Section I What the analysis found
The Partnership for Public Service published an analysis on 20 August 2026, built on OPM's own data. It matched separations against later hiring and found that by June 2026, nearly 20,600 people had been hired into the same agency subcomponents and the same occupational series that deferred resignation separations had left.
Its summary was blunt: many employees were paid to resign only to be replaced just months later, often at a more junior level.
One series, in and out
An equal number were later hired into it.
Note what the analysis does and does not claim. It does not say a named person was swapped for another named person. It matches organisational unit and occupation — which is the only thing personnel data can show at scale.
Section II What OPM says back
OPM Director Scott Kupor disputed it on a specific point: that an occupational series is too broad to work as a proxy for replacement, because one series can hold quite different jobs across a large subcomponent.
His position was that 85 per cent of the reductions were real and lasting, putting the replaced share at around 15 per cent.
Section III Why both are describing the same number
Here is the part the coverage mostly skipped. Run the arithmetic.
The two claims, side by side
Which is what OPM means by around 15 per cent replaced and 85 per cent lasting.
The dispute is definitional, not arithmetic.
So the question is not whose number is right. It is whether refilling the same occupation in the same organisational unit within months means the work came back.
If you think it does, 20,600 reads as a substantial partial reversal of a reduction the government paid for twice — once in leave, once in salary. If you think a series is too coarse a unit to prove it, the same figure reads as ordinary churn around a real cut.
It supports the count. It cannot settle the label. Anyone telling you the figure proves the programme failed, or proves it worked, is adding a judgement the data does not contain.
Section IV The finding that survives
Set the argument aside and one finding holds either way.
Where positions were refilled, the people hired came in at an average of about 1.4 General Schedule grades below those who left.
| Where | Average grade gap |
|---|---|
| Across refilled positions | 1.4 grades below |
| FBI criminal investigation series | 3.1 grades below |
This does not depend on accepting the word replacement. It describes only the people actually hired into those series. A gap of that size is a real loss of experience in the role whatever the hiring is called — and in an investigative series, three grades is the difference between someone running complex cases and someone learning to.
Section V If you took the DRP
The practical reading is that agencies are recruiting into the occupations that were vacated. Whether you can go back turns on reinstatement rules, which have nothing to do with the programme itself.
Reinstatement is generally open without time limit to preference eligibles and to those who completed career tenure. For everyone else it runs three years from separation from the last qualifying appointment, with certain intervening federal service extending the window.
Most participants separated around the end of September 2025. For anyone on the three-year clock, that window runs toward late 2028.
Coming Back to Federal Service has the rules in full and an eligibility checker.
Two cautions. Reinstatement conveys status eligibility to be considered for vacancies — not a job, and not your old grade. And the backfill data shows hiring into these series running well below the departing grade, so treat the grade as a negotiation rather than a restoration. Check whether the announcement carries promotion potential back to your former level, because that is often the real term worth arguing over.
Section VI What it cost
The Government Accountability Office, in GAO-26-108477, estimated that salary for paid administrative leave for participants came to about $6.7 billion in 2025 — roughly 70 per cent of the $9.5 billion the government spent on paid administrative leave that year.
Two things to hold alongside that figure. It excludes continuing health and life insurance and accruing retirement benefits, so it understates the total. And the GAO cautioned it is not precise, because some agencies recorded programme leave together with other leave types and holiday pay — and OPM does not plan to correct those historical records retroactively.
One number, two labels
20,600 hires against 144,000 participants.
If you left, or are weighing an offer like this
- Check your reinstatement eligibility before assuming the door is closed.
- If you are on the three-year clock, date it from your actual separation, not your last working day.
- Preference eligibles and those with career tenure have no time limit — confirm which you are.
- Search your own occupational series; it may be hiring again in your former agency.
- Expect to be offered below your old grade, and read the promotion potential on the announcement.
- Keep your SF-50s and last ratings — you will need them to evidence status and prior grade.
- If a similar offer comes round again, price the paid-leave period against what returning costs you in grade.
Section VII Questions
The Government Accountability Office puts participation at roughly 144,000 employees across 2025. About 126,000 of them were on paid administrative leave for at least two months and nearly 100,000 for at least three months before their service formally ended, which for most participants was the end of September 2025. For scale, the civil service shrank by more than 300,000 people over that period, and the deferred resignation accounted for something like 41 per cent of those separations. The rest came through retirements, reductions in force, probationary terminations and ordinary attrition.
Its analysis, released on 20 August 2026 and built on OPM's own data, matched separations against subsequent hiring and found that by June 2026 nearly 20,600 people had been hired into the same agency subcomponents and the same occupational series that deferred resignation separations had left. The Partnership's summary was that many employees were paid to resign only to be replaced just months later, often at a more junior level. The analysis does not claim that a named individual was swapped for another named individual; it matches organisational unit and occupation.
Yes, and on a specific point. OPM Director Scott Kupor argued that occupational series are too broad to work as a proxy for replacement, since a single series can contain quite different jobs across a large subcomponent, and he wrote that 85 per cent of the reductions were real and lasting, putting the replaced share at around 15 per cent. That objection goes to the label rather than the count. He is describing the same arithmetic the Partnership describes, and disagreeing about whether hiring into the same unit and the same occupation should be called replacement.
On the numbers, both. Roughly 20,600 hires measured against roughly 144,000 participants is about 14 per cent, which is what OPM means by around 15 per cent replaced and 85 per cent lasting. The disagreement is definitional, not arithmetic. If you think refilling the same occupation in the same organisational unit within months means the work came back, the figure reads as a substantial reversal. If you think a series is too coarse a unit to prove that, it reads as noise around a real reduction. The honest answer is that the data supports the count and cannot settle the label.
The grades. Where positions were refilled, the people hired came in at an average of about 1.4 General Schedule grades below those who left. In the Federal Bureau of Investigation's criminal investigation series the gap averaged 3.1 grades. That finding does not depend on whether you accept the word replacement, because it describes only the people who were in fact hired into those series, and a grade gap of that size represents a real loss of experience in the role regardless of what the hiring is called.
The Government Accountability Office, in report GAO-26-108477, estimated that salary for paid administrative leave for participants came to about 6.7 billion dollars in 2025, within a total of roughly 9.5 billion dollars spent on paid administrative leave across the government that year. The deferred resignation therefore accounted for something near 70 per cent of it. The estimate excludes continuing health and life insurance costs and accruing retirement benefits, so it understates the full figure, and the GAO cautioned that it is not precise because some agencies recorded programme leave together with other leave types and holiday pay.
Possibly, and the hiring pattern suggests agencies are in fact recruiting into the occupations that were vacated. Whether you are eligible turns on reinstatement rules rather than on the programme: reinstatement is generally open without time limit to preference eligibles and to those who completed career tenure, and for everyone else it runs for three years from separation from the last qualifying appointment, with certain intervening federal service extending that window. Because most participants separated around the end of September 2025, the three-year window for those who hold one runs toward late 2028.
Plan on not getting it. Reinstatement conveys status eligibility to be considered for vacancies, not an entitlement to a job or to the grade you held, and the backfill data shows hiring into these series running well below the departing grade level. Treat the grade as a negotiation rather than a restoration, look closely at whether the announcement carries promotion potential back to your former level, and remember that a lower grade on return affects your high-3 only if you stay long enough at that grade for it to enter the calculation.