFA “cited” a “competing” study, he implies that plaintiffs cherry-picked an outlier to undermine USNA’s otherwise strong evidence. In reality, however, this study was introduced by USNA’s own expert, who—when confronted with its findings during cross-examination—was forced to admit that they contradicted his broader claims about diversity and retention. Moreover, the term “competing” presumes the existence of other comparable studies that affirmatively support USNA’s claims—yet none of the studies Bennett cites fit that description.

Bennett’s distortions don’t stop there. He goes on to assert that the study “at least partly supports Defendants’ national security interest” because it found a “positive correlation between racial diversity and retention of Black male enlisted servicemembers.”²¹¹ Like Lyall before him, he either failed to read the study carefully or lacked the statistical proficiency to recognize its glaring limitations. As outlined in Section 3.2.3, everything about the study reeks of statistical noise.

For one, the supposed “positive” correlations Bennett cites were minuscule in magnitude and, crucially, generally failed to reach statistical significance at the conventional 95% threshold—meaning they were indistinguishable from zero. As noted in Section 3.2.3, these correlations were derived from a model with a massive Black sample size (N=23,494) and few control variables. In a dataset of this scale, a marginally significant result is far from compelling; if anything, it suggests random variation rather than a meaningful causal relationship. Even more damning, the same analysis found no corresponding effect for Hispanic servicemembers—despite USNA making identical arguments about the need for Hispanic representation in leadership. If racial representation truly drives retention, why would its effects appear in one group but be entirely absent in another?

Ultimately, the idea that this study “at least partly supports” the Defense’s claims is laughable. Even if the finding in question were statistically robust (which it isn’t), it would still be irrelevant to the issue at hand—officer retention. The entire justification for racial preferences at USNA is that increasing minority representation in the officer corps enhances officer retention and, eventually, minority representation among senior leadership. The enlisted ranks are an entirely separate issue, and their retention rates have no bearing on the constitutionality of USNA’s race-conscious admissions policies.

More importantly, Bennett applies a glaring double standard in how he treats the study’s findings. He readily cites a weak, statistically ambiguous correlation between racial diversity and Black enlisted retention as evidence that diversity improves retention. Yet, when confronted with the same study’s finding that greater Black representation among senior officers correlates with lower first-term Black officer retention, he downplays it rather than treating it with the same weight. If the former is strong enough to “at least partly” support USNA’s case, then by that same standard, the latter should “at least partly” undermine it. Instead of applying consistent scrutiny, Bennett inflates one result while brushing off the other—an approach that betrays his bias rather than a genuine engagement with the evidence.

Role-Model Theory in Disguise

This pattern of selective reasoning continues in Bennett’s discussion of recruitment, where he attempts to sidestep SFFA’s argument that the government’s justification for racial preferences amounts to a legally impermissible role-model rationale. According to SFFA, the logic underlying USNA’s position rests on the notion that minority recruits need to see officers who “look like them” to feel motivated to enlist—a claim the Supreme Court has consistently rejected. For instance, in Wygant, the Court explicitly rejected the idea that racial classifications could be used to ensure that students “see others like them,” a rationale nearly identical to the government’s justification for racial preferences in military admissions.²¹²

Rather than grappling with this precedent, Bennett simply asserts that DoD’s policy does not rely on a role-model theory. “Defendants evince no such theory,” he writes.²¹³ Instead, he claims, DoD’s goal is to ensure that recruits “see opportunities for success” in the military regardless of their background. But this is a distinction without a difference. The entire argument for racial diversity in the officer corps hinges on the assumption that increasing minority representation makes military service more attractive to potential minority recruits. If that is not a role-model theory, then what exactly is it?

Bennett’s attempt to reframe the argument also contradicts the testimony he relies on elsewhere. Throughout his opinion, he repeatedly cites defense witnesses and DoD officials who emphasize the importance of minority recruits being able to “see themselves” in the officer corps as a justification for racial preferences. For instance, he cites remarks by Secretary of the Navy Carlos Del Toro who stated that “we need a diverse force, so every child in America can see themselves wearing the uniform or working in our civilian ranks tomorrow.”²¹⁴ Likewise, he cites Deputy Assistant Secretary Miller, who testified that a “diverse officer corps, particularly in special forces units, enables individuals to envision themselves succeeding in the Navy and Marine Corps.”²¹⁵

These statements make it abundantly clear that the government’s rationale is, at its core, a role-model theory—even if Bennett refuses to call it that.

Domestic and International Legitimacy

As the final pillar of Judge Bennett’s deference to the government’s claimed compelling interest, he asserts that Defendants have “shown diversity in the officer corps furthers the domestic and international legitimacy of the Navy and Marine Corps, which in turn serves a compelling national security interest.”²¹⁶ While acknowledging that courts have generally rejected racial classifications justified solely by “subjective evidence” like the need to gain public confidence (e.g., Hayes, Christian, Wygant), he proceeds to carve out an exception by arguing that the military’s legitimacy is not a stand-alone justification but is instead intertwined with other national security interests—namely, recruitment, retention, cohesion, and lethality.²¹⁷

But as with the other justifications, this claim is entirely unsubstantiated, relying on anecdotal testimony, vague historical references, and unsupported assumptions rather than empirical evidence. Notably, neither USNA nor the government provides any quantifiable link between racial diversity in leadership and public confidence in the military. In fact, as discussed in Section 3.2.3, long-term survey data contradict this premise, showing no meaningful relationship between the racial composition of the officer corps and either minority enlistment propensity or overall public trust in the military.

With no supportive survey data to lean on, Bennett relies heavily on testimony from defense witnesses who simply declare that diversity strengthens the military’s legitimacy, as if saying it makes it so. For instance, he cites Captain Birch’s personal experience as a Black Navy SEAL Captain, claiming that his presence enhanced the legitimacy of his unit in Somalia and a U.S. Naval delegation in China.²¹⁸ He also again references Secretary of the Navy Carlos Del Toro’s claim that a diverse force is necessary so that “every child in America can see themselves” in military service. But none of this constitutes evidence that racial representation meaningfully impacts recruitment or institutional credibility. These are anecdotes, not data.

Of course, Bennett also ties the claim to “historical evidence that a lack of diversity in the officer corps decreased the legitimacy of the military during the Vietnam War.”²¹⁹ Yet this proposition, too, rests on assertion rather than actual research demonstrating that the officer corps diversity—or lack thereof—played a decisive role in shaping public perceptions of the military during Vietnam.

Here, as elsewhere, Bennett’s ruling does not merely defer to military judgment—it uncritically accepts a narrative crafted to justify racial preferences without meeting the constitutional burden of proof.

4.2.3. Conclusion of Law: Does racial/ethnic diversity advance a compelling national security interest?

If the reader held out hope that Judge Bennett would apply greater scrutiny in his final analysis of the ‘compelling interest’ test, what follows will bring severe disappointment. His reasoning in this section is indefensible, reflecting either a fundamental misunderstanding of how evidence and causality work or a deliberate attempt to sidestep the evidentiary burden required under strict scrutiny. Neither alternative is reassuring.

First, Bennett acknowledges that the Supreme Court requires institutions using race-based admissions to do so in a manner “sufficiently measurable to permit judicial [review]” under strict scrutiny—meaning they must quantify how much a compelling interest would suffer in the absence of racial classifications.²²⁰ Yet he attempts to carve out an exception for military academies by invoking Holder v. Humanitarian Law Project.²²¹ But Holder had nothing to do with racial classifications; it concerned restrictions on providing material support to foreign terrorist organizations. The Court’s deference in that case stemmed from national security concerns wholly unrelated to equal protection. Strict scrutiny of racial classifications, by contrast, requires a rigorous evidentiary showing—one that SFFA v. Harvard and Fisher make clear is not subject to such deference. Bennett’s reliance on Holder is not just a misreading of precedent—it is an attempt to circumvent strict scrutiny by misapplying an unrelated case.

Bennett also cites Trump v. Hawaii to argue that a “searching inquiry into the persuasiveness” of national security interests is inconsistent with judicial deference to the executive branch. But this reliance is entirely misplaced.²²² Trump dealt with executive authority over immigration policy—not racial classifications. The Court’s ruling there was based on the principle that national security decisions related to foreign policy warrant heightened deference. But as the Supreme Court has repeatedly made clear, racial classifications are subject to an exacting form of strict scrutiny. Bennett’s use of Trump suggests either a failure to grasp this distinction or a deliberate attempt to obscure it. Under equal protection doctrine, the government must provide specific, measurable evidence that racial preferences serve a compelling interest and that no race-neutral alternatives would suffice. Invoking Trump v. Hawaii to sidestep this requirement is legally indefensible.

By carving out this “national security” exception to strict scrutiny, Bennett essentially declares that USNA and the government do not need to provide empirical validation for their claim that racial preferences improve military effectiveness. Having argued that the government is entitled to a lower evidentiary burden, he notes that USNA has nonetheless “proven” its compelling interest is measurable.²²³ If the evidence were as strong as he suggests, there would be no need to dilute the standard. This contradiction exposes the weakness of the ruling—it simultaneously lowers the evidentiary bar while falsely claiming that USNA has cleared it.

So what “evidence” does Bennett offer to substantiate his claim that USNA’s racial preferences advance a compelling national security interest? As he acknowledges, under the Harvard precedent, the measurability requirement demands answers to two empirical criteria:
(1) how many fewer leaders would an institution produce without racial preferences?²²⁴
(2) how much poorer would the institution’s core function—education, in Harvard’s case—be without racial preferences?²²⁵

Bennett asserts that because USNA’s interest lies in national security rather than educational diversity, only the first criterion applies.²²⁶ But this is an artificial distinction—if racial diversity is essential to military readiness, the proper analogy would be to ask how much worse military performance would be without racial preferences. Recognizing that USNA cannot meet this empirical demand, Bennett conveniently limits the focus to the first criterion, which he claims USNA “can easily meet” in the context of its graduates.²²⁷

Bennett’s Misuse and Misunderstanding of Statistics

So what kind of “proof” does Bennett offer? Astonishingly, his answer is little more than a pair of raw data points.

To begin, Bennett presents a table—reproduced verbatim below—comparing the number of officers from different racial groups who commissioned from USNA in 2001 and 2023 (which he erroneously labels as 2024).²²⁸ This methodological approach is deeply flawed and profoundly misleading on multiple levels. Most glaringly, this table does not satisfy the Supreme Court’s measurability requirement in any meaningful sense. No one disputes that USNA can count the number of minority officers among its graduates. But the core question is not whether USNA can tally racial demographics—it is how many fewer officers it would have commissioned without racial preferences. Since USNA has never conducted a counterfactual analysis to answer this question, Bennett settles for something far weaker: a crude comparison of two arbitrary time points, which he misleadingly presents as proof of racial preferences’ causal impact.

Table 1. Judge Bennett’s table comparing the number of Navy officers who graduated in 2001 vs. 2024 by race/ethnicity.

Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Findings of Fact and Conclusions of Law, at 150 (D. Md. Dec. 6, 2024).

Beyond the fact that these data do not allow for any causal inference, Bennett’s decision to report only absolute numbers—and to do so for just two time points—raises serious red flags. This suggests either a deliberate attempt to mislead or a fundamental misunderstanding of statistical analysis.

First, absolute numbers alone tell us nothing about whether minority representation in the officer corps actually increased in a meaningful way. USNA’s total number of graduates fluctuates from year to year, meaning that a proper analysis would normalize these figures as percentages rather than raw counts. Yet Bennett fails to do this, leaving the reader with no way to determine whether these changes reflect genuine increases in minority representation or are merely artifacts of class-size fluctuations.

