The proposed research will use state-of-the art statistical methods to (a) describe relationships between gateway behaviors (e.g., cigarettes and alcohol, dropping out of school) and illicit drug use, and (b) test Elliott's integrated theory of delinquency and illicit drug use. We will use data from the National Youth Survey (NYS) which is a multi-wave longitudinal study that is based on a probability sample of 1 ,725 American youth aged Il to 17 in 1977. The NYS study includes six waves of assessment over a seven year period. We will use log-linear models for multi wave dichotomous non-repeatable event data to describe how change in potential gateway behaviors (e.g., dropping out of school, graduating from school, criminal behavior, spending more time with deviant peers) relates to change in illicit drug use. We also seek to identify predictors of users from low risk backgrounds (e.g., high SES) and non-users living in neighborhoods where there is high drug use. We are interested in both the change to the use of different illicit drugs and the cessation of use in response to the life transition from adolescence to young adulthood. In addition, we will use random effects models for continuous and ordinal outcome variables to describe the patterns of increased use and cessation of use of marijuana. We will also use random effects models to test Elliott's theory. The major constructs in Elliott's theory that we will examine are demographic variables relating to social disorganization, strain produced by blocked opportunities to achieve personal goals (e.g., a high school graduate who is unemployed), conventional external bonding (e.g., time spent at school and time spent with one's family), conventional internal bonding (e.g., attachment to family, school and societal norms) and deviant peer bonding (e.g., having friends who use illicit drugs). We will test the validity of all our models in different subpopulations by performing subgroup analyses on each ethnic (Black vs. White) and gender subgrouping. Our focus throughout will be to identify predictor variables that have the potential to guide successful prevention efforts.
