Baseline, one-year and two-year follow-up interviews were obtained from 164 arrested youths processed at a juvenile assessment center in a prospective longitudinal study. A structural equation model that included cross-sectional and longitudinal associations among drug (alcohol and marijuana), drug sales and index offenses was supported by the data. Research, theoretical, and policy implications of the results are drawn. (Contains 38 references, 4 tables, and 3 figures.) (Author)