Research Summary: Multi-Scale Spatio-Temporal Exploratory Analysis of Urban Vehicle Theft

A longitudinal multi-scale analysis of 521,098 Los Angeles Police Department records from 2000 to 2022 demonstrates that urban vehicle theft is highly concentrated geographically, with over half of all incidents clustered in just 10.8 percent of 250-meter cells and specific census tracts persisting as hotspots for over two decades. The findings underscore that leveraging administrative public safety data requires extensive curation, including separating original offenses from recoveries and harmonizing shifting jurisdictional boundaries.

Why it matters

This analysis demonstrates that vehicle theft is structurally entrenched in specific micro-locations over decades rather than occurring as a diffuse or transient phenomenon. Recognizing this chronic spatial concentration allows public and private stakeholders to focus interventions and security investments precisely where risk is highest. Additionally, it highlights that raw administrative data cannot guide strategy effectively without rigorous cleaning and validation.

Key insights

• Analysis of 521,098 LAPD stolen-vehicle reports (2000-2022) revealed extreme spatial and temporal concentration across multi-scale resolutions.
• Vehicle theft is persistently concentrated, with half of all recorded thefts occurring within 10.8 percent of 250-meter grid cells.
• Six census tracts remained statistically significant theft rate hotspots continuously across all 23 years after false discovery rate control.
• Administrative records required substantial data hygiene to be analytically viable, including decoupling thefts from recoveries and traffic reports, correcting timestamps, and validating geocoding.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2610.36478