Second, the choice to report only two time points (2001 and 2023) is itself highly suspect. While it is true that USNA only provided data beginning in 2001, Bennett’s decision to present only the earliest and latest available years creates the illusion of steady progress, ignoring the likely fluctuations that occurred in the intervening years.

But Bennett doesn’t stop there. He applies the same flawed logic to the DoD’s retention data, which I previously visualized and discussed in Section 3.2.3. Again, he treats crude comparisons between the earliest and latest available data points as definitive proof that race-conscious admissions have improved officer retention. Bennett claims that 10-year retention rates for American Indian or Alaskan Native, Asian, Black, and White officers have increased between 2001 and 2024 and that 15-year retention rates have risen for Asian and Black officers. Yet, as noted earlier, these data appear to be highly volatile, which is precisely why the USNA witness who presented them (Stephanie Miller) did so by collapsing all non-white racial groups into a single ‘minority’ category.

Even more revealing is Bennett’s inclusion of White officers in his retention analysis. He asserts that their 10-year retention rates have also increased since 2001. But if racial preferences were driving retention gains, why would a group that does not benefit from them exhibit the same trends? This directly undermines the premise that diversity policies, rather than broader institutional or economic factors, are responsible for improved retention. Bennett never addresses this contradiction, either because he does not recognize it or because his argument would unravel if he did.

But even if one were to accept the unsubstantiated causal claim that USNA’s racial preferences have increased minority representation and retention among its graduates, Bennett’s analysis is fundamentally incomplete. He never considers their impact on the broader Navy officer pipeline. Recall that USNA commissions just 20% of all naval officers, meaning that any increase in minority representation at USNA affects only a fraction of the officer corps. If racial diversity is truly a compelling national security interest, then the relevant question isn’t whether USNA has increased its minority graduate numbers, but whether this translates into a meaningful shift in the overall composition of the officer corps. On this, Bennett provides no evidence whatsoever.

Nor does he consider whether USNA’s racial preferences have actually diversified the most selective—and least diverse—warfare communities, where diversity is supposedly needed most. As detailed in Section 3.3, Black and Hispanic graduates remain heavily underrepresented in communities such as Submarine Warfare, Aviation, and Special Warfare—precisely the pipelines that produce the Navy’s senior leadership. If race-conscious admissions at USNA were achieving the Navy’s stated diversity objectives, we would expect to see minority officers gaining greater representation in these elite communities. Instead, the evidence suggests the opposite: USNA’s racial preferences have done little to alter the demographic composition of the most selective warfare communities, which in turn limits their potential impact on the diversity of the senior leadership ranks.

Setting aside these many issues, another question remains: Is USNA’s use of race in admissions narrowly tailored to achieving its stated goal of a more diverse officer corps? As the next section will show, Bennett’s deference does not stop at the compelling interest test—it extends to a wholesale acceptance of USNA’s admissions practices, no matter how tenuous the justification.

4.4. Is USNA’s Consideration of Race Narrowly Tailored?

Even if the government can demonstrate that race-conscious admissions serve a compelling national interest, the Supreme Court has been unequivocal: such policies must also meet the rigorous standard of narrow tailoring. This requires more than vague assertions—it demands concrete evidence that the policy is both necessary and limited in scope. To satisfy this requirement, institutions must adhere to three fundamental principles. First, race may only be used as a “plus” factor to achieve a compelling interest; it cannot impose a penalty or systematic disadvantage on other applicants. Second, institutions must seriously consider and exhaust race-neutral alternatives before resorting to explicit racial preferences. Third, and perhaps most crucially, the use of race must be temporary—it cannot persist indefinitely, and institutions employing such policies must conduct periodic reviews to ensure that racial preferences are being phased out over time.

Yet, as will become clear, Judge Bennett’s analysis of USNA’s compliance with these standards consists largely of an uncritical endorsement of Gurrea’s speculative and misleading critiques of Arcidiacono’s findings. Rather than engaging with the evidence on its merits, Bennett selectively defers to USNA’s arguments while dismissing Arcidiacono’s analysis outright. In doing so, he sidesteps the exacting scrutiny that narrow tailoring demands, setting the stage for his conclusion that USNA’s use of race passes constitutional muster.

4.4.1 Does USNA Use Race as a Negative?

Arcidiacono’s analysis documented stark racial disparities in USNA’s admission outcomes, showing that White applicants had only a fraction of the probability of admission compared to Black and other minority applicants with the same qualifications. If taken seriously, his findings constitute compelling evidence that USNA’s use of race extends beyond a mere “plus” factor for underrepresented minorities—it imposes a significant penalty on more qualified White applicants, violating the Supreme Court’s clear prohibition against using race as a negative factor in admissions.

Judge Bennett, however, does not take Arcidiacono’s findings seriously. His treatment is not just one-sided—it is a wholesale regurgitation of Gurrea’s talking points, with little effort to engage Arcidiacono’s evidence on its own terms. The imbalance is striking: Bennett devotes twice as many pages (14) to summarizing and repeating Gurrea’s critiques as he does to engaging Arcidiacono’s findings. Even more troubling, he entirely ignores the subsequent data-driven rebuttal Arcidiacono submitted, which directly responded to and refuted Gurrea’s claims. This omission raises a fundamental question: did Bennett even read Arcidiacono’s full reports, or did he simply accept USNA’s rebuttal at face value?

A key claim Bennett borrows from Gurrea is that omitted variable bias (OVB) inflates Arcidiacono’s estimates. As outlined in Section 3.1.1, this argument asserts that Arcidiacono’s models fail to account for unobserved factors—such as personal statements, teacher recommendations, and life experiences—that USNA allegedly considers in admissions. Because these omitted factors may correlate with race, Gurrea speculates that the racial preference coefficients in Arcidiacono’s models could be inflated. Yet this claim is entirely speculative—Gurrea never actually tests whether including these variables would diminish the estimated racial preferences. More critically, every time Arcidiacono introduces additional control variables—proxies for these omitted factors—the magnitude of the racial preferences grows rather than shrinks. If omitted variables were truly driving the observed disparities, one would expect the estimated racial preferences to decrease, not increase, as more controls are added.

Bennett entirely ignores this empirical reality and makes no mention of Arcidiacono’s analyses demonstrating the implausibility of the OVB argument. Even worse, he misrepresents Arcidiacono’s own testimony on the matter. For instance, he seizes on Arcidiacono’s response to a purely hypothetical question posed by USNA’s lead attorney—who asked whether, if unobserved factors were “massive and perfectly correlated with race,” his racial preference estimates would disappear—as if it were an admission that his findings lack reliability.²²⁹

But the Judge either misunderstands or has a selective reading of Arcidiacono’s response.²³⁰ His point was not that such omitted variables actually or even plausibly exist, but rather that for them to fully explain the observed racial disparities, they would need to meet two implausible conditions simultaneously. First, they would all have to be strongly and positively correlated with admission, meaning that these unobserved factors—such as personal statements or life experiences—must consistently and significantly increase an applicant’s likelihood of being admitted. Second, they would all have to be either uncorrelated or strongly and negatively correlated with all observable factors in the model, including being White, such that they systematically disadvantage White applicants while favoring Black and other minority applicants in a way that perfectly offsets the documented racial preferences.

This is an extraordinary assumption, one that neither Gurrea nor Bennett provide any evidence for. In fact, when additional control variables—proxies for these unobserved factors—were introduced into Arcidiacono’s models, the estimated racial preferences grew rather than shrank. If omitted variables were truly responsible for the disparities, we would expect the opposite effect: the racial preference estimates should diminish as more relevant controls are added. Yet rather than engaging with this evidence, Bennett simply dismisses Arcidiacono’s findings as “unreliable,” uncritically accepting Gurrea’s speculation as fact.²³¹

Predictably, Bennett also fully embraces Gurrea’s erroneous claim that logistic regression models are inappropriate for analyzing USNA’s admissions data—a claim that, if accepted seriously, would make USNA’s use of race immune to empirical scrutiny.²³² Yet, as discussed in Section 3.1.3, this critique is both incorrect and unprecedented in affirmative action litigation.

Even if we momentarily entertained Gurrea’s concerns about interdependence, they are wildly overstated and directly accounted for in Arcidiacono’s models. His analysis explicitly controls for the very factors Gurrea claims introduce interdependence—slate characteristics, nomination types, and congressional district representation—thereby addressing any such concerns directly. It also incorporates class-year fixed effects, ensuring that variations in admissions standards across different application cycles are controlled for. Bennett, however, ignores all of this, endorsing Gurrea’s argument without seriously engaging the body of contrary evidence.

Most revealing is Bennett’s failure to recognize the broader implication of Gurrea’s argument: that rolling admissions and fixed class sizes somehow prevent meaningful statistical analysis. If this were true, Harvard and UNC—both of which use zero-sum admissions, rolling evaluations, and fixed class sizes—would have been immune from empirical scrutiny in SFFA v. Harvard and SFFA v. UNC. Even Harvard’s own expert, David Card—a Nobel Prize-winning economist—never challenged the validity of logistic regression in that context. Bennett’s failure to acknowledge this precedent is telling. If logistic regression was valid for Harvard and UNC, why is it suddenly invalid for USNA? The answer is clear: Gurrea’s critique is a last-ditch effort to shield USNA from empirical scrutiny. If rolling admissions truly invalidated econometric modeling, this argument would have surfaced in every prior affirmative action case. The fact that it didn’t proves that this is not a legitimate methodological concern but a contrived objection designed to insulate USNA from statistical and legal analysis.

Beyond the methodological distortions, the practical consequences of USNA’s racial preferences is undeniable: White applicants with identical credentials are admitted at significantly lower rates than their Black counterparts. Yet, rather than confronting this reality, Bennett attempts to minimize the effect by uncritically adopting Gurrea’s claim that eliminating racial preferences would “only” increase White admission rates by 4.2%.²³³ Yet where does this number come from? Not from Gurrea’s written report. Instead, it surfaced for the first time during his testimony—without any accompanying explanation of its derivation.²³⁴ More troubling still, Bennett does nothing to verify its accuracy. Instead, he blindly accepts it at face value—while also misquoting it as “4.02%”—despite its dubious origins.

More importantly, the 4.2% figure does not reflect Arcidiacono’s actual estimates. As discussed in Section 2.7, removing racial preferences from both direct and indirect (prep school) admissions would increase the White share of overall admits by 9.4 percentage points, translating to 649 additional White admits over the 2023–2027 admission cycles.

Even if racial preferences remained in the prep school pipeline and were only eliminated in direct USNA admissions, the White admit share would still increase by 6.3 percentage points—far exceeding Gurrea’s 4.2% claim—resulting in 438 additional White admits. That Bennett entirely ignores these figures and instead relies on an unsubstantiated number from Gurrea speaks volumes about the uncritical deference he affords USNA’s defense.

Finally, even if we were to accept the 4.2% figure for argument’s sake, Bennett’s framing of it as “only” a small effect is deeply misleading. A 4.2% increase in the White admit rate still translates to hundreds of additional White admits—a fact that is anything but trivial. His reasoning suggests that racial discrimination is permissible so long as its impact remains below an undefined threshold. But Supreme Court precedent is clear: any use of race in admissions must be narrowly tailored and cannot impose undue burdens on disfavored groups. The evidence shows that racial preferences at USNA impose significant costs on White applicants, undermining the notion that race is used as nothing more than a “plus” factor.

