NIH budget instability shows up far more in grant counts and trial starts than in total dollars

Summary

Using NIH — the federal government's largest funder of extramural biomedical research (~82% of its ~$47B annual budget goes to outside grantees) — as the case study: year-to-year instability in what NIH can fund shows up much more sharply in the number of grants and projects funded than in the total dollar figure Congress appropriates. Total NIH appropriations have been nearly flat since FY2024 (within about 1% year to year), yet NIH's own published estimate is that the number of new competing research grants it could support fell from 10,086 (FY2024) to 6,095 (FY2025) — a 39.6% drop — driven mainly by a policy shift toward committing multi-year funding up front rather than a change in the topline number. Independently, AskAmerica's own award-level NIH data (RePORTER) shows total award count falling 9.9% from its 2023 peak (80,844) to 2025 (72,804) while total award dollars moved by less than 2% over the same span, and NIH-sponsored clinical-trial starts fell from 217 in 2021 to 138 in 2025 (-36%). The published literature on the 2011-2013 sequestration era finds the same pattern historically: funding volatility depresses publication counts and grant-application success rates even when it doesn't show up as a dramatic single-year budget cut. In short: the transmission from budget volatility to research output is real, measurable, and often larger in the 'number of things funded' than in the dollar total — because agencies protect ongoing commitments first and let new starts absorb the swing.

What 'agency budget volatility' means here, and why the natural denominator is grant/project counts, not dollars

NIH's total appropriated 'program level' has actually been fairly stable in nominal dollars in the last few years: $47.31B (FY2024, program level including transfers) to $47.035B (FY2025, -0.6%) to $47.493B (FY2026 enacted, +1.0%). Looking only at that series, one would conclude NIH funding is not volatile at all recently. But CRS's own history of the series (FY1996-FY2026) documents real instability further back: NIH's budget doubled from FY1998 to FY2003, then grew only gradually for over a decade, with outright nominal-dollar decreases in FY2006, FY2011, and FY2013 (the debt-ceiling/sequestration era), before resuming steady 3-9% annual growth from FY2016-FY2023. Adjusted for the Biomedical R&D Price Index (a sector-specific deflator, not general CPI), NIH's real purchasing power peaked in FY2003 and has still not recovered that peak as of the FY2026 enacted level (about 10.3% below the FY2003 peak in real terms) — meaning 'volatility' at NIH includes multi-decade real erosion, not just single-year swings.

The key finding for the current period is that dollar-total stability can mask real volatility in what gets funded. NIH itself projected that its FY2025 policy change — reserving half of its competing-grant budget for grants that are fully funded across all their years up front, instead of funding them incrementally each year — would cut the number of new competing grants nearly in half (10,086 to 6,095) even though the topline budget barely moved. That is a mechanism-level volatility (how the money is allocated across years and grant vintages) rather than an appropriations-level one, and it lands directly on researchers: fewer new projects get started, independent of the total dollar figure.

What we computed directly from NIH's own award-level and trial data

Using NIH RePORTER's award microdata (research.nih_award_projects, FY2022-2025, ~372K award rows) we counted total awards (all mechanisms, new and continuing) per fiscal year: 79,707 (2022) -> 80,844 (2023) -> 79,243 (2024) -> 72,804 (2025). Total award dollars over the same span moved by less than 2% ($46.86B to $47.93B to $46.99B to $46.09B), so essentially all of the swing shows up as fewer, larger awards rather than fewer dollars — the same pattern CRS documents nationally for FY2025's competing-grant cliff.

Using ClinicalTrials.gov (health.clinical_trials, filtered to funder_type='NIH'), we counted trials by start year among the 3,797 of 4,219 NIH-funded trial rows that carry a recorded start_date (422 rows, 10%, have no start_date and were excluded from every by-year count below — those excluded rows are not a random subset of NIH trials, since a missing start_date is somewhat more common for older or administratively-closed records, so the true year-by-year counts could differ slightly from what is shown). Among the trials with a start_date, the count with a given start year rose from 194 (2020) to a peak of 217 (2021), then declined each year to 138 (2025), a 36% drop from the 2021 peak — consistent in direction and rough timing with the funding pattern above, though this table is also restricted to studies updated on or after 2020, so earlier-year counts (pre-2021) are additionally understated by survivorship and the true peak-to-trough decline for trial initiations could be somewhat different from this 36% figure. We are not claiming this proves causation between the FY2025 grants policy and trial starts specifically — funding for a trial that starts in a given year usually derives from an award made one or more years earlier — but the co-movement is directionally consistent with less total new-project funding flowing through to fewer new trials with a lag.

