Chicago Youths’ Exposure to Community Violence: Contextualizing Spatial Dynamics of Violence and the Relationship With Psychological Functioning

Abstract

This study explores where and when community violence exposure (CVE) matters for psychological functioning in a sample of low-income, racial/ethnic minority youth (M) age = 16.17, 55% female, 69% Black, and 31% Non-Black/Latinx) living in Chicago. CVE was measured with violent crime data that were geocoded in terms of distance from youths’ home and school addresses, and then calculated in terms of three distinct spatial dynamics: chronicity, pervasiveness, and spatial proximity. These measures reflect indirect/objective CVE across different conceptualizations of time, space, and neighborhood context. We tested the relationship between each CVE measure and trait anxiety and behavioral and cognitive dysregulation while controlling for youth-reported, direct violent victimization (e.g., being attacked) to examine how indirect/objective CVE occurring within youths’ neighborhood contexts matters beyond direct/subjective violence exposure. Results revealed that long-term chronic, pervasive, and spatially proximal CVE was related to higher levels of behavioral dysfunction. In contrast, CVE within home- and school-based neighborhoods interacted to predict trait anxiety; youth living in low-crime neighborhoods and attending schools in high-crime neighborhoods had the highest rates of trait anxiety. Measuring CVE within both home and school neighborhoods at specific spatial measurements and time frames is critical to understand and prevent the consequences of CVE.

Introduction

Violence exposure and its negative impacts on American youth is recognized as a national crisis (Finkelhor, Turner, Shattuck, Hamby, & Kracke, 2015). Nationally, 18% of youth witnessed violence in the community over the previous year and 28% witnessed community violence in their lifetime (Finkelhor et al., 2015). In Chicago, youth exposure to community violence is alarmingly prevalent; one study found that 21% of youth witnessed someone being attacked with a weapon, and 56% reported hearing gunshots (Zimmerman, 2014). In 2016, Chicago saw a surge of gun violence and homicides, despite two decades of reduction in rates of gun violence (University of Chicago Crime Lab, 2017). Another study found that 72 violent crimes occurred in the average Chicago Public School’s (CPS) neighborhood annually (Burdick-Will, 2016).

Community violence exposure (CVE) is defined as violence experienced directly or indirectly in or near the home and school neighborhoods (Scarpa, 2003). This operationalization best captures how violence is experienced broadly within relevant intersections of youth’s social ecologies. A growing body of literature has linked CVE to the presence of psychological dysfunction (Fowler, Tompsett, Jacques-Tiura, & Baltes, 2009; McDonald & Richmond, 2008), but research has yet to delineate spatial and temporal elements of CVE. That is, where and when CVE occurs might have substantial implications to guide future research, interventions, and public policies. To this end, this study investigates where and when CVE relates to psychological outcomes in a sample of youth living in low-income, high-crime, predominately south, and west-side communities in Chicago. It proposes that CVE is positively associated with two key areas of youths’ psychological dysfunction: anxiety and dysregulation. Three distinct spatial dynamics of CVE are explored to test in what ways home and school neighborhood-based CVE matters for youths’ psychological functioning, including to what extent CVE is chronic (e.g., repeated exposure across time and space), pervasive (e.g., experienced in multiple neighborhood contexts), and spatially proximate (e.g., closeness to youths’ home or school). This is accomplished by linking violent crime statistics with youths’ home and school addresses to calculate precise measures of each spatial dynamic. We then estimate relationships with psychological outcomes, adjusting for youths’ self-reported direct violent victimization (DVV), therefore exploring how youths’ indirect exposure to objective indicators of violent crime is related to psychological functioning beyond direct, subjective reports of direct violence exposure.

Evidence Linking CVE With Psychological Dysfunction in Youth

There is a relatively large body of research linking CVE with psychological dysfunction in youth. A meta-analysis by Fowler et al. (2009) examined 114 studies which tested relationships between varying forms of CVE and psychopathological symptoms. The authors found that direct victimization most strongly predicted symptoms of post-traumatic stress disorder (PTSD) and externalizing symptoms (e.g., delinquency, aggressive behavior). In contrast, witnessing violence most strongly predicted externalizing symptoms and both witnessing and “hearing about” violence predicted internalizing symptoms (e.g., depression, anxiety). Another meta-analysis by McDonald and Richmond (2008) examined 26 studies focusing on CVE in urban communities and found robust relationships with PTSD and aggressive symptoms among youth.

