Racial and Ethnic Disparities in Structural Disadvantage and Crime: White, Black, and Hispanic Comparisons

Abstract

Objectives

The objective of this study is to advance knowledge on racial/ethnic disparities in violence and the structural sources of those disparities. We do so by extending scarce and limited research exploring the relationship between race/ethnic gaps in disadvantage and differences in violent crime across groups.

Methods

Using census place-level data from California and New York, we construct White, Black, and Hispanic “gap” measures that take as a given the existence of disparities across race/ethnic groups in structural disadvantage and crime and subsequently utilize seemingly unrelated regression models to assess the extent to which gaps in disadvantage are predictive of gaps in homicide and index violence.

Results

Our results suggest that (1) there is considerable heterogeneity in the size of White-Black, White-Hispanic, and Black-Hispanic gaps in structural disadvantage and crime and (2) that race/ethnic disparities in structural disadvantage, particularly poverty and female headship, are positively associated with race/ethnic gaps in homicide and index violence.

Conclusion

In light of recent scholarship on the racial invariance hypothesis and on the relationship between structural inequality and crime, the current study demonstrates that disparities in disadvantage, particularly family structure and poverty, are important in driving racial and ethnic disparities in crime.

1. Introduction

Violent crime, especially homicide, is a burden that falls more heavily on African American (and to a lesser extent) Hispanic than White communities. The most common explanations of the relatively high levels of violent crime among Blacks and Hispanics focus on the effects of poverty and unemployment, educational inequality, residential segregation, social disorganization, subcultural adaptations to disadvantage, and the legacy of racism and discrimination on behavior (Steffensmeier, Ulmer, Feldmeyer, and Harris, 2010; Kubrin and Weitzer, 2003). Besides lacking economic resources, Blacks and other minorities are subject to relatively greater social isolation and resource deprivation (Blau and Blau, 1982; Hipp and Yates 2011; Lee, 2000; Parker and McCall, 1999; Tittle, 1995) and some observers suggest that these conditions, in turn, help to spawn subcultural patterns, such as a “Code of the Streets” which fosters violence, that distinguish minority from White communities in particular (Anderson, 1999; Harer and Steffensmeier, 1992).

Recent research has made substantial advances in identifying the extent to which race/ethnic-specific rates of (violent) crime, particularly homicide, are driven by structural disadvantage and other key macro-structural traits. However, there remain gaps in the empirical literature on racial/ethnic disparities in crime and the sources of those disparities. As several reviews show (e.g., Steffensmeier et al., 2010; Ousey, 2000; Parker, 2008; Shihadeh and Shrum, 2004), the degree to which differences across groups in structural disadvantage predict racial or ethnic differences in violence is far from settled. For one, most studies are limited to Black-White comparisons and little is known about how differences in violent crime between Whites, Blacks, and especially Hispanics are driven by disparities in the structural circumstances of these groups (for recent exceptions, see Martinez, Stowell, and Lee, 2010; Wadsworth, 2010; Xie 2010; Cancino, Martinez, and Stowell, 2009; Jones-Webb and Wall, 2008). This oversight is concerning in light of the proposed “Latino paradox,” in which Hispanic populations experience surprisingly low levels of violence in the face of extreme disadvantage levels (Feldmeyer 2009; Martinez, 2003; Sampson, 2008; Steffensmeier et al. 2011), and evidence that Hispanic levels of violence are more similar to White levels, even controlling for disadvantage (e.g., Krivo, Peterson, and Kuhl, 2009; Martinez et al. 2010; Steffensmeier et al. 2010; but see Shihadeh and Barranco 2010b). Second, the preponderance of research to date has focused on homicide – a small part of the broader criminal landscape – and questions remain about the relationship between racial/ethnic disparities in disadvantage and gaps in violent crime more generally.

The fact that predominantly White localities rarely approach the levels of disadvantage found in predominantly Black (and to a lesser extent, Hispanic) areas makes it difficult to compare the race- and ethnic-specific effects of structural disadvantage on race- and ethnic-specific levels of violence (Feldmeyer 2009; McNulty, 2001; Phillips, 2002; Velez, Krivo, and Peterson, 2003). Yet, considerable variation between places exists in the size of racial/ethnic differences in levels of structural disadvantage and in racial/ethnic differences in violence rates. That is, the size of the gap in violence (and disadvantage) between Whites, Blacks, and Hispanics varies across geographic units. As such, gaps in disadvantage might explain the relative size of violence gaps (see Peterson and Krivo, 2005). Treating the race-ethnic gap in violence itself as a dependent variable (Velez et al. 2003) renders the lack of overlapping structural conditions among race and ethnic groups analytically less important by taking disparity in violence and structural disadvantage as given and examining how variation in the size of the race/ethnic gaps in disadvantage affects race-ethnic gaps in violence.

