Harvard received 47,893 applications for the class that entered in Fall 2025. Three cycles earlier, for the class that entered in Fall 2022, it received 61,221. The gap between those two figures is 13,328 applications, and it opened across three consecutive years of decline with no year of recovery in between.
Selectivity moved with it. Harvard’s admit rate bottomed at 3.24 percent for Fall 2022 and has risen every year since, to 3.45, then 3.65, then 4.18 percent. A rising admit rate at Harvard is unusual enough on its own. A rising admit rate driven entirely by a shrinking pool, during a period when applications at comparable universities were climbing, is the central fact of this five-year record.
Through all of it, one number barely moved. Harvard’s yield, the share of admitted students who enroll, landed between 82.96 percent and 84.17 percent in every one of the five filings. No other figure in this report is that stable, and none explains more about how the university operates.
How to read this record
Five filings carry the analysis: 2021-22 through 2025-26. Harvard published the most recent one on July 13, 2026. Readers should hold three qualifications in mind.
The Fall 2021 entering class of 1,951 students is larger than any that follows, and the reason is the pandemic. Students who deferred admission from Fall 2020 arrived alongside the new cohort. First-year retention that year reads 96 percent against 98 percent in every other filing. Treat 2021-22 as a distorted baseline rather than a normal year, and note that the application peak in the next cycle sits on top of a period of unusual application volume across selective admissions generally.
Several fields are blank. The 2025-26 filing reports no residency breakdown, no early action counts, and no cost of attendance at all, because Harvard’s 2026-27 prices were not set when the form was filed. Residency detail appears in only two of the five years. Early action counts appear in only one.
One field cannot be read reliably, and the honest thing is to say so rather than guess. Harvard marks C7, the table rating the importance of each admission factor, with checkboxes in four of the five filings, and the machine-readable text of those files preserves the marks but not which column they sit in. This report presents C7 only for 2023-24, the year Harvard filed written ratings instead. What is legible across all five years is which rows appear on the form, and that turns out to carry its own news.
Applications, admits, and the number that never moves
| 2021-22 | 2022-23 | 2023-24 | 2024-25 | 2025-26 | |
| Entering class | Fall 2021 | Fall 2022 | Fall 2023 | Fall 2024 | Fall 2025 |
| Applications | 57,786 | 61,221 | 56,937 | 54,008 | 47,893 |
| Admitted | 2,318 | 1,984 | 1,965 | 1,970 | 2,003 |
| Admit rate | 4.01% | 3.24% | 3.45% | 3.65% | 4.18% |
| Enrolled | 1,951 | 1,646 | 1,645 | 1,647 | 1,675 |
| Yield | 84.2% | 83.0% | 83.7% | 83.6% | 83.6% |
| First-year retention | 96.0% | 98.0% | 98.0% | 98.0% | 98.0% |
Source: Harvard University Common Data Set filings, sections C1 and B22, entering classes of Fall 2021 through Fall 2025. Admit rate = students admitted divided by applications. Yield = students enrolled divided by students admitted. Gender detail cross-foots to the filed totals in every year.
Applications and admit rate, Harvard University, five Common Data Set filings
The admitted group is remarkably steady. Setting aside the inflated 2021-22 figure, Harvard admitted 1,984, then 1,965, then 1,970, then 2,003 students. Across four cycles the largest year-over-year change in the number of offers was 38 students. Harvard is not responding to a smaller pool by admitting fewer people, and it is not chasing a lower admit rate by admitting fewer people either. It decides how many students it wants and issues close to the same number of offers regardless of how many applied.
Students admitted and yield rate, Harvard University, five Common Data Set filings
That is only possible because yield holds. When roughly 84 percent of admitted students enroll every year, the university can predict its class size from its offer count with a margin of a few dozen students. Duke, examined earlier in this series, peaked at 60.1 percent yield and had to move its admitted group by more than 200 students across five years to hold enrollment flat. Harvard has the opposite problem, which is to say almost no problem at all.
The waiting list confirms it. Harvard reported no waiting list activity in four of five filings and admitted 75 students from the list for Fall 2025. A university that converts five of every six offers has little need for one.
