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

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The author(s) test a wide array of socioeconomic, psychological, and behavioral correlates of weapon carriage among Black youth who reside in an urban center. The paper’s strength is that it evaluates a broad range of correlates associated with weapon carriage in an understudied population. It also invokes a substantial body of relevant theory, which I appreciate. However, I believe the manuscript could be improved by attending to several areas:

• The author(s) measured weapon carriage, which includes, but is not limited to, firearms (e.g., knives, bats, and other non-firearm weapons). The author(s) acknowledge this as a limitation, which is helpful; however, the introduction and discussion focus too heavily on firearm carriage and firearm violence (e.g., firearm access/acquisition). Although this literature is relevant, the manuscript should incorporate additional empirical studies, theories, and discussion that examine youth weapon carriage more broadly, given that this is not specifically a firearm carriage paper. For example, discussions of firearm acquisition, access, and firearm-related beliefs are relevant, but the outcome assessed was not limited to firearm carriage. As currently framed, the manuscript suggests a focus on firearm-related behaviors even though the outcome encompasses both firearm and non-firearm weapon carriage.
• Related to the first point, given the scope of the intervention, did the author(s) capture any information on firearm acquisition, firearm carriage, or related firearm-specific behaviors? If the intervention sought to address firearm violence, it is somewhat surprising that no measures appear to capture firearm-specific behaviors.
• The Measures section is somewhat difficult to follow. It would be helpful if, for each variable, the author(s) clearly reported the number of items used to measure the construct, inter-rater or internal reliability (as applicable), a sample item for multi-item scales (or the full item if a single-item measure), and the corresponding citations.
• Was there missing data in the analytic sample, and if so, how was it addressed?
• Education level is used as an indicator of socioeconomic status; however, some participants may still be enrolled in school. For example, adolescents who are still in high school (e.g., ages 16–17) may appear to have lower socioeconomic status simply due to their current educational stage, which could affect interpretation of this measure.

Author Response

We thank the reviewer for these thoughtful observations, and have revised as suggested in the following ways:

Comment #2: We agree that our outcome reflects weapon-carrying behavior broadly rather than firearm-specific carriage. In response, we revised the Introduction and Discussion to clarify that the study examines weapon carriage as a general behavioral response to perceived threat and safety appraisal, within which firearm carriage represents one possible manifestation. We reduced firearm-specific framing and incorporated literature on youth weapon carriage, school-based weapon carrying, and protection-motivated behavior to better align the narrative with the measured outcome. We now explicitly state that the YRBS item captures general weapon-related risk rather than firearm-specific behavior and interpret findings accordingly.

Comment #2: We appreciate this important question. The present manuscript is a secondary analysis of baseline screening data collected before intervention delivery. The baseline screening battery was adapted from the YRBS to identify youth at elevated violence risk and therefore assessed general weapon carriage rather than firearm-specific acquisition or access. The intervention itself addresses firearm violence prevention; however, firearm-specific behaviors were not assessed at screening to reduce participant burden and minimize disclosure concerns among at-risk youth. We now clarify this distinction in the Methods section.

Comment #3: We appreciate this suggestion and have substantially revised the Measures section. For each construct, we now report (under subheadings) the number of items, provide a sample item, describe response options, and include citations to the original instruments (though most items are from the YBRS). 

Comment #4; We have added a Missing Data subsection to the Statistical Analysis section. Missingness was low and handled using listwise deletion in regression models. Sample size variation across models reflected item-level nonresponse rather than attrition. We now report this clearly.

Comment #5: We agree that educational attainment among adolescents primarily reflects developmental stage rather than SES. We have revised the manuscript to interpret education as an indicator of school engagement and developmental positioning rather than SES and have clarified this in the Methods and Discussion.

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript examines weapon carrying among high risk youth in violence exposed contexts, an important topic and one that school based surveillance systems often miss. The paper can make a stronger contribution with clearer documentation of sample construction and missingness, sharper discussion of measurement comparability to YRBS including item wording and survey mode, and more careful model specification and diagnostics given the small sample and correlated predictors. None of these issues are fatal, but addressing them would materially improve interpretability and credibility.

