Social and Economic Correlates of Weapon-Carrying in Violence-Exposed Urban Young Black Males

Abstract

Firearm homicide is a leading cause of death among children and young men in the U.S. (ages 1–19), with young Black males in urban environments facing rates 18-to-24-fold higher than their non-Hispanic White peers in 2023. A key precursor to firearm violence victimization is weapon-carrying behavior (WCB), defined as carrying, concealing, or displaying firearms or other weapons in community or social contexts that elevate risk for injury, interpersonal threats, or law enforcement contact. Several structural, behavioral, and trauma-based risk factors fuel weapon-carrying. Yet these WCBs are rarely studied in tandem, leaving a critical gap in our understanding of these high-risk behaviors for youth. This cross-sectional study leveraged baseline data from a convenience sample of 226 violence-exposed urban young Black males, ages 15–24 (M_age = 18.3 years; SD = 3.1) enrolled in a trauma-informed digital firearm violence prevention pilot study. Eligibility required prior personal or witnessed experience of youth violence; reported prevalence therefore characterizes a high-risk subgroup rather than urban young Black males as a whole. Past-30-day weapon-carrying frequency was measured across five YRBS-aligned categories (0, 1, 2 to 3, 4 to 5, and 6+ days) and modeled as a categorical index under negative binomial regression. Associations with peer and community violence exposure, substance use, sociodemographic, and socioeconomic factors were estimated as incidence rate ratios (IRRs) with 95% CI. Past-30-day weapon carrying was reported by 42.5% of participants, with carrying frequency ranging from 1 day to 6 or more days. Participants reported high levels of direct victimization (64.8%), witnessing community violence (76.4%), and use of nonprescribed medications, including in instances preceding violence. In the fully adjusted model, indicators of violence exposure were the most consistent correlates of carrying. Direct victimization (IRR = 1.15, p < 0.05), general exposure to violence or aggression (IRR = 7.82, p < 0.01), and physical fighting (IRR = 1.11, p < 0.05) remained independently significant. Conversely, associations with substance use, dating aggression, and employment were attenuated, suggesting shared ecological vulnerability rather than independent causal pathways. Findings underscore the central role of chronic violence exposure and support the need for trauma-informed, multilevel prevention strategies in clinical and community settings.

1. Introduction

Weapon carriage among children and adolescents remains a critical public health challenge. In contexts where chronic community violence intersects with structural disinvestment, carrying a weapon may emerge as a calculated survival strategy, not solely an act of delinquency (Anderson, 2000; Jay, 2023; Sampson & Levy, 2022). Adolescents exposed to chronic neighborhood violence are significantly more likely to carry a gun compared to those not exposed, even after controlling for mental health, substance use, and prior victimization (Baiden et al., 2024). In these environments, limited access to protective resources may compel youth to navigate a landscape of perceived and actual threat, where weapon possession is viewed as a necessary means of self-protection (Jay, 2023; Sampson & Levy, 2022). Non-recreational weapon possession by minors is widely considered a high-risk behavior given its legality and association with elevated risk of unintentional injury, retaliatory violence, suicidal behavior, and escalation during interpersonal conflict (Ruggles & Rajan, 2014).

Weapon-carrying behaviors (WCBs) are defined in this study as the possession, carrying, display, or use of weapons (e.g., firearms, knives, blunt objects) in non-recreational, high-risk contexts. These contexts include public streets, schools, unattended home settings, and online spaces, particularly in situations involving substance use or retaliation where the risk of harm, threats, or legal consequences is elevated. While recent national surveillance data from the Youth Risk Behavior Survey (YRBS) indicates that weapon carrying among U.S. high school students on school property remained stable at about 3% between 2021 and 2023, other risk indicators have changed. Notably, the proportion of youth missing school due to safety concerns reached 13%, and reports of being threatened or injured with a weapon rose from 7% to 9% (Centers for Disease Control and Prevention [CDC], 2024; National Center for Education Statistics, 2024), underscoring that weapon carrying remains a critical component of broader patterns of youth safety risk.

Excluding recreational use (e.g., sustenance hunting and sports), past-12-month non-recreational gun carrying among U.S. high school students was reported by 6.8% of males and 1.9% of females, with persistent disparities by race and ethnicity; non-Hispanic Black males reported the highest carrying rate at 10.6%, followed by Hispanic males (7.2%) and non-Hispanic White males (6.1%) (Simon, 2022). Notably, rates of firearm carriage are substantially higher in disinvested and racially segregated communities than in national samples. For example, in the Flint Youth Injury Study, 23% of assault-injured youth ages 14–24 presenting to an urban emergency department reported firearm possession in the prior six months (Carter et al., 2020), and community-recruited samples in comparable settings have documented firearm carriage rates approaching 20–25% (Carter et al., 2020; Sokol et al., 2022). These estimates underscore the importance of contextualizing firearm carriage within the broader discussion of weapon carrying, particularly in structurally disadvantaged urban environments. Beyond chronic violence exposure, firearm carrying is highly prevalent among youth reporting firearm victimization, suicidal ideation or attempts, and substance use (Simon, 2022). These behaviors are part of a larger cluster of risks, with gun carrying showing strong associations with physical fighting, bullying, substance use, and dating violence; yielding adjusted prevalence ratios between 1.5 and 10.1 across these domains (Simon, 2022). Overall, the structural and behavioral determinants documented in firearm-focused studies extend to non-firearm weapons as well, mainly within urban contexts where youth may carry multiple weapon types, and do so situationally (Harms & Bush, 2022; Oliphant et al., 2019; Vaughn et al., 2012).

Weapon-carrying patterns are responsive to perceived risk. While the national firearm homicide rate for Black Americans is approximately seven times higher than for White Americans across all age groups (e.g., 26.6 vs. 3.9 per 100,000), these consequences are most profound among youth; for young Black males ages 15–24, the disparity in lethal victimization exceeds 24 times that of their White peers (Violence Policy Center, 2025). Firearms were implicated in 86% of homicides involving Black Americans, compared to 70.1% of homicides involving White Americans (Violence Policy Center, 2025). This disparity underscores the racially patterned lethality of firearm injuries and violence in the United States, where conflicts in disinvested urban environments are significantly more likely to involve high-lethality firearms (Violence Policy Center, 2025).

Despite the scale of these disparities, relatively few studies have examined the antecedents of WCB among young Black males (YBM) with sufficient contextual depth, in part due to methodological challenges (Baiden et al., 2024). Weapon carrying is a statistically low-prevalence behavior in general youth population datasets, resulting in sparse data, unstable model estimates, and limited capacity for subgroup analyses. Measurement approaches often rely on broad categorizations or high-risk convenience samples, which can mask heterogeneity tied to perceived threat, trauma exposure, peer networks, and community disadvantage (Baiden et al., 2024). In addition, self-report data remains prone to underreporting in ways that may complicate interpretation. Adolescents may conceal weapons because of fear of legal consequences, mistrust of institutions, or concerns about anticipated injustice. Conversely, some youth may exaggerate or valorize weapon possession to signal protection, power, or social status.

Weapon carrying among youth often reflects overlapping motivations, ranging from perceived self-protection and retaliatory intent to reputation management and online identity signaling (e.g., weapon display on social media) (Carter et al., 2020; Patton et al., 2017, 2019; Sokol et al., 2022). Although firearms represent a high-lethality subset of this behavior, youth frequently carry knives or other implements in response to situational vulnerability or barriers to firearm access.

Access pathways are equally critical to this context. While federal law prohibits licensed dealers from selling handguns to individuals under 21, no uniform federal ban addresses all forms of juvenile possession. Consequently, state laws vary widely, leaving many adolescents to access firearms through household sources, private transfers, stolen weapons, or illicit markets (Webster et al., 2014). Against this backdrop, it remains crucial to understand the social and contextual dynamics that shape why and how young people carry weapons (Patton et al., 2013, 2017).

