The Impact of Early Racial Discrimination on Illegal Behavior, Arrest and Incarceration among African Americans

Abstract

The prospective relations between perceived racial discrimination (PRD), assessed at four different time periods from childhood through adolescence, along with assessments of PRD from the police (“hassling”), and self-reports of arrest and incarceration at mean age 24.5, were examined in a sample of 889 African Americans from the Family and Community Health Study. Multiple covariates were included in the analyses (e.g., academic orientation, SES, self-control). Structural equation modeling revealed relations between PRD, especially that assessed in childhood, and both arrest and incarceration reported in adulthood. Mediators of these relations included deviant affiliation and self-reports of both substance use and illegal behavior. PRD from the police directly predicted subsequent illegal behavior. Racial pride moderated reactions to both types of PRD: Persons high in racial pride reported more illegal behavior after PRD from police, but less illegal behavior in the absence of perceived police discrimination, and less illegal behavior overall. Finally, childhood PRD, but not adolescent PRD, directly predicted incarceration that occurred up to 14 years later, and it did so controlling for arrest, self-reported illegal behavior, and other covariates. The importance of childhood PRD experiences and possible avenues of intervention suggested by the pattern of results are discussed.

Rates of criminal behavior and incarceration in the US are higher among African Americans (Blacks) than European Americans (Whites) (Kochel, Wilson, & Mastrofski, 2011). Early explanations of this disparity, almost all from sociologists and criminologists, focused on structural factors, including SES, neighborhood risk, and inequities in the criminal justice system, all of which have been associated with higher levels of offending and arrests among Blacks (Wikström & Loeber, 2000). More recent studies have indicated that perceived racial discrimination (PRD) is an important and previously overlooked social factor associated with Black adolescent delinquency and young adult offending. In fact, Unnever, Cullen, Mathers, McClure, and Allison (2009) claimed that underestimation of the impact of PRD on offending in earlier research (e.g., Hirschi, 1969), in essence, misled a generation of criminologists who, as a result, did not focus on PRD as a contributing factor.

Since the publication of Unnever et al., links between PRD and delinquency among Black adolescents have been found in many studies (Jones & Greene, 2016), including several from the Family and Community Health Study (FACHS; Burt & Simons, 2015; Burt, Simons & Gibbons, 2012; Evans, Simons & Simons, 2016). These studies have suggested that the PRD/delinquency relation may be stronger for Blacks than for other minority groups (Unnever, Barnes, & Cullen, 2016; Unnever & Gabbidon, 2011). In addition, although comparisons of PRD’s criminogenic effects with those of other stressors are rare, one study has suggested that the impact of PRD on criminal behavior may exceed that of environmental factors, such as neighborhood risk, which is a strong predictor of criminal behavior (Martin et al., 2011).

Critical Periods

Timing of the discriminatory experiences appears to be an important factor in predicting future behavior. Black children are aware of their racial status as early as age 6 or 7, and have reported discriminatory experiences by the age of 9 (Dulin-Keita, Hannon, Fernandez, & Cockerham, 2011). Moreover, these early experiences appear to be especially consequential: PRD before age 13 has a stronger negative relation with well-being (mental health) in Black children than do similar experiences later in life (Lee & Ahn, 2013), including cumulative PRD effects throughout adolescence (Benner et al., 2018). Schmitt, Branscombe, Postmes, and Garcia (2014) have suggested this may reflect the fact that children lack the resources necessary to effectively cope with threats to their well-being and their identity—at a time when that identity is still developing (Sellers, Copeland-Linder, Martin, & Lewis, 2006). Others have made similar arguments regarding the mental, physical, and behavioral impact of early PRD on Black adolescents (Brody et al., 2013; Mays, Cochrane & Barnes, 2007).

Source of Discrimination

Discrimination is endemic in the US (and other countries), and it exists at multiple levels—economic, institutional, societal. As recent events (e.g., the highly-publicized deaths of multiple Black men and boys) have made salient, however, one particularly consequential source of discrimination for Black adolescents is the police. In fact, Unnever, Owusu-Bempah and Deryol (2019) recently reported that Black adolescents have more trouble with the police than do adolescents of other minority groups—controlling for multiple factors that usually predict such troubles (e.g., previous offending, impulsivity, gang membership, SES level). Moreover, several studies have documented an “iatrogenic” effect of early contact with the criminal justice system: being stopped by police may increase the likelihood of future delinquent / criminal behavior and arrest, especially among Blacks (Bernburg & Krohn, 2003; Del Toro et al., 2019; Wiley & Esbensen, 2013). This prospective effect also maintains controlling for personality factors, such as impulsivity and aggressiveness (Lopes et al., 2012). Why this is the case is not clear, however. The current study looked at short- and long-term reactions to early perceived police “hassles” that the adolescent attributed to his/her race. This was done controlling for prior illegal behavior and self-control, and also more general PRD (not from police), thereby allowing for comparison of the effects of two different sources of PRD on criminal behavior and arrest/incarceration.