Taken together, these distortions and omissions reveal a fundamental flaw in Bennett’s ruling: what he presents as strict scrutiny is, in reality, deference disguised as analysis. By downplaying or outright ignoring key aspects of Arcidiacono’s analysis, Bennett sidesteps the uncomfortable but unavoidable reality: USNA’s racial preferences do not operate as a mere “plus” factor—they impose a significant disadvantage on more qualified White applicants. That, on its face, is incompatible with the principles of narrow tailoring as articulated by the Supreme Court.

4.4.2 Has USNA Exhausted Race-Neutral Alternatives?

Judge Bennett contends that USNA has undertaken “serious, good faith considerations of race-neutral alternatives,” yet his analysis amounts to little more than listing USNA’s outreach and recruitment efforts—treating these as sufficient substitutes for meaningful, race-neutral reforms to the admissions process itself.²³⁵ But the core legal question is not whether USNA engages in diversity outreach; it is whether the Academy can achieve its stated diversity goals via race-neutral admissions policies.

On this count, Bennett asserts that USNA has “incorporated numerous race-neutral factors to its WPM and RABs to award points to candidates based on race-neutral characteristics, including socioeconomic status.”²³⁶ Deferring to the testimony of USNA witnesses, he further claims that the Academy “also considers and highly values the same characteristics” beyond the formal point system for socioeconomically disadvantaged candidates.²³⁷

But if this were true, we would expect indicators of low socioeconomic status (SES) to have significant, positive, and independent effects on candidates’ probabilities of admission. As Figure 4.4A below illustrates, Arcidiacono’s admissions model shows otherwise. Whereas the coefficients for racial groups are large and statistically significant, those for indicators of low SES are indistinguishable from zero—suggesting that socioeconomic disadvantage confers little to no real benefit in USNA admissions. Worse still, private high school attendance and the share of a student’s high school class that matriculates to four-year colleges both exert strong, positive effects on admission. The latter effect is particularly striking: The latter effect is particularly striking: it rivals the size of the Hispanic racial preference, indicating that USNA is more inclined to admit students from privileged, college-bound environments than those who have overcome hardship.

Figure 4.4A. Race vs. Socioeconomic Status in USNA Admissions: Estimated Logit Coefficients

Note. Bars represent logit coefficients, with error bars indicating 95% confidence intervals. Bars and error bars that cross the dashed vertical line are not statistically significant at the 95% confidence threshold. Estimates derived from Arcidiacono’s preferred USNA admissions model (Model 6). Sample size is 12,300.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rebuttal Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 222, Table D.82R, at D70–D74.

If USNA truly valued socioeconomic diversity on par with racial diversity, its admissions patterns would reflect that priority. Instead, as Arcidiacono’s model shows, race—not SES—is the decisive factor. This finding is corroborated by USNA’s own Dean of Admissions, Stephen Latta, who acknowledged in his pre-trial declaration that SES-based considerations failed to significantly increase racial diversity because “many non-diverse applicants also benefit from such considerations.”²³⁸ In plain terms: SES was deprioritized because it benefited too many low-income White applicants.

This admission is squarely at odds with Bennett’s assertion that USNA pursues racial diversity not for its own sake but as a means of enhancing the cognitive and experiential diversity necessary for national security.²³⁹ If that were true, low SES should be at least as significant an admissions factor as race—especially given its broader connections to the many other forms of diversity (e.g. geographic, experiential) that USNA purports to value.²⁴⁰

Instead, USNA marginalizes SES, revealing that its goal is not functional diversity but racial balancing.

Moreover, simulations conducted by Arcidiacono for SFFA’s race-neutral alternatives expert, Richard Kahlenberg, demonstrate that a race-neutral system prioritizing SES would still yield significant racial diversity while dramatically increasing the share of economically disadvantaged students at USNA. Figure 4.4B presents results from one such simulation, which eliminates racial preferences and instead grants a 1.5x admissions boost to students from socioeconomically disadvantaged families, schools, and neighborhoods—the same advantage currently given to Black applicants under race-conscious admissions. The model also removes admissions advantages for students from wealthy schools and reserves non-athlete prep school slots for low-income applicants.

Under this SES-based race-neutral system, the overall share of underrepresented minority students (Black, Hispanic, and Native American applicants) reaches 36.5%—a figure significantly higher than the 31.0% share that would result from a purely race-neutral system that eliminates racial preferences without adding SES-based boosts. Hispanic representation actually increases beyond the status quo (from 12.6% to 13.4%), while Black representation remains substantial (8.7% vs. 6.9% under a purely race-neutral model). Meanwhile, the percentage of admitted students from families earning below $80,000 more than doubles, from 18.8% to 40%. These findings demonstrate that race-neutral alternatives can produce both substantial racial diversity and dramatically improved socioeconomic diversity—all without resorting to constitutionally suspect classifications.

Figure 4.4B. The Impact of Race-Neutral SES-Based Admissions on Racial and Socioeconomic Diversity at USNA

Note. This simulation is based on modified coefficients from Arcidiacono’s preferred USNA admissions model (Model 6). Teal bars represent a subgroup’s share of USNA admits under the current race-conscious admissions system. Blue bars represent a subgroup’s predicted share of USNA admits when all non-White racial group coefficients are set to zero (i.e., racial preferences are removed). Gray bars represent a subgroup’s predicted share of USNA admits under a race-neutral SES-based model, in which non-White racial group coefficients are set to zero, the coefficient on ‘Income < $80k’ and ‘First Generation College’ is increased to 1.5 times the status quo coefficient on African American applicants, the coefficient on ‘Percent Free/Reduced Lunch’ is set to 0.75 times the African American coefficient, and the coefficients on ‘Private HS’ and ‘Average Zip Code Salary’ are set to negative 0.75 times the African American coefficient. In the NAPS admission model, the coefficient on ‘Income < $80k’ is set to an exceedingly high value, effectively making low-income status determinative of admission for non-Blue Chip athletes.
Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Richard D. Kahlenberg, Plaintiff's Exhibit 219, Tables 1 and 7, Appendix D, at 6 and 9.

Bennett, however, is not persuaded. He asserts that “the Academy’s race-conscious admissions policy has a significant impact on the racial diversity of its class” and argues that any race-neutral alternative must “allow the Government to achieve the benefits that it derives from its current degree of diversity within a given class year, while also being practicable, affordable, and not requiring a material decline in academic quality or any of the other measures of excellence valued by the Naval Academy.”²⁴¹

This reasoning is flawed on multiple levels. First, Bennett’s assertion that racial preferences have a “significant impact” is fundamentally incompatible with USNA’s claim that race is merely a limited “plus” factor. If race truly plays a minor role, eliminating it should have little effect on the composition of admitted students. But if it often proves decisive, then USNA’s admissions process violates the Supreme Court’s requirement that race be used only in a narrowly tailored way. Of course, as noted in Section 1.6, USNA has never attempted to model its admitted class under a race-neutral policy. Despite this, Bennett concludes that such a policy would make its diversity goals unattainable. His ruling thus allows race to be simultaneously too modest to measure and too critical to abandon, depending on what best supports USNA’s case.

Second, Bennett sets an impossibly high bar for race-neutral alternatives. Rather than assessing whether they could achieve sufficient diversity to serve USNA’s mission, he demands that they precisely replicate the racial outcomes of race-based admissions.²⁴² But the Supreme Court has never required perfect racial matching as the benchmark for race-neutral alternatives. The relevant legal question is whether such alternatives could achieve USNA’s stated diversity goals without resorting to racial classifications—not whether they can mechanically reproduce the same racial percentages. By requiring race-neutral policies to produce an identical racial composition to race-based admissions, Bennett sets an impossible standard—one that effectively guarantees the continued use of racial preferences regardless of viable alternatives.

Third, Bennett’s reasoning is internally inconsistent. Earlier in his ruling, he adopts Gurrea’s claim that Arcidiacono’s estimates exaggerate the role of race in USNA admissions. Yet when justifying racial preferences, he reverses course—citing those same estimates to claim that eliminating race-conscious admissions would drastically reduce Black and Hispanic admits.²⁴³ If Arcidiacono’s numbers are too flawed to show racial discrimination, they are also too flawed to justify racial preferences. Bennett cannot have it both ways.

Fourth, and perhaps most damning, neither USNA nor the government has ever defined what specific “level of diversity” is essential for national security—nor have they conducted any empirical study demonstrating that racial preferences are both effective and the only viable means of achieving it. Under cross-examination, both USNA Dean of Admissions Stephen Latta and Under Secretary of Defense Ashish Vazirani admitted that the DoD has never quantified how much racial representation is required to foster unit trust, improve cohesion, or encourage minority recruits to see themselves as future leaders.²⁴⁴ Nor has it provided any rationale for whether the Navy’s officer corps should mirror the general population’s racial composition—or why failing to do so would diminish military effectiveness.

Finally, Bennett’s dismissal of race-neutral alternatives on academic grounds is unpersuasive.²⁴⁵ The SES-based model outlined above yields only marginal declines in total SAT scores (1344 vs. 1323 points) and high school class rank (585 vs. 575.1)—differences that are negligible compared to the far larger academic disparities that already exist under race-based admissions. Yet Bennett ignores this inconsistency, raising concerns about academic standards only when race-neutral policies are on the table. His ruling thus exposes a clear double standard: racial preferences are excused despite their impact on academic credentials, while race-neutral alternatives that would dramatically increase socioeconomic diversity are rejected on the same grounds.

In sum, Bennett’s approach to race-neutral alternatives is not just deferential—it is designed to entrench racial preferences. By demanding that race-neutral policies precisely replicate the racial outcomes of race-based admissions, he imposes an impossible standard, ensuring their inevitable failure. Worse, he shifts his evidentiary standards at will—dismissing data that hurts USNA’s case while embracing it when it supports the status quo. The result is not judicial review, but judicial abdication: a ruling that exempts USNA from constitutional scrutiny and allows racial preferences to continue indefinitely without real oversight.

4.4.3 Is USNA’s Use of Race Time-Limited?

Even if one accepts Judge Bennett’s conclusions regarding USNA’s compelling interest in racial diversity and the purported inadequacy of race-neutral alternatives, a final crucial question remains: Is USNA’s use of race in admissions genuinely time-limited, as strict scrutiny requires?

The Supreme Court has repeatedly held that governmental use of racial classifications cannot be indefinite. In Grutter v. Bollinger (2003), the Court explicitly ruled that all race-conscious admissions policies must have a “logical end point” and be subject to periodic review to ensure they do not persist beyond necessity. The Court reaffirmed this principle in Harvard, striking down race-conscious admissions at Harvard and UNC in part because neither institution provided a clear stopping point for their racial preferences.

Judge Bennett acknowledges this precedent but proceeds to carve out yet another “national security” exception, arguing that strict scrutiny’s “logical end point” requirement has been applied “largely in the context of civilian institutions” and that USNA is “distinct from a civilian college” because “strict scrutiny is context-dependent.”²⁴⁶ By drawing this distinction, Bennett effectively exempts USNA from the constitutional constraints that govern civilian universities, allowing its use of racial preferences to continue indefinitely.

