Results indicated that the NCAR demonstrated different levels of predictive validity by gender and race/ethnicity. Specifically, the NCAR predicted recidivism for all groups except White females and underpredicted recidivism for African-American youths. One problem may be the brevity of the NCAR, previously considered an administrative advantage, because it includes so few predictors. One strategy for improving the predictive validity of the NCAR is to expand the scope of measured risk. The administrative records of 9,534 adjudicated offenders were examined for their risk assessment scores, ratings of the seriousness of the intake offense, dates of disposition, disposition levels, and demographic information. Data were analyzed using the Kaplan-Meier Product Limit estimator and Cox regression. The author recommends caution in generalizing these findings to other juvenile justice risk assessment instruments. Tables, notes, references