1.1. Current Study

This cross-sectional analysis utilizes baseline data from the BrotherlyACT Study, a pilot feasibility trial evaluating a culturally adapted, trauma-informed web- and app-based intervention designed to reduce firearm violence risk and provide pre-crisis mental health resources for urban young Black males (n = 301 recruited; ages 15–24). We focus on past-30-day weapon carrying, capturing a broad repertoire of behaviors rather than limiting analysis to firearms alone. The current analysis is based on n = 226 participants who completed baseline screening with non-missing values on the primary outcome.

We address three specific objectives: (1) To estimate the prevalence and frequency of 30-day WCB and identify sociodemographic and socioeconomic correlates; (2) assess the co-occurrence of WCB with violence exposure, substance use, and aggression; and (3) examine multivariable predictors of WCB frequency via negative binomial regression. Throughout, we interpret these behaviors through the Phenomenological Variant of Ecological Systems Theory (PVEST) (Spencer et al., 1997), which centers youth meaning-making, cumulative risk, and contextually shaped coping within intersecting structural systems.

1.2. Literature Review

Weapon carrying among under-resourced youth is attributed to a broader ecology of trauma, structural exclusion, and social disconnection beyond individual factors (Culyba et al., 2016). Recent sociodemographic differences warrant further consideration. National epidemiological data show significantly higher rates of firearm carrying among male (6.8%) than female (1.9%) students in grades 9–12 (Simon, 2022). When disaggregated by ethnicity, non-Hispanic Black males reported the highest carriage rates (10.6%), followed by Hispanic males (7.2%) and non-Hispanic White males (6.1%) (Simon, 2022). Among female students, Hispanic girls reported the highest carriage rate (3.5%), followed by Black girls (2.0%) and White girls (1.1%) (Simon, 2022). Longitudinal trend data indicate that handgun carriage among adolescent girls more than doubled since 2002, with the most pronounced increases occurring among non-Hispanic White and Hispanic subgroups (Vaughn et al., 2019). Similar disparities in general weapon-carrying frequency have been documented among minority male youth, particularly in school and community settings characterized by elevated violence exposure and structural disadvantage (Jewett et al., 2021; McCuddy et al., 2024; Oliphant et al., 2019).

Early initiation of weapon- or firearm-carrying- behaviors is associated with elevated risk of subsequent firearm violence perpetration and victimization (Lanfear et al., 2024; Teplin et al., 2021). Prospective cohort studies have demonstrated this trajectory. A 16-year study of 1829 detained youth (ages 10–18) found that nearly all types of firearm involvement during adolescence, including access, use, and victimization, were associated with higher odds of adult firearm perpetration and ownership (Teplin et al., 2021). Complementing this, a within-individual longitudinal study of 1216 adolescent males found that exposure to gun violence, peer gun carrying, and engagement in antisocial or illegal behaviors predicted fluctuations in gun carrying over five years, indicating that carrying reflects dynamic, context-sensitive processes rather than a fixed trait (Beardslee et al., 2021). Although much of the existing epidemiological literature focuses specifically on firearm carriage, the risk factors identified, including violence exposure, victimization, structural disadvantage, and substance use, are likely to extend to general weapon carriage more broadly, given the documented co-occurrence of firearm and non-firearm weapon carrying in high-risk urban contexts (Harms & Bush, 2022; Vaughn et al., 2012).

1.2.1. Emerging and Contemporary Weapon-Carrying Behaviors

Prior studies have contextualized proximal individual-level and sociocontextual predictors of gun violence among young boys and men, including witnessing non-gun violence, fatalism (the belief that little can change), street-code frameworks, pro-gun beliefs (McCuddy et al., 2024), negative future orientation (Emezue, 2026), and perceived personal rewards for crime (Rowan et al., 2019). Emerging studies situate WCB within digital and performative contexts, where weapons are displayed on social media to signal identity, masculinity, or power (Lane & Stuart, 2022; Patton et al., 2017). Contemporary forms include posting or flashing weapons online (i.e., cyberbanging), gun sharing or borrowing, reckless discharge (e.g., firing into the air), and acquisition through informal markets (Brunson et al., 2022; Kachel et al., 2024; Patton et al., 2017). Other high-risk behaviors include brandishing firearms online (e.g., firearm flossing), modifying handguns with illegal automatic switches, carrying unserialized ghost guns or 3D-printed firearm components, carrying while intoxicated, and using knives or blunt objects for intimidation (Harms & Bush, 2022; Patton et al., 2019).

Contemporary weapon-carrying conversations also extend to mass shootings and school shootings, including how and where attackers source their weapons. Evidence indicates that over 76% of attackers obtain weapons from parents or close relatives (U.S. Secret Service National Threat Assessment Center [NTAC], 2019). These events serve as salient environmental exposures that shape adolescents’ perceptions of safety, heighten anxiety, and influence behavioral coping responses, including the decision to carry a weapon for self-protection (Ruggles & Rajan, 2014; Sokol et al., 2022).

1.2.2. Youth Motivations for Weapon-Carrying

Weapon-carrying motivations are multifaceted and shaped by a complex interplay of identity, victimization, and environmental context (Mattson et al., 2020). These motivations vary across racial and geographic contexts. School-based surveillance indicates that while non-Hispanic White boys in suburban or rural settings may report higher rates of overall weapon carrying (primarily knives or recreational tools), firearm-specific carriage remains more strongly associated with violence exposure and perceived threats (Jewett et al., 2021; Mukherjee et al., 2022), as well as recreational use such as sport shooting or hunting (Oliphant et al., 2019). In some contexts, attitudes supportive of gun violence are linked to rigid gender norms and grievance-oriented worldviews, including racial resentment, particularly among men (Everytown for Gun Safety Support Fund et al., 2023).

For Black youth, however, weapon-carrying behaviors emerge as a proximal coping response to the net vulnerability created by structural disinvestment and chronic community violence (Carey & Coley, 2022). In these settings, limited institutional trust and constrained access to safety resources shape decision-making, positioning weapon possession as a calculated strategy for self-protection. Firearm mortality among urban adolescents is highly spatially concentrated and closely tied to structural disadvantage, underscoring that these behaviors are not simply expressions of individual delinquency or recreational choice (Chadha et al., 2020).

Emerging research further highlights the role of digital environments in shaping motivations for weapon carrying. Youth increasingly engage in forms of “reputation management” that span online and offline contexts, where displays of weapons may signal status, deterrence, or affiliation (Lane & Stuart, 2022; Levine, 2025; McCuddy et al., 2024; Patton et al., 2013, 2019). Across these interconnected spaces, motivations often reflect a blend of protection, retaliation, and identity performance (Patton et al., 2013; Wilkinson, 2021). Studying urban Black male adolescents within high-violence environments provides a critical opportunity to understand weapon carrying as a situational and contextually grounded safety strategy, rather than a uniform behavior across populations.

1.3. Conceptual Framework

This study is grounded in the Phenomenological Variant of Ecological Systems Theory (PVEST) (M. Cunningham et al., 2023; R. M. Cunningham et al., 2015; Spencer et al., 1997; Velez & Spencer, 2018), which emphasizes how the lived experiences of children and youth and their meaning-making processes interact with structural and contextual forces to shape development and behavior. PVEST is well-suited for examining weapon-carrying behaviors among YBM, as it centers on how youth identity formation, reactive coping, risks, and net vulnerability unfold within intersecting systems of oppression and opportunity, factors often beyond the youth’s immediate control. At its core, PVEST recognizes that young people continuously appraise and respond to environmental stressors through processes that can foster either adaptive or maladaptive coping strategies. Within this framework, weapon carrying is interpreted as a contextually shaped coping response to cumulative threats and perceived risks in unsafe environments (Emezue et al., 2025b).