Most studies on CVE and youths’ psychological functioning to date have operationalized CVE using subjective measures, asking youth to report whether they have witnessed, heard about, or been directly exposed to violence. Only a small number of studies have used violent crime statistics to operationalize CVE and test its relationship to psychological functioning (Bingenheimer, Brennan, & Earls, 2005; Osypuk et al., 2012). In a study set in Bogotá, Colombia, Cuartas and Roy (2019) found that exposure to local homicides in adolescents’ neighborhoods significantly predicted poor mental health symptoms. Sharkey, Tirado-Strayer, Papachristos, and Raver (2012) examined the effect of local, recent homicides in Chicago, finding that greater numbers of homicides predicted deficits in impulse control and attention in preschoolers. Other research in Chicago found that fifth and sixth graders living in high-crime neighborhoods were faster to pay attention to emotionally negative stimuli compared with children living in lower-crime neighborhoods (McCoy, Roy, & Raver, 2016). Finally, one study found an association between zip-code level crime rates and adolescent behavioral health outcomes after adjusting for self-reported crime exposure (Grinshteyn, Xu, Mantueffel, & Ettner, 2018).

Why Explore the Spatial Dynamics of CVE?

With increased attention being paid to both indirect and direct measures of CVE, researchers have also begun to recognize the variability in what we mean by “exposure.” Bronfenbrenner’s bioecological model of development highlights that youth are embedded in, and develop within, intersecting social systems such as family, school, neighborhood, and culture (Bronfenbrenner & Morris, 2006). Therefore, to understand how CVE relates to youth development, research must embody a theoretical and empirical approach that addresses multiple settings of youth daily life (i.e., home and school neighborhoods). This perspective postulates that the three spatial dynamics of CVE (chronicity, pervasiveness, and spatial proximity) may be unique in how they affect individuals and communities (McCoy, 2013).

Firstly, CVE may vary in chronicity, or the repeated experience of CVE across specific periods of time (i.e., lifetime vs. 1 year) in space (where it occurs). Chronic CVE has been described as “frequent and continual exposure” to violence (Osofsky, 1999, p. 34). Previous research suggests that youth may become normalized to chronic and long-term CVE (Ng-Mak, Salzinger, Feldman, & Stueve, 2004). Several studies demonstrate that while CVE is linearly related to increases in aggressive behavior, youth report curvilinear relationships with psychological distress and depression, such that emotional duress increases initially but declines as CVE becomes more chronic (Gaylord-Harden, So, Bai, Henry, & Tolan, 2017; Gaylord-Harden, So, Bai, & Tolan, 2017; Kennady & Ceballo, 2016; Mrug, Loosier, & Windle, 2008; Ng-Mak et al., 2004). However, less research has addressed how indirect, chronic CVE (i.e., living in a high-crime neighborhood) relates to this process. Therefore, we conceptualize chronicity in terms of frequency of exposure to crime over varying time frames.

Secondly, youth may experience CVE differently across multiple neighborhood contexts, and as such, the accumulation and interaction of CVE in these spaces may be important. This highlights the second element of CVE, pervasiveness, such that individuals may encounter community violence in multiple environments, such as experiencing violence in both home and school neighborhoods. Context-specific exposure is particularly important in urban communities where youth are embedded in several neighborhood spaces. In Chicago, many youth do not attend schools in their immediate residential communities (Chicago Public Schools, 2017). Youth may find themselves traveling to neighborhoods where they feel more or less familiar and are exposed to different contextual supports (e.g., teachers, community centers) and threats (e.g., traversing gang territory). Mrug et al. (2008) found interactive effects when examining youth exposure to violence across three contexts—family, school, and neighborhood—such that youths’ psychological functioning was worse when they reported low neighborhood-based violence exposure but high home- and school-based violence. As such, it is possible that the youth do not become desensitized when they develop in contexts with incongruent levels of violence.

Finally, the bioecological model of development suggests that spatial proximity of exposure must be explored. Proximity of CVE is often conceptualized by levels of personal victimization, which vary from direct victimization (i.e., most proximal exposure), to witnessing and to simply “knowing about” or “hearing about” violence (i.e., most distal exposure). However, CVE may be spatially proximal (i.e., on the street outside one’s home) or more distal (i.e., two blocks away). This is particularly important given the spatial concentration of violence in Chicago (Morenoff, Sampson, & Raudenbush, 2001). In 2016, nearly half of the increase in Chicago’s homicides occurred in only five south and west-side neighborhoods (University of Chicago Crime Lab, 2017). Research indicates that spatial proximity to crime can be particularly detrimental for psychological outcomes; Weisburd et al. (2018) found that residents living in crime “hot spots” reported significantly more symptoms of depression and PTSD compared with residents living in spaces with less concentrated crime. However, the majority of youth CVE research has only examined how differentiating levels of personal victimization relate to functioning leaving the influence of spatial proximity of CVE largely unexamined.

