Answers are grounded in this paper and cite their sources.
Crime clusters in a few places, but missing coordinates make it hard to tell which components of a place host which crimes, and when. Treating São Paulo as a tissue, we take 1.65 million parcels as cells, census tracts as spots of known composition and 0.69 million police records of 15 crime types as the phenotype. Each class of cell has a distinct crime profile that holds at finer scale (median Spearman ρ = 0.78) and was replicated in 19 held-out districts tested once against criteria fixed beforehand. Daily rhythms of phone crime vary by class and do not follow the crowd: robbery risk per pedestrian is highest at night. Beyond composition, persistent place effects account for at least as much variation in tract crime counts. Month-to-month variation is concentrated in a few high-crime tracts and was poorly forecast. The approach requires only a cadastre, a business registry and police records.
No takes yet. Share an insight, caveat, or question.
Toquetão et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: