The Enduring Significance of Racism: Discrimination and Delinquency Among Black American Youth

Abstract

Prominent explanations of the overrepresentation of Black Americans in criminal justice statistics focus on the effects of neighborhood concentrated disadvantage, racial isolation, and social disorganization. We suggest that perceived personal discrimination is an important but frequently neglected complement to these factors. We test this hypothesis with longitudinal data on involvement in general and violent juvenile delinquency in a sample of Black youth from a variety of communities in 2 states. We examine the direct effects of concentrated disadvantage and racial isolation and the direct and mediating effects of social organization, support for violence, and personal discrimination. Consistent with our hypothesis, perceived personal discrimination has notable direct effects on both general and violent delinquency and is an important mediator between neighborhood structural conditions and offending; moreover, its effects exceed those associated with neighborhood conditions.

The overrepresentation of Black American youth in crime statistics is well documented. In the 2000 census, approximately 16% of American youth identified themselves as Black, and according to the 2006 Uniform Crime Reports, African American youth accounted for 59% of arrests for murders, 51% of arrests for violent crimes, 31% of property offenses, and 30% of arrests for drug use (Federal Bureau of Investigation, 2006). This disproportionality is not limited to offending: Adolescent Black American males’ homicide victimization rate is about seven times higher than the rate for White adolescents (Baum, 2005).

Many of the adolescents represented in these statistics are from the inner city, and much of the research on race and juvenile crime has focused on the conditions of these neighborhoods; yet, many of the young Black Americans arrested and incarcerated live outside of the inner city. Several studies document considerable similarity between urban and nonurban areas in the patterning of crime and delinquency (see Osgood & Chambers, 2000), and scholars have hypothesized that theories that explain inner-city crime may account for crime in nonurban settings (Laub, 1983). Many criminologists attribute the overrepresentation of Black Americans in criminal justice statistics to structural racism. Structural racism refers to an array of historical and contemporary conditions that have helped create inner-city communities characterized by racial segregation, poverty, residential instability, and low levels of social control, conditions that contribute to high rates of offending in these communities.

While structural racism is certainly important for understanding Black Americans’ offending, we add to a small group of studies that examine the consequences of perceived personal discrimination (Caldwell, Kohn-Wood, Schmeelk-Cone, Chavous, & Zimmerman, 2004; Gibbons, Gerrard, Cleveland, Wills, & Brody, 2004; Martin, 2005; McCord & Ensminger, 2003; Simons, Chen, Stewart, & Brody, 2003; Simons et al., 2006; Stewart & Simons, 2006). Personal discrimination refers to the unequal, harmful treatment of a person because of their minority status, by an individual or individuals from a dominant group (Feagin & Eckberg, 1980).

We extend previous research on the positive association between perceived personal discrimination and delinquency in several ways. First, although some prior research on personal discrimination and crime controls for neighborhood structural disadvantage and the structural racism it may engender, we provide a theoretical argument to support the claim that the effect of perceived personal discrimination on delinquency will typically exceed that associated with structural racism. Second, we examine the direct effect of perceived personal discrimination on delinquency as well as its role as a mediator between neighborhood conditions and delinquency. Third, we evaluate the importance of personal discrimination in a model that controls for neighborhood social organization. Fourth, we assess the extent to which the relationship between discrimination and delinquency is reciprocal and reflects the differential treatment youth may receive, not because of their race, but because of their prior involvement in crime. Finally, we investigate these relationships with three waves of data from the Family and Community Health Study (FACHS), a longitudinal study of Black American families living in suburban and rural areas in two states. We use these data to examine the connections between neighborhood structural conditions and crime among a group often neglected in research: Black American youth who live outside of the inner city.

MACROLEVEL FACTORS: CONCENTRATED DISADVANTAGE AND RACIAL ISOLATION

Several theories of crime and delinquency focus on neighborhood economic disadvantage and racial composition (Pratt & Cullen, 2005). For example, Shaw and McKay (1942) emphasize neighborhood socioeconomic status and ethnic heterogeneity, while Blau and Blau (1982) highlight relative deprivation and racial and economic inequality. Sampson and Bean (2006) and Sampson and Wilson’s (1995) racial invariance thesis builds on these works and underscores the role of two factors: concentrated disadvantage and racial segregation. Concentrated disadvantage has two dimensions: the increasing convergence of an array of negative economic and social conditions that includes joblessness, welfare dependency, poverty, family disruption, and residential instability and the concentration of these conditions in specific geographical areas (Wilson, 1987). Sampson and Wilson note that concentrated disadvantage increased dramatically in inner-city communities in the post-1970 period, especially in Black American communities.

Sampson and Wilson argue that these and other historical and contemporary conditions have created many segregated inner-city communities. This racial isolation is a product of a structural racism—or what some scholars calls racism without racists (Bonilla-Silva, 2006; Ford, 2008)—that restricts Black Americans’ access to housing, employment, and other neighborhood resources. According to Sampson and Wilson, higher crime rates, and in particular higher violent crime rates, in some inner-city Black American neighborhoods are consequences of this racial segregation and deleterious economic conditions.