1. Please provide a clear sample flow from enrollment to the final analytic sample and summarize missingness by variable. A comparison of included versus excluded participants on baseline demographics and key risk measures would help assess whether listwise deletion likely induces selection bias. If feasible, consider multiple imputation. If not, at minimum show that key results are similar when restricting to a common sample across models.
2. Generalizability claims should be tightly bounded to the sampling frame. Because eligibility required prior violence exposure and recruitment occurred through specific channels, the results should be framed as applying to a high risk convenience sample rather than broader youth or national trends. This is largely a wording issue and can be fixed throughout the abstract, discussion, and conclusions. Given the design, I recommend consistently describing estimates as conditional associations within this sampling frame.
3. The outcome is 30 day weapon carrying, while many predictors are past year measures and some may reflect longer screening or eligibility windows. That structure can be reasonable, but the manuscript should spell out what it implies for interpretation. Past year victimization could precede, co occur with, or follow recent carrying in a cross sectional snapshot. Please avoid language that implies temporal ordering or causal pathways unless the timing supports it. Where possible, add a sensitivity analysis emphasizing predictors with comparable recall windows or add a short interpretive note clarifying that these are associations across mismatched time windows.
4. Please include an appendix or supplement with the exact wording and response options for key items, especially those adapted from YRBS, and clearly note any modifications such as whether “on school property” was removed and whether response options were categorical versus 0 to 30. Please also address that YRBS now includes gun specific content and that newer work uses these items. Even if your dataset cannot reproduce that measure, it matters for comparability. Please cite a few recent papers that use the gun specific item and discuss how your measure maps onto that literature.
5. Because survey mode can affect reporting of sensitive behaviors, please describe how the survey was administered, for example phone or app self administered versus staff present, and describe privacy protections and anonymity. If details are in a blinded parent study paper, it would still help to include a short summary here. Differences relative to YRBS administration could plausibly shift reporting and should be discussed explicitly as a comparability and limitations issue.
6. With N around 226 and predictors spanning multiple overlapping domains such as victimization, witnessing violence, aggression, partner violence, substance use, and socioeconomic factors, there is a real risk of multicollinearity and overfitting, especially given that many constructs are likely correlated. Please discuss this directly and note potential instability of estimates. Empirically, I suggest a more parsimonious strategy such as domain by domain models or a pre specified core model plus a final model, and consider a robustness check that reduces dimensionality using a domain index or standardized composite. Reporting basic collinearity diagnostics such as correlations could also help.
7. Given you already describe a binary any carrying outcome defined as at least one day, please estimate a logit or probit model and report marginal effects as a robustness check, either in the main text or in the appendix with a brief discussion in the results. This can also address concerns about nonlinearities that a single count specification may not capture. Please add a simple figure such as a bar plot or histogram of the weapon carrying distribution to show the mass at zero and the right skew, which will help motivate the chosen model and the overdispersion assumptions. If zeros are substantial, a two part approach separating any carrying from intensity among carriers would be a helpful sensitivity check.
8. The manuscript’s value added beyond what YRBS can already answer is not fully explicit. If the key advantage is capturing youth not represented in school based surveillance, a violence exposed high risk population, an intervention relevant setting, or additional measures not available in YRBS, please state this clearly and early. It would also strengthen the paper to compare your descriptive prevalence and correlates to YRBS estimates from the same year, noting where results are consistent versus divergent and offering plausible explanations such as sampling frame and survey mode.
9. Please update the literature review and reference list through 2025 since this area is moving quickly. Incorporate recent empirical work on youth weapon and gun carrying and exposure to firearm violence, including nationally representative studies and newer longitudinal or contextual analyses. It may also help the framing to briefly discuss mass and school shootings as environmental violence exposures that can shift perceived safety and mental health, potentially affecting protective behaviors including weapon carrying. If the intervention context is central, consider citing recent work on relevant digital interventions targeting aggression, safety perceptions, or firearm related attitudes and behaviors, only where it genuinely fits the manuscript’s aims. The literature review would read more smoothly as an integrated narrative with transitions rather than segmented sections.
10. Please ensure each table is self contained with variable definitions, coding decisions especially recodes and bins, time windows, and the estimation sample used. Include N for each model.
11. Please report enough model information for readers to evaluate fit, including fit statistics, the overdispersion parameter, how zeros were handled, and whether influential observations were assessed.
12. Please describe findings consistently as associations rather than effects, given the cross sectional design and potential simultaneity in timing.
13. A quick editorial pass to reduce redundancy between the introduction and discussion would improve readability.