By applying PVEST, this study moves beyond deficit-based perspectives, individual pathology models, and ‘general strain’ frameworks, which have been critiqued for overemphasizing individual-level risk at the expense of structural violence (Levine, 2025). Furthermore, it rejects behavioral choice models that reduce complex systemic exposures to narrow, individual risk-benefit calculations (Li et al., 2021; Lee et al., 2022; McCuddy et al., 2024). Instead, we foreground the embeddedness of WCB within intersecting systems of structural racism, social disadvantage, and chronic community-level violence. This perspective is further contextualized by Krieger’s Ecosocial Theory (Krieger, 2011, 2012), which emphasizes how the embodiment of structural conditions (e.g., systemic violence, disinvestment, marginalization) influences health and behavioral risk over the life course. It also cannot be separated from intersectionality-informed approaches that examine how overlapping identities (e.g., race, gender, access, opportunities, age, class) confer differential exposure to harm and access to protective resources (Crenshaw, 1990; M. Cunningham et al., 2023; McCrea et al., 2019; Velez & Spencer, 2018). We draw on PVEST as an interpretive lens for situating weapon-carrying decisions within structurally constrained environments. The present analysis does not directly operationalize PVEST constructs such as net vulnerability, reactive coping, or emergent identity, and we therefore use the framework to scaffold interpretation rather than to test its propositions. Subsequent waves of BrotherlyACT data will permit more direct operationalization of these constructs through measured indicators of perceived threat, coping appraisal, and identity-relevant outcomes. Jointly, these frameworks shift the analytical lens from risky individuals to risky contexts, providing a more precise foundation for identifying mechanism-specific intervention targets.

2. Materials and Methods

2.1. Study Design and Data Source

This cross-sectional study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (von Elm et al., 2007). The parent study tested a culturally adapted, trauma-informed mobile app to reduce firearm violence risk and early substance use among urban young Black males ages 15–24 (Emezue et al., 2023, 2024, 2025a). Recruitment occurred from July 2022 to July 2023 through collaborations with local educational institutions, youth-oriented organizations, and community outreach initiatives, guided by the Strategies to Recruit, Engage, and Retain Children in Behavioral Health Risk Factor Studies (REACH) model (Schoeppe et al., 2013). Interested youth completed a sociodemographic form and eligibility screening via REDCap (Harris et al., 2009, 2019). To minimize social desirability bias, surveys were self-administered on participants’ own devices or provided tablets in private settings, with research assistant support available remotely for technical issues or distress referrals. Enrolment screening items were adapted from the CDC’s Youth Risk Behavior Surveillance System (YRBS), a nationally validated survey monitoring self-reported health risk behaviors, including weapon carrying, among U.S. adolescents. Items were assessed across varying recall windows. Accordingly, analyses evaluated co-occurring associations within a cross-sectional snapshot rather than temporal ordering.

2.2. Ethical Considerations

The study received IRB approval with a waiver of parental consent, allowing youth aged 15–17 to provide informed assent and those aged 18–24 to provide informed consent. This waiver was critical given the sensitive nature of weapon-carrying disclosures and the practical challenges of obtaining parental consent for out-of-school or system-involved youth (Vaughn et al., 2012). The assent and consent process emphasized voluntary participation, youth-friendly language and confidentiality safeguards. Only minimal contact information (phone, email, and optional social media handle) was collected for study communication and gift card disbursement. Mandatory reporting obligations were disclosed, and immediate referral resources (e.g., crisis hotlines, trauma-informed counseling) were made available irrespective of risk level; no participant required immediate referral. In the parent study, participants received incentives after completing each of the three phases of that study: the needs assessment (Emezue et al., 2023, 2024, 2025a), (2) usability testing of the app (Emezue et al., 2025a), and (3) the pilot feasibility study (Emezue et al., 2025a). To minimize potential coercion, no incentives were provided for the eligibility screening utilized in the current analysis.

2.3. Eligibility Criteria

Eligible participants met the following criteria: (1) self-identified as male; (2) Black or African American; (3) ages 15–24 years; (4) having experienced or witnessed any form of youth violence in their lifetime (e.g., physical fighting, dating or relationship violence, school violence, bullying, threats with weapons, or gang-related violence); (5) English fluency; and (6) ability to consent or assent to participate.

2.4. Dependent Variable: Weapon Carrying

Weapon carrying was assessed using an adapted item from the CDC Youth Risk Behavior Surveillance System (YRBS), “During the past 30 days, on how many days did you carry a weapon such as a gun, knife, or club?”, capturing general and recent weapon-carrying frequency. Although the YRBS did not distinguish gun-specific from other weapons until 2017, prior research demonstrates substantial overlap between firearm and non-firearm weapon carrying in high-violence contexts, supporting its use as a proxy indicator of elevated weapon-related risk in urban youth samples (Vaughn et al., 2012). The original YRBS item specifies “on school property”; this wording was broadened to include non-school contexts to capture behavior among out-of-school youth. To minimize recall burden and align with the YRBS reporting convention, responses were collected across five grouped frequency categories (0 days, 1 day, 2 to 3 days, 4 to 5 days, and 6 or more days) rather than as an open 0 to 30 count. For descriptive and bivariate analyses, the variable was categorized as dichotomous (0 days vs. ≥1 day) and as grouped frequency (0 days, 1 day, 2–3 days, 4–5 days, 6+ days). For primary regression analyses, the variable was retained in its five-level ordinal form (0, 1, 2 to 3, 4 to 5, and 6+ days). For primary regression analyses, the grouped variable was coded as an ordinal bin index (0, 1, 2, 3, 4) and modeled under negative binomial regression to accommodate the right-skewed distribution and excess zeros. Similar bin index coding was applied to frequency-based predictors including physical fighting, threats or injury with a weapon, and binge drinking.

2.5. Direct Victimization

Other items adapted from the YRBS assessed direct victimization. Participants were asked whether they had been threatened or injured with a weapon in the past 12 months, with responses collected across the YRBS frequency grouping (0, 1, 2 to 3, 4 to 5, 6 to 7, 8 to 9, 10 to 11, and 12 or more times). For regression modeling, this grouping was coded as an ordinal bin index (0 through 7). Items measuring past-12-month physical fighting frequency were collected and coded analogously.

2.6. Witnessing Violence

Witnessed community violence was assessed with a binary item asking the participant “Have you ever seen someone in your neighborhood being attacked, beaten, stabbed, or shot?” (yes/no).

2.7. Dating and Partner Abuse

Additionally, we assessed psychological aggression toward a romantic or dating partner (e.g., threats, derogatory language, hostile tone, weapon threats, threats of harm or throwing objects) and physical aggression towards a dating partner (e.g., slapping, kicking, hitting, punching, or weapon threats). Specifically, the item asked, “In the past 12 months, have you used threatening or hurtful language toward a dating or romantic partner (for example, insulting, yelling, or threatening harm)?” Response options were Yes/No.

2.8. Substance Use

We evaluated three independent substance use items: binge drinking (“During the past 30 days, what is the highest number of alcoholic drinks you had on one occasion?”), nonprescribed drug or medication use (Yes/No to “Have you used any drugs or medications not prescribed for medical reasons?”), and substance-related violence/aggression post-consumption (Yes/No to “Have you been involved in, perpetrated, or committed any violence or aggression after using drugs or alcohol?”).

2.9. Covariates

Covariates included age (continuous) and relationship status (categorized as single, in a relationship, or married/other). Educational attainment and employment status were included as indicators of developmental and role status rather than static measures of socioeconomic status. Specifically, educational level captured current school enrollment stages and engagement with non-school activities. Employment was primarily treated as a contextual indicator, with participants grouped into the following categories (full-time, part-time [<35 h/week], student, and laid off or unemployed). Because identifying as Black/African American was an eligibility criterion, race was not modeled; however, ethnicity was included to account for within-group variation in structural exposures. Detailed study measures (including specific item wording, response options, recall windows, and analytic coding decisions) are provided in Supplementary Materials.

