The current study uses data from the first two waves (1994 and 1996) of the
National Longitudinal Study of Adolescent Health (Add Health). Add Health is a school-
based national representative study of adolescents in grades 7 through 12 and contains
information on the individual, family, school and community levels. More information
about the data and sample design can be obtained at
Individuals’ suicide attempt reports are available in the second wave of data collection.
All other individual measures are drawn from the first wave of data collection. Tract and
county level measures are drawn from the Add Health-wave-one-contextual-database.
We have a final sample size of 6,406 respondents within 314 neighborhoods.
To capture the particular effect of individual and structural level predictors on
adolescents’ number of suicide attempts, we use multilevel analysis. This methodological
technique, known also as Hierarchal Linear Modeling, allows more elaborative and
precise analysis of individuals’ behaviors within primary structural units (for instance,
neighborhoods, schools, organizations, etc). In this work we use two levels Poisson
model (HGLM) to simultaneously estimate individual and neighborhood level predictors
of adolescents’ suicide attempts (Raudenbush & Bryk, 2002).
RESULTS & CONCLUSIONS
Table 2 presents results from the multilevel analysis. Model 1 results support the
Durkheimian assumptions, indicating that high proportion of conservative adherents
within county significantly reduce adolescents’ suicide attempt event rate. In addition
(and contrary to the social disorganization perspective) high population density is
significantly and negatively related to lower suicide attempt event rate. In particular, one
unit increase in tract’s population density reduces adolescent’s suicide attempt event rate
by nearly 8%.
Convention