Mediation

Having established a link between PRD and offending behavior, more recent examinations by psychologists have focused on variables that mediate and moderate the relation. As might be expected, an important factor is social: affiliation with others who are involved in delinquent behavior. PRD has been associated with increases in the likelihood that Black adolescents will seek out the company of peers who are engaging in deviant behavior (Roberts et al., 2012; see Benner et al., 2018, for a review). One explanation for this is that rejection by the majority culture (i.e., discrimination) increases a desire to affiliate with others who have also been rejected, and who are rejecting that culture themselves (Unnever & Gabbidon, 2011; Whitbeck, Hoyt, McMorris, Chen, & Stubben, 2001). This “social path” to illegal behavior and criminal involvement is common in adolescence (Dodge, Dishion, & Lansford, 2006), including among Black adolescents (Simons & Burt, 2011; Unnever et al., 2016).

Once again, timing is relevant: Deviant affiliation early in adolescence has been shown to have a stronger prospective connection with delinquent behavior than does similar affiliation later on (Evans et al., 2016; Monahan, Steinberg, & Cauffman, 2009). Similarly, Del Toro et al. (2019) reported that the younger a minority male was the first time he was stopped by police, the more likely he was to engage in delinquent behavior six months later. In short, discriminatory experiences, including discrimination from the police, that occur early--when both self- and group-identity are developing and coping resources are limited-- may have a disproportionate impact on the illegal behavior of Black children. This hypothesis has not been directly tested in longitudinal research.

Moderation: Racial Pride (RP)

Racial pride is a multidimensional construct typically defined as maintaining a positive perception of one’s racial group and having the belief that members of that group are worthy of respect from others (Sellers, Smith, Shelton, Rowley, & Chavous, 1998). African American RP is usually instilled during adolescence through racial socialization, and has been associated with several positive outcomes, including academic success, increased self-esteem, improved psychological well-being, and reduced substance use (Miller-Cotto & Byrnes, 2016; Phinney & Chavira, 1992; Thompson & Gregory, 2010). High RP has also been shown to buffer against the harmful effects of PRD (Sellers, Caldwell, Schmeelk-Cone, & Zimmerman, 2003), in part, because it attributes the cause of racism more to the perpetrator than the victim (Major, Quinton, & McCoy, 2002). However, RP also appears to be associated with more vigilance for PRD (Sellers & Shelton, 2003), especially among males (Lee & Ahn, 2013), and also more reaction to slights and offenses directed against the in-group (Marino, Negy, Hammons, McKinney, & Asberg, 2007). In short, it appears that RP can be both a buffer and a risk factor: acting as a psychological resource, in general, but also potentially increasing the likelihood of problems after PRD.

Arrest vs. Incarceration: Accountability

For a variety of reasons, the relation between criminal behavior and criminal justice is less than perfect, and that is especially true for Blacks (Leung, 2005). Not only are Blacks arrested at higher rates than Whites (Kochel et al., 2011), but, after arrest, they are more likely to be incarcerated (Starr & Rehavi, 2013), and for longer periods of time, controlling for type of crime (Mustard, 2001; Spohn, 2013). According to the National Academy of Science (Travis, Western, & Redburn, 2014), racial differences in accountability--the extent to which arrests account for (match) incarceration--have increased since 2000. Put simply, Blacks are more likely to be jailed after arrest. This is especially true for low level crimes, such as drug trafficking and assault, suggesting discrimination may also play a role in sentencing after arrest. The current study examines the relation between PRD and self-reports of both arrests and incarceration, controlling for self-reports of “deviant” values, and multiple self-reports of illegal behavior, which have been shown to be valid—i.e., they correlate with and/or predict actual arrest and incarceration (Leung, Woolly, Tremblay, & Vitaro, 2005; Roberts & Wells, 2010).

