Many scholars argue that collective efficacy is not relevant to understanding crime at the microgeographic level. We examine variation in collective efficacy across streets with different levels of crime in Baltimore City, MD, and, then, employ multilevel modelling to assess this relationship. We find that people who live in crime hot spots have much lower levels of collective efficacy than people who live in non-hot spot streets and that this relationship persists when controlling for a large number of potential confounders both at the street and community levels. These findings suggest the importance of collective efficacy both in understanding and controlling crime at microgeographic units.