Author Response

1. Comment 1: Please provide a clear sample flow from enrollment to the final analytic sample and summarize missingness by variable. A comparison of included versus excluded participants on baseline demographics and key risk measures would help assess whether listwise deletion likely induces selection bias. If feasible, consider multiple imputation. If not, at minimum show that key results are similar when restricting to a common sample across models.
   1. Response: We added a clearer description of the sample flow from enrollment through the final analytic sample and documented reasons for exclusions. To assess whether listwise deletion may have induced selection bias, we compared included vs excluded participants on baseline demographics and key risk measures, and we reported standardized differences alongside p-values to emphasize magnitude, not only statistical significance.
2. Comment 2: Generalizability claims should be tightly bounded to the sampling frame. Because eligibility required prior violence exposure and recruitment occurred through specific channels, the results should be framed as applying to a high risk convenience sample rather than broader youth or national trends. This is largely a wording issue and can be fixed throughout the abstract, discussion, and conclusions. Given the design, I recommend consistently describing estimates as conditional associations within this sampling frame.
   1. Response: We revised generalizability statements throughout the abstract, discussion, and conclusions to align with this sample. We described the study as a "high-risk convenience" sample recruited through specified channels with eligibility requiring prior violence exposure, and we removed language suggesting broader youth or national inference. We also reframed the interpretation of estimates as conditional associations within this sampling context, explicitly noting that the estimates should not be interpreted as population-level prevalence or causal effects.
3. Comment 3: The outcome is 30 day weapon carrying, while many predictors are past year measures and some may reflect longer screening or eligibility windows. That structure can be reasonable, but the manuscript should spell out what it implies for interpretation. Past year victimization could precede, co occur with, or follow recent carrying in a cross sectional snapshot. Please avoid language that implies temporal ordering or causal pathways unless the timing supports it. Where possible, add a sensitivity analysis emphasizing predictors with comparable recall windows or add a short interpretive note clarifying that these are associations across mismatched time windows.
   1. Response: We clarified the implications of mismatched recall windows, given the past-30-day weapon carrying outcome and predictors measured over the past year (and, for some items, longer windows). We added an interpretive note stating that past-year victimization could precede, co-occur with, or follow recent weapon carrying in a cross-sectional snapshot, and we removed any language implying temporal ordering or causal pathways that the timing cannot support. 
4. Comment 4: Please include an appendix or supplement with the exact wording and response options for key items, especially those adapted from YRBS, and clearly note any modifications such as whether “on school property” was removed and whether response options were categorical versus 0 to 30. Please also address that YRBS now includes gun specific content and that newer work uses these items. Even if your dataset cannot reproduce that measure, it matters for comparability. Please cite a few recent papers that use the gun specific item and discuss how your measure maps onto that literature.
   1. We added a supplement in the appendices that lists the exact wording and response options for key items, including those adapted from YRBS. For each adapted item, we documented modifications, such as location qualifiers “on school property” were removed, and whether response options were categorical versus 0–30 days.
   2. We also added a short sentence acknowledging that the YRBS has included gun-specific items in more recent cycles (since 2017). We cited recent studies using the gun-specific item and positioned our measure as related but not interchangeable, which we treated explicitly as a comparability limitation.
5. Comment 5: Because survey mode can affect reporting of sensitive behaviors, please describe how the survey was administered, for example phone or app self administered versus staff present, and describe privacy protections and anonymity. If details are in a blinded parent study paper, it would still help to include a short summary here. Differences relative to YRBS administration could plausibly shift reporting and should be discussed explicitly as a comparability and limitations issue.
   1. Response: We expanded the methods to describe survey administration mode and privacy protections, given the sensitivity of weapon carrying and related exposures. We specified whether completion occurred via self-administered phone/app/web survey versus staff-present administration, and we summarized procedures used to enhance privacy and reduce social desirability bias (for example, private completion space, minimal staff visibility of responses, confidentiality language, and secure data handling).
   2. We also added a limitations note explaining how differences in survey mode and context relative to YRBS could plausibly shift reporting, and we framed this explicitly as a comparability issue when interpreting prevalence and correlates.
6. Comment 6: With N around 226 and predictors spanning multiple overlapping domains such as victimization, witnessing violence, aggression, partner violence, substance use, and socioeconomic factors, there is a real risk of multicollinearity and overfitting, especially given that many constructs are likely correlated. Please discuss this directly and note potential instability of estimates. Empirically, I suggest a more parsimonious strategy such as domain by domain models or a pre specified core model plus a final model, and consider a robustness check that reduces dimensionality using a domain index or standardized composite. Reporting basic collinearity diagnostics such as correlations could also help.
