Abstract
Urban crime, especially property crime, is one of the main challenges
facing public safety in Brazil, with significant territorial heterogeneity.
In this context, understanding the spatial distribution of theft is
essential to support more effective and targeted public policies. This
article investigates the spatial patterns of theft rates in the state of São
Paulo in the year 2024, with the aim of identifying the existence of spatial
clusters as well as possible effects of temporal inertia. To this end,
Exploratory Spatial Data Analysis (ESDA) is used, through Global and Local
Moran’s I statistics (LISA), in addition to the application of a spatial
ANOVA model, with and without time lags. The results show a strong positive
spatial autocorrelation, with a concentration of high theft rates in the São
Paulo Metropolitan Region and Baixada Santista. Additionally, temporal
persistence of these spatial patterns is observed, especially in high crime
clusters. Diagnostic tests indicate the presence of autocorrelation in the
residuals and heteroscedasticity, suggesting the need to employ more robust
spatial models, such as SAR and SEM, in future analyses.