Our primary goal is to extend scarce research to explicitly explore the association between race/ethnic disparities in structural disadvantage and Black-White, Hispanic-White, and Black-Hispanic gaps in homicide and violent arrest rates. To do this, we use arrest data and census estimates disaggregated by race and Hispanic ethnicity from incorporated census places in California and New York. These census place-level data display meaningful overlap in Black, White, and Hispanic levels of structural disadvantage and disorganization, as well as meaningful variation in the size of the Black, White, and Hispanic gaps in violence. In particular, we build on and extend the prior work of Velez, Krivo, and Peterson (2003), which (to our knowledge) is the only empirical assessment of the association between gaps in structural disadvantage and gaps in crime.

2. Structural Disadvantage and Race/Ethnic Differences in Violence

That differences in exposure to macro-structural characteristics across racial and ethnic groups drive differences in crime has a long history. For example, Blau and Blau (1982: 118-119) argue that “socioeconomic inequalities that are associated with ascribed positions, thereby consolidating and reinforcing ethnic and class differences, engender pervasive conflict in a democracy. Great economic inequalities generally foster conflict and violence, but ascriptive inequalities do so particularly.” They further argue that pronounced racial/ethnic disadvantage creates resentment, frustration, hopelessness, and alienation, which are expressed in greater violence. Thus, racial/ethnic group differences in the bottom of the inequality distribution (that is, differences in disadvantage) seem to be especially troubling in terms of the production of violent crime. Similarly, Agnew (1999) proposes that group-level structural disadvantage produces greater social psychological strains among group members, which then produces between-group differences in rates of crime, including violence. Likewise, structural disadvantage (such as poverty, unemployment, and female headship) is said to erode local systems of informal social control and collective efficacy, and may additionally foster subcultural adaptations favorable to violence (Sampson et al., 2005; Sampson and Bean, 2006; Matsueda, Drakulich, and Kubrin, 2006; Harer and Steffensmeier, 1992; Shihadeh and Steffensmeier 1994). Notably, Wilson (1987, 2009) and especially Anderson (1999) argue that racial segregation and concentrated disadvantage have fueled subcultural adaptations in which violence is tolerated and expected as a daily part of life in black underclass neighborhoods.

In sum, in what has come to be known as the racial invariance hypothesis, it is implied that differences in structural disadvantage should strongly predict race/ethnic differences in violence (see review in Steffensmeier et al 2010). As Velez (2006: 93) describes, “factors that affect crime, such as employment opportunities, concentrated disadvantage, and informal social control, should apply equally regardless of whether the neighborhood [or other locality] is predominantly White, Hispanic, or Black. Therefore, the relatively high levels of structural disadvantage in African American neighborhoods are responsible for their high levels of crime, especially in comparison to White neighborhoods.”

A sizable body of empirical research now exists showing that poverty and other measures of structural disadvantage, family structure, and residential stability are indeed correlated with rates of violence (especially homicide) both across a variety of ecological units and over time (see reviews by Crutchfield et al., 2006; Peterson and Krivo, 2005; Parker and McCall, 1999), and that structural disadvantage strongly impacts both White and Black violence rates (Peterson and Krivo, 2005; Kubrin and Wadsworth, 2003; Parker, 2001; McNulty, 2001; Krivo and Peterson, 2000). A handful of recent studies also point to a prominent role of structural disadvantage variables in explaining Hispanic homicide (Martinez et al., 2010; Steffensmeier et al. 2010; Xie 2010; Shihadeh and Barranco 2010b) and robbery (Cancino et al., 2009).

As some of these reviews highlight, there are shortcomings and ambiguities in extant research surrounding the racial invariance hypothesis and its empirical status remains unsettled (Steffensmeier et al. 2010). For our purposes, we focus on two main ambiguities. First, there is an overall lack of research that compares Hispanic, Black, and White violence (though a handful of recent studies have begun to address this shortfall). The scarcity of studies that incorporate Hispanics represents a serious weakness in our ability to evaluate the racial invariance claim that “structural variables predict [violent] crime in the same way for all racial and ethnic groups [emphasis added]” (Peterson and Krivo 2005: 338), and thus how structural variables explain race/ethnic differences in violence. This scarcity is especially problematic because Hispanics are the largest and fastest growing ethnic minority in the U.S. and research that helps us understand Hispanic violence relative to Black and White violence is vital for both theory and policy (Harris et al., 2009; Shihadeh and Barranco 2010a; 2010b; Steffensmeier et al., 2011). The few extant studies of Hispanic violence, while important, have focused on only one or a few cities and observe that Hispanic communities experience levels of disadvantage comparable to Blacks but have lower crime rates (Martinez, 2003; Sampson et al., 2005; Velez, 2006; Crutchfield et al., 2006; for an exception, see Phillips, 2002). Moreover, Black and Hispanic disadvantage are likely not independent of each together; rather, Hispanics may displace blacks from low-skilled jobs, exacerbating black unemployment and poverty (Shihadeh and Barranco 2010a). Thus, whether gaps in structural conditions between Hispanics and other race/ethnic groups have as strong of effects on gaps in violence compared to the effect of Black-White gaps in disadvantage on Black-White violence rates remains an important empirical question.

