Federal research funding per doctorate: ~$1.1-1.7M nationally, but life-sciences PhDs draw 15-17x more federal R&D money than business, education, or humanities PhDs

Summary

Nationally, federal research money runs from $1.1 million to $1.7 million for every research doctorate the US produces, depending on which federal-funding total is used as the numerator. Using NCSES's all-performer federal research obligations (universities, industry, national labs, federal agencies combined: $99.6B in FY2024) against the 58,131 research doctorates awarded in 2024, the ratio is $1.71 million per doctorate. Using only the federal R&D money that actually flows to universities -- the institutions that produce doctorates -- the ratio is $1.11 million per doctorate ($64.6B FY2024 HERD federal expenditure / 58,131 doctorates). Neither number is a real per-student training cost: most federal research funding pays for faculty, staff, equipment, and non-degree research activity, not tuition or stipends for the students who eventually graduate.

The variation by field is large: life sciences (biology, health sciences, agricultural sciences combined) draws about $2.67 million in federal university R&D funding per doctorate -- roughly 17 times the ~$158,000 per doctorate that non-S&E fields (business, education, humanities) draw. Engineering, physical sciences, and computer science sit in the middle, at roughly $0.87M-$1.09M per doctorate.

Two different, defensible ways to define the numerator

"Federal research funding" can mean two genuinely different things, and they produce ratios nearly 60% apart:

• All-performer federal research obligations (NCSES Survey of Federal Funds for R&D, Table 91): total federal obligations for basic + applied research, across every performer type -- universities, industry, federal intramural labs, FFRDCs, nonprofits. FY2024 = $99,556 million. This is the true "federal research spend" figure, but roughly a third of it never touches a university and therefore has no direct bearing on doctoral training.
• University-performed federal R&D expenditure (NCSES Higher Education R&D Survey, HERD): federal dollars actually expended at degree-granting universities (research + development combined; HERD does not split research from development by field). FY2024 federal share = $64,640 million. This is the money reaching the institutions that award doctorates, and its field taxonomy lines up much better with SED's own field categories, so it is used for the field-level breakdown below.

A third possible denominator -- federal S&E obligations to "universities and colleges" as a performer class in the Federal Funds survey -- was reported by NCSES at roughly $47 billion for FY2024 (a 4.3% decline from FY2023), which would put the ratio closer to $0.81 million per doctorate. The $47B and $64.6B figures differ because they come from two different NCSES surveys (the agency-reported Federal Funds survey vs. the institution-reported HERD survey), which have known, published reconciliation differences. All three numbers are legitimate; none of them is "the" federal cost of producing a PhD, because federal research funding is not allocated with doctoral production as its stated purpose.

Field-level detail: what drove the field ranking

Life sciences' dominance reflects NIH's outsized role in federal research: NIH funds large wet-lab operations (equipment, core facilities, postdocs, technicians) that support many more researchers than doctoral students per dollar, but the field still produces a large volume of PhDs, so the ratio stays high on both sides. Geosciences shows the highest per-doctorate ratio of any field despite modest total funding, because it produces very few doctorates (1,229, the smallest field in this comparison) relative to the capital-intensive nature of oceanographic and atmospheric research infrastructure. At the other end, business, education, and the humanities receive comparatively little federal research money -- these fields are funded overwhelmingly by tuition, private philanthropy, and institutional funds rather than federal grants -- yet still produce over 12,000 doctorates a year combined, driving their per-doctorate ratio to a small fraction of the STEM fields'.

Why this is an association, not a training-cost estimate

This ratio is a description of two independently-produced statistics divided by each other, not a causal or accounting relationship. Federal research grants pay for a lab's operations -- principal investigators, postdocs, staff scientists, equipment, and materials -- of which doctoral student support (tuition remission and stipends via research assistantships) is only one line item. A field can have a high dollars-per-doctorate ratio because it is genuinely capital- and staff-intensive (geosciences, life sciences) rather than because it is expensive to train a given student. Conversely, a low ratio in the humanities does not mean those PhDs cost little to produce -- it means their funding overwhelmingly comes from non-federal sources (institutional funds, teaching assistantships, tuition waivers), which this ratio does not count at all in the denominator's cost side.

The SED and HERD/NCSES field taxonomies also do not align 1:1. SED's "biological and biomedical sciences," "health sciences," and "agricultural sciences and natural resources" were combined to match HERD's single "Life sciences" bucket; SED's "education," "humanities and arts," "business," and "other non-S&E fields" were combined to match HERD's "Non-S&E" bucket. NCSES's own "Other sciences" HERD category (~$499M) has no clean SED counterpart and was excluded from the field table rather than force-matched.

Every query behind this report

4 warehouse calls ran in this session, in order. Each is reproducible against the same snapshot.

SELECT "year", rd_field, field_level, obligations_usd_million FROM research.nsf_rd_by_field WHERE "year" = (SELECT MAX("year") FROM research.nsf_rd_by_field WHERE field_level=1) AND field_level IN (0,1) ORDER BY obligations_usd_million DESC NULLS LAST

SELECT "year", rd_field, field_level, obligations_usd_million FROM research.nsf_rd_by_field WHERE "year" = 2024 AND field_level IN (0,1) ORDER BY obligations_usd_million DESC NULLS LAST

SELECT DISTINCT rd_field FROM research.nsf_herd_by_institution WHERE "year" = 2024 ORDER BY rd_field

SELECT rd_field, SUM(rd_expenditure_usd_thousand)/1000.0 AS federal_rd_usd_million
FROM research.nsf_herd_by_institution
WHERE "year" = 2024 AND funding_source = 'Federal' AND federal_agency = 'Total'
  AND rd_field IN ('All','Computer and information sciences, all','Engineering, all','Geosciences, atmospheric sciences, and ocean sciences, all','Life sciences, all','Mathematics and statistics, all','Non-S&E, all','Other sciences, all','Physical sciences, all','Psychology, all','Social sciences, all')
GROUP BY rd_field
ORDER BY federal_rd_usd_million DESC