A class with no early decision
Harvard answers no to C21 in every filing. There is no binding early decision plan and never has been in this window. What Harvard runs instead is restrictive early action, a non-binding plan that limits where else a student may apply early. Applicants admitted in December are free to compare offers and decline.
This makes the round-selection advice that governs Duke, Vanderbilt, and Villanova inapplicable here. There is no commitment to trade for a higher admit rate, and no aid comparison to forfeit. Applying early to Harvard costs an applicant the ability to apply early to most other private universities, and returns a decision in December rather than March.
Harvard reported early action counts in exactly one of the five filings. For Fall 2023 it received 9,544 early applications, admitted 721, and enrolled 677 of them. That single year is worth sitting with. The early admit rate of 7.6 percent runs more than double the overall rate of 3.45 percent. Ninety-four percent of early admits enrolled, a yield far above the university’s already extraordinary average. And those 677 students filled 41 percent of the entering class.
| Fall 2023 entering class | Early action | All applicants |
| Applications | 9,544 | 56,937 |
| Admitted | 721 | 1,965 |
| Admit rate | 7.56% | 3.45% |
| Enrolled | 677 | 1,645 |
| Yield on admits | 93.9% | 83.7% |
Source: Harvard University Common Data Set 2023-24, sections C1 and C22. Harvard reported early action application, admit, and enrollment counts in this filing only; the other four filings in this window leave those fields blank. Early action at Harvard is restrictive and non-binding.
Whether that pattern held in the years Harvard left blank cannot be determined from the Common Data Set. The university stopped releasing admissions figures during the cycle after 2023 and moved to a single annual release each fall, so the December early action numbers that once appeared in press coverage no longer do.
Testing came back, and the score bands fell
| 2021-22 | 2022-23 | 2023-24 | 2024-25 | 2025-26 | |
| Submitted SAT | 54% (1,049) | 55% (903) | 52% (862) | 54% (892) | 79% (1,325) |
| Submitted ACT | 31% (604) | 28% (466) | 22% (356) | 19% (321) | 27% (450) |
| SAT reading, 25th to 75th | 730-780 | 730-780 | 740-780 | 740-780 | 720-770 |
| SAT math, 25th to 75th | 750-800 | 760-800 | 760-800 | 770-800 | 730-790 |
| ACT composite, 25th to 75th | 33-36 | 34-36 | 34-36 | 34-36 | 33-35 |
| Average high school GPA | 4.20 | 4.20 | 4.20 | 4.21 | 4.22 |
Source: Harvard University Common Data Set filings, sections C9 and C12. Percentages express submitters as a share of the enrolled first-year class; students submitting both tests appear in both rows. Harvard reported no SAT composite range in any of the five filings, reporting section scores only.
Share of enrolled first-year students submitting each test, Harvard University
Harvard was test-optional for the first four filings in this window and reinstated a testing requirement for students applying to enter in Fall 2025. The effect is visible in a single year. SAT submission jumped from 54 percent of the enrolled class to 79 percent, and the count rose from 892 students to 1,325.
SAT section scores at the 25th percentile of the enrolled class, Harvard University
The score bands went the other way. The 25th percentile SAT math score had climbed from 750 to 770 across the test-optional years, then dropped to 730 the year submission became mandatory. Reading followed the same path, from 740 down to 720. Both 75th percentile figures fell by 10 points as well.
That reversal is a reporting effect rather than a decline in the class. Under a test-optional policy, students choose whether to submit, and students with lower scores mostly do not. Requiring scores adds those students to the reported distribution for the first time. The published range widened at the bottom because the bottom finally became visible. Applicants comparing a 1490 to last year’s band should understand they are comparing against two different populations.