Bennett’s National Security Justification and USNA’s Contradictory End-Point

Bennett claims that because “there is no end date on national security,” USNA is not obligated to specify a precise timeline for discontinuing its consideration of race.²⁴⁷ Nevertheless, he asserts that USNA satisfies the time-limited requirement by promising to abandon race-conscious admissions once its incoming classes of midshipmen allow it to produce a Navy and Marine officer corps that better “represents racial and ethnic diversity among enlisted servicemembers and the American population.”²⁴⁸

But there are several fundamental problems with this logic. First, USNA’s stated ‘end-point’ extends beyond merely creating a representative officer corps. USNA’s own Dean of Admissions testified—on both days 2 and 3 of the trial—that the Academy’s goal is to achieve and maintain a level of racial/ethnic diversity within USNA’s student body that is “comparable to the ethnic and racial diversity of the general population”.²⁴⁹ This, of course, is textbook racial balancing, which the Supreme Court has repeatedly and unequivocally rejected as unconstitutional (Harvard, Parents Involved, Fisher, Grutter).

Note that this was not an offhand remark. It was repeated, deliberate, and reaffirmed in the Defendants’ Post-Trial Proposed Findings of Fact and Conclusions of Law, which state:

“The Naval Academy anticipates that, once it both achieves and maintains the racial and ethnic diversity of its student body at a level comparable to the ethnic and racial diversity of the general population, it will no longer need to continue its limited consideration of race in the admissions process.”²⁵⁰

Despite this explicit testimony and formal policy statement, Bennett asserts—without irony—that USNA “does not engage in racial balancing” because it allegedly applies its admissions policy “without reference to the ethnic or racial makeup of its midshipmen class as compared to that of the United States.”²⁵¹ His primary evidence for this claim? That both racial minority enrollment numbers and offers to racial minority candidates fluctuate year to year.²⁵²

Fluctuations in Enrollment Do Not Disprove Racial Balancing

This “fluctuations” argument is profoundly misguided. For starters, Bennett once again (see Section 4.2.3) relies on raw counts of minority matriculants rather than examining their share of the overall class. Fluctuations in total class size naturally produce corresponding fluctuations in absolute numbers without revealing anything about racial balancing. A meaningful analysis would look at the proportion of each racial group across cohorts—not raw totals.

But even fluctuations in shares would not necessarily disprove racial balancing. There is no reason to expect perfect stability under a racial balancing regime. Harvard made this same argument, pointing to fluctuations in racial group shares to claim it was not engaged in racial balancing—yet the Supreme Court rejected this reasoning, deeming the relatively narrow fluctuations in group shares indicative of “outright racial balancing.”

The pattern in USNA’s data is strikingly similar. As Figure 4.4C demonstrates, individual racial minority group shares of USNA’s classes have either trended upward or fluctuated within a relatively narrow band, while the White share has trended downward. Moreover, increases in the overall non-White share and decreases in the White share have occurred far too quickly to be explained by national demographic trends, which evolve gradually. For the USNA Classes of 2016 to 2020 (roughly corresponding to the 2011–2015 admission cycles), the non-White share hovered around 35–36%. By the Class of 2021, it had jumped to 38.4%, surpassed 41% by 2023, and reached 46% by the Class of 2027. Over the same period, the White share declined from 66% in 2017 to just 53.9% in 2027—a nearly 13 percentage point drop.

Figure 4.4C. Racial Composition of USNA Classes (2016–2027)

Note. Lines represent racial/ethnic group shares of USNA’s incoming classes. Dashed lines represent shares for those identifying as a member of a listed racial group in conjunction with other racial groups. In left panel, gray line represents the share of USNA matriculants who did not identify exclusively as non-Hispanic White. The right panel excludes this line and the line for non-Hispanic Whites to provide a clearer view of trends among individual racial/ethnic minority groups.
Source: USNA Class Snapshots (2016–2027).

The magnitude and pace of these changes over a short window suggest a deliberate policy shift rather than organic demographic trends. While one could reasonably question whether these data single-handedly “prove” an attempt at racial balancing, it would be untenable to claim—as Bennett does—that they prove its absence.

Bennett’s Census Comparison: Flawed Premises and Misleading Benchmarks

To be sure, Bennett doesn’t rest his case on the “fluctuations” alone. He additionally notes—and provides tables showing (reproduced below)—that the racial group shares of USNA’s Class of 2023 do not match and generally fall short of their shares in the wider population. For Bennett, these data suggest that “Defendants are not engaging in racial balancing, or, if they are, they are failing to produce anything close to racially balanced classes of midshipmen.”²⁵³

This argument too is deeply misconceived. First, it presumes that racial balancing is only unconstitutional if it succeeds. But partial or failed attempts at racial balancing do not make an admissions system any less unconstitutional. A school could be actively engaging in racial preferences while failing to meet its target racial composition—just as a quota system remains unlawful even if it doesn’t always produce a perfectly proportional outcome.

Table 2. Judge Bennett’s Comparison of USNA Class of 2023 Racial Composition to 2023 U.S. Census Population Shares

Source: Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Findings of Fact and Conclusions of Law, at 159.

Further undermining Bennett’s argument is his choice of benchmark. USNA’s Class of 2023 corresponds to the 2018 admissions cycle, meaning that comparing it against 2023 Census data is inappropriate. A valid analysis would not use general Census data but rather population benchmarks of Americans who are actually eligible for USNA admission, such as 17–22-year-olds who are unmarried and physically fit. In any case, the relevant question is not whether the USNA student body perfectly mirrors the entire U.S. population, but whether racial preferences are being used to engineer a specific racial composition under the guise of national security.

Even if one were to accept Bennett’s comparison at face value, his conclusion remains unconvincing. As discussed in Section 2.2 regarding Arcidiacono’s decile analysis, racial groups differ significantly in academic preparedness. Achieving proportional representation across all groups would require USNA to further lower its already reduced admissions standards for certain minority groups. Doing so would come with substantial risks: admitting even less qualified candidates—for instance, those with combined SAT scores averaging around 1000—would inevitably increase attrition, reduce graduation rates, and undermine the very national security goals USNA claims to pursue. This dynamic also explains why proportional representation has already been met and even exceeded for Asian applicants, who generally present stronger academic credentials. Ultimately, Bennett fails to acknowledge that without racial balancing, the representation gaps for certain minority groups would likely be even larger.

Racial Preferences Without a Clear End-Point

Bennett’s effort to disclaim racial balancing at the student level serves a clear purpose: he seeks to reframe USNA’s race-conscious admissions not as an attempt to mirror the broader US population, but as a pipeline to produce a more representative officer corps. This distinction is central to his reasoning. Bennett appears to accept that racial balancing in admissions would be constitutionally indefensible—but treats racial balancing in officer composition as justified on national security grounds. He even concedes that this objective “comes close to racial balancing,” yet allows it to stand because it is ostensibly directed at a compelling government interest rather than imposed mechanically at the admissions stage.²⁵⁴

But even if one accepts Bennett’s preferred characterization—namely, that USNA’s goal is “creating an officer corps that represents the people it protects and the servicemembers it leads”—that goal must still have a clear and measurable end point, as strict scrutiny requires. Yet Bennett tries to have it both ways. On the one hand, he insists that “there is no end date on national security.” On the other, he claims that USNA’s race-conscious admissions will terminate once “the incoming classes of midshipmen enable Defendants to develop a Navy and Marine officer corps that better represents racial and ethnic diversity among enlisted servicemembers and the American population.”²⁵⁵

This supposed end point, however, is anything but concrete. Neither Bennett nor USNA defines what “better represents” actually means, or what data would demonstrate that the objective has been met. Without measurable benchmarks or criteria for success, this “goal” becomes an abstraction—an ever-receding horizon that allows racial preferences to persist indefinitely under the open-ended assertion that “more diversity” is always needed.

Even if this goal were defined, a deeper problem remains: it is not merely to achieve a certain level of diversity, but to maintain it.²⁵⁶ This renders the “end point” a moving target. Even if a particular racial group reaches proportional representation at one moment in time, subsequent demographic shifts, changes in enlistment patterns, or broader social trends could be used to justify the continued—or reactivated—use of racial preferences whenever those numbers begin to drift. This directly conflicts with the Supreme Court’s holding in Harvard, which rejected open-ended justifications for race-conscious admissions and made clear that such policies cannot persist indefinitely.

The problem is further deepened by the fact that USNA’s stated interest does not stop at the officer corps level. As Bennett’s ruling makes clear, the government justifies racial preferences not merely to diversify the pool of commissioned officers, but to address racial disparities within the chain of command.²⁵⁷ For example, although Asian Americans are proportionally represented—or even overrepresented—among midshipmen and junior officers, the DoD claims that preferences remain necessary because Asians are underrepresented at senior ranks, such as O-6 (Captain) and above. This logic ensures that racial preferences will not only persist, but evolve into a perpetual mandate—subject to constant recalibration whenever disparities emerge at any level of leadership.

In the end, this is not a “time-limited” policy—it is a self-renewing cycle in which race remains embedded in admissions, commissioning, and promotion processes indefinitely. Bennett’s reasoning rests on shifting goalposts: he dismisses evidence of racial balancing at the student level, while defending racial balancing at the officer level as a national security imperative. Yet in either case, his framework sanctions the permanent use of racial preferences under the guise of a temporary measure. By embracing an undefined and endlessly flexible “end point,” Bennett does not apply strict scrutiny—he nullifies it.

5. Policy Options

SFFA vs. USNA began under the Biden administration and reached its district court conclusion in December 2024—just weeks before President Trump returned to office. Since then, the Trump administration has moved swiftly to dismantle DEI initiatives and race-based preferences across the federal government. Through a series of executive orders and directives, the administration has prohibited racial and sex-based considerations in hiring, contracting, and education—including admissions at the military service academies.²⁵⁸

One such order, EO 14173 (“Ending Illegal Discrimination and Restoring Merit-Based Opportunity”), rescinded President Lyndon B. Johnson’s 1965 EO 11246, which had long served as a foundation for affirmative action in federal contracting. Another, EO 14185 (“Restoring America’s Fighting Force”), directed the Department of Defense to eliminate all race- or sex-based goals in admissions, staffing, and training. In compliance with these directives, the U.S. Naval Academy formally eliminated race-conscious admissions on February 14, 2025—a change confirmed in Senate testimony and DOJ filings.²⁵⁹ Meanwhile, similar changes are under review at West Point and the US Air Force Academy (USAFA), though the policy remains in effect at these service academies as of this writing.²⁶⁰

These developments represent a major policy shift—but they are not self-enforcing, nor are they guaranteed to last. Without stronger institutional safeguards, future administrations could reinstate racial preferences with little resistance. Moreover, Judge Bennett’s ruling in SFFA v. USNA, which upheld race-based admissions, remains on the books and could be cited by future courts or agencies unless formally vacated. A more permanent solution is needed.

This section outlines a multi-pronged strategy—judicial, legislative, and executive—to secure a lasting end to race-based admissions at the service academies. It focuses not only on eliminating current preferences but on raising the political, legal, and institutional costs of reinstating them in the future.

5.1 Judicial Strategy

Despite recent developments in SFFA vs. USNA, litigation remains a crucial front in the effort to permanently eliminate race-based admissions at the service academies. The most promising opportunities lie in the ongoing cases against the U.S. Military Academy at West Point and the U.S. Air Force Academy (USAFA).

