Variations in drug use have been found across individual-level factors and community characteristics, and by type of drug used. Relatively little research, however, has examined this variation among juvenile offenders. Based on a sample of 924 newly arrested juvenile offenders, two multilevel logistic regression models predicting marijuana test result and cocaine test result were separately examined. The results highlighted a strong association between individual-level socio-demographic characteristics, most notably age and seriousness of arrest charge, and drug use. Residential stability was related to cocaine use, yet none of the community variables were related to marijuana use. These findings suggest that different risk factors are related to adolescent marijuana and cocaine use. The research implications of the findings are discussed. (Contains 4 notes, 3 tables and 1 figure.)