What the published literature adds

The pattern we compute directly for the most recent NIH data matches what peer-reviewed and preprint literature has found about earlier NIH funding volatility episodes. Following the 2011 Budget Control Act and 2013 sequestration, the count of NIH-funded 'fundamental' and 'mixed mostly fundamental' basic-science publications declined (Absolute and Relative Declines in NIH-Funded Basic Science Publications, bioRxiv). A separate literature strand shows the mechanism: when NIH funding is flat or falls, per-application success rates drop, so investigators submit more applications to stay funded, which further dilutes success rates — a volatility-amplifying feedback loop, not a one-time shock (Modeling the impacts of budget reductions at the U.S. NIH on research and innovation, PMC). Funding interruptions specifically — gaps between one grant ending and a renewal starting — measurably reduce a lab's research activity in the interim, with the effect partially buffered by a PI holding multiple simultaneous grants (Science, interrupted: Funding delays reduce research activity but having more grants helps, PLOS ONE). And modeling work on the proposed FY2026 NIH cuts finds that even temporary funding disruptions create output losses (fewer trainees, delayed drug-development milestones) that persist for years after funding is restored, because postdocs leave the field and multi-year projects cannot simply resume where they left off (PMC13490408 above; see also Effects of Government Spending on Research Workforce Development, PMC, on how funding swings shape the postdoctoral pipeline specifically).

Caveats and what this analysis does not show

• This is a single-agency (NIH) case study, not a cross-agency estimate of a general elasticity between budget volatility and output; NIH was chosen because it is the largest, best-documented extramural research funder and because CRS and NIH itself publish detailed year-by-year figures.
• The AskAmerica warehouse's own agency-level total-obligations table (fiscal.usaspending_by_agency) reports Department of Health and Human Services totals, not NIH specifically — HHS's ~$2.5-2.8 trillion annual obligations are dominated by Medicare/Medicaid entitlement spending, so that table cannot isolate NIH's research budget and was not used for the volatility computation; we used CRS's NIH-specific appropriations history and NIH's own RePORTER award microdata instead.
• research.nsf_federal_rd_obligations (an alternative NIH-specific source in this warehouse) only has 2 years loaded (2022-2024, with gaps), too short to compute a volatility series on its own; research.nih_award_projects (2022-2026) supplied the longer, more complete award-level series used above.
• The FY2026 competing-grants figure (4,312) is the President's budget request, not enacted funding — FY2026 was ultimately enacted close to FY2025 levels, so the actual FY2026 competing-grant count is very unlikely to be as low as the request implied; we show it only as NIH's own stated sensitivity of grant output to budget level, not as a realized outcome.
• The clinical-trials co-movement with award counts is a directional, small-n correlation (6 points), not a regression-tested causal estimate; the trial counts also exclude 422 of 4,219 NIH-funded trials (10%) with no recorded start_date, and are drawn from a table restricted to studies updated since 2020, both of which understate true trial counts, especially in earlier years, and could distort the apparent 2020-2021 uptick.

What This Report Does Not Answer

• Distinguish this from a cross-agency general answer: Answered as a single-agency (NIH) case study rather than an all-agency average, because NIH is the only agency in this warehouse with award-level output-adjacent data (grant counts, trial starts) at sufficient granularity; a cross-agency version of this question would need award-level microdata for NSF, DOE, NASA etc., which this warehouse does not carry at the same grain

Every query behind this report

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

SELECT DISTINCT agency_name FROM fiscal.usaspending_by_agency WHERE agency_name ILIKE '%health%' OR agency_name ILIKE '%science%' OR agency_name ILIKE '%energy%' ORDER BY 1

SELECT CAST(LEFT(start_date,4) AS INTEGER) AS start_year, COUNT(*) AS trial_count
FROM health.clinical_trials
WHERE funder_type = 'NIH' AND start_date IS NOT NULL AND LENGTH(start_date) >= 4
GROUP BY CAST(LEFT(start_date,4) AS INTEGER)
ORDER BY 1

SELECT year, obligated_amount FROM fiscal.usaspending_by_agency
WHERE agency_name = 'Department of Health and Human Services' ORDER BY year

SELECT fiscal_year, COUNT(*) AS award_count, COUNT(DISTINCT project_num) AS distinct_projects, SUM(award_amount) AS total_award_amount
FROM research.nih_award_projects
WHERE fiscal_year BETWEEN 2022 AND 2025
GROUP BY fiscal_year ORDER BY fiscal_year

SELECT funder_type, COUNT(*) AS total, SUM(CASE WHEN start_date IS NULL THEN 1 ELSE 0 END) AS null_start_date
FROM health.clinical_trials WHERE funder_type = 'NIH' GROUP BY funder_type