Unlike USNA—which has formally adopted a race-neutral admissions policy—neither of the latter two service academies have formally changed their admissions policies as of this writing. In both cases, the government has cited President Trump’s executive orders banning race- and sex-based preferences across federal institutions, including the military, and has requested time to assess their implications.²⁶¹ In the West Point case, the court denied the government’s request to stay proceedings for 90 days, opting instead to extend discovery deadlines through the end of 2025.²⁶² The court also ordered the government to provide regular status updates—beginning in April 2025—on how it intends to implement the new policies. A settlement or mediation report is due in June, and a case management conference is scheduled for February 2026. These steps suggest the case is at least procedurally active and moving forward, though its future remains uncertain. The USAFA case is in earlier stages, with the government recently granted a 30-day extension to respond to the complaint.²⁶³

These cases now offer the clearest path to judicial review of the constitutional question left unresolved in SFFA v. Harvard: whether the federal service academies can lawfully consider race in admissions. If either case proceeds through the district and appellate levels, and is ultimately taken up by the Supreme Court, the resulting ruling could definitively prohibit such policies nationwide. That outcome would foreclose future attempts to reinstate race-based admissions through executive or administrative action alone.

However, there are significant risks. First, even if a case survives through the district and appellate levels, there is no guarantee the Supreme Court will grant certiorari—or, if it does, that it will issue a decisive ruling. In SFFA v. Harvard, the Court expressly declined to apply its holding to the military academies, citing their “potentially distinct interests” in Footnote 4. That caveat, authored by Chief Justice Roberts, leaves open the possibility that some justices—perhaps including Roberts and Justice Barrett—may favor granting special deference to the military. A fractured or equivocal ruling could leave room for narrow exceptions, delaying or undermining the prospect of a clear constitutional ban.²⁶⁴

Second, the litigation could become moot if the academies voluntarily revise their policies to align with the executive order. Such strategic concessions have already occurred at the Naval Academy and may be replicated at West Point and USAFA as litigation advances. To prevent this outcome, it is essential that courts remain skeptical of mootness claims based solely on informal or facially neutral policy changes. In FBI v. Fikre (2024), the Supreme Court clarified that government defendants face the same “formidable burden” as private parties when seeking to moot a case through voluntary cessation.²⁶⁵ The decision rejected any presumption of good faith for government actors and emphasized that claims of compliance must be backed by clear, structural guarantees against subsequent backsliding.

As SFFA argued in its litigation against the University of Texas at Austin, such guarantees require more than policy statements—they demand concrete action.²⁶⁶ Institutions must take technological and procedural steps to eliminate any possibility that race can factor into admissions decisions, including removing checkbox data from application files, firewalling access to aggregate racial statistics, and ensuring that no person involved in admissions has access to race-related information during the decision-making process. Without such safeguards, the risk of reversion remains, and courts should not consider the matter moot.

If, despite these efforts, all three service academy cases are deemed moot, attention must turn to vacating Judge Bennett’s district court ruling, which would stand as the only judicial opinion upholding race-based admissions in the military context. As argued, that ruling was unusually deferential—accepting speculative claims of military necessity without applying meaningful strict scrutiny. If left intact, it could serve as a blueprint for future attempts to reintroduce racial preferences under similar rationales, especially in times of political or institutional flux. Vacatur would eliminate that risk by stripping the ruling of any precedential force, ensuring it cannot be cited to justify such policies in the future. While the request for vacatur must be made by SFFA, the Trump Department of Justice could signal its support in court filings or settlement discussions. Doing so would reinforce the administration’s broader commitment to eliminating racial preferences not only as a matter of policy, but as a matter of law.

5.2 Legislative Action

A second avenue for ensuring the permanent abolition of race-based admissions at service academies is congressional legislation. Unlike executive orders, which can be reversed by a future administration, a statutory ban offers far greater durability. Repealing such a law would require new legislation or a successful constitutional challenge—both of which present significant hurdles.

Still, pass such a statute will not be easy in today’s closely divided and ideologically polarized Congress. As of this writing, Republicans hold narrow majorities in both chambers—218–215 in the House and 53–47 in the Senate. A standalone bill explicitly banning racial preferences at the service academies would likely face strong opposition from Democrats and possibly some moderate Republicans. Overcoming a Senate filibuster would require either 60 votes or strategic use of budget reconciliation, assuming the measure can be framed to satisfy procedural rules. Given the historical pattern of midterm losses for the president’s party, Republicans must act quickly; unified control of Congress beyond 2026 is far from assured.

Success on this front will likely depend on active leadership from the Trump administration. By formally proposing a legislative ban—or publicly endorsing it and rallying congressional allies—the administration can help overcome political resistance and shape the terms of debate. Presidential support would also signal that this is not a partisan wedge issue but a national security reform aimed at restoring meritocratic standards in officer selection.

An additional and perhaps more politically viable strategy would be to embed the prohibition within a broader, must-pass bill—most notably, the National Defense Authorization Act (NDAA). This annual legislation governs defense funding and is rarely allowed to fail. Including a ban on race-based admissions in the NDAA would force opponents into the politically fraught position of opposing military funding in order to preserve affirmative action—an unpopular stance with the general public. Framing the measure as a merit-based military reform, rather than a front in the culture war, would also make it more palatable to centrist lawmakers.

That said, embedding the measure in the NDAA is not without risk. Opponents may attempt to strip it out during committee negotiations or floor debate, arguing that Congress should not use defense legislation to settle controversial social issues. Even if it passes, the law could face legal challenge. Critics might argue that Congress is improperly micromanaging military admissions or infringing on the executive branch’s authority to shape personnel policy. While such arguments are unlikely to succeed under the current administration, they could be used by a future administration seeking to challenge or circumvent the statute.

There is also a potential tradeoff with ongoing litigation. Passing a statutory ban before the resolution of existing service academy cases could risk mooting them, thereby depriving the judiciary of the opportunity to issue a binding constitutional ruling on racial preferences in military admissions. On the other hand, legislation could provoke new legal challenges that offer an alternate path to Supreme Court review. But these cases would likely center on separation-of-powers issues—i.e., whether Congress can dictate admissions policy to military institutions—rather than the underlying question of whether race-based admissions violate equal protection. A decision on those narrower grounds could leave the constitutional question unresolved. In that scenario, the statute—while effective in the short term—would remain vulnerable to repeal by a future Congress or to invalidation by courts on procedural or structural grounds. Only a clear constitutional holding can definitively foreclose future attempts to reinstate race-based admissions through executive or legislative means.

In sum, legislation remains a powerful tool for locking in recent policy gains and raising the political costs of reinstating racial preferences. But it does not provide an airtight safeguard. Without a definitive Supreme Court ruling, legislative prohibitions—like executive orders—remain vulnerable to reinterpretation, legal challenge, or future repeal. For that reason, legislative action is best pursued in tandem with other strategies, especially those that generate empirical and legal ammunition to bolster the case for race-neutral admissions as a matter of enduring national policy.

5.3 Executive Action

With race-based admissions at USNA—and likely other service academies--officially banned under the Trump administration’s 2025 executive orders, the immediate policy landscape has shifted. But while this represents a significant victory for merit-based military admissions, executive actions alone do not guarantee permanence. They are inherently reversible—susceptible to rescission or revision by future administrations. As such, further executive measures are needed to solidify the ban, create institutional safeguards, and challenge the assumptions underlying the diversity rationale that once justified racial preferences in officer selection.

5.3.1 Enforcing and Monitoring the Ban on Racial Preferences

Ongoing enforcement of the Trump ban will be critical. An executive order could expand upon existing directives by mandating periodic compliance audits across all service academies. These audits would examine whether race-neutral criteria are being applied consistently and identify any residual or covert practices that may reintroduce race-consciousness under the guise of holistic review or “whole person” metrics.

Additionally, the administration could require each service academy to publish anonymized admissions data and demographic breakdowns annually, enabling independent watchdogs and civil rights organizations to monitor compliance. This kind of transparency mechanism would not only deter backsliding but also build public trust in the fairness of military academy admissions. However, to ensure that such reporting does not itself compromise race neutrality, it is essential that admissions officers be structurally barred from accessing any race-related data—whether individual checkbox identifiers or aggregate demographic dashboards—during the admissions cycle. As SFFA has argued in ongoing litigation against the University of Texas, even passive access to such information poses a “cognizable danger” that race could influence decision-making, undermining claims of compliance and exposing institutions to constitutional challenge.²⁶⁷

Still, enforcement mechanisms—however robust—remain vulnerable to political reversal. A future administration could simply rescind oversight requirements or reinstate race-based criteria. To make such reversals more difficult and politically costly, the deeper ideological premises that once justified racial preferences must also be challenged. Central among these is the military’s long-invoked “diversity rationale”—the untested assumption, often presented as settled wisdom, that racial diversity is essential to cohesion and effectiveness. By shattering this manufactured consensus—or at minimum, inducing serious dissensus—policymakers can erode the basis on which courts have granted deference and ensure that race-conscious admissions cannot quietly return under a different name.

5.3.2 Establishing a Commission to Reassess the Diversity Rationale

One way to challenge the deeper premises behind race-based admissions is to establish an independent commission tasked with reassessing the evidence, origins, and assumptions underlying race-conscious policymaking in military contexts.

This commission would serve three primary functions:

First, it would systematically evaluate claims that racial diversity enhances military performance. This includes reviewing studies cited by government lawyers, military leaders, and policymakers in defense of race-conscious admissions. Each study would be assessed against stringent methodological standards, with credibility granted only to findings derived from controlled experiments, quasi-experimental methods, or statistical models that isolate the effect of racial diversity from confounding variables. Studies based on correlation alone, anecdotal evidence, extrapolations from civilian institutions, or untested theoretical claims—i.e., evidence of the kind submitted by USNA’s expert witnesses—would not meet the threshold for justifying racial preferences.

Second, given the Department of Defense’s admission that it has never systematically studied the relationship between racial diversity and military performance, the commission would fill this gap by conducting an independent analysis of DoD personnel and operational data. This research would assess whether racial composition influences key military outcomes—such as unit cohesion, combat readiness, mission success, and leadership effectiveness. As discussed briefly in Section 3.2.1, one promising approach would be to exploit natural variation in unit or officer demographics to test for performance differentials. Any observed differences would have to survive extensive statistical controls and be replicated across datasets, time periods, and operational environments to be considered meaningful. Findings that fall short of this evidentiary bar would be classified as merely suggestive—not sufficient to justify race-based admissions.

Finally, the commission would critically examine the historical origins of racial preferences in service academy admissions. While DoD reports cite Vietnam-era racial unrest and black underrepresentation in the officer corps as justification for affirmative action, this narrative has not been subjected to serious empirical scrutiny. The commission would investigate whether these claims stemmed from credible analysis or whether alternative factors—such as changes in recruitment standards, military culture, or broader social tensions—were more relevant causes of unrest. Additionally, it would assess how the justification for racial preferences evolved, shifting from a conflict-mitigation rationale to broader claims about diversity’s role in military performance. Foundational documents, such as the 2011 Military Leadership Diversity Commission report, would be reassessed—particularly their assumption that minority underrepresentation in the officer corps was inherently problematic and required policy intervention. Though the report framed diversity as a potential strategic resource, it provided no empirical evidence that greater officer racial diversity improves military performance—or that its absence threatens it. By scrutinizing these policy justifications, the commission would determine whether racial preferences were ever rooted in military necessity or were instead shaped by ideological and political pressures.

Ultimately, the commission’s findings would establish an authoritative government record that challenges the empirical and historical foundations of the military’s diversity rationale. By rigorously evaluating the justification, implementation, and consequences of race-conscious policymaking, it would provide a durable, evidence-based counterpoint to decades of institutional orthodoxy. These conclusions would inform future military and legislative decisions, while also serving as a powerful evidentiary resource in litigation—undermining any future attempt to revive racial preferences under the banner of “military necessity.”

