How Does Admissions at Harvard Work?
Every selective college keeps its evaluation process private. Harvard is the exception, and not by choice.
Between 2014 and 2023, the lawsuit brought by Students for Fair Admissions forced Harvard to turn over its internal admissions manuals, its confidential rating rubrics, six years of applicant-level data, its own unpublished institutional research, and thousands of internal emails. A federal judge then spent three weeks in open court taking testimony from the dean of admissions, the director of admissions, the dean of the college, and rank-and-file admissions officers. She issued a 130-page factual account of how the office actually operates. The First Circuit and the Supreme Court added more.
The result is that we know more about how Harvard reads applications than we know about any other college in the country. Not from an anonymous former admissions officer selling a course. From sworn testimony and produced documents.
This article walks through what that record shows. It covers the path a file takes through the office, the numeric ratings, the confidential lists that most of the public has never heard of, what the case revealed about the treatment of Asian American applicants, what changed after the Supreme Court ruled in 2023, and the small oddity that started this piece for me, which is Harvard's strange answer to the Common Data Set.
Then, at the end, the part that matters most to students, which is how much of this you can actually act on.
The Path Your Application Takes
Step one, the docket
Harvard does not read applications in one national pile. It divides them into roughly twenty geographic "dockets" based on the location of your high school. Each docket is staffed by a subcommittee led by a senior admissions officer who serves as docket chair, along with three to six additional officers. Each of those officers is responsible for reading every application from a specific subset of the docket's high schools.
Two consequences follow. First, the person reading your file has probably read files from your school before and knows how your school's grading, course offerings, and counselor letters compare to others nearby. Second, at the first real decision point you are being compared against other applicants from your region before you are ever compared against the national pool.
Step two, the first and second reads
The first reader records a set of data points from your file. For the Class of 2023, those included citizenship, race, legacy status, recruited athlete status, socioeconomic and disadvantaged status, financial aid indicators, and test scores. The reader then assigns the numeric ratings described below and writes a summary.
A second reader, always an admissions office employee, verifies the factual record and assigns their own scores. Faculty sometimes review submitted academic work, research, music, or art. Most applications receive two reads.
There is a documented cutoff at this stage, and it is worth pausing on how Harvard writes its scores, because the notation runs backwards from what most people expect.
Harvard rates applicants from 1 to 6, where 1 is the best score and 6 is the worst. Readers then sharpen a score with a plus or a minus, and those work exactly the way they do in letter grades. A 2 plus is a strong 2. A plain 2 is a solid one. A 2 minus is a weak 2, and a weak 2 still outranks the strongest 3. The only thing to keep straight is that a smaller number is a better result, the opposite of a GPA.
With that in hand, the rule from the Class of 2023 reading procedures is simple. A file scoring anywhere in the 1 or 2 range overall, down to and including a weak 2, should go forward to committee. A strong 3 goes forward only if the first reader argues for it. A solid 3 or anything below it usually ends there. That one score, assigned before any committee has met to discuss anyone, is where the majority of applications stop.
Step three, the subcommittee
Regional subcommittees of roughly five to eight people, mixing admissions officers, faculty readers, and the docket chair, meet in multi-day sessions. Early Action subcommittees begin in November. Regular Decision subcommittees run from late January into February. Harvard's own description of these meetings notes that a single applicant can be discussed for up to an hour. Decisions are made by simple majority and forwarded to the full committee.
Step four, the full committee
The full committee has forty members. Discussion centers on the applicants the regional subcommittees have recommended. Every member votes, and a majority is required to admit.
The trial record includes a detail worth knowing. At the start of the committee process, the group would discuss the racial composition of the admitted pool, with the stated goal, in the director of admissions' words, of avoiding a dramatic drop-off in minority admissions from the previous class. This practice ended after 2023, and I will come back to it.
Step five, the lop
This is the stage almost no applicant knows about.
