Using automated vehicle locator data to classify discretionary police patrol across space

Place-based policing strategies assume officers have sufficient discretionary time to engage in proactive crime prevention activities, yet little research examines the spatial distribution of uncommitted patrol time. This study analyzes automated vehicle locator (AVL) data from Manchester, New Hampshire (2022−2023) to classify discretionary police patrol patterns across 5878 street segments. Using over 9.7 million GPS coordinates recorded every 10 s from patrol vehicles, we measure the proportion of patrol time that is discretionary (not assigned to a call for service) versus committed (assigned to a call for service). Group-based trajectory modeling identified three distinct groups: 55 % of street segments had high discretionary time (averaging 50 % of monthly patrol time), 26 % had medium discretionary time (averaging 38 % of monthly patrol time), and 19 % had low discretionary time (averaging 36 % of monthly patrol time). Anselin Moran's I revealed significant clustering of both high and low discretionary time segments. Multinomial logistic regression examined factors predicting discretionary time clusters. Notably, none of the crime measures significantly predicted high discretionary time hot spots, suggesting a potential misalignment between patrol availability and crime prevention needs. Traffic-related calls for service were the only police measure positively associated with high discretionary time clusters. Areas with higher concentrated disadvantage and ambient population were more likely to be discretionary time hot spots, while theft was associated with lower odds of high discretionary clustering. These findings challenge assumptions underlying place-based policing interventions and suggest that discretionary patrol time may not naturally concentrate where it is most needed for crime prevention.