Federal academic research funding is moderately and slowly increasingly concentrated: the top 10 universities take about 21%, the top 50 about 61%, of all federal R&D dollars — up modestly from 2010

Summary

Federal research funding to universities is heavily concentrated by dollar share but spread across a long tail of institutions — the two standard concentration metrics tell different parts of the story and both are needed. In FY2024, the top 10 recipient universities (led by Johns Hopkins, at $3.6B, more than 3x the #2 school) took 20.8% of all $64.6B in federal academic R&D; the top 50 took 60.8%; the top 100 took 81.2% — out of roughly 650-900 institutions that report any federal R&D spending each year. The Gini coefficient for the full distribution is 0.82-0.83 (extreme inequality), essentially unchanged since 2010. Concentration at the very top has drifted up modestly, not dramatically: the top-10 share rose from 19.7% (2010) to a peak of ~21.2% (2016-2020) and has eased slightly to 20.8% (2024); the top-50 share rose more steadily, from 57.8% to 60.8%. The Herfindahl-Hirschman Index (HHI) — which is dominated by the single largest players rather than the whole distribution — rose from 98 to a 2019-2020 peak of ~115 and has since eased to ~111; by the DOJ/FTC merger-guideline scale this stays in the "unconcentrated" band (<1000) throughout, because even the #1 recipient never holds more than about 5.6% of the national total. In short: concentration is real and persistent (a small set of research universities dominates, and Gini says it always has), but the trend over 2010-2024 is one of slow, modest drift toward more concentration at the top, not a sharp or ongoing consolidation.

Method

Data: NSF NCSES Higher Education R&D (HERD) Survey, institution-level microdata (research.nsf_herd_by_institution), filtered to funding_source='Federal' AND federal_agency='Total' AND rd_field='All' — the true federal-funding grand total per institution per year (avoiding double-counting from the table's field/agency rollup hierarchy). Coverage is FY2010-FY2024 (15 contiguous years; the table's declared 2006-2009 window is not actually loaded, per data_coverage). National totals reproduced from this table ($37.5B in 2010 rising to $64.6B in 2024, federal share of ~55% of the $117.5B FY2024 total) match NSF's own December 2025 HERD release and independent press coverage (Forbes/NCSES), which is a genuine external cross-check, not just an internal consistency check.

On the NOT NULL filter used in the HHI/Gini calls: every institution-level total query included a rd_expenditure_usd_thousand IS NOT NULL (and, for HHI/Gini specifically, > 0) predicate. Checked directly: for both 2010 and 2024, the (funding_source='Federal', federal_agency='Total', rd_field='All') slice contains zero NULL rows and zero rows equal to 0 out of 737 (2010) and 674 (2024) total institution rows — so this predicate excluded nothing in practice for either year reported here; the top-N-share and HHI/Gini figures above are computed on the full population of institutions reporting a federal total that year, with no units silently dropped.

Two complementary concentration measures were computed for every year 2010-2024, across the full population of reporting institutions each year (no top-N sampling):

• Top-N cumulative share (top 10 / 20 / 50 / 100 institutions' share of the national federal-R&D total) — answers "how much sits with the leaders."
• HHI (sum of squared percentage-point shares, DOJ/FTC 0-10,000 scale) — sensitive mainly to the single largest players.
• Gini coefficient (0=perfectly even, 1=one unit holds everything) — sensitive to the shape of the whole distribution, including the long tail of low- or no-federal-funding institutions.

HHI and Gini diverge sharply here (HHI ~100-115, "unconcentrated"; Gini ~0.82, "extreme") precisely because they measure different things: no single university ever holds more than ~5.6% of the federal total (so HHI stays low), but hundreds of institutions split a tiny remainder among them (so Gini stays high). Both are reported because a single number would mislead in either direction.

Level: how concentrated is it today (FY2024)

Johns Hopkins alone (largely driven by its management of the Applied Physics Laboratory, a federally funded R&D center) holds roughly 5.6% of the entire national federal academic R&D total — more than 3x the #2 institution, Georgia Tech. The top 10 together hold 20.8%; top 20, 34.6%; top 50, 60.8%; top 100, 81.2%. The remaining ~550-800 institutions that report federal R&D split the final ~19% of the total, and Gini = 0.82-0.83 confirms the bulk of institutions receive a very small amount each.

Trend: how much has this changed since 2010

All three top-N shares rose from 2010 through roughly 2019-2020, then flattened or eased slightly through 2024. The top-50 share shows the steadiest, most persistent rise (+3.0 points over the full period) — concentration in the sense of "a broader elite tier pulling share from everyone outside it" is the more durable pattern than "the single top school pulling away," which peaked and partly reversed. Gini moved from 0.8253 (2010) to 0.8234 (2024) — essentially flat, meaning the underlying inequality of the whole distribution (including the hundreds of low-funded institutions) has not materially changed even as the elite tier's share drifted up and back down.

Context from outside the warehouse

External research corroborates both the headline level and a structural cause worth naming: federal R&D funding is also geographically concentrated in a handful of states and metro clusters (East Coast: Johns Hopkins, Duke, Columbia; West Coast: Stanford, UC San Diego, U Washington), and NCSES-adjacent research on the NSF EPSCoR program (aimed at states historically underfunded for federal research) finds non-EPSCoR states receive roughly 5x the federal research funding of EPSCoR states, even though EPSCoR-state institutions are found to be more productive per federal dollar invested in several output measures. This means the concentration measured here is not simply "the biggest, best-resourced research universities winning on merit alone" — it partly reflects a persistent, policy-relevant geographic/historical funding pattern that EPSCoR was created specifically to address and that a national top-N or Gini figure does not itself explain.

