Racial and Ethnic Disparities in Violent Crime

A Closer Look at the Neighborhood Context

by Logan Folger

Photo by Patrick Schreiber on Unsplash
Photo by Patrick Schreiber on Unsplash

This study tests two hypotheses derived from the racial invariance thesis in criminology: (1) Structural features of neighborhoods have common criminogenic effects across race and ethnicity and (2) Controlling for structural features of neighborhoods should explain racial and ethnic disparities in neighborhood violent crime. In short, race and ethnic disparities in violence reflect the relative exposure of groups to criminogenic influences in the neighborhoods in which they are embedded. Using recent data from the Atlanta Police Department and the American Community Survey (ACS) on Atlanta (n = 323), results show that higher violent crime rates in Black and Mixed neighborhoods relative to their White counterparts are fully explained by an index of concentrated disadvantage. Findings also show that disadvantage increases violent crime similarly in Black, Mixed, and White neighborhoods. Consistent with the invariance thesis, racial and ethnic disparities in violence are not a product of unique differences but rather of structural disadvantages within neighborhoods.

Keywords: racial disparities, concentrated disadvantage, racial invariance thesis, violent crime


Numerous studies document racial and ethnic disparities in violent crime. Light and Harris (2012) examined the violent crime rate in 1,315 counties in the United States (U.S.) and found that neighborhoods with majority Black residents had a much higher violent index rate (306.33) compared to their White counterparts (70.95). Hannon, Knapp, and Defina (2005) examined Black and White homicide rates and found that the homicide rate for the Black population is 40.99, which was nearly five times that of the White population (8.76). Despite recent declines in violent crime nationally, racial disparities in violent victimization and offending remain pronounced. 

Early schools of thought argued that people of color are more involved in violent behavior compared to their White counterparts because they are genetically predisposed to crime-prone behavior. Others argue that distinctive cultural values, in opposition to mainstream values, may account for group differences in violent crime (Wilcox, Cullen, & Feldmeyer, 2018). There is a notable lack of empirical support for both of these ideas. More recently, attention has shifted to neighborhood structures and social contexts with disproportionate group representations as an explanation for racial/ethnic disparities in violence (Sampson, Wilson, & Katz, 2018).

Sampson and Wilson (1995) emphasize the impact of macrosocial patterns of residential inequality by race/ethnicity in the United States and its effect on the social dislocation and ecological concentration of the truly disadvantaged, especially African Americans. Indeed, even decades later, it is near impossible to reproduce in White neighborhoods the same circumstances under which many Black Americans live in the U.S., especially the historical and contemporary impact of prolonged exposure to racial discrimination and disadvantage across generations. The argument implies that if Whites were disproportionately embedded in neighborhoods with concentrated structural disadvantage, they would display behavioral and cultural adaptations similar to that documented in poor, predominantly Black neighborhoods (Sampson, Wilson, & Katz, 2018). This is actually an old idea stemming from Clifford Shaw and Henry McKay’s (1942) theory of social disorganization (Wilcox, Cullen, & Feldmeyer, 2018). 

Shaw and McKay’s (1942) social disorganization theory is among the first to argue for a contextual interpretation of the racial disparity in crime and violence. Two particular findings from their extensive delinquency research in Chicago in the early to mid-twentieth century led them to this inference: (1) Crime rates within groups varied according to community context, and (2) Despite neighborhoods’ racial/ethnic composition, the same ones maintained high crime rates over time (Shaw & McKay, 1969). They thus conclude that personal (e.g., biological) and cultural (e.g., distinctive values) characteristics of race/ethnic/nationality groups are not factors underlying crime; rather, neighborhood structures are.

