Neighborhood Racial Composition and Unequal Exposure to Violent Crime in Everyday Contexts

Abstract

Exposure to violence is a critical aspect of contemporary racial inequality in the United States. While extensive research has examined variations in violent crime rates across neighborhoods, less attention has been given to understanding individuals’ everyday exposure to violent crimes. This study investigates patterns of exposure to violent crimes among neighborhood residents using cell phone mobility data and violent crime reports from Chicago. The analysis reveals a positive association between the proportion of Black residents in a neighborhood and the level of exposure to violent crimes experienced by residents. Controlling for a neighborhood’s level of residential disadvantage and other neighborhood characteristics did not substantially diminish the relationship between racial composition and exposure to violent crimes in everyday life. Even after controlling for violence within residents’ neighborhoods, individuals residing in Black neighborhoods continue to experience significantly higher levels of violence in their day-to-day contexts compared to those living in White neighborhoods. This suggests that racial segregation in everyday exposures, rather than residential segregation, plays a central role in racial inequality in exposure to violence. Additionally, the analysis suggests that neighborhoods with more Hispanic and Asian residents are exposed to less and more violent crime, respectively, compared to neighborhoods with more White residents. However, this is only observed when not adjusting for the volume of visits points of interest receive; otherwise, the finding is reversed. This study offers valuable insights into potentially novel sources of racial disparities in exposure to violent crimes in everyday contexts, highlighting the need for further investigation.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11524-024-00868-6.

Introduction

Exposure to violence negatively affects health and is strongly linked to elevated cortisol and self-reported stress levels [1, 2]. It increases the risk of adverse mental health outcomes, including post-traumatic stress disorder, depression, and suicidal ideation [3] and is associated with adverse birth outcomes [4]. Violence impacts not only direct victims but also bystanders. Many Americans die from homicides, while others suffer severe physical injuries, ultimately resulting in costs that run into the billions [5].

In the American context, violence has been framed as central to understanding contemporary racial inequality. Violence tends to be more prevalent in poor Black neighborhoods, adversely impacting the life chances of youth who grow up in these neighborhoods [6]. However, this perspective is incomplete and fails to consider exposure to violence that occurs in everyday contexts or even racial heterogeneity within “safe” neighborhoods. Thus, although it is well-documented where violence takes place, it is also important to understand who is exposed to violence and how those patterns of exposure diverge from what would be expected based on residential context alone.

Neighborhood socioeconomic disadvantage has been considered focal for understanding neighborhood violence [7, 8]. Neighborhood exposure to violence is a central mechanism through which neighborhood inequality shapes the life chances of adolescents in the United States [9]. Neighborhood effects explain a large share of racial differences in adolescent outcomes, highlighting the importance of neighborhoods in the American social stratification process [10]. Neighborhood exposure to violent events, such as police brutality, shootings, and homicides, can cause substantial detriment to adolescents’ academic performance, health, and mental well-being [11–14].

Understanding the impact of violence on adolescents requires an examination of racial disparities in how exposure to violence is patterned. Previous research has shown that Black Americans are more likely to be exposed to violence [15], and urban sociological theories exploring the relationship between race, neighborhood, and crime [16] have highlighted racial segregation and economic conditions as the central drivers of adverse neighborhood outcomes, including violence. Massey’s [17] theory of segregation argues that changes in Black neighborhoods’ socioeconomic conditions have led to significant increases in crime rates. Sampson’s social disorder and collective efficacy theories suggest that a lack of informal social control and stable social ties in poor, non-White neighborhoods contributes to high rates of violence [18, 19]. Additionally, recent research suggests that White Americans are much more likely to move out of and less likely to move into neighborhoods with high or increasing crime rates, which may contribute to racially unequal exposure to neighborhood violence [20].

Historically, neighborhoods have been considered rigid constructs studied in terms of residential processes. However, more recently, data on everyday mobility patterns has substantially expanded how neighborhoods are studied. Studies show that mobility-based measures of disadvantage, which consider people’s out-of-neighborhood exposure in their daily lives, are better predictors of various neighborhood outcomes compared to traditional residential measures [21]. Mobility patterns can better predict several neighborhood outcomes, such as violence, infectious diseases, and birth weight [8, 21, 22].