Consistent with Sampson and Wilson’s explanation, Pratt and Cullen’s (2005) meta-analysis of research on macrolevel variables or “neighborhood effects” finds that racial composition (i.e., the percent non-White or Black), economic deprivation, and family disruption are among the strongest and most stable predictors of crime rates. Although Sampson and Wilson focus on variation in community characteristics and crime rates, they (1995, p. 44) suggest that the neighborhood factors they highlight also contribute to the race–crime link among individuals. Research by Sampson, Morenoff, and Raudenbush (2005) and by others offers evidence of these macromicro links. For example, in an analysis of data from the National Longitudinal Study of Adolescent Heath, McNulty and Bellair (2003) found that significant differences between Black and White adolescents’ involvement in violence disappear with controls for neighborhood disadvantage (unemployment, poverty, and female-headed households). Other studies found evidence of the negative effects of neighborhood disadvantage in settings other than the inner city (Liska, Logan, & Bellair, 1998; Osgood & Chambers, 2000). These results are not surprising; as McNulty and Bellair (2003, p. 714) note, the logic of theories of racial isolation and concentrated disadvantage “is rooted in general principles of social relations—not in distinctions between urban and other settings and hence should apply to all types of communities.”

MEDIATING VARIABLES: SOCIAL ORGANIZATION AND CULTURAL ADAPTATIONS

Sampson and Wilson highlight several factors that intervene between racial segregation, concentrated disadvantage, and crime. Consistent with Shaw and McKay, they argue that economic deficiencies decrease a community’s social organization, that is, its ability to realize the common values of its residents and maintain effective social control. Key components of social organization include informal connections with neighbors and neighborhood organizational participation (Sampson & Groves, 1989), as well as collective efficacy (Sampson, Raudenbush, & Earls, 1997). Collective efficacy involves two interrelated community-level social processes: mutual trust or a sense of connectedness (social cohesion) and a willingness to intervene in community problems (informal social control). Sampson and Wilson argue that these neighborhood disadvantages also contribute to Black Americans’ social isolation from normative society. This seclusion has encouraged the development of another mediating variable: a cognitive landscape that accepts offending, including violent crime as viable means for addressing inequality or expressing frustration about it.

Sampson and Wilson do not detail the specific features of the cognitive landscape that supports crime; however, their brief description suggests that it shares many features of the “code of the street” described in Anderson’s (1999) ethnographic research in Philadelphia. According to Anderson, neighborhood racial segregation, economic disadvantages, and social isolation provide few opportunities for residents and especially youth, to gain status or respect in conventional arenas such as school or work. Many of these neighborhoods are further characterized by high levels of criminal victimization and low probabilities that authorities will provide protection. As a result, many inner-city youth, and, in particular, many Black American males, exaggerate the importance of being respected as a source of status. The emphasis on respect encourages a code of the street that legitimizes the use of violence not only as a means of self-protection, but also as a way to gain and maintain status. This code of the street also accepts that nonviolent crime and delinquent behavior (e.g., drug selling and theft) are acceptable means of achieving income, respect, and other desirable outcomes. As Anderson (1999, p. 75) notes, “[o]ne way to campaign for status is to take the possessions of others.” Anderson’s argument about crime and the code of the street resonates with Sampson and Wilson’s description of an inner-city cognitive landscape and other explanations of crime that focus on cultural or subcultural adaptations (see Brezina, Agnew, Cullen, & Wright, 2004; Matsueda, Drakulich, & Kubrin, 2006).

Several studies examine the extent to which Anderson’s thesis can be generalized to other communities. Matsueda et al. (2006) found that in Seattle neighborhoods support for the code is strongly related to the percentage of Black residents in a community; however, Black respondents did not significantly differ from Whites in their support for the code. Moreover, the effect of community economic disadvantage on the code is nonsignificant with controls for neighborhood racial composition. In their analysis of males surveyed in the first three waves of the National Youth Survey, Brezina et al. (2004) found a significant relationship between violence and a scale that contains items that measure respondents’ support for the use of physical retaliation. Yet, Brezina and colleagues also reported a nonsignificant coefficient for the relationship between urbanism and support for these beliefs, suggesting that support for the code may develop in nonurban and urban areas (also see Stewart & Simons, 2006).

PERCEIVED PERSONAL DISCRIMINATION AND DELINQUENCY

Perceived personal discrimination may contribute to crime directly and as a factor that intervenes between neighborhood conditions and offending. Personal discrimination differs from structural or group discrimination in that the latter refers to the negative treatment of members of one’s group (Bourguignon, Seron, Yzerbyt, & Herman, 2006): Being called a racist name is an example of personal discrimination, whereas the belief that members of one’s racial group are discriminated against in job interviews reflects an awareness of group discrimination. Sampson and Wilson (1995) and Anderson (1999) recognize the pernicious effects of discrimination, but focus mostly on the effects of structural discrimination.

Our hypothesis that perceived personal discrimination influences offending is based on three observations. First, perceived personal discrimination increases negative emotions and has deleterious consequences for an array of outcomes, including depression, anxiety disorders, high blood pressure, and other mental and physical health outcomes (Schnittker & McLeod, 2005; Williams, Neighbors, & Jackson, 2003). Although the relationship between perceived personal discrimination and crime is not widely researched, a handful of studies document a connection. McCord and Ensminger (2003) found a positive association between adolescent reports of racism (e.g., harassed by police and been in trouble with teachers) and adult arrests for robbery, assault, and other violent crimes in a sample of 1,242 grade nine Chicago youth. Using data from 325 adolescents surveyed in Flint, Michigan, Caldwell et al. (2004) reported that for both females and males, involvement in violence significantly increases with perceived discrimination. Previous research that uses the data that we analyze also documents a positive association between discrimination and delinquency (Gibbons et al., 2004; Martin, 2005; Simons et al., 2003; Stewart & Simons, 2006). These findings are consistent with several criminological theories. Agnew’s General Strain Theory (2001) argues that negative treatment from others is one type of strain that can generate negative emotions such as anger, frustration, and resentment. Shoplifting, vandalism, violence, and other crimes may relieve these feelings by fulfilling a need for revenge. Sherman’s (1993) theory of defiance argues that treatment perceived as unjust can lead to defiant crimes that reassert one’s individuality and independence. Bernard’s Angry Aggression Theory (1990) argues that situations that engender anger, combined with the inability to respond to the actual sources of that stress, increase the tendency to transfer aggression to more accessible and immediate targets. Although these theories differ in important ways, their logic suggests that personal discrimination will increase negative emotions that contribute to offending.