2.10. Statistical Analysis

Primary analyses were conducted using Stata 18.1/SE (StataCorp, College Station, TX, USA). Descriptive statistics summarized categorical variables as proportions and continuous variables as means with standard deviations (SD). Bivariate associations were assessed using Pearson chi-square (χ²) tests for categorical variables and unadjusted negative binomial regression for continuous or ordinal predictors. Because weapon carrying is a low-frequency, overdispersed count outcome, multivariable negative binomial regression was employed to estimate incidence rate ratios (IRRs) with robust standard errors. Ninety-five percent confidence intervals are reported for the fully adjusted (Model 13) specification. This approach provides superior analytic power for skewed count distributions compared to traditional logistic regression. As this analysis draws on convenience-recruited pilot data without a defined sampling frame, weighted estimation was not appropriate, and unweighted estimates are reported throughout. Bollen et al. (2016) note that the decision to apply weights in regression analysis should follow from diagnostic comparison of weighted and unweighted estimates rather than from convention. The absence of design weights in this trauma-informed pilot precludes such a diagnostic and is acknowledged as a limitation of inference from a convenience sample (Bollen et al., 2016).

We estimated 13 hierarchical regression models. First, individual violence, substance use, and sociodemographic predictors were examined in unadjusted analyses (Models 1–2, 4–10) to assess crude associations with weapon-carrying behavior (WCB). Three intermediate models (Models 3, 11, and 12) grouped conceptually related predictors—for example, Model 3 examined the joint effect of weapon-related injury and past-12-month violence exposure. The final fully adjusted model (Model 13) included all sociodemographic and socioeconomic covariates. Statistical significance was set at p ≤ 0.05. To ensure robustness, we also estimated logistic regression models predicting any weapon carrying (0 vs. ≥ 1 day) to determine if findings differed when modeling binary occurrence versus frequency.

2.11. Missing Data

The baseline screening records were restricted to eligible participants (male-identifying youth ages 15–24 years) who provided complete data for the primary outcome of past-30-day weapon-carrying frequency. Regression models employed listwise deletion, resulting in an analytic sample size that varied by specification based on covariate missingness. The fully adjusted model utilized a sample of n = 204; the reduction from the initial n = 226 was driven primarily by missing responses regarding employment and relationship status. To evaluate potential selection bias, we compared participants included in the fully adjusted model with those excluded due to missing covariates and found no meaningful differences in baseline demographics or primary risk measures. Furthermore, a sensitivity analysis restricting all models to the common complete-case sample yielded substantively similar estimates, supporting the robustness of the findings to model-specific missingness.

2.12. Negative Binomial Regression Models

Negative binomial regression was utilized to model the frequency of weapon carrying in the past 30 days. The outcome exhibited significant overdispersion (where the variance exceeds the mean) rendering negative binomial regression more appropriate than a Poisson model, which assumes equidispersion and can produce underestimated standard errors in skewed distributions.

While zero-inflated negative binomial (ZINB) models were considered, they were not selected; the excess zeros in this sample likely reflect a combination of true non-carriers and intermittent, situational risk rather than a distinct structural zero-generating process. To confirm model selection, fit was compared across Poisson and negative binomial specifications using the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and sample-size-adjusted BIC. Negative binomial models consistently demonstrated superior fit in unadjusted and intermediate models.

Primary results are presented as Incidence Rate Ratios (IRRs). An IRR of 1.00 indicates no association, while values <1.00 and >1.00 indicate a decrease or increase in expected carrying days, respectively. For example, an IRR of 1.20 corresponds to a 20% increase in the expected frequency of weapon-carrying days. Because the outcome and several frequency-based predictors (past-12-month threats or injury with a weapon, past-12-month physical fights, past-30-day binge drinking) were measured as grouped frequency categories and coded as ordinal bin indices, reported IRRs reflect the expected multiplicative change in weapon-carrying category per single-step increase in the predictor category. This per-category interpretation is a property of the YRBS-aligned measurement instrument and applies uniformly across these items. Binary and categorical predictors retain their conventional interpretation. Robustness checks using a common analytic sample (n = 204) confirmed that findings remained substantively unchanged across model variations. Because both the outcome and key frequency predictors were analyzed using categorical bins rather than raw counts, the reported IRRs do not reflect changes per individual event. Instead, they indicate the expected proportional shift from one weapon-carrying category to the next for each single-step increase in the predictor’s category.

3. Results

3.1. Participant Characteristics

The final analytic sample included n = 226 young Black males (YBM), ages 15–24, who provided complete data for the primary weapon-carrying outcome. Participants had a mean age of 18.32 years (SD = 3.10), with ages ranging from 15 to 24 and 52.7% falling within the 15 to 17-year age range. While the sample predominantly identified as African American, 6.6% also identified as Hispanic or Latino. Regarding educational and role status, 78.2% held a high school diploma or less, while 21.8% had some college experience. Employment data revealed that 53.1% were students, 22.3% worked part-time, 16.5% were unemployed or unoccupied, and 8.0% held full-time positions. Most of the sample (65.5%) identified as single, while 33.6% were currently partnered.

Weapon-carrying behaviors (WCB) were prevalent, with 42.5% of participants reporting carrying a weapon at least once in the past 30 days. This frequency was distributed among those who carried for one day (8.8%), 2–3 days (17.3%), 4–5 days (7.1%), and 6 or more days (9.3%). These rates occurred against a backdrop of high violence exposure; approximately 64.8% reported direct victimization in the past year (e.g., threatened, injured, or assaulted), and 76.4% reported witnessing community violence. Additionally, 19.1% reported threatening someone with a weapon at least 2–3 times in the past year, indicating elevated weapon-related aggression beyond the 30-day window. Because eligibility required prior experience or witnessing of youth violence, baseline prevalence figures reflect this enriched-risk recruitment frame and are not intended to estimate population-level prevalence among urban young Black males.

Furthermore, one-third reported general partner violence (30.4%), 36.7% reported psychological aggression and 28.3% reported physical aggression toward a romantic partner within the past year (categories not mutually exclusive).

Substance use and its intersection with conflict were also notable within the sample. About 58.7% reported no past-30-day alcohol use, 10.2% consumed 1–2 drinks on a single occasion, and 20.8% met binge drinking criteria (≥5 drinks on one occasion for males; NIAAA/CDC). Nonprescribed drug or medication use was reported by 43.6% of participants, and 37.2% reported engaging in violent or aggressive behavior after alcohol or drug use.

3.2. Bivariate Analysis

Bivariate associations between sociodemographic and socioeconomic variables and 30-day weapon carrying are presented in

Table 1. Sociodemographic and Socioeconomic Characteristics of Young Black Males by Past-30-Day Weapon-Carrying Frequency (n = 226).

	In the Past 30 Days, How Many Days Did You Carry a Weapon?
	0 Day	1 Day	2–3 Days	4–5 Days	6+ Days	Total	χ2
Race/Ethnicity
Non-Hispanic Black	120 (92.31)	19 (95.00)	36 (92.31)	14 (87.50)	20 (95.24)	209 (92.48)	2.68
Hispanic Black	8 (6.15)	1 (5.00)	3 (7.69)	2 (12.50)	1 (4.76)	15 (6.64)
Non-Hispanic Multiracial	2 (1.54)	0 (0.00)	0 (0.00)	0 (0.00)	0 (0.00)	2 (1.54)
Age
15–17 years	94 (72.87)	5 (25.00)	11 (28.21)	0 (0.00)	9 (45.00)	119 (53.12)	82.56 ***
18 years	11 (8.52)	1 (5.00)	1 (2.56)	0 (0.00)	1 (5.00)	14 (6.25)
19–21 years	8 (6.20)	1 (5.00)	8 (20.51)	6 (37.50)	6 (30.00)	29 (12.95)
22–24 years	16 (12.40)	13 (65.00)	19 (48.72)	10 (62.50)	4 (20.00)	62 (27.68)
Education
HS Diploma or Less	105 (88.98)	12 (60.00)	25 (64.10)	9 (56.25)	14 (77.78)	165 (78.20)	21.00 ***
Any College	13 (11.02)	8 (40.00)	14 (35.90)	7 (43.74)	4 (22.22)	46 (21.80)
Relationship Status
[A] Single	91 (70.00)	13 (65.00)	23 (58.97)	7 (43.75)	14 (66.67)	148 (65.49)	10.23
[B] Partnered (Dating, Married, Hooking Up, Common-Law)	38 (29.23)	6 (30.00)	16 (41.03)	9 (56.25)	7 (33.33)	76 (33.63)
[C] No Longer Partnered (Divorced, Separated, Widowed)	1 (0.77)	1 (5.00)	0 (0.00)	0 (0.00)	0 (0.00)	2 (0.88)
Employment Status
Full-Time (35+ h/week)	5 (3.88)	1 (5.00)	6 (15.79)	2 (12.50)	4 (19.05)	18 (8.04)	41.28 ***
Part-Time (less than 35 h/week)	17 (13.18)	11 (55.00)	13 (34.21)	6 (37.50)	3 (14.29)	50 (22.32)
Student	84 (65.12)	7 (35.00)	15 (39.47)	4 (25.00)	9 (42.86)	119 (53.12)
Laid Off, Unemployed, Disabled, Homemaker, Other	23 (17.83)	1 (5.00)	4 (10.53)	4 (25.00)	5 (23.81)	37 (16.52)

