Neighborhood Predictors of Concealed Firearm Carrying Among Children and Adolescents: Results From the Project on Human Development in Chicago Neighborhoods

Background  Previous studies of concealed firearm carrying among children and adolescents have focused on individual risk factors.

Objective  To identify features of neighborhoods associated with concealed firearm carrying among a representative sample of youth from Chicago, Ill.

Design  Cross-sectional analysis of individual- and neighborhood-level data collected by the Project on Human Development in Chicago Neighborhoods.

Setting  Families and neighborhoods in Chicago.

Participants  Population-based sample of 1842 multiethnic youth aged 9 to 19 years and the 218 neighborhoods in which they resided.

Main Outcome Measure  Whether youth had ever carried a concealed firearm.

Results  Lifetime estimates for concealed firearm carrying were 4.9% for males and 1.1% for females. We found that youth in safer and less disordered neighborhoods were less likely than youth in unsafe and more disordered neighborhoods to carry concealed firearms. Specifically, multilevel nonlinear regression models identified a positive association between concealed firearm carrying and (1) community members' ratings of neighborhoods as unsafe for children; (2) neighborhood social disorder; and (3) neighborhood physical disorder. Neighborhood collective efficacy was negatively associated with concealed firearm carrying. Models controlled for neighborhood economic indicators and individual and family factors associated with the carrying of concealed firearms by youth.

Conclusions  Youth are less likely to carry concealed firearms in areas where there is less violence and increased safety. Interventions to improve neighborhood conditions such as increasing safety, improving collective efficacy, and reducing social and physical disorder may be efficacious in preventing firearm use and its associated injuries and death among youth.

Under federal law, individuals younger than 18 years are prohibited from owning handguns and in most states from carrying concealed weapons.¹ Nonetheless, data from the Youth Risk Behavior Survey^2,3 (YRBS) and other studies indicate high levels of self-reported gun carrying among American youth. During the last several years, the YRBS found that of high school students in the United States, 7.6% in 1995² and 5.7% in 2001³ reported having carried a firearm in the preceding 30 days. Not surprisingly, studies that ask if a child or adolescent has ever carried a firearm yield higher estimates than those that ask about current firearm carrying.⁴ For example, the 1994-1995 National Longitudinal Study of Adolescent Health⁵ provided a national estimate of 9.3% of 7th through 12th graders having ever carried a firearm at school.

Almost all of the published literature about concealed firearm carrying by youth focuses on how individuals who carry concealed firearms differ from those who do not. In aggregate, these studies have provided a detailed picture of individual-level predictors of carrying, including male sex,^6-10 minority background,¹¹ cigarette smoking,⁶ alcohol consumption,^6,8,10,12 drug use,^7,12-14 gang membership,¹⁵ criminal activity,⁷ fear of victimization,^{9,12,13,16,17} poor academic performance,^6,10 fighting,^4,8,18 lack of confidence in being able to stay out of fights,^6,13 and exposure to violence.^5,19,20

Neighborhood characteristics associated with violence may play an additional and possibly synergistic role in a youth's decision to carry a firearm. Consistent with this possibility, fear of victimization^{9,12,13,16,17} in one's neighborhood or school has been offered as a primary reason why youth carry concealed firearms. In addition, behaviors associated with carrying such as drug and alcohol use,^21,22 lower academic achievement,²³ delinquency,²⁴ and violent victimization²⁵ are known to be influenced by community factors.

Levels of social and physical disorder are 2 features of urban neighborhoods associated with fear of victimization. When strangers behave in public spaces in ways that are considered threatening, these behaviors constitute social disorder.²⁶ Examples include interactions such as drug dealing, prostitution, loitering, or public intoxication. In contrast, physical disorder describes the environmental deterioration of urban landscapes, such as abandoned properties, graffiti, and discarded drug paraphernalia and condoms. In addition to making people fearful of being victimized, observation of disorder can influence the attributions that are made about a neighborhood and may affect real estate prices, investors, and community residents' pride or willingness to actively contribute to its improvement.