Recent research on the racial invariance thesis has been quite mixed with respect to whether structural disadvantages have similar effects on violence across neighborhoods and therefore could explain racial/ethnic disparities in neighborhood violent crime (Hernandez, Velez, & Lyon, 2018; Laurence, 2015; Sampson, Wilson, & Katz, 2018). Indeed, while studies report significant reductions in race/ethnic disparities in violence with controls for structural disadvantage, few fully explain those disparities. In addition, there have been inconsistent results with respect to whether structural disadvantage measures increase violence similarly across race/ethnic neighborhoods. Some have taken such findings as evidence against the invariance thesis (e.g., Unnever, 2018), leading to calls for theories that address the unique experiences of specific groups, especially Black people (see also Unnever, Barnes, and Cullen 2016). Yet, Sampson et al. (2018) argue that partial elimination of race/ethnic differences does not necessarily invalidate the invariance thesis. In particular, partial explanation may reflect the unmeasured influence of disadvantages deriving from structural and historic racism that have impeded the accumulation of wealth in the Black community. Sampson et al. (2018) also argue that evidence of identical effects of disadvantage measures on violence rates across race/ethnic equations or neighborhoods is not necessary to validate the invariance thesis. Structural disadvantages, however, should increase violence similarly across race/ethnic neighborhoods.

This study draws on very recent data (circa 2017) for neighborhoods (i.e., Census block-groups) in the city of Atlanta, Georgia (n = 323). Drawing on data from the Atlanta Police Department and the American Community Survey five year estimates, I distinguish between non-Hispanic White, non-Hispanic Black, and racially/ethnically mixed neighborhoods. The former are defined as greater than 70% non-Hispanic White and Black, respectively; the remainder are defined as race/ethnic mixed. Two hypotheses are tested with Ordinary Least Squares (OLS) regression: (1) Concentrated disadvantage has common criminogenic effects on violent crime rates in White, Black, and Mixed neighborhoods; and (2) Controlling for concentrated disadvantage should explain racial and ethnic disparities in neighborhood violent crime (Hernandez, Velez, & Lyon, 2018). Specifically, I regress a total violent crime index, comprising murder, robbery, and aggravated assault rates per 1,000 population (averaged over 2015-17), on indicators of concentrated disadvantage and residential stability. To establish causal order, the latter are lagged by a time period using the ASC estimates for 2011-2015. I assess the extent to which race/ethnic disparities in neighborhood violent crime are accounted for, and derive the average effect of concentrated disadvantage across the neighborhood types as means to assess whether it increases violence similarly in White, Black, and Mixed neighborhoods. In short, I test whether race/ethnic disparities in violence reflect the relative exposure of groups to criminogenic neighborhood structures.

Literature Review

Shaw and McKay’s social disorganization theory grew out of the Chicago school research in the early 20th century on the adaptation of immigrant populations to an “industrial, large, heterogeneous, and transient” space (Wilcox, Cullen, & Feldmeyer, 2018, p. 16). Chicago’s city life was a recipe for social disorganization as there was high community disadvantage and residential turnover, especially in neighborhoods in “transition zones” near the city center. Immigrant groups would eventually assimilate into the American cultural and economic structure, however, and would “reorganize and move to new communities, where they would once again flourish” (Wilcox, Cullen, & Feldmeyer, 2018, p. 17). In addition, poor, unstable neighborhoods maintained high crime rates, irrespective of racial/ethnic composition. These observations led Shaw and McKay (1942) to conclude that people were not criminogenic but were simply living in socially disorganized areas. Shaw and McKay also specified that the causes of crime are racially invariant, and therefore when disadvantage differences are accounted for, racial/ethnic gaps in violence should be minimized (Hernandez, Velez, & Lyon, 2018).

McNulty and Bellair (2003) examine racial/ethnic differences in serious adolescent violence with a contextual model that considers both neighborhood and individual level factors. Consistent with previous research, their data show that racial/ethnic minorities reside in substantially more disadvantaged neighborhoods, reflected in heightened involvement in violence among Black, Hispanic, and Native American adolescents compared to their White counterparts. Consistent with the invariance thesis, findings show that the Black-White disparity in serious adolescent violence is fully explained by a concentrated disadvantage index. Remaining race/ethnic disparities are explained by correlates or outcomes of growing up in disadvantaged neighborhoods (e.g., gang involvement, exposure to violence, weaker bonds).

