Abstract
Predictive policing algorithms have emerged as prominent technological tools in modern law enforcement, purporting to optimize patrol allocation and deter property crimes through spatio-temporal forecasting. This study evaluates the empirical efficacy of predictive patrol deployment algorithms in mitigating residential burglaries within the North Precinct of the Seattle Police Department from 2018 through 2022. Utilizing a quasi-experimental difference-in-differences design paired with spatial displacement analyses, we examine crime incident logs across treated patrol beats subjected to algorithm-directed directed patrol boxes versus synthetic control beats employing traditional hot-spot tactics. The findings reveal a statistically significant initial reduction of 7.4% in residential burglaries within primary target grids during the first eighteen months of implementation. However, this deterrent effect decayed substantially in subsequent years, accompanied by noticeable spatial displacement into contiguous buffer micro-zones and diminished patrol dosage fidelity. Furthermore, contextual shifts during the COVID-19 pandemic altered baseline residential occupancy patterns, attenuating the algorithmic model's predictive accuracy. We conclude that while predictive policing provides marginal short-term gains in tactical resource allocation, its sustained efficacy against residential burglary is constrained by dosage decay, spatial displacement, and an inability to account for dynamic sociological determinants of acquisitive crime.