Who enrolls
| First-year students | 2021-22 | 2022-23 | 2023-24 | 2024-25 | 2025-26 |
| White | 650 (33%) | 476 (29%) | 458 (28%) | 457 (28%) | 480 (29%) |
| Asian | 431 (22%) | 403 (25%) | 411 (25%) | 375 (23%) | 462 (28%) |
| International | 287 (15%) | 252 (15%) | 250 (15%) | 265 (16%) | 251 (15%) |
| Hispanic or Latino | 228 (12%) | 193 (12%) | 179 (11%) | 183 (11%) | 139 (8%) |
| Black or African American | 186 (10%) | 165 (10%) | 169 (10%) | 112 (7%) | 97 (6%) |
| Two or more races | 139 (7%) | 123 (7%) | 137 (8%) | 127 (8%) | 139 (8%) |
| Race or ethnicity unknown | 19 (1%) | 20 (1%) | 33 (2%) | 117 (7%) | 102 (6%) |
Source: Harvard University Common Data Set filings, section B2, degree-seeking first-time first-year column. Percentages calculated against first-year totals of 1,949, 1,644, 1,644, 1,641, and 1,673. American Indian or Alaska Native and Native Hawaiian or Other Pacific Islander counts, in single digits throughout, are omitted for space. The Fall 2021 column reflects an enlarged pandemic-era class.
The unknown row has to be read first. Students declining to report race numbered 19 and 20 in the first two filings, then 33, then 117, then 102. That is a sixfold increase concentrated in the two years following the 2023 Supreme Court decision on race-conscious admission, and it is large enough to distort any single category.
With that stated, the movement in two rows exceeds what reporting change alone accounts for. Black first-year enrollment fell from 186 students to 97, a decline of 48 percent across the window, with the sharp break arriving in the Fall 2024 class. Hispanic enrollment fell from 228 to 139. Asian enrollment rose to 462, its highest figure in the period. International enrollment held between 15 and 16 percent every single year.
Residency: the two years Harvard showed its work
| Applications | Admits | Admit rate | Enrolled | |
| Fall 2023, Massachusetts | 3,083 | 241 | 7.82% | 210 |
| Fall 2023, out of state | 37,920 | 1,423 | 3.75% | 1,167 |
| Fall 2023, international | 15,934 | 301 | 1.89% | 268 |
| Fall 2024, Massachusetts | 3,032 | 265 | 8.74% | 227 |
| Fall 2024, out of state | 34,216 | 1,380 | 4.03% | 1,125 |
| Fall 2024, international | 16,760 | 325 | 1.94% | 295 |
Source: Harvard University Common Data Set filings 2023-24 and 2024-25, section C1 residency table, reported by known physical address at time of application. Admit rate = admits divided by applications. The 2021-22 and 2022-23 forms did not collect this breakdown and the 2025-26 filing left it blank.
Admit rate by residency category, Harvard University, Fall 2023 and Fall 2024 entering classes
International applicants faced an admit rate below 2 percent in both years Harvard reported the split. That is less than a quarter of the rate for Massachusetts residents and roughly half the rate for domestic applicants from other states. The international pool grew from 15,934 to 16,760 while the out-of-state pool fell by more than 3,700, meaning international applicants made up a rising share of a shrinking total.
Home-state applicants hold a real advantage at a university with no in-state mandate of any kind. Harvard is private and receives no state appropriation tied to Massachusetts enrollment, yet Massachusetts residents were admitted at more than twice the out-of-state rate in both reported years.
Price, aid, and borrowing
| 2021-22 | 2022-23 | 2023-24 | 2024-25 | 2025-26 | |
| Costs reported for | 2022-23 | 2023-24 | 2024-25 | 2025-26 | not reported |
| Tuition | $52,659 | $54,269 | $56,550 | $59,320 | n/a |
| Required fees | $4,602 | $4,807 | $5,126 | $5,476 | n/a |
| Housing and food | $19,502 | $20,374 | $21,190 | $22,130 | n/a |
| Total published cost | $76,763 | $79,450 | $82,866 | $86,926 | n/a |
| First-years with need | 797 | 1,091 | 938 | 920 | 932 |
| Average need-based grant | $61,818 | $65,595 | $67,949 | $73,497 | $72,749 |
| Average total aid package | $62,252 | $67,898 | $70,478 | $74,387 | $75,166 |
| Share of need met | 100% | 100% | 100% | 100% | 100% |
Source: Harvard University Common Data Set filings, sections G1 and H2. Each filing reports costs for the following academic year. Aid figures cover full-time first-time first-year students. The 2025-26 filing left G1 blank because Harvard’s 2026-27 prices were not set at the time of filing. The 2022-23 filing reports an aid cohort of 1,982 first-year students against an enrolled class of 1,646, an inconsistency carried here as filed.
