INTRODUCTION AND MOTIVATION
Just over twenty-two years ago, we published “Toward a Theory of Race, Crime, and Urban Inequality” (Sampson and Wilson, Reference Sampson, Wilson, Hagan and Peterson1995). As stated in the first sentence of that paper, our motivation was to “address one of the most central yet difficult issues facing criminology—race and violent crime” (p. 37). As further indicated by the title, this effort was first and foremost theoretical, relying on overarching assumptions, introducing theoretical concepts to make sense of empirical findings, and presenting testable hypotheses.Footnote 1
The current paper reassesses and updates the theory in light of evidence that has accumulated since 1995. We also address new empirical challenges and chart a path forward in the form of a research agenda. At the time of our original publication, we expressed dismay that “the discussion of race and crime is mired in an unproductive mix of controversy and silence” (Sampson and Wilson, Reference Sampson, Wilson, Hagan and Peterson1995, p. 37). Although controversy continues, we are gratified that there is no longer silence on these matters. The past two decades have witnessed an upsurge in empirical studies that have explicitly or implicitly tested the theoretical tenets of what has come to be known as the thesis of “racial invariance”—the assertion that racial disparities in rates of violent crime ultimately stem from the very different social ecological contexts in which Blacks and Whites reside, and that concentrated disadvantage predicts crime similarly across racial groups.
We are motivated to revisit the racial invariance thesis for more than just intellectual reasons, however. Important social changes and political realities bear heavily on any contemporary discussion of race. Indeed, in the summer of 2017, the nation became embroiled once again in ferocious debates over race and racism. A full-throated defense of White supremacy, for example, is no longer merely a relic of the past. Race, crime, and concentrated disadvantage have also framed the recent push for criminal justice reform. These frames are most visible in the Black Lives Matter movement’s focus on the policing crisis, and in bipartisan opposition to the unprecedented and racially tinged mass incarceration of the last few decades. Moreover, we are motivated to consider other major social changes from the past twenty years that we either did not anticipate or did not fully consider in the original paper. These changes include the dramatic drop in violent crime—including among African Americans—and the sharp rise in immigration that has challenged the “Black-White” framework that might be read into the racial invariance thesis.
To anticipate our conclusion: we believe that the core elements of Sampson and Wilson’s theory find broad empirical support. Large racial disparities in both violent crime and ecological contexts continue; structural ecological factors are strong predictors of violent crime and account for a substantial proportion of racial disparities; and the predictive power of these factors transcends racial boundaries. In support of this assessment, we clarify the meaning of “invariance” and also consider different supraindividual level units of analysis, the applicability of our thesis in light of widespread immigration—especially from Latin American countries—and the mechanisms underlying the linkages among ecological disadvantage, race, and crime. Before assessing the challenges to and evidence for our argument, we first step back and briefly review the basic tenets of the theory and the context within which valid empirical tests of the theory should be evaluated.
THEORETICAL SYNOPSIS
In the early 1990s, Black neighborhoods experienced considerably more violence than White neighborhoods; in fact, the leading cause of death among young Black males was homicide. In cities like New York and Chicago, homicide rates hit record levels; in racially isolated neighborhoods of concentrated poverty, gun deaths hit epidemic proportions. Despite these stark empirical developments, many criminologists faced with the topic of race and crime tended to revert to simplistic arguments that pitted culture against social structure. Others denied the existence of race-related differentials in violence and focused instead on police bias and the potential invalidity of official crime statistics (Sampson and Wilson, Reference Sampson, Wilson, Hagan and Peterson1995). Discrimination was (and is) real, of course, but behavioral differences in violent victimization and offending were (and are) also real. Residents of poor Black neighborhoods demanded action on both fronts, but the response of the state was severely and tragically lopsided in favor of “law and order” responses to violence and drug use in the Black community (Forman Reference Forman2017; Travis et al., Reference Travis, Western and Redburn2014).
Within this intellectual context, we advanced a theoretical strategy that was primarily structural but that also incorporated cultural arguments regarding race, crime, and inequality in American cities. In contrast to psychologically-based relative deprivation theories and the subculture of violence thesis, we viewed the race and crime linkage from a contextual framework that highlighted the very different community contexts of Blacks and Whites. Our basic thesis was that macrosocial patterns of residential inequality by race gave rise to the social isolation and ecological concentration of the truly disadvantaged, which in turn led to structural barriers and behavioral adaptations that undermined social organization and hence the control of crime and violence.
Our contextual argument traces its roots in criminology back to at least the mid-twentieth century, when Clifford Shaw and Henry McKay (Reference Shaw and McKay1969 [1942]) highlighted three key facts: 1) overall rates of delinquency were higher for Black boys than for White boys in Chicago; 2) rates of delinquency among both Black and White boys varied by neighborhood; and yet 3) it was impossible to determine empirically if Black boys’ rates were higher than those for Whites living in comparable neighborhoods. “Even if it were possible to parallel the low economic status and the inadequacy of institutions in the white community,” Shaw and McKay wrote elsewhere, “it would not be possible to reproduce the effects of segregation and the barriers to upward mobility” (Reference Shaw and McKay1949, p. 614). These scholars thus recognized the structural predicates of racial segregation that included racially discriminatory laws and housing policies. In their formulation, political economy and structural ecology constituted a self-amplifying feedback loop that produced exceptional disadvantage in African American neighborhoods.
In our 1995 paper, we extended Shaw and McKay’s ideas logically and theoretically by posing two central questions. First, we asked to what extent do Black and White rates of crime vary by type of ecological area? Second, in American cities during the late twentieth century, is it possible to find in White communities the structural circumstances under which many Blacks live?
The first question is crucial, for it signals that Blacks are no more monolithic than Whites. To assign to Blacks (or to Black neighborhoods) a distinct and homogeneous “criminogenic” character is an act of racial stereotyping or essentializing. We wished to move away from this essentialism and toward the analytic point that there is variation in the crime rates of both Blacks and Whites—and that this within group variation might correspond similarly to variation in community context. Our review of the evidence suggested that Whites’ and Blacks’ rates of violence varied similarly with specific ecological conditions (e.g., concentrated poverty, family disruption). We thus proposed that the ultimate causes of Black crime were not unique but rather were rooted in structural differences among communities, cities, and states (Sampson and Wilson, Reference Sampson, Wilson, Hagan and Peterson1995). According to this perspective, “race” is not a direct cause of violence (or of any other social behavior, for that matter). Instead, race is a marker for the accumulation of social and material adversities that both follow from and constitute racial status in America (Sampson Reference Sampson2012). For better or worse, this proposition has been labeled the “racial invariance thesis.”
