NSF's Proposal Workload Rose 8%. Its Federal Workforce Fell 36%.

The National Science Foundation received more competitive proposals in fiscal 2025 than it had in any of the previous three years. Its staff still had to organize nearly 200,000 expert reviews, make funding decisions, monitor tens of thousands of active awards and oversee a portfolio of research facilities that includes some of the most complex scientific instruments in the world.

The federal workforce carrying that responsibility was becoming much smaller.

Office of Personnel Management data show NSF with 1,719 covered federal employees in December 2024 and 1,104 in June 2026. That is a reduction of 615 employees, or 35.8%, in 18 months. The decline continued after January 2026 rather than reversing: the count moved from 1,153 in January to 1,104 in June.

During fiscal 2025, competitive proposal actions rose 8% to 43,532, according to NSF's Agency Financial Report. Competitive award actions fell 21% to 8,377. Average decision time rose from 5.87 months to 6.49 months, while the share decided within 182 days fell from 66% to 55%.

Those facts describe a narrowing capacity margin. They do not establish that workforce reductions caused a particular proposal to be rejected, delayed or canceled. Appropriations, program priorities, proposal complexity and policy changes can all alter awards and timing. Public workforce records also omit many non-federal scientists who rotate through NSF and do not identify which employees worked on a particular grant or facility.

What the records do establish is more basic: NSF entered a higher-volume proposal year with fewer federal employees across scientific, engineering, analytical, administrative and technology occupations. Its public recruiting signal then remained near a historic low through Aug. 24, 2026.

A proposal pipeline can grow even when awards fall

NSF's grant workload does not begin or end with an award. Staff receive proposals, check eligibility and compliance, recruit outside experts, manage conflicts, assemble review panels, document recommendations, communicate decisions and oversee awards after they are made. A proposal that is declined still consumes review and decision work.

The agency's five-year operating table shows that the 2025 increase was not simply a return to a tiny base. Competitive proposal actions moved from 40,454 in 2024 to 43,532 in 2025, an increase of 3,078. That put the count just 85 below the 43,617 recorded in 2021.

Competitive award actions moved in the opposite direction. They fell from 10,591 to 8,377, a reduction of 2,214. The funding rate declined from 26% to 19%. Active awards fell from 58,081 to 54,773, while proposal reviews rose from 187,829 to 197,242.

The 54,773 active awards are a stock of continuing obligations, not another count of new decisions. An active award can require progress review, payment administration, amendments, compliance work and eventual closeout. The intensity varies widely: a small individual award and a multi-institution research center should not be treated as equal units of labor. Even so, the active portfolio is evidence that work continues after the headline funding decision.

OPM's occupation detail offers a narrow view of that stewardship layer. Grants Management series 1109 moved from 48 NSF employees in January 2025 to 40 in June 2026. Contracting fell from 25 to 17 and Auditing from 23 to 14, while Financial Administration and Program changed from 36 to 35. Those are small groups, and their codes do not reveal assignments or the amount of work provided by staff in other series. They also caution against interpreting fewer new awards as equivalent to less oversight work. A smaller incoming class can coexist with a large inventory of active agreements that still need federal judgment and control.

The funding rate is not a workforce productivity measure. NSF cannot award money it does not have, and proposals differ in requested amount, field, complexity and merit. An 8% rise in submissions does not require an 8% rise in staff, just as a 21% fall in awards does not mean award officers did 21% less work.

The less visible review count matters. NSF coordinated 197,242 reviews in 2025, up 5% from the previous year. Much of the scientific judgment comes from outside experts, but federal staff remain responsible for constructing and administering a defensible process. Peer reviewers are not a substitute for the employees who handle conflicts, documentation, program decisions and award stewardship.

The federal headcount fell below its recent operating range

The contraction stands out against a longer history. NSF's covered federal headcount rose from 894 in September 2000 to 1,482 in September 2009. It stayed near 1,500 for much of the next decade, then climbed to 1,717 in September 2024.

