De-identified U.S. Fatal Crash Events with Standardized Operational, Environmental, and Participant Classifications (2004, 2014, 2024)
Description
This dataset consists of nine comma-separated value (CSV) files and one codebook file: • cleaned_2004_accident.csv • cleaned_2004_person.csv • cleaned_2004_vehicle.csv • cleaned_2014_accident.csv • cleaned_2014_person.csv • cleaned_2014_vehicle.csv • cleaned_2024_accident.csv • cleaned_2024_person.csv • cleaned_2024_vehicle.csv
Files
Steps to reproduce
1. Data acquisition Download raw Fatality Analysis Reporting System (FARS) data for years 2004, 2014, and 2024 from NHTSA. Obtain the three primary tables for each year: accident, vehicle, and person. 2. File standardization Rename files to include year identifiers (e.g., 2004_accident.csv, 2014_vehicle.csv). Ensure consistent column headers across years (e.g., ST_CASE, YEAR, MONTH). Create a unique crash identifier by concatenating year and case number if needed. 3. Variable selection Retain only variables required for standardized classification: Accident-level lighting condition weather manner of collision harmful event route type urban/rural month and year fatality count pedestrian count person count Person-level age person type (for minor indicator) Vehicle-level body type model year speed damage severity alcohol involvement 4. Harmonization across years Align categorical values that differ across years into a unified structure. Resolve inconsistencies in naming conventions and coding schemes. Remove duplicate or conflicting variables across tables. 5. Standardized classification mapping Transform all retained variables into the AAA_ZZZZZ coding system. Apply consistent categorical bins for: year and month speed ranges vehicle age damage severity alcohol involvement collision type harmful event lighting and weather route and urban/rural classification fatality, pedestrian, and person counts age groups and minor presence Mappings are defined in the included codebook 6. Derived variables Construct additional indicators: vehicle age categories from model year speed bins from raw speed values minor presence indicator from person-level ages aggregated person counts per crash pedestrian involvement categories 7. Table reduction Accident table retain only standardized environmental and outcome variables Person table reduce to age group and minor presence indicators Vehicle table retain standardized operational and mechanical variables 8. Data cleaning Handle missing or unknown values using standardized “unknown” categories. Remove null or invalid records where necessary. Ensure all categorical outputs conform to the coding system. 9. Output generation Export final datasets as: cleaned_2004_accident.csv cleaned_2004_person.csv cleaned_2004_vehicle.csv cleaned_2014_accident.csv cleaned_2014_person.csv cleaned_2014_vehicle.csv cleaned_2024_accident.csv cleaned_2024_person.csv cleaned_2024_vehicle.csv