The Federal Workforce Was Ordered Back to Offices. Washington-Area Headcount Fell by 47,998.

On Jan. 20, 2025, the White House ordered executive-branch agencies to end remote-work arrangements and require employees to work in person at their duty stations, with exemptions left to agency heads. The policy was national. The geography many people associated with it was Washington.

The public workforce record moved in the opposite direction.

FederalHiringData found 222,670 covered employees with published duty stations in the Washington-Arlington-Alexandria metropolitan area in January 2025. By June 2026, the count was 174,672. That was a loss of 47,998 employees, or 21.6%. Published U.S. headcount outside the region fell 12.7% over the same endpoints.

The Washington region therefore did not gain share. It fell from 17.88% of covered employees with a published U.S. duty-station geography to 16.37%, a decline of 1.51 percentage points. In the 17 months after January 2025, the region recorded 60,632 separation actions against 14,868 accessions. The simultaneous hiring freeze and workforce-reduction program overwhelmed any geographic recentralization visible in the aggregate data.

That finding is not an office-attendance claim. OPM's duty station is the official place of work or assignment used for personnel administration. It does not show where someone sat on a Tuesday, how often they teleworked, whether they entered a building, whether they moved, or whether they received an exemption. It also does not establish that RTO caused the headcount decline.

A national return order was not an order to move to Washington

The presidential memorandum told agencies to require full-time in-person work at employees' respective duty stations. It did not tell agencies to transfer the federal workforce to Washington.

OPM's Jan. 22 implementation guidance did address official locations. It recommended that agencies take steps to move the duty station of an employee more than 50 miles from an existing agency office to the most appropriate agency office based on duties and job function. That could produce individual reassignments. The public file does not mark which changes occurred under that paragraph, what distance someone moved, or whether the new office was in the same metropolitan area.

Two other directives arrived at nearly the same time. The Jan. 20 hiring freeze restricted civilian hiring, with specified exemptions. The Feb. 11 workforce-optimization order directed preparations for large-scale reductions in force and, subject to exceptions, a one-hire-for-four-departures ratio after the freeze.

That overlap is central to the result. A return-to-office rule can change work arrangements while a hiring freeze and separation program reduce both Washington and field staffing. Aggregate duty-station data can test whether Washington's share increased. They cannot isolate the causal effect of one order from the others.

The Washington region lost headcount and share

FederalHiringData used OPM's code 47900 for the Washington-Arlington-Alexandria core-based statistical area. It includes duty stations across the District, Maryland, Virginia and West Virginia that OPM maps to the metropolitan region.

The red line fell faster. That difference matters more than the raw national decline for the central question. If RTO had geographically recentralized the published workforce, the NCR could have lost people yet gained share because other regions shrank faster. It did not.

The denominator needs unusual care. OPM's June 2026 file represented 1,953,300 covered employees, but geography was redacted for 883,756. The 174,672 published NCR count was 8.94% of all covered records and 16.37% of the published U.S. subset.

Neither percentage is a full governmentwide Washington estimate. Security-sensitive and other redacted records can include employees inside or outside the region. FederalHiringData therefore uses the observable U.S. subset for the share comparison and labels it throughout. The finding is about the direction of the published series, not an assertion that every federal employee's location is known.

Separations replaced an earlier accession surplus

The personnel-action files show when the employment balance changed.

Accessions are not simply new hires from outside government. OPM's category includes personnel actions that bring employees into the covered population through multiple pathways. Separations also span several exit types. The counts are actions, not a linked employee-level ledger.

Even with that limitation, the equal-window comparison is stark. Before the order, NCR accessions exceeded separations by 6,370. After January 2025, separations exceeded accessions by 45,764. September 2025 alone recorded 17,057 NCR separations, and December recorded 6,086. Those spikes align with the step-down in monthly employment but do not assign motive to any individual departure.

The action balance also answers a common misconception. Moving an existing employee from a remote duty station to an agency office can change a location code without creating an accession. The aggregate files cannot distinguish that reassignment from other location-record changes. Yet a large inflow of employees into Washington would still tend to support the region's headcount or share. The observed balance did not.

The decline reached large departments

Every large department with at least 1,000 published NCR employees in January 2025 lost regional headcount by June 2026.

