Abstract
Aker's (1998) theory of social structure and social learning (SSSL) argues that structural variations in deviant behavior, such as gender or race/ethnic differences in underage or heavy drinking, are mediated by social learning variables. However, longitudinal analyses of deviant drinking in an urban sample of white, black, and Hispanic adolescents fail to support the SSSL mediation hypothesis. Significant gender and race/ethnic differences persist after controls for social learning variables as well as for social bonding variables. Interaction effects involving two social bonding variables—family attachment and moral belief—point to theoretically important conditions that maintain a gender gap in underage drinking and relatively low levels of deviant drinking among African American adolescents.
Acknowledgments
We are grateful for the advice and assistance of Jay Turner, Richard Wilsnack, Don Lloyd, and John Reynolds.
Notes
1Unlike Monitoring the Future and other national surveys, the Add Health data analyzed by Watt and Rogers (Citation2007) do not show significant race differences in heavy drinking or in the 30-day prevalence of illegal drug use.
2See Gil and colleagues (Citation2002) for details on sampling and survey administration. Gil and colleagues report that there were no statistically significant differences in 28 behaviors and family factors between wave 3 respondents who did not complete the wave 4 survey versus those who did. Also, the high school drop-out rate among young adults in the wave 4 sample is similar to that of the population of students in the Miami-Dade public school system.
3Although a few frequent drinkers drink one to two drinks three or more times a week, the vast majority drinks three or more per occasion. The distribution of wave 4 respondents across these six quantity frequency categories is as follows: 39.9% non-drinkers, 9.3% low-infrequent drinkers, 6.8% low-weekly drinkers, 12.9% high-infrequent drinkers, 16.8% high-weekly drinkers, and 14.2% are frequent drinkers.
Note. Means are followed by subsets of the sample in parentheses (n).
a Main effect of gender significant at p < .01.
b Main effect of race/ethnicity significant at p < .01.
c Main effect of race/ethnicity significant at p < .05.
4In preliminary analyses, we also examined a scale of Familism, which is comprised of seven items dealing with family, trust, loyalty, and pride (e.g., “My family shares values and beliefs,” “I can express my feelings with my family”). However, this scale did not yield any significant main effects or interactions in equations for our three measures of deviant drinking. Given these negative results and the conceptual overlap of this scale with our measure of Family Attachment, we do not include it in this report. Similar to other research on social bonding theory, our wave 3 data do not include a measure of the temporal bond of “involvement.”
5We used the SPSS Binary Logistic Regression program to estimate equations for our analyses of Underage Drinking and DSM-IV Alcohol Abuse. The equations for the Q-F scale were estimated with the SPSS Linear Regression program. All analyses employed list-wise deletion of missing data. We used product terms to estimate two-way and three-way interaction effects, retaining those interactions that made a statistically significant improvement (p < .05) in the fit of logistic regression equations or variance explained in OLS regression equations.
6The two-way interaction effect of gender × race/ethnicity on Underage Drinking was marginally significant at p = .078 as was the main effect of gender on Alcohol Abuse at p = .061.
*p ≤ .05. **p ≤ .01.
*p ≤ .05. **p ≤ .01.
*p ≤ .05. **p ≤ .01.