The study, conducted by Bomi Jin and Lonn Lanza-Kaduce of the Department of Sociology and Criminology & Law at the University of Florida, is published in the American Journal of Criminal Justice. Rather than treating schools simply as protective institutions that supervise young people, the authors reframe them as structured social environments that organize adolescents’ face-to-face interactions — including interactions with marijuana-using peers. When schools closed, that organizing structure dissolved; when they reopened, it snapped back, with measurable consequences for the transmission of substance use behavior through peer networks.

The theoretical foundation draws on Akers’ social structure and social learning framework, an extension of Edwin Sutherland’s classic differential association theory. In its simplest form, differential association holds that criminal and deviant behavior is learned through interaction with intimate personal groups, where individuals acquire not only the techniques of deviance but also the attitudes, definitions, and reinforcement patterns that sustain it. Social learning theory, developed by Ronald Akers and colleagues, added mechanisms of differential reinforcement, imitation, and definitions favorable or unfavorable to law-violating behavior. Meta-analytic reviews have consistently ranked peer associations among the strongest predictors of adolescent delinquency and substance use. What the structural branch of the theory adds is the insight that macro-level and organizational contexts — neighborhoods, communities, and schools — pattern the frequency, duration, priority, and intensity of those associations.

Jin and Lanza-Kaduce seized on a rare analytical opportunity embedded in the pandemic. Because Florida school districts resumed in-person instruction at different points during the 2020–2021 academic year, students across the state faced sharply different levels of in-person peer contact at the same moment in time. The researchers used this variation in school return timing as a proxy for the degree of face-to-face interaction adolescents had with their peers. Students in early-returning districts were reimmersed in hallways, classrooms, lunchrooms, and after-school spaces where friendships are enacted bodily — through conversation, shared routines, and what interactionist scholars call interaction rituals — while late returnees remained more dependent on digital communication, which prior research suggests carries weaker or at least different socialization signals than physical co-presence.

The empirical analysis drew on statewide survey data from Florida middle and high school students in grades 6 through 12, with an analytical sample of 7,656 adolescents. The data came from the Florida Youth Substance Abuse Survey, an instrument designed to measure risk and protective factors for adolescent substance use and problem behaviors. The outcome of interest was adolescents’ own marijuana use, and the central explanatory variable was friends’ marijuana use, alongside the school return timing that distinguished early returnees from later ones. The models also controlled for established individual and family-level predictors, including low self-control, measured through items adapted from Grasmick and colleagues’ widely used scale, and parental monitoring, with full item wording provided in the study’s online supplement.

Methodologically, the study confronted several challenges familiar to survey researchers. Missing data were substantial and, importantly, not missing completely at random: additional analyses following the approach of Meldrum and colleagues revealed significant mean differences on study variables between cases with and without missing responses, undermining the assumption of randomness. The authors therefore employed multiple imputation using chained equations, a modern technique that preserves uncertainty while filling in missing values based on observed relationships. Because marijuana use in the sample was a relatively low-prevalence behavior, which can bias maximum-likelihood estimates in logistic regression, the researchers also ran penalized Firth logistic regressions, a technique designed to stabilize estimates for rare events; the substantive conclusions were unchanged. To interpret the interaction between school return timing and friends’ use in a nonlinear probability model, the study drew on contemporary best practices for presenting and testing interaction effects in generalized linear models.

The results split cleanly into two parts. Evidence for mediation — the idea that returning to school earlier indirectly increases marijuana use by increasing friends’ marijuana use — was, in the authors’ words, limited. The structural timing of school reopening did not simply manufacture new marijuana-using friendships that then produced use. Instead, the striking finding was moderation: the association between friends’ marijuana use and adolescents’ own use was substantially stronger among early returnees than among later returnees. In other words, having marijuana-using friends predicted a student’s own marijuana use much more powerfully when those friendships were being enacted in person at school. The effect was particularly pronounced among adolescents with multiple marijuana-using friends, suggesting a dose-response pattern in which exposure to several using peers, combined with daily physical co-presence, created the strongest gravitational pull toward use.

These findings carry weight for how scientists understand the mechanics of peer influence. Social learning researchers have long distinguished between what peers think and what peers do, and laboratory and field work on ‘deviancy training’ has shown that adolescent friendships can become engines of reinforcement for rule-breaking talk and behavior. The Florida results imply that this training accelerates in the environments where it can be rehearsed and rewarded most consistently — the structured but only partially supervised social spaces of the school day and its surrounding routines. They also align with research on unstructured socializing, which finds that unsupervised time with peers is a robust situational trigger for delinquency, and with studies showing that school-adjacent settings shape patterns of adolescent alcohol and marijuana use. The virtual substitutes adolescents relied on during remote schooling — messaging, social media, gaming — appear to have sustained friendships but not to have replicated the intensity of in-person reinforcement, consistent with emerging evidence that online and offline peer influences operate differently.

The study is not without limitations, and the authors are candid about them. The data are cross-sectional, which constrains causal inference; perceived friends’ use may partly reflect projection, the well-documented tendency of adolescents to assume their friends behave as they do. The marijuana measure did not separately capture edible forms of the drug. And because the sample is statewide rather than nationally representative, generalization requires caution. Nevertheless, the pandemic-generated variation in school return timing offers a form of leverage that ordinary survey designs rarely achieve, and the consistency of the moderation result across estimation strategies strengthens confidence in the core conclusion.

The practical implications reach beyond criminology into public health and school policy. Prevention programs have traditionally targeted individual adolescents — educating them about drug risks, building refusal skills, strengthening self-control. The Florida study suggests that such individual-level interventions, while valuable, miss a structural layer: the social settings through which marijuana-using friends exert their influence. Schools, in this framing, are not merely backdrops but active organizers of peer exposure. Prevention efforts, the authors argue, may benefit from extending into the social environments where in-person interactions with marijuana-using peers occur — the rhythms of the school day, the transitions between classes, the informal gatherings before and after instruction — with particular attention to students who report multiple friends who use marijuana. As districts worldwide continue to weigh the costs and benefits of remote, hybrid, and in-person instruction, the study adds an unexpected variable to the calculus: the timing of a school’s reopening may shape not only learning loss and mental health, but the very social channels through which adolescent substance use spreads.

Subject of Research: How variation in school return timing during the COVID-19 pandemic moderated the association between deviant peer association and adolescent marijuana use.

Article Title: Structural Context, Deviant Peer Association, and Adolescent Marijuana Use: Evidence from Variation in School Return Timing during the Pandemic

Article References: Jin, B., & Lanza-Kaduce, L. (2026). Structural Context, Deviant Peer Association, and Adolescent Marijuana Use: Evidence from Variation in School Return Timing during the Pandemic. American Journal of Criminal Justice. https://doi.org/10.1007/s12103-026-09947-7

Image Credits: AI Generated

DOI: 10.1007/s12103-026-09947-7

Keywords: adolescent marijuana use, peer influence, differential association, social learning theory, school return timing, COVID-19, Florida Youth Substance Abuse Survey, criminology, substance use prevention, deviant peers, Structural, Context