Don’t Follow Harvard on Grade Caps (opinion)

On May 19, Harvard University’s Faculty of Arts and Sciences voted 458 to 201 to cap flat A grades in undergraduate courses, limiting the total number of A grades to 20 percent of course enrollment, plus four students. Other institutions watching elite peers for cover will be tempted to follow. They shouldn’t. The closest institutional analogues have produced documented harm. Even Harvard’s own 2023 grading report warned against policies of this kind.

Harvard’s case for the cap might feel intuitive: If everyone gets an A, an A means nothing. But that intuition confuses meaning with scarcity. If we’re going to use grades at all, a grade should report what a student has demonstrated against a defined standard—not where they finished in a class ranking. If Harvard wants grades to mean more, the answer is pedagogical: clearer standards, not rarer A’s. Rarity is not rigor.

The cap moves the opposite way: away from standards and toward peer ranking. It installs a fixed top-end quota on the highest letter grade in undergraduate courses, with the maximum number of seats at the top decided in advance. Harvard’s grading subcommittee predicts the cap will reduce anxiety and encourage intellectual exploration. Research on similar systems suggests otherwise. A preregistered experiment found that norm-referenced grading—where students are ranked against their classmates instead of measured against a fixed standard—produced lower mastery orientation, weaker self-efficacy and less help-seeking and help-giving among students. When Cornell University began publishing median course grades online, follow-up research found increased enrollment in more leniently graded courses—a pattern consistent with strategic, grade-protective course selection. Neither study supports the prediction that students would become less strategic or more exploratory. Both describe incentives that pull students toward positioning rather than learning.

The clearest warning, though, comes from equity research. Wellesley College already ran the closest grade-capping experiment: a ceiling of B-plus for mean grades in 100- and 200-level courses with at least 10 students, starting in fall 2004. Wellesley’s own economists analyzed what the cap did to grades, enrollments, majors and faculty evaluations. The overall grade decline in departments affected by the policy was about 0.17 grade points. For Black students, it was 0.36—more than double. Lower-level enrollments in those departments fell about 18 percent; majors fell about 30 percent.

The percentage of students graduating magna cum laude in treated departments dropped from 20 to 16 percent. Course evaluations of the affected faculty fell. The costs were concentrated on Black students, students with lower incoming test scores and the faculty in affected departments. Harvard’s policy isn’t identical to the one that was in place at Wellesley, which the college rescinded in 2019. But the basic logic—a numerical ceiling on top grades—is the same. Harvard would apply it more broadly.

The evaluation hit to faculty wouldn’t land evenly, either. A study using random assignment of students to instructors found that male students rate female instructors about 21 percent of a standard deviation lower than male instructors, even though the instructor’s gender impacted neither students’ grades nor the number of hours they spent studying.

Harvard’s own Office of Undergraduate Education has flagged an adjacent mechanism—that expected grades and student evaluations of teaching move together. Its 2023 grading report noted, “There is a direct correlation between expected grades and Q [teaching evaluation] ratings that may be contributing to trends towards higher grades.” The same report concluded that “institution-wide policies such as capping the percentage of A grades or publishing median course grades on transcripts tend to lead to unintended negative consequences.”

Three years later, Harvard voted to install such a policy anyway. If a cap forces instructors to give lower grades than they otherwise would, enforcement will be paid in evaluation scores by faculty already exposed to documented bias in those evaluations. The cap can reproduce this inequality without intending to.

Princeton University tried the same governing logic—numerical targets—and abandoned it. Princeton set numerical A-range targets in 2004 and removed them in 2014, after the review committee said the targets had been “too often misinterpreted as quotas” and had added “a large element of stress to students’ lives, making them feel as though they are competing for a limited resource of A grades.”

Princeton’s faculty saw this. So have Harvard’s own students. In a self-selected Harvard Undergraduate Association survey of nearly 800 respondents, 84.9 percent “definitely” opposed the cap on A grades. More than 72 percent definitely or probably opposed the separate percentile-rank system Harvard will use to decide internal honors.

There are alternatives that don’t ration the A. Criterion-referenced grading ties grades to demonstrated standards rather than to peer rank. Transparent rubrics, standards-based assessment and mastery-oriented course design all preserve rigor without treating the classroom as a tournament. Each keeps the focus where teaching ought to be: on what students have learned, not on where they finished in line.

Harvard’s grading subcommittee said the cap would mean “an A- need no longer be a source of anxiety, encouraging students to explore new subjects and take intellectual risks.” It is hard to read the Wellesley data, the Princeton reversal, the Cornell median-grades experiment and Harvard’s own 2023 OUE report and believe that a quota system makes students less anxious or less strategic. The closest evidence has already given the warning. Harvard’s own report gave the same warning three years ago. We know how this ends.

Don’t follow.