Universities now pay for a quarter of their own research

The institution's own share has more than doubled since 2010

American universities spent $117.7 billion on research and development in fiscal 2024. Of that, $30.2 billion came from the institutions' own funds, roughly a quarter of the total (share computed; National Science Foundation, Higher Education Research and Development survey, fiscal 2024, Tables 1 and 3). In fiscal 2010 the institutional share was $11.9 billion. It has grown to two and a half times that size since then (multiple computed).

Part of the institutional total is money universities choose to spend on purpose: seed funding for new laboratories, cost sharing written into award terms, and early support for projects they want to exist before any sponsor does. Another $7.07 billion was earned under negotiated rates and never recovered (NSF HERD fiscal 2024, Tables 3 and 12). That money is indirect cost: the shared expenses of running research that no single grant pays for directly, such as laboratory buildings, utilities, research computing, compliance staff, and the people who administer awards. Universities recover indirect costs through the overhead rate they negotiate with the federal government on their grants, called the facilities and administrative rate, or F&A rate. When recovery falls short of what the rate allows, the difference comes out of the institution's own budget.

The sponsor buys the research once, and the institution buys a share of it again.

The gap shows up at closeout

A research administrator closing out a five-year federal award in June works from a small stack of documents: the notice of award, the institution's negotiated rate agreement, the effort certifications that attest how people's time was spent, and the final financial report. The award carried the negotiated rate on paper, and the ledger recovered less. Equipment purchases sat outside the base the rate applies to, along with the tuition support for the graduate assistants. A subcontract carried the rate on only part of its invoiced amount. One sponsor program capped overhead below the negotiated figure, and the institution took the award anyway, because the science mattered and the team was ready. The administrator books the shortfall to an institutional account and moves to the next file.

Survey data measures that distance. The negotiated rate is the number on the agreement with the federal government. The effective rate is what an institution actually recovers for each direct research dollar once exclusions, sponsor caps and waivers have been applied. Across 120 institutions surveyed in 2023 by the Council on Governmental Relations, or COGR, an association of research universities, the average negotiated rate was 56.9 percent and the average effective rate was 44.3 percent. COGR reports the gap as 12.7 points (COGR 2023 survey, n = 120, published December 2024).

The administrative cap has stayed at 26 percent since 1991

The overhead rate has two halves. The facilities half covers buildings, utilities and equipment, and it moves with documented costs; when a new laboratory building depreciates into the cost pool, the facilities component can rise to reflect it. The administrative half covers department administration, sponsored programs offices and general administration, and since 1991 it has been capped at 26 percent of modified total direct costs, the portion of a grant's direct spending the rate is allowed to apply to. The cap entered federal cost rules under Office of Management and Budget Circular A-21 in 1991 and stands today in the uniform guidance at 2 CFR 200, Appendix III. Rate negotiations since then have happened entirely inside that constraint. A university can document administrative costs above the cap in its rate proposal, and the negotiators can accept the documentation as sound, and the number still comes back 26. COGR's survey work puts actual administrative cost near 35 percent. The difference, roughly nine points of the base on every award (computed from the two figures above), is carried by the institution.

Public and private institutions carry the shortfall in different proportions. Across all institutions filing the standard HERD form, 26.9 percent of all earned indirect cost went unrecovered. Public institutions left 30.5 percent uncollected; private institutions left 21.3 percent (computed from NSF HERD fiscal 2024, Table 12).

Unfunded proposals are paid for too

The National Institutes of Health made 8,161 research project grant awards from 62,592 applications in fiscal 2025, a success rate of 13.0 percent. In fiscal 2023 the rate was 21.3 percent (NIH Data Book, fiscal 2025). The 54,431 applications that went unfunded (count computed) were still written, budgeted, routed and submitted by faculty and staff whose time the institution had already paid for.

The time inside funded projects is measured too. In the Federal Demonstration Partnership's 2018 Faculty Workload Survey, principal investigators, the faculty members who lead grants, reported that 44.3 percent of the time they spend on federally funded research projects goes to administrative tasks. The survey's definition is precise: it measured time inside those funded projects and asked nothing about the rest of a professor's week. It is a coincidence that the same figure, 44.3 percent, appears above as the effective recovery rate; the two numbers come from different surveys and measure different things. The survey covered 11,167 investigators at 111 member organizations, covering 149 institutions.

