The goal of this study was to assess longitudinal, predictive relationships between community violent crime and reactive and proactive aggression. Community violent crime data were gathered from local law enforcement agencies and combined with an existing dataset of at-risk youth. Aggression was assessed by parents using the Reactive and Proactive Aggression Questionnaire (RPQ). Data were examined over four time points. Autoregressive cross-lagged modeling was used to test two models: one for proactive aggression and one for reactive aggression. Results revealed a positive relationship between community violent crime and proactive aggression, whereas the model including reactive aggression had poor model fit. Therefore, results support reactive and proactive aggression as distinct constructs. Findings also demonstrate that publicly accessible violent crime data can be used to predict children's behavior over time. Finally, results have important implications for preventive interventions for at-risk youth.