By June 2026, the count was 1,104. Apart from the unusually low 2000 observation, that was below every annual September count in the assembled 2000-2025 series.

The OPM series is headcount, not full-time-equivalent staffing. It covers employees assigned to NSF subelement NF00 and does not measure productive hours, workload or contractor support. The long-run chart uses September observations through 2025 and June 2026 as the latest available month.

NSF's inspector general had already identified workforce management as a challenge. In the fiscal 2025 financial report, the inspector general described a proposed 25% staffing reduction, from 1,735 in fiscal 2024 to 1,297 in fiscal 2026, and warned that reductions could affect institutional knowledge and grant oversight. That staffing projection is not directly comparable to the OPM headcount series, but both point to the scale of the operating change.

The decline was not confined to one administrative category. From January 2025 through June 2026, general physical science fell from 166 covered employees to 90. Management and program analysis fell from 256 to 200. General engineering moved from 98 to 44, and miscellaneous administration and program from 163 to 111.

Education and training fell from 84 to 39. Natural resources and biological science fell from 99 to 56. Information Technology Management moved from 100 to 64. Computer science fell from 47 to 15, social science from 66 to 35 and mathematics from 28 to 11.

An occupation code is not an assignment. A physical scientist may work in a program office, a laboratory-related function or another role. A management analyst may or may not support grants. The losses nevertheless span the kinds of capacity an award-making science agency uses: domain expertise, engineering, program administration, analysis and information technology.

One category grew. Program Management series 0340 increased from 42 to 52 employees. That gain is real counterevidence to a claim of uniform collapse, but it was too small to offset the broader decline.

The experience profile changed in an unexpected way

The public data do not support a simple story in which the contraction consisted only of older employees retiring. The largest loss occurred among employees with less than five years of federal service.

That group fell from 511 in January 2025 to 94 in June 2026, a reduction of 417. Employees with five to nine years of service fell from 320 to 208. The 10-to-19-year group increased slightly, from 448 to 465, while employees with at least 20 years of service fell from 420 to 337.

Employees age 55 or older also declined in absolute terms, from 662 to 398. Their share of the workforce fell only modestly, from 39.0% to 36.1%, because younger groups were shrinking too.

These categories cannot reveal whether a particular scientific specialty or relationship was lost. They do show why institutional knowledge is not only a retirement issue. When the newest cohort falls from more than 500 employees to fewer than 100, the agency has a thinner pipeline of people learning its programs, systems and oversight practices.

OPM personnel-action files point in the same direction. They record 36 NSF accession actions and 601 separation actions in calendar 2025. Through June 2026, they record three accessions and 63 separations. Actions are not necessarily unique people, so these figures should not be treated as an exact hiring-to-exit ratio. They do show recorded inflow remaining low after the large 2025 outflow.

Rotating scientists are essential, but they are a separate workforce

NSF is unusual because outside scientists and engineers routinely serve temporary appointments at the agency. Any analysis that treats the OPM count as the entire intellectual workforce would miss that operating model.

A 2018 Government Accountability Office review found that Intergovernmental Personnel Act rotators averaged about 12% of NSF's workforce and 17% of its workforce cost from fiscal 2008 through 2017. Many came from universities and brought current knowledge of research communities. They helped shape funding decisions and program direction.

IPA rotators generally remain employees of their home institution and are not federal employees. They are therefore excluded from the OPM headcount in this article. Visiting Scientist, Engineer and Educator appointees are temporary federal employees and can appear in federal personnel data. The categories should not be combined casually.

NSF's fiscal 2025 budget request listed 232 IPA full-time equivalents in fiscal 2023 and requested 316 for fiscal 2025. A budget request is not an actual 2025 headcount, and those FTE figures cannot fill the missing months in the OPM series. They do show that NSF expected rotators to remain a substantial part of its model.

