Purpose: Although much of the literature shows a linkage between socioeconomic status and violent behavior, there is a lack of investigations that attempt to sort out etiologies of violent behavior. In this study, we attempt to test a theory that more frequent and intense exposure to violence predicts higher rates of violent behavior. This project was funded by two National Institute on Drug Abuse grants to examine the demographic, psychosocial, mental health, and cognitive characteristics of youth incarcerated in the Missouri State Division of Youth Services (DYS) residential treatment system.

Methods: The Missouri DYS residential treatment system is comprised of 32 facilities; approximately 716 youth are in residence at any one time. The study was completed in June 2004, with 723 youth interviewed and a 98.5% participation rate. Youth completed a 60-to-90 minute interview including standardized measures of trauma exposure, substance use, mental health symptoms, suicidality, delinquent activity, and psychological and cognitive characteristics. The study was approved by the Washington University IRB and federal Office of Human Research Protection. Multiple linear regression was used to account for variance in continuous variables of violent and non-violent offending.

Results: Of the 723 youth interviewed, 629 male youth (87%) are included in this analysis. Mean age of all male youth interviewed was 15.5 years. Previous-year exposure to violence predicts increased violent offending more strongly (b= .78, p<.001) than any other continuous variables included in the analysis even when potential confounds such as drug and alcohol use and SES are controlled for. The variable shows high tolerance (.70) in the regression analysis, and only moderate correlation (.50) with the dependent variable, thus ruling out the possibility multicollinearity. Other significant predictors of increased previous-year violence include African American race/ethnicity (b= 3.66 p< .001), previous-year alcohol use (b= .37, p<.05) younger age at first crime committed (b= -.37, p<.05) and previous-week self-reported hostility (b= .46, p<.001). Predictors of decreased previous-year violence included previous week depression (b= -.28, p<.05) and, surprisingly, previous-year cigarette smoking (b=-.35, p<.05).

Implications for practice and policy: Understanding the predictive role of exposure to violence in increased offending, and its interaction with other variables such as alcohol use, supports the idea that practitioners and policymakers must work cooperatively to design and apply multi-systemic interventions – from national policy down to the clinical level – to effectively address problems of violence among adolescents.