Ambient Population and Crime: The Fragility of Causal Links in Urban Environments

The relationship between ambient population and crime has long been studied, with competing theories suggesting that pedestrian activity may enhance guardianship and informal surveillance, but also increase suitable targets. Yet, empirical evidence remains mainly correlational, despite causal identification being essential for policy. We assess how robustly this link can be causally identified using high-resolution smartphone-derived footfall data from Baltimore, Chicago, and Philadelphia (2023–2025), estimating increasingly rigorous models—from naïve baselines to two-stage residual inclusion with distributed lags, fixed effects, and spatial spillovers. The estimated associations remain small and essentially unchanged after addressing endogeneity, while precision drops. The few significant effects are sensitive to crime type, temporal granularity, and context, and none survives multiple-testing correction. This fragility reflects operationalization, not theory: competing mechanisms move with headcounts, but in opposite directions, making null and offsetting effects indistinguishable. Ambient population should therefore be measured by who is present, where, and why, not how many.