Substance use in the college-age population is an important public health and educational concern. This study compared rates of use among college students and nonstudents, including high school dropouts, from a single data source representative of the nation. Rates of use were estimated from the combined National Household Surveys on Drug Abuse from 1991 to 1993. Logistic regression models were used to test the effects of educational status and living arrangement. Educational status and living arrangement were found to be significant predictors of substance use. Rates of illicit drug and cigarette use were highest among high school dropouts, while current and heavy alcohol use were highest among college students who did not live with their parents. Substantial variation in substance use patterns within the college-age population suggests that overall rates of use for young adults should not be used to characterize specific subgroups of young adults. These data from a single source will thus help planners more clearly distinguish the service needs of the diverse subgroups within this population.