Rotators add scientific reach, but they do not make permanent federal capacity irrelevant. GAO has examined the need for conflict-of-interest controls, cost oversight and continuity when temporary appointees cycle through the agency. NSF's inspector general reported that, in October 2025, federal employees supervised by non-federal rotators were reassigned to federal supervisors. That management change reinforces the distinction between temporary expertise and federal supervisory responsibility.

The safest conclusion is not that NSF lost 36% of everyone doing its work. It is that its covered federal workforce lost 36%, while the available public records do not provide a matching monthly series for non-federal rotators or contractors.

Decision time worsened, but timing alone does not explain why

NSF's annual performance data show a measurable change in proposal timing. Average time from receipt of a complete proposal to notification of a funding decision was 5.60 months in fiscal 2023 and 5.87 months in 2024. It rose to 6.49 months in 2025.

The share decided within 182 days moved from 70% in 2023 to 66% in 2024 and 55% in 2025.

The change occurred during the workforce contraction and the proposal-volume increase. That chronology is relevant, but it is not a causal test. The public measure does not separate waiting for reviewers from internal processing, budget decisions, program holds, policy review or unusually complex proposals. It cannot show what decision time would have been with a different workforce.

NSF's performance plan describes efforts to keep decisions within six months and streamline award processes. It also reported 99.9% IT systems availability in fiscal 2025. That operating evidence argues against depicting the agency as unable to function.

The more precise concern is resilience. An average of 6.49 months means the system had less room against a six-month benchmark at the same time that proposal and review counts increased. A smaller federal workforce may have fewer people available to absorb a surge, cover a vacancy or redesign a process. Whether that reduced margin becomes a sustained delay will require later performance years, not a verdict from one data point.

Research facilities show both performance and strain

NSF's responsibilities extend beyond grants to individual investigators. The agency funds and oversees major research infrastructure, including observatories, ships, polar facilities and advanced instruments. These projects involve long schedules, specialized construction, international partners and risks that cannot be summarized by proposal counts.

A 2026 GAO review identified 21 projects in NSF's major- and midscale-infrastructure portfolio as of July 2025: 13 major projects and eight midscale projects. Four of the seven major projects then in construction had schedule delays of four to 27 months compared with their June 2024 schedules. Two major projects and three midscale projects had reduced scope.

GAO cited multiple contributors, including labor availability, contractor performance and budget uncertainty. All projects remained within their authorized cost levels at the time of its snapshot. The report does not assign those delays or scope changes to NSF staffing.

NSF's own fiscal 2025 performance measures provide important counterevidence. Seventy-five percent of the tracked major-facility construction projects met the agency's cost-and-schedule goal, up from 50% in fiscal 2024. All five tracked midscale projects met the goal, up from 89% a year earlier.

The GAO and NSF measures are not contradictions. They use different baselines and answer different questions. A project can experience a schedule change relative to an earlier plan and still remain within a currently authorized cost or performance threshold. Together, the records describe a portfolio that was still delivering against some formal measures while carrying meaningful schedule and scope pressure.

Related research

Put this finding in context

All articles

Data investigations/August 18, 2026

FEMA's Public Workforce Count Fell 22% as GAO Warned of Readiness Risks

FEMA had 5,793 fewer covered civilian employees in June 2026 than in January 2025. GAO says the agency cut staff without the workforce planning needed to assess mission capacity.

Data investigations/August 18, 2026

IRS Workforce Fell 29% as GAO Warned of Filing-Season Risks

The IRS had 29,663 fewer covered civilian employees in June 2026 than in January 2025. GAO says workforce-planning gaps and post-filing pressures pose risks to future operations.

That distinction matters for workforce reporting. A smaller federal team can oversee contractors and facilities successfully. The public evidence does not prove otherwise. But complex portfolios need program officers, engineers, contracting staff, financial analysts and technical reviewers who can identify problems early and preserve institutional memory across projects lasting many years.