On the second question of cross race comparability or reproduction of conditions, our answer was a clear “no.” In 1995, more than forty years after Shaw and McKay’s assessment, Blacks and Whites still did not share similar neighborhood environments—even after controlling for family income. Racial differences in concentrated urban poverty were so enormous that the most deprived urban contexts in which Whites resided were considerably better off than the average context of Black communities. The State took center stage in our explanation of this disparity. We wrote: “no discussion of concentration effects is complete without recognizing the negative consequences of deliberate policy decisions to concentrate minorities and the poor in public housing” (Sampson and Wilson, Reference Sampson, Wilson, Hagan and Peterson1995, p. 43). Thus, we argued that macrostructural factors—both historical and contemporary—combined to concentrate urban Black poverty and its associated social dislocations (see also Hirsch Reference Hirsch1983; Massey Reference Massey1990; Sugrue Reference Sugrue1996).Footnote 2
The theory went on to propose structural and cultural mediating mechanisms, drawing especially from social disorganization theory (Bursik Reference Bursik1988; Sampson and Groves, Reference Sampson and Byron Groves1989), Ruth Kornhauser’s (Reference Kornhauser1978) systemic model of community, and William Julius Wilson’s ([1987] 2012) theory of the truly disadvantaged. Mechanisms included social isolation (i.e., segregation from resources and networks); a diminution of informal neighborhood controls or what we now call “collective efficacy” (Sampson et al., Reference Sampson, Raudenbush and Earls1997); the disruption of institutional and organizational strength (e.g., diminished networks of connectivity among institutions, lower density of organizations); and the emergence of a peer control system that facilitated gang formation.
We also posited that structural conditions shaped what we termed “cognitive landscapes”: ecologically structured norms regarding the role of legal institutions and appropriate standards and expectations of conduct. Under conditions of severe and persistent concentrated poverty, high crime, and ineffective policing, residents come to expect crime, disorder, and the illegal economy to be a part of their daily lives. These shared expectations formed the basis of what was later termed legal cynicism (Kirk and Papachristos, Reference Kirk and Papachristos2011; Sampson and Bartusch, Reference Sampson and Bartusch1998). We thus asserted the power of ecologically structured shared expectations and schemas—in interaction with structural features of the urban environment—to influence the probability of criminal outcomes and harmful deviant behavior. In short, our theory was: 1) structurally based but incorporated cultural mechanisms, and 2) probabilistic, rather than deterministic, in nature.
The Context of Validation
In discussing the logical structure of our theory, it is important to underline the context of validation—namely, the structure of explanation, the meaning and significance of concepts, and the nature of evidence.
Turning first to the structure of explanation, we must clearly state the conditions necessary to adequately test a theory’s validity, and therefore to determine the extent to which it has withstood critical scientific scrutiny over time. Aside from the stipulation that empirical tests of our original thesis include disadvantaged groups in urban areas, the other important condition that has to be satisfied is the unit of analysis. Our theoretical discussion of the linkage between race and crime associates macrosocial patterns of residential inequality with the emergence of social isolation and ecological concentration among disadvantaged groups. As noted, this relationship leads to structural barriers and cultural adaptations that ultimately weaken social organization and thereby decrease crime control. As Sampson and Wilson pointed out, “This is a distinctly sociological viewpoint, for it focuses on the proximate structural characteristics and mediating processes of community social organization that help explain crime, while also recognizing the larger historical, social, and political forces shaping local communities” (1995, p. 45). Accordingly, the primary unit of analysis in our theory was at the macrostructural/ecological level, which is represented by analyses of meso- and macrolevel data (e.g., aggregate level data), not individual level data, to test its validity.Footnote 3 Questions might be raised about the most suitable macrostructural/ecological level data set to address our theory (e.g., neighborhoods, incorporated census places, or cities), but, in the context of validation, such questions represent the separate domain of assessing the explanatory and predictive import of the theory.
Related to the structure of explanation is the meaning and significance of concepts. If a concept is included in a theoretical proposition, its meaning derives from the role it plays in clarifying or helping to make theoretical arguments explicit. In addition, the theoretically derived definitions of key concepts in our theory of race, class, and urban inequality—such as racial invariance, cognitive landscapes, and social isolation—reflect our macrostructural/ecological unit of analysis. Such concepts enable researchers to devise empirical measures or operational definitions that are theoretically based and therefore can play a role in identifying causality. A concept is considered significant if the theoretical proposition within which it is embedded is supported by empirical research studies, as revealed in an assessment of the nature of evidence.
In reflecting on the context of validation, we are cognizant of two potential problems to confront when trying to test our theory. One is the misinterpretation of racial invariance. In the original formulation, we put forth a qualitative rather than quantitative conceptualization of differences by race—our argument was that the ultimate sources of crime are the same across racial groups, not that their observed effects are necessarily equivalent in magnitude within each group. Consider the only appearance of the language of invariance in the entire paper: “The sources of violent crime appear to be remarkably invariant across race and rooted instead in the structural differences among communities, cities, and states in economic and family organization” (1995, p. 41). The primary emphasis was thus on equivalence in the community level sources of crime. To insist on statistically identical coefficients in race-specific models (e.g., Ousey Reference Ousey1999), especially in light of measurement error and random fluctuations across studies, is in our view both unrealistic and a misreading of our claims. Take concentrated poverty as a concrete example. If neighborhood poverty has a significant positive effect on rates of violence in both Black and White communities, we would consider this evidence in favor of the theory; the fact that one coefficient is larger than another does not undermine the logic of the original theory. By contrast, if one factor consistently reduces crime in the Black community but consistently increases it in the White community, then that would be evidence against the theory, i.e., there would be different sources of crime. In our original assessment and in the updated review below, we apply this qualitative, and falsifiable, perspective.