Second, although race relations have improved considerably in the last 50 years in the United States, personal discrimination remains widespread. Studies on housing searches, employment applications, requests for mortgages and credit, insurance applications, the provision of medical care, and the submission of discrimination claims (Pager & Shepherd, 2008; Quillian, 2006), as well as self-report surveys (Schnittker & McLeod, 2005), indicate that personal racial discrimination still occurs with an alarming frequency. This discrimination is not directed solely toward adults, but affects male and female Black youth of all ages from a wide array of neighborhoods and socioeconomic classes: Caldwell et al. (2004) reported that 82% of the Black American youth they surveyed had experienced at least one racially discriminatory episode during the past year (seven was the median value). As Essed (1991) noted, the frequency of these acts of suggests that many Black Americans experience what she calls “everyday racism.”

Third, the negative effects of everyday racism on crime may exceed those associated with other forms of racism, especially structural racism. Attributional research finds that discounting negative treatment from others decreases as hostility increases in frequency, duration, and intensity (Branscombe, Schmitt, & Harvey, 1999). Personal discrimination conveys an intense desire to exclude or reject and typically provokes an array of negative feelings about one’s self. In contrast, group discrimination can enhance an individual’s self-image, in part because it can strengthen identification with a persecuted group (Branscombe et al., 1999). In this way discrimination may operate like deprivation: personal deprivation typically tarnishes an individual’s self-image, whereas group deprivation encourages solidarity and collective responses (Bourguignon et al., 2006). In a survey of African immigrants, Bourguignon et al. (2006) found that self-esteem is negatively related to perceived personal discrimination, but positively associated with group discrimination. In a study of African Americans, Branscombe et al. (1999) reported that witnessing discriminatory treatment has a direct, negative effect on personal well-being, but this effect is reduced by minority group identification (also see Sellers & Shelton, 2003). They (1999, p. 138) conclude that “[b]ecause exclusion by the dominant group is painful, inclusion and identification with one’s minority in-group may serve as an alternative means of protecting well-being.”

Although we posit a positive association between personal discrimination and delinquency, we recognize that other factors may moderate discrimination’s negative consequences. For example, Hughes and Chen (1997) reported that many parents encouraged their children to identify with positive aspects of their African American culture. This racial socialization may counteract the negative effects of discrimination, while other processes, such as supporting a street code, may put youth at greater risk for these effects.

INTEGRATING STRUCTURE, CULTURE, AND PERCEIVED PERSONAL DISCRIMINATION

Figure 1 presents a conceptual model of the hypothesized relationships suggested by the preceding discussion. Our model begins with two exogenous structural variables: neighborhood concentrated disadvantage and racial isolation. The second set of variables in our model highlights the three mediating processes described earlier: social organization, support for a code of the street, and personal discrimination. Consistent with Sampson and Wilson’s work, we include paths from neighborhood concentrated disadvantage to social organization (also see Sampson, Morenoff, & Gannon-Rowley, 2002). We include a connection between concentrated disadvantage and personal discrimination because prior research reports a positive association between neighborhood poverty and racism (Franzini, Caughy, Spears, & Fernandez Esquer, 2005). Following Anderson and Sampson and Wilson, we expect concentrated disadvantage to have a positive effect on endorsement of a code of the street and on delinquency.

FIGURE 1.

Conceptual model of community conditions, discrimination, and delinquency.

Although some analyses include neighborhood racial characteristics in their measure of concentrated disadvantage (Sampson et al., 1997), other research suggests that these conditions may have opposing effects (Peterson & Krivo, 1993; Shihadeh & Shrum, 2004). We separate these variables because they likely have different consequences for the variables in our model. In contrast to concentrated disadvantage, living in a neighborhood in which Black Americans are in the majority may encourage more frequent interactions and greater participation in activities that promote social organization; at the same time, it may reduce interactions with non-Black Americans and thereby reduce opportunities for discrimination (Hunt, Wise, Jipguep, Cozier, & Rosenberg, 2007). Thus, our model includes a positive effect of racial isolation on social organization and a negative relationship between racial isolation and discrimination. We expect racial isolation to increase support for a code of the street and delinquency. We anticipate that the social organization will have the opposite effect and will negatively influence offending.

The remainder of our analytical model is a three-wave, three-variable autoregressive model with cross-lagged effects between personal discrimination, street code, and delinquency. The cross-lagged model allows us to control for prior delinquency and street code in our test of discrimination’s effect on change in delinquency over time. We also assess the extent to which support for a code of the street predicts delinquency while controlling for personal discrimination and prior delinquency. This part of the model tests our key theoretical prediction that discrimination uniquely contributes to delinquency and that it is an important mediator of the effects of structural conditions on offending. Drawing on Anderson’s thesis that alienation increases support for attitudes that legitimize violence, we expect that discrimination will predict later support for a street code. In addition to the hypothesized paths in the conceptual model (Figure 1), we test competing explanations by estimating paths from delinquency to subsequent discrimination and endorsement of a street code as well as street code’s effect on discrimination. Offending may encourage youth to develop a code that justifies their prior actions, and it may increase the likelihood of being in situations where youth are more likely to encounter discrimination. Likewise, youth who adopt a street code may be more likely to experience discrimination. Our analytic model assumes, and exploratory analyses confirmed, that concentrated disadvantage and racial isolation affect second and third wave discrimination and delinquency primarily through their influence on first wave measures of these variables. Additionally, we estimated correlations between all concurrent constructs (e.g., the correlation between concentrated disadvantage and racial isolation in 1990 and the correlations between wave 1 variables).

