Multilevel modeling approach to adolescent risk perception in the neighborhood setting

Using multilevel modeling, this study will examine how risk propensity during earlier adolescent years can determine deleterious health behavior, including co-morbid substance use and risky sexual behavior in young adulthood. Data from the National Longitudinal Survey on Youth Young Adult waves of 1998, 2000 and 2002 will be analyzed to investigate between and within group variation by region of study subjects—an application of Hierarchical Linear Modeling. A single level model will first test the effects of personal characteristics on adolescents' risk perception, and second the adolescents' neighborhood scale ratings on risk perception, as outcomes using linear regression, run cross-sectionally for each year and then longitudinally. The model in the second step will evaluate the different effects of adolescent personal characteristics and adolescents' perception of their neighborhoods on (1) alcohol use, (2) drug use, (3) tobacco and (4) sexual behavior in four separate regressions, again both cross-sectionally and longitudinally. The level 1 model will assess the association between personal and neighborhood characteristics. The level 2 model will evaluate how the adolescents' own risk perception varies within and among urban and rural areas according to neighborhood characteristics as appraised by adolescents themselves. This research addresses an understudied area in the literature as defined by Bronfenbrenner's ecological model of the interaction between environment—the neighborhood—and the individual—the adolescent. Mother's educational attainment—as a resource for information and appraisal about risks--will be introduced in this analysis as a social support proxy and buffer to offset the effects of neighborhood quality.