Public recruiting has not shown a recovery

FederalHiringData's historical USAJOBS archive offers a limited view of NSF's external recruiting. We counted distinct NSF announcements opened from Jan. 1 through Aug. 24 in each year from 2018 through 2026.

The archive contains 192 announcements in the 2018 window, 206 in 2020, 213 in 2021 and 214 in 2024. The count fell to 13 in 2025 and remained at 13 in 2026.

An announcement is not a hire, vacancy or person. One posting can cover several grades or locations, and NSF can use temporary appointments and other authorities that are not visible as ordinary competitive announcements. The archive also cannot reveal how many selections were made.

The measure is still useful as a public signal. Through Aug. 24, 2026, it showed no broad announcement rebound from the 2025 low. Three accession actions in the OPM file through June point in the same direction, although action counts and announcements are different measures.

Readers can check current federal job listings separately. The historical announcement counts in this article do not mean that any of those positions remain open today.

What the evidence supports

NSF still processed a large national research portfolio in fiscal 2025. It coordinated nearly 200,000 reviews, maintained 54,773 active awards and improved its reported construction-performance rate for both major and midscale facilities. Those are substantial operating outcomes.

It did so while its covered federal workforce entered a sharp contraction. By June 2026, the headcount was 36% below December 2024. Losses reached scientific, engineering, administrative and technical occupations. The under-five-year service cohort absorbed the largest numerical decline, and public recruiting remained close to its 2025 low.

At the same time, proposal volume rose, competitive awards fell and average decision time moved above six months. These overlapping trends are evidence of pressure, not proof of cause. A careful assessment should watch what happens next: whether decision times recover, whether hiring resumes, whether facility schedules stabilize, and whether permanent staff and rotating experts together can sustain review and oversight quality.

The central public-interest question is not how many grants each employee produces. That ratio would be analytically false because grant difficulty, peer review, funding, post-award monitoring, facilities and temporary personnel all vary. The question is whether NSF retains enough durable federal capacity to make defensible decisions and manage a changing research portfolio.

Methodology and limitations

FederalHiringData analyzed OPM legacy FedScope agency files for September 2000 through 2014 and OPM Federal Workforce Data for September 2015 through 2025, plus June 2026, using NSF agency subelement NF00. December 2024 and June 2026 support the 18-month headline comparison. January 2025 and June 2026 support detailed occupation, age and length-of-service comparisons.

The OPM measure is covered federal headcount, not full-time equivalents, workload, contractor staffing or total people contributing to NSF programs. It excludes non-federal IPA rotators. Occupation codes do not identify assignments to a grant, review panel, program or facility.

Personnel-flow figures come from OPM public accessions and separations files. They count actions, not necessarily unique people. The analysis does not convert those actions into an exact replacement rate.

Proposal, award, review and active-award figures come from NSF's fiscal 2025 Agency Financial Report. Decision-time and facility-performance figures come from the fiscal 2027 Annual Performance Plan and fiscal 2025 Annual Performance Report. Fiscal years do not align exactly with the calendar-month workforce snapshots.

Infrastructure findings come from GAO-26-107842. Rotator context comes from GAO-18-533, NSF budget materials and the inspector general's management discussion. These sources do not provide a continuous monthly count of every IPA rotator, visiting scientist or contractor through June 2026.

The USAJOBS analysis counts distinct NSF announcement control numbers opened from Jan. 1 through Aug. 24 in each year. FederalHiringData historical coverage begins in March 2017, so the same-window chart starts in 2018. Announcements are not hires, vacancies or selections.

Sources and public records were checked through Aug. 24, 2026. The analysis does not establish that workforce reductions caused a specific proposal outcome, facility delay, scope change or decision-time result. It does not measure the scientific merit of funded or declined proposals.

Readers can explore federal workforce statistics, the agency directory, the historical jobs archive and more FederalHiringData investigations.