Nascent research indicates that the types of neighborhoods that people tend to visit vary substantially. Studies have shown that demographically similar neighborhoods have stronger neighborhood ties [23, 24]. While some studies emphasize the role of spatial proximity, some nationwide studies have considered racial similarity a key driver of neighborhood ties [25]. This suggests that understanding a neighborhood’s racial composition and mobility patterns in conjunction is critical for predicting neighborhood outcomes and understanding the perpetuation of residential segregation.

Racial homophily in neighborhood mobility patterns may especially influence individual outcomes. Residential racial segregation and its effects on neighborhoods have been studied extensively [25]. Neighborhoods can determine people’s access to resources, whereas segregation perpetuates the context in which people feel comfortable and have access to, thereby limiting future opportunities to reside in other neighborhoods [26, 27]. As Black neighborhood residents tend to spend much of their lives in Black neighborhoods, it is reasonable to hypothesize that the cumulative disadvantages they experience in their lifetime tend to become concentrated in these neighborhoods.

Athey and colleagues’ [28] analysis of racial segregation in different everyday contexts showed that people experience less segregation in everyday contexts than they experience residential segregation. Browning et al. [29] investigated youths’ everyday experiences of neighborhood racial compositions and found that their home neighborhood’s racial composition is not reproduced in their visited neighborhoods. However, adolescents living in Black neighborhoods spend much more out-of-neighborhood time in other Black neighborhoods compared with those living in other neighborhoods [29]. Browning et al. [29] also found that Black youth tend to be disproportionately exposed to neighborhoods with high violent crime rates.

This study analyzes the relationship between neighborhood racial composition and everyday exposure to violent crimes in the city of Chicago. While most past research has measured exposure to violence through residential assumptions or self-reported data on exposure, this study uses cell phone mobility data to estimate people’s actual exposure to violence in everyday contexts. This study additionally decomposes racial disparities in everyday exposure to violence in terms of between-neighborhood differences, within-neighborhood differences, and between-time differences. Finally, this study highlights the potential for further research to further develop a better understanding of the precise individual-level and neighborhood-level mechanisms that shape everyday exposure to violence.

Methods

Data Collection and Variables

Data on mobility patterns consists of monthly data on the number of visits made by residents (devices) of census block groups in Chicago to all points of interest (PIs) in Chicago. Geocoded crime data comes from publicly available records published by the Chicago Police Department and was categorized in line with past research [30]. Census block group demographic and socioeconomic information comes from the 2015–2019 American Community Survey 5-year estimates, and variables of interest were selected based on past scholarship on violence [31]. For more information on the data involved in this study, see the Supplementary Information.

80,420 reported violent crimes were identified from the crime dataset in 2019. 41,587 PIs were identified as being located in the city of Chicago. PIs in which a violent crime had been reported within a 100-m radius at least once in 2019 were specifically identified. Consequently, 36,737 unique PIs in Chicago were involved in this analysis, where 52,747 unique violent crimes had occurred within a 100-m radius in 2019.

Next, the total number of visits by residents of neighborhood i to all PIs was estimated as follows:

$$V_i=\sum _{j=1}^J\sum _{w=1}^W{v}_{\mathit{ijw}}$$

where $v_{\mathit{ijw}}$ is the number of visits residents of neighborhood i made to PI j in week w.

A secondary analysis also considered the possibility that PIs that generally attract more visitors from residents of different neighborhoods may thus be predisposed to experiencing more violent crimes [32]. However, the average visitor may be less likely to be a victim, witness, or bystander exposed to a violent crime. Based on this assumption, the total number of weighted visits by residents of neighborhood i to all PIs were estimated as follows:

$$V_i=\sum _{j=1}^J\sum _{w=1}^W{v}_{\mathit{ijw}}/\sum _{i=1}^I{v}_{\mathit{ijw}}$$

where $v_{\mathit{ijw}}$ represents the number of visits by neighborhood-i residents to PI j in week w. Earlier, the same number of visits was estimated using a second approach, restricting them to unique sets of PI and weeks, accounting for one or more violent crimes reported within 100 m of PI j during week w.