Health and Human Services had the largest published NCR decline, down 9,286. Homeland Security fell 3,914, Treasury 3,322, Justice 3,069 and Commerce 2,929.

Share and count did not always move together. HHS's NCR share of its published U.S. geography changed little because its field headcount also fell. Education's NCR count declined while its regional share rose, again because the visible workforce outside the NCR contracted more. A rising agency share can therefore coexist with fewer Washington employees.

Component data show where some of the largest losses sat, but reorganizations can change component codes and should not be mistaken for physical moves.

NIH lost 4,042 published NCR employees and FDA lost 2,868. USAID fell from 2,974 to 131. The IRS lost 2,328 NCR employees. The Department of State component fell 2,571. These are endpoint changes, not a list of office closures or proof that each employee left the region.

Administrative and technology occupations fell hardest

The occupational mix points toward a thinning of headquarters-heavy functions rather than a uniform retreat from every kind of work.

Management and program analysis lost 7,535 NCR employees. Miscellaneous administration and program lost 5,796, information technology management 4,538, and general attorneys 3,581.

The counterexamples are important. Patent examining added 507 NCR employees, and air traffic control added 166. The national air traffic control series grew by 976 as well. Those gains show that the regional contraction was selective. They do not overturn the governmentwide result, but they prevent a false claim that every occupation moved away from Washington.

Occupation codes also do not identify a building, telework schedule or RTO exemption. A patent examiner officially assigned within CBSA 47900 can work in a different pattern from an attorney with the same regional code. The chart measures official geography, not presence at a desk.

Recruiting shifted away from Washington even faster

The historical USAJOBS archive supplies a separate measure of where agencies advertised work. FederalHiringData counted distinct announcements opened from Jan. 1 through Aug. 14 in each year, using Aug. 14 because that was the latest 2026 open date in the local research warehouse at analysis time.

The 2024 window contained 231,953 announcements, including 36,634 with at least one NCR proxy location. The 2026 window contained 94,066, including 7,339 NCR announcements. National volume fell 59.4%; the NCR count fell 80.0%.

That reduced the NCR share from 15.79% to 7.8%. The same comparison for 2025 was already lower at 9.67%.

An announcement is not a vacancy or hire. One notice can list many locations and support multiple selections, or none. FederalHiringData counted each control once and marked it NCR if any location matched the city/state pairs OPM assigned to CBSA 47900 in June 2026. That is a consistent recruiting-footprint proxy, not an estimate of Washington jobs filled.

Structured remote and telework fields do not support a comparable 2018-2026 trend because historical extraction is incomplete. In the 2026 same-window data, 1,687 of 7,339 NCR announcements, or 23.0%, carried a telework-eligible flag. None was classified as confirmed remote while also carrying the NCR proxy. Telework eligibility is not full remote work and does not establish how often a selectee would report in person.

The contraction was national, but Washington was not protected

The state table reinforces the distinction between a national workforce reduction and a Washington-specific shift.

The District of Columbia had the largest state-level loss, down 24,481. Maryland was close behind at 24,297. Virginia fell 7,919. Those state totals include duty stations outside CBSA 47900, so they should not be added to reproduce the NCR count.

California, Texas, Georgia, Florida, New York and Pennsylvania also lost thousands. Only Wyoming, up 222, and Montana, up 27, posted positive changes among published state geographies. The cuts were not a simple Washington-to-field redistribution. Most visible state workforces contracted.

Related research

Put this finding in context

All articles

Workforce/August 28, 2026

Federal Bargaining-Unit Coverage Fell by 579,903. Most of That Was Not Job Loss.

Federal represented bargaining-unit coding fell by 579,903 through May 2026. Defense drove 45% of the change, but the count is not job loss or union membership.

Data investigations/August 25, 2026

The Coast Guard Plans to Add 15,000 Military Members. Only 26% of Its Workforce Requirements Are Current.

Recruiting rebounded, but cutter vacancies widened and covered civilian employment fell 9.9% as Force Design accelerated.

Washington's decline predates the current RTO debate

The longer D.C.-only series adds perspective. FedScope recorded 149,905 covered employees with District of Columbia duty stations in September 2000. The series reached 167,893 in September 2010, stood at 123,552 in September 2024 and fell to 96,552 by June 2026.