By the end of committee, Harvard has tentatively admitted more students than it can enroll, somewhere around two thousand. The class then has to be trimmed. Applicants under consideration for cutting are placed on what the office internally calls the "lop list."
According to the Supreme Court's description of the trial record, the lop list contained only four pieces of information about each applicant. Legacy status. Recruited athlete status. Financial aid eligibility. Race. Judge Burroughs's findings describe a slightly fuller version that could include Harvard Financial Aid Initiative status, athletic rating, legacy status, gender, and race.
Everything else, the essays, the grades, the recommendations, the interview, is gone by this point. The committee is cutting on a handful of institutional variables. This stage was central to the plaintiff's case, and the Supreme Court majority later pointed to it as evidence that race functioned as more than a soft consideration.
The Ratings
The 1 to 6 scale described above is applied separately in each of the categories below, and each category has its own written definition of what a given number means. Those definitions come from the reading manuals Harvard produced in discovery.
Academic. The manual defines each level by how the student would eventually graduate from Harvard. A 1 means summa cum laude potential, meaning someone who would finish at the very top of a Harvard class, and it requires near-perfect grades and scores and sometimes original scholarship. Fewer than one percent of applicants receive it. A 2 means magna cum laude potential, the tier just below that, and calls for superb grades with mid to high 700s on SAT sections or a 33 or higher on the ACT. A 3 means cum laude potential, meaning graduating with honors, with mid 600s to low 700s or a 29 to 32. A 4 is described as adequate preparation, low to mid 600s or a 26 to 29. A 5 is marginal potential.
Extracurricular. A 1 indicates possible national-level achievement or genuine professional experience. A 2 indicates strong state or regional involvement, and the manual's own examples are class president, newspaper editor, and concertmaster. That is worth sitting with. Being student body president of your high school is a 2, not a 1.
Athletic. A 1 indicates national, international, or Olympic-level potential.
Personal. A 1 is reserved for what the Class of 2023 procedures call truly outstanding qualities of character, enormous courage, singular ability to lead or inspire, extraordinary concern or compassion. A 2 is very strong. A 3 is generally positive. A 4 is described as bland or somewhat negative or immature. Readers are instructed to weigh humor, sensitivity, grit, leadership, integrity, helpfulness, courage, and kindness. Personal 1s are given to fewer than ten applicants in a typical year.
School Support. This scores your teacher and counselor recommendations. Language along the lines of "best of a career" or "one of the best in many years" pushes toward a 1. "One of the best" or "the best this year" pushes toward a 2.
Alumni interview. More than fifteen thousand alumni conduct interviews each year. They assign their own personal and overall ratings and never see your application. Harvard testified that it weighs the interviewer's written comments more heavily than the numbers. One line from the record deserves emphasis, which is that applicants who do not interview are rarely admitted.
Overall. This is not an average. It is an explicit "stepping back" judgment. An overall 1 signals an exceptional candidate with better than a ninety percent chance of admission, and applicants with an overall 1 are essentially always admitted. Applicants at 3 minus or worse are almost always rejected.
What the ratings data actually showed
Harvard's own expert, David Card, produced the most useful finding for applicants in the entire case. He looked at strength across the four profile ratings rather than at any single one.
Applicants who earned a 1 or 2 in three or four of the four categories were admitted at roughly seventy percent, and they made up forty-six percent of all admits. Applicants who earned a 1 or 2 in one category or none were seventy-three percent of the applicant pool and only fifteen percent of admits.
Fewer than one percent of admitted students had an academic, extracurricular, or personal rating of 4 or lower.
The Lists
Here is where the case became genuinely revelatory, because these mechanisms had never been documented in public before.
The Dean's Interest List and the Director's Interest List
These are confidential lists of applicants flagged as being of special institutional interest, in practice largely applicants connected to donors and prominent alumni. Dean William Fitzsimmons testified that the University Development Office sometimes offers names to the dean, and that he met regularly with development staff to discuss specific high school students. He defended the practice as important to the long-term strength of the institution.