Analysis Procedure

Individuals’ level of exposure to violent crimes in a given neighborhood were estimated by considering the number of visits to PIs in which a violent crime occurred within its 100-m radius during the same week. The primary racial variables used for the estimation included the proportion of non-Hispanic Black and Hispanic (of any race) individuals and individuals of other races, with non-Hispanic White individuals serving as the omitted reference group.

The main model can be expressed as follows:

$$\text{ln}\left({\mu }_i\right)={\theta }_1\ast {\text{BLACK}}_i+{\theta }_2\ast {\text{HISP}}_i{+\theta }_3\ast {\text{ASIAN}}_i{+\theta }_4\ast {\text{OTHER}}_i+{\theta }_c\ast c_i+\text{ln}({\text{VISITS}}_i)+\epsilon$$ 4

where ${\mu }_i$ represents the number of visits by neighborhood i residents to PIs that had experienced a violent crime in the same week, and ${\text{BLACK}}_i$, ${\text{HISP}}_i$, ${\text{ASIAN}}_i$, and ${\text{OTHER}}_i$ represent the proportion of non-Hispanic Black, Hispanic, non-Hispanic Asian, and “other” (non-White, non-Black, non-Asian, and non-Hispanic) residents of neighborhood i, respectively. $c_i$ represents the vector of controls for neighborhood i, and ${\text{VISITS}}_i$ represents the total number of visits made to PIs by neighborhood i residents.

Furthermore, the exposure to violence was estimated by adjusting for the total number of visitors to a PI, because individuals from specific neighborhoods may disproportionately visit PIs that attract many visitors and, consequently, are more likely to experience a violent crime.

Finally, as a robustness check, all prior analyses were also performed by operationalizing violent crimes strictly as homicides. This practice conforms with a well-established perspective in criminology research that police-reported homicides are less subject to sample and selection biases. However, unlike other violent crimes, some neighborhoods in Chicago had no observed exposure to homicides. To accommodate this, half the minimum non-zero exposure value was added to all observations, as zero values cannot fit into the logged model. Table 1 presents the summary statistics and Table S1 in the Appendix presents the correlation matrix.

Table 1.

Summary statistics

Variable	N	Mean	Std. Dev	Min	Pctl. 25	Pctl. 75	Max
Violent crime exposure	2157	2729	2394	202	1514	3250	46,483
Total visits	2157	13,772	9688	1206	8480	16,404	18,5114
Weighted violent crime exposure	2157	1.4	0.94	0.08	0.78	1.7	15
Weighted total visits	2157	8.5	4.5	0.67	5.5	10	58
Homicide exposure	2157	9	11	0	0	12	76
Weighted homicide exposure	2157	0.0062	0.02	0	0	0.0069	0.59
Residential disadvantage	2157	0.13	1.2	-2.5	-0.76	1	3.3
Prop. White	2157	0.32	0.31	0	0.023	0.61	1
Prop. Black	2157	0.34	0.4	0	0.011	0.84	1
Prop. Hispanic	2157	0.26	0.3	0	0.029	0.43	1
Prop. Asian	2157	0.055	0.1	0	0	0.066	0.97
Prop. other	2157	0.021	0.032	0	0	0.031	0.27
Violent crimes per capita	2157	0.035	0.039	0	0.0097	0.048	0.51
Logged population density	2157	9.7	0.79	5.9	9.2	10	14
Median age	2157	37	8.3	16	31	41	86
Prop. young males	2157	0.16	0.074	0	0.11	0.2	0.63
Prop. owner occupied	2157	0.48	0.24	0	0.3	0.64	1
Prop. long-term residents	2157	0.26	0.16	0	0.13	0.36	0.83

