All covariates are measured antecedent to the developmental process under investigation. We control for key demographic factors including age, race, gender, and socio-economic status. Age is a truncated measure of age at Wave 7, and age-squared was included to capture any curvilinear associations.6 Race is measured using a series of dummy variables (Black, Hispanic, Other), where White serves as the reference group. Sex is a dummy indicator of male. Finally, SES is measured at baseline using the parent index of social position, which is based upon occupational and educational scores (see Hollingshead, 1971 (link)); higher scores indicate greater SES. Peer antisocial behavior is measured by asking youth to report the proportion of their friends that engaged in 12 antisocial behaviors on a five-point Likert scale spanning from “none of them” to “all of them” (Thornberry et al., 1994 (link)). To measure youth’s perceptions of success, a six-item measure of expectations for work, family, and law-abiding behavior adapted from the NYS prediction of adult success scale is used (see Menard & Elliott, 1996 ); higher scores indicate more predicted success. To control for differences across locations, we include a dummy indicator for site (1 = Maricopa County, AZ; 0 = Philadelphia County, PA). Street time is measured as the proportion of time on the street (i.e., not in a secure facility) and is included as a covariate where appropriate.7 Finally, as noted above, we account for psychosocial maturity to examine whether health predicts changes in psychosocial maturity. Likewise, we control for substance use and offending when modeling those outcomes as well to examine whether health states and/or psychosocial maturity levels influence changes in behavioral outcomes.