Crash Report Sampling System (CRSS) Processed Data – rfars

The General Estimates System (GES), redesigned as the Crash Report Sampling System (CRSS) in 2016, is a nationally representative sample of police-reported motor vehicle crashes in the U.S., maintained by the National Highway Traffic Safety Administration (NHTSA). While the raw GES/CRSS data are publicly available, they are distributed in formats that require significant preprocessing before analysis. This dataset provides a processed version of the GES/CRSS database (2014-2023), optimized for immediate use in R, Python, Excel, and other modern data tools. It was prepared by the rfars R package, enabling researchers, policymakers, and educators to access and analyze crash data more efficiently and reproducibly.

This dataset is provided in three different file formats to support a wide range of users and analysis environments:

• CSV (.csv) - a plain text format, with each of the five tables saved separately. These files are the most universally compatible and can be opened directly in Excel, though they are larger.
• Parquet (.parquet) - a modern, compressed, cross-platform columnar format. Each of the five tables is provided separately (fars_accident.parquet, fars_vehicle.parquet, etc.). This version is recommended for Python, SQL, and advanced R users, as it is smaller than CSV and loads quickly in most data science environments.
• RDS (.rds) - a native R format containing the full dataset as a list of five related tables (accident, vehicle, person, drugs, distract). This version is recommended for R users, since it can be loaded in one step with readRDS() and preserves all variable types exactly as processed.