Hernandez, Velez, and Lyon (2018) examine racial disparities in violent crime among White, Black, and Hispanic neighborhoods (i.e., Census tracts). They create a concentrated disadvantage index comprising the poverty rate, joblessness, low-wage jobs, professional jobs (this was reverse coded), high school graduates, and female-headed households. Of the 11 tract-level comparisons of violent crime rates in Black and White neighborhoods, 7 are found to be non-significant once concentrated disadvantage is controlled. In addition, concentrated neighborhood disadvantage and residential instability heighten violent crime similarly in Black, Hispanic, and White neighborhoods, although effects appear somewhat more pronounced in White neighborhoods. Overall, after comparing the magnitude and direction of coefficients across the different neighborhoods, Hernandez et al. (2018) find that 80% of the neighborhoods had non-significant comparisons. Additionally, after adjusting for the difference in disadvantage distributions, the level of non-significance went up by 5% (85% total). This study provides support for the racial invariance thesis that race/ethnic disparities in neighborhood violent crime largely reflect the structure of urban inequality. 

Not all studies, however, have been supportive of the racial invariance thesis. James Laurence (2015) tested the thesis using communities in the United Kingdom, specifically London boroughs. He argues that London boroughs provide more overlap in disadvantage across race/ethnicity than in the United States and hence a more definitive test of the thesis. In his analysis, Laurence (2015) find that at equal levels of extreme disadvantage, disadvantage had a significantly larger effect on violent crime in Black neighborhoods than White neighborhoods. At similar levels of low and high disadvantage, however, the effect of disadvantage on violent crime was similar across Black, White, and South Asian boroughs. These findings partially contradict the racial invariance thesis because the theory argues that structural disadvantage should have a consistently similar effect across neighborhood types. 

Steffensmeier, Ulmer, Feldmeyer, and Harris (2010) consider issues that may arise in testing the racial invariance thesis and the range of its scope. Steffensmeier et al. (2010) assess the scope of the thesis by using homicide rates as well as an overall violent crime index as dependent variables and extending the analysis to White, Black, and Hispanic comparisons. They draw two conclusions from their findings: First, their results are partially consistent with the thesis in that structural disadvantage helps explain a portion of violent crime rate differences across racial/ethnicity. Second, they find that although the disadvantage index has a positive relationship with the total violent crime index across the White, Black, and Hispanic equations, the magnitude of the effects often varied. The effects of disadvantage on homicide rates, however, displayed more racial/ethnic invariance than variance. Indeed, only four of the 15 homicide models produced significant (or near significant) differences in the effects of disadvantage across the White, Black, and Hispanic equations. These findings provide some support for the racial invariance thesis, but they conclude that it may yet be provisional and have more limited scope (e.g., homicide) than originally conceived.

Data & Methods


Data for this study are derived from the Atlanta police department and the U.S. Census Bureau’s American Community Survey (ACS). The units of analysis are Census block groups, widely used to indicate neighborhoods in the urban and criminological literature. The Atlanta police department data comprise 330 block groups within the city of Atlanta. Six block groups were excluded from the study because, according to the ACS, they have no population (e.g., park or airport). Another block group was excluded due to its potential to skew the results; this is a predominately Black neighborhood with a small population, low concentrated disadvantage, yet very high violent crime rate. The findings are more conservative without it with respect to the race and ethnic comparisons. Thus, the final multivariate analysis includes 323 block groups within the city of Atlanta. Table 1 displays descriptive statistics for variables included in the multivariate analysis disaggregated by neighborhood race/ethnic composition. The table reveals that non-Hispanic Black and Mixed neighborhoods, on average, appear quite disadvantaged compared to their non-Hispanic White counterparts. 

Violent Crime

Crime data were collected through the Atlanta police department. Rates of murder, aggravated assault, and robbery were combined into a total violent crime index per 1,000 population. To ensure the calculation of reliable rates for small census block groups and to minimize the impact of annual fluctuations, violent crime rates were averaged over three years covering 2015-2017. The total violent crime index does show some skew (skewness = 1.5). Models with a logged violent crime index were generated to try to reduce skew; however, the results were substantially the same as the results presented below with the unlogged outcome. Table 1 shows that Black (12.52) and Mixed (6.79) neighborhoods have significantly (p < .001) higher violence rates than their White counterparts (2.54). 