Finally, a decomposition method was used to decompose the differences in exposure to violence between White and Black neighborhoods. First, differences in exposure to violence was decomposed into differences in visit rates to certain neighborhoods and PIs using the following formulas:

$$D_1=\sum \left(C_i^B,-,C_i^W\right)\ast \frac{M_i^B\ast G_i^B+M_i^W\ast G_i^W}{2}$$

$$D_2=\sum \left(M_i^B,-,M_i^W\right)\ast \frac{C_i^B\ast G_i^B+C_i^W\ast G_i^W}{2}$$

$$D_3=\sum \left(G_i^B,-,G_i^W\right)\ast \frac{M_i^B\ast C_i^B+M_i^W\ast C_i^W}{2}$$ 5

where (1) $C_i^B$ and $C_i^W$, ${\left(2\right)M}_i^B$ and $M_i^W,\mathrm{and}(3)$ $G_i^B$ and $G_i^W$ are vectors of the proportion of visits by Black- and White-neighborhood residents to (1) all neighborhoods in Chicago; (2) all PIs in Chicago, as a fraction of the total visits to the PIs’ neighborhood; and (3) certain PIs in different weeks of 2019, as a fraction of the total visits to that PI, respectively.

Results

Main Analysis

Table 2 presents the results of the first measure of exposure to violent crimes by neighborhood. Model 1 includes only an offset of the total number of visits made to all PIs and provides a baseline for model fit. Model 2 includes proportions Black, Hispanic, and other. Proportion White constitutes the omitted reference category. The model results indicate that the proportion of Black residents in a neighborhood is positively and significantly (p < 0.001) associated with exposure to violent crimes, whereas surprisingly, the opposite is true for the proportion of Hispanic residents (p < 0.001). This finding is surprising given that Hispanic neighborhoods, on average, experience more violent crimes than White neighborhoods do. The model results also indicate that the proportion of Asian residents in a neighborhood is positively and significantly (p < 0.001) associated with exposure to violent crimes. This finding is notable, as past research has historically found that Asian Americans are exposed to less violence than other racial groups [33]. Finally, the proportion-other coefficient suggests that the proportion of “other” residents in a neighborhood is positive and insignificantly associated with exposure to violent crimes. These results suggest that a 100% Black neighborhood experiences 46% more exposure to violent crimes in everyday contexts than a 100% White neighborhood. Distinctly, a 100% Hispanic, 100% Asian, and a 100% “other” neighborhood experience 10% less, 34% more, and 4% more exposure to violent crimes in everyday contexts compared to a 100% White neighborhood, respectively.

Table 2.

Predicting violent crime exposure by neighborhood

	Model 1	Model 2	Model 3	Model 4	Model 5
Prop. Black		0.38***	0.26***	0.38***	0.41***
		(0.01)	(0.01)	(0.02)	(0.02)
Prop. Hispanic		− 0.10***	− 0.12***	0.03	0.04*
		(0.02)	(0.02)	(0.02)	(0.02)
Prop. Asian		0.29***	0.28***	0.36***	0.18***
		(0.04)	(0.04)	(0.04)	(0.03)
Prop. other		0.04	0.08	0.05	− 0.19
		(0.13)	(0.12)	(0.12)	(0.10)
RND				− 0.05***	− 0.05***
				(0.01)	(0.01)
Violent crime rate			1.75***	1.97***	1.91***
			(0.12)	(0.12)	(0.11)
Logged population density					0.09***
					(0.00)
Median age					0.00*
					(0.00)
Prop. young males					0.13**
					(0.05)
Prop. owner occupied					− 0.18***
					(0.02)
Prop. long-term residents					− 0.12***
					(0.03)
N	2157	2157	2157	2157	2157
AIC	− 18.60	− 1320.61	− 1512.96	− 1579.32	− 2501.84
BIC	− 7.25	− 1286.55	− 1473.22	− 1533.91	− 2428.05
Adj. R2 Adj. R2	0.00	0.45	0.50	0.52	0.69