Composition and Structure of the Commission

The commission’s effectiveness will hinge on its empirical rigor and independence, ensuring that its findings carry lasting legitimacy across administrations and policy domains. To that end, its core membership should consist primarily of social scientists and economists with demonstrated expertise in causal inference, labor markets, military personnel systems, and organizational performance. These experts would lead the review of existing scholarship and conduct original statistical analyses, grounding the commission’s conclusions in evidence rather than assumption-driven reasoning. Their research should be held to the highest standards of peer review to ensure that its findings are both credible and legally durable.

To supplement this empirical foundation, the commission should also include retired military leaders and national security professionals with firsthand experience in unit cohesion, operational performance, and officer development. While such members would provide vital context for interpreting statistical patterns, their role should be advisory rather than directive. The commission’s analytical core must remain methodologically rigorous and intellectually independent. Legal and policy scholars should also participate to assess the constitutional and administrative implications of racial preferences, particularly in light of evolving equal protection jurisprudence and doctrines of judicial deference in military affairs.

To guard against institutional capture—especially given the extent to which the “diversity rationale” has been internalized within the Department of Defense and elite policymaking circles—membership must be determined by analytical qualifications, not ideological litmus tests. Selection should prioritize empirical competence, methodological integrity, and professional independence over preordained policy agendas. By placing data at the center and treating institutional context as an interpretive layer, the commission would reorient the debate around facts rather than narratives.

A major structural challenge is the risk of political disruption. A future administration favoring race-based admissions could attempt to defund, obstruct, or disband the commission before its work is complete. To protect against this, the commission could be established through the annual defense appropriations process, ideally embedded in the National Defense Authorization Act (NDAA). This is the same legislative vehicle used to create the 2009 Military Leadership Diversity Commission (MLDC), and it would provide statutory recognition, dedicated funding, and a measure of insulation from executive interference. Institutionalizing the commission in this way would ensure it has the authority, independence, and operational runway to complete a full and credible reassessment of the diversity rationale in military admissions.

5.3.3 Strategic Implementation

While a commission and an executive order banning racial preferences could each stand alone, linking them offers important strategic advantages. Given that the Trump administration has already issued a ban, it could now reframe the policy as a temporary suspension pending the commission’s findings. Presenting the measure as part of a broader, evidence-based reassessment—rather than a permanent ideological shift—would bolster its legitimacy, reduce institutional resistance, and make it harder for critics to cast it as politically motivated. It would also shift the debate from normative assertions to empirical claims, where the diversity rationale is at its weakest.

During the suspension, the commission would evaluate both existing research and new evidence submitted by military officials, academic experts, and outside stakeholders. Proponents of race-conscious admissions would be invited to present quantitative studies showing a causal link between racial diversity and improved military outcomes. However, all submissions would be subject to rigorous evidentiary standards. Only research using valid methods—such as randomized experiments, quasi-experimental designs, or multivariate models that isolate the effect of racial diversity—would be considered. Studies relying on correlation, anecdote, or extrapolation from civilian institutions would not suffice.

If the commission finds no compelling statistical evidence that racial diversity enhances military performance, the administration could then convert the suspension into a permanent ban—grounded not only in constitutional concerns but in an official, government-commissioned analysis discrediting the diversity rationale. Any future effort to reinstate racial preferences would then face a dramatically higher bar. Proponents would need to:

1. Produce equally rigorous empirical evidence showing a direct and meaningful performance benefit;

2. Overcome the commission’s formal findings, which would carry significant weight in both judicial and policymaking forums; and

3. Demonstrate that any measurable effect is large enough to justify an explicit departure from equal protection principles—something the Supreme Court has emphasized must be both “compelling” and narrowly tailored.

This sequencing strategy would eliminate race-based admissions in the near term while erecting long-term institutional and legal barriers against their return. It would also shift the burden of proof to advocates of racial preferences, who would no longer be able to rely on untested assumptions or vague appeals to cohesion and legitimacy. By embedding the policy shift within a transparent and methodologically credible process, the administration would not only achieve immediate reform but also lay the groundwork for its enduring defense.

Conclusion: The Stakes of SFFA v. USNA

The district court ruling in SFFA v. USNA was not merely an isolated legal misstep—it marked a dramatic departure from decades of Supreme Court jurisprudence on equal protection. By uncritically accepting the Naval Academy’s justifications for racial preferences, Judge Bennett reduced strict scrutiny—the judiciary’s highest standard of review—to a mere procedural formality. His decision set a dangerous precedent: that the military can exempt itself from core constitutional constraints simply by invoking national security.

Recent developments have shifted the policy landscape. President Trump’s January 2025 executive orders, followed by directives from the Department of Defense, have formally ended race-conscious admissions at the Naval Academy. This move brings USNA into alignment with the Supreme Court’s decision in SFFA v. Harvard and marks a substantial policy victory for equal treatment.

Yet this reversal, while important, does not resolve the broader constitutional question: Can federal institutions sidestep equal protection guarantees simply by invoking national security? Judge Bennett effectively said yes. If that view is allowed to prevail, it would gut the Equal Protection Clause of its meaning—precisely in the context where constitutional limits matter most: the government’s power to allocate opportunity and authority.

The risks are compounded by the possibility that ongoing legal challenges at West Point and the Air Force Academy could be mooted by strategic policy changes. If all three cases end without definitive judicial review, Judge Bennett’s opinion may stand as the lone precedent addressing race-based admissions in the military. That outcome would leave a deeply deferential ruling on the record, ready to be revived when political conditions change.

To guard against that outcome, policymakers must act across all fronts—judicial, legislative, and executive. If the existing service academy cases are mooted, SFFA should seek to vacate Judge Bennett’s ruling to ensure it carries no precedential weight—and the courts, perhaps with support or encouragement from the Trump DOJ, should grant it. Codifying race-neutrality into statute—ideally through must-pass legislation such as the National Defense Authorization Act—should be a top priority for Congress, with the Trump administration actively proposing and marshalling support for the measure among its allies. Finally, the Trump administration must fortify its ban with structural safeguards: transparency, auditing, and independent oversight, including a commission to scrutinize the untested claims behind the military’s “diversity rationale.”

Ultimately, SFFA v. USNA is not just a dispute about affirmative action. It is a test of whether constitutional principles hold firm when they are least convenient. If vague appeals to national security are allowed to override the guarantee of equal protection, then that guarantee becomes a privilege—granted or withdrawn at the discretion of government officials. That is a precedent too dangerous to ignore. Ending race-conscious admissions at the service academies is not only sound policy—it is a constitutional imperative.

1. Davids, Y. M. (2025, March 26). Statement of Vice Admiral Yvette M. Davids, USN, Superintendent of the United States Naval Academy [Testimony]. Senate Subcommittee on Personnel, Committee on Armed Services. https://www.armed-services.senate.gov/imo/media/doc/davids_testimony.pdf; Students for Fair Admissions, Inc. v. United States Naval Academy, et al., No. 24-2214 (4th Cir. Mar. 28, 2025), Unopposed Motion to Hold Briefing in Abeyance, PACER Doc. 24.

2. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Declaration of Stephen Bruce Latta, Plaintiff's Exhibit 259, at 10 (D. Md. Sept. 17, 2024).

3. Ibid., at 11.

4. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 2, at 189-190 (D. Md. Sept. 17, 2024).

5. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, at 14 (D. Md. Sept. 18, 2024).

6. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Declaration of Stephen Bruce Latta, Plaintiff's Exhibit 259, at 20.

7. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, Table 3.14, at 45.

8. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Declaration of Stephen Bruce Latta, Plaintiff’s Exhibit 259, at 18.

9. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, Table 3.14, at 45.

10. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Declaration of Stephen Bruce Latta, Plaintiff’s Exhibit 259, at 19.

11. Ibid., at 18.

12. Ibid., at 20.

13. Ibid., at 14.

14. Ibid.

15. Ibid., at 14–15.

16. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 3, at 95. (D. Md. Sept. 18, 2024). See also Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Declaration of Stephen Bruce Latta, Plaintiff’s Exhibit 259, at 21–22.

17. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 1, at 77. (D. Md. Sept. 16, 2024).

18. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Declaration of Stephen Bruce Latta, Plaintiff’s Exhibit 259, at 17.

19. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, at 13.

20. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Declaration of Stephen Bruce Latta, Plaintiff’s Exhibit 259, at 30.

21. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 3, at 12–13.

22. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, Table 3.1, at 26.

23. Ibid., Table C.1, at C3.

24. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Declaration of Stephen Bruce Latta, Plaintiff’s Exhibit 259, at 5–6.

25. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, at 19.

26. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 5, at 94 (D. Md. Sept. 20, 2024).

27. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, at 55.

28. Ibid., Table 3.8, at 35.

29. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 1, at 69.

30. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, Table 3.9, at 36.

31. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 1, at 77.

32. Ibid. See also Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Declaration of Stephen Bruce Latta, Plaintiff’s Exhibit 259, at 23–24.

33. Ibid., at 24.

34. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, Table 3.9, at 36.

35. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Declaration of Stephen Bruce Latta, Plaintiff’s Exhibit 259, at 24.

36. Ibid., at 24–25.

37. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, at 52 n.56.

38. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 3, at 22–23.

39. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 2, at 214.

40. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 3, at 23–24 and 30.

41. While this threshold reduction was not mentioned or featured in any of the case-related documents, it becomes apparent when one compares archived NAPS brochures from the 2019–2020 academic year and earlier to those from more recent years (See under ‘Appointment to the U.S. Naval Academy’).

42. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, at 19.

43. Ibid., at 36.

44. Ibid., Table D.3, at D4.

45. Ibid., at E21.

46. Ibid., Table 3.9, at 36.

47. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Declaration of Stephen Bruce Latta, Plaintiff’s Exhibit 259, at 29.

48. Ibid., at 26.

49. Ibid., at 26–27.

50. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, Table 6.1, at 96.

51. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 2, at 238.

52. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 1, at 54.

53. Ibid., at 79–80.

54. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 3, at 37–38.

55. Ibid., at 38–39.

56. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 1, at 224–225.

57. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 2, at 26–27.

58. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 5, at 59.

59. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Declaration of Stephen Bruce Latta, Plaintiff’s Exhibit 259, at 21.

60. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, at 17.

61. Ibid.

62. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 2, at 85–94.

63. Note that unlike traditional R² values in linear regression, Pseudo R² values in logistic regression do not represent the proportion of variance explained in the same way. However, the predictive accuracy of these models remains high, underscoring the strength of the underlying relationships.

64. Arcidiacono explains that he treats the estimates in Model 6 as his preferred model for three key reasons. First, it eliminates concerns that RAB points might be influenced by racial preferences, ensuring that the estimated effects are not artificially inflated. Second, it preserves a larger sample size, avoiding unnecessary data loss that could weaken the statistical power of the analysis. Finally, it provides a conservative estimate of racial preferences—meaning that, if anything, it understates the actual extent of racial preferences, as demonstrated by the larger estimated effects in Models 7 and 8. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, at 64.

65. For technical details regarding this transformation, see Ibid., at B3.

66. Ibid., Tables 3.10–3.13, at 38–42.

67. Ibid., Table 4.5, at 73.

68. Ibid., Table D.3, at D4.

69. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 1, at 78–79.