Collective efficacy, on the other hand, is a neighborhood-level construct that has been linked to reduced levels and perceptions of violence. Sampson et al²⁵ defined this construct as a combination of social cohesion, represented by mutual trust and solidarity between neighbors, and informal social control, or the capacity of neighborhood residents to regulate the members of their community and reach collective goals.

Few studies to date have had the opportunity to explore youth firearm carrying within an overarching sociological framework emphasizing potential neighborhood-level determinants of individual development. Ours provides such an opportunity because questions about firearm carrying were embedded within a large multilevel study of neighborhood-level determinants of violence. To our knowledge, this study represents the first attempt to assess the independent effects of individual-, family-, and neighborhood-level characteristics on the probability that a youth has carried a concealed firearm.

Data come from the Project on Human Development in Chicago Neighborhoods (PHDCN), a multidisciplinary study of families and neighborhoods. Study families are residents of Chicago, Ill, and are of diverse races and ethnicities, socioeconomic status, and family structures. The PHDCN includes a comprehensive community design as well as a longitudinal cohort study embedded within the community design. Data used in this study are cross-sectional and come from 4 sources: (1) wave 2 of the longitudinal cohort study; (2) systematic social observation of study neighborhoods; (3) an independent community survey of Chicago residents; and (4) 1990 US census data. A detailed description of the PHDCN's design and methods is available elsewhere.²⁷ The institutional review boards of Harvard Medical School and the Harvard School of Public Health (Boston, Mass) approved all procedures. Informed consent was obtained from both caregivers and youth prior to each interview.

Sampling was accomplished in 2 stages. First, using 1990 US census data, 343 neighborhood clusters that were geographically compact as well as similar in ethnic composition, socioeconomic status, housing density, and family structure were identified and grouped into 7 strata of race and ethnic composition according to tertiles of socioeconomic level. Each neighborhood cluster contained approximately 8000 residents. Among these 21 strata were 3 empty cells; no neighborhood clusters were more than 75% Hispanic with high socioeconomic status, more than 75% white with low socioeconomic status, or of mixed Hispanic and African American race with high socioeconomic status. Eighty neighborhood clusters were chosen for intensive study from the remaining 18 cells using stratified random sampling, except for 3 cells with fewer than 5 neighborhood clusters that were sampled with certainty.

The second goal of sampling was to select families residing within each of the 80 selected neighborhood clusters to compose the longitudinal cohort study sample. Households were screened for eligible children and adolescents within 6 months of 7 target cohort ages (0 [in utero through age 6 months], 3, 6, 9, 12, 15, and 18 years). Within a selected household, all children and adolescents of eligible ages in 1994-1995 were invited to participate along with at least 1 primary caregiver for those younger than 18 years.

Data used in our investigation were obtained from youth in the 9-, 12-, and 15-year-old cohorts for whom the response rate was 74%. For these cohorts, baseline assessments were conducted with 2345 youth beginning in 1995 and ending in 1997, and follow-up interviews were done with 1921 youth approximately 24 months later, beginning in 1997 and ending in 2000. The analytic sample for this investigation consisted of everyone in this follow-up group who completed a questionnaire on firearm carrying (n = 1842). For the multilevel analyses (n = 1751), we excluded an additional 91 youth who moved outside the city limits of Chicago between baseline and follow-up for whom we no longer had neighborhood data. Mean values of the neighborhood variables calculated for this analytic sample were the same (to the first decimal place) as those for the entire baseline sample. The analytic sample had approximately the same demographic composition as the baseline one.

A random sample of adults were surveyed in 1995 to characterize residents' perceptions of Chicago neighborhoods, with oversampling done in the selected 80 neighborhood clusters. There was a 75% response rate, resulting in 8782 study participants representing all 343 neighborhood clusters. The survey focused on the social and organizational characteristics of neighborhoods, including assessments of violence and perceptions of neighborhood danger. This sample was independent of the longitudinal cohort sample of families.

For a description of natural social phenomena occurring in neighborhood environments, a systematic social observation study²⁸ was conducted using block-by-block videotaping and trained observers rating their observations as they were driven down city streets at approximately 5 mph. This study was conducted in 1995 and covered all 80 sampled neighborhood clusters, recording data from 23 816 face blocks (the left and right sides of a street when one is in the middle of the road) for a mean of 298 per neighborhood cluster. A random sample of 15 141 videotaped face blocks was selected for coding owing to budgetary constraints. Detailed information on physical conditions, housing characteristics, types of businesses, and negative social interactions was coded.

