The Risk of Informal Socializing with Peers: Considering Gender Differences Across Predatory Delinquency and Substance Use

Previous research has demonstrated that adolescents who socialize with peers in unstructured and unsupervised settings are more likely to engage in deviant behavior. Research on this front has generally pooled adolescents together, suggesting that it is a risk for nearly all youth across a wide array of deviant outcomes. The current study instead hypothesizes that the strength of the relationship between time use and different forms of deviance varies for male and female adolescents. Specifically, it proposes that unstructured and unsupervised socializing with peers will be a significantly stronger risk for predatory delinquency (i.e. violent and property crime) for male adolescents than for females, whereas it will be an equivalent risk across gender for substance use. Analyses using the AddHealth data support this hypothesis. The discussion considers the implications of these results. Keywords: time useroutine activitiesgenderdelinquencysubstance use Acknowledgments This research uses data from Add Health, a program project directed by Kathleen Mullan Harris and designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris at the University of North Carolina at Chapel Hill, and funded by grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, with cooperative funding from 23 other federal agencies and foundations. Special acknowledgment is due to Ronald R. Rindfuss and Barbara Entwisle for assistance in the original design. Information on how to obtain the AddHealth data files is available on the Add Health website (http://www.cpc.unc.edu/addhealth). No direct support was received from grant P01-HD31921 for this analysis. Notes 1. Scholars also suggest there are type-scripts according to race, sexual identity, SES, and other attributes—as well as the intersection of these characteristics (e.g. an African-American woman from a disadvantaged neighborhood; see Simpson & Gibbs, Citation2006). 2. For a detailed description of the AddHealth research design see the project’s website at: http://www.cpc.unc.edu/projects/addhealth/design.html. 3. This attrition is primarily due to the peer deviance measure. Though accepting so much data loss for a control variable may seem odd, omitting it from the analysis could reduce the validity of our core measure, informal socializing. There are alternative perceptual measures of peer deviance—which ask the respondent how many of his or her three best friends have engaged in deviant behavior—but these measures are severely limited for two main reasons: they only ask about substance use (i.e. cigarettes, alcohol and marijuana); and, they are only in reference to the three closest friends, not the peer group. Thus, the peer deviance measure used here, though missing data, is theoretically more valid and meaningful. Information on our measure of peer deviance could be missing because of any one of the following reasons: the subject did not complete the in-school survey; s/he did not identify any in-school friends; and/or, these identified friends did not provide a valid response to all of the seven items from which we construct the peer deviance factor. Individuals with missing data on this measure were different from those who did have valid data. Those with missing data were more likely to be male, were slightly older (about two months, on average), were less likely to be White, and tended to report more delinquency and substance use at Wave II. Though the amount of data loss when using the AddHealth to study peer processes is unfortunate and greatly restricts generalizability, the availability of various peer and deviance measures provides a unique opportunity for empirical inquiry. Even so, as a form of sensitivity analysis, we ran our models excluding this measure (therefore with a larger sample size). Our results are substantively the same; gender conditions unstructured socializing when predicting predatory delinquency, but not substance use. We thank a reviewer for this recommendation. 4. We considered constructing similar dependent variables based on Wave I data, with the idea that cross-sectional models may provide an important test of the immediate “situational” risk under consideration (i.e. informal socializing with peers). Unfortunately, doing so would pose significant problems for temporal order. To be specific, our independent variable from Wave I asks about behavior during the previous week, yet the measures on delinquency and substance use at Wave I query about behavior either during the last 30 days or the previous 12 months. Thus, because it is very plausible that a cross-sectional model would reflect dependent variables that preceded the independent variable, we opted to rely only on longitudinal analysis. 5. In part, we opted for a factor score as an outcome because answers to the 11 predatory delinquency items are not on the same scale, which would make a simple summed scale inappropriate. 6. It is also reasonable to view this measure as an ordinal variable, therefore suggesting the use of ordinal logistic regression for relevant analyses. Doing so violated the proportional regression assumption (p < .01), however. Because of this, as well as the fact that one can rightly claim this measure is a count variable, we view it as such and employ negative binomial regression. 