The D.C. share of all covered records fell from 8.50% in 2000 to 4.94% in June 2026. That long decline is not a clean causal trend. FederalHiringData joins legacy FedScope files through 2014 to modern OPM files beginning in 2015, and public geography coverage changes at the source-era boundary. The modern series also redacts many records.

The defensible conclusion is narrower: Washington was not on an uninterrupted path toward greater concentration before 2025. RTO arrived after years of declining D.C. duty-station share, then overlapped with an unusually large national workforce contraction.

Office capacity and workforce geography answer different questions

The return order followed years of concern about underused federal buildings. GAO found that 17 of 24 agency headquarters buildings operated at 25% of estimated capacity or less during selected weeks in early 2023. It also warned that agencies lacked common utilization benchmarks and that office configurations could complicate space decisions.

Those findings establish a real property-management problem. They do not convert OPM duty stations into attendance records. A regional employee can spend the day at an agency building, in the field, traveling, or under an approved arrangement while retaining the same official duty station. A building can add occupants without adding headcount if existing Washington employees report more often.

The reverse is also possible: a Washington building can feel fuller while regional headcount falls. Attendance frequency and employee count are different numerators. Public badge-swipe, seat-assignment and employee-level relocation records are not available governmentwide, so this article does not claim to measure them.

What the record says about recentralization

The observable test produces a clear answer. The 2025 return-to-in-person policy did not coincide with a higher Washington-region share in OPM's published workforce geography. NCR headcount fell 21.6%, compared with 12.8% outside the region among covered U.S. records with visible locations. Separations exceeded accessions by 45,764 after January 2025. The NCR share of same-window USAJOBS announcements fell too.

That does not mean RTO had no geographic effects. OPM expressly contemplated changing some remote employees' duty stations to agency offices. Individual offices may have consolidated, and a small set of occupations or components grew in Washington. The public data simply do not show those changes adding up to governmentwide recentralization.

The stronger conclusion concerns scale. Whatever location changes RTO produced were smaller than the simultaneous reduction in Washington-area employees and recruiting. The order changed where many remaining employees were expected to appear. It did not preserve, let alone expand, the capital region's share of the covered workforce.

Methodology and limitations

FederalHiringData analyzed OPM Federal Workforce Data employment snapshots from January 2023 through June 2026. NCR means core-based statistical area 47900, Washington-Arlington-Alexandria. Endpoint comparisons use January 2025 and June 2026. The NCR share denominator is covered employment with a published U.S. duty-station country code. Records with redacted geography are excluded from that share but remain visible in the all-covered denominator.

The long-run D.C. series uses September FedScope employment-cube files for 2000-2014, September OPM Federal Workforce Data snapshots for 2015-2025 and June for 2026. It measures District of Columbia duty stations, not the full NCR. The source transition, exclusions and changing redactions prevent treating every level shift as workforce movement.

Accessions and separations use OPM action files from September 2023 through June 2026. Equal 17-month windows are September 2023-January 2025 and February 2025-June 2026. Counts sum OPM's published record count. They do not link a person's accession to a later separation, distinguish all external hires from other accession paths, or identify RTO causation.

Historical USAJOBS results use the FederalHiringData archive, whose coverage begins in March 2017. The chart starts in 2018 and compares Jan. 1-Aug. 14 each year through 2026. The archive was current through announcements opened Aug. 14, 2026 at analysis time. The NCR proxy is the set of D.C., Maryland, Virginia and West Virginia city/state pairs that OPM mapped to CBSA 47900 in June 2026. Each control number counts once. Announcements are not vacancies, applications, referrals, interviews, selections, hires or employees.

OPM-covered employment excludes or redacts some security-sensitive and non-covered populations and does not include contractors. Duty station is not daily work location, office capacity, residence or commute. This analysis does not estimate relocations, exemptions, telework days, building attendance, costs, productivity or causal effects.

The downloadable research data include the monthly employment, personnel-action period, department, occupation, state and USAJOBS series used in the analysis.

Photo: Senior Airman Madelyn Keech, U.S. Air Force, via DVIDS, public domain. The photo shows Pentagon employees and families at an April 2025 workplace event; it is illustrative and does not represent the governmentwide sample.