In sum, we argue that perceived personal discrimination contributes to delinquency, directly and as a mediator of the effects of structural conditions. We test these assertions in a model that also considers social organization and cultural adaptations as additional mediators of the effects associated with neighborhood structural disadvantage. In other words, our model examines how structural conditions, social processes, cultural adaptations, and personal experiences influence changes in involvement in delinquency over time.

Building on Previous FACHS Findings

As previously noted, research that uses the data we analyze documents positive associations between discrimination and delinquency (Gibbons et al., 2004; Martin, 2005; Simons et al., 2003; Stewart & Simons, 2006). Our analysis builds on these earlier studies, especially those of Stewart and Simons, in several ways. First, we provide a theoretical explanation for why the effects of perceived personal discrimination may exceed the effect associated with structural discrimination, an explanation absent in earlier work. Second, we hypothesize that two neighborhood characteristics, racial isolation and concentrated disadvantage, will have contrasting rather than similar consequences for racial discrimination and other variables in our model. Third, we examine the relationship between discrimination and delinquency with measures from three points in time. Only two waves of data were available at the time of Stewart and Simons study. Fourth, we use formal tests to examine the strength and significance of mediating effects rather than relying on change in main effects as evidence of mediation. Fifth, we examine general and violent delinquency rather than focusing on one type of delinquency. Finally, we use structural equation modeling (SEM) to conduct our analyses. The estimated regression coefficients in SEM are corrected for measurement error in the observed variables, errors that may have unduly influenced the results obtained in prior regression analyses, especially when the variables differ in the extent of measurement error (Blalock, 1982).

METHOD

Participants and Procedure

The FACHS is a longitudinal study of Black American families in Iowa and Georgia. The FACHS used characteristics of census block groups—clusters of blocks within a census tract—to identify neighborhoods in which Black American families lived. Most census tracts include four to five block groups. During the 1990 census, block groups averaged 452 housing units with 1,100 residents. The communities included in the FACHS varied considerably in demographic characteristics, particularly racial composition, economic level, and urbanism (from metropolitan areas and suburbs to rural farm communities). Despite the variation across the neighborhoods in the sample, the families from Iowa and those from Georgia were very similar. There were no significant differences between the families from Iowa and Georgia in either gross household or per-capita family income or in the percent of the community residing below the poverty line; moreover, where significant differences were found, the differences were relatively small in magnitude. For instance, the percent of single mother households in the neighborhoods of study families averaged 13.2% in Iowa and 12.1% in Georgia, while the average years of education for primary caregivers was 12.8 and 12.2, respectively.

In Iowa, school officials provided names and addresses of all Black Americans in the fifth grade living in these neighborhoods. Georgia schools did not participate in the study; instead, community members used information from parents, teachers, pastors, and community organizations to compile rosters of youth aged 10 through 12 within each block group. In 1997, a random sample of about 650 youth and a caregiver were selected in each state; just under 75%of Iowa families (475) and 65% of Georgia families (422) participated.

In each state, focus groups of 10 Black American women reviewed the study’s instruments and suggested changes. Eight families in each state also participated in a pilot test of the revised instrument, and their feedback resulted in additional modifications. A group of Black American university students and community members trained to collect the data met respondents in their homes to obtain informed consent, and a week later returned to conduct the interview. Interviewers administered self-report questionnaires separately to youth and to a primary caregiver. The interviewers read questions aloud and recorded participants’ responses on laptop computers. Over 85% of respondents and caregivers completed the second (n = 781) and third (n = 767) waves of the study in 1999 and 2002. Analyses indicated that the families who did not participate in 1999 or 2002 did not differ significantly from those who did with regard to family income, primary caregiver educational attainment, or youth age, gender, or delinquency. The youth respondents were between 10 and 12 years of age at the first wave (M = 10.51), between 12 and 14 years of age at the second wave (M = 12.46), and between 14 and 17 years old when they completed the third wave (M = 15.61). Fifty-five percent of the respondents were female.

Measures

Neighborhood variables

We used census data to measure our two exogenous variables. We measured concentrated disadvantage as a latent variable with indicators that link a respondent’s home address to a census block group. We used 1990 sample weighted census data at the block group level to create percent scores for four commonly used indicators of neighborhood concentrated disadvantage (Pratt & Cullen, 2005; Sampson et al., 2005): percent unemployed adult males, percent families with children living below the poverty level, percent single mother households, and percent families receiving public assistance. Block group is the smallest geographic unit for which the Census Bureau tabulates 1990 sample data, and the smaller the geographic area, the more closely it resembles what most would consider a neighborhood. As Hipp’s (2007) analysis suggests, block groups may be the best level for examining the consequences of neighborhood economic and related conditions.

We created our second neighborhood measure, racial isolation, with data on the percentage of Black Americans who lived in the block group where the respondent resided. Percent Black is a commonly used measure of neighborhood racial composition in studies of neighborhood effects on crime (Pratt & Cullen, 2005). Massey and Denton (1988) used it as a measure of isolation, one of the five dimensions of racial segregation they identified. Researchers commonly aggregate data from smaller geographical areas to create segregation indicators for a larger area; however, we used data from the block group level, so there are no smaller units to aggregate. City-level racial isolation is typically measured with an isolation index (i.e., the percentage of Blacks living in the tract of the average Black American); our measure of racial isolation used information only on the respondent’s actual neighborhood (at the block group level).