^***p < 0.001; **p < 0.01; *p < 0.05

Model 3 includes a control term for the number of per-capita violent crimes in a neighborhood. Past research has noted that violent crime is patterned unequally across neighborhoods in terms of racial composition. Consequently, racial inequalities in neighborhood exposure to violent crimes likely exist, as individuals often visit PIs that are in or near their neighborhoods. If substantial racial neighborhood inequalities in the exposure to violent crimes in everyday contexts persisted even after controlling for the neighborhood violent crime rate, this would suggest the presence of a novel pathway through which exposure to violent crimes is racially unequal. Specifically, this finding would suggest that individuals of different races are not simply unequally exposed to violent crimes because their neighborhoods have different violent crime rates. Instead, this unequal exposure can also be attributed to spending time in places where violent crimes occur at unequal rates, even if these places are outside their own neighborhoods.

Interestingly, the racial coefficients in Model 3 are not substantially different from those in Model 1. The coefficient for proportion Black is the most substantially attenuated, falling to 0.26 from 0.38. However, the significance of the coefficient remains unchanged (p < 0.001). Considering exposure net of a neighborhood’s violent crime rate, a 100% Black neighborhood is exposed to 30% more violent crimes in everyday contexts than a 100% White neighborhood. The proportion-Hispanic, proportion-Asian, and proportion-other coefficients remain mostly unchanged from Models 2 to 3, shifting to − 0.12 from − 0.10, 0.28 from 0.29, and 0.08 from 0.04, respectively. The coefficient of the neighborhood violent crime rate was positive and highly significant (p < 0.001).

Model 4 incorporates a measure of residential disadvantages, a major predictor of neighborhood violent crime rates [8]. Mobility-based measures of residential disadvantage also strongly predict a neighborhood’s violent crime rate, suggesting that a neighborhood’s measure of residential disadvantage may explain its level of exposure to everyday violent crimes. Model 4 aimed to test whether residential disadvantage could explain racial inequalities in exposure to violent crimes. Previous research has found that residential disadvantages can explain racial inequalities in neighborhood violent crime rates; however, it may not necessarily explain residents’ exposure to everyday violent crimes.

The racial coefficients in Model 4 are substantially different from those in Model 3 but not in the expected direction. Instead, the coefficients for proportions Black, Hispanic, Asian, and other are skewed more positively compared with those in Model 3. The coefficients for proportions of Black, Hispanic, Asian, and other are now 0.38 (previously 0.26), 0.03 (insignificant, previously − 0.12), 0.36 (previously 0.28), and 0.05 (previously 0.08), respectively. Although it would make sense to expect the coefficient to be positive, as the baseline correlation coefficient between residential disadvantage and violent crime exposure (offsetting for total visits) is weakly positive (0.21), the residential disadvantage coefficient in Model 4 was negative. The coefficient of − 0.05 suggests that a one-standard-deviation increase in residential disadvantage is associated with 5% less exposure to violent crimes. This strange result can likely be explained by the strong multicollinearity of racial composition and residential disadvantage and the surprisingly weak predictive power of residential disadvantage. Ultimately, the model’s results suggest that residential disadvantages cannot explain racial neighborhood disparities in exposure to violence in everyday contexts.

Finally, Model 5 incorporates a set of controls for other neighborhood characteristics that could be associated with violent crime. This set of controls includes logged population density, median age, and proportions of young men, owner-occupied households, and long-term residents. The racial-composition coefficients of Model 5 closely resemble those of Model 4, but are more positively skewed. As the controls failed to substantially attenuate the racial composition coefficients, other clear neighborhood characteristics do not explain the observed racial inequalities. The control variables exhibit expected associations with violence exposure: population density, median age, and more young males are positively associated with violence exposure, while more owner-occupied households and long-term residents are negatively associated, aligning with past research [8].