Second is the problem of “restricted distributions” that frequently arises in aggregate-level research comparing the effects of structural variables on violence rates disaggregated by race and ethnicity. It is well known that predominantly Black, White, and Hispanic areas experience very disparate levels of disadvantage (Shihadeh and Flynn, 1996; Krivo and Peterson, 2000; McNulty, 2001; Phillips, 2002; Krivo et al., 2009). This creates a thorny analytical problem whereby it is difficult to find spatial units that are comparable in levels of White, Black, and Hispanic disadvantage (particularly Black and White disadvantage) in order to compare the race- and ethnic-specific effects of disadvantage and disorganization factors on race- and ethnic-specific levels of violence (McNulty, 2001; Velez et. al., 2003).

Velez, Krivo, and Peterson (2003) approached this question in an innovative manner that attempted to address the problem of the lack of Black-White overlap in levels of structural disadvantage. By taking as given the existence of gaps in levels of disadvantage and homicide, Velez et al. (2003) examined how the size of gaps in disadvantage explained variation in the size of gaps in homicide. As Velez et al. (2003: 646) explain, the “lack of overlap in the separate distributions that is critical when comparing the sources of race-specific homicide becomes irrelevant in studying the link between Black-White inequality in social conditions and the racial gap in violence.” To our knowledge, theirs is the only study to date that has utilized this method to explicitly assess the extent to which disparities in macro-structural conditions are associated with race/ethnic gaps in violence. Indeed, their 1990 analysis of city-level Black-White gaps in homicide indicates that Black-White homicide disparities are driven significantly by differences across groups in structural characteristics.

3. The Present Study

We extend the Velez et al. (2003) study and contribute to extant knowledge on race/ethnic disparities in crime and their relationship to social structure in several ways. First, there is scant research on variation in race/ethnicity differences (or gaps) in violence, and another study that extends the research direction begun by Velez and colleagues is needed. Second, we move beyond the study by Velez et al. (2003) – which focuses only on homicide – to examine more recent data and include broader violence. Third, we employ a novel unit of analysis – the census place – which brings distinct analytical advantages over previous research using cities, neighborhoods, or counties. Fourth, we incorporate more extensive measures of structural disadvantage, as well as several different control variables. Fifth, Velez et al. (2003) focused only on Black-White differences (as does most research on race differences in violence) while our racially and ethnically disaggregated arrest data enable us to examine Hispanic-White and Hispanic-Black differences in homicide and violence rates, as well.

3.1 Data and Methods

To address the issues discussed above, we use racially and ethnically disaggregated arrest data from California and New York. We examine race and ethnic gaps in both homicide and the violent crime index (which combines homicide, forcible rape, aggravated assault, robbery). We use incorporated census places as the ecological unit of analysis, which has key advantages. Census places include non-overlapping geographic units (cities, villages, towns, boroughs) tracked by the U.S. Census bureau (U.S. Census Bureau, 1994), offering a unique and relatively untapped ecological unit of analysis. Census places are well-suited for our analysis because they provide a diverse set of spatial units that vary widely between each other in size, structural characteristics, and rates of violence. They also generally provide large enough numbers of each race-ethnic group for meaningful statistical analysis. Though they are larger than neighborhoods, census places are smaller and more spatially homogeneous units compared to states, counties, SMSAs, or cities (it is common for large cities to contain more than one census place). Additionally, we use 1999-2001 data, which provide a relatively more timely reflection not only of contemporary racial-ethnic differences in violent crime but also the association between contemporary structural conditions and violence. We include only those census places that have a total population of 10,000 or above in the year 2000 and have at least 1,000 residents of the race-ethnic group under consideration, yielding 232 places. We use these selection criteria in order to provide reliable measures of violence and structural characteristics disaggregated by race/ethnicity across census places.

We draw data from two main sources. First, information on race/ethnic-disaggregated homicide and index violence arrests is drawn from California’s and New York’s crime reporting programs. Second, we use 2000 US Census summary files 1 and 4 for data on social and economic characteristics of the White, black, and Hispanic populations in California and New York.

A key advantage of using arrest data from California and New York is that they classify arrestees into “White” “Black” and “Hispanic” groups (as well as “American Indian” and “Asian”), thus allowing us to examine White, Black, and Hispanic differences in the effects of disadvantage indicators on both homicide and violent offending (Demuth, 2003). A second advantage is that their populations are (a) large – together are home to more than 27 million Whites, 5 million Blacks, and 14 million Hispanics, and account for about 14 percent of all Whites, 16 percent of all Blacks, and 40 percent of all Hispanics living in U.S. (U.S. Census Bureau 2008); (b) diverse and representative of White, Black, and Hispanic populations for the nation as a whole.ⁱ A third advantage of the California and New York data is that arrests for violent index crimes in these two states make up a sizable share (about 20%) of all arrests for violent crime in the U.S.