Note: *** p ≤ 0.001. Outcome definition and binning: Weapon carrying was assessed with the YRBS item “During the past 30 days, on how many days did you carry a weapon such as a gun, knife, or club?” Responses (0–30 days) were grouped for descriptive and bivariate analyses as 0 days, 1 day, 2–3 days, 4–5 days, and 6+ days. Cell entries: Values are n (%) within weapon-carrying category. χ² tests compare distributions across weapon-carrying categories. Coding for covariates (as displayed): Age group: 15–17, 18, 19–21, 22–24 years (categorical, mutually exclusive). Race/ethnicity: Non-Hispanic Black, Hispanic Black, Non-Hispanic Multiracial (categorical). Education: HS diploma or less vs any college (binary). Relationship status: Single, Partnered, No longer partnered (categorical). Employment status: Full-time (35+ hours/week), Part-time (<35 h/week), Student, Other (categorical).

. Among those who reported carrying a weapon in the previous 30 days, frequencies varied from one day to six or more days. Significant associations were observed for age, educational attainment, and employment status, while differences by relationship status and ethnicity were not statistically significant (χ² = 2.68, p > 0.05), likely reflecting sample homogeneity (99.1% Black-identified across non-Hispanic and Hispanic categories combined).

Weapon carrying varied substantially by age group (χ² = 82.56, p ≤ 0.001). Among participants aged 15–17 years, 21.0% reported any past-30-day weapon carrying. This proportion increased sharply among older participants, with 72.4% of those aged 19–21 years and 74.2% of those aged 22–24 years reporting past-30-day carrying. Frequent carrying (4 or more days) was similarly concentrated among older cohorts. No adolescent aged 15–17 reported carrying on 4–5 days, compared to 20.7% of those aged 19–21 and 16.1% of those aged 22–24. In unadjusted negative binomial regression, each additional year of age was associated with a 22% increase in weapon-carrying frequency (IRR = 1.220, p ≤ 0.05), though this association attenuated to non-significance in the fully adjusted model (IRR = 1.041), suggesting that age-related risk is not an independent driver but likely operates through correlated environmental exposures such as direct victimization and violence exposure.

Educational attainment was significantly associated with weapon-carrying frequency (χ² = 21.00, p < 0.001). Participants with a high school diploma or less were overrepresented among those reporting more frequent carrying, while those with any college education were more likely to report no carrying in the past 30 days.

Employment status also showed a strong association with past-30-day weapon carrying (χ² = 41.28, p ≤ 0.001). Contrary to typical expectations, full-time workers reported the highest prevalence of any carrying (72.2%), followed by part-time workers (66.0%), laid off or unemployed participants (37.8%), and students (29.4%). Frequent carrying of 4 or more days was also concentrated among full-time and part-time workers. In unadjusted negative binomial regression (Model 12), part-time workers, students, and laid off or unemployed participants all carried significantly less frequently than full-time workers (IRRs = 0.490, 0.486, and 0.502, respectively; all p ≤ 0.05). These associations attenuated to non-significance in the fully adjusted model (Model 13), suggesting that employment-related differences in carrying frequency operate through correlated exposures rather than employment status alone.

Relationship status was not significantly associated with weapon-carrying frequency (χ² = 10.23, p > 0.05). Distributions across carrying categories were similar between single and partnered youth, suggesting that relationship status alone did not differentiate risk patterns in this sample.

Bivariate associations with substance use, violence exposure, and interpersonal aggression are summarized in

Table 2. Frequency Table for Other Covariates.

	In the Past 30 Days, How Many Times Did You Carry a Weapon?
	0 Day	1 Day	2–3 Days	4–5 Days	6+ Days	Total	χ2
Number of Drinks You’ve Had in a Row in Last 30 days
Did not drink alcohol in last 30 days	106 (82.17)	3 (15.00)	12 (30.77)	1 (6.25)	10 (47.62)	132 (58.67)	132.20 ***
1–2 drinks	10 (7.75)	6 (30.00)	4 (10.26)	1 (6.25)	2 (9.52)	23 (10.22)
3 drinks	5 (3.88)	3 (15.00)	2 (5.13)	0 (0.00)	3 (14.29)	13 (5.78)
4 drinks	3 (2.33)	2 (10.00)	3 (7.69)	2 (12.50)	0 (0.00)	10 (4.44)
5 drinks	4 (3.10)	1 (5.00)	4 (10.26)	3 (18.75)	0 (0.00)	12 (5.22)
6–7 drinks	1 (0.78)	2 (10.00)	6 (15.38)	4 (25.00)	1 (4.76)	14 (6.22)
8–9 drinks	0 (0.00)	0 (0.00)	4 (10.26)	4 (25.00)	1 (4.76)	9 (4.00)
10+ drinks	0 (0.00)	3 (15.00)	4 (10.26)	1 (6.25)	4 (19.05)	12 (5.33)
Drug Use
No	100 (76.92)	6 (30.00)	10 (26.32)	1 (6.25)	10 (47.62)	127 (56.44)	58.96 ***
Yes	30 (23.08)	14 (70.00)	28 (73.68)	15 (93.75)	11 (52.38)	98 (43.56)
Ever involved with or committed violence after drinking or using drugs?
No	111 (85.38)	6 (30.00)	14 (35.90)	2 (12.50)	9 (42.86)	142 (62.83)	70.60 ***
Yes	19 (14.62)	14 (70.00)	25 (64.10)	14 (87.50)	12 (57.14)	84 (37.17)
Psychologically aggressive/violent to romantic partner in the past year
No	117 (90.00)	6 (30.00)	10 (25.64)	2 (12.50)	8 (38.10)	143 (63.27)	96.74 ***
Yes	13(10.00)	14 (70.00)	29 (74.36)	14 (87.50)	13 (61.90)	83 (36.73)
Physically aggressive/violent to romantic partner in past year
No	122 (93.85)	9 (45.00)	16 (41.03)	2 (12.50)	13 (61.90)	162 (71.68)	85.13 ***
Yes	8 (6.15)	11 (55.00)	23 (58.97)	14 (87.50)	8 (38.10)	64 (28.32)
Impacted by any violence or aggression in past 12 months
No	76 (58.46)	0 (0.00)	2 (5.13)	0 (0.00)	1 (4.76)	79 (35.11)	73.93 ***
Yes	54 (41.54)	19 (100.00)	37 (94.87)	16 (100.00)	20 (95.24)	146 (64.89)
Ever seen someone physically attacked?
No	48 (36.92)	1 (5.00)	2 (5.13)	1 (6.25)	2 (9.52)	54 (23.89)	28.74 ***
Yes	82 (63.08)	19 (95.00)	37 (94.87)	15 (93.75)	19 (90.48)	172 (76.11)

Note: *** p ≤ 0.001. Outcome definition and binning: Weapon carrying was measured over the past 30 days and grouped as 0 days, 1 day, 2–3 days, 4–5 days, and 6+ days. Values are n (%) within weapon-carrying category. χ² tests compare distributions across weapon-carrying categories. Predictor coding and recall windows (as displayed): Binge drinking intensity: “During the past 30 days, what is the highest number of alcoholic drinks you had on one occasion?” categorized as did not drink, 1–2, 3, 4, 5, 6–7, 8–9, 10+ drinks (past 30 days). Illicit drug use: Yes/No to “Have you used any drugs or medications not prescribed for medical reasons?” (item recall as administered, reported here as binary). Substance-related violence/aggression: Yes/No to “Have you been involved in, perpetrated, or committed any violence or aggression after using drugs or alcohol?” (item recall as administered, reported here as binary). Partner psychological aggression and partner physical aggression: Yes/No items referencing the past 12 months (as labeled in the table).