Overview

Six waves of FACHS data were used to examine the prospective relations of self-reports of PRD--overall and from the police--with illegal behavior, and then arrest and also incarceration (hereafter referred to as “jail”), the latter two being reported at age 24/25. The first analysis compared the predictive effect of PRD with four other common sources of stress previously linked with jail: low SES, low parental monitoring, neighborhood poverty, and victimization. The second analysis included Structural Equation Models of the PRD → arrest/jail relations with multiple controls, including the four stressors, plus individual differences known to be related to arrest (e.g., risk-taking, low academic orientation, deviant values). Self-reported illegal behavior at ages 21/23 was included as an endogenous construct. Finally, racial pride was examined as a moderator of reactions to PRD and police hassle. Hypotheses:

H₁) Self-reports of PRD—especially early PRD-- will predict arrest and jail, controlling for multiple stressors shown to be predictive of these outcomes among adolescents (Ou & Reynolds, 2010).

H₂) This PRD effect will be mediated by: affiliation with deviant peers (the social path), more severe self-reported illegal behavior, and more substance use--all three enhanced by the PRD.

H₃) Reports of police PRD (hassle) will predict subsequent arrest, controlling for the mediators. This effect will be moderated by RP: more negative reaction to police hassle for those high in RP, especially males. Less illegal behavior overall and after PRD (buffering) was also expected from High RP persons.

Finally, exploratory analyses examined the pathways to arrest and jail separately, with a goal of examining a possible role of PRD in incarceration after arrest (i.e., accountability).

Method

Sample, Recruitment, and Procedure

Sample.

FACHS is an ongoing study of psychosocial factors related to the mental and physical health of African Americans. The original sample comprised a panel of 889 Black families, half from Iowa and half from Georgia. Each family had an adolescent who was in 5^th grade at W1 (M age = 10.5 at W1, 24.5 at W6) and self-identified as African American or Black, along with his/her primary caregiver (“parent”). Most of the parents (92%) and 54% of the adolescents were female; 84% of the parents were the adolescents’ mothers; 55% of them were single mothers. The sample was representative of the demographic: Black families in lower/middle SES, nonurban neighborhoods in Iowa and Georgia.

Recruitment.

Families were recruited in 1997 from 259 block group areas in small metropolitan areas and suburbs in Iowa and Georgia that varied in racial composition. School liaisons compiled lists of all families in the area that included a fifth-grade Black child; families were then chosen randomly from those lists. Data were gathered from 72% of the families on the lists. Those who declined to participate usually cited the amount of interview time (up to 3 hours per wave) as the reason. More description of the FACHS sample can be found in Gerrard, Gibbons, Stock, VandeLune, and Cleveland (2005).

Procedure.

All interviewers were Black; most lived in the communities where the study took place. They received extensive training (up to four weeks) in interview technique. The interviews were conducted in participants’ homes or nearby locations and required two interviewers and one or two visits. For privacy reasons, questions were presented using the Computer Assisted Personal Interview (CAPI) technique; there was also a structured diagnostic assessment (the DISC-IV). Compensation ranged from $100 at W1 to $195 at W6. Average time between interviews was 24 months for W1-W2, 36 months for W2-W3, W3-W4, and W4-W5, and 29 months for W5-W6. Assent was obtained from participating adolescents at each wave until age 18; after that, informed consent was obtained from them. All procedures were approved by the relevant university IRBs.

Measures: Focal (Measurement wave noted in parentheses for each construct).¹

PRD (W1 to W4).

Participants completed a 12-item modified version of the Schedule of Racist Events (Landrine & Klonoff, 1996). This measure, which is commonly used in the PRD literature (Pascoe & Richman, 2009), describes various discriminatory events and asks how often the respondent has experienced each event because of their race; e.g., “How often has someone said something insulting to you just because you are African American?” (from 1 = never to 4 = several times; αs for all waves > .86). The modifications involved altering some vocabulary in the early waves due to the age of the children.

PRD from police (i.e., “hassle;” W3, W4).

Because of its direct relation with the hypotheses and the outcomes, one item from the original PRD scale was removed and included as a separate (discrimination) construct: “How often have the police hassled you just because you are African American?” (from 1 = never to 4 = several times). The W3 and W4 versions were used in analyses because, unlike PRD, few participants reported experiencing any police hassle at W1 or W2 (see below).

Illegal behaviors (W5, W6).