3.2. Dependent Variables

The dependent variables in this study are Black-White, Hispanic-White, and Black-Hispanic gaps in homicide and violent index (sum of arrests for homicide, aggravated assault, forcible rape, robbery) offending rates per 100,000 at-risk persons for our sample of census places.ⁱⁱ Arrest data are subject to many well-known criticisms, but homicide is viewed as the most serious of index crimes and is the most likely to be reported to the police and result in an arrest, whereas violent index crimes are defined as more serious than (most) other crimes and have a high likelihood of being reported to the police. Though they overlap, homicide (lethal violence) and the violent index (mostly non-lethal) are typically seen as two substantively different measures of violence and represent two widely-used yardsticks for assessing patterns of violent offending in criminological research (Mosher, Miethe, and Phillips, 2002). Both violence measures are calculated using 3-year averaged arrest figures for 1999-2001 to add stability to the rates and also ensure adequate arrest counts for statistically rare offenses (particularly for homicide). We calculated the race/ethnic gaps in homicide and violence by subtracting the race/ethnic-disaggregated arrest rates from one another (Black–White, Black–Hispanic, etc.).

3.3. Primary Independent and Control Variables

We focus on black-White, black-Hispanic, and Hispanic-White gaps in three indicators of disadvantage – poverty, unemployment, and female headship – that have emerged as important structural predictors in macro-level research and macro-structural theories of violence. These disadvantage indicators are disaggregated by race-ethnicity (White, Black, Hispanic), as are several important control variables. Poverty is measured as the percentage of census-place residents below the poverty line. Unemployment is measured as the percentage of the civilian labor force between the ages of 16 and 59 that is unemployed. Female headship is measured as the percentage of families with children under 18 years old that are headed by a female.

In addition, because of their demonstrated relevance in previous studies, we include as controls: population density (residents per square mile) of a census place (logged); residential instability (percentage of Black, White, or Hispanic households that experience housing turnover during the 1995-2000 period); entropy as measure of racial/ethnic heterogeneity;ⁱⁱⁱ young male population (the percentage of the Black, White, or Hispanic population aged 15-24 and male); last, we include police per capita as a control for variations in law enforcement activity.

A concern with aggregate data is that measures such as poverty, unemployment, and female headship tend to be highly correlated and multicollinearity can bias estimates of specific relationships between gaps in our independent measures and gaps in violence. However, variance inflation factors (VIFs) for our full models including all of our measures (see Table 3) were well below the suggested value of 2.0, indicating that multicollinearity was not a concern in our models (Allison 1999). Additionally, we examine the effects of gaps in our three disadvantage measures separately (see Table 2) with a full set of controls. This strategy addresses potential multicollinearity and allows us to identify the relative importance of gaps in key disadvantage measures in predicting variation in Black, White, and Hispanic gaps in violence.

Table 3.

Seemingly Unrelated Regression of Homicide and Violent Index Gaps on Gaps in Structural Disadvantage (Poverty, Unemployment, and Female Headship) and Other Structural Characteristics (N=232)

	(A) Homicide						(B) Violent Index
	B-W Gap		B-H Gap		H-W Gap		B-W Gap		B-H Gap		H-W Gap
	b	B	b	B	b	B	b	B	b	B	b	B
Structural Disadvantage:
Poverty	.367* (.156)	.111	.294† (.153)	.211	.335** (.112)	.088	6.547 (4.132)	.072	5.36 (4.126)	.136	3.987 (3.057)	.041
Unemployment	−.230 (.216)	−.049	−.237 (.207)	−.095	−.197 (.167)	−.043	8.395 (5.652)	.065	8.452 (5.566)	.119	6.783 (4.484)	.057
Female Headship	.090 (.136)	.034	.100 (.129)	.074	.137 (.107)	.049	8.394* (3.537)	.117	8.853* (3.449)	.231	5.871* (2.847)	.082
Controls:
Residential Instability	−.076 (.094)	−.030	−.095 (.095)	−.088	−.121 (.069)	−.049	−7.470** (2.492)	−.107	−7.049** (2.564)	−.230	−7.494*** (1.875)	−.117
Young-male pop.	.342 (.296)	.044	.378 (.325)	.100	.468* (.230)	.057	−1.605 (7.960)	−.008	.539 (8.792)	.005	1.574 (6.336)	.007
Pop. density (ln)a	−1.543 (1.931)	−.051	−1.591 (1.928)	−.139	−.076 (.774)	−.003	34.094 (51.828)	.041	10.447 (49.640)	.032	20.135 (21.715)	.026
Entropya	−1.740 (9.087)	−.012	−5.035 (9.250)	−.090	3.234 (3.719)	.022	−274.236 (244.014)	−.068	−279.001 (238.319)	−.176	−13.053 (104.242)	−.003
Police per capitaa	8.786*** (1.714)	.333	9.861*** (1.690)	.990	−1.189 (.794)	−.046	56.596 (46.004)	.079	138.499** (43.593)	.491	−59.383** (22.100)	−.089
Constant	10.801 (15.779)		12.337 (15.607)		−.398 (6.478)		660.523 (423.448)		628.191 (401.831)		78.791 (181.383)
R2		.155		.114		.075		.146		.114		.079
Breusch-Pagan	X2 = 186.402, p<.001						X2 = 193.001, p<.001