The Current Study

The current study addresses gaps in the literature by exploring the relationships between CVE, operationalized using objective rates of violent crime and calculated across varying conceptualizations of space, time, context, and youths’ anxiety and self-regulation. A major strength of this study is the multidimensional way in which different measures of objective, indirect CVE are compared while adjusting analyses for subjective DVV. Grounded in a bioecological perspective, the use of objective, violent crime data, in addition to youths’ subjective, self-reports of DVV, recognizes that both have unique implications for youth development. That is, this study explores how living and attending schools in violent neighborhoods is related to adolescent psychological functioning beyond the influence of DVV alone. This study is the first of its kind to delineate and compare how each distinct spatial dynamic of CVE (chronicity, pervasiveness, and spatial proximity) relate to youths’ psychological outcomes. By comparing precise measures of CVE, this study tests which measures are most predictive across different domains of psychological functioning.

This study poses the question of, how do varying timeframes of chronicity, additive and interactive contexts of exposure, and ranges of spatial proximity relate to youths’ psychological functioning? We propose four hypotheses. First, we hypothesize that longer periods of CVE (i.e., more chronic exposure) will be most predictive of behavioral dysregulation; long-term chronicity may be a more accurate reflection of how often CVE occurs in youths’ neighborhoods and research on pathologic adaptation would suggest that while youth may not report increased emotional duress (i.e., anxiety), they report more maladaptive behavior (Ng-Mak et al., 2004). Likewise, we hypothesize that youth with more pervasive CVE (i.e., accumulation of home- and school-based CVE) will strongly predict behavioral dysregulation. Keeping with prior work examining violence across settings (Mrug et al., 2008), we also anticipate that the interaction between home- and school-based CVE may be predictive of youths’ anxiety outcomes because experiencing less violence in one context may protect against emotional desensitization. Third, we hypothesize that the most spatially proximal exposures (closest to the home and school) will be most strongly related to youth psychological functioning, based on theoretical assumptions that more proximal processes have a more direct relationship with development (Bronfenbrenner & Morris, 2006). Finally, we hypothesize that DVV will be detrimental for each psychological outcome, such that DVV will be related to worse psychological functioning.

Methods

Sample and Procedures

This study uses data from the Chicago School Readiness Project (CSRP), a longitudinal sample of low-income, predominately African American (65.8%) and Latinx (26.9%) Chicago youth assessed six times over the course of a decade (for additional details, see Raver et al., 2008). Two cohorts of children (N = 602, 53.3% female) and caregivers were recruited from Head Start Centers located in seven of Chicago’s most disadvantaged neighborhoods. These neighborhoods were located on Chicago’s south and west-sides and, to date, the majority of participating families continue to reside in these neighborhoods. Children and families were assessed when children were in preschool (Wave 1, N = 602), kindergarten (Wave 2, N = 398), third (Wave 3, N = 505), fifth (Wave 4, N = 491), ninth/tenth (Wave 5, N = 469), and tenth/eleventh (Wave 6, N = 437) grades. At waves 5 and 6, youth completed computerized assessments in their schools by a trained team of project staff. The analytic sample includes 314 youth who resided in Chicago and had completed assessments at wave 6, enabling their home and school address data to be geocoded and linked to violent crime data. Specifically, out of the 437 youths recruited at wave 6, 330 lived in Chicago and had addresses that could be geocoded and linked to crime data, and of this sample, 314 also had valid data on wave 6 psychological assessments. Only a small number of youth in the analytic subsample did not have covariate data at wave 5 (6%); therefore, we used single mean imputation for these cases of missing data. Additional analyses investigated demographic differences between the analytic subsample, the full wave 6 subsample, and the wave 5 subsample, finding no significant differences. The mean age for the analytic sample at wave 6 was 16.17 years old (SD = 0.79; see Table 1 for demographic characteristics).

Table 1.

Wave 6 analytic sample descriptives

Variable	N = 314
Gender	55% Female, 45% Male
Ethnicity	69% Black, 31% Non-Black
Family income-needs ratio	M = 0.96, SD = 0.82
Neighborhood poverty	M = 33%, SD = 13%
Chicago School Readiness Project (CSRP) treatment	48% Treatment, 52% Control
CSRP cohort	54% Cohort 1, 46% Cohort 2
Long-term chronic community violence exposure (CVE) (Sum)	M = 298.31, SD = 157.76
Mid-term chronic CVE (Sum)	M = 137.81, SD = 71.03
Short-term chronic CVE (Sum)	M = 5.76, SD = 3.91
Direct violent victimization (Sum)	M = 1.05, SD = 1
Trait anxiety (Mean)	M = 1.74, SD = 0.41
Behavioral dysregulation	M = 0.27, SD = 0.17
Cognitive dysregulation	M = 0.36, SD = 0.16