Table 3 presents the results of the same models as Table 2, but using a weighted version of visits, accounting for the total number of visitations in PIs. The results are relatively similar to those of Table 2, except for the significant increase and decrease in the proportion-Hispanic and proportion-Asian coefficients, respectively. In Models 2, 3, and 5, the proportion-Hispanic coefficient changed from negative and significant in Table 2 to positive and insignificant in Table 3, while the proportion-Asian coefficient changed from positive and significant in Table 2 to negative and insignificant in Table 3. This finding is relatively more logical, as past research has documented that Hispanic neighborhoods are generally more exposed to violent crime than White neighborhoods. This significant shift in coefficients suggests that Hispanic neighborhoods’ residents are less exposed to violent crimes in everyday contexts and Asian neighborhoods’ residents are exposed to more (compared to White neighborhoods’ residents) only because the PIs they visit tend to attract different volumes of visitors and, subsequently, tend to experience different rates of violent crimes. While it remains unclear which measure of violence exposure (visit-adjusted or visit-unadjusted) most meaningfully reflects the degree to which individuals are harmed by exposure to violence, the findings of Table 3 clarify the race-specific patterns observed in Table 2.

Table 3.

Predicting violent crime exposure by neighborhood (weight-adjusted)

	Model 1	Model 2	Model 3	Model 4	Model 5
Prop. Black		2.69***	2.29***	1.77***	1.76***
		(0.13)	(0.15)	(0.21)	(0.22)
Prop. Hispanic		1.72***	1.66***	1.01***	1.00***
		(0.17)	(0.17)	(0.25)	(0.25)
Prop. Asian		− 0.07	− 0.11	− 0.46	− 0.59
		(0.43)	(0.43)	(0.44)	(0.44)
Prop. other		− 0.84	− 0.73	− 0.62	− 0.77
		(1.25)	(1.24)	(1.24)	(1.24)
RND				0.21***	0.19**
				(0.06)	(0.07)
Violent crime rate			5.84***	4.89***	5.93***
			(1.25)	(1.27)	(1.38)
Logged population density					0.18**
					(0.06)
Median age					0.00
					(0.01)
Prop. young males					− 0.52
					(0.62)
Prop. owner occupied					− 0.16
					(0.26)
Prop. long-term residents					0.38
					(0.37)
N	2157	2157	2157	2157	2157
AIC	9051.24	8516.67	8496.83	8486.93	8484.53
BIC	9062.60	8550.72	8536.57	8532.35	8558.32
Adj. R2	0.00	0.22	0.23	0.23	0.23

^***p < 0.001; **p < 0.01; *p < 0.05

Tables 4 and 5 replicate the findings of Tables 2 and 3 with homicide exposure. The results of these models align with those of Tables 2 and 3, with extremely stark racial disparities. For instance, Model 1 results in Tables 4 and 5 suggest that residents of a 100% Black neighborhood are 394% and 1474% more exposed to homicides in everyday contexts compared to residents of a 100% White neighborhood, respectively.

Table 4.

Predicting homicide exposure by neighborhood

	Model 1	Model 2	Model 3	Model 4	Model 5
Prop. Black		1.37***	1.22***	0.77***	0.72***
		(0.06)	(0.07)	(0.09)	(0.09)
Prop. Hispanic		0.75***	0.72***	0.16	0.19
		(0.07)	(0.07)	(0.11)	(0.11)
Prop. Asian		− 0.79***	− 0.80***	− 1.10***	− 1.02***
		(0.19)	(0.19)	(0.19)	(0.19)
Prop. other		− 0.76	− 0.72	− 0.63	− 0.46
		(0.54)	(0.54)	(0.54)	(0.54)
RND				0.18***	0.16***
				(0.03)	(0.03)
Violent crime rate			2.27***	1.46**	1.93**
			(0.55)	(0.55)	(0.60)
Logged population density					0.01
					(0.03)
Median age					0.00
					(0.00)
Prop. young males					− 0.62*
					(0.27)
Prop. owner occupied					− 0.07
					(0.11)
Prop. long-term residents					0.40*
					(0.16)
N	2157	2157	2157	2157	2157
AIC	5765.23	4947.77	4932.54	4888.42	4873.34
BIC	5776.58	4981.83	4972.27	4933.83	4947.13
Adj. R2	0.00	0.32	0.32	0.34	0.34

^***p < 0.001; **p < 0.01; *p < 0.05

Table 5.