. Strong, significant associations emerged across all domains (all χ² > 28, p < 0.001). Higher-frequency weapon carrying clustered among youth reporting alcohol use (especially binge or heavy episodic drinking), nonprescribed drug use, substance-related aggression, and both psychological and physical dating aggression. Furthermore, carrying was strongly linked to direct victimization and witnessing community violence; conversely, participants reporting no exposure or substance use predominated among non-carriers.

Collectively, these bivariate patterns indicate that weapon-carrying frequency clusters along developmental (e.g., age), structural (e.g., education, employment), and behavioral lines. These findings underscore the necessity of situating weapon carrying within broader socioeconomic and life-course “risk constellations” rather than treating it as an isolated behavioral trait.

3.3. Negative Binomial Regression Models of Weapon Carrying Frequency

Negative binomial regression models examined associations between violence exposure, behavioral factors, and sociodemographic characteristics and 30-day weapon-carrying frequency (

Table 3. Negative binomial regression models of past-30-day weapon-carrying category (IRR) among eligible male-identifying youth ages 15–24 (Models 1–13).

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7	Model 8	Model 9	Model 10	Model 11	Model 12	Model 13
Number of times threatened or injured with a weapon, past 12 mos.	1.550 **		1.299 **										1.146 *
	(0.084)		(0.065)										(0.063)
Impacted by violence or aggression, past 12 mos. (yes/no)		14.948 **	8.801 **										7.821 **
		(5.573)	(3.346)										(4.168)
Seen someone physically attacked or beaten in the community (yes/no)				4.199 **									1.721
				(1.288)									(0.548)
number of times in a physical fight, past 12 mos.					1.378 **								1.105 *
					(0.059)								(0.051)
Largest number of drinks consumed in a row, last 30 days						1.270 **							0.952
						(0.047)							(0.049)
Used prescription medication that was not prescribed to you (yes/no)							2.986 **						1.140
							(0.584)						(0.249)
Involved with violence/aggression after drinking/drugs (yes/no)								3.425 **					0.996
								(0.631)					(0.212)
Psychologically aggressive/violent toward partner in past year (yes/no)									4.469 **		3.739 **		1.572
									(0.774)		(0.915)		(0.370)
Physically aggressive/violent toward partner in past year (yes/no)										3.349 **	1.276		1.060
										(0.621)	(0.302)		(0.247)
Age in years (Min = 15, Max: 24)												1.220 **	1.041
												(0.046)	(0.039)
Hispanic Ethnicity (yes/no)												0.616	0.675
												(0.225)	(0.194)
Education (ref: High School Diploma or less)												0.847	0.986
												(0.195)	(0.179)
Marital Status: Relationship Status (ref: Single)
Partnered (Dating, Married, Common-Law)												1.427	1.075
												(0.269)	(0.186)
No Longer Partnered (Divorced, Separated, Widowed)												0.272	0.350
												(0.347)	(0.365)
Employment Status (Ref: Full-time, 35+ hours/week)
Part-Time (Less than 35 h/week)												0.490 *	0.744
												(0.160)	(0.186)
Student												0.486 *	0.659
												(0.164)	(0.178)
Laid off/Unemployed/Disabled/Homemaker/Other												0.502 *	1.025
												(0.173)	(0.274)
Intercept	0.406 **	0.101 **	0.090 **	0.296 **	0.477 **	0.478 **	0.543 **	0.535 **	0.448 **	0.611 **	0.445 **	0.036 **	0.027 **
	(0.060)	(0.037)	(0.032)	(0.085)	(0.065)	(0.071)	(0.081)	(0.027)	(0.061)	(0.074)	(0.061)	(0.029)	(0.024)
Number of observations	226	225	225	226	226	225	225	226	226	226	226	207	204
AIC	572.86	547.44	523.04	611.64	585.21	596.69	601.93	596.10	573.77	598.53	574.72	551.85	463.70

Notes: ** p ≤ 0.001; * p ≤ 0.05; standard errors in parentheses. Table notes: Outcome was weapon-carrying days in the past 30 days (0–30). Predictors were coded as follows: direct victimization (past 12 months), witness of neighborhood violence (ever, Yes/No), partner psychological aggression and partner physical aggression (past 12 months, Yes/No), binge drinking intensity (maximum drinks on one occasion, past 30 days, categorical), nonmedical drug/medication use (Yes/No), and violence/aggression after alcohol or drugs (Yes/No). Covariates included age (years), Hispanic ethnicity (Yes/No), education (HS diploma or less versus any college), relationship status (single, partnered, no longer partnered), and employment status (full-time, part-time, student, other). Models used listwise deletion; “Number of observations” reports the column-specific analytic N based on variables included in that model.

). Results are presented for unadjusted, intermediate, and fully adjusted (Model 13) specifications. Multicollinearity diagnostics indicated no problematic overlap among predictors, supporting stable coefficient estimates.

In unadjusted models, violence exposure indicators were consistently associated with increased weapon-carrying frequency. Past-year threats or injury with a weapon was associated with higher expected carrying frequency (IRR = 1.55, p < 0.01), direct exposure to violence or aggression showed the largest association (IRR = 14.95, p < 0.001), witnessing a physical attack or beating in the community was also significant (IRR = 4.19, p < 0.01), and each one-category increase in physical fighting frequency was associated with higher carrying frequency (IRR = 1.38, p < 0.001). These associations persisted, though attenuated in magnitude, in intermediate models grouping conceptually related violence indicators.

Weapon carrying varied substantially by age group (χ² = 82.56, p ≤ 0.001). Among participants aged 15–17 years, 21.0% reported any past-30-day weapon carrying. This proportion increased sharply among older participants, with 72.4% of those aged 19–21 years and 74.2% of those aged 22–24 years reporting past-30-day carrying. Frequent carrying (4 or more days) was concentrated among older participants. No adolescent aged 15–17 reported carrying on 4–5 days, compared to 20.7% of those aged 19–21 and 16.1% of those aged 22–24. In unadjusted negative binomial regression, each additional year of age was associated with a 22% increase in weapon-carrying frequency (IRR = 1.220, p ≤ 0.05), though this association attenuated to non-significance in the fully adjusted model (IRR = 1.041), suggesting that age-related risk operates through correlated exposures such as direct victimization and violence exposure rather than as an independent predictor.

Substance use variables were significant in unadjusted analyses. Each one-category increase in past-30-day binge drinking intensity (IRR = 1.27, p < 0.01), past-year nonprescribed drug or medication use (yes vs. no, IRR = 2.99, p < 0.01), and substance-related aggression (yes vs. no, IRR = 3.43, p < 0.01) were each associated with increased weapon-carrying frequency. These associations were substantially attenuated and lost statistical significance in the fully adjusted model, suggesting shared variance with broader environmental violence exposure. Similarly, psychological aggression toward a dating partner was associated with more than four-fold higher weapon-carrying frequency (IRR = 4.47, p < 0.001), as was physical aggression toward a partner (IRR = 3.35, p < 0.001). Both associations weakened and lost statistical significance in the fully adjusted model, consistent with shared variance with broader violence exposure.