In this study, models that included measures from the systematic social observation study had a smaller sample size (n = 1606) than those including data from the community survey because 236 participants had moved away from the 80 initial study neighborhoods before the follow-up interview. As described earlier, the 1995 community survey covered the entire city of Chicago; thus, participants who moved to other Chicago neighborhoods were able to be kept in their particular models using data from the community survey. This increased the neighborhood sample size (n = 218).

Items regarding gun use were administered in the individual-level cohort study. Youth were asked (1) whether they had ever carried a concealed or hidden gun anywhere; (2) whether anyone in their home kept a gun; and (3) whether any family members had ever been shot or killed by a gun.

A composite measure of family socioeconomic status was created based on principal components analysis²⁹ of 3 variables—parental income, parental education, and parental occupational code—taking the maximum value for each if there were 2 parents in the household. A neighborhood-level index of concentrated poverty was derived from 1990 US census data for the proportion of residents who had an income lower than the poverty line, were unemployed, or were receiving public assistance.

Individual-level exposure to violence was measured by asking youths about their experiences during the past year.³⁰ Confirmatory factor analysis³¹ showed that 3 separate constructs were present. Hierarchical item response theory models³² confirmed which items fit within each construct, and 3 scales were developed: (1) victimization, (2) witnessing violence, and (3) hearing about violence. These scales showed good reliability and validity (R. T. Brennan, EdD, B.E.M., and F. Earls, MD, unpublished data, 2004). Respondents were categorized as having been victimized by violence if they reported being attacked with a weapon, beaten up, chased, shot at, or seriously threatened. If they had witnessed these same acts perpetrated on somebody else and/or saw someone being shot or killed, they were categorized as having witnessed violence. They were categorized as having heard about violence if they reported hearing about someone they knew being sexually assaulted, shot, or killed. Because the vast majority of the youth who carried concealed firearms were in the highest quartile of exposure to violence, dichotomous variables were also constructed for which 1 represented the highest quartile and 0 constituted the remainder of the sample.

Subjects were also administered a reduced version of the Youth Self-Report³³ externalizing behavior problems scale, which assesses delinquent and aggressive behaviors in the past 6 months. The scale scores are a sum of 21 items for which 0 = not true, 1 = somewhat true, and 2 = very true. To lessen response burden in the follow-up interviews, a reduced version of the Youth Self-Report scale was created with factor analysis using the sample's full-scale data from baseline interviews. Factor loadings for individual items contributing to the various subscales were examined, and items that loaded least strongly on the scale were dropped. The Cronbach α was calculated before and after dropping these items to ensure that the reliability of the scale measurement was not dramatically reduced. This was based both on published values for the Youth Self-Report scale normative samples³³ and on the baseline results. The reduced instrument included 8 of 13 items for the delinquent behavior subscale and 13 of 20 items for the aggressive behavior subscale.

Neighborhood conditions related to fear of victimization were measured on 4 scales with established psychometric properties. Individual items in each of the scales are listed in Table 1. The unsafe-to-play scale reflects residents' perceptions of the safety of neighborhood clusters based on the 1995 community survey. Collective efficacy, a summary measure of informal social control and social cohesion, was also derived from the community survey.²⁵ Each scale was formed by finding the mean for the item responses. Although most respondents in the community survey answered all of the items, anyone responding to at least 1 item on a scale provided data for the analysis. Using an unconditional 3-level hierarchical linear item-response model that took into account the number and difficulty of the answered items, a person-specific standard error was calculated and was used to adjust for missing responses.²⁵ Scales of social disorder and physical disorder were derived from the systematic social observation study; descriptions of the methods and psychometric properties of these scales have been published previously.^26,34 Ranges of values for each scale are presented in Table 1. Prior to using these scales in the models described as follows, they were all standardized to have a mean of 0 and an SD of 1.