The 2025 cap never took effect

In 2025, indirect cost recovery reached board agendas at institutions where it had rarely been discussed outside the research office. On February 7, 2025, the National Institutes of Health issued guidance proposing a cap on the indirect cost rates applied to its awards. On April 4, 2025, a federal court blocked the policy from taking effect, an order lawyers call an injunction. The block was upheld on appeal on January 5, 2026, and the policy was withdrawn on April 6, 2026. The recovery gap described above predates the episode and remained in place when it ended. No negotiated rate changed because of the policy at any point between February 7, 2025 and April 6, 2026.

One point of recovery is worth $2 million a year on a $200 million base

The arithmetic a chief financial officer needs is short. One percent of $200 million is $2 million. An institution with a $200 million research base that raises its effective recovery by a single point keeps $2 million a year, and the amount recurs for as long as the base holds (computed; the base is annual research expenditure as defined in NSF HERD). An institution with a $400 million base doubles the figure, and one with a $100 million base halves it (computed).

Effective recovery moves through operational work as much as through negotiation: applying the correct rate to the correct base on every award, catching sponsor caps at the proposal stage, and recording each waiver as a decision someone made. The negotiated rate is set in one negotiation every few years. The effective rate is produced by thousands of award-level events in between. Precision protects the institution in the other direction as well. In one publicly reported case, a university repaid $9.5 million to the federal government in 2016 after applying its on-campus rate to work performed off campus, across 423 grants, over twelve years.

The question a chief financial officer can put to the research office is what the institution's effective recovery rate was last year and how far it sat below the negotiated rate. Across the 120 institutions in COGR's 2023 survey, that distance averaged 12.7 points.

Methodology

Here is every computed figure in the piece, with its arithmetic:

• Institutional share of total academic R&D, "roughly a quarter" / 25.7%: $30.2 billion ÷ $117.7 billion = 25.66%, rounded to 25.7%. Base figures from NSF HERD fiscal 2024, Table 1.
• "Two and a half times" growth of institutional spending: $30.2 billion ÷ $11.9 billion = 2.54. Base figures from NSF HERD fiscal 2024 and fiscal 2010.
• Other funding sources in Figure 2, $87.5 billion: $117.7 billion − $30.2 billion = $87.5 billion (residual).
• Unfunded NIH applications, 54,431: 62,592 applications − 8,161 awards = 54,431. Base figures from the NIH Data Book, fiscal 2025.
• Unrecovered shares in Figure 5, 26.9% / 30.5% / 21.3%: 7,064,076 ÷ 26,236,219 = 26.9% (all institutions); 4,895,834 ÷ 16,059,752 = 30.5% (public); 2,168,242 ÷ 10,176,467 = 21.3% (private). Figures in thousands of dollars from NSF HERD fiscal 2024, Table 12, standard-form institutions.
• "Roughly nine points" of administrative shortfall: approximately 35% actual administrative cost − 26% cap = approximately 9 points. Approximate because COGR states the actual figure as "near 35 percent."
• $2 million per point on a $200 million base: 1% × $200,000,000 = $2,000,000 per year. Recurs each year the research base holds; scales proportionally with the base.
• The 12.7-point gap between negotiated and effective rates is COGR's published figure and was not recomputed here. The published averages of 56.9 and 44.3 are rounded; subtracting them does not reproduce COGR's unrounded gap, so the gap is cited as published.
• FDP definition note: the 44.3 percent figure is the share of time principal investigators spend on administrative tasks within their federally funded research projects (FDP 2018 Faculty Workload Survey, n = 11,167 at 111 member organizations, covering 149 institutions). It is a share of federally funded research time only. It is never presented in this piece as a share of total faculty time.
• All other figures are as published by their named sources: NSF HERD fiscal 2024 (Tables 1, 3, 12), COGR 2023 survey (published December 2024), NIH Data Book fiscal 2025, FDP 2018, and 2 CFR 200 Appendix III.