While there is abundant research on correlates and predictors of crime and delinquency in general, there is virtually no research on whether individual characteristics affect patterns of criminal specialization, escalation, and de-escalation. The current study addressed this gap in the literature by exploring the impact of relevant background, behavioral, and social offender characteristics on patterns of offending over the course of the criminal career. Data were drawn from 2 study cohorts of the Predicting Parole Performance in the Era of Crack Cocaine study conducted by the California Youth Authority (CYA) in the 1980's, resulting in a sample of 2,294 of juvenile wards for the current analysis. Variables under examination included offense type at arrest, offense seriousness, measures of family and environment and relationships, and school performance. Results of multinomial logit models indicated that, overall, the correlates of crime were also the correlates of criminal specialization and escalation. Individual characteristics that predicted any criminal behavior also predicted the type of criminal behavior. The findings illustrate how patterns of crime can be tested using multinomial logit models to move beyond aggregate analyses of crime type data. Future research should focus on the potential impact of race on offense patterns using multinomial logit models. Tables, figures, references, appendix