Published cost of attendance by component, Harvard University
Published tuition rose 12.6 percent across the four years Harvard reported it, from $52,659 to $59,320. Total published cost rose 13.2 percent to $86,926. Both are modest increases by the standards of this series. Duke’s tuition rose 22 percent over a comparable span.
Published tuition against average need-based grant for first-year aid recipients, Harvard University
The average need-based grant to a first-year aid recipient rose from $61,818 to $72,749, an increase of 17.7 percent that outpaced tuition growth. In every year of the window that average grant exceeded published tuition, meaning the typical aided first-year student received more in grant money than the university charged for instruction.
The affordability thresholds moved twice. The 2023-24 filing describes attendance becoming free for families with annual incomes below $85,000, up from $75,000, plus a $2,000 grant to juniors from those families. The 2025-26 filing raises the free threshold to $100,000 and adds full tuition coverage for families under $200,000. Harvard met 100 percent of demonstrated need in all five years and awarded no merit aid worth counting, with non-need-based grants going to between one and nine first-year students annually.
Share of graduating class borrowing and average cumulative amount borrowed, Harvard University
Borrowing is rare and getting more expensive for those who do it. The share of graduates leaving with any loan debt fell from 21 percent to 16 percent, while the average cumulative amount borrowed rose from $16,366 to $22,926. Roughly one Harvard graduate in six borrows anything at all.
Enrollment, faculty, and class size
Total undergraduate enrollment and bachelor’s degrees conferred, Harvard University
Undergraduate enrollment declined across the window, from 7,153 to 6,738, a loss of 415 students or 5.8 percent. Bachelor’s degrees conferred moved the other way, from 1,292 to 2,014, as the enlarged pandemic-era cohorts reached graduation and the deferral backlog cleared.
Full-time instructional faculty held between 1,774 and 1,849 in every year, ending at 1,796. Undergraduate class sections ranged from 1,014 to 1,105. Sections enrolling fewer than 20 students accounted for 70.6 percent of the total in Fall 2021 and 73.6 percent in Fall 2025, while sections of 100 or more fell from 54 to 49.
The reported student-to-faculty ratio reads 7 to 1 in the first four filings and 11 to 1 in the fifth. That change reflects a revision to the form, which now asks institutions to include graduate students and the faculty who teach them. Harvard noted an undergraduate ratio of 7 alongside the new figure. Comparing the 2025-26 ratio to earlier years measures a form revision rather than a classroom.
What Harvard students study
| Share of bachelor’s degrees | 2021-22 | 2022-23 | 2023-24 | 2024-25 | 2025-26 |
| Social sciences | 26% | 26% | 28.8% | 27% | 27% |
| Mathematics and statistics | 12% | 12% | 11.1% | 12% | 14% |
| Biological and life sciences | 14% | 14% | 12.2% | 13% | 13% |
| Computer and information sciences | 11% | 11% | 9.8% | 11% | 11% |
| History | 9% | 9% | 8.8% | 8% | 7% |
| Physical sciences | 8% | 6% | 8.2% | 8% | 6% |
| Psychology | 4% | 4% | 5.2% | 4% | 5% |
| Engineering | 3% | 5% | 4.5% | 5% | 4% |
| English | 3% | 3% | 2.3% | 2% | 2% |
Source: Harvard University Common Data Set filings, section J1, degrees conferred in the year preceding each filing. Percentages use majors rather than headcount, so double majors appear twice. Four of the five filings report whole percentages; 2023-24 reports one decimal place.
