The risk assessment instrument developed in the current study from a pooled sample worked efficiently for males. Predictor variables associated with female recidivism, such as emotional abuse, sexual abuse, truancy, and running away, were available in the secondary dataset. However, due to the smaller size of the female strata, these variables did not attain statistical significance and as such were not included in the risk assessment instrument. The area under the curve (AUC) for the females in the separate validation sample was .611, which suggests that the instrument was only a marginal improvement over chance prediction of recidivism for youthful female probationers. Although the current study incorporated dynamic variables of substance abuse, family, school, and peers, needs assessment variables were not available, thus limiting the current assessment to a second-generation risk prediction. Using secondary data to create the risk assessment instrument likely contributed to a lower overall level of predictability for both girls and boys in the model developed than would have been possible if information had been gathered from multiple data sources to create a comprehensive assessment of risk factors and specific predictors. Data were collected from 8,293 juvenile probation referrals from the Texas Juvenile Probation Commission database. Tables and references