Note: Standard errors in parentheses

^aNot a race-specific measure (no gaps calculated)

^†p<.10,

^*p<.05,

^**p<.01,

^***p<.001 (two-tailed)

Table 2.

Seemingly Unrelated Regression of Homicide and Violent Index Gaps on Race-Specific Gaps in Structural Disadvantage (Poverty, Unemployment, and Female Headship), N = 232

		(A) Homicide						(B) Violent Index
			B-W Gap		B-H Gap		H-W Gap		B-W Gap		B-H Gap		H-W Gap
Model		b	B	b	B	b	B	b	B	b	B	b	B
1) Controls only	R2	.121		.128		.023		.033		.042		.070
(2) Poverty		.328** (.123)	0.099	.267* (.131)	0.191	.320** (.094)	0.084	12.466*** (3.297)	0.138	10.718** (3.628)	0.271	7.733** (2.580)	0.079
	R2	.155		.108		.064		.112		.070		.083
(3) Unemployment		.044 (.169)	0.009	.004 (.168)	0.002	.048 (.150)	0.010	13.455** (4.465)	0.104	12.777** (4.584)	0.181	10.864** (3.989)	0.092
	R2	.123		.128		.024		.082		.060		.083
(4) Female headship		.205† (.109)	0.078	.196† (.108)	0.145	.216* (.100)	0.078	9.877** (3.003)	0.137	9.986** (2.959)	0.260	8.205** (2.670)	0.114
	R2	.131		.118		.045		.088		.094		.065

Note: Each model is estimated with only one structural disadvantage gap at a time and a full set of controls (not shown); standard errors in parentheses

^†p<.10,

^*p<.05,

^**p<.01,

^***p<.001 (two-tailed)

3.4. Analytic Strategy

We use seemingly unrelated regression (SUR) to estimate separate models for each racial/ethnic gap comparison: Black-White, Hispanic-White, and Black-Hispanic. SUR is more appropriate than ordinary least-squares (OLS) procedures for estimating our models because SUR takes into account that the race-ethnic comparisons are based on data for the same census places and therefore that the Black, White, and Hispanic arrests are not independent samples (see Ousey, 1999, Phillips, 2002, Schwartz, 2006 for details and similar applications of SUR).

Two approaches were used to deal with those census places with values of zero for homicide rates (low frequency or zero values are not problematic for index violent offending, since almost all places are likely to report at least one violent crime). For the homicide SUR models, we assigned a value of 0.1 to the rates in those census places not reporting any homicides over the three-year period (see Steffensmeier and Haynie 2000; Schwartz, 2006). We replicated the results using negative binomial methods, a strategy that is appropriate when a relatively large number of zero values exist across aggregate units (as is the case for homicide), and the results (available from the authors) closely parallel those derived from the SUR models reported below.^iv

4. Findings

First, we present descriptive statistics comparing both levels and gaps in our key disadvantage indicators and violent crime in order to help establish the levels of exposure of Whites, Blacks, and Hispanics to disadvantage and crime as well as the relative size of and variation in differences between each race/ethnic group. Second, we present SUR models regressing gaps in homicide and violent index crime separately on gaps in poverty, unemployment, and female headship in order to assess their individual contribution to explaining gaps in crime. Third, we present SUR models regressing gaps in homicide and violence on gaps in poverty, unemployment, and female headship on gaps in homicide and violent index crime in order to establish their combined effect net of other key macro-structural characteristics.

4.1. Descriptive Statistics

Table 1 shows descriptive statistics for both the levels and race/ethnic gaps in the dependent and independent variables used in our analyses.

Table 1.