70. See Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, Table 3.21, at 54.

71. According to the Latta Declaration, the Secretary of the Navy “may nominate 85 midshipmen per year from enlisted members of the Regular Navy and Regular Marines; 85 midshipmen per year from enlisted members of the Navy Reserve and the Marine Corps Reserve; and 20 midshipmen per year that graduated as honor graduates from schools designated as honor schools and from members of the Naval Reserve Officer's Training corps.” Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Declaration of Stephen Bruce Latta, Plaintiff’s Exhibit 259, at 16–17.

72. USNA refused to share individual-level performance data for 2023–2027 matriculants, and Judge Bennett ultimately upheld its position, ruling that the Academy was not required to provide the data. Students for Fair Admissions v. The United States Naval Academy, et al., Civil No. RDB-23-2699, Order on Motion for Other Relief (D. Md. Apr. 11, 2024).

73. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rebuttal Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 222, Table 6.1, at 96.

74. Ibid.

75. Ibid., Table D.127R, at D140–D141.

76. Applicants who fail USNA’s medical exam have the option to appeal for a waiver, but not all appeals are resolved. Because passing the medical exam or obtaining a waiver is a prerequisite for admission, disparities in waiver approvals directly influence which applicants remain eligible. While USNA denies considering race/ethnicity when deciding whether to grant a waiver, the data reveal significant racial differences in waiver approval rates.

77. Ibid., Table 6.3, at 99.

78. Ibid., at 100–101.

79. Ibid., at 101–102.

80. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 7, at 190–193 (D. Md. Sept. 24, 2024).

81. Ibid, at 156.

82. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Witness Rebuttal Report of Stuart Gurrea, Ph.D., Defendant's Exhibit 200, at 49–50 (D. Md. Sept. 24, 2024).

83. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 4, at 34 (D. Md. Sept. 19, 2024).

84. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Witness Rebuttal Report of Stuart Gurrea, Ph.D., Defendant's Exhibit 200, at 44.

85. Ibid., at 72.

86. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Reply Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 518, at 9 (D. Md. Sept. 18, 2024).

87. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Witness Rebuttal Report of Stuart Gurrea, Ph.D., Defendant's Exhibit 200, at 60–61.

88. Ibid., at 61.

89. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, at G1.

90. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Witness Rebuttal Report of Stuart Gurrea, Ph.D., Defendant's Exhibit 200, at 61–71.

91. Ibid., at 62.

92. Ibid., at 61.

93. Ibid., at 65–66.

94. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Supplemental Expert Report of Peter Arcidiacono, Plaintiff's Exhibit 518, at 14.

95. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Witness Rebuttal Report of Stuart Gurrea, Ph.D., Defendant's Exhibit 200, at 69.

96. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Supplemental Expert Report of Peter Arcidiacono, Plaintiff's Exhibit 518, at 14.

97. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Witness Rebuttal Report of Stuart Gurrea, Ph.D., Defendant's Exhibit 200, at 67.

98. Ibid., at 75.

99. Ibid., at 36.

100. Ibid., at 38–39.

101. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Supplemental Expert Report of Peter Arcidiacono, Plaintiff's Exhibit 518, at 3–4.

102. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Witness Rebuttal Report of Stuart Gurrea, Ph.D., Defendant's Exhibit 200, at 38.

103. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Supplemental Expert Report of Peter Arcidiacono, Plaintiff's Exhibit 518, at 5.

104. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 7, at 219–220.

105. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, July 15, 2024 Lyall Expert Report, Defendant's Exhibit 195, at 6 (D. Md. Sept. 23, 2024).

106. Ibid., at 6 n.16 (citing Goodwin, Gerald, Nikki Blacksmith & Meredith Coats, The Science of Teams in the Military: Contributions From Over 60 Years of Research, 73 AM. PSYCHOLOGIST 322, 322–333 (2018)).

107. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rebuttal Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 222, Figure 2.1N, at 7.

108. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, July 15, 2024 Lyall Expert Report, Defendant's Exhibit 195, at 6 n.14 (D. Md.) (citing Hong, Lu, Groups of Diverse Problem Solvers Can Outperform Groups of High-Ability Problem Solvers, 101 PROC. NAT'L ACAD. SCI. 16385, 16385-16389 (2004)).

109. Ibid. (citing Almaatouq, Abdullah, Mohammed Alsobay, Ming Yin & Duncan Watts, Task Complexity Moderates Group Synergy, 118 PROC. NAT'L ACAD. SCI. 1, 1-9 (2021)).

110. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rule 26(a)(2)(C) Disclosures of Jeannette Haynie, PhD, Defendant's Exhibit 194, at 4 n.6 (D. Md. July 15, 2024) (citing David Rock & Heidi Grant, Why Diverse Teams Are Smarter, HARV. BUS. REV. (Nov. 4, 2016), https://hbr.org/2016/11/why-diverse-teams-are-smarter).

111. Mackintosh, J. (2024, June 28). Diversity was supposed to make us rich. Not so much. The Wall Street Journal. https://www.wsj.com/finance/investing/diversity-was-supposed-to-make-us-rich-not-so-much-39da6a23

112. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rule 26(a)(2)(C) Disclosures of Jeannette Haynie, PhD, Defendant's Exhibit 194, at 6 n.13 (citing Daphne S. L. Wong, Exploring the Impact of Team Building on Group Cohesion of a Multicultural Team (2015) (Master's thesis, Pepperdine University), https://digitalcommons.pepperdine.edu/cgi/viewcontent.cgi?article=1606&context=etd).

113. Ibid., at 9 n.26 (citing Col. Suzanne M. Streeter, The Air Force and Diversity: The Awkward Embrace, AIR & SPACE POWER J., May-June 2014, at 104–132, https://www.airuniversity.af.edu/Portals/10/ASPJ/journals/Volume-28_Issue-3/F-Streeter.pdf).

114. Erik G. Helzer, Paul B. Lester & Simona Tick, Assessing the Relationship Between Diversity and Navy Unit Performance, at 12 (Naval Postgraduate School, NPS-23-N093-A, Mar. 2024), https://hdl.handle.net/10945/72655. See also Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 6, at 167–168 (D. Md. Sept. 23, 2024).

115. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 6, at 65–66.

116. Ibid., at 217.

117. Ibid., at 106.

118. Ibid., at 126.

119. Ibid., at 147–151

120. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 1, at 53.

121. For Haynie’s remarks on this point, see Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 6, at 124–125. For Lyall’s remarks, see Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 7, at 49–50.

122. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 4, at 227.

123. Erik G. Helzer, Paul B. Lester & Simona Tick, Assessing the Relationship Between Diversity and Navy Unit Performance, at 12 (Naval Postgraduate School, NPS-23-N093-A, Mar. 2024), https://hdl.handle.net/10945/72655.

124. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 6, at 179–180.

125. Frances J. Milliken & Luis L. Martins, Searching for Common Threads: Understanding the Multiple Effects of Diversity in Organizational Groups, 21 ACAD. MGMT. REV. 402, 402-433 (1996).

126. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 7, at 45–48.

127. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Deposition of Peter S. Arcidiacono, at 476 (D. Md. Aug. 20, 2024).

128. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 2, at 61–62.

129. Ibid., at 74–77

130. Ibid., at 81.

131. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, July 15, 2024 Lyall Expert Report, Defendant's Exhibit 195, at 7–21.

132. Ibid., at 14–15.

133. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rebuttal Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 222, at 7–9.

134. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 7, at 19.

135. Ibid., at 20.

136. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, July 15, 2024 Lyall Expert Report, Defendant's Exhibit 195, at 18–21.

137. Ibid., at 18.

138. Ibid., at 19.

139. Ibid., at 20–21. For Haynie’s citation and discussion of this study, see Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rule 26(a)(2)(C) Disclosures of Jeannette Haynie, PhD, Defendant’s Exhibit 194, at 11–12.

140. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 1, at 62–63. See also Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 9, at 56 (D. Md. Sept. 26, 2024).

141. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, July 15, 2024 Lyall Expert Report, Defendant's Exhibit 195, at 19.

142. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 1, at 62–63. See also Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 8, at 75–76 (D. Md. Sept. 25, 2024). This point is also echoed by USNA witness Dr. Beth Bailey in her expert report: “Racial divisions continue in American society today, and the visible diversity of leadership serves as a bulwark—symbolic and potentially practical—against renewed racial conflict.” Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Beth Bailey, PhD, Defendant's Exhibit 196, at 41 (D. Md. July 15, 2024).

143. Goldberg, Z. [@ZachG932]. (2020, October 14). [Tweet]. X. https://x.com/ZachG932/status/1316471938156507136

144. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 1, at 180.

145. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 6, at 41–44.

146. Specifically, the authors conclude their analysis of discretion-related disparities between black and white service members by noting that the “results provide no specific evidence of discretion-related racial bias in the USA MJS.” Amanda Kraus, Elizabeth Clelan, Dan Leeds, Sarah Wilson with Cathy Hiatt, Jared Huff & Dave Reese, Exploring Racial, Ethnic, and Gender Disparities in the Military Justice System, at 65 (CNA 2023).

147. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 5, at 164–165.

148. U.S. House of Representatives Committee on Armed Forces. (1973). Report by the Special Subcommittee on Disciplinary Problems in the US Navy (H.A.S.C. 92-81). U.S. Government Printing Office.

149. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, July 15, 2024 Lyall Expert Report, Defendant's Exhibit 195, at 22.

150. Hernandez Rodriguez, J. M., & Serna, C. (2020). The effects of diversity among peers and role models on U.S. Navy retention [Master's thesis, Naval Postgraduate School]. Naval Postgraduate School.

151. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 7, at 53–59.

152. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 8, at 55–75.

153. Ibid., at 59.

154. Ibid., at 60, 64, and 66.

155. Ibid., at 66.

156. Ibid.

157. Ibid., at 62–64.

158. Ibid., at 138.

159. Ibid., at 147–148.

160. USNA admits to considering race as an admissions factor since at least 2002—when the current Dean of Admissions (Stephen Latta) was assigned to the Office of Admissions. However, it also says it’s been “tracking race/ethnicity in admissions since 1980”, and that “diversity recruiters have been assigned to the Office of Admissions dating back to the early 1970s”. Hence, it is likely that race began to be considered much earlier than 2002. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Memorandum for the House Armed Services Committee: Use of "Affirmative Action" in Admissions Decisions at the U.S. Naval Academy (Nov. 9, 2022), Plaintiff's Exhibit 31 (D. Md. Sept. 16, 2024).

161. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 8, at 55–56.

162. Ibid., at 108–109.

163. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rule 26(a)(2)(C) Disclosures of Jeannette Haynie, PhD, Defendant's Exhibit 194, at 11 n.30. (citing Slapakova, L., Caves, B., Posard, M., Muravska, J., Dascalu, D., Myers, D., Kuo, R., & Thue, K. (2022). Leveraging Diversity for Military Effectiveness: Diversity, Inclusion and Belonging in the UK and US Armed Forces. RAND Corporation, p. 33 n.166, which cites McPherson, M., Smith-Lovin, L., & Cook, J. M. (2001). Birds of a feather: Homophily in social networks. Annual Review of Sociology, 27(1), 415-444).

164. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 6, at 64.

165. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, July 15, 2024 Lyall Expert Report, Defendant's Exhibit 195, at 23 n.79. (citing Bove, V., & Ruggeri, A. (2016). Kinds of Blue: Diversity in UN Peacekeeping Missions and Civilian Protection. British Journal of Political Science, 46(3), 681-700).

166. Ibid., at 23 n.77.

167. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 7, at 61–67.

168. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Rule 26(a)(2)(C) Disclosures of Jeannette Haynie, PhD, Defendant’s Exhibit 194, at 15.

169. Slapakova, L., Caves, B., Posard, M., Muravska, J., Dascalu, D., Myers, D., Kuo, R., & Thue, K. (2022). Leveraging Diversity for Military Effectiveness: Diversity, Inclusion and Belonging in the UK and US Armed Forces (p. vii, 7–8). RAND Corporation.

170. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 7, at 119–122.

171. Ibid., at 126–127.

172. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 6, at 14–15, 19–20, 28–29.

173. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 9, at 37–38.

174. U.S. Navy. (2023, January). Guide 5: Physical readiness test (PRT). Naval Personnel Command. https://www.mynavyhr.navy.mil/Portals/55/Support/Culture%20Resilience/Physical/Guide_5-Physical_Readiness_Test_PRT_JAN_2023.pdf

175. See Figure 2.2B as well as Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Expert Report of Peter S. Arcidiacono with Appendices, Plaintiff's Exhibit 218, at 45. To be sure, the PRT data reflect midshipmen from USNA classes 2014–2020, whereas the CFA data pertain to admitted applicants for USNA classes 2023–2027. While direct comparisons should be made with caution, the consistency of group rank order in both datasets suggests that the disparities observed in CFA scores for recent cohorts likely extend to earlier class years as well.

176. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Email with “Midshipmen Service Assignment” and “Class OOM” Documents attached, Plaintiff's Exhibit 202, USNAINST 1531.51B, at 4 (D. Md. Sept. 16, 2024).

177. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Plaintiff's Exhibit 148, Email from Steve Vahsen to David Forman re: “Any available time?” with attachment, Summary Assessment of Midshipmen Equity by Race/Ethnic Group, at 14.

178. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 6, at 50–51. See also Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 8, at 50 and 132–134.

179. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 5, at 46–48, 81–82, and 88–89.

180. See page 8 of Switzer, W. S. (2020, October 30). USNA service assignment [Conference presentation]. Annual Parent Club Officer Conference, U.S. Naval Academy. https://www.usna.com/file/2020-pcoc/Updated-2020-Parents-Club-Service-Assignment-Brief.pdf

181. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Email with “Midshipmen Service Assignment” and “Class OOM” Documents attached, Plaintiff's Exhibit 202, USNAINST 1301.5L, at 3.

182. Ibid., at 3–4.

183. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 8, at 38.

184. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Task Force One Navy: Our Navy Team – Navigating A Course to True North, Plaintiff's Exhibit 600, at 102–103.

185. Insler, M. A., & Karam, J. (2019). Do sports crowd out books? The impact of intercollegiate athletic participation on grades. Journal of Sports Economics, 20(1), 115-153.

186. Lim, N., Mariano, L. T., Cox, A. G., Schulker, D., & Hanser, L. M. (2014). Improving demographic diversity in the U.S. Air Force officer corps (Report No. RR495). RAND Corporation. https://www.rand.org/content/dam/rand/pubs/research_reports/RR400/RR495/RAND_RR495.pdf; see also Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 6, at 181–206.

187. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 6, at 59–62.

188. Wygant v. Jackson Bd. of Educ., 476 U.S. 267, 280, n.6 (1986); Fisher v. University of Texas at Austin, 570 U.S. 297, 312–13 (2013) (Fisher I)

189. Students for Fair Admissions, Inc. v. President & Fellows of Harvard Coll., 600 U.S. 7, 20, 22, 27–28 (2023).

190. Grutter v. Bollinger, 539 U.S. 333 (2003)

191. Students for Fair Admissions, Inc. v. President & Fellows of Harvard Coll., 600 U.S. 7–8, 21, 30, 33.

192. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Findings of Fact and Conclusions of Law, at 2 (D. Md. Dec. 6, 2024).

193. Congressional Record. (2008, September 17). 110th Congress, 2nd Session, 154(148), S8931-S8932. U.S. Government Printing Office.

194. Duncan Hunter National Defense Authorization Act for Fiscal Year 2009, Pub. L. No. 110-417, 122 Stat. 4356, 4476-4478 (2008).

195. Military Leadership Diversity Commission. (2011, March 15). From Representation to Inclusion: Diversity Leadership for the 21st-Century Military (Final Report). Department of Defense. https://apps.dtic.mil/sti/pdfs/ADA539297.pdf

196. Esper, M. T. (2020, June 19). Actions for improving diversity and inclusion in the Department of Defense [Memorandum]. Department of Defense. https://media.defense.gov/2020/Jun/22/2002319394/-1/-1/1/ACTIONS-FOR-IMPROVING-DIVERSITY-AND-INCLUSION-IN-THE-DOD.PDF

197. Department of Defense Board on Diversity and Inclusion. (2020, December 18). Department of Defense Board on Diversity and Inclusion report: Recommendations to improve racial and ethnic diversity and inclusion in the U.S. military. Department of Defense. https://s3.documentcloud.org/documents/20433039/dod-diversity-and-inclusion-final-board-report.pdf

198. Department of Defense. (2020, December 18). Department of Defense Board on Diversity and Inclusion report: Recommendations to improve racial and ethnic diversity and inclusion in the U.S. military. https://media.defense.gov/2020/Dec/18/2002554852/-1/-1/0/DOD-DIVERSITY-AND-INCLUSION-FINAL-BOARD-REPORT.PDF

199. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Findings of Fact and Conclusions of Law, at 17.

200. Ibid., at 18–20.

201. Ibid., at 126–132.

202. Ibid., at 135.

203. Bennett writes: “Similarly, expert testimony elucidated numerous studies that support DoD’s military judgment regarding the importance of diversity to unit cohesion and thus to national security. See, e.g., (PX275; DX170; PX597).”. Note that these exhibit labels correspond to the studies discussed here and in subsequent paragraphs. Ibid., at 136.

204. Ibid., at 136.

205. Ibid., at 137.

206. Ibid., at 136–137.

207. Ibid., at 137–18.

208. Ibid., at 141–142.

209. Ibid., at 142.

210. Ibid., at 143.

211. Ibid.

212. Wygant v. Jackson Bd. of Educ., 476 U.S. 267 (1986).

213. Ibid.

214. Ibid., at 139.

215. Ibid., at 141.

216. Ibid., at 144.

217. Ibid.

218. Ibid., at 145.

219. Ibid.

220. Ibid., at 146.

221. Ibid., at 147.

222. Ibid.

223. Ibid., at 147–148.

224. Ibid., at 147.

225. Ibid., n. 82.

226. Ibid.

227. Ibid., at 147.

228. Ibid., at 149.

229. Ibid., at 80.

230. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 4, at 37.

231. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Findings of Fact and Conclusions of Law, at 82.

232. Ibid., at 82–84.

233. Ibid., at 86.

234. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 7, at 181.

235. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Findings of Fact and Conclusions of Law, at 89–98.

236. Ibid., at 89–90.

237. Ibid., at 90.

238. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Declaration of Stephen Bruce Latta, Plaintiff's Exhibit 259, at 39.

239. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Findings of Fact and Conclusions of Law, at 164.

240. Carnevale, Rose, & Strohl, “Achieving Racial and Economic Diversity with Race-BlindAdmissions Policy,” in The Future of Affirmative Action.

241. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Findings of Fact and Conclusions of Law, at 89.

242. Ibid., at 171–172.

243. Ibid., at 168–169.

244. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 2, at 125–126. See also Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 6, at 64–65.

245. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Findings of Fact and Conclusions of Law, at 171–172.

246. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Findings of Fact and Conclusions of Law, at 165–167.

247. Ibid., at 164.

248. Ibid., at 166.

249. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 2, at 95. See also Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 3, at 45.

250. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Defendants’ Post-Trial Proposed Findings of Fact and Conclusions of Law, at 69 (D. Md. Oct. 2, 2024).

251. Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Findings of Fact and Conclusions of Law, at 158.

252. Ibid., at 156.

253. Ibid., at 159.

254. Ibid., at 167 n. 98.

255. Ibid., at 166.

256. Ibid., at 9–10 and 20.

257. Ibid., at 155. See also Students for Fair Admissions v. The United States Naval Academy, et al., 1:23-cv-02699-RDB, Trial Transcript Day 8, at 114.

258. Trump, D. J. (2025, January 21). Ending illegal discrimination and restoring merit-based opportunity [Executive order]. The White House. https://www.whitehouse.gov/presidential-actions/2025/01/ending-illegal-discrimination-and-restoring-merit-based-opportunity/

259. Davids, Y. M. (2025, March 26). Statement of Vice Admiral Yvette M. Davids, USN, Superintendent of the United States Naval Academy [Testimony]. Senate Subcommittee on Personnel, Committee on Armed Services. https://www.armed-services.senate.gov/imo/media/doc/davids_testimony.pdf; Students for Fair Admissions, Inc. v. United States Naval Academy, et al., No. 24-2214 (4th Cir. Mar. 28, 2025), Unopposed Motion to Hold Briefing in Abeyance, PACER Doc. 24.

260. Students for Fair Admissions v. United States Military Academy at West Point, et al., No. 23-CV-08262 (PMH) (S.D.N.Y. Feb. 25, 2025), Order, ECF No. 127; Latest document in the USAFA case (courtesy of Pacer) is ‘Order on Motion for Extension of Time to Answer or Otherwise Respond’. While the document itself cannot be viewed, the full docket text reads: “MINUTE ORDER granting [19] Consent Motion to Extend Defendants' Deadline to Respond to Complaint and to Submit Proposed Scheduling Order. The deadline for Defendants to respond to Plaintiff's Complaint is extended to April 16, 2025. The deadline to file a Proposed Scheduling Order is extended to May 7, 2025. By Magistrate Judge Maritza Dominguez Braswell on 3/10/2025. Text Only Entry (mdblc5)”. Thus, as of now, the case is still ongoing.

261. Letter from U.S. Department of Justice to Judge Philip M. Halpern, Students for Fair Admissions v. United States Military Academy at West Point, et al., No. 23-CV-08262 (PMH) (S.D.N.Y. Feb. 20, 2025), ECF No. 125; Students for Fair Admissions v. United States Air Force Academy, et al., No. 1:24-cv-03430-NYW-MDB (D. Colo. Mar. 10, 2025), Consent Motion to Extend Defendants' Deadline to Respond to Complaint and to Submit Proposed Scheduling Order, ECF No. 19.

262. Students for Fair Admissions v. United States Military Academy at West Point, et al., No. 23-CV-08262 (PMH) (S.D.N.Y. Feb. 25, 2025), Order, ECF No. 127.

263. Students for Fair Admissions v. United States Air Force Academy, et al., No. 1:24-cv-03430-NYW-MDB (D. Colo. Mar. 10, 2025), Minute Order Granting Consent Motion to Extend Defendants' Deadline to Respond to Complaint and to Submit Proposed Scheduling Order, ECF No. 21 (text-only entry).

264. Just recently, Roberts and Barrett joined the Court’s three liberal justices in a 5-4 ruling against the Trump administration on foreign aid spending, a move that underscores the unpredictability of the Court’s current alignment.

265. FBI v. Fikre, 601 U.S. 234, 241 (2024)

266. Students for Fair Admissions, Inc. v. University of Texas at Austin, et al., No. 24-50631 (5th Cir. Feb. 5, 2025), Reply Brief of Appellant Students for Fair Admissions, Inc., ECF No. 57.

267. Ibid., at 16.