Predicting homicide exposure by neighborhood (weight-adjusted)

	Model 1	Model 2	Model 3	Model 4	Model 5
Prop. Black		2.69***	2.29***	1.77***	1.76***
		(0.13)	(0.15)	(0.21)	(0.22)
Prop. Hispanic		1.72***	1.66***	1.01***	1.00***
		(0.17)	(0.17)	(0.25)	(0.25)
Prop. Asian		− 0.07	− 0.11	− 0.46	− 0.59
		(0.43)	(0.43)	(0.44)	(0.44)
Prop. other		− 0.84	− 0.73	− 0.62	− 0.77
		(1.25)	(1.24)	(1.24)	(1.24)
RND				0.21***	0.19**
				(0.06)	(0.07)
Violent crime rate			5.84***	4.89***	5.93***
			(1.25)	(1.27)	(1.38)
Logged population density					0.18**
					(0.06)
Median age					0.00
					(0.01)
Prop. young males					− 0.52
					(0.62)
Prop. owner occupied					− 0.16
					(0.26)
Prop. long-term tesidents					0.38
					(0.37)
N	2157	2157	2157	2157	2157
AIC	9051.24	8516.67	8496.83	8486.93	8484.53
BIC	9062.60	8550.72	8536.57	8532.35	8558.32
Adj. R2	0.00	0.22	0.23	0.23	0.23

^***p < 0.001; **p < 0.01; *p < 0.05

Other notable findings in Tables 4 and 5 are that proportion-Hispanic and residential disadvantage coefficients are significantly positive across most models. This finding aligns more closely with past research, suggesting that violent crimes may be especially underreported in places visited by residents of Hispanic and disadvantaged neighborhoods. While homicide is generally considered a measure of violence that is subject to less measurement error compared to violent crime, there is no way of determining if neighborhood disparities in the ratio of homicide to violent crime are attributable to measurement error or if they represent real differences in violence.

Decomposition

The results of the decomposition shed light on the sources of racial disparities in exposure to violence in everyday contexts. First, within Black neighborhoods, operationalized as neighborhoods where more than 50% of the population is Black, the average rate of exposure to violence in everyday contexts is 23.4%. Thus, 23.4% of the visits made by residents of Black neighborhoods were to PIs where a violent crime occurred during the same week. In White neighborhoods, operationalized as neighborhoods where more than 50% of the population is White, the average rate of exposure to violence in everyday contexts was 19.6%.

The decomposition analysis reveals that a gap of 2.5 absolute percentage points (or 65.1% of the entire gap) can be explained by differences in residents’ neighborhood visit patterns. Thus, if Black-neighborhood residents visited the same set of neighborhoods at the same rate as White-neighborhood residents did, their exposure to violence would be expected to decrease from 23.4 to 20.9%. Furthermore, 1.3 absolute percentage points (or 32.5% of the entire gap) can be explained by differences in residents’ within-neighborhood visit rates to different PIs. Therefore, despite visiting the neighborhoods at the same rate, if Black-neighborhood residents visited PIs within those neighborhoods at the same rate as White-neighborhood residents, their exposure to violence would be expected to decrease from 23.4 to 22.1%. Finally, the gap of 0.1 absolute percent (or 2.4% of the entire gap) could be explained by the differences in the residents’ time-specific visit rates at different PIs. Thus, despite visiting neighborhoods and PIs at the same rate, if Black-neighborhood residents visited these PIs at the same time as White-neighborhood residents, their exposure to violence would be expected to decrease from 23.4 to 23.3%.

Ultimately, the analysis results suggest that racial disparities in exposure to violence are primarily driven by differences in neighborhood destinations. However, neighborhoods alone do not explain the racial disparity; within-neighborhood differences in visits to PIs also play a substantial role. Segregation has traditionally been studied between neighborhoods and not within them; however, this study suggests that within-neighborhood racial segregation can exacerbate inequalities in exposure to violence.