The second interpretive issue concerns counterfactuals. It is difficult or impossible to reproduce in White communities the structural circumstances under which many Black Americans live, especially the historical legacy of extended racial discrimination across generations. From our perspective, race in the United States—specifically, being African American—embodies both historical and concurrent disadvantage. That Whites have not experienced this structural reality does not negate the power of our theory or the postulate about racial invariance, however, because we argue that had Whites been exposed to the same structural conditions, they would exhibit similar responses. To argue otherwise is to posit that there is something unique about the Black race, unmoored from structurally experienced conditions, that is criminogenic. Moreover, just as no subject in a randomized experiment can simultaneously be in the treatment and control group, the causal effect of structural disadvantage is an unobserved but logical counterfactual.
Although no study can provide an absolute test of the full theory, one can still provide relative tests of the theory by examining those situations where the structural conditions of Whites begin to approximate the structural conditions of Blacks. Irish-American “ghettoes” in the early twentieth century were rife with deep poverty, ethnic discrimination, violence, and cynicism toward the law (Laub and Sampson, Reference Laub and Sampson2003; Muhammad Reference Muhammad2011), for example, and pockets of rural White Appalachia today are economically and socially dislocated—constituting a kind of “white underclass” (Vance Reference Vance2016). More generally, in White Trash, the historian Nancy Isenberg (Reference Isenberg2016) shows that when poor Whites live in a caste like system of class disadvantage and stigmatization, their responses are far from virtuous. This is not to say, of course, that poor White communities have experienced the same disadvantages as poor Black communities, but rather that within race responses to concentrated disadvantage can yield similar patterns. Researchers might draw upon ethnographic or archival data to uncover other conditions in White communities that provide meaningful even if not exact comparisons. The crucial question, to which we now turn, is the extent to which empirical studies—those that satisfy the conditions for testing the theory that we outlined above—lend support to our theoretical arguments. We must also bear in mind that the findings from any given study neither refute nor prove a theory (Lieberson and Horwich, Reference Lieberson and Horwich2008).
THE EVIDENCE
We highlight here the two basic observations that motivated Sampson and Wilson (Reference Sampson, Wilson, Hagan and Peterson1995): the existence of, and connections between, racial disparities in violent crime rates and concentrated deprivation. Even a cursory review of the evidence suggests that these two sets of disparities persist to this day and remain closely linked.
We focus mainly on homicide, as a reflection of broader trends in violent crime. Homicide is of great interest not only because of its severity but also because it is widely accepted as a reliable indicator of violent crime. In the United States, the overall homicide rate nearly doubled from the mid-1960s to the late 1970s, reaching a peak of 10.2 per 100,000 population in 1980. It subsequently fell off but then rose to a second and nearly equal peak of 9.8 per 100,000 in 1991—just four years before we published our racial invariance thesis. In Chicago, the homicide rate reached its twentieth-century peak in 1993. Since then, the homicide rate in the United States has declined sharply to 4.8 per 100,000 in 2010, though it has risen slightly over the last couple of years and markedly in Chicago (Cooper and Smith, Reference Cooper and Smith2011; Rosenfeld Reference Rosenfeld2016).
Despite the significant and unexpected crime drop in America, racial disparities in violent offending and victimization continue. The annual rate of homicide victimization was approximately six times greater among African Americans than among Whites for each year from 1999 through 2008.Footnote 4 Parallel disparities in homicide offending have fluctuated between 6.7 and 8.1, landing on 7.26 in 2008 (Cooper and Smith, Reference Cooper and Smith2011). These trends in homicide victimization and offending suggest that the racial disparity in violent crime that we noted in 1995 constitutes a continuing empirical reality. The same conclusion holds for violent victimization (Lauritsen et al., Reference Lauritsen, Heimer and Lang2018).
As is true of disparities in violent crime, racial disparities in residential and other social contexts continue to plague the United States. Large American cities remain highly segregated by race (Massey et al., Reference Massey, Rothwell and Domina2009), and recent declines in residential segregation among smaller cities stem largely from the growth of multiethnic neighborhoods rather than integration between Whites and Blacks (Friedman Reference Friedman2008). These trends interact with racial discrimination in the housing and mortgage markets (Pager and Shepherd, Reference Pager and Shepherd2008), and rising income inequality in general (Piketty Reference Piketty2014) to create a feedback loop of concentrated racial disadvantage.
Concomitantly, extant research points to enduring racial differences in the experience of neighborhood poverty and related disadvantage (Sampson Reference Sampson2012; Sharkey Reference Sharkey2013). Recently, Kristin Perkins and Sampson (Reference Perkins and Sampson2015) found that in Chicago these differences have survived several dramatic upheavals in the American city—increases in immigration and in suburban poverty; the growth of the Black middle class; the macroeconomic blow of the Great Recession; and the aforementioned decline in violence. Not only do individually poor Blacks (i.e., those in low-income households) have higher mean neighborhood poverty than any other sub-group, but “even blacks who are not individually poor have higher mean neighborhood poverty rates than Latinos and whites who are individually poor” (Perkins and Sampson, Reference Perkins and Sampson2015, p. 44). A similar pattern holds for both neighborhood unemployment and a summary index of concentrated disadvantage. Perkins and Sampson also find that Blacks face much greater odds than Whites of experiencing compounded (i.e., simultaneous individual and neighborhood level) poverty, even controlling for individual differences in rarely measured factors such as self control, aggression, anxiety, depression, and cognitive ability. Controlling for these background characteristics and initial poverty, the factor most strongly predictive of later compounded poverty was being Black rather than White. These findings underscore the second fact that motivated our 1995 piece: race-based contextual disadvantage—regardless of individual characteristics—continues to characterize the U.S. demographic landscape. In support of this general thesis and based on a nationwide longitudinal study of over 20 million children, Chetty and colleagues (2018, p. 5) find that very few Black children grow up in environments that foster upward economic mobility across generations. They report that “Fewer than 5 percent of black children currently grow up in areas with a poverty rate below 10 percent and more than half of black fathers present. In contrast, 63 percent of white children grow up in areas with analogous conditions.”