The relevant donor body is the Committee on University Resources, membership in which requires a gift of one million dollars or more.
The numbers are stark. Across the Classes of 2014 through 2019, applicants on the dean's or director's list were admitted at roughly 42.7 percent, against a baseline of under 5.5 percent for everyone outside the preference categories. They made up about 9.34 percent of all admits. In the Class of 2019 alone, 192 admitted students, more than ten percent of the class, were on one of these lists.
Trial exhibits included internal emails that made the mechanism concrete, including a thank-you note from a Harvard dean to Fitzsimmons noting that an applicant's family had already committed to funding a building, and an exchange concerning an applicant whose late grandfather had given 8.7 million dollars.
The Director's List operates similarly but is maintained through the admissions office by the director of admissions rather than keyed as directly to the dean's development contacts.
There is no application step that puts you on either list. That is the point.
The Z-List
The Z-list is Harvard's back door, and it is small. Roughly fifty to sixty students per year are offered admission on the condition that they take a gap year before enrolling. The name apparently came from computer technicians, because these are the last admits entered each cycle.
Its composition is the story. For the Classes of 2014 through 2019, the Z-list was about seventy percent white and about 46.5 percent legacy. Black students made up roughly two percent of it.
Functionally, it is a pressure valve. It lets Harvard admit academically borderline but institutionally desirable applicants without counting them in the on-time class or its reported statistics.
ALDC
This acronym came out of Harvard's own expert rebuttal, and it now dominates the academic literature on the case. It stands for Athletes, Legacies, Dean's and Director's interest list, and Children of faculty and staff.
ALDCs are about five percent of Harvard's applicant pool and about thirty percent of its admitted class. The admit rates, from Arcidiacono, Kinsler, and Ransom's analysis of the Classes of 2014 through 2019, are as follows.
Recruited athletes, about 86 percent
Children of faculty and staff, about 46.7 percent
Dean's interest list, about 42.7 percent
Legacies, about 33.6 percent
Everyone else, under 5.5 percent
Over forty-three percent of white admitted students were ALDC. For Black, Hispanic, and Asian American admits, the figure was under sixteen percent for each group. Arcidiacono estimated that roughly three-quarters of white ALDC admits would have been rejected had they been evaluated as ordinary applicants.
Recruited athletes
The Ivy League offers no athletic scholarships, so the mechanism runs through admissions. Coaches vet transcripts, send promising recruits to the admissions office for an informal early verdict on whether the file would survive, which is known internally as a pre-read, and then endorse the ones who clear it. Endorsed athletes receive "likely letters," which Harvard sends as early as October, months before regular decisions.
The single most extreme number in the trial record concerns this group. Among applicants with an academic rating of 4, recruited athletes were admitted at roughly seventy percent. Non-athletes with the same academic rating were admitted at 0.076 percent. That is a difference of nearly a thousandfold. The only applicants admitted with an academic rating of 5 were recruited athletes.
Recruiting is constrained by the Academic Index, a composite of academic record and test scores on a 60 to 240 scale that the Ivy League has used since the 1980s. The rule is that a team's recruits, taken as a group, have to average reasonably close to the wider student body rather than drifting far below it, and individual recruits are sorted into bands according to how far below the campus average they sit. The admitted-class average at Harvard, Yale, and Princeton sits near 220. The commonly cited floor for recruited athletes is around 176. The exact current formula is not officially published.
One-pagers
At key points in the cycle, at the close of Early Action and again at the close of Regular Decision, senior officials review "one-pagers," which are demographic snapshots of the admitted pool. They track gender, geography, intended academic interest, legacy status, disadvantaged status, recruited athlete status, citizenship, and, before 2023, race, each compared against the prior year.
Tips, and How Harvard Recruits
Internally, Harvard's officers speak of "tips," meaning advantages rather than points. Tips run to underrepresented minorities before 2023, first-generation and low-income applicants through the Harvard Financial Aid Initiative, recruited athletes, ALDCs, and applicants from what the office calls "sparse country," meaning states from which few students attend.
