In examining the relationship between substance use among juvenile offenders, this study had three hypotheses: (1) the relationship between substance use and recidivism would be free from interactions with ethnicity, gender, or data source; (2) the relationship would depend on interactions with ethnicity, gender, or data source; and (3) there would be both a main-effect relationship between substance use and recidivism and interactions with ethnicity, gender, or data source. Utilizing the Colorado Longitudinal Youth Study (CLYS), an ongoing study of 505 participants born between 1979 and 1990 each with at least 1 police arrest before age 18, this study accumulated the study sample between 1997 and 2001. Results indicate: (1) parent reports that youths often use substances more than doubles first rearrest risk; (2) averaged youth and parent substance use reports predict recidivism better than a single source; (3) parent or youth denial of youth substance use predicts recidivism; (4) age at first arrest does not predict recidivism; (5) non-White/non-Asians have a 79 percent higher recidivism risk than peers; (6) parent-reported delinquency predicts recidivism with declining accuracy; and (7) substance use robustly predicts recidivism despite prior reported delinquency, gender, ethnicity, age, follow-up, or data source. These findings were evaluated on their support of the host-provocation (HP) theory which is designed to integrate multiple mechanisms that explain harmful adolescent behavior and repeated involvement with law enforcement. References