Means and Standard Deviations for Levels and Gaps by Race/Ethnicity (N = 232)

	Levels			Gaps
	White	Black	Hispanic	B-W	B-H	H-W
Dependent Variables:
Homicide rate	2.85 (5.43)	15.81 (26.75)	6.39 (8.94)	12.96 (26.65)	9.42 (26.25)	3.55 (10.06)
Violent index rate	331.52 (221.66)	1311.48 (818.36)	505.63 (334.68)	979.96 (725.30)	805.85 (675.12)	174.11 (285.16)
Key Independent Variables:
Poverty	8.65 (4.89)	18.62 (10.39)	18.93 (9.89)	9.96 (8.03)	−.31 (6.89)	10.27 (7.21)
Unemployment	5.46 (2.80)	10.82 (6.26)	9.28 (5.08)	5.36 (5.63)	1.54 (5.71)	.381 (4.03)
Female Headship	8.19 (3.45)	24.57 (10.95)	15.08 (8.21)	16.38 (10.09)	9.49 (9.42)	6.88 (7.43)
Residential Instability	44.32 (9.39)	56.62 (12.31)	56.43 (8.72)	12.30 (10.39)	.18 (10.53)	12.11 (9.29)
Entropya	.71	.71	.71	-	-	-
Control Variables:	(.18)	(.18)	(.18)	-	-	-
Young Male Population	5.86 (2.50)	7.97 (3.83)	9.71 (3.06)	2.11 (3.46)	−1.74 (3.18)	3.85 (2.67)
Population Density (ln)a	8.24 (.88)	8.24 (.88)	8.24 (.88)	- -	- -	- -
Police Per Capitaa	1.41 (1.01)	1.41 (1.01)	1.41 (1.01)	- -	- -	- -

Note: Standard deviations in parentheses

^aNot a race-specific measure (no gaps calculated)

Key findings are as follows. First, there are notable differences in levels of homicide and violent index crime across race/ethnic groups. The first three columns of Table 1 indicate that Blacks have the highest homicide and violent index crime rates (rates per 100,000 of 15.81and 1311.48, respectively), followed by Hispanics (6.39 and 505.63) and then Whites (2.85 and 331.52). Hispanic crime rates tend to fall between the relatively low levels of Whites and the higher rates of blacks, though they are closer to the White than the Black rates.

Second, we observe differences in our key independent structural characteristics, as well. Mean Black (18.62) and Hispanic (18.93) poverty levels are greater than twice that of Whites (8.65). Unemployment and female headship show similar disparities (in fact, Black female headship is three times greater than White female headship).

Third, given differences in levels of crime and structural disadvantage, we observe sizable gaps in both our dependent and independent measures. For homicide and the violent index, the mean Black-White gaps are significantly larger than the Hispanic-White and Black-Hispanic gaps. Among our key independent measures, the mean Black-White and Hispanic-White gaps in poverty are roughly comparable, while the mean Black-Hispanic gap is essentially zero. Thus, the mean poverty rates of Blacks and Hispanics here are nearly identical. For unemployment and female headship, the mean Black-White gap is significantly greater than either the mean Hispanic-White or Black-Hispanic gaps.

As noted above, an advantage of looking at racial/ethnic gaps is that they explicitly take into account the considerable variation between places in the size of racial/ethnic differences in levels of structural disadvantage and violence rates. The minimum and maximum sizes of the gaps (not shown) confirm this. For example, in several census places we observe Black-White homicide gaps where the White rate exceeds the Black rate (minimum = -5.22), while other census places have homicide gaps that are much greater (maximum = 11.49) than the mean Black-White gap of 1.39, indicating that the Black rate is dramatically greater than the White rate. Similar variation is observed for the racial/ethnic disparities in our key independent measures, where minimum values of Black-White, Hispanic-White, and Black-Hispanic gaps in poverty, unemployment, and female headship are in the opposite direction from the mean gap and the maximum gap values are double or triple the mean gap.

Overall then, Table 1 suggests greater Black and Hispanic violent offending (both for homicide and violent index crime) and exposure to structural disadvantage than Whites, reflected in both the levels and gaps in our key measures. Additionally, we note that our racial/ethnic gaps in violence and structural characteristics vary greatly in size across our sample of census places.

4.2. Race/Ethnic Gaps in Homicide and Violence

Table 2 shows results from SUR models regressing homicide and violent index gaps on the racial-ethnic gaps in individual components of disadvantage, along with control variables (not shown).

For homicide, gaps in poverty significantly predict racial/ethnic gaps in homicide, net of controls. In each comparison, greater racial/ethnic gaps in poverty are associated with larger gaps in homicide arrest rates. That is, the more Black poverty rates exceed those of Whites, the more Black homicide rates exceed those of Whites. The same is true for Hispanic–White and Black-Hispanic poverty gaps, both of which have statistically significant, positive relationships with their respective gaps in homicide. Specifically, poverty has a small to moderate-size standardized effect of about .08 on Hispanic-White homicide gap, about .10 on the Black-White gap, and about .19 on the Black-Hispanic homicide gap. Race/ethnic gaps in unemployment are unassociated with gaps in homicide, while race/ethnic gaps in female headship are positively associated with gaps in homicide. However, female headship’s effect is only marginally significant for the Black-White and Black-Hispanic homicide gaps.

