Neighborhood Disadvantage, Social Groups, and Adolescent Violence: Assessing Mechanisms in Structural-Cultural Theories

Abstract

Structural-cultural perspectives link contextual characteristics to interactions with associates who transmit definitions favorable to crime, thus influencing behavior. Drawing on this, we predict that: (1) adolescents in disadvantaged neighborhoods will be more likely to have parents, friends, and schoolmates who hold violent rationalizations, (2) exposure to associates who hold violent rationalizations will increase an adolescent’s own violent rationalizations, and (3) one’s willingness to rationalize violence will be associated with self-reported violent conduct. Using panel data from the University of Missouri – St. Louis Comprehensive School Safety Initiative and structural equation models, our findings indicate that respondents residing in disadvantaged neighborhoods are more likely to be exposed to associates who report a greater willingness to rationalize violence, which in turn predicts an adolescent’s own rationalizations. Further, one’s willingness to rationalize violence is predictive of violent conduct and mediates the effects of associates’ attitudes on self-reported violence.

Keywords:

• Neighborhoods
• rationalizations
• structural-cultural theories
• violence

Acknowledgements

We would like to thank Shaina Herman and Holly Nguyen for their helpful comments on earlier versions of this manuscript. This research was supported in part by Award No. 2015-CK-BX-0021 awarded by the National Institute of Justice, Office of Justice Programs, U.S. Department of Justice. The opinions, findings, and conclusions or recommendations expressed in this publication are those of the authors and do not necessarily reflect the views of the Department of Justice.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Notes

1 In Akers (1998) theory, differential reinforcement (as opposed to definitions) is the proximate cause of crime. Wikstrom (2014) incorporates a complex conditional interplay between morality and self-control. Beyond the theoretical issues with these perspectives (e.g., Hirschi, 1996), extant research has indicated that Akers’ additions of imitation and reinforcements offer little in terms of increased explanatory power (Pratt et al., 2010) and that the processes outlined by Wikstrom (2014) generally have weak empirical support (Kammigan, 2023).

2 We also estimated models with a binary indicator of violence (=1 if a respondent committed any violent act at wave 3 and 0 otherwise) as well as a Rasch model capturing violent tendencies as a latent trait (Raudenbush et al., 2003) using logit and tobit models, respectively. The substantive results are identical to those discussed in the main text. We present these supplemental analyses in the Supplemental Information Table S1.

3 One concern may be that there is substantial theoretical and empirical overlap between the perceived peer rationalizations scale and the measure of schoolmate rationalizations since schools can be a primary source by which youth find and develop friendships. While it is certainly true that schools offer a social context in which individuals forge friendships, intimate peer groups and broader schoolmates are conceptually and empirically distinct. First, research using data from the National Longitudinal Study of Adolescent Health has indicated that, on average, many respondents report at least one close friend who does not attend their same school, with some respondents even reporting that all of their close friends are from outside of their school (Haynie & Payne, 2006). Second, research has also indicated that the pool of schoolmates is larger, distinct, and more diverse than more intimate friendship groups, and that schoolmates still have an impact on individuals’ behaviors and attitudes (McGloin et al., 2014). Finally, the correlation between our estimated schoolmate rationalization measure and friends’ rationalizations in our data is weak to moderate (r = .38). Taken together, these suggest that the two measures do not substantially overlap conceptually or empirically.

4 The overwhelming majority of respondents identify as either white (44.56%) or black (47.69%), which is consistent with the racial-ethnic demographics of St. Louis County.

5 VanderWeele (2016) has demonstrated that conventional mediation analyses often conflate a “true” mediation effect as well as a moderating effect (e.g., that the impact of X depends on the value of M), which could potentially bias estimates. Accordingly, VanderWeele has encouraged scholars interested in mediation to examine potential moderation and how it, in turn, impacts estimates of direct and indirect effects. We explored moderated-mediation in a number of ways, including using a continuous interaction (VanderWeele & Vansteelandt, 2014) and splitting the samples across various levels of our mediators (e.g., +/- .5 SD, +/- 1SD) (VanderWeele & Vansteelandt, 2014). In no instances were the estimated moderations significant. More importantly, the size of the direct and indirect effects were similar across all models.

6 Scholars have demonstrated that conventional approaches—like the product of coefficient method—used to decompose direct and indirect effects can be particularly problematic with a non-linear models, such as Poisson estimation, when moderation between X and the mediators are present and have encouraged the use counterfactual decompositions that allow for moderating effects (VanderWeele, 2016). However, “For linear models and log-linear models, they will coincide when there is no exposure–mediator interaction” (Valeri & VanderWeele, 2013, p. 142). We tested and found no evidence for exposure-mediator interactions and thus proceed with a conventional product of coefficient approach.