Neighborhood social organization

Our first intervening variable was measured using primary caregivers’ reports of community deviance, informal social control, and social cohesion as indicators of a neighborhood social organization latent construct. We assessed a neighborhood’s risk for deviance with a scale composed of seven items that asked caregivers to evaluate how much of a problem things such as drinking in public, gang violence, and graffiti are in their community (Cronbach’s α = .90; a full list of items and response categories for all our measures is available upon request). Collective efficacy involves two interrelated community-level social processes: informal social control and social cohesion. We measured informal social control with a scale of three items based on caregivers’ reports of monitoring children and youth in their community (Cronbach’s α = .82). Our measure of social cohesion was based on caregivers’ responses to 16 dichotomous questions about their ties with neighbors (Cronbach’s α = .89).

Perceived personal discrimination

We measured our second intervening variable with questions adapted from the Schedule of Racists Events (Landrine & Klonoff, 1996). Members of minority and dominant groups often disagree about whether particular actions are discriminatory. This conflict reflects the inherently subjective nature of discrimination, each group’s members’ motivations for interpreting behaviors in a particular way and the tendency to attribute negative outcomes to unequal treatment. As a result, researchers use measures of discrimination that ask specific questions about a variety of life experiences (Schnittker & McLeod, 2005). At each wave, youth reported how often (never, once or twice, a few times, several times) they had experienced each of 10 types of discrimination. Items included questions such as, “How often has someone yelled a racial slur or racial insult at you just because you are African American?” and “How often has a store owner, sales clerk, or person working at a place of business treated you in a disrespectful way just because you are African American?” We randomly parceled the discrimination items into three indicators of the latent construct at each wave (Cronbach’s α = .83, .87, and .88). Parcels offer three advantages over the use of individual items: They typically produce more stable solutions, they are less likely to share specific sources of variance that may not be of primary interest, and they reduce the likelihood of spurious correlations (Little, Cunningham, Shalar, & Widaman, 2002).

Code of the street

We measured our third intervening variable with an eight-item scale developed by Simons et al. (2003) and Stewart and Simons (2006). The scale assessed the extent to which youth agreed (strongly agree to strongly disagree) with statements about the necessity of using violence in a variety of contexts, including to achieve respect, resist exploitation, defend one’s rights, and avoid appearing weak. We randomly parceled the items into three indicators of the latent construct at each of the three waves (Cronbach’s α = .73, .78, and .79).

General and violent delinquency

Our analysis focuses on two self-report measures of delinquency: a 17-item scale of general delinquency and a subscale of seven violent offenses. Our measure of general delinquency covers a broad range of acts, from serious violent behaviors, such as hurting someone with a weapon, to violations of rules and status offenses, such as staying out past curfew and skipping school. The nonviolent items in the scale ranged in seriousness from lying about money or responsibilities, truancy, and breaking curfew to vandalism, shoplifting, breaking and entering, and fraud. The seven violence items included bullying, fighting, assault, extortion, and assault with a weapon. At each wave, youth indicated the extent of their involvement in each activity in the year before the survey. We randomly parceled all 17 items into three indicators of general delinquency at each wave (Cronbach’s α = .66, .65, and .70; a general delinquency scale that excluded lying, truancy, and breaking curfew produced effects comparable to those reported below; however, the effect of social organization on general delinquency was not significant with the smaller general delinquency scale). For the violence subscale, the subset of violence items were recoded as dichotomies because of their more limited variation and then summed into three parcels at each wave (Cronbach’s α = .65, .53, and .66).

Analyses

We used SEM to test the conceptual model depicted in Figure 1. The FACHS data contain individuals nested within Census block groups; thus, families are clustered within communities, and observations within any one community are not independent. Additionally, discrimination and both general and violent delinquency were positively skewed. Thus, we estimated our models using Mplus 5.21 (Muthén & Muthén, 1998–2007) and maximum likelihood estimation with standard errors robust to nonindependence of observations and violations of normality (MLR). Although Mplus’s MLR estimator produces estimates of standard errors robust to violations of normality, observed variables must still be continuous, unless otherwise specified. Our general delinquency and discrimination parcels were continuous variables and thus estimated in this manner. However, violent delinquency consisted of fewer items, and the violence items were among the least likely to be endorsed; thus creating parcels for violent delinquency in the same manner that we did for general delinquency resulted in variables with distributions that were not continuous. Because of their more limited variation, we recoded these items as dichotomies and summed the items into three parcels at each wave. These parcels have only three or four possible categories and are highly skewed toward zero; thus they are ordered categorical, rather than continuous, variables and were treated as categorical variables (Kline, 2005). Each violent delinquency parcel was specified as an observed ordered categorical dependent variable, and the MLR estimator was used to fit the violent delinquency models.

We used full-information maximum likelihood to estimate parameters. This approach fits a covariance structure model to each observation’s raw data rather than to a covariance matrix among observed variables, and thus no information is lost because of missing data. This approach provides more consistent, less biased estimates than listwise or pairwise deletion (Arbuckle, 1996). We used the delta method to calculate the standard errors of mediating effects (Sobel, 1982).

RESULTS

Descriptive Results

The youth in the sample lived in a wide variety of neighborhoods: The percent of families living below the poverty level in the Census block groups in which the youth lived ranged from 0% to almost 64%; single mothers comprised between 0% and 40% of families in these neighborhoods; and the proportion of families on public assistance varied between 0% and 46% (complete descriptive statistics are available upon request). Likewise, the racial composition of the communities in which the respondents resided ranged from 1% to 97% Black American. Eighteen percent of respondents lived in communities that were highly racially isolated (i.e., more than 70% Black).