For the violent index, all of the gaps in the disadvantage variables have positive, statistically significant relationships with gaps in violent crime (net of controls). Poverty gaps, again, consistently predict each of the racial ethnic gaps in violent crime, as do gaps in unemployment and the proportion of households that are female-headed. Each of these three measures (poverty, unemployment, and female-headship) suggests that as Black-White, Hispanic-White, and Black-Hispanic disparities in disadvantage increase, so too do disparities in violent index crime rates. Gaps in poverty have particularly notable effects on Black-White (Beta = .14) and Black-Hispanic (Beta = .27) gaps in violent crime.

We now move beyond estimating the separate effects of poverty, unemployment, and female headship gaps to assess how race/ethnic gaps in each of these disadvantage measures together impact race/ethnic gaps in homicide and violent index rates. Table 3 shows results for SUR models that simultaneously include poverty, unemployment, and female headship rather than including these individual variables separately.

In contrast to the effects in Table 2, net of controls, only race/ethnic gaps in poverty significantly explain race/ethnic gaps in homicide (and this effect is marginal for the Black-Hispanic gap). Gaps in unemployment and female headship are unassociated with Black-White, Black-Hispanic, or Hispanic-White gaps in homicide. The only other measure which predicts Black-White and Black-Hispanic gaps in homicide is police per capita, while the Hispanic-White gap in the young-male population significantly predicts the Hispanic-White homicide gap.

For violent index crime, our SUR models indicate that race/ethnic gaps in female headship are positively and significantly associated with each of the race/ethnic gaps in violent index rates. That is, the greater the race/ethnic gap (Black to White, Black to Hispanic, or Hispanic to White) in female headship, the greater the gap between groups in violent index rates, net of other important gaps and levels of structural characteristics. In contrast, once all of the disadvantage indicators are entered together, gaps in poverty and unemployment are unassociated with gaps in violence, though these relationships are in the expected positive direction. It is likely that, since these disadvantage indicators are correlated with each other, there is just enough shared explanatory power to weaken the other gap coefficients (and the poverty coefficient, in particular).

Among our other independent and control variables, gaps in residential instability are associated with decreased violence. That is, greater Black-White, Hispanic-White, and Black-Hispanic disparities in residential turnover predict smaller Black-White, Hispanic-White, and Black-Hispanic gaps in violent index crime rates. We note that this might be the result of areas where disparities in population mobility favor Blacks and Hispanics and create greater racial/ethnic diversity (with associated crime-reducing effects) and is consistent with the finding for our entropy measure (negative effect), though the entropy-violence gap relationships are not significant at p<.10. Levels of police per capita are also associated with gaps in violent crime for the Black-Hispanic and Hispanic-White violence gaps. That police per capita is associated with racial/ethnic gaps in overall violence likely reflects the reactive stance in policing to target high crime (violence) localities, as is more often characteristic of Hispanic and especially black neighborhoods (Maltz, 1994).

5. Discussion and Conclusion

We have extended research by Velez et al. (2003) by investigating the relationships between gaps in Black, White, and Hispanic disadvantage (rather than just Black-White differences) and between-group gaps in homicide and index violence (rather than just homicide), with more recent data and a novel unit of analysis. As Velez et al. (2003) argue, treating the race-ethnic gap in violence itself as a dependent variable accounts for the differential distribution of disadvantage between race and ethnic groups, especially Blacks and Whites, and allows us to further examine the impact of inequality on crime across at the community-level.

As proponents of the racial invariance perspective contend, group differences in structural disadvantage should strongly predict racial and ethnic differences in homicide and violence. We find moderate, but not unqualified, support for this position. Our analyses show that, among the structural disadvantage variables, gaps in poverty appear to significantly affect homicide. That is, there are greater Black-White, Black-Hispanic, and Hispanic-White differences in homicide where there are greater between-group differences in poverty. In addition, there is evidence that when it is considered alongside other key macro-structural controls (but without the other disadvantage indicators), gaps in female headship predict White-Hispanic gaps in homicide and (to a lesser extent) Black-White and Black-Hispanic gaps. In addition, gaps in the relative size of the young, male population helps to explain Hispanic-White homicide gaps. Regarding overall index violence, racial/ethnic gaps in poverty and unemployment are associated with greater racial/ethnic disparities in violence individually (net of controls). Gaps in female headship, however, consistently and positively predict violent crime gaps even when poverty and unemployment are simultaneously included in the models, especially the gap between Black and Hispanic violence.

In sum, gaps in various structural disadvantage predictors, especially poverty and female headship, exert important influences on the relative size of homicide and violence gaps. However, gaps in disadvantage do not fully explain Black-White and Black-Hispanic violence gaps. Much variation remains unexplained. Cultural and social organizational differences may be at play, too, in explaining racial and ethnic differences in violence. In particular, even though gaps in poverty and female headship significantly explained Black-Hispanic homicide and violence gaps, differences in dimensions of disadvantage between Blacks and Hispanics may not be the full explanation of differences in violence between blacks and Hispanics. Note that in Table 1, Black-Hispanic poverty and unemployment gaps are minimal compared to gaps of both groups vis a vis Whites, though Black-Hispanic gaps in female headed households are more substantial. Yet, Black-Hispanic gaps in violence and homicide are substantially greater than Hispanic-White crime gaps. Although Hispanic communities experience similar levels of disadvantage to blacks, their cultural and social capital may insulate them at least somewhat from the worst violence-producing effects of structural disadvantage.