Participants were asked if they had engaged in 11 common illegal behaviors in the past year. These behaviors were categorized into 6 groups according to severity; e.g., 0 = no behavior; 1 = steal something cheap; 2 = damage property… 5 = shot/stabbed someone (the full list is provided in the Supplemental Material). The value for the most severe behavior reported in each wave was used and then severity of those behaviors in the two waves was averaged.²

Arrest and jail (W6).

Arrest and jail were lifetime self-reports (adapted from Copes & Hochstetler, 2003). The scale for arrest was: 0 = never, 1 = once, 2 = twice, 3 = 3–5 times, 4 = ≥ 6; for incarcerated, it was: 0 = never, 1 = once, 2 = twice, 3 = three times, 4 = ≥ 4 times. In addition, although it couldn’t be included in the SEMs (due to sample size and its redundancy with arrest / jail), those who had been sent to jail were also asked about the crime for which they had been sent. An index was then created by coding the offense in terms of severity (e.g., 1 = probation/parole violation; … 4 = assault or violent act). This index was included in other analyses (see below).

Measures: Controls (W1, W2, W3)

Stress.

In an effort to isolate the effects of PRD and also to control for the impact of confounding factors, 11 constructs that have been previously linked with PRD, arrest/jail, and/or illegal behavior were included as covariates. The first four were stress-related: W1 SES was the parent’s education and income (α = .61); parents’ reports of child monitoring included nine items, e.g., how often they know where the child is and what s/he is doing (α = .61); environmental stress had two factors: nine items on parent and adolescent criminal victimization; e.g., been.. robbed, molested, a victim of a violent crime (α = .52 and .40, respectively); Neighborhood poverty comprised six questions from the census (e.g., vacant housing, unemployment, % families below the poverty line; α = .73).

Individual differences.

Three personality constructs frequently linked with delinquency were included at W2: risk-taking (six items adapted from Eysenck & Eysenck, 1978; e.g., “You enjoy taking risks;” α = .75); academic orientation (eight items; e.g., “School bores you,” “Grades are very important to you” α = .71), low self-control, which is a strong predictor of criminal behavior (Gottfredson & Hirschi, 1990) (10 items from Kendall & Wilcox, 1979; e.g., “you usually think before you act,” “you can deliberately calm down when you are excited…” α = .72). Deviant values (W3) were assessed by asking participants how wrong they thought it was for someone to engage in seven illegal behaviors (e.g., steal something, hit someone with intent to injure; from 1 = very wrong to 4 = not at all wrong (Menard & Elliot, 1994); α = .92. Prior lifetime delinquent behavior (W3) was measured with 13 items from the Diagnostic Interview Schedule for Children (DISC-IV; Shaffer et al., 2000); e.g., ever engaged in: property damage, stealing, assault (α = .71). “Offending variety” scales like this (that include multiple behaviors) have been shown to be effective predictors of future criminal behavior (Sweeten, 2012). Controlling for these W3 items meant the W5/W6 reports provided some evidence of change in illegal behaviors from age 15 to 21/24. Early puberty (W1), which has been linked with both PRD (Seaton & Carter, 2019) and delinquency (Cota-Robles, Neiss, & Rowe, 2002), was self-reported with five items (e.g., body hair, growth spurt; Ge et al., 2006; αs = .68 for girls, .62 for boys). Gender was also included as a control.

Measures: Mediators and Moderators

Mediators.

A core list of mediators of the PRD → arrest/jail relation was created based on previous research examining both PRD and arrest/jail.³ Those mediators included the following. For deviant affiliation (W3, W4), participants were asked how many of their close friends engaged in 12 different deviant behaviors; e.g., stole something expensive, attacked someone with a weapon (1 = None of them, 2 = Some, 3 = All; αs: W3 = .83 and W4 = .80). Substance use (W4) included four items about use in the last year: two drinking items, “have you had a drink in the last year” and “during the past 12 months, how often had you a lot to drink, that is 3 or more drinks at one time;” plus one smoking item (“have you smoked cigarettes in the last year?”) and one marijuana item (“have you used marijuana in the last year?”) (all coded as yes/no, α = .61). Finally, self-reported illegal behaviors were included as an endogenous construct, serving as both an outcome of PRD (as mentioned earlier) and as a mediator of PRD effects on jail/arrest.

Moderator.