Scores for general delinquency indicated that a sizable proportion of the sample reported some delinquency: 38% at wave one, 66% at wave two, and 64% at wave three. The average score for delinquency increased significantly across the three waves of the study (M = .06, .12, and .14), and a nontrivial minority of the sample admitted to three or more delinquent acts at each wave: 13% in 1997, 26% in 1999, and 31% in 2002. Violent delinquency was less common: 11% of respondents reported using any violence in 1997, as did 16% in 1999, and 12% in 2002.

Reports of perceived personal discrimination were common for these youth (M = .54, .55, and .64), and, although the means for females and males differed significantly for individual items, there were no significant mean gender differences on the discrimination scales at any of the three waves. Percentage distributions revealed that at the first wave about 66% of these 10–12-year-olds reported that someone had said something insulting to them because they were African American, 40% said that someone had yelled a racial slur or insult at them, and a third indicated that they had been treated disrespectfully in a store or business. Less common experiences reported at the first wave included being threatened by a person (17%) and being harassed by police (6%) because of one’s race. In 2002 a smaller proportion of respondents indicated that they had been threatened because of their race (13%), but about one-quarter of the sample reported a negative experience with the police. The proportion of the sample that experienced a specific type of discrimination rose significantly from the first to the second wave of the study for 3 of 10 types, for 8 types from the second to the third wave, and for 7 types from the first to third wave. For example, reports of police harassment increased from 6% to 11% and 26% across the three waves of the study.

Collectively, these findings suggest that the majority of Black American youth experience some form of perceived personal discrimination and that for over 80% of these youth, some form of discrimination occurred before they reached the age of 13. Although some types of discrimination occurred more frequently and to a larger proportion of Black American youth than to others, none of the experiences we measured occurred to only a small minority of the youth who were surveyed. Moreover, these percentages may be conservative: Prior research finds that people perceive lower levels of personal discrimination than group discrimination, perhaps because individuals minimize personal experiences to maintain positive views about themselves (Bourguignon et al., 2006).

Structural Equation Models of General and Violent Delinquency

Our tests of the model described earlier progressed in three steps. First, we examined the factor loadings. All factor loadings were significant, in the expected direction and of relatively large magnitude. These findings affirmed the usefulness of the variables selected to measure our latent constructs. Second, we estimated full models for general and violent delinquency, freeing all of the aforementioned parameters (results available upon request). In our third step we dropped paths that were not statistically significant (p < .05, one-tailed test) and used model fit tests for competing nested models to assess improvement in model fit (Satorra & Bentler, 1999, provide the formula and procedures for computing a χ² difference test when using MLR). In contrast to earlier analyses of these data (Simons et al. 2003; Stewart & Simons, 2006), our results indicated that aside from its autoregressive effects, support for a street code did not significantly influence violent or general delinquency at either wave, or any of the other variables in the model. Further modeling revealed that dropping the street code construct did not reduce the fit of the model, and the Bayesian Information Criterion (BIC) and Akaike Information Criterion (AIC) were considerably improved with this change (e.g., for the general delinquency model the BIC was 50,322.248 for the model with code of the street included and 38,422.445 when it was dropped). Given this construct’s nonsignificant effects we dropped support for a street code from our final models.

Our final trimmed models are presented in Figures 2 (general delinquency) and 3 (violent delinquency). Fit indices affirm that the model of general delinquency fit the data well: The Root Mean Square Error of Approximation (RMSEA) is less than .06 and the Comparative Fit Index (CFI) is greater than .95 (Hu & Bentler, 1999). RMSEA, CFI, and TLI are not available for models with categorical indicators estimated with MLR; however, the BIC and AIC indicated the trimmed violence model fit the data better than did the full violence model. An alternative violence model that treated violent delinquency as a manifest variable with one indicator also demonstrated good fit to the data (RMSEA = .025; CFI = .982).

FIGURE 2.

Trimmed model of community conditions, perceived personal discrimination, and general delinquency (standardized coefficients).

As predicted by Sampson and Wilson’s thesis, our results for general delinquency indicated that neighborhood concentrated disadvantage in 1990 had a significant negative effect on social organization in 1997 (β = −.18; Figure 2) and was positively and significantly related to discrimination (β = .23) and delinquency in 1997 (β = .09). As well, social organization in 1997 had a significant and negative effect on general delinquency in 1999 (β = −.08); however, this was its only significant effect. Thus, while greater social organization appears to decrease future delinquency, it did not substantially deter discrimination or endorsement of a street code. There was also support for our hypothesis that concentrated disadvantage and racial isolation have different consequences. Although concentrated disadvantage and racial isolation were significantly correlated (r = .49), community racial isolation in 1990 had a significant, negative effect on discrimination in 1997 (β = −.14), but its effects on social organization and on delinquency were negligible.

The stability paths for both perceived personal discrimination and delinquency were sizable and significant. For example, discrimination in 1997 significantly predicted discrimination in 1999 (β = .55), and delinquency in 1997 was significantly associated with delinquency in 1999 (β = .42). Consistent with our hypothesis, our cross-lagged effects revealed that discrimination in 1997 was associated with significant relative increases in delinquency from 1997 to 1999 (β = .19); moreover, discrimination in 1999 predicted relative increases in delinquency from 1999 to 2002 (β = .11). The reverse was not true: The cross-lagged effects of delinquency on discrimination were not statistically significant and thus were dropped from the model. Collectively, these results suggest that personal discrimination is an important influence on delinquency among African American youth.