Measures

Community Violence Exposure

Crime statistics were obtained from the Chicago Data Portal (Chicago Data Portal, 2019a, 2019b). Data from this source are updated daily and include all incidents that involve police, even if an arrest was not made. Each crime report includes the date, time, location, type of crime, whether an arrest resulted or not, and whether it was considered a domestic incident. We downloaded all violent crime, including murder, rape, assault, robbery, and battery, that occurred during the 12 months prior to the exact date of the wave 6 assessment. These data were mapped and geocoded onto youth’s wave 6 home and school addresses using ArcGIS version 10.4.1 (Environmental Systems Research Institute, Inc. [Esri], Redlands, CA, USA). First, we drew spatial boundaries at one and two-block radii around each home and school address. An average block in Chicago is approximately 660 feet long (Chicago Department of Transportation, 2007); therefore, the one-block boundary was drawn with a radius of 660 feet around each home and school address and the two-block boundary was drawn with a radius of 1,320 feet. We also calculated exposure at the level of census tracts to compare the more proximal boundaries to a spatial range used more commonly in research. Geocoding techniques joined all violent crimes that occurred within the one-block, two-block, and census tract perimeters around home and school addresses to the respective boundaries. The count of all geocoded violent crime in each spatial boundary was divided by the number of days during that time frame to capture the temporal elements of each CVE measure.

In order to delineate the individual influence of home- and school-based CVE relative to their combined influence, spatial dynamic measures were calculated at the level of the school, the level of the home, and the combination of the two. When home- and school-based measures were combined, there was a small percentage of duplicates, in which the same crime was counted as occurring twice because it fell within both the home and school spatial boundaries. The percent of duplicate violent crimes ranged from 1.2% at the total one-block boundary, 2.9% at the total two-block boundary, and 2.4% at the total census tract boundary, revealing that the majority of youth did not attend schools in spatial proximity to their homes. We reduced these duplicates to a single crime exposure only in these combined measures so that crime frequencies were not inflated. This process recognized that a single exposure to a violent crime that occurred in two overlapping neighborhood contexts does not equate to “double exposure.”

Data were then aggregated to create distinct measures of CVE chronicity, pervasiveness, and spatial proximity. Chronic CVE was operationalized by quantifying the frequency of exposure across different periods of time in both home and school contexts. Long-term chronicity was calculated by taking total violent crime in the home and school two-block boundaries and dividing that by 365 days, exactly 1 year prior to the assessment date at wave 6. Mid-term chronicity follows the same calculation with a 6-month time frame, and short-term chronicity follows the same process for a 1-week time frame. Pervasive CVE compared the unique home and school context. Home-based CVE was calculated by taking the frequency of all violent crime in the previous year at the home two-block spatial boundary, 1 year prior to assessment at wave 6, whereas school-based CVE measures the 1-year frequency of violent crime within the school two-block boundary. Lastly, proximal CVE captured differing closeness to crime. Most distal CVE included the frequency of violent crime of both home and school census tracts, 1 year prior to assessment. Most proximal CVE measured frequency at one-block radii of both home and school addresses, 1 year prior to assessment (see Table 2 for a review).

Table 2.

Descriptions of each community violence exposure spatial dynamic measure

Measure	Neighborhood context	Spatial boundary	Timeframe
Long-term chronicity	Home and school	Two-block (1,320 ft.)	1 year
Mid-term chronicity	Home and school	Two-block (1,320 ft.)	6 months
Short-term chronicity	Home and school	Two-block (1,320 ft.)	1 week
Home-based	Home	Two-block (1,320 ft.)	1 year
School-based	School	Two-block (1,320 ft.)	1 year
Most distal	Home and school	Census tract	1 year
Most proximal	Home and school	One-block (660 ft.)	1 year

Self-Report Direct Violent Victimization

Direct violent victimization was assessed at wave 5 using three questions: In the last 6 months, have you been hit, kicked, or hurt by another kid? Have you been hit, kicked, or hurt by an adult? Have you been in a physical fight? Youth responded with either yes or no (yes = 1, no = 0), and a sum of these data was computed as a measure of DVV. Although this measure differs from other measures of subjective CVE in that it does not ask where the violence occurs, the focus on direct victimization allows us to partial out the relative contributions of direct victimization and indirect crime exposure on youths’ psychological functioning.

Trait Anxiety

At waves 5 and 6, anxiety was measured using the State-Trait Anxiety Inventory for Children (STAIC; Spielberger, Edwards, Montuori, & Lushene, 1973). Youth responded to 20 items that tapped stable, general levels of anxiety (α = .72). An example item states, “I worry about making mistakes… hardly ever, sometimes, or often.” Responses were scored on a Likert scale from 1 to 3. The items were aggregated into a mean score, with a higher value indicating worse trait anxiety.

Self-regulation

Self-regulation outcomes were computed by averaging items from two different measures of psychological functioning, the Barratt Impulsiveness Scale (BIS-11; Patton, Stanford, & Barratt, 1995) and the Behavior Rating Inventory of Executive Function (BRIEF; Gioia, Isquith, Guy, & Kenworthy, 2000). These assessments were taken at waves 5 and 6. Two subscales were calculated.

Behavioral dysregulation.