Discussion

This study analyzed the relationship between neighborhood racial composition and exposure to violent crimes in everyday contexts. The results showed that the proportion of Black and Asian residents in a neighborhood was positively and significantly associated with exposure to violent crime, whereas the proportion of Hispanic residents was negatively or not associated. Controls for the neighborhood’s own violent crime rate, residential disadvantage, and other potential confounding variables failed to substantially attenuate this relationship. Controlling for the volume of visits PIs receive reversed the specific findings for neighborhood proportions Hispanic and Asian. Thus, proportion-Hispanic was positively associated with violence exposure, and proportion-Asian was negatively associated. Overall, the results suggest that racial inequalities in exposure to everyday violence cannot be explained solely by other neighborhood characteristics. This suggests that racially unequal exposures in everyday life are a central pathway through which exposure to violence persists unequally, and this pattern exists beyond residential neighborhood characteristics.

Using a three-way decomposition, the gap in exposure to violence between Black and White neighborhoods was examined in terms of differences in neighborhood visit patterns, within-neighborhood visit rates to different PIs, and time-specific visit rates to different PIs. Notably, while differences in the neighborhoods visited by residents of Black and White neighborhoods explain the majority of the gap, nearly one-third of the gap can be attributed to differences in the PIs visited within the same neighborhoods. While the role of between-neighborhood segregation is frequently implicated as central to unequal exposures, this finding highlights the potentially underestimated role of smaller-scale segregation in shaping inequalities in exposure to violence.

A central limitation of this study is that the actual exposure to violent crimes was not perfectly observed but was estimated. Consequently, the true exposure to violent crimes was substantially overestimated, as individuals’ visits to PIs do not last for an entire week. Regardless, this approach provides a reasonable estimate of how exposure to everyday violence varies by neighborhood. Another limitation is that the timescale of this analysis was relatively coarse. One week is a long period; thus, this study’s time-based decomposition of exposure to violence is likely not accurate. However, time-based decomposition does add more nuance while explaining the racial disparity in exposure to violence in daily life.

Nevertheless, this research sheds substantial light on the nature of racial inequality in the exposure to violent crimes. The results reveal that even after controlling for within-neighborhood violence, residents of Black neighborhoods are exposed to disproportionality more violence in everyday contexts than residents of White neighborhoods. This suggests a novel pathway through which violence is racially unequal. While past research has emphasized residential neighborhood inequalities and residential racial segregation in explaining racial disparities in exposure to violence, this research indicates that racial differences in everyday exposures are crucial. Furthermore, spending time in different neighborhoods is not the sole source of this inequality; rather, spending time in different places within neighborhoods also plays a substantial role, accounting for a third of the disparity between Black and White neighborhoods.

The findings also suggest that, regarding exposure to violence in everyday contexts, residents of neighborhoods with more Asian residents but not more Hispanic residents face more exposure than residents of neighborhoods with more non-Hispanic White residents. These findings are inconsistent with past research—that has documented that Hispanic Americans, but not Asian Americans, face greater exposure to violence [33]. While conditioning on the volume of visits to PIs appears to explain these findings and looking at homicide produces different results, future research should explore whether visit-adjusted or visit-unadjusted measures of violence exposure most meaningfully reflect the degree to which individuals are exposed to the harmful effects of violence. Future research exploring contexts other than Chicago could further clarify how exactly neighborhood racial composition is associated with exposure to violence. It remains unclear whether the descriptive racial patterns observed in Chicago could be found in other cities. Generally, these findings are notable, as exposure to violence is highly detrimental to the health and development of adolescents and adults.

This study provides a valuable descriptive portrait of racial disparities in the exposure to violent crimes in everyday contexts. However, there is considerable scope for further research to build on this. One important area for future research is to investigate the timescale of exposure to violence and its association with racial disparities. Additionally, research could explore individual-level rather than neighborhood-level factors of violence exposure to better understand the causes of inequalities in exposure to violence in everyday life. By identifying these driving forces, future research could potentially inform the development of effective interventions and policies to reduce overall exposure to violence in addition to specific racial disparities.

Supplementary Information

Below is the link to the electronic supplementary material.