   1. Response: We have addressed multicollinearity and overfitting directly, reporting basic collinearity diagnostics, including correlation structure among key predictors and VIF (below threshold). We also completed a robustness check that reduced dimensionality using a domain composite (unadjusted, intermediate, and full), and noted interpretive caution for unstable parameters.
7. Comment 7: Given you already describe a binary any carrying outcome defined as at least one day, please estimate a logit or probit model and report marginal effects as a robustness check, either in the main text or in the appendix with a brief discussion in the results. This can also address concerns about nonlinearities that a single count specification may not capture. Please add a simple figure such as a bar plot or histogram of the weapon carrying distribution to show the mass at zero and the right skew, which will help motivate the chosen model and the overdispersion assumptions. If zeros are substantial, a two part approach separating any carrying from intensity among carriers would be a helpful sensitivity check.
   1. Response: Because we already defined an “any carrying” outcome (≥1 day), we further estimated a logit model as a robustness check and report marginal effects,  presented in the main results. We added a Distribution of weapon-carrying days histogram in the appendices, appendix 2. 
8. Comment 8: The manuscript’s value added beyond what YRBS can already answer is not fully explicit. If the key advantage is capturing youth not represented in school based surveillance, a violence exposed high risk population, an intervention relevant setting, or additional measures not available in YRBS, please state this clearly and early. It would also strengthen the paper to compare your descriptive prevalence and correlates to YRBS estimates from the same year, noting where results are consistent versus divergent and offering plausible explanations such as sampling frame and survey mode.
   1. Response: We strengthened the articulation of the manuscript’s value early in the introduction. We state explicitly that the contribution was not to replicate school-based surveillance, but to characterize weapon carrying and correlates within a violence-exposed, intervention-relevant, high-risk convenience sample that may not be well represented in school-based national surveillance.
   2. We also added a brief contextual comparison to YRBS estimates from a comparable time period, focusing on directional consistency of correlates rather than direct prevalence, and in the Discussion, we offered plausible explanations for divergence (sampling frame and survey mode).
9. Comment 9: Please update the literature review and reference list through 2025 since this area is moving quickly. Incorporate recent empirical work on youth weapon and gun carrying and exposure to firearm violence, including nationally representative studies and newer longitudinal or contextual analyses. It may also help the framing to briefly discuss mass and school shootings as environmental violence exposures that can shift perceived safety and mental health, potentially affecting protective behaviors including weapon carrying. If the intervention context is central, consider citing recent work on relevant digital interventions targeting aggression, safety perceptions, or firearm related attitudes and behaviors, only where it genuinely fits the manuscript’s aims. The literature review would read more smoothly as an integrated narrative with transitions rather than segmented sections.
   1. Response: We updated the literature review and reference list through 2025, incorporating recent empirical work on youth weapon and gun carrying, exposure to firearm violence, and newer longitudinal or contextual analyses. We also tightened the structure of the literature review, converting segmented sections into a more integrated narrative with transitions that lead directly to the study’s aims and rationale. 
10. Comment 10: Please ensure each table is self-contained with variable definitions, coding decisions especially recodes and bins, time windows, and the estimation sample used. Include N for each model.
    1. Response: Tables are now self-contained (with definitions, coding, time windows, estimation sample, N per model)
       We revised each table to be self-contained by adding table notes with variable definitions, coding decisions (including recode decisions), recall windows, and the estimation sample used. 
11. Comment 11: Please report enough model information for readers to evaluate fit, including fit statistics, the overdispersion parameter, how zeros were handled, and whether influential observations were assessed.
    1. Response: We expanded model reporting so readers could evaluate fit and assumptions. We reported fit statistics (AIC, and BIC, where available), described how zeros were handled (retained and modeled directly), and documented checks for influential observations using standard residual and leverage diagnostics, noting that conclusions were not meaningfully altered by excluding influential cases.
12. Comment 12: Please describe findings consistently as associations rather than effects, given the cross sectional design and potential simultaneity in timing.
    1. Response: We revised the results and discussion to describe findings consistently as associations, not effects, reflecting the cross-sectional design and potential simultaneity in timing. We also added a brief interpretive statement reinforcing that estimates reflect conditional associations within the study’s sampling frame and measurement windows.
13. Comment 13: A quick editorial pass to reduce redundancy between the introduction and discussion would improve readability.
    1. Response: We completed an editorial pass to reduce redundancy between the introduction and discussion.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I appreciate the authors’ revisions to the manuscript. However, the introduction continues to read as though the study is focused primarily on firearm carriage. Because the study appears to assess weapon carriage more broadly, the manuscript would benefit from incorporating empirical studies that focus on the determinants of general weapon carriage and violence exposure (e.g., homicide beyond firearm-specific incidents).