Behavioral dysregulation includes six items from the BIS and eight items from the BRIEF (α = .72). An example item from the BIS is “I act on impulses… Rarely/Never, Occasionally, Often, or Almost always/Always.” An example from the BRIEF is “I do not think of consequences before acting…. Never, Sometimes, or Often.” Responses across the two scales were standardized on a scale of 0–1 so that each item contributes equally to a given scale. The mean of all items was calculated. Higher scores indicate greater dysregulation.

Cognitive dysregulation.

Cognitive dysregulation was calculated using seven items from the BIS and seven items from the BRIEF (α = .72). An example item from the BIS is, “I plan things carefully… Rarely/Never, Occasionally, Often, or Almost always/Always” (reverse-coded). An example from the BREIF is “I have a short attention span… Never, Sometimes, or Often.” Items were standardized and averaged. A higher score indicates more dysregulation.

Covariates

Sample and study demographic characteristics were utilized as covariates. Youths’ race/ethnicity (Black = 1, Latinx/other = 0), gender (female = 1), CSRP study treatment (treatment = 1), and CSRP study cohort (cohort 1 or 2) were all collected at wave one. Due to the high correlation of participant age and CSRP cohort status (r = −.63, p < .001, Table 3), age was not included as a covariate. Families’ income-to-needs ratio (INR), calculated as the ratio of family income relative to the federal income standard, normed for family size where a ratio of 1 reflects the cutoff for living in poverty, was collected at each wave of data collection. We use the average INR across all waves for families who had at least four waves of valid data. Neighborhood poverty was calculated using data obtained from the 2012–2016 American Community Survey 5-year estimate and defined as the percent of all individuals living at or below the federal poverty line within youths’ wave 6 residential census tracts (U.S. Census Bureau, 2018). We found that the sample was highly mobile with 170 youth changing homes and 225 youth changing schools between waves five and six. Therefore, we included covariates for residential mobility and school mobility (1 = moved or changed schools). Importantly, all models also include parallel measures of psychological functioning, measured at wave five allowing us to examine the residual change in outcomes at wave six.

Table 3.

Pearson correlations among all variables of interest

	Treatment	Cohort	Female	Black	Neighborhood poverty	Family income	Age	DVV	TA	BD	CD	School-based
Treatment	1
Cohort	0.04	1
Female	0.08*	0	1
Black	0.17	0.55**	0	1
Neighborhood Poverty	−0.07	.26**	0	0.38**	1
Family Income	0.05	−0.06	0.02	−0.02	−0.12**	1
Age	0	−0.63**	0	−0.36**	−0.18**	0.11*	1
DVV	0.03	0.10*	0.05	0.13**	−0.02	0	−.11*	1
TA	−0.04	−0.02	−0.22**	−0.09+	−0.14*	0.07	−0.03	.20**	1
BD	0	0.07	0.11*	0.04	−0.07	−0.02	−0.16**	.28**	0.36**	1
CD	0.02	−0.05	0.01	−0.10*	−0.14*	−0.03	−0.05	.25**	.41**	0.06**	1
Long-term Chronic	−0.03	0.36**	−0.04	0.49**	.52**	−0.07	−.23**	0.11*	−0.05	0.05	−.09+	N/A
Mid-term Chronic	−0.01	.34**	−0.04	0.49**	.51**	−0.06	−.21**	0.11*	−0.06	0.05	−.09+	N/A
Short-term Chronic	−0.06	0.25**	−0.02	0.35**	.43**	−0.03	−.18**	0.08	−.09+	0	−0.08	N/A
Home-based	−0.02	0.33**	−0.04	0.45**	.57**	−0.09	−.23**	−0.02	−.12*	0.02	−.09+	0.34**
School-based	−0.05	0.27**	−0.01	0.38**	.27**	−0.07	−.19**	.21**	0.01	.11*	−0.03	1
Most Distal	0.05	0.288**	0	0.46**	.37**	−.1+	−.15**	0.05	−0.07	0.06	−.09+	N/A
Most Proximal	0	0.33**	0	0.45**	.49**	−0.05	−.21**	0.14*	−0.08	0.06	−0.03	N/A

Displays correlations between wave 6 CVE measures, wave 6 psychological outcomes, wave 5 DVV, and all other covariates. “N/A” is in place of some of CVE measures; these correlations are not appropriate because they represent spatial/temporal ranges that overlap with each other. It was of interest to test the correlation between home- and school-based CVE, which is included in the table.

BD, behavioral dysregulation; CD, cognitive dysregulation; CVE, community violence exposure; DVV, direct violent victimization; TA, trait anxiety.

⁺p < .1.

^*p < .05.

^**p < .01.