Every query behind this report

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

SELECT year, COUNT(*) AS n_inst, SUM(rd_expenditure_usd_thousand) AS total_federal_thousand
FROM research.nsf_herd_by_institution
WHERE funding_source='Federal' AND federal_agency='Total' AND rd_field='All'
GROUP BY year ORDER BY year

SELECT institution, rd_expenditure_usd_thousand AS val FROM research.nsf_herd_by_institution WHERE funding_source='Federal' AND federal_agency='Total' AND rd_field='All' AND year=2010 AND rd_expenditure_usd_thousand IS NOT NULL AND rd_expenditure_usd_thousand > 0

SELECT institution, rd_expenditure_usd_thousand AS val FROM research.nsf_herd_by_institution WHERE funding_source='Federal' AND federal_agency='Total' AND rd_field='All' AND year=2024 AND rd_expenditure_usd_thousand IS NOT NULL AND rd_expenditure_usd_thousand > 0

SELECT institution, rd_expenditure_usd_thousand AS val FROM research.nsf_herd_by_institution WHERE funding_source='Federal' AND federal_agency='Total' AND rd_field='All' AND year=2010 AND rd_expenditure_usd_thousand IS NOT NULL AND rd_expenditure_usd_thousand > 0

SELECT institution, rd_expenditure_usd_thousand AS val FROM research.nsf_herd_by_institution WHERE funding_source='Federal' AND federal_agency='Total' AND rd_field='All' AND year=2024 AND rd_expenditure_usd_thousand IS NOT NULL AND rd_expenditure_usd_thousand > 0

WITH ranked AS (
  SELECT year, institution, rd_expenditure_usd_thousand AS val,
         RANK() OVER (PARTITION BY year ORDER BY rd_expenditure_usd_thousand DESC) AS rnk,
         SUM(rd_expenditure_usd_thousand) OVER (PARTITION BY year) AS yr_total
  FROM research.nsf_herd_by_institution
  WHERE funding_source='Federal' AND federal_agency='Total' AND rd_field='All'
    AND rd_expenditure_usd_thousand IS NOT NULL AND year IN (2010,2024)
)
SELECT year,
  SUM(CASE WHEN rnk<=10 THEN val ELSE 0 END)/MAX(yr_total) AS top10_share,
  SUM(CASE WHEN rnk<=20 THEN val ELSE 0 END)/MAX(yr_total) AS top20_share,
  SUM(CASE WHEN rnk<=50 THEN val ELSE 0 END)/MAX(yr_total) AS top50_share,
  SUM(CASE WHEN rnk<=100 THEN val ELSE 0 END)/MAX(yr_total) AS top100_share
FROM ranked
GROUP BY year ORDER BY year

SELECT institution, rd_expenditure_usd_thousand AS val
FROM research.nsf_herd_by_institution
WHERE funding_source='Federal' AND federal_agency='Total' AND rd_field='All' AND year=2024
ORDER BY val DESC LIMIT 10

SELECT institution, rd_expenditure_usd_thousand AS val
FROM research.nsf_herd_by_institution
WHERE funding_source='Federal' AND federal_agency='Total' AND rd_field='All' AND year=2010
ORDER BY val DESC LIMIT 10

WITH base AS (
  SELECT year, institution, rd_expenditure_usd_thousand AS val
  FROM research.nsf_herd_by_institution
  WHERE funding_source='Federal' AND federal_agency='Total' AND rd_field='All'
    AND rd_expenditure_usd_thousand IS NOT NULL AND rd_expenditure_usd_thousand > 0
),
tot AS (
  SELECT year, SUM(val) AS yr_total FROM base GROUP BY year
),
ranked AS (
  SELECT b.year, b.val, RANK() OVER (PARTITION BY b.year ORDER BY b.val DESC) AS rnk, t.yr_total
  FROM base b JOIN tot t ON b.year=t.year
)
SELECT year,
  ROUND(SUM(CASE WHEN rnk<=10 THEN val ELSE 0 END)/MAX(yr_total),4) AS top10_share,
  ROUND(SUM(CASE WHEN rnk<=20 THEN val ELSE 0 END)/MAX(yr_total),4) AS top20_share,
  ROUND(SUM(CASE WHEN rnk<=50 THEN val ELSE 0 END)/MAX(yr_total),4) AS top50_share,
  ROUND(SUM(POWER(val/yr_total*100,2)),2) AS hhi
FROM ranked
GROUP BY year ORDER BY year

SELECT year,
  COUNT(*) AS total_rows,
  COUNT(*) FILTER (WHERE rd_expenditure_usd_thousand IS NULL) AS null_rows,
  COUNT(*) FILTER (WHERE rd_expenditure_usd_thousand = 0) AS zero_rows
FROM research.nsf_herd_by_institution
WHERE funding_source='Federal' AND federal_agency='Total' AND rd_field='All'
  AND year IN (2010,2024)
GROUP BY year ORDER BY year