A 2013 internal chart ranked the size of these tips roughly in this order: a top athletic rating first, then a high personal rating, then legacy status, then African American identity.
The recruiting operation is enormous and, as the trial revealed, differentiated. Harvard buys score and contact data on well over a hundred thousand students annually and sent more than 114,000 recruitment letters for the Class of 2018 alone. Fitzsimmons testified to the score thresholds used for those mailings. Black, Hispanic, and Native American students received letters at a combined math and verbal SAT of 1100 or above. White students in sparse-country states such as Montana and Nevada received them at 1310. Asian American men needed 1380, and Asian American women 1350. Asian American students in sparse-country states did not receive the lower sparse-country threshold.
Fitzsimmons defended the lower thresholds as recognizing that economic disadvantage limits access to test preparation. Arcidiacono's separate research argued that the recruiting drew applications from students who were then almost never admitted.
What Else the Case Revealed
Harvard's own internal research
In 2012 and 2013, Harvard's Office of Institutional Research produced internal analyses, reportedly prompted in part by public criticism, that found Asian American applicants had lower admission rates than their credentials predicted and that legacy and athlete preferences outweighed anything Harvard did for low-income students.
One OIR model estimated that on academic credentials alone, Asian American students would have made up about forty-three percent of the admitted class. Adding athlete and legacy preferences dropped that to about thirty-one percent. Adding the extracurricular and personal ratings dropped it to about twenty-six percent. The actual figure was about nineteen percent.
The reports went to Fitzsimmons and to the dean of the college. They were not acted on and were not published. Harvard characterized them at trial as incomplete and preliminary, and disputed their methodology. OIR staff testified they had no explanation for the negative association other than the obvious one.
The personal rating
Among applicants outside the preference categories, personal ratings of 1 or 2 went to 21.3 percent of white applicants and 17.6 percent of Asian American applicants, with Black applicants at roughly nineteen percent and Hispanic applicants at roughly 18.7 percent. Asian American applicants scored highest as a group on the academic and extracurricular ratings.
The comparison that drew the most attention at trial is that alumni interviewers, who actually met the applicants, gave Asian American students personal ratings comparable to white students. The disparity appeared in the ratings assigned by admissions officers reading files.
The battle of the experts
Peter Arcidiacono of Duke, testifying for the plaintiffs, found a penalty that was statistically significant, meaning large enough and consistent enough across years that random chance was an unlikely explanation. His headline estimate was that the typical Asian American applicant's admission rate would rise by about nineteen percent, roughly one percentage point in absolute terms, if that applicant were treated as white. His model excluded ALDCs, on the grounds that they go through a fundamentally different process, and excluded the personal rating, on the grounds that the rating was itself the mechanism in question.
David Card of Berkeley, testifying for Harvard and later a Nobel laureate, included both ALDCs and the personal rating and modeled each year separately. He found no statistically significant penalty. The plaintiffs countered that removing the personal rating from Card's own model produced a significant penalty.
The Supreme Court's majority opinion cited one Harvard-specific figure that captures why the case reached the Court. The analysis sorted applicants into ten equal groups by academic strength, from the weakest tenth of the pool to the strongest tenth. A Black applicant in the fourth-weakest of those ten groups had a marginally better chance of admission, 12.8 percent, than an Asian American applicant in the strongest group, 12.7 percent.
Two cautions on that number. The far more widely circulated figures of 83, 58, and 47 percent belong to the University of North Carolina, not to Harvard, and are misattributed constantly online. And a comparison built this way sorts students only by grades and scores. It says nothing about how the same applicants scored on the extracurricular, personal, and school support ratings, which is precisely what the two sides spent the trial arguing about.