Share of bachelor’s degrees conferred by discipline, 2021-22 filing against 2025-26 filing
Harvard’s distribution is the most stable this series has examined. Social sciences, which at Harvard means economics, government, sociology, and related concentrations, has held between 26 and 29 percent of degrees in every year. No category moved more than three percentage points across five filings. Mathematics and statistics gained the most, rising from 12 to 14 percent. History and English both declined, from 9 to 7 percent and from 3 to 2 percent.
Two absences say as much as the numbers. Harvard confers no degrees at all in business, health professions, education, communication, or public administration. The concentrations that dominate undergraduate enrollment at most American universities do not exist here. Computer science has held at 11 percent without the surge that carried it to 18 percent of Duke’s degrees, which reflects a curriculum that routes quantitative students through applied mathematics and statistics as readily as through computer science.
Beyond the numbers
Harvard admits to one undergraduate college. There is no separate engineering school admission, no direct entry to a business program, and no application by intended major. Students declare a concentration in the fall of sophomore year. This removes an entire layer of strategy that shapes applications to universities with divided undergraduate colleges, where the choice of school can move an applicant’s odds by several percentage points. At Harvard the intended field of study carries no admissions weight of its own, and an applicant cannot improve or damage their chances by selecting one.
The application fee rose from $75 to $85 to $90 across the window, and fee waivers remain available on the basis of financial need.
Transfer admission is effectively closed. Across five years Harvard received between 1,735 and 2,256 transfer applications annually and admitted between 13 and 16 students. The transfer admit rate ran from 0.64 percent to 0.86 percent, well below the first-year rate in every year. Applicants who treat transferring as a fallback should understand the arithmetic.
The application decline invites explanation the Common Data Set does not provide. The filings record what happened without recording why, and the period covered was one of sustained national attention to Harvard, including federal actions concerning research funding and international student visas during 2025. Whether and how much those events moved application volume cannot be established from this data, and the decline began before them, with the drop from Fall 2022 to Fall 2023 arriving first. What the record does establish is that the fall was steady rather than sudden, and that it did not change the size or the yield of the class Harvard enrolled.
One more absence deserves mention. Harvard’s 2023-24 filing, the only year with written factor ratings, rates every academic and non-academic factor it considers at the same level. Rigor of record, grade point average, essays, recommendations, extracurricular activities, talent, and character all sit at considered. Nothing is marked very important. Four factors are marked not considered: class rank, state residency, religious affiliation, and level of applicant’s interest. Where Duke distinguishes six factors as very important and moved test scores down a tier in a traceable way, Harvard declines to rank at all. Applicants looking to that table for guidance on what to emphasize will find none.
Across the full window one row did change. Racial and ethnic status appears as a factor on the 2021-22 and 2022-23 forms and disappears from the 2023-24 filing onward, following the Supreme Court decision issued in June 2023.
What the five-year record means for applicants
Start with yield, because it governs everything else. Harvard enrolls roughly 84 percent of the students it admits, so it issues almost exactly as many offers as it wants students. There is no large admitted group waiting to be whittled down by competing offers, and no reason for the university to over-admit. An applicant is competing for one of about 2,000 offers, and that figure has not varied by more than 40 in four years no matter what happened to the applicant pool.
A rising admit rate is not a softening. Harvard’s rate went from 3.24 to 4.18 percent because 13,328 fewer people applied, not because standards moved. The class that resulted was the same size, with the same retention and the same academic profile, as every other class in this record.
Send scores, and read the published ranges carefully. Testing is required again, so the choice has been removed. The reported bands fell when submission became mandatory, which means the current 25th percentile of 730 in math describes a fuller and more representative population than the 770 reported two years earlier. The older figures were never a target; they were a filtered view.
International applicants should calibrate against 1.9 percent rather than the headline rate. That gap, more than four percentage points below the Massachusetts figure and half the out-of-state figure, is the largest structural difference in the data and it held steady in both years Harvard reported it.
Do not build a strategy around the concentration you intend to study. Harvard admits to a single college and asks students to choose a field sixteen months after they arrive. And do not build one around transferring in later, at a rate under 1 percent. What remains is the application itself, evaluated against a factor table that tells applicants almost nothing about which parts count more.
