GAO traces $6.7bn of federal administrative leave pay to DOGE's resignation offer

Scott Kupor's rebuttal sets one $9.5bn outlay against $40bn a year, and at that rate the money comes back in about 2.9 months ^[1]. "That 400% return on investment is a massive benefit to the taxpayer," the OPM director told Fortune ^[7]. The $40bn does not appear in the GAO report, which counted salaries agencies paid to employees who were off the clock ^[1]. Civilian salaries and benefits are about 5.5% of the federal budget, and were among the first items DOGE cut ^[16].

The cost side is easier to trace. February's offer paid full salary and benefits through the end of September ^[3], and the nearly 140,000 people who took it ^[4] account for $6.7bn of the leave total ^[2], about $48,000 each ^[2], which over those eight months annualizes to roughly $72,000 ^[3]. Kupor's $40bn spread across the 270,000-employee reduction he cites works out to about $148,000 saved per departure per year ^[4]. The two figures describe different groups: deferred resignation covered about 52% of the more than 271,000 people who have left since January 2025 ^[5].

Whether the annual saving survives hiring is the open question. The Partnership for Public Service reported last month that agencies have backfilled or are seeking to backfill more than 20,000 roles emptied by the program ^[9], with new hires averaging 1.4 GS grades below the people they replaced ^[10] and a 6.1-grade gap in ATF criminal investigation work ^[11]. Each backfill puts a salary back on the books, at a lower grade.

GAO said differences in how agencies report administrative leave may affect the accuracy of the numbers, a problem that spans administrations ^[12]. In 2014 the watchdog found the Defense Department and USAID granting leave on different terms and asked OPM to write guidance standardizing how the time is recorded ^[13]. OPM is the office now contesting the count.

I think the GAO number is the better-evidenced of the two, because payroll data records money that has already left the Treasury, while the $40bn describes fiscal years that have not closed ^[1][6]. Two ways that could be wrong: if most of the 271,000 departures stay unfilled, even a heavily discounted saving clears $9.5bn within a year or two ^[8]; and a voluntary buyout at roughly $48,000 a head ^[2] may be cheaper than fighting involuntary separations. DOGE has claimed $215bn in savings from cancelled grants, contracts and workforce reductions, which, according to Fortune, experts have disputed as much less ^[14]. That claim equals about 5.7% of the $3.8 trillion added to the federal debt since January 2025 ^[6].