OSHA Citations — Q1 2026 (Citation-Level)

Overview

This dataset contains every OSHA citation issued during the first quarter of 2026 (January 1 through March 31), joined to inspection-level and employer-level context. The unit of observation is the individual citation — not the inspection or the employer — so each row represents exactly one cite. An inspection that produced six citations contributes six rows; an employer cited at five separate inspections during the quarter contributes one row per citation across them.

OSHA issues citations after an inspection finds a workplace safety violation. Each citation references a specific OSHA standard (the "standard cited" field is not in this release; see Known limitations). The four most consequential fields for downstream analysis are viol_type (the severity class — Serious, Willful, Repeat, Other), gravity (1–10 inspector-assessed severity), current_penalty (the post-settlement penalty), and agencies_alltime (the cross-agency footprint of the cited employer).

Data sources

• U.S. Department of Labor, OSHA — Inspection and Violation Records (public OIS extracts published via the DOL v4 enforcement API). Refreshed daily at FastDOL.
• FastDOL employer-resolution pipeline. Each OSHA inspection's establishment-name + state + zip5 tuple resolves to a deterministic UUID (employer_id), making citations joinable to other federal datasets keyed on the same identity.
• SEC EDGAR Exhibit 21 + curated parent overrides — populates the parent_name column when an OSHA-cited establishment is owned by a publicly-disclosed parent.

All source data is public-records information released by the Department of Labor under FOIA and posted by the agency in machine-readable form.

Methodology

Citations are pulled from FastDOL's violation_detail table, which mirrors OSHA's IMIS violation feed at the citation grain. Each violation is joined to its parent inspection (inspection_detail, on activity_nr), and the inspection is joined to the resolved employer (employer_profile, on employer_id). The Q1 2026 filter is applied on vd.issuance_date >= 2026-01-01 AND vd.issuance_date < 2026-04-01.

The employer_name column is lightly cleaned with REGEXP_REPLACE to strip OSHA's legacy case-prefix patterns (e.g. "CPX2026XJW184X0206 - B & F CONTRACTING" → "B & F CONTRACTING"). The agencies_alltime column is an in-database aggregation: the full set of federal agencies with any recorded activity against the same employer_id across the entire FastDOL history, joined as a pipe-delimited string for CSV compatibility.

Refresh cadence

This release is a fixed Q1 2026 snapshot. The upstream OSHA bronze feed refreshes daily; future releases will issue quarterly snapshots (osha-citations-2026-q2, osha-citations-2026-q3, etc.) following the same methodology so quarter-over-quarter comparisons are point-in-time clean.

Known limitations

• Federal-OSHA jurisdiction only. OSHA shares enforcement with 22 state-plan states (California, Oregon, Washington, North Carolina, and others). State-plan citations are reported separately by each state agency and are not included here. Coverage gaps are largest in those state-plan states.
• standard_cited and contest_date are not included in this release. Both are spec'd in the OSHA feed schema but not currently bound by the FastDOL loader. A follow-up release will add them.
• Penalty amounts reflect assessment, not collection. OSHA's collection records are not part of the public extract; the current_penalty field is post-settlement but pre-collection.
• Inspection-level context is duplicated across citation rows. When an inspection produced multiple citations, the inspection-level columns (inspection_open_date, inspection_type, state, zip5, employer columns) repeat across each row. Sum or count carefully — joining at the inspection level can multiply counts.
• agencies_alltime is all-time, not Q1-specific. Use it as a cross-agency footprint signal, not as a Q1-period filter.
• Employer resolution is high-confidence but not perfect. False positives (two distinct employers folded into one cluster) and false negatives (one employer split across two clusters) both occur and are most common for generic LLC names and DBAs.

Use cases

• Journalism and policy research on Q1 2026 OSHA enforcement patterns — which industries, geographies, employer sizes saw the most serious / willful / repeat citations.
• Industry benchmarking — pair agencies_alltime with viol_type to surface multi-agency repeat offenders.
• Compliance and risk — underwriting or procurement diligence workflows can filter on current_penalty > X or viol_type IN ('W', 'R') for high-signal employer flagging.
• Academic study of OSHA enforcement gravity scoring, penalty-to-violation patterns, or post-settlement adjustment behavior across NAICS sectors.
• Cross-walk with the WHD wage-theft dataset (linked under Related datasets) — the employer_id column is the same across both, so employers appearing in both feeds in the same quarter are a particularly high-signal cohort.