For variables that were categorical, the χ² test of proportions was used to assess bivariate relationships between risk factors and the carrying of concealed weapons. Spearman correlation coefficients were used to test for bivariate relationships between carrying firearms and risk factors that were continuous, including the 4 neighborhood-level scales in the study. Spearman correlation coefficients were chosen because the dichotomous outcome variable is not normally distributed, requiring a nonparametric test.

Because of the multilevel design, hierarchical nonlinear modeling³⁵ was used to estimate the relationship between individual- and neighborhood-level independent variables and youth gun carrying. The outcome was dichotomous, so the analysis involved a standard hierarchical linear model with the introduction of special weighting to linearize the outcome at level 1. After convergence, the weights were recalculated. This method resembles logistic regression, but by using multilevel methods to account for clustering by neighborhoods, the result is a proper estimation of standard errors. It also makes it possible to estimate contextual factors that may influence behavior independent of individual characteristics. Datasets were prepared for analysis using SAS version 8.0 statistical software (SAS Institute Inc, Cary, NC) and were analyzed with the statistical program HLM 5.³⁵

The goal of the data analysis was to estimate the magnitude of the association between 4 central neighborhood-level constructs and youth firearm carrying, controlling for established individual- and family-level risk factors. Initial models included the neighborhood-level constructs, a fixed set of individual-level variables (sex, age, race and ethnicity, and family socioeconomic status), and neighborhood-level concentrated poverty. Theoretically, we considered the 4 neighborhood-level constructs to measure distinct but overlapping aspects of the environment. Because of high correlations between them, they were not able to be modeled simultaneously. Instead, the models that contained multilevel potential confounders were used 4 times with each of the primary neighborhood exposures added one at a time. Finally, individual-level risk factors previously identified in the literature were added to each model: whether the youth knew that a gun was kept at home, whether a family member had been shot with a gun, whether the youth had high exposure to violence in the past year (either witnessing violence, being victimized, or hearing about others being victimized), and whether he or she had committed other acts of delinquency and/or aggression the past 6 months.

Fifty-six youth (3%) indicated that they had carried a concealed firearm. In the bivariate analyses in Table 2, whereas only half (50%) of the sample were male, they represented 82% of those carrying a concealed firearm. More firearms were carried by older youth, with 71% of the firearm carriers aged 16 to 19 years. There were no significant differences in concealed firearm carrying by race or ethnicity or between families of different socioeconomic status.

Although 10% of respondents reported living in a household with a firearm, they accounted for 19% of those who carried a firearm themselves. One third (32%) of respondents reported that they had a family member who had been shot with a gun, and these accounted for 76% of participants who carried concealed firearms.

Of the 56 youth who reported having carried a concealed gun, most had been exposed to high levels of violence: 93% were in the highest quartile on the scale of witnessing violence in the past year, 63% were in the highest quartile of having heard about violent victimization of acquaintances during the past year, and 91% were in the highest quartile reporting victimization by violence in the previous year. Similarly, most firearm carriers (63%) were in the highest quartile on the scale of delinquent and aggressive behaviors.

At the neighborhood level, there were significantly higher rates of concealed firearm carrying by youth in neighborhoods that were less safe and in those characterized by greater social disorder, physical disorder, and concentrated poverty. Youth living in neighborhoods with higher levels of collective efficacy reported less firearm carrying (Table 2). In multilevel multivariate models, all of the neighborhood characteristics included in this study were significantly associated with youth carrying concealed firearms when controlling for economic conditions at both family and neighborhood levels as well as for age, sex, and race and ethnicity.

The primary focus of this investigation was whether these neighborhood characteristics were associated with concealed firearm carrying above and beyond individual and family correlates. The final step was to include previously identified risk factors at the level of the individual. Table 3 presents the results of models that included variables measuring whether a gun was kept in the home, whether a family member had been shot with a gun, individual witnessing of violence, victimization, hearing about others' victimization by violence, and delinquency or aggression by youth. For neighborhoods rated as unsafe to play, the odds of youth carrying a concealed firearm were increased (odds ratio [OR] = 5.8; 95% confidence interval [CI], 1.4-24.5); for collective efficacy, the odds were decreased (OR = 0.3; 95% CI, 0.2-0.7); for social disorder, the odds were increased (OR = 1.9; 95% CI, 1.2-3.2); and for physical disorder, the odds were increased (OR = 2.3; 95% CI, 1.2-4.2). Given that all of the scales were standardized prior to analysis to have a mean of 0 and an SD of 1, these findings can be interpreted as the odds of gun carrying associated with a 1-SD increase in each neighborhood measure. There were no significant changes in either the magnitude or direction of the odds of youth carrying concealed firearms associated with neighborhood characteristics after individual-level risk factors were included.