It is clear, then, that race still matters for both violent crime and for experiences of structural disadvantage. Our racial invariance thesis is not merely descriptive, however; rather, it integrates two basic claims. First, we posit that community level structural disadvantage is an important cause and thus predictor of violent offending. Second, we argue that this relationship accounts for a considerable portion of the observed racial differences in violent crime and similarly explains variations in violent crime rates within racial groups. In the following sections, we review the empirical assessments of these two claims that have emerged over the last twenty years.
Structural Disadvantage and Crime
The preponderance of evidence strongly supports our claim that structural disadvantage at the community level is an important predictor of violence. This claim emerges intact from “virtually all prior research” (Krivo et al., Reference Krivo, Peterson and Kuhl2009, p. 1794) that align with the appropriate context of validation. Although extant tests rely on observational data rather than experiments for ethical and pragmatic reasons, the findings are nonetheless highly consistent with a casual interpretation. Excellent reviews of these studies may be found in Travis Pratt and Frances Cullen’s (2005) systematic meta-analysis and Ruth Peterson and Lauren Krivo’s (Reference Patterson, Patterson and Fosse2005) more narrative overview of the literature.Footnote 5 In general, the hypothesized link between a community’s structural disadvantage and violent crime is robust across multiple indicators of disadvantage: poverty, female family headship, low education and skill, joblessness, and differing combinations thereof (Peterson and Krivo, Reference Peterson and Krivo2005; Pratt and Cullen, Reference Pratt and Cullen2005).
This linkage is geographically robust as well; empirical work conducted in a wide range of cities points to a strong relationship between disadvantage and violent crime. These cities include Chicago (Morenoff et al., Reference Morenoff, Sampson and Raudenbush2001); Atlanta (McNulty Reference McNulty2001); Miami, San Diego, and El Paso (Lee et al., Reference Lee, Martinez and Rosenfeld2001); St. Louis (Kubrin and Wadsworth, Reference Kubrin and Wadsworth2003); Cleveland, Seattle, and Washington, DC (Crutchfield et al., Reference Crutchfield, Glusker and Bridges1999); Cincinnati (Wooldredge and Thistlethwaite, Reference Wooldredge and Thistlethwaite2003); Columbus, Ohio (Krivo and Peterson, Reference Krivo and Peterson1996); and Baton Rouge, Louisiana (Shihadeh and Shrum, Reference Shihadeh and Shrum2004). These studies also cover a range of mesolevel units: census tracts and block groups, cities and the surrounding metropolitan areas, resident-defined neighborhoods, national time series, or simultaneous combinations of the above.
Finally, existing empirical tests find a link between structural disadvantage and various types of crime: homicide (Krivo and Peterson, Reference Krivo and Peterson2000; Vélez et al., Reference Vélez, Krivo and Peterson2003); other violent offenses, e.g., forcible rape, robbery, aggravated assault (Kubrin and Wardsworth, 2003; McNulty Reference McNulty2001; Shihadeh and Shrum, Reference Shihadeh and Shrum2004; Wooldredge and Thistlethwaite, Reference Wooldredge and Thistlethwaite2003); and even nonviolent property crimes (Chamberlain and Hipp, Reference Chamberlain and Hipp2015).Footnote 6
Disaggregation by Race
Our theory predicts not only that ecological disadvantage contributes significantly to violent crime, but that it does so for both African Americans and for Whites. As we discuss in this section, the empirical evidence largely supports this second hypothesis.
An enormous number of studies test one or two of the three bipartite relationships discussed above between: 1) race and crime; 2) race and structural disadvantage; and 3) structural disadvantage and crime. Perhaps stymied by the lack of overlap in the contextual experiences of African Americans and Whites (Peterson and Krivo, Reference Peterson and Krivo2010a), until recently few studies have examined the intersection of all three of these relationships. An exemplar of this recent body of work is Darnell Steffensmeier and colleagues’ (2010) study of the interaction between macrostructural factors and race (Black, White, and Latino) in predicting homicide and violent crime. Although this study focuses on literal invariance in the values of the coefficients linking structural predictors with crime, a more relevant finding emerges: structural disadvantage is linked significantly with homicide and with violent crime more generally—and these linkages transcend racial boundaries. Put differently, Steffensmeier and colleagues provide no consistent evidence that one set of structural factors leads to violent crime in Black neighborhoods and that a different set is operative in White neighborhoods. Their coefficients by race lie in the same direction and are often statistically indistinguishable from one another. This suggests, in harmony with Sampson and Wilson (Reference Sampson, Wilson, Hagan and Peterson1995), that community level sources of violent crime are consistent across races. When differences do emerge, they are typically in the magnitude of association.Footnote 7 In qualitative terms, therefore, we interpret this evidence as favoring the thesis of racial invariance.
Julie Phillips (Reference Phillips2002) also finds support for racial invariance in her investigation of the White-Latino homicide gap, though she notes that structural factors explain only half of the Black-White homicide gap. Importantly, Phillips’ method of regression decomposition quantifies the relative weights of: 1) the effects of by race differences in ecological circumstances, and 2) racial invariance in the effects of these circumstances. Her approach, however, is still plagued by the lack of overlap in community level predictors between the different racial groups.
Krivo and Peterson (Reference Krivo and Peterson1996) tackle this overlap issue by focusing on Columbus, Ohio—a city with both Black and White communities of high disadvantage. Their findings are consistent with the racial invariance thesis. This is also true of Edward Shihadeh and Wesley Shrum (2004), who find that the relationship between percent Black and violent crime rates among Baton Rouge block groups is eliminated once structural disadvantage and social disorganization are taken into account. Krivo and Peterson (Reference Krivo and Peterson2000) add nuance to these findings, arguing “that theoretically important structural factors may have weaker effects on violent crime [across U.S. cities] when disadvantage is particularly widespread because further increases, above already high levels, may not appreciably differentiate communities” (p. 340). Thomas McNulty (Reference McNulty2001) replicates Krivo and Peterson’s city level findings at the neighborhood level. Maria Vélez and colleagues (2003) adopt a different solution for the lack of ecological overlap; they integrate racial disparities into their independent and dependent variables and find that Whites’ socioeconomic advantages contribute to the homicide gap.
