Harvard’s Faculty of Arts and Sciences (FAS) announced major PhD admissions cuts, including a 75% cut in Sciences, later revised to 50% after strong faculty pushback and restored federal funding.

Even at 50%, Harvard’s Science cuts remain far steeper than peer institutions’ (many aren’t cutting STEM).

We estimate Harvard’s 50% Science admissions cut eliminates ~400 PhD seats over two cycles, and — under a worst-case assumption of no new federal research funding — produces present-value savings of only ~$89 million by 2033 (~5% of its annual budget). 

For context, Harvard could save a similar amount by temporarily or permanently not refilling a small number of administrative roles.

If federal funding continues even partially, Harvard’s savings from PhD cuts would likely be smaller than our estimate, and in some cases, could even reduce the University’s income since Science PhDs are often supported by external grants.

Despite the rollback to 50%, shrinking Science cohorts still risks major academic consequences, including reduced faculty research support, strained undergraduate teaching, and a weakened PhD pipeline.

Hedging against a worst-case world where federal funding dries up and students arrive to fewer viable research opportunities

Limiting Harvard’s downside risk if it must suddenly fund a larger share of Science stipends itself

Recalibrating programs for a weaker academic job market, where tenure-track openings have been shrinking for years

Using this moment to rethink the PhD model itself, including what FAS and the Graduate School of Arts and Sciences (GSAS) have described publicly as an unsustainable funding structure

Our model assumes a worst-case scenario in which Harvard receives no new federal research grants while these cohorts get their PhDs, meaning the University bears the full cost of PhD funding. This reflects the lean federal grant landscape and ongoing government actions that could restrict Harvard’s eligibility for future funding. 

We model the PhD costs Harvard truly avoids: stipends and health benefits for ~400 PhDs not admitted. Harvard guarantees health benefits for six years and living stipends for five. When FAS pays for its students, it is “money out the door,” but in the Sciences, federal grants sometimes cover these costs. Assuming no federal funding, Harvard bears the full cost for all students.

Because Harvard wouldn’t generate revenue off of Science PhDs in a no-grant scenario, we exclude PhD tuition from our model. 

We exclude research infrastructure and support costs (labs, facilities, research staff, etc.) because those would likely not change substantially if the cuts last only two years.

We include the cost of replacing PhD-led discussion sections. Harvard PhDs often serve as Teaching Fellows (TFs) for small discussion sections as part of their stipend or graduation requirements. With fewer PhDs, there are fewer TFs, but the number of sections stays the same. Our model assumes Harvard maintains all sections and backfills by paying remaining PhDs the standard ~$6,000-$7,000 per extra section. Replacing ~2,200 sections would add about $13.5 million in instructional costs through FY33. While Harvard could stop paying per-section rates beyond teaching requirements, it has not announced any such change.

FAS’s deficit is large, long-standing, and mostly predates federal actions. In 2024, before any intervention, Harvard economics professors Jeremy Stein and John Campbell estimated a $327 million annual structural deficit, with a present value of $6.54 billion. FAS recently raised that estimate to $350 million, factoring in revised capital costs and the new 8% endowment tax passed through the One Big Beautiful Bill Act. While the tax alone is expected to cost Harvard $300 million annually, the updated figure adds just $25 million to FAS’s annual deficit — highlighting that the core issue is a long-standing imbalance between FAS’s recurring costs and income, not federal pressure.

Because PhD cuts last only two years, they can’t fix a recurring gap this large. Admissions return to normal after two years, which means the savings end once the smaller cohorts graduate. While temporary cuts can relieve short-term budget pressure, they do not address the ongoing, structural mismatch between FAS’s annual costs and income.

The cuts more plausibly address the $700 million research funding risk, but only slightly. Harvard received $686 million in FY24 and $629 million in FY25 from federal grants, so even partial losses would hurt. While a judge restored Harvard’s access to frozen funds, future support is uncertain amid shrinking grants nationally and federal efforts to bar Harvard from future funding. In a worst-case no-grant scenario, annual PhD savings would offset just ~1-3% of the potential $700 million annual exposure, or about $10-20 million per year.

