The Government Accountability Office has put a number on the deferred resignation programme: about $6.7bn in 2025.
That is what it cost to keep roughly 144,000 federal employees on the payroll while they were on administrative leave, after accepting an offer to resign and before actually leaving. It accounts for around 70 percent of the $9.5bn the federal government spent on paid administrative leave that year.
The number that is missing
The watchdog's more serious finding is not the cost. It is that nobody can check the other side of the ledger.
Federal payroll systems do not separate administrative leave taken under this programme from administrative leave taken for every other reason — investigations, weather, building closures. Without that split, GAO wrote, "there is no way to accurately determine to what extent the governmentwide cost-savings goal is being met."
The spread between independent estimates shows what that means in practice. The Partnership for Public Service put the cost at $4.5bn. Public Citizen put it at $11bn or more. GAO landed between them, and recommended that the Office of Personnel Management disclose the data's limitations and create a category that tracks this leave separately.
The case for it, stated fairly
OPM's director, Scott Kupor, called the $6.7bn a one-time expense set against $20bn in expected annual savings — a "400 percent return on investment."
That arithmetic is not absurd on its face. A federal salary plus benefits is a recurring cost, and buying out a post once to remove it permanently can pay back quickly. Private employers run the same calculation, and governments have used voluntary separation incentives for decades precisely because forced reductions are slower, uglier and legally fraught.
The question is whether the posts stay gone, and whether the people who left were the ones the government could afford to lose.
What this desk has found before
Neither question is answered by a headcount. This desk reported in early September that saving an agency is not the same as keeping its capability: FEMA had shed more than a thousand experienced staff and was rehiring into the same functions. We reported last week on a court finding that DHS could not explain the reasoning behind halving FEMA's workforce, and earlier that agencies cannot hire fast enough to spend what Congress appropriated.
Those three stories describe one mechanism. Experience leaves faster than headcount, and the cost of replacing it does not appear in the year the saving is booked.
Why the accounting gap is not an accident
It is tempting to read the missing category as bureaucratic sloppiness. The likelier explanation is sequencing.
The programme was designed and offered quickly. Payroll systems across dozens of agencies were not modified to accommodate it, because modifying them would have required a decision about what to call the new leave, guidance to every agency payroll office, and time the programme did not have.
The cost of that speed is precise and it is the cost GAO identified: a year later, the only figures available are estimates built by outsiders from partial data, and they differ by a factor of more than two. A programme can be a good idea and still be unmeasurable, and the second fact makes the first impossible to prove.
What to watch
Whether OPM adopts GAO's tracking recommendation, which is the only way the $20bn claim ever becomes checkable. Then the rehiring data: how many of the posts vacated under this programme are refilled within two years, and at what grade. A saving that reappears as a contract or a re-advertised job was a timing difference, not a saving.
The GAO estimate of roughly $6.7bn in 2025 and its share of the $9.5bn total spent on paid administrative leave; the participation figure of about 144,000 employees; the quoted finding on cost-savings measurement; the recommendations that OPM disclose data reliability issues and create a separate tracking category; the alternative estimates from the Partnership for Public Service and Public Citizen; and the response from OPM director Scott Kupor describing a one-time expense against $20bn in expected annual savings and a "400% return on investment" are as reported by Federal News Network and Politico on 15 and 16 September 2026. This publication has not independently reviewed the GAO report. The analysis is our own.
Topics nationalfederal workforceopm