What the trial judge concluded
Judge Allison Burroughs ruled for Harvard on every count in September 2019. She found a statistically significant and negative relationship between Asian American identity and the personal rating. She also found no persuasive documentary evidence of racial animus or conscious prejudice, and could not identify any individual applicant who had been harmed. She raised the possibility of implicit bias while noting it was unsupported by direct evidence, and recommended that Harvard adopt implicit bias training and monitor its rating disparities. Her closing framing was that the program is not perfect but passes constitutional muster, and that she declined to dismantle a very fine admissions program solely because it could do better.
What Changed in 2023, and What Did Not
On June 29, 2023, the Supreme Court held that Harvard's and UNC's race-conscious admissions violated the Equal Protection Clause and Title VI, the federal statute barring racial discrimination by any institution that accepts federal money. The majority pointedly did not resolve the separate claim that Harvard had intentionally discriminated against Asian American applicants.
What changed. Race is no longer considered in application review. Students may still self-report race, but that information is released only after decisions. Harvard replaced its supplemental essay with five required 150-word short answers, the first of which asks how the life experiences that shaped you will enable you to contribute to Harvard. That prompt is the legal channel the Court left open, which permits a college to consider how race affected an applicant's life as conveyed through an essay, while forbidding consideration of race as such.
What the numbers did. The Class of 2027, the last pre-ruling class, was roughly eighteen percent Black and roughly 29.9 percent Asian American. The Class of 2028 was fourteen percent Black, sixteen percent Hispanic, and thirty-seven percent Asian American. The Class of 2029 was 11.5 percent Black, eleven percent Hispanic, and forty-one percent Asian American, with the acceptance rate rising to about 4.18 percent from 3.65 percent. Harvard also changed its reporting methodology and now publishes demographics in the fall, which makes clean year-over-year comparison harder.
What did not change. Everything else in this article. The dockets, the readers, the ratings, the subcommittees, the full committee, the lop, the Dean's Interest List, the Z-list, legacy preference, donor preference, and athletic recruiting all survive.
That asymmetry has drawn attention. In July 2023, within weeks of the ruling, the Department of Education's Office for Civil Rights opened a Title VI investigation into Harvard's legacy and donor preferences following a complaint from Lawyers for Civil Rights, noting that roughly seventy percent of legacy and donor-connected applicants are white. Massachusetts has weighed legislation on the issue. Wesleyan, Amherst, Johns Hopkins, and Carnegie Mellon have ended legacy preference. Harvard has not.
Why Harvard Says Everything Is Merely "Considered"
Now the small thing that prompted this article.
Every college files a Common Data Set. Section C7 asks the institution to sort seventeen admissions factors into four buckets: Very Important, Important, Considered, and Not Considered. It is the closest thing that exists to a standardized, self-reported statement of what a college weighs.
Most selective colleges answer it the way you would expect. Rigor of secondary school record and academic GPA land in Very Important. Essays, recommendations, and character land in Important. Interview, first-generation status, and geography land in Considered. Across roughly a hundred selective institutions, eighty-two rate course rigor as Very Important.
Harvard marks essentially everything "Considered." Rigor, GPA, class rank, test scores, essay, recommendations, character, talent, first generation, alumni relation, geography, race, volunteer work, work experience. All in the same bucket. The one factor Harvard affirmatively places in Not Considered is level of applicant's interest, which is accurate, since Harvard does not track demonstrated interest.
It reads like a dodge. I think the truth is more interesting than that, and it comes in three parts.
First, and most importantly, it may be the most literally accurate answer available to them. The Common Data Set format assumes that a factor carries roughly the same weight for every applicant. At Harvard, the trial record shows that assumption is false. An academic rating of 4 ends the process for a typical applicant, at an admit rate of 0.076 percent, and is nearly irrelevant for a recruited athlete, at about seventy percent. If Harvard marked academic record as Very Important, that would misdescribe thirty percent of its admitted class. If it marked it as Considered, that misdescribes the other seventy percent. The overall rating is explicitly a stepping-back judgment rather than a formula, and readers are told as much in the manual. When weights genuinely vary by applicant, the only universally true answer to a fixed-weight question is the vaguest one.