Some writers have argued that Hispanics have lower rates of violence than Blacks because elements of their culture and social organization differ from those of underclass African-Americans (Martinez, 2003; Velez, 2006; Steffensmeier et al. 2011). Hispanics have experienced different forms of discrimination (Portes and Rumbaut, 2006), and Mexican Americans, for example, often maintain close ties with family and friends at home and have strong cultural identities, heritage, and values (Massey 2009). The shared culture, in turn, provides the basis for group cohesion, social organization, and economic opportunities (Light and Gold, 2000). Hispanics’ comparatively communal culture is seen as a potential buffer that (partly) insulates them from the effects of disadvantage on crime (Martinez, 2003). By contrast, Wilson (1987, 2009) and especially Anderson (1999) argue that racial segregation, concentrated disadvantage, and social isolation have fueled sub-cultural adaptations in which violence is tolerated and expected as a daily part of life in Black underclass neighborhoods (see also Harer and Steffensmeier, 1992; Steffensmeier and Ulmer, 2005). All this suggests that disadvantage may not be the only factor that explains White, Black, and Hispanic differences in violence. Future research should try to incorporate cultural and social organizational measures in addition to disadvantage, and investigate how both cultural and structural factors affect race/ethnic differences in violence.

Future research needs to build on the current study in at least two important ways. First, while our research represents an important contribution, alternative methods might also be employed to assess race/ethnic gaps in violence and whether structural disadvantage accounts for those gaps. We have chosen to analyze gaps in violence, the advantage of which is that this method accounts for differences in distributions of disadvantage across groups. However, one limitation of this approach is that it is unable to definitively test whether disadvantage variables have the same effects across groups, as argued by the racial invariance hypothesis.^v Recent research (Steffensmeier et al., 2010) suggests that disadvantage indicators have significantly different effects between Blacks, Whites, and Hispanics, especially for index violence. The implication of this for the current analysis is that, if disadvantage has differing effects across groups, then the impact of race/ethnic gaps in disadvantage on gaps in violence might not be constant across the distribution of disadvantage. This is likely to be another important part of the story of race/ethnic gaps in violence that future research should investigate.^vi

Second, more studies are needed that focus on macro-social predictors of Hispanic violence both because of the shortage of research including Hispanics and because of the substantial growth of the Hispanic population, in particular. Most needed, perhaps, is research that takes into account some nuances of this Hispanic growth and recent immigration patterns. For example, Shihadeh and Barranco (2010b) have raised serious questions about the “Latino paradox,” and have shown that the notion that Hispanics experience relatively low violence given their levels of disadvantage applies only to places that are traditional Hispanic immigrant destinations. By contrast, Hispanics are victimized by violence at considerably higher levels in new immigration destinations. Shihadeh and Barranco (2010b) attribute this to “linguistic isolation” from the larger community, which both directly makes Hispanic victimization more likely and increases structural disadvantage. In addition, Shihadeh and Barranco (2010a) find that Hispanic immigration increases black violence in contexts where Hispanics displace Blacks from low-skill jobs.

Unfortunately, our analysis focused only on California and New York includes many traditional, urban Hispanic immigrant destinations. Therefore, one promising direction for future research would be to investigate whether effects of structural disadvantage (and gaps therein) varies in effect on differences in both Hispanic and Black violence relative to Whites across established/traditional versus new Hispanic immigrant destinations. Such research would shed light on the role of cultural influences (as suggested above) in that it would help us adjudicate the veracity of the “Latino paradox” and illuminate the role of Hispanic linguistic isolation in new immigration destinations highlighted by Shihadeh and Barranco (2010b).

The degree to which gaps between groups in structural disadvantage explain racial or ethnic differences in violence is not settled. Furthermore, we need more research assessing the link between differences in White, Black, and Hispanic violent crime are driven by disparities in the structural circumstances of these groups. We have sought to contribute to our understanding of the role of disadvantage in producing differences between racial and ethnic groups in homicide and violence. We find that group differences in various disadvantage factors, especially poverty and female headship, do indeed partially explain gaps in violence. Such differences in disadvantage are not the whole story, however, and future research should investigate other social organizational and cultural factors that may explain racial and ethnic gaps in violent crime. As Shihadeh and Barranco (2010a; 2010b) imply, the growth of Hispanic populations and the dynamics of their migration are the most significant population shifts in the U.S. in more than a century, and it is crucial that we understand the implications of these shifts for structural disadvantage and crime for White and racial and ethnic minorities.