McNulty (Reference McNulty2001) and Peterson and Krivo (Reference Peterson and Krivo2010a) further suggest that any studies that identify weaker relationships between Black ecological disadvantage and crime rates than between White disadvantage and crime rates (e.g., Ousey Reference Ousey1999) have not accounted for the fact that Black neighborhoods lie far above White neighborhoods on the ‘disadvantage distribution.’ A similar point holds at the city level (Krivo and Peterson, Reference Krivo and Peterson2000). Although not focused on racial disaggregation, John Hipp and Daniel Yates (2011) examine the functional form that best captures the positive relationship between ecological disadvantage and crime rates. Using a sample of census tracts from twenty-five U.S. cities and a range of serious crimes, they find that a diminishing positive effect, rather than a threshold effect, most nearly approximates this relationship. These findings may initially seem counter to the functional form of Wilson’s (1987) concept of “concentration effects.” We maintain, however, that Wilson’s concept invokes any contextual effect of neighborhood poverty over and above individual level poverty. Threshold and other nonlinear effects are consistent with the idea of concentration or neighborhood effects but not a necessary implication.
Explaining Racial Gaps in Violence
A clear implication of the racial invariance thesis is that the gap between Black and White rates of violence is socially explicable and not due to intrinsic racial features. Some of the studies reviewed above point to a complete elimination of the racial gap in community level violence when disadvantage is controlled (e.g., Shihadeh and Shrum, Reference Shihadeh and Shrum2004). Others point to a partial explanation. For example, Krivo and colleagues (2009) find that after controlling for neighborhood disadvantage, racial gaps are diminished but rates of violence are still higher in non-White (except Latino) than White U.S. neighborhoods.
Partial explanations also emerge at other levels of analysis. Janet Lauritsen and colleagues (2018) find that the national Black-White gap in serious violent victimization drops considerably (from 71% to 22% greater) but is not entirely eliminated once controls for poverty, urban residence, age, and employment are introduced.Footnote 8 A somewhat larger racial gap in violence, 40%, remains unexplained in Sampson and colleagues’ contextual analysis (2005) after controlling for a range of criminogenic factors including neighborhood disadvantage. Ben Feldmeyer and colleagues (Reference Feldmeyer, Steffensmeier and Ulmer2013) find that both Black and Latino composition effects for census-defined places (e.g., towns and metropolitan areas) in New York and California are only partly explained by controlling for structural conditions (especially structural disadvantage).Footnote 9 Most recently, Michael Light and Jeffrey Ulmer (2016) point to structural disadvantage and segregation as strongly but not perfectly predictive of both the cross-sectional racial disparities in homicide rates and the changes in those disparities over time at the metropolitan level.
James Unnever (Reference Unnever, Unnever, Gabbidon and Chouhy2018) interprets any unexplained “race effect” on crime as evidence against the racial invariance thesis. We disagree with his interpretation for four reasons. First, virtually all studies reviewed evince the predicted and significant drop in the racial gap once theoretically specified structural factors are controlled, with several studies eliminating this gap entirely. Second, even if a racial gap or unexplained residual remains, this does not necessarily imply the presence of racially unique factors. All studies must confront measurement error in the observed covariates (which attenuates their effect), and more importantly, potential omitted variable bias. Central to Sampson and Wilson’s thesis is the argument that race is a marker for multiple forms of disadvantage, including historical and intergenerationally transmitted contexts of poverty. No study of which we are aware has been able to capture the full range of these factors, and yet we still see consistent and large reductions in racial disparities once basic indicators of disadvantage are taken into account. Third, as explained in the above section on context of validation section, we conceive of our thesis in a qualitative spirit rather than in a strictly quantitative one. Fourth, Unnever’s (2018) analysis flows ultimately from the individual level (see also Unnever et al., Reference Unnever, Barnes and Cullen2016) and thus does not constitute a strong challenge to our ecological argument.
QUALIFICATIONS AND QUESTIONS
Our thesis eschews criminology’s traditional focus on individual behavior and instead traces a causal path from macro- to mesolevel phenomena. The relative suitability of various supraindividual units, however, remains an open and fair question: is it most appropriate to apply our theory to violent crime and ecological disadvantage in neighborhoods, block groups, communities, cities, counties, or some other unit? We have cited studies that take on each of these units, and it is not immediately clear which approach is best. In their review of the literature, Peterson and Krivo (Reference Peterson and Krivo2005) point out that neighborhood level analysis has produced the most consistent support for strict racial invariance (i.e., the claim that structural disadvantage predicts crime among Blacks and Whites not only in the same direction but also to a similar degree). They conclude that “cities are poor units for assessing community level processes” (p. 338).
We do not endorse a single answer to the unit of analysis question. Rather, we believe that the focal mesolevel unit of analysis should be driven by the particular research hypothesis—that is, the selection of unit follows from the mechanism(s) driving the theorized relationship between structural covariates and violent crime. For instance, scholars who focus primarily on the mediating impact of the political economy have targeted cities, states, and other bureaucratic units. On the other hand, scholars who drill down on factors such as friend or acquaintanceship networks tend to focus on those units most cognitively salient to residents—contiguous neighborhoods, gang-defined turfs, and the like (e.g., Papachristos et al., Reference Papachristos, Hureau and Braga2013). As we discuss below, these guiding mechanisms are under-examined and thus constitute fertile ground for further research.
Multilevel Analysis
The persistence of a direct association between race and crime in several community level studies that control for structural conditions raises an important question: is there an ecological explanation for this residual racial disparity in violence?
Krivo and colleagues (2009) provide one possible explanation: they show that multilevel or nested analysis—that is, analysis that incorporates variation at multiple supraindividual units—may help to explain at least some of these residual racial disparities. Their study incorporates not only neighborhood level structural variables but also city level measures of racial segregation—and finds that these latter measures account for some of the unexplained racial gap across neighborhoods. The authors posit that city segregation contributes to neighborhood violent crime directly by making it difficult for separate and unequal groups to address shared problems and indirectly by creating structural disadvantage and isolation in predominantly minority areas (Krivo et al., Reference Krivo, Peterson and Kuhl2009, p. 1792). Vélez and colleagues (2015) adopt a similarly nested analytic approach, with two key findings: 1) there is by city variation in the relationship between a neighborhood’s percentage of Black residents and the prevalence of violence; and 2) this “average positive relationship often is attenuated, and reduced to statistical insignificance, in cities with favorable political contexts” (p. 93). These favorable contexts, Vélez and colleagues argue, yield benefits that are both substantive and symbolic, providing a footing for neighborhood based organization against violence.