Analytic Plan

We employed residualized change, hierarchical multiple regression models to test the relationships between each aspect of CVE and outcomes. Given high correlations between predictors, each CVE spatial dynamic measure was analyzed separately with the exception of the pervasiveness models which tested for additive and interactive effects of home- and school-based CVE. Models tested the unique relationship between each CVE measure (calculated based on wave 6 assessment date) and each psychological outcome (collected at wave 6) while adjusting corresponding measure of psychological functioning at wave 5, DVV at wave 5, and demographic covariates. Given prior work demonstrating curvilinear relationships between CVE and psychological distress (Ng-Mak et al., 2004), we ran supplemental analyses with quadratic terms in models predicting anxiety, but no significant relationships were found. Results are available upon request from first author.

Results

This section summarizes significant findings on CVE and DVV measures, but all results are available from first author. Chronic CVE was significantly related to behavioral dysregulation (Table 4). Specifically, long-term chronicity significantly, positively predicted behavioral dysregulation (B = 0.07, SE = .02, p < .01, see Model 1). Interestingly, mid-term chronicity yielded a nearly identical relationship (B = 0.07, SE = .02, p < .01, Model 2). There was relatively no difference between the coefficients and predictive power across models, suggesting that long-term and mid-term chronicity are comparable metrics.

Table 4.

Community violence exposure (CVE) spatial dynamics and direct violent victimization (DVV) predicting behavioral and cognitive dysregulation

	Behavioral dysregulation							Cognitive dysregulation
	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	Model 14
Long-term	0.069**							0.023
	(0.025)							(0.023)
Mid-term		0.073**							0.015
		(0.027)							(0.025)
Short-term			0.015							−0.003
			(0.017)							(0.016)
Home-based				0.060	0.064+						0.036	0.038
				(0.039)	(0.039)						(0.035)	(0.036)
School-based				0.064+	0.072+						0.007	0.011
				(0.037)	(0.037)						(0.033)	(0.034)
Home x School					−0.171							−0.080
					(0.129)							(0.118)
Most Distal						0.050*							0.004
						(0.021)							(0.019)
Most Proximal							0.159+							0.127+
							(0.083)							(0.076)
DVV	0.028**	0.028**	0.031***	0.028**	0.029**	0.030***	0.028**	0.020*	0.020*	0.021**	0.020*	0.021*	0.021**	0.019*
	(0.009)	(0.009)	(0.009)	(0.009)	(0.009)	(0.009)	(0.009)	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)
R2	0.282	0.280	0.256	0.277	0.281	0.277	0.272	0.297	0.296	0.293	0.303	0.304	0.295	0.302
Adj. R2	0.255	0.254	0.228	0.247	0.249	0.250	0.245	0.272	0.270	0.267	0.274	0.273	0.269	0.276
Num. obs.	312	312	303	304	304	312	312	313	313	304	305	305	313	313
RMSE	0.141	0.141	0.144	0.143	0.142	0.142	0.142	0.131	0.131	0.132	0.130	0.130	0.131	0.130

Displays wave 6 CVE measures predicted wave 6 behavioral and cognitive dysregulation. Standard error in parentheses. All models are controlled for wave 5 behavioral and cognitive dysregulation, gender, race/ethnicity, family level income-to-needs ratio, neighborhood level poverty, Chicago School Readiness Project treatment, cohort, residential, and school mobilities.

⁺p < .1.

^*p < .05.

^**p < .01.

^***p < .001.

When considering the unique relationships between home- and school-based CVE, there was a significant interaction between home- and school-based CVE and trait anxiety (B = −0.69, SE = .33, p < .05, see Table 5, Model 19), visually depicted in Fig. 1. Graphing the data at the interquartile range revealed that when home- and school-based CVE were low, trait anxiety was also low. However, when home-based CVE was lowest, at the 25th percentile, and school-based CVE was highest, at the 75th percentile, trait anxiety was the highest (B = 0.27, SE = .13, p < .05). There were no significant additive effects of home- and school-based CVE on any psychological outcome.

Table 5.

Community violence exposure (CVE) spatial dynamics and direct violent victimization (DVV) predicting trait anxiety

	Trait anxiety
	Model 15	Model 16	Model 17	Model 18	Model 19	Model 20	Model 21
Long-term	−0.017 (0.064)
Mid-term		−0.031 (0.070)
Short-term			−0.066 (0.044)
Home-based				−0.095 (0.100)	−0.077 (0.100)
School-based				0.072 (0.095)	0.108 (0.096)
Home x School					−0.694* (0.333)
Most Distal						−0.018 (0.054)
Most Proximal							−0.191 (0.211)
DVV	0.081*** (0.022)	0.082*** (0.022)	0.082*** (0.022)	0.077*** (0.023)	0.080*** (0.023)	0.081*** (0.022)	0.084*** (0.022)
R2	0.263	0.263	0.266	0.268	0.279	0.263	0.265
Adj. R2	0.236	0.236	0.238	0.238	0.247	0.236	0.238
Num. obs.	309	309	300	301	301	309	309
RMSE	0.363	0.363	0.365	0.366	0.364	0.363	0.362

Displays wave 6 CVE measures predicted wave 6 anxiety. Standard error in parentheses. All models are controlled for wave 5 anxiety, gender, race/ethnicity, family level income-to-needs ratio, neighborhood level poverty, Chicago School Readiness Project treatment, cohort, residential and school mobilities.