In intermediate models including sociodemographic characteristics (Model 12), each additional year of age was associated with a higher expected rate of weapon-carrying days (IRR = 1.22), though this association was reduced after adjustment for co-occurring exposures. Compared to full-time employed youth, those working part-time (IRR = 0.49), students (IRR = 0.49), and those who were unemployed, laid off, or otherwise not working (IRR = 0.50) all showed lower expected weapon-carrying rates, though these associations did not persist in the fully adjusted model. Overall, direct exposure to violence remained the most consistent correlate of weapon-carrying frequency, whereas substance use and interpersonal aggression appeared to operate primarily through shared variance with broader violence exposure.

3.4. Fully Adjusted Model

In the fully adjusted model (Model 13, n = 204), three violence-related predictors remained independently associated with weapon-carrying frequency. Past-year exposure to violence or aggression (yes vs. no) was the strongest correlate (IRR = 7.82, 95% CI [2.75, 22.23], p < 0.001), followed by each one-category increase in past-year threats or injury with a weapon (IRR = 1.15, 95% CI [1.03, 1.28], p < 0.05) and each one-category increase in past-year physical fighting frequency (IRR = 1.11, 95% CI [1.01, 1.21], p < 0.05). Each of these effects persisted after adjustment for substance use and sociodemographic characteristics. Witnessing community violence and all substance use indicators were no longer significant in the final model. While age was positively associated with weapon-carrying frequency in intermediate models (IRR = 1.22, p < 0.01), it attenuated to non-significance in the fully adjusted model. Similarly, Hispanic ethnicity, educational attainment, relationship status, and employment status did not independently predict carrying frequency once the model accounted for direct violence exposure. The wide confidence interval for past-year violence exposure reflects modest cell sizes in the unexposed reference group and is interpreted as evidence of direction and rank rather than as a calibrated magnitude.

3.5. Model Robustness

Weapon-carrying frequency was modeled as an ordinal bin index (0 through 4, corresponding to the five YRBS frequency categories), with zero values retained and modeled directly. Negative binomial models were retained as the primary approach to accommodate a right-skewed outcome with substantial zeros; model fit was summarized using AIC (Table 3). Zero-inflated negative binomial models were considered but not selected because zeros plausibly reflect a mix of true non-carrying and intermittent risk rather than a distinct structural-zero process. Given the cross-sectional design and mismatched recall windows across measures, results are interpreted as conditional associations rather than causal effects.

To assess multicollinearity and model stability in the fully adjusted model (Model 13), we computed variance inflation factors (VIF). VIF values for primary predictors ranged from 1.44 to 3.59, below commonly used thresholds (e.g., 5 or 10), suggesting that correlated constructs did not introduce problematic collinearity. The past-year violence exposure association remained large in magnitude and statistically significant across model specifications (Model 13 IRR = 7.82, p < 0.001), and estimates were substantively similar when restricting analyses to the common complete-case sample (n = 204).

3.6. Binary Logistic Models of Any Weapon Carrying

As a robustness check, we estimated logistic regression models predicting the presence of any weapon carrying (≥1 day in the past 30 days) to examine whether associations observed in count models reflected general risk of carrying rather than only frequency among carriers. The binary outcome was coded 0 = no carrying and 1 = any carrying.

In unadjusted models, multiple indicators of violence exposure were strongly associated with weapon carrying. Youth reporting prior exposure to violence had higher odds of carrying a weapon (OR = 43.16, p < 0.001) compared with those without recent violence exposure. Being injured or threatened with a weapon in the past year also significantly increased the likelihood of carrying (OR = 2.817, p < 0.001). Involvement in physical fights and witnessing serious community violence were each associated with higher probability of carrying (OR = 2.307 and OR = 8.780, p < 0.001 respectively). Interpersonal violence behaviors showed similarly large effects, with youth reporting psychological (OR = 24.231, p < 0.001) or physical aggression (OR = 21.350, p < 0.001) toward a dating partner having markedly higher odds of weapon carrying.

In the fully adjusted logistic model, several associations remained statistically significant. Youth who reported being exposed to violence or aggression in the past 12 months had higher odds of weapon carrying (adjusted OR = 7.88, p = 0.003). Being injured or threatened with a weapon in the past year was also associated with increased odds of carrying (adjusted OR = 1.55, p = 0.008). Additionally, past-year psychological aggression toward a dating or romantic partner was associated with higher odds of weapon carrying (adjusted OR = 3.71, p = 0.011). Other indicators, including fighting, witnessing violence, and physical partner aggression, were attenuated and no longer statistically significant after adjustment. The overall model fit was strong (Nagelkerke R² ≈ 0.65), and classification accuracy was 84%.

These binary findings closely paralleled the negative binomial models predicting carrying frequency, suggesting that violence exposure is primarily associated with whether youth carry a weapon at all, whereas frequency of carrying reflects additional behavioral escalation among carriers.

4. Discussion

This study offers a contemporary assessment of weapon-carrying behaviors (WCB) among a high-risk sample of violence-exposed urban young Black males (YBM) recruited outside school settings. Across multivariable negative binomial and logistic models, direct and proximal violence exposure emerged as the most consistent and robust predictor of both WCB frequency and presence, even after controlling for substance use, interpersonal aggression, and sociodemographic factors. Because the outcome captured both firearm and nonfirearm weapons, the findings reflect general weapon-carrying rather than firearm-specific behavior. These patterns align with protective or situational interpretations of youth weapon carrying, wherein carrying functions as a response to perceived threat or prior victimization rather than solely instrumental or premeditated intent. Given the cross-sectional design and differing recall windows across measures, results represent conditional associations rather than causal relationships.

Interpreted through PVEST, weapon carrying emerges as a contextually embedded coping strategy shaped by high net vulnerability, repeated threat appraisals, and structurally constrained environments. Chronic violence exposure recalibrates perceptions of safety, making carrying a functional (though potentially (mal)adaptive) response aimed at protection, deterrence, or identity management. Krieger’s Ecosocial Theory complements this by highlighting how cumulative exposures to structural violence become embodied in behavior, while PVEST foregrounds how youth make meaning of these conditions. Together, these frameworks position weapon carrying not as a stable individual trait but as a dynamic response to intersecting structural and developmental forces.

Direct victimization and broad exposure to violence remained strongly associated with weapon-carrying frequency, each contributing large effect sizes, while physical fighting indicated additional vulnerability linked to recent interpersonal conflict. Witnessing community violence, however, was significant only in unadjusted models, suggesting that proximal and direct exposures carry greater weight in shaping weapon-carrying decisions than vicarious exposure alone.

Substance use and dating aggression showed strong unadjusted associations with WCB but attenuated after adjusting for prior violence exposure, reflecting their co-occurrence within shared high-risk ecologies rather than independent causal pathways. From a PVEST lens, these behaviors may represent parallel coping responses to chronic threat and structural constraint. Although cross-sectional data limit causal interpretation, the attenuation patterns point toward intervention strategies that address ecological risk conditions rather than targeting individual behaviors in isolation.

These results corroborate emerging critiques cautioning against overreliance on individual-level strain or rational choice models to explain firearm violence exposure, without sufficient attention to sociostructural and multilevel determinants such as segregation, racialized and targeted policing, lack of education and livable job prospects, neighborhood disinvestment, and trauma exposure (Levine, 2025). Consistent with prior work documenting elevated gun-carrying risk among Black youth exposed to violence (Kemal et al., 2018; McCuddy et al., 2024; Wallace, 2017), findings support the need for prevention strategies that address these upstream structural drivers rather than downstream behaviors alone. Because weapon carrying may function as a protective response to chronic threat (Patton et al., 2013, 2019; Wilcox et al., 2006), interventions should avoid punitive framing and instead prioritize trauma-informed, culturally grounded approaches that center youth safety, healing, future orientation, and structural change.