These nested analyses connect to a broader body of work that focuses on the role of social and spatial contiguity among the focal unit(s) of analysis (Light and Harris, 2012). In particular, Daniel Mears and Avinash Bhati (Reference Mears and Bhati2006, p. 509) argue that resource deprivation in one community promotes more violence, including instrumental violence, in communities that are socially proximate. Peterson and Krivo (Reference Peterson and Krivo2009) expand upon these findings, using neighborhood data from thirty-six U.S. cities to show that proximity to more disadvantaged neighborhoods and especially to racially privileged White areas is critical in accounting for differences in violence found across neighborhoods of different colors. Lallen Johnson and Robert Kane (Reference Johnson and Kane2016) also draw upon spatial data to argue that highly disadvantaged communities that are positioned at the center of a contiguous ghetto have significantly higher violent crime rates than other very underprivileged areas.
Peterson and Krivo’s (2010b) National Neighborhood Crime Study (NNCS) has facilitated such multilevel investigations. NNCS nests neighborhood level contextual data—including racial composition and indicators of ecological disadvantage—within city- and metropolitan-level variables for 9,593 census tracts in ninety-one cities in sixty-four metropolitan areas. Numerous recent studies have leveraged the NNCS to probe the relationships among city level political economy, neighborhood level structural disadvantage, race, and violent crime (Lyons et al., Reference Lyons, Vélez and Santoro2013; Vélez et al., Reference Vélez, Lyons and Santoro2015).
Immigration from Latin America
The explosion of immigration from Latin America and around the world since the 1990s has revitalized inner city neighborhoods and institutions, presenting an important extension of our largely dichotomous Black/White racial invariance thesis (Sampson Reference Sampson2015b). Particularly relevant is the so-called “Latino Paradox” —the counterintuitive coupling of generally high disadvantage with low rates of violence in Latino immigrant communities (see also Krivo et al., Reference Krivo, Peterson and Kuhl2009). It seems that this Paradox deviates from the structural conditions that inspired our original thesis, motivating us to ask: does our thesis apply to immigrants from Latin America?
Research over the last decade suggests a qualified “yes” to this question. First, within Latin American communities, ecological disadvantage seems to predict violent crime in a similar manner as it does for Whites and for African Americans (MacDonald and Sampson, Reference MacDonald and Sampson2012; Martinez Reference Martinez and Hawkins2003). Most recently, Light and Ulmer (Reference Light and Ulmer2016) find that “structural disadvantage is one of the strongest predictors of levels and changes in racial/ethnic [i.e., White, Black, and Hispanic] disparities” in homicide rates (p. 290). Relatedly, Christopher Lyons and colleagues (2013) suggest that the same multilevel analysis that enhanced our understanding of Black/White racial disparities may also help in the area of immigration. By incorporating city level measures into their fundamentally neighborhood level analysis, Lyons and colleagues discover that the “often-found inverse relationship between immigration and crime at the neighborhood level…is generally enhanced in cities with favorable immigrant political opportunities” (2013, p. 604).
Second, a number of scholars add nuance to these findings. Casey Harris and Ben Feldmeyer (2013) argue that the relationship between immigration and violence varies between traditional and non-traditional destinations for new immigrants. Krivo and colleagues (2009) note that it is unclear whether immigration’s protective effect against violence is limited to Latinos—perhaps because of the social capital associated with their ethnic enclaves and community institutions (Vélez Reference Vélez, Peterson, Krivo and Hagan2006)—or whether this protective effect extends to a broader range of racial and ethnic neighborhoods (Martinez and Nielsen, Reference Martinez, Nielsen, Peterson, Krivo and Hagan2006). Sampson (2013, p. 20) adds a longitudinal question to this litany: “Will lower crime rates in the first and second generation endure?” And Light and Ulmer’s (2016) analysis suggests that increased immigration diminishes community level disparities between Black and White rates of homicide.Footnote 10
Underlying Mechanisms
Sampson and Wilson (Reference Sampson, Wilson, Hagan and Peterson1995) provided a theoretical discussion of the social organizational and cultural mechanisms that may explain the effects of structural disadvantage. In their review of the literature, however, Peterson and Krivo (Reference Peterson and Krivo2005) lament the relative dearth of empirical work that tests the mechanisms that link race, ecological disadvantage, and crime. Although a detailed assessment of mediating factors is beyond the scope of the present paper, we briefly review such work and ultimately echo Peterson and Krivo’s call for further investigation (see also Sampson Reference Sampson2013).
Consistent with the implications of our original thesis, empirical research has focused largely on the neighborhood based organizational and cultural mechanisms that link structural disadvantage with violence. Elijah Anderson’s (Reference Anderson2000) Code of the Street provides ethnographic evidence for a cultural code of legitimized interpersonal violence and delegitimized criminal justice processes. This cultural code, Anderson argues, mediates the relationship between poor urban Blacks’ ecological disadvantage and their violent behavior. Robert J. Sampson and Dawn Bartusch (1998) supplement these findings by showing that variations in collective tolerance of crime align with neighborhood characteristics rather than with race per se. Charis Kubrin and Ronald Weitzer (2003) and Baumer and colleagues (Reference Baumer, Horney, Felson and Lauritsen2003) triangulate Anderson’s qualitative research with quantitative data, though in both cases the authors infer the existence of a “code of the street” from the relationship between a neighborhood’s disadvantage and its most prevalent types of crime. Ross Matsueda and colleagues (Reference Matsueda, Drakulich, Kubrin, Peterson, Krivo and Hagan2006) circumvent this limitation by directly operationalizing “codes of violence,” finding that neighborhood impoverishment and racial composition significantly predict the prevalence of such codes. They also find that mistrust of the police—an orientation we more generally conceive of as legal cynicism (Kirk and Papachristos, Reference Kirk and Papachristos2011; Sampson and Bartusch, Reference Sampson and Bartusch1998)—mediates a significant proportion of the effects of percent Black and percent Hispanic on the prevalence of codes of violence. David Kirk and Andrew Papachristos (2011) and Sampson (Reference Sampson, Patterson and Fosse2015a) argue that legal and moral cynicism, which are strongly related to racial segregation and concentrated poverty, are major factors in the persistence of violence in Chicago neighborhoods.Footnote 11
Additional social organizational mechanisms from Sampson and Wilson (Reference Sampson, Wilson, Hagan and Peterson1995) have been tested empirically. A body of evidence provides qualified support for the mediating role of diminished informal social controls or collective efficacy in explaining the effects of concentrated disadvantage on violence (Sampson et al., Reference Sampson, Raudenbush and Earls1997). Research has also engaged with the explanatory power of institutional and/or organizational strength, such as the density of organizations and the networks that connect institutions (Pattillo Reference Pattillo1998; Peterson et al., Reference Peterson, Krivo and Harris2000). These factors contribute to, and are in turn exacerbated by, the documented attenuation of interpersonal trust in majority-Black inner city neighborhoods (Smith Reference Smith2010). Christopher Browning and colleagues (Reference Browning, Feinberg and Dietz2004) complicate matters, however, by presenting evidence that strong local networks may attenuate the regulatory function of collective efficacy (see also Vélez et al., Reference Vélez, Krivo and Peterson2003).