Second, and this is inference rather than documented fact, litigation history rewards caution. Harvard has never publicly explained its C7 answers, so nobody can say this with certainty. But consider what happened to the documents Harvard did produce. Internal research reports, reading manuals, and casual emails all became trial exhibits, quoted in a federal opinion and then in a Supreme Court opinion. A differentiated C7 grid would be a signed, dated, institutional statement about the relative weight of admissions factors, published annually, and available to any future plaintiff who wanted to argue that Harvard's practice contradicts its own disclosures. Filing everything under one heading creates nothing to impeach.
Third, some of those boxes would be uncomfortable to check. The Common Data Set includes "alumni relation" as a factor. Given a legacy admit rate of 33.6 percent against a baseline under 5.5 percent, the empirically honest label is not "Considered." No institution is eager to publish that in a standardized form while a federal civil rights investigation into that exact preference is open.
I would note that Harvard's flat grid is longstanding. It predates the lawsuit and has been stable through it, which argues for the first explanation carrying more weight than the second.
Which brings me to the practical part.
What Applicants Should Actually Do With All of This
I wrote a piece a while back arguing that students should ignore the published admissions rubrics at Harvard, Yale, and Stanford. Nothing in the trial record changes my view. If anything, the record is the strongest evidence for it, and Harvard's own Common Data Set is the closest thing to an official confirmation. When a college declines to tell you what it weighs most, the most likely explanation is that the honest answer is "it depends on the applicant."
Read back through this article and sort it into two columns.
Things you cannot influence: which docket you land in, who reads your file, whether the development office knows your family, whether a coach wants you, whether your parent went to Harvard or works there, and which four variables the committee sees when it trims the class in March. That is most of the article. It includes every mechanism with a large documented effect on admission.
Things you can influence: the rigor of your coursework, the depth and level of your accomplishments, the teachers who write for you, the interview, and your own voice on the page.
The second column is short. It is also, conveniently, exactly what you would work on if the lists had never been disclosed at all. The record does sharpen four specific points.
1. Breadth across categories beats a single spike. This is the most useful actionable finding in the entire case, and it comes from Harvard's own expert. Strength in three or four rating categories produced a seventy percent admit rate. Strength in one or none produced fifteen percent of admits out of seventy-three percent of applicants. The popular strategy of building one hyper-specialized hook and letting everything else slide is contradicted by Harvard's own data. Be genuinely strong academically, genuinely accomplished in something, and someone teachers write about with warmth.
2. Calibrate what the extracurricular ratings actually mean. Harvard's own manual gives class president, newspaper editor, and concertmaster as examples of a 2, not a 1. A 1 means national-level achievement or real professional work. This is a scale for describing outcomes, so use it as a sanity check on where you stand rather than as a to-do list. It measures what you have accomplished, not the path you took to accomplish it.
3. The people writing about you are a scored category. School Support is its own rating, driven entirely by counselor and teacher letters, and the distinction between a 1 and a 2 lives in whether a teacher can say you are the best they have taught in years and then prove it with specifics. That comes from being the student a teacher actually knows, over time, in a class where you did something memorable.
4. Do the interview. The record is unusually blunt on this, in that applicants who do not interview are rarely admitted. Alumni interviewers are also, on the evidence of the case, the least biased evaluators in the process, since their personal ratings showed none of the disparities the office's own ratings did. Take the meeting.
Everything else in this article is worth understanding, because knowing how a system works is better than guessing at it, and because it should relieve some of the pressure students put on themselves when a decision goes the wrong way. A great deal of what happens to your file has nothing to do with you.
Then close the tab and go be excellent at something. The applicants who do best at Harvard are the ones who pursued a real interest far enough to become genuinely distinguished in it, took the hardest curriculum available to them, and gave their teachers something specific and true to write about. That was the answer before the lawsuit, and it is still the answer now that we have seen the manuals.
If you would like help figuring out what that looks like for you, and an honest read on where you stand, schedule a free consultation with us.
