When employing survey data on adolescents the most common methodological problem is the underestimation of prevalence rates because the most severe cases are not usually recruited. Fortunately, the sample used in the current analyses was recruited at a young age and did not exclude respondents with behavioral problems or psychiatric disorders. Second, because additional data sources (such as parents and teachers) were unavailable we had to assume that respondent’s reports were truthful. Interestingly, only 2.4–4% of the sample violated our model of progression (i.e., reported a lower level of lifetime involvement at Wave 2 than at Wave 1), suggesting a high level of accuracy in reporting. Third, other covariates of substance abuse (e.g., conduct disorder, attention deficit hyperactivity disorder, depression) (Button et al., 2007; Crowley et al., 1998; Ferdinand et al., 2001; Riggs et al., 1999; Tarter, 2002; Whitmore et al., 1997) were not included in the logistic regression models. While we recognize that omitting these predictors is a limitation of this study, the focus of this paper was determining whether the risks for substance problems in young adulthood were