Independent Variables 

To ensure causal order, independent variables are lagged by a time period and are assessed at the block-level with the ACS five-year estimates. Neighborhood racial composition is determined based on the 2013-2017 ACS five-year estimates. Black (N=151) and White (N=83) neighborhoods are defined as comprising greater than 70% non-Hispanic Black and White, respectively. The remaining neighborhoods are defined as racially/ethnically mixed (N=89). These cutoffs are consistent with previous research and clearly define predominantly Black and White neighborhoods in Atlanta. For example, the mean percent Black in predominately Black neighborhoods is 92%, whereas the mean percent White in White neighborhoods is 85%. For Mixed neighborhoods, the mean percent Black and White is 35% and 47%, respectively, with 8% Hispanics and 6% Asians making up the rest. Dummy variables distinguish the race/ethnic neighborhoods types, with non-Hispanic White neighborhoods serving as the referent.

Concentrated disadvantage data is derived from the ACS five-year estimates for 2011-2015. The scale includes three items: (1) the percent of residents below the poverty threshold, (2) the percent of families headed by females with children less than 18 years of age, and (3) the percent of civilian noninstitutionalized males (16 years and older) who are unemployed or not in the labor force. The three items were standardized (i.e., z-scored) and then averaged. Higher scores indicate a more disadvantaged neighborhood. Consistent with the literature, concentrated disadvantage is highly prevalent in Black and to a lesser extent mixed neighborhoods compared to their White counterparts (shown in Table 1). There is also less variation on the disadvantage index among White neighborhoods compared to Black and Mixed neighborhoods. 

Residential stability is a standardized measure averaging the percentage of neighborhood residents residing in the same house for at least one year and the percentage of owner-occupied housing units. Residential stability is more problematic in Black and especially Mixed neighborhoods relative to White neighborhoods (shown in Table 1). 


A weighted least-squares (WLS) regression procedure was used to counter the heteroscedasticity in the models, which violates the assumption that random regression errors have a constant variance over all observations (i.e., homoscedastic). With a WLS regression, the effects are weighted by block group population size. This reduces the heteroscedasticity in the models by reducing the undue influence of block groups with small populations, resulting in more efficient estimates. Multicollinearity is not a problem in the analysis as indicated by Variance Inflation Factors (VIFs) that are well below critical thresholds. A descriptive overview of Atlanta is given, followed by WLS regression models that test the central hypotheses of the invariance thesis that the relative exposure to structural causes of violence explains the higher violent crime rates evident in Black and Mixed neighborhoods compared to their White counterparts.

Results & Analysis

Data were analyzed using Stata 16. Descriptive statistics for the relevant variables are shown in table 1. The violent crime rate, disadvantage index, and residential stability index are all significantly different between the three neighborhood types. Black neighborhoods experience substantially more violent crime than White or Mixed neighborhoods, but this is expected because the concentrated disadvantage index is significantly higher in Black neighborhoods. This similar pattern can be seen in Mixed neighborhoods as well. Another important variable to consider is residential stability. Compared to White neighborhoods, instability appears most problematic in Mixed neighborhoods.

Table 1: Descriptive Statistics for Variables, Atlanta

Violent Crime Rate (per 1,000)2.54*^~(3.13)12.52(9.37)6.79(7.86)8.31(8.78)
Concentrated Disadvantage-1.07*^~(0.28).76(.77)-.22(.72).00(1.00)
Residential Stability0.04*^~(.75)-.67(.70)-1.03(.95).00(1.00)
N of Neighborhoods8315189323

Notes: Weighted means displayed. Standard deviations in parentheses. *^~ Bonferroni multiple comparison tests of mean differences by neighborhood type:

*White-Black neighborhoods differ p < .01
^White-Mixed neighborhoods differ p < .01
~Black-Mixed neighborhoods differ p < .01