Consistent with the idea that social and organizational characteristics of neighborhoods explain further variations than those accounted for by aggregated characteristics of individuals and families,³⁶ we found that neighborhood characteristics likely to increase fear of victimization predicted concealed firearm carrying among Chicago youth. Specifically, carrying of concealed firearms was positively related to both a neighborhood's level of social and physical disorder and the scarcity of safe places for children to play. These neighborhood-level associations held after controlling for individual-level and family characteristics of respondents such as sex, age, exposure to various forms of violence, household firearm ownership, and other delinquent or aggressive behaviors and after accounting for economic disadvantages of families and neighborhoods.

Additionally, youth who lived in neighborhoods with higher levels of collective efficacy were significantly less likely to carry a concealed firearm. Youth in these neighborhoods may feel that the community provides a measure of vigilance and protection as well as a level of oversight that makes the act of carrying a firearm no longer necessary or attractive. Many of the youth who lived in neighborhoods with higher collective efficacy were still exposed to violence; the correlation between collective efficacy and witnessing violence was low at 0.11. Informal social control of children, a component of collective efficacy, was associated with reduced delinquent behavior of adolescents in Chicago in a previous study.³⁷ Perhaps the social forces and norms found in these neighborhoods protected youth from taking the extreme step of arming themselves with guns even if they were exposed to violence.

Previous studies have shown that most adults (91%) who carry concealed firearms obtain them from their home.⁶ Given that handgun ownership is illegal in Chicago,¹ youth who report guns kept at home may come from families with a greater propensity for illegal behavior. However, only 2 (4%) of 56 firearm-carrying youth in our sample reported carrying a gun from home. In contrast, 64% had received or borrowed their gun from friends or family, 4% had stolen or traded for the gun, 5% had acquired the gun through unspecified means, and 23% had bought the gun themselves, consistent with similar findings by other investigators.³⁸

Several individual-level risk factors previously reported to predict concealed firearm carrying among youth appear to predict carrying among Chicago youth, including male sex,^{6-8,10,11,15,16} older age,¹⁵ having a family member who has been shot,⁶ and other delinquent or aggressive behaviors.^4,7,8,18 We found a strong association between individual exposure to violence and carrying a concealed firearm, a relationship that held whether the youth had experienced, witnessed, or heard about violence. These results are consistent with literature that found that the primary reason youth carry concealed firearms is for self-protection.^15,38,39 Although gun carriers in our survey overwhelmingly reported that they carry guns for self-protection, some gun carrying may be by youth who are engaged in a variety of criminal and antisocial activities. We found an association between carrying a concealed firearm and individual delinquency and aggression.

In this study, 4.9% of males and 1.1% of females between the ages of 9 and 19 years reported having carried a concealed firearm. These levels are low relative to those reported elsewhere,^4,6,13,15,40 perhaps because of low rates of gun ownership in Chicago, where handguns are illegal.¹ In the 1997 YRBS conducted in Chicago, the 30-day prevalence rates were reported to be 2.5% among females and 11.5% among males.⁴¹ Restricting the current analysis to youth aged 14 to 18 years, the prevalence of having carried a gun was 9.3% among males and 1.8% among females. Thus, the prevalence in this study is still lower than the YRBS estimates. One possible explanation may be social desirability bias due to the differences in study methods. Carrying a concealed gun is a rare and extreme activity; the YRBS is self-administered and anonymous, whereas our study was interviewer administered in subjects' homes. Therefore, the prevalence of firearm carrying measured in this study may be an underestimate.