A final set of mechanisms that emerged from our thesis of racial invariance incorporates state policies and the political economy. These mechanisms include the ways in which segregation and concentrated poverty are “embodied in public policy and historical patterns of racial subjugation” (Sampson and Wilson Reference Sampson, Wilson, Hagan and Peterson1995, p. 43; see also Hirsch Reference Hirsch1983). The city- and multi-level studies cited above provide some empirical evidence for these sorts of “higher order” or interlocking structures (Sampson Reference Sampson2012). Support also comes from Shihadeh and Graham Ousey (1998), who find that, between 1970 and 1990, diminished access to low skill jobs augmented economic deprivation in central cities, which in turn increased rates of homicide. Crucially for our purposes, these proximal and distal effects of macrolevel deindustrialization were similar for Blacks and Whites.
Taken as a whole, research on mediating social organizational and cultural mechanisms in the racial invariance thesis is not as extensive as the research on structural covariates. However, these two bodies of research share an important feature: neither points convincingly to the existence of distinct sets of sources for violent crime in Black and White communities.
A RESEARCH AGENDA FOR THE FUTURE
Sampson and Wilson’s (1995) thesis of racial invariance has provoked a vigorous body of research, one that has advanced our understanding of race and crime. As this review has revealed, however, many questions remain unanswered. Although space considerations preclude a full delineation, we would like to briefly summarize the areas that we believe warrant further research.
Foremost on our proposed agenda is more careful elaboration, measurement, and testing of the social mechanisms that connect race with variations in crime at the neighborhood level. We have asserted that race is not a direct cause of violence but rather a marker for the cluster of social and material disadvantages that both follow from and constitute racial status in America. Concepts such as legal cynicism, social isolation, political economy, and collective efficacy have been proposed and measured to test intervening mechanisms that flow from structural disadvantage, but this research remains in its early phases. We must also explore more fully the role of the criminal justice system itself (e.g., policing strategies, incarceration) as an institutional mechanism that potentially exacerbates or diminishes both racial inequality and the racial gap in crime. For example, mass incarceration is argued to have increased racial disparities by imposing barriers to employment and other collateral consequences upon former prisoners (e.g., Travis and Western, 2014)—but it may have diminished other kinds of racial inequality by reducing, even if modestly, the violence that disproportionately takes its toll on Black communities (Sharkey Reference Sharkey2018). Teasing out these complex and perhaps countervailing processes should be high on the research agenda.
We must also carefully explicate the structural disadvantages and cumulative adversities that members of different racial and ethnic groups face. Although Blacks experience concentrated poverty much more frequently and intensely than Whites, we might still ask:
• What are the features and effects of analogous structural dislocations (e.g., family instability, joblessness, drug addiction) in poor White communities?
• Does the racial invariance thesis apply in whole or in part to non-urban contexts, particularly to rural pockets of concentrated White disadvantage?
There is a further need to move beyond the Black-White dichotomy. For example:
• What is the applicability of the thesis to other racial/ethnic groups that have suffered historical disadvantages and discrimination, such as Native Americans?
• What about racial or ethnic groups that might be considered relatively advantaged compared to Blacks, such as Asian Americans?
• Does the supposedly singular “Latino Paradox” apply equally well to other immigrant groups? Will lower crime rates among Latino immigrants persist beyond the first and second generations?
Finally, a number of measurement related matters deserve further scrutiny, including:
• More consistent or precise definitions of “invariance”—what sort of evidence would convince us that the structural predicates of crime vary qualitatively by racial or ethnic group?
• Further research that integrates multiple mesolevel units of analysis (e.g., neighborhoods, census tracts, and cities) in interrogating the racial disparities in crime that remain even after accounting for one of these mesolevel units.
• Is there a linear relationship between ecological disadvantage and the crime rate? Or, given the strong connection of race/ethnicity to neighborhood disadvantage, is the relationship between ecological disadvantage and crime nonlinear, reflecting a threshold, exponential, or diminishing effect?
A NOTE ON SOCIAL POLICY
Sampson and Wilson’s (1995) effort was aimed at theoretical development, devoting only two sentences, at the very end, to policy implications. We concluded that “our framework suggests the need to take a renewed look at social policies that focus on prevention. We do not need more after-the-fact (reactive) approaches that ignore the structural context of crime and the social organization of inner cities” (p. 54). We thus set aside a detailed discussion of policy options and the evidence of their effectiveness. Such an effort demands a separate treatment; there is, in fact, a voluminous literature in criminology devoted to policy evaluations. It is nonetheless worth emphasizing here that the logical perspective of the theory implies the pursuit of both meso- and macro-level social policies in crime prevention rather than solely person based approaches. It also implies policies that are general in nature, even though their results may predictably and disproportionately benefit African Americans in poor neighborhoods.