Table 2 shows the regression of violent crime in Atlanta neighborhoods considering racial composition and social disorganization variables. Non-Hispanic White neighborhoods are used as the referent. Looking at racial composition alone in the first model, Black (b = 9.99) and Mixed (b = 4.25) neighborhoods evidence significantly (p < .001) higher violent crime rates relative to White neighborhoods. However, consistent with the invariance thesis, when concentrated disadvantage is accounted for in the second model, the magnitude of the Black and Mixed coefficients are reduced to non-significance. Concentrated disadvantage is a significant factor in explaining the variation in violent crime in Atlanta neighborhoods (p < .001). The adjusted R-squared increased from .208 to .320, meaning that the model is explaining 32% of the variance in the violent crime rate. For every unit increase in the concentrated disadvantage index, the violence crime rate increases by 4.16 violent crimes/1000. Most importantly, concentrated disadvantage fully explains racial/ethnic disparities in neighborhood violent crime. When residential stability is added to the equation in model 3, the race/ethnic disparities remain non-significant and, although in a negative direction, stability does not reach significance once concentrated disadvantage is already controlled for. 

Table 2: WLS Regressions of Violent Crime 2015-17, Atlanta 

Racial/Ethnic Composition^
Social Disorganization Variables
—Concentrated Disadvantage4.16***(.57)3.90***(.61)
—Residential Stability -.62(.55)
Adjusted R-Squared.208.320.321

Notes: *(p < .05); **(p < .01); ***(p < .001). N = 323. Standard Errors in parentheses.

^White neighborhoods are the Referent. 

Table 3 shows the average marginal effects of concentrated disadvantage in Mixed, White, and Black neighborhoods. The average marginal effects were significant (p < .001) in the positive direction for both Mixed (b = 3.15) and Black (b = 4.53) neighborhoods but not White (b = .74, ns) neighborhoods. Perhaps this is due to the lack of variability on the disadvantage index among White neighborhoods. As Table 1 shows, the standard deviation for the disadvantage index (.28) and violent crime rate (3.13) for White neighborhoods are both smaller indicating considerably less variability than in Black neighborhoods (.77 & 9.37, respectively). 

Table 3: Average Marginal Effects of Concentrated Disadvantage

AME95% Confidence Interval
Racial/Ethnic Composition
Mixed3.15***(.97)[1.23, 5.06]
White.74(2.80)[-4.77, 6.25]
Black4.53***(.760)[3.03, 6.03]

Notes: *** (p < .001). N = 323. Standard Errors in parentheses. 

Discussion and Conclusion

Racial disparities in crime and violent behavior have been well documented in the United States. However, explanations for the racial differences are still the subject of debate and research. There is a racial difference in both criminal activities and crime victimization, and both are more likely to be committed by or happen to African Americans than Caucasians (Sampson & Lauritsen, 1997). One of the most prominent explanations for the racial disparity in violence is concentrated disadvantage and its influence on family. The racial invariance thesis argues that there are structural features in neighborhoods that have common criminogenic effects across race and ethnicity, and that when these features are controlled for, the racial and ethnic disparities in violence should disappear (Hernandez, Velez, & Lyon, 2018).

This study shows that concentrated disadvantage has a positive relationship with violence in all three types of neighborhoods (but not significant for White neighborhoods). The effect is more pronounced in Black and Mixed neighborhoods as indicated by the average marginal effects of concentrated disadvantage. This could be due to the disproportionate distribution of disadvantage, especially in Black neighborhoods. This indicates that concentrated disadvantage is part of the reason why there is a disparity in violence rates between White, Black, as well as Mixed neighborhoods. This study provides support for the racial invariance thesis in that neighborhood concentrated disadvantage fully explains the race/ethnic disparities in neighborhood violence. Lastly, though residential stability was a supposed protective factor that could also provide some explanation, the results indicate that the effect of residential stability on violence rates is not significant when disadvantage is controlled.

Further research could be done on public policies and how that could break down the barriers that arise from concentrated disadvantage. Racial/ethnic disparity in violence does not only come from within the neighborhood context but also from the public policies that foster and reinforce the disadvantaged conditions of neighborhoods. For example, government housing is often overcrowded and lacks security supervision. It is easy to take advantage of neighborhoods like this because the tenants lack the resources to fight crime, and their voices are often ignored. These residents are already disadvantaged in some ways because they are living on government support. It is imperative to start making changes within the system and educate people on structural disadvantages and why crime is so concentrated in some areas more than others. 