⁺p < .1.

^*p < .05.

^**p < .01.

^***p < .001.

Fig. 1.

Delineated home- and school-based community violence exposure predicting trait anxiety.

When testing indicators of CVE proximity, we found the most distal measure of CVE, calculated at the level of the census tract to be positively related to behavioral dysregulation (B = 0.05, SE = .02, p < .05, Model 6), such that higher levels of census tract level CVE were related to greater dysregulation.

Direct violent victimization consistently, positively predicted all three outcomes with only slight variation in the predictive power of each parameter (Tables 4 and 5), such that higher levels of DVV were related to higher anxiety, behavioral dysregulation, and cognitive dysregulation.

Discussion

As one of the first studies to precisely measure and compare CVE spatial dynamics in terms of chronicity, pervasiveness, and spatial proximity, this study makes major contributions to our understanding of the relationship between CVE and adolescent psychological functioning. By also including a self-report measure of DVV in our models, we estimate these relationships independent of the influence of direct exposure to violence. Our findings reveal that specific spatial and temporal conceptualizations of CVE have unique implications for adolescent anxiety and self-regulation. Our indicators of CVE are most robustly related to youths’ behavioral dysregulation, and these relationships are strongest when exposure is characterized using the broadest parameters of time and two-block parameters of space. Additionally, anxiety was highest among youth living in low-crime neighborhoods and attending school in high-crime neighborhoods. We demonstrate that indirect exposure to violent crime occurring in home and school neighborhoods, even beyond direct victimization, is detrimental for psychological functioning.

Across all measures of psychological functioning, behavioral dysregulation had the strongest relationship with long-term chronic CVE, which aligns with previous research indicating that CVE increases behavioral maladaptation (Ng-Mak et al., 2004). While previous work has demonstrated that emotional desensitization mediates the relationship between CVE and behavioral maladaptation (Gaylord-Harden, So, Bai, & Tolan, 2017), it is unclear how behavioral regulatory abilities relate to this process. One study found that youth reporting high levels of CVE had poor self-regulatory behavioral skills and more aggressive beliefs in the following year (Goldweber, Bradshaw, Goodman, Monahan, & Cooley-Strickland, 2011). Therefore, it could be that behavioral dysregulation plays a moderating role in desensitization. Moreover, while behavioral dysregulation was robustly related to CVE, cognitive dysregulation was not highlighting how self-regulatory abilities may be context-dependent. This notion is understudied, but research has found that in highly disadvantaged communities, adolescent’s self-regulatory abilities can relate to specific environments and social contexts (Mason et al., 2010).

While most community violence research only measures violence occurring in the home or does not probe for specific neighborhood context, this study successfully delineated and compared CVE in home and school neighborhoods. This approach revealed that the cumulation of home- and school-based CVE predicted behavioral dysregulation while CVE in one setting did not. In comparison, trait anxiety was highest when adolescents lived in neighborhoods with lower CVE than where they went to school. These findings indicate that only examining CVE occurring within the residential neighborhood may over-look the full impact that CVE can have on youth. Future research should assess both home- and school-based CVE to better understand how experiences across these important contexts influence development.

Why did youth living in lower-crime neighborhoods while attending school in higher crime neighborhoods have higher anxiety? It is possible that when a student must travel to a dangerous neighborhood to attend school, their anxiety increases because they are aware that they are comparatively less safe than they are at home. One study in Philadelphia found that adolescents perceived less safety when traveling to school through high-crime areas (Wiebe et al., 2013) while Cuartas and Roy (2019) found that perceived safety moderated the relationship between homicide exposure and youth psychological functioning, such that youth reported less psychological dysfunction when they perceived their neighborhoods to be safe. Therefore, anxiety may increase because youth are aware that they are in, or traveling through, a relatively less safe space.

Keeping with the bioecological model of development (Bronfenbrenner & Morris, 2006), it was hypothesized that the most proximal measure of CVE would produce the most robust relationship with psychological impairment. However, we found that it was the second most proximal boundary, the two-block aggregate that was most predictive of behavioral dysregulation. This distinction may be a function of how adolescents inhabit neighborhood space. A two-block distance is still quite close to the home or school, likely encompassing neighborhood space that youth physically, habitually navigate, but also allowing for youths’ independence that a one-block boundary may fail to capture. Other research has found that youth spend active and leisure time in spatial ranges comparable to the two-block boundary (Chambers et al., 2017; Colabianchi et al., 2014). Unsurprisingly, many researchers note that census tracts do not accurately reflect how adolescents inhabit and develop within neighborhood space (Chambers et al., 2017; Colabianchi et al., 2014; Morenoff et al., 2001). Future research investigating other spatially linked neighborhood attributes should draw spatial boundaries carefully to reflect habitual youth behavior and the urban landscape.

