01 Two point three million
people work for
the federal government.

This page reads the U.S. federal civilian executive-branch workforce as of the September 2024 snapshot. The source is OPM FedScope Employment, in its record-level release on data.opm.gov - one row per federal civilian employee. The pipeline aggregates that file directly (see scripts/build-data.ts).

What is real vs. a stand-in

Real, from the record-level file: every headcount, the veteran and STEM shares, the year-over-year change, the GS-grade distribution, the geography section (by locality pay area), and most salary medians. Curated stand-ins: only the salary cells for groups OPM redacts at source. OPM stores REDACTED in place of pay for ~45% of records - 100% of the Department of Defense (a third of the workforce), plus 72% of DHS and 64% of DOJ. Where more than half of a group's pay is redacted, no honest median can be computed, so those cells keep a curated placeholder that is marked with an est badge everywhere it appears, never shown as a real number: the agency medians for DoD, DHS and DOJ, and the occupation medians for Contracting (1102), General Engineer (0801), Criminal Investigator (1811) and Police (0083). Every other salary on the page is a real FY2024 median over the unredacted records.

What FedScope captures

Executive-branch civilian employees, which is the lion's share of the federal workforce visible in this view: 2.31 million people. Not in this dataset: active-duty military (1.3M), Postal Service (~600K, a separate accounting), the Legislative branch (Congress, GAO, CBO, Library of Congress), the Judicial branch, and several intelligence agencies that are statutorily redacted.

Pictograms as data marks

Each Aicher-style figure represents a fixed unit (25,000 people for agencies, 15,000 for geography, 100,000 for the masthead total). Counts are rounded to the nearest whole figure; long tails are summarized with a "+N" overflow. Different occupational series get different pictograms (lawyer / doctor / engineer / agent / manager) but the figure-as-unit remains constant.

Pay distributions

OPM's only pay column is annualized adjusted basic pay - base pay plus locality, the figure on an employee's SF-50. For each occupational series the bar shows the real 10th-to-90th-percentile spread over the unredacted records, with the median as a vertical mark; physician (0602) and air traffic controller (2152) break the GS ceiling via Title 38 and premium pay. The four series where OPM redacts most of the pay - Contracting (1102), General Engineer (0801), Criminal Investigator (1811), Police (0083) - instead carry an est-badged stand-in, because a percentile spread over only the unredacted minority would be biased. Headcounts are always real.

Geography and locality pay

Geography groups the workforce by locality pay area - the unit OPM actually sets pay against. Headcounts and medians per area are real, but they cover only the ~55% of employees with a disclosed duty station: OPM redacts location for the same records it redacts pay (all of DoD), so these are disclosed-duty-station counts, not full-population totals, and they understate exactly the places where DoD civilians concentrate. Treat each area's figure as a floor. The locality percentages are the published 2024 OPM rates (Rest of U.S. 16.82%, Washington-Baltimore 33.26%, San Jose-San Francisco 45.41%); a GS-13 nets those points above base pay.

What's still curated

The seven est-badged salary cells above (redaction, non-random). Also absent: career length / tenure trends, separation rates, time-in-grade, and per-agency occupation/grade splits. The compare page reads the real headcount / salary (where real) / veteran / STEM fields; the searchable "any agency × any agency" picker is the next pass.