Safety event reporting narratives, safety outcomes, and operational data across four industries

Data are curated versions of publicly available safety event reports, safety outcomes, and operational measures across rail, nuclear power, coal mining, and aviation. Data are organized by industry. For each industry the collection provides three linked components: (i) event reports, including the free-text incident narratives from which organizational-safety-climate measures are derived; (ii) safety outcomes, i.e. counts and severities of accidents, injuries, and other adverse events; and (iii) operational measures that quantify each organization's exposure and activity, used as modeling offsets and covariates. All sources are U.S. federal safety-reporting systems; the curated files standardize identifiers, dates, and outcome fields and aggregate operational data to the analytic panel cadence. Where noted, source files extend earlier than the modeling window, which is bounded by the availability of matching operational-exposure data. Rail (Federal Railroad Administration, FRA). Event narratives and casualty outcomes are drawn from the FRA Rail Equipment Accident/Incident reporting system (Form 6180.54); operational exposure — train-miles, passenger-miles, and employee-hours — is from the FRA Injury/Illness Summary operational source data (Form 55). Event reports span 1982–2025; operational data span 1975–2025. The unit of analysis is the railroad-month. Nuclear power (Nuclear Regulatory Commission, NRC). Event narratives and outcomes are from the NRC Licensee Event Report (LER) system, retrieved via the Idaho National Laboratory LER Search (lersearch.inl.gov); these yield licensee event reports, reactor scrams, emergency-classification declarations, and power-loss measures. Operational exposure — reactor-days at power and daily power variability, aggregated across units to the site level — is from NRC daily power-reactor status data. LER reports span 1994–2026; operational data span 2000–2025, which bounds the nuclear analytic window. The unit of analysis is the site-quarter. Coal mining (Mine Safety and Health Administration, MSHA). Event narratives and injury outcomes are from the MSHA accident/injury reporting system (Form 7000-1) via MSHA Open Data; operational exposure (employee-hours worked) is from the MSHA quarterly mine employment/production data, and regulatory covariates (citations, significant-and-substantial violations) from the MSHA violations/inspections databases. The analysis is restricted to coal operations (a mines master table provides commodity classification). Accident and operational records span 2000–2025 (events through mid-2026). The unit of analysis is the mine-quarter. Aviation (four sources). Event narratives supplying the climate measure are from the NASA Aviation Safety Reporting System (ASRS), a voluntary, confidential reporting system, covering 1988–2025. Safety outcomes are drawn from two independent registries: the National Transportation Safety Board (NTSB) aviation accident database (serious and fatal accidents; combined pre-2008 and post-2008 records, source coverage 1948–2023) and the FAA Accident/Incident Data System (AIDS; no-harm and property-damage events, source coverage extending to 1945–2026). Operational exposure — departures, seats, and passengers by airport — is from the U.S. Bureau of Transportation Statistics T-100 Segment data, covering 1990–2023, which bounds the aviation analytic window. Outcome and operational records are linked to the nearest FAA-coded airport within 5 nautical miles, and the analysis is restricted to 1988 onward. The unit of analysis is the airport-quarter. Each industry directory contains the curated analysis-ready files together with the corresponding raw source extracts, so that the processing from public source data to modeling inputs is fully reproducible. Retrieval dates are indicated in source-file names where applicable (e.g., FRA Form 55 and NRC enforcement exports retrieved February 2026).