Whether Harvard cuts 50% of Science PhDs or the originally proposed 75%, the financial difference is relatively modest. We estimate the revised cut saves Harvard a present value of $89 million over seven years, compared to $108 million under the original plan, a total difference of ~$20 million.

Harvard might want to limit its worst-case financial exposure if federal grants stop covering Science PhD stipends. If grants disappear, Harvard must cover five years of stipends and health benefits (together about $56,000 annually) for every PhD across FAS. In the Sciences, these stipends have often been grant-funded beginning in year 2 of PhD programs (G2). Reducing cohort size mitigates this worst-case cost.

With $700 million in annual federal funds at risk, Harvard may want to avoid enrolling Science PhDs whose labs or job outcomes could be compromised. While FAS will be supported by an extra $250 million this year in “continuity funding” for research, without new federal funding, students may still lack the research opportunities and funding to produce strong work. The lack of such opportunities could weaken their post-program placements, in turn discouraging future applicants and creating a feedback loop that erodes Harvard’s brand. Shrinking cohorts may hedge against this risk.

Harvard is responding to a long-run slowdown in tenure-track openings for new PhDs. Science fields now produce roughly 10 PhDs for every faculty opening. In the humanities, declining undergraduate enrollments have reduced tenure-track opportunities. At Harvard, more than 30% of recent Science PhDs now move directly into industry, where salaries can absorb them; Humanities PhDs lack a comparable high-paying alternative. With a mismatch of supply and demand for PhDs, and uneven industry safety nets, Harvard may be reconsidering how much it spends to subsidize future professors (though this wouldn’t explain its outsized cuts in the Sciences). 

Harvard has already linked smaller cohorts to a broader review of its PhD model. In 2021-22, the University cut PhD admissions to help reverse its “unsustainable” use of unrestricted endowment funds to cover PhD tuition. In announcing this year’s cuts, FAS Dean Hopi Hoekstra wrote that this year’s cuts would coincide with a review of “the future model of Ph.D. education.” Given that no peer institutions have cut anywhere near Harvard’s 50-60% across multiple PhD divisions, the University’s outlier move might signal it sees this as a chance to reset its PhD programs, cutting now and rebuilding later on its own terms.

Faculty retention may suffer if they lose graduate-student support. Many faculty, especially in the Sciences and Social Sciences, rely on graduate students to sustain their labs and research. Losing that support, even temporarily, may disrupt projects and drive faculty to look elsewhere (e.g., to MIT), especially as FAS signals possible broader changes to the PhD model. A decline in faculty output and reputation may make it harder to recruit top faculty going forward. 

Shrinking PhD cohorts risk weakening Harvard’s PhD student body. Smaller cohorts may help ensure the remaining PhDs have research opportunities and funding, but they may still lag behind peers who are not cutting at similar rates. That could deter top applicants across divisions, who look for robust academic communities, research opportunities, and mentorship networks.

Harvard’s national academic influence could erode. A 2022 study found that just five universities, including Harvard, produce nearly 14% of U.S.-trained tenure-track faculty. While Harvard may overproduce Humanities PhDs relative to academic jobs available (partly because its grads may avoid accepting roles at less prestigious institutions), it may still underproduce Science PhDs relative to demand. Cutting across the board could undermine Harvard’s unique position in training top-tier faculty; arguably, even if weaker programs contract to meet reduced market demand, Harvard could continue investing in academic leadership.

Shrinking PhD cohorts leave hundreds of course sections without TFs. Covering those gaps requires trade-offs that can harm both graduate training and undergraduate learning. These include: 

On FAS PhDs: Dean Hoekstra noted that the cohort reductions would accompany a broader review of “the future model of Ph.D. education,” and that the decision to continue admissions came only “after careful deliberation.”

On FAS Administration: She also described FAS as entering a phase of “strategic redesign and long-term sustainability,” with a task force formed to reconsider the administrative structure. These developments point to more extensive changes ahead, including potential layoffs and longer-term reductions to PhD programs. Yale announced today that departments may need to downsize or lay off staff  in response to the same 8% endowment tax facing Harvard.