In keeping with existing literature (Fowler et al., 2009), we found that DVV robustly predicted all the outcomes we examined. However, it is important to note that long-term chronic, CVE had a relationship with behavioral dysregulation that was comparable to that found with DVV. This highlights how simply living and attending schools in highly violent neighborhoods is just as critical as subjective, direct experiences of violence for behavioral self-regulation. This is important because far more youth live and attend schools in violent neighborhoods than youth who are directly victimized. Furthermore, given the importance of self-regulation for many long-term outcomes (McCoy, 2013), there is the potential for a cascade of unfortunate negative consequences for adolescents who develop in chronically violent neighborhoods. Future research must continue to assess these indices separately to understand how CVE shapes development.

Implications and Limitations

This study has important implications for the operationalization and measurement of CVE and highlights some of the complexities that researchers must consider when doing geospatial research. We found no difference in how long- or mid-term chronic CVE were related to behavioral dysregulation; therefore, studies employing a similar geospatial methodology could measure chronic CVE at either 1-year or 6-month timeframes. Furthermore, future research should continue to examine both home and school neighborhoods separately and in combination to test how each neighborhood setting influences youth development. The two-block spatial boundary had the strongest relationship with behavioral dysregulation, demonstrating that the commonly used census tract may not be the best boundary for determining exposure.

We suggest that this study is replicated in locations beyond Chicago and with nonparametric statistical methods. Chicago experienced a surge in gun violence and homicides during the time frame of this study (University of Chicago Crime Lab, 2017); therefore, it is possible that the sudden increase in violent crime helped produce these findings. Additional research replicating this study in other communities with more stable or lower rates of violent crime would address this consideration. Because of neighborhood interdependence, or the possibility that what happens in one neighborhood can influence other neighborhoods (Morenoff et al., 2001), parametric testing may not be appropriate for this line of inquiry. Diagnostic tests revealed that there was no variance inflation that would have significantly violated parametric assumptions, but previous research has successfully conducted spatial dynamic research using first-order Markov models and Hierarchical Bayesian estimation which do not require an assumption of independence (Verbitsky & Raudenbush, 2009). This research found that the first-order Markov, Bayesian approach was the best method to estimate the parameters in comparison with other approaches. Other research utilizes nested, hierarchical models in spatial dynamic research, which was not possible with the geospatial techniques used in this research design.

Other limitations come at the expense of one of this study’s major strengths and relate to the usage of crime to operationalize indirect CVE. Firstly, it must be noted that crime statistics are not bias free due to under-reporting of crime, particularly in Black communities affected by police violence and misconduct (Desmond, Papachristos, & Kirk, 2018). It is also possible that the youth in this study were directly exposed to the crimes in their neighborhoods, which we are not capturing with our current measures. However, because all crime data used for this study were deidentified, it is not possible to know whether youths were directly victimized by these crime events. Secondly, while the use of precise, geocoded data allows one to examine spatial dynamics of CVE, it does not capture youths’ perceptions of violence. It is imperative to consider that youth may live in a high-crime neighborhood but may not themselves consider it “dangerous.” Indeed, qualitative work has found that youth living in high-crime communities report their neighborhoods to be both “safe” and “unsafe” (Teitelman et al., 2010).

Conclusion

These findings assert that all youth living in high violent crime neighborhoods need mental health support, highlighting a need for universal prevention strategies. Practitioners working with youth who reside in high-crime communities should be aware that even if youth were not directly exposed to violence, they may be in need of mental health supports. Moreover, mental health services must be prioritized in neighborhoods with more chronic, spatially clustered violent crime and should incorporate ways to improve self-regulatory abilities and reduce anxiety. School-based mental health support services have been found to be particularly effective at buffering against the negative consequences of CVE (Gaias, Johnson, White, Pettigrew, & Dumka, 2019) and may be prime locations to administer universal prevention initiatives (Dodington et al., 2012). Furthermore, given the import of school neighborhoods, we encourage policy makers to promote programs such as Safe Passage, which places adults on routes that youth frequently take to schools and parks (Curran, 2018; McMillen, Sarmiento-Barbieri, & Singh, 2019). Research indicates that this program may potentially reduce crime (Curran, 2018; McMillen et al., 2019); therefore, the results of this study suggest that such programs may subsequently improve the mental health of youth who traverse these neighborhoods.

Highlights.

• Tested community violence spatial dynamics (chronicity, pervasiveness, proximity) and youth outcomes

• Behavioral dysregulation is related to home- and school-based chronic community violence.

• Trait anxiety is uniquely predicted by the interaction of home- and school-based community violence.

• One of the few studies to objectively measure community violence, comparing home and school settings