This study investigates whether experience with cocaine and the perception of risk associated with cocaine use might tend to cluster within neighborhoods and cities in the US. Population-based data from six years of the National Household Surveys on Drug Abuse public use files are employed. The alternating logistic regressions model is used to quantify the extent of geographic concentration. Perceptions of the harm associated with cocaine use and actual experience with cocaine tend to cluster within neighborhoods; once within-neighborhood concentration is taken into account, there is little evidence of residual concentration within cities.