Assessments
Youths and their parents were surveyed at recruitment in
the fall of the fifth grade (mean age 10.8 years, SD 0.5), in
the spring of the fifth grade, and annually thereafter in the
spring through 1991. Assessment was through a classroom
survey at ages 10 and 11 and through face-to-face interviews
from age 13 onwards. Participants, but not their
parents, were interviewed again in the spring of 1993, and in
1996 when the participants were 21 years of age. Signed
parental consent was obtained from youths’ caretakers and
assent was obtained from youths for data collected before
age 18, and consent was obtained from youths themselves
after age 18. Data collection procedures were approved by
a university human subjects internal review board. Annual
participation rates were consistently high; more than 93% of
the original sample during the last five waves of interviews.
A variety of overlapping methods are used to locate hardto-
find respondents including: a list of people identified by the participant who will always know where they are; online
address update, telephone, and reverse directory services;
Department of Licensing records; and address correction
mailings. Nonparticipation at age 21 was not related
to gender, ethnicity, or use of tobacco or alcohol, or delinquency
at any age 10 to 21.
The Seattle Social Development Project is a theorydriven
study with a preventive intervention nested within it
during the elementary school grades (see Hawkins et al. [21]
for a complete description of the intervention). Special care
must be taken when conducting studies of etiology on samples
that contain interventions, because analyses that do not
take the intervention into account may be subject to threats
to validity. Although analyses of intervention effects have
found differences in the levels and prevalence of risk and
protective factors and outcomes between groups [21], we
have found little evidence of differences among the intervention
groups in the relationships among variables related
to the etiology of substance use and other child outcomes
(e.g., [22,23]). However, before analyzing the effects of the
predictors, we first examined the interactions between
SSDP intervention and the predictors to see whether the
intervention significantly changed the relationships between
these predictors and the outcome. When no significant interactions
were found, we continued the analysis without
including intervention; otherwise, we incorporated the intervention
effect into the model.