Our mediation analyses further demonstrated the importance of personal discrimination in understanding African American adolescents’ delinquency. Four of six indirect effects involving discrimination mediating the structural effects of racial isolation and concentrated disadvantage on delinquency were significant: the indirect effect of concentrated disadvantage on general delinquency in 2002 through discrimination in 1997 and general delinquency in 1999 (p = .003), concentrated disadvantage’s indirect effect on general delinquency in 1999 through discrimination in 1997 (p = .002), racial isolation’s indirect effect on general delinquency in 2002 through discrimination in 1997 and general delinquency in 1999 (p = .036), and the indirect effect of racial isolation on general delinquency in 1999 through discrimination in 1997 (p = .029). In comparison, none of the indirect effects involving social organization mediating the effect of concentrated disadvantage were significant, although social organization did have a significant indirect effect on general delinquency in 2002 through delinquency in 1999 (p = .049).

Parallel to the results for general delinquency, racial isolation was significantly and negatively associated with discrimination in our model of violent delinquency (β = −.15; Figure 3); as well, concentrated disadvantage was significantly and positively associated with discrimination (β = .22). However, unlike our results for general delinquency, neither concentrated disadvantage nor social organization significantly predicted violent delinquency.

FIGURE 3.

Trimmed model of community conditions, perceived personal discrimination, and violent delinquency (standardized coefficients).

Similar to general delinquency, our model of violent delinquency revealed a notable relationship between discrimination and violence over time: discrimination in 1997 significantly predicted relative increases in violent delinquency from 1997 to 1999 (β = .17), and discrimination in 1999 predicted relative increases in violent delinquency from 1999 to 2002 (β = .21). While discrimination had a sizable effect on violence, the reverse was not true: Violence did not significantly predict experiences of discrimination.

The three indirect effects involving discrimination’s mediation of the effects of concentrated disadvantage on violent delinquency were all significant: the indirect effect of concentrated disadvantage on violence in 2002 through discrimination in 1997 and in 1999 (p = .009), concentrated disadvantage’s indirect effect on violent delinquency in 2002 through discrimination in 1997 and violent delinquency in 1999 (p = .047), and concentrated disadvantage’s indirect effect on violence in 1999 through discrimination in 1997 (p = .006). There was only one significant mediating effect of discrimination for racial isolation’s effect on violent delinquency: the effect of racial isolation on violence in 1999 through discrimination in 1997 (p = .047).

We conducted three additional analyses as a way of gauging the robustness of our findings. First, we examined the correlations between the manifest variables of interest across the three time points by gender. Sampson and Wilson focused mostly on male offending, and the inclusion of females in our analyses may have led to weaker effects of some of the variables in our models. Yet, the correlations revealed considerable similarity between females and males (correlations not displayed, but available upon request). We then conducted multiple group analyses to determine if any of the paths in our models differed by gender. Equality constraints were placed on sets of parameters in a cumulative fashion, χ² values for competing nested models were compared, and a nonsignificant difference in χ² values indicated that the constrained parameters were not statistically different in magnitude for females and males. The results of these multiple group analyses indicated there were no significant gender differences in the structural parameters of the models. Finally, we conducted another set of multiple group analyses to test for differences between the families from Iowa and those from Georgia in the same manner as our tests for gender differences. We found no significant differences between the families from Iowa and Georgia in the structural parameters of the models of general or violent delinquency.

DISCUSSION

The overrepresentation of Black American adolescents in crime statistics demands explanation. Previous research and theory have provided important insights and highlight the ways in which structural racism, economic conditions, community characteristics, and cultural responses have contributed to the alienation that many Black Americans feel and to the crime that plagues many inner-city communities. This research suggests that Black Americans’ higher crime rates are a function of their greater exposure to criminogenic conditions: concentrated disadvantage, low collective efficacy, and a cultural landscape divorced from mainstream society. Thus, some scholars predict that the crime rates for Black Americans would more closely resemble those of White Americans if a greater number of Black Americans lived in communities that more closely resembled those of White Americans.

Improving the structural conditions of Black American communities will likely diminish Black Americans’ overrepresentation in crime statistics. However, it will not address a condition that continues to distinguish Black and White Americans: Most White Americans rarely, if ever, experience racial discrimination. Ford (2008, p. 337) reminds us that although the last 50 years have seen dramatic shifts in race relations in the United States, racism is still with us:

And old-school racism of the virulent and hateful variety, unafraid of its own shadow, [still] rears its unwelcome head in posh uptown boulevards, leafy suburban lanes, and gritty ghetto alleyways alike more than occasionally … [This] post-racism … can look an awful lot like racism of the pre-civil rights era vintage.

We have argued that perceived personal discrimination is an important but frequently neglected component in understanding the overrepresentation of Black Americans in criminal justice statistics. Our analysis supports our key hypothesis: Discrimination was significantly related to violent as well as general delinquency and was an important mediator of the effects of concentrated disadvantage and racial isolation on delinquency. Importantly, discrimination predicted change in delinquency over time in our models, whereas delinquency did not significantly influence later experiences of discrimination.

Our study found that processes emphasized in research on inner-city crime also influence delinquency among Black American youth who live in nonurban areas. Concentrated disadvantage appears to operate similarly, but not identically, in nonurban environments as in the inner city: in our study concentrated disadvantage was associated with greater involvement in general but not violent delinquency. Although racial isolation was strongly correlated with concentrated disadvantage, we did not find a significant effect of racial isolation on offending.

We also investigated the mediating effects of social organization. Consistent with the ideas of Sampson and Wilson, our results reveal that economically disadvantaged communities have more limited possibilities for social organization. We also found a negative relationship between social organization and general delinquency but not violence. Overall, these results suggest similarities between nonurban and inner-city environments.