7. As we later note, this is one reason why we include a measure of direct parental supervision as a control. 8. The AddHealth survey also queried respondents about whether they met up with each nominated in-school friend to “hang out or go somewhere” (0 = no, 1 = yes) as well as whether they had spent time with each friend the previous weekend (0 = no, 1 = yes). Because these measures do not necessarily speak to informal socializing with a group of friends, because they could capture such behaviors as going to dinner or a sporting event with family members and friends, and because they only reflect in-school friends, we were reticent to rely on them as our main proxy for informal socializing with peers. Even so, we created an alternative measure as a form of sensitivity analysis. Responses to these two questions were summed across all nominated friends and then divided by the number of friends in order to obtain an average of how much each respondent hung out with his or her peers after school and socialized with peers over the weekend. This score was then standardized and summed with the standardized score for the measure used in our main analyses (i.e. how often did you just hang out with friends?; see also Haynie & Osgood, Citation2005, for a similar measure). When this measure was substituted for the original measure, the findings largely remained the same. Informal socializing with peers was a statistically significant predictor of substance use for both males and females and this risk was not significantly different across gender. Alternatively, informal socializing with peers was a statistically significant predictor of predatory delinquency for male adolescents, but not for females (and the slopes were significantly different from one another). 9. As with the predatory delinquency outcome, we relied on a factor score rather than a summed scale in part because the answers to the individual items were not all on the same scale. 10. Admittedly, this measure captures the deviance of in-school friends. Even so, we believe it is important to include a proxy for peer deviance as a control (see Haynie & Osgood, Citation2005 who used a similar peer deviance measure). It should be noted that prior work has examined whether or not unstructured socializing with peers is moderated by delinquent peers, with some studies revealing such a relationship (Bernburg & Thorlindsson, Citation2001; Thorlindsson & Bernburg, Citation2006) and others not (Haynie & Osgood, Citation2005). Additional work on this premise would be theoretically informative and consistent with the spirit we adopt here of questioning whether informal socializing is a larger risk for some people or under certain conditions, but is not central to our current investigation. 11. For a detailed description on the sampling strategy and the creation of the grand sample weights, see “Grand Sample Weights” by Tourangeau and Shin available at: http://www.cpc.unc.edu/projects/addhealth/pubs/guides 12. The substance use count measure demonstrated overdispersion. 13. Some scholars may question whether this gender difference is driven by the inclusion of violent crime measures in the predatory delinquency factor. Even though this inclusion is fully consistent with our argument and focus, we completed supplemental analyses in which our outcome was a factor score based only on the property crime items (i.e. excluding fighting, robbery, brandishing a knife or gun, and shooting or stabbing someone). With this factor as the outcome, the risk of informal socializing with peers is still significantly stronger for male adolescents than it is for female adolescents. 14. As a form of sensitivity analysis, we also ran the regression models for male and female subsamples and then subsequently determined whether the slopes for informal socializing were different based upon the equality of coefficients test (see Brame, Paternoster, Mazerolle, & Piquero, Citation1998; Paternoster, Brame, Mazerolle, & Piquero, Citation1998). These results were consistent with the findings from our main models—the slopes were significantly different for predatory delinquency, but not when substance use was the outcome. 15. Given the arguments made here, some scholars may question whether the gender composition of the peer group impacts or conditions the effects of time use on delinquency and substance use. Though one can construct a measure capturing the ratio of females to males for the nominated in-school friendship network only, we did complete supplemental analyses investigating this premise. Results demonstrate that controlling for the gender composition of the peer group has no impact on the time use coefficients (i.e. the substantive pattern of results remains the same to those reported here). It is also worth noting that the risk of unstructured socializing with peers may shift and change as a consequence of peer group size, with larger peer groups providing more motivation and opportunities to engage in delinquency, as well as a more powerful sense of diffusion of responsibility (see McGloin & Piquero, Citation2009). In order to determine whether differences in peer group size can account for any of the results observed here, we also ran supplementary models in which we controlled for the number of nominated friends. The inclusion of this measure did not alter the findings. 16. Again, the hypothesis was also supported when controlling for the gender composition and size of the nominated peer group.