At the broadest level, the logical structure of our theory points to efforts to directly combat the ecological concentration of disadvantage and break its link to racial segregation. Pursuing this logic, Sampson (Reference Sampson2016) has argued for policies that intervene holistically at the community level by infusing new resources into the existing but disinvested neighborhoods in which the poor reside. Community level interventions, particularly crime based ones (e.g., community policing, neighborhood watch), have produced uneven evaluation results. Certainly, there is no magic bullet. But few neighborhood policies take the long view; most interventions are single site or time constrained, with outcomes measured locally and in the short run. As Patrick Sharkey (Reference Sharkey2013) has argued, there is a need for durable investments in disadvantaged urban neighborhoods to counteract the persistent institutional disinvestments that such neighborhoods have endured over many years.
We also note that the community based enhancements to urban public safety that arose in response to the violence crisis of earlier decades may have disproportionately benefited inner city Black neighborhoods. Sharkey (Reference Sharkey2018) points out that this transformation of public spaces was the most fundamental change that took place in American cities in the last two decades, especially in those neighborhoods that were hardest hit by violence. “Streets that had been abandoned for decades were taken over by police officers, security guards, and community groups,” he states, while “opportunities for criminal activity began to shrink, and violence began to fall” (2018, p. 75). He points in particular to the upsurge in community organizations in the 1990s, and documents the impact of nonprofit organizations on crime reduction in the nation’s largest cities. This endogenous effect of community based institutions on violence—a manifestation of the effect of collective efficacy—is an encouraging sign and may point the way toward new community specific interventions, especially those that work in tandem with the police and other agencies of public control (Bursik Reference Bursik1988; Sampson et al., Reference Sampson, Raudenbush and Earls1997).
A more radical “non-crime” policy is to give cash assistance or to reduce the tax rate for those experiencing compounded deprivation—that is, poor individuals who also live in poor or historically disinvested areas (Sampson Reference Sampson2016). Cash assistance or tax relief (e.g., a negative income tax) could also be combined with job training and sustained efforts to enforce housing laws and physical upkeep of the community. The logic behind such “affirmative action for neighborhoods” is consistent with Sampson and Wilson (Reference Sampson, Wilson, Hagan and Peterson1995): poor individuals who have lived for an extended period in poor neighborhoods have accumulated a set of disadvantages more extreme than those amassed by poor individuals who have been surrounded by the resources of better off neighborhoods. Though technically race neutral, our proposed neighborhood-level investments would disproportionately benefit minorities. And unlike housing voucher programs, this policy approach would allow poor residents to remain in place, if they so desire, while at the same time increasing their available income (Sampson Reference Sampson2016).
Because joblessness is positively associated with violent crime in poor neighborhoods, we also ought to consider macro- or mesolevel policies such as public sector jobs for those who have difficulty finding employment in the private sector. When we speak of public sector jobs, we mean the types of jobs provided by the Works Progress Administration (WPA) during the Great Depression—jobs that would improve the infrastructure in our communities, including the underfunded National Park Service and state and local park districts. Public sector jobs programs would be especially helpful to young Black adults who have been stigmatized by prison records and who thus find it virtually impossible to gain employment in the private sector (Wilson Reference Wilson2009).
We are under no illusions that these programs would garner widespread support in the current political climate. But theoretical analysis does not demand that policy implications be realistic—rather, it requires that such policies be logically implied by and consistent with the theory. Especially given the worrisome spikes in violence in the last few years, we feel that we have to start thinking seriously about what should be done when we have a more favorable political climate, and when people from both parties are willing to consider seriously the type of meso- and macro-level policies of crime prevention we have discussed.
CONCLUSION
In the two decades since “Toward a Theory of Race, Crime, and Urban Inequality” appeared, a growing body of criminological research has tested relevant hypotheses posed by the theory. Based on the evidence that we have assessed in this paper, the sources of violent crime continue to appear “remarkably invariant across race and rooted instead in the structural differences among communities, cities, and states in economic and family organization” (Sampson and Wilson, Reference Sampson, Wilson, Hagan and Peterson1995, p. 41). To be sure, nuances are present: indicators of race continue to have residual “effects” in many studies, there is variability in the magnitude of association between structural ecological conditions and violence by race, and there is recent evidence that some neighborhood factors differentially predict intergenerational economic mobility by race and gender (Chetty and colleagues, 2018).Footnote 12 We have also witnessed a number of new and relevant developments, including the significant crime decline in the United States; the rise of the Black Lives Matter movement and claims for White supremacy; and the influx of immigrants from Latin America and elsewhere. However, we maintain that there is no systematic evidence that one set of neighborhood-level factors explains crime rates for Blacks, while a distinct set applies to Whites—or, for that matter, to Latinos.Footnote 13 What drives crime (and other human behaviors), in our view, remains rooted in fundamental historical and structural conditions that are differentially experienced by racial groups. This contextual assertion anchors our ecological explanation of crime.
We thus continue to place our faith in general theory rather than in the idea that each racial (or ethnic) group necessarily implies a unique set of causal factors and thus a separate theory of crime. That African Americans have endured discrimination and extraordinary hardship, especially as inflicted by the criminal justice system, is indisputable. However, we continue to be impressed by just how much of the racial disparity in violence is explained by a general approach to community-level structural variations. Furthermore, we worry about the dangers of race-specific crime policy in politicized hands. Indeed, the United States has already witnessed the face of such race-specific policy in the late twentieth century “war on crime” (Hinton Reference Hinton2016). Although these concerns make us reluctant to advocate for separate theories of crime by racial group, we welcome serious consideration of the historical and contemporary links between race, racism, and crime. We look forward to the next generation of research that tests the agenda that we have highlighted, and to the design of creative policies that tackle the enduring structural inequalities that disproportionately haunt African Americans.
ACKNOWLEDGMENT
We thank Ruth Peterson for organizing the Presidential Plenary at the American Society of Criminology meeting in New Orleans (2016) upon which this paper is based, and María B. Vélez for serving as a discussant. We also thank Ruth along with Lauren Krivo, Patrick Sharkey, James Quane, and James Unnever for their helpful comments on a previous draft. Finally, we acknowledge the suppport of the Project on Race, Class and Cumulative Adversity at Harvard University, funded by the Ford Foundation and the Hutchins Family Foundation.