Terms and PhrasesDefinition
Social disorganization theory Socially disorganized neighborhoods tend toward weak informal social controls and hence high crime rates. High poverty, unemployment, etc. undermine neighborhood collective efficacy.
Multivariate analysisAnalysis that involves multiple statistical outcome variables
SkewnessThe variable’s distribution departs from a normal distribution
Logging to reduce skewnessLog transformations can help reduce skew in the outcome, resulting in less biased estimates.  
HeteroscedasticityPrediction errors do not have constant variance over observations
Homoscedasticity Prediction errors have equal variances over observations.
Weighted-Least SquaresA regression technique used to account for the heteroscedasticity in the data
MulticollinearityOccurs when there is high correlation between two or more predictor variables in a regression.
Variance Inflation FactorMeasure of the impact of multicollinearity
Average Marginal EffectsThe average effects of disadvantage across race/ethnic neighborhood types. 


Hannon, L., Knapp, P., & Defina, R. (2005). Racial similarity in the relationship between poverty and homicide rates: Comparing retransformed coefficients. Social Science Research34(4), 893–914.

Hernandez, A. A., Velez, M. B., & Lyons, C. J. (2018). The racial invariance thesis and neighborhood crime: Beyond the Black-White divide. Race and Justice, 8(3), 216–243. 

Laurence, J. (2015). Community disadvantage and race-specific rates of violent crime: An investigation into the “racial invariance” hypothesis in the United Kingdom. Deviant Behavior36(12), 974–995.

Light, M. T., & Harris, C. T. (2012). Race, space, and violence: Exploring spatial dependence in structural covariates of White and Black violent crime in US Counties. Journal of Quantitative Criminology28(4), 559–586.

McNulty, T. L., & Bellair, P. E. (2003). Explaining racial and ethnic differences in adolescent violence: Structural disadvantage, family well-being, and social capital. Justice Quarterly,20(1), 1-31. 

Sampson, R. J., & Lauritsen, J. L. (1997). Racial and ethnic disparities in crime and criminal justice in the United States. Crime and Justice21, 311–374.

Sampson, R. J., & Wilson, W. J. (1995). Toward a theory of race, crime, and urban inequality. In J. Hagan & R. D. Peterson (Eds.), Crime and Inequality (pp. 37-54). Stanford, CA: Stanford University Press.

Sampson, R. J., Wilson, W. J., & Katz, H. (2018). Reassessing “Toward a Theory of Race, Crime, and Urban Inequality”: Enduring and new challenges in 21st century America. Du Bois Review: Social Science Research on Race15(1), 13–34. 

Shaw, C. R., & McKay, H. D. (1942). Juvenile delinquency and urban areas: A study of rates of delinquents in relation to differential characteristics of local communities in American cities. Chicago, IL: University of Chicago Press.

Shaw, C. R., & McKay, H. D. (1969). Juvenile delinquency and urban areas. Chicago, IL: University of Chicago Press.

Steffensmeier, D., Ulmer, J. T., Feldmeyer, B., & Harris, C. T. (2010). Scope and conceptual issues in testing the race-crime invariance thesis: Black, White, and Hispanic comparisons. Criminology48(4), 1133–1169.

Unnever, J. D. (2018). The racial invariance thesis in criminology: Toward a Black criminology. In J. D. Unnever, S. L. Gabbidon, & C. Chouchy (Eds.) Building a Black Criminology: Race, Theory, and Crime (Advance in Criminological Theory, Vol 24., p. 77-100). New York, NY: Routledge. 

Unnever, J. D., Barnes, J. C., & Cullen, F. T. (2016). The racial invariance thesis revisited: Testing an African American theory of offending. Journal of Contemporary Criminal Justice, 32(1), 7–26. 

Wilcox, P., Cullen, F. T., & Feldmeyer, B. (2018). Communities and crime: An enduring American challenge. Philadelphia, PA: Temple University Press.

Acknowledgements: I would like to thank Dr. Thomas McNulty for providing me with valuable feedback and guidance in the writing and editing processes of this paper.

Citation Style: APA