Our study has several limitations. These data come from Chicago, a large city that is heavily segregated and has particular laws that make handguns illegal to own. The results may not be generalizable either to other urban areas without similar gun laws or segregation or to nonurban areas. To the extent that youth who carry guns contribute to adults' perceptions of the social impoverishment of their neighborhoods, it is possible that the relationship between gun carrying and neighborhood is bidirectional (predominantly the reverse of what we describe) or not causally related. We were unable to evaluate this possibility because we used cross-sectional data and could not estimate neighborhood-level gun carrying using data from our sample, having only 56 gun carriers distributed across 218 neighborhoods. A bidirectional, causal relationship between gun carrying and neighborhood safety offers the hope that an intervention targeted at either firearm carrying or neighborhood disorder might break or at least mitigate a seemingly vicious cycle. In addition, we asked only whether youth in our survey had ever carried a gun, not how often or how recently. As a result, we are unable to determine whether there is a relationship between the frequency of gun carrying and our neighborhood scales. Finally, we did not assess whether peers of the youths in the study carried guns, a potentially important determinant of whether a youth carries a firearm, although we did control for other delinquent and aggressive behaviors by the respondents.

Despite these limitations, an important strength of this study is the large, multilevel design that allows us to examine the association between neighborhood social processes and a rare outcome such as concealed firearm carrying among adolescents. An additional advantage is that our questions about firearm carrying and our measures of neighborhood characteristics were obtained from separate samples. Because these data on neighborhood features do not come from the youth participants themselves, they provide an independent source of information about perceived and observed levels of disorder, safety, and collective efficacy that exist across neighborhoods.

Each of these neighborhood characteristics is potentially modifiable and thus important when considering both primary and secondary prevention of youth violence. The neighborhood-level constructs identified as strongly associated with gun carrying by youth in this cohort have not previously been described and represent potential targets for public health interventions. Decreasing the degree to which a community is burdened by threatening or hostile behavior by adults, reclaiming the number of public spaces available for children to play in, and supporting adults who are willing to intervene in the lives of their community's children may positively alter the social dynamics that currently contribute to concealed firearm carrying by youth.

Article

Efficacious interventions are necessary to reduce the carrying of firearms by adolescents, a population at high risk for gun-related injuries and death. Previous research has identified individual-level risk factors for youth carrying firearms, especially concerns regarding personal safety in their communities. However, to our knowledge, no studies have used a multilevel approach to identify processes in those communities that may increase or decrease the risk of carrying firearms. This study uses a multilevel approach to identify potentially modifiable neighborhood-level correlates of firearm carrying by adolescents. Neighborhood social disorder, physical disorder, and the rating of neighborhoods as unsafe for children to play in were each independently associated with increased firearm carrying by youth. In addition, a positive social process, collective efficacy, was found to be independently associated with decreased firearm carrying. All of these neighborhood-level correlates remained significant when individual-level risk factors were included in the models, demonstrating that reducing social and physical disorder, increasing safety, and promoting collective efficacy among neighbors have the potential to decrease the prevalence of youth carrying concealed firearms and to reduce gun-related injuries.

Accepted for publication March 4, 2004.

This study was supported by grant R49/CCR115279-05 from the Centers for Disease Control and Prevention (Atlanta, Ga) to the Harvard Injury Control Research Center. Funding for the PHDCN was provided by the John D. and Catherine T. MacArthur Foundation (Chicago, Ill), the National Institute of Mental Health (Bethesda, Md), and the National Institute of Justice (Washington, DC).

We thank Kathy McGaffigan for her assistance with data analyses, Harvard Injury Control Research Center director David Hemenway, PhD, for his guidance, and the field staff in Chicago that tirelessly collected these data. We acknowledge the roles of Jeanne Brooks-Gunn, PhD, Felton Earls, MD, Stephen Raudenbush, PhD, and Robert Sampson, PhD, the scientific directors of the PHDCN and reviewers of this manuscript. We are especially grateful to the families participating in the PHDCN who gave us their time and energy to help us better understand their lives and communities.

This article is solely the responsibility of the authors and does not represent the official views of the Centers for Disease Control and Prevention.

Correspondence: Beth E. Molnar, ScD, Department of Society, Human Development, and Health, Harvard School of Public Health, 677 Huntington Ave, Sixth floor, Boston, MA 02115 (bmolnar@hsph.harvard.edu).