The sharp increase in weapon carrying between adolescence and young adulthood (rising from 21.0% among participants aged 15–17 to more than 72% among those aged 19 and older) warrants attention in both research and intervention design. This pattern aligns with the developmental criminology literature documenting that weapon carrying peaks in late adolescence and early adulthood, particularly among males in high-violence urban environments (Beardslee et al., 2018, 2021; Shulman et al., 2021). From a PVEST perspective, accumulated stress exposures during adolescence may intensify perceived threat and identity-based motivations for carrying. Although age effects diminished in fully adjusted models, the concentration of frequent carrying among 19–24-year-olds suggests that interventions targeting mediating factors around this transition are especially important.

The association between full-time employment and elevated weapon carrying is among the most theoretically generative findings of this study. Rather than reflecting economic desperation or idleness, weapon carrying in this sample appear concentrated among those most embedded in the labor force, a pattern consistent with routine activity theory (Cohen & Felson, 1979). Young Black males working full-time in high-violence urban environments may traverse dangerous transit corridors, work late shifts, and carry cash or valuables in ways that heighten perceived vulnerability and motivate protective carrying. This interpretation aligns with qualitative evidence that weapon carrying among working youth is often framed as a rational safety strategy rather than an expression of aggression or criminal intent (Beardslee et al., 2018). From a PVEST lens, employment may paradoxically amplify net stress engagement by increasing exposure to environmental threat without providing sufficient social or institutional protection. These findings challenge deficit-oriented framings that locate weapon-carrying risk primarily in unemployment or school disengagement, and suggest that workplace safety, transit safety, and employer-based violence prevention programs merit serious consideration as ancillary intervention entry points for this population. Educational attainment should be interpreted with caution in adolescent samples. For younger participants, lower educational level primarily reflects developmental stage rather than socioeconomic disadvantage. Observed associations are therefore more likely to capture differences in school engagement and daily structure, rather than underlying socioeconomic status.

Finally, prevention efforts must account for the evolving peer and masculinity dynamics shaping weapon-carrying risk among young Black males. Contemporary masculinity contexts, including algorithmic masculinities (Abbas & McNeil-Willson, 2025), manosphere-influenced masculinities (Nicholas, 2024), hybrid and care-oriented masculinities (Moura, 2024), and the digitally mediated performances of masculinity within peer networks (O’Rourke & Haslop, 2025), interact with structural risk in ways that generalized prevention frameworks often miss in today’s contemporary landscape. Trauma-informed interventions must therefore attend to how digital environments, peer identity dynamics, and evolving norms around manhood shape youth decisions to carry weapons, offering affirmative pathways toward safety, healing, and non-violent coping that resonate with contemporary youth experience.

Limitations and Study Strengths

This study has several limitations. The cross-sectional design permits estimation of associations but not causation or temporal ordering. Generalizability is limited to violence-exposed urban young Black males recruited through a trauma-informed intervention study, and findings should not be extrapolated to other youth populations without caution.

Although the CDC YRBS includes both a 30-day weapon-carrying item (on school property) and a 12-month non-recreational gun-carrying item introduced in 2017, we analyzed an adapted 30-day measure to capture recent behavior and reduce recall bias. The school-property framing was broadened to include non-school contexts, which improves ecological validity for out-of-school youth but limits direct comparability with national YRBS estimates. Past-30-day weapon carrying in this sample (42.5%) reflects the violence-exposure eligibility criterion of the parent intervention trial and the broadened weapon definition. The YRBS-adapted weapon-carrying item, along with the past-year fighting, threat or injury, and binge drinking items, was administered with grouped response categories rather than open counts. Reported IRRs for these predictors are therefore interpreted as expected change per category step, which limits direct comparability with studies that collected open counts.

Comparison to the YRBS school-property estimate of approximately 3% (Centers for Disease Control and Prevention [CDC], 2024) illustrates the depth of risk in this recruitment frame rather than a secular trend, and findings should not be read as a population estimate for young Black males more broadly. Recruitment occurred between July 2022 and July 2023, a period of elevated post-pandemic community violence in Chicago and comparable urban settings. Prevalence and exposure estimates reported here should be read within that base-rate context and are not directly comparable to pre-2020 estimates. Beyond the general limits of cross-sectional inference, the temporal direction of the association between violence exposure and weapon carrying cannot be adjudicated here. Youth who carry are plausibly more likely to enter contexts in which they are threatened, witness retaliation, or become involved in fights, so a portion of the observed association may reflect carrying as a driver of subsequent exposure rather than as a response to prior exposure. Longitudinal designs with measured carrying onset are required to disentangle these pathways.

Two substance use items did not specify a recall window in the screening battery, which limits precision in interpreting those associations and should be addressed in future iterations of the measure. Dating violence items also excluded sexual violence, a known correlate of weapon possession (Ruggles & Rajan, 2014), limiting our ability to capture the intersection of this underreported and severe form of partner violence with weapon-carrying behavior. Of note, past-year psychological partner aggression remained associated with any carrying in the adjusted logistic specification (adjusted OR = 3.71, p = 0.011) but attenuated in the negative binomial specification predicting carrying frequency. This asymmetry suggests interpersonal hostility may be more closely tied to whether a young man carries at all than to how often, consistent with a threshold rather than dose-response pattern, and warrants targeted follow-up in studies with adequate power to disentangle initiation from intensification of carrying.

The study also did not assess long-term or time-varying weapon-carrying patterns, including seasonal and episodic carrying, a recognized gap in the literature (Oliphant et al., 2019). Future research should examine weapon-carrying prevalence during peak violence periods (nights, weekends, summer months) and clarify the motivations driving carrying across those contexts. Self-reported weapon-carrying is subject to social desirability bias in both directions, but the direction is unlikely to be symmetric in this recruitment frame. Youth recruited through a violence-prevention pilot, in which carrying may be normalized within peer networks, may be more likely to over-report than under-report, while youth concerned about legal consequences or institutional mistrust may move in the opposite direction (Dijkstra et al., 2010). The net direction of bias cannot be determined without external validation, and future studies should incorporate objective or multi-informant verification where feasible.

Despite these limitations, the study has notable strengths. Recruitment outside traditional school settings enabled inclusion of a high-risk population chronically underrepresented in school-based surveillance, including youth who are chronically absent, suspended, or disconnected from formal institutions. Confidential digital survey administration may have enhanced disclosure of sensitive behaviors relative to in-class formats. The 30-day recall period reduced potential recall bias compared to longer retrospective windows common in prior research.

Analytically, negative binomial regression provided a strong fit for the low-frequency, overdispersed count outcome, offering more precise estimates than traditional logistic approaches. Using both count and binary models enhanced robustness by distinguishing factors associated with any weapon carrying from those predicting carrying frequency among carriers. Multicollinearity diagnostics indicated stable coefficient estimates across model specifications. Finally, the emphasis on recent, proximal behaviors and lived experiences strengthens the relevance of findings for trauma-informed prevention planning in disproportionately affected communities.

5. Conclusions

Weapon carrying among urban young Black males in this sample is closely tied to cumulative perceived threats, chronic victimization, and environmental danger rather than to substance use, dating aggression, or socioeconomic characteristics in isolation. The strong and consistent associations between violence exposure and weapon-carrying frequency underscore the need for interventions that prioritize immediate safety, trauma recovery, and structural support. Prevention strategies are most effective when they target the upstream conditions generating violence exposure, strengthen community safety nets, and expand access to youth-centered resources that address both immediate safety concerns and long-term structural inequities, rather than focusing narrowly on individual strain or decision-making capacities.

Longitudinal designs are needed to clarify the temporal dynamics of weapon carrying, including seasonal and episodic patterns, and to better understand the motivations and contextual factors shaping these behaviors over time. This work is critical for informing context-relevant policy, resource allocation, and intervention development that centers the lived realities of young people in high-violence urban communities.