Consistent with the framing presented in the discussion, the introduction should more explicitly situate firearm carriage research within the broader context of youth weapon-carrying behaviors. For example, after presenting firearm-specific findings, the authors could acknowledge that these associations may plausibly extend to other forms of weapon carriage.

Author Response

We thank the reviewer for this thoughtful and constructive feedback. We agree that the introduction, as previously written, placed disproportionate emphasis on firearm carriage specifically, which risks misrepresenting the broader scope of the study.

In response, we have revised the introduction to more explicitly situate firearm carriage research within the wider context of youth weapon-carrying behaviors. Specifically, we have taken the following steps:

1. We added empirical literature on the determinants of general weapon carriage and violence exposure, including homicide data that extends beyond firearm-specific incidents, to provide a more comprehensive evidentiary foundation.
2. Following the presentation of firearm-specific findings, we now include transitional language that acknowledges these associations may plausibly extend to other forms of weapon carriage. This language mirrors the framing already present in the discussion section, creating greater coherence across the manuscript.
3. We have reframed key introductory statements to signal from the outset that the study addresses weapon carriage broadly, with firearm carriage as one important dimension of that broader construct.

We believe these revisions strengthen the conceptual alignment between the introduction and the study's actual scope, and we are grateful to the reviewer for identifying this inconsistency. The revised introduction appears in the introduction/background section of the updated manuscript.

Reviewer 2 Report

Comments and Suggestions for Authors

I want to thank the authors for their diligent work addressing all my comments. Excellent, very important work. It was a pleasure reviewing your work.

Author Response

No comment from reviewer. We would like to thank the reviewer for their thoughtful and constructive feedback throughout this process. Their insights significantly strengthened the manuscript, and we are grateful for their time and expertise. It was a pleasure to engage with such a diligent review.

Round 3

Reviewer 1 Report

Comments and Suggestions for Authors

The introduction aligns well with the study’s focus on weapon carriage.

I have one final suggestion. If possible, it would help further contextualize the firearm-related component if the author(s) noted that rates of firearm carriage are often substantially higher in disinvested and segregated communities than in national samples. I believe estimates in some studies are closer to 20-25% (for example, work by Patrick Carter and Rebecca Cunningham in their Flint Youth Injury study). Including this point would help underscore the significance of discussing firearm carriage within the broader context of weapon carriage, particularly in sections such as “Emerging and Contemporary Weapon-Carrying Behaviors,” where the emphasis is more heavily on firearm carriage.

Author Response

We appreciate the reviewers’ insightful and constructive comments. In response to the request to better contextualize firearm carrying within the broader ecology of weapon carrying, particularly in relation to differential patterns across marginalized communities, we have added the following text to the intro section - lines 65-70:

"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., 2013), 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."

Thank you