An influential literature has used the aftermath of terrorist attacks to estimate large effects of police street deployment on crime. However, the elasticities obtained in these settings may not easily extrapolate to more standard circumstances. This paper exploits a natural experiment that aimed to increase police presence in more than 6,000 well-defined areas, by economically-realistic amounts and under relatively normal circumstances. Using data transmitted by GPS devices worn by police officers, we first document exogenous and discontinuous changes in patrolling intensity. We do not find that these increases in patrolling were accompanied by corresponding decreases in crime. The standard errors are small enough to reject relatively small elasticities. We discuss and empirically evaluate explanations for our findings.