Methodology
How the prototype dataset is built
The pipeline is designed to block uncertain rows rather than invent or smooth missing values.
Sources
Crime data comes from FBI NIBRS state bulk files provided as official source archives. Population context uses Census Vintage 2025 subcounty estimates for the matching data year where available.
Selected offense definition
The current property-crime definition uses NIBRS codes 200, 220, 240, and 23A through 23H: arson, burglary/breaking and entering, motor vehicle theft, and larceny/theft offenses.
Publication gates
A city page must have a publishable city-type FBI agency, NIBRS participation, a unique Census match, non-covered agency status, positive offense counts, and a minimum Census 2024 population of 5,000 for automatic indexation.
Small towns
Small places remain in the review queue. They can be valuable for search, but they should be handled as noindex drafts, county/state rollups, or manually reviewed pages until the match quality and denominator risk are clear.
Trends
The pipeline supports multiple FBI years. A trend is only shown when at least two validated years exist for the same city; otherwise the page explicitly says that trend data is pending.