Counting vs. weighting: five agencies write 77% of federal rules, but only 53% of the economically significant ones

Summary

Rule counts and rule weight tell very different stories about which federal agencies drive regulatory burden. Over the last five presidential years (Feb 2021 - Jan 2026), the Department of Transportation, DHS, EPA, Commerce and HHS together issued 76.9% of all final rules published across the 17 executive-branch agencies subject to OIRA's EO 12866 significance review. But weight those same rules by whether OMB/OIRA designated them "economically significant" (an anticipated annual effect on the economy of $100 million or more, or other major impact) and that same top-5-by-volume group accounts for only 53.1% of significant rules. The reshuffling is dramatic at the agency level: HHS writes only 4.7% of all rules (5th place by raw count) but a full 27.5% of economically significant ones (1st place) — driven by its annual Medicare/Medicaid payment-rule cycle. DOT, the single largest rule-writer by volume (29.8% of all rules, mostly routine FAA airworthiness directives and pipeline/hazmat notices), drops to roughly 4.0% of significant rules. Labor and Energy likewise punch far above their raw-count weight once significance is applied.

Method and data

Two figures were needed for every agency: (1) the total count of RULE-type Federal Register documents, and (2) the count of those rules OIRA flagged as economically significant under Executive Order 12866. The AskAmerica connector's fedregister schema has both concepts as tables (fr_documents and fr_significance), but a direct query of fr_significance returned only 822 total rows, all dated 2026, with just 6 rows flagged significant=true — despite the catalog's own metadata describing the table as holding 167,652 rows spanning 2013-2026. This is a live, verified data-quality defect in the connector (reported via report_issue during this session), and it makes the connector unusable for a multi-year significance-weighted analysis. fr_documents itself is populated and confirms the raw rule-count side of the story (e.g., DOT issued 4,492 RULE documents 2020-2025, more than any other agency), but without a working significance table there was no way to compute the weighted comparison from AskAmerica data alone.

The analysis instead uses the George Washington University Regulatory Studies Center's RegStats project, which independently ingests and republishes the Federal Register API's own `significant` field (the same source field fr_significance was supposed to carry) as clean, downloadable CSVs, broken out by agency and presidential year back to 1981 (2003 for DHS, which did not exist earlier). Two GWU files were fetched directly and parsed: the full Federal Register rules-by-agency file (proposed and final rule counts per agency per year, 1995-2025) and the economically-significant-rules-by-agency file. Figures reported here are final-rule counts only, summed across presidential years 2021-2025, for the 17 executive-branch departments/agencies GWU's own significance file covers.

Independent regulatory agencies (FCC, SEC, NRC, FTC, CFTC, FDIC, FHFA, etc.) are excluded from the significance side of the comparison, and this is a real scope limit, not an oversight: those agencies are broadly exempt from EO 12866 review and neither GWU's dataset nor the Federal Register API's own field assigns them a significance determination. Several of them are large rule-writers by raw count (FCC alone issued 508 rules over 2021-2025, more than half of HHS's total) whose economic weight simply cannot be assessed on this metric. Any claim about "regulatory burden concentration" below is therefore scoped to EO-12866-covered executive agencies.

The reshuffle, agency by agency

The two rankings barely resemble each other. Only EPA holds a top-3 spot on both measures. DOT and Commerce are dominant by volume and marginal by weight; HHS, Labor and Energy are the reverse. A Spearman-style eyeball check: of the 5 agencies ranked highest by raw count, only 2 (DHS, EPA) remain in the top 5 by significance.

Why the divergence, and what it means

The mechanism is well documented in the regulatory-process literature: DOT's rule count is inflated by thousands of narrow, individually mandated actions — FAA airworthiness directives for specific aircraft models, PHMSA hazardous-materials exemptions, and Coast Guard local notices — each of which is a separate Federal Register document but rarely clears the $100 million economic-effect threshold. HHS's rule count is comparatively small, but a handful of its rules — the annual Medicare Hospital Inpatient (IPPS) and Physician Fee Schedule updates, Medicaid/CHIP program rules, and (in recent years) ACA marketplace rules — each redirect tens of billions of dollars and are almost automatically economically significant. Labor's rise on the weighted measure reflects OSHA and wage-and-hour rules (e.g., the 2024 overtime-threshold rule) that are comparatively rare in count but large in scope; Energy's reflects appliance and vehicle efficiency standards.

The practical upshot: a headline like "DOT and DHS are the biggest rule-writers" is true of paperwork volume but misleading about where the compliance cost and OMB oversight burden actually falls. On the weighted measure, HHS alone accounts for more economically significant rulemaking (27.5%) than the volume leader DOT, DHS and Commerce combined (13.0%). Anyone assessing regulatory burden by simple rule counts — a common shorthand in political and press coverage — is measuring administrative throughput, not economic impact.

What This Report Does Not Answer

• Cover independent regulatory agencies (FCC, SEC, NRC, etc.) in the weighted comparison: Not possible: independent regulatory agencies are broadly exempt from EO 12866 review and carry no significance determination in any available source (GWU dataset or the Federal Register API's own significant field). Their raw rule counts are reported and flagged, but they cannot be included on the significance-weighted side of the comparison. This is a scope limit of the underlying government data, not an omission in the analysis.

Every query behind this report

3 warehouse calls ran in this session, in order. Each is reproducible against the same snapshot.

SELECT "year", count(*) AS n, sum(case when significant then 1 else 0 end) AS sig_true, sum(case when significant is null then 1 else 0 end) AS sig_null FROM fedregister.fr_significance GROUP BY "year" ORDER BY "year"

SELECT count(*) AS total FROM fedregister.fr_significance

SELECT agency_names, count(*) AS n_rules
FROM fedregister.fr_documents
WHERE doc_type = 'RULE' AND "year" BETWEEN '2020' AND '2025'
GROUP BY agency_names
ORDER BY n_rules DESC
FETCH FIRST 20 ROWS ONLY