We did not find significant associations between support for a street code and offending, violent or otherwise. Earlier analyses of the data we used demonstrated that code of the street does have a significant, cross-sectional relationship with violent delinquency (Stewart & Simons, 2006) and conduct disorder (Simons et al., 2003); however, we did not find any longitudinal effect of support for street-code on general or violent delinquency. Simons et al.’s (2003, 2006) analyses included other independent variables (e.g., family type, violent peers) that we exclude, and it is possible that the effect of code of the street is suppressed unless these variables are controlled. Alternatively, the association between code of the street and offending may be more short-lived when contrasted to the relationship between discrimination and delinquency.

Although the results of this study did not support our hypotheses regarding the code of the street, the consistent effect of discrimination on delinquency highlights the necessity of programs that aim to reduce discrimination. As well, efforts to increase awareness among community leaders, merchants, teachers, and police of the detrimental effects of discrimination are critical. There is a long history of negative sentiment and tension between police and Black American communities in this country (see Hagan, Payne, & Shedd, 2005). Perhaps a greater understanding of how personal discrimination contributes to offending—that we truly do reap what we sow—would help to reduce discriminatory practices. As well, we must continue to research and develop more effective strategies for dealing with discrimination and the negative outcomes it engenders.

Subsequent research will hopefully build on our findings. Our study focuses on Black American youth who live in suburban and rural areas of two states. Future research that uses nationally representative data will be able to assess the extent to which the patterns we observe apply to youth in other communities and from other racial backgrounds. Measures of concentrated disadvantage and racial isolation from different geographical units (we use Census block groups) and with a shorter time lag between surveys may refine the effects of neighborhood concentrated disadvantage and racial isolation on crime.

Studies that involve more families from each neighborhood will also extend our results. Our social organization indicators consist of parents’ reports of their neighborhood; thus, our analysis treats scores from a small number of families as characteristic of their entire neighborhood—and in some cases from only one family. This may have reduced the variation in our explanatory variables and the power to detect their effects (see Osgood & Chambers, 2000).

Our measure of perceived personal discrimination relies on adolescents’ interpretations of their interactions with others; studies with more complete data on the perceptions of the discriminators, as well as details about the discriminator and the context in which the discrimination occurred, will improve the measurement of personal discrimination. Future research should also explore the various factors that contribute to personal discrimination.

Our understanding of the effects of discrimination on delinquency will also be enhanced by research that includes measures of the variables that intervene between perceived personal discrimination and delinquency. We argue that discrimination contributes to an array of emotions from anger to shame that often contribute to crime, but we have not examined the influence of these mediating factors. Previous research that examines the data used here found that anger intervenes between discrimination and offending (Simons et al., 2003, 2006); however, additional research is necessary if we are to improve our understanding of the connections between discrimination, emotions, and offending.

Subsequent research should also investigate factors that may moderate the relationship between personal discrimination and delinquency. The data we present indicate that most of the Black American youth surveyed believe that they had experienced some type of racial discrimination; yet for most youth, this treatment does not lead to delinquency. Instead, many youth find other, perhaps less destructive ways to cope with the discrimination they experience. Future research should examine factors such as racial socialization that interact with personal discrimination to intensify or weaken its effect on offending.

Our focus on the relationship between perceived personal discrimination and crime does not deny that the repeal of Jim Crow laws, the passing of civil rights legislation, and the end of formal racist practices in government, at school, and at work dramatically improved the lives of many Black Americans. The last 50 years have witnessed notable increases in the proportion of Black Americans who complete high school, attend and graduate from college, obtain middle-class jobs, and own homes. Moreover, public opinion surveys suggest that the United States has become less racist over time. For example, in 1963, 62% of European Americans surveyed by the National Opinion Research Center supported laws prohibiting marriage between European Americans and Negroes and 60% agreed that European Americans had the right to keep Negroes out of their neighborhoods; by 1996, only 13% of respondents agreed with such restrictions (Schuman, Steeh, Bobo, & Krysan, 1997). In many settings it is now socially unacceptable to act in overtly racist ways (Ford, 2008; Jackman, 1994).

Although these changes in attitudes toward race are impressive, surveys of Black Americans report that personal racism remains part of the American landscape. Some critics argue that this individual racist behavior typically arises unconsciously because of an implicit-bias based on racial stereotypes (Lane, Kang,& Banaji, 2007) or the use of an inadvertently racist shorthand in situations of incomplete information (Loury, 2002). Others maintain that particular issues or situations explain racism, as in the case when people endorse stereotypical racist beliefs when they argue against programs that address racial inequality (Bobo & Kluegel, 1993) or discriminate when they feel threatened by a racial minority (St. John & Heald-Moore, 1996). Yet it is unlikely that unconscious stereotypes, misinformation, and fear can account for the frequency and extent of this discrimination; instead, racial animosity likely continues to play a significant role in encouraging discrimination (Jackman, 1994).

Approaches to crime that ignore this reality may overestimate the benefits from increased contact with representatives of the majority culture because they underestimate how individuals from the dominant culture contribute to Black American offending. Unless attitudes and behaviors continue to change, contact with these individuals may increase the frequency of personal discrimination and may contribute to rather than discourage crime. Indeed, focusing solely on structural racism minimizes the role that individuals and racial animus play in offending among Black Americans and other racial minorities. Barak Obama’s (2008) inspiring speech on race relations in the United States reminds us of the reality of everyday racism:

[W]hat ails the African-American community does not just exist in the minds of black people; … the legacy of discrimination—and current incidents of discrimination, while less overt than in the past—are real and must be addressed

Racism of this sort is clearly different from the racism of Jim Crow; but this does not mean its effects are benign, particularly when it is an “everyday” experience.