Racial pride (W3) was assessed with the Multi-Construct African American Identity Questionnaire (Smith & Brookins, 1997), which contains 21 items; e.g., Black people should be proud of their race; I like living in a Black neighborhood (1 = strongly disagree to 4 = strongly agree; α = .79).⁴

Results

Analysis Plan

There are four analysis sections: a) Means and correlations of the primary measures. b) A path analysis of relations between W6 jail and the four stressors: (low) parental monitoring, (low) SES, neighborhood poverty, and victimization, plus PRD, all measured at W1. c) Structural Equation Models (SEMs) predicting arrest and jail by PRD. PRD was assessed at all time periods; however, full results are presented only for the SEM with the first two PRDs: W1 (PRD1) in childhood and W2 (PRD2) in early adolescence. Also included were social/behavioral reactions to PRD and self-reported illegal behavior as mediators of the PRD → jail/arrest relations. These were followed in some cases by sensitivity analyses (including regressions) to further examine hypothesized paths. d) Moderation of the relations found in the SEMs by RP. The path analysis and SEMs both used MLR estimators to address potential violation of multivariate normality, and FIML to handle missing values (MPlus, V. 7; Muthen & Muthen, 2012). Modification indices were used to identify significant paths that were not originally hypothesized (specified) in the SEMs. Indirect (mediated) effects are presented when informative.⁵

A. Means and Correlations (see Table 1)

Table 1:

Means and Correlations of Primary Measures

	VARIABLE	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21
1	PRD 1
2	PRD 2	.47***
3	PRD 234	.42***	.79***
4	PRD-POL 3	.20***	.24***	.40***
5	RP 3	.03	.05	.12***	−.03
6	GENDER	−.05	−.08*	−.04	.22***	−.13***
7	SES 1#	−.11***	.00	.03	.02	.07	−.03
8	MONIT 1#	.16***	.00	.02	.10*	−.23***	.12***	−.09*
9	NEIPOV 1#	.09*	.01	−.04	.00	.00	−.02	−.34***	.02
10	VICTIM 1#	.17***	.13***	.14***	.03	−.09*	.00	−.08*	.03	.06
11	PUBTY 1#	.09**	.10**	.13***	.03	.10**	.00	.08*	−.04	−.02	.01
12	RISK 2	.11***	.09*	.11***	.15***	−.02	.09*	.14***	.10**	−.11***	.08*	.05
13	ACAD 2	−.10**	−.01	.01	−.11**	.13***	−.12***	.14***	−.27***	−.07	−.03	.02	−.18***
14	POORSC 2	.20***	.17***	.13***	.04	−.11***	.02	−.04	.18***	.01	.07	.06	.39***	−.32***
15	DELBEH 3	.07*	.08*	.14***	.19***	−.05	.05	.01	.14***	−.11***	.08*	.03	.32***	−.21***	.20***
16	DEVVAL 3	.07	.10*	.11***	.19***	−.11***	.13***	−.01	.05	−.05	.01	.01	.17***	−.12***	.11***	.23***
17	DEVAFF 3	.17***	.15***	.24***	.22***	−.05	−.01	.01	.12***	−.07	.05	.00	.28***	−.17***	.22***	.44***	.31***
18	SUBUSE 4	.07	.06	.18***	.12***	.03	−.01	.09*	.04	−.08*	.06	.05	.18***	−.09*	.11***	.22***	.15***	.20***
19	ILBEH 56	.18***	.14***	.17***	.20***	−.08*	.15***	−.02	.11***	.00	.09*	.00	.15***	−.15	.12***	.20***	.15***	.28***	.18***
20	ARREST 6	.11**	.08*	.13***	.25***	−.04	.24***	−.12***	.11***	.06	.08*	−.03	.17***	−.20***	.13***	.19***	.17***	.18***	.25***	.31***
21	JAIL 6	.14***	.11**	.12***	.23***	−.06	.22***	−.11**	.09*	.05	.05	−.03	.17***	−.20***	.13***	.18***	.14***	.16***	.23***	.32***	.89***
	M	1.67	1.67	1.74	1.41	3.56	0.46	0.00	0.00	0.00	0.00	0.01	1.46	3.12	1.68	0.84	1.33	1.36	0.21	0.97	0.97	0.88
	sd	0.56	0.59	0.50	0.77	0.32	0.50	0.84	0.76	1.00	0.73	1.00	0.43	0.40	0.36	1.54	0.60	0.29	0.27	1.38	1.24	1.30
	Range	1–4	1–4	1–4	1–4	1–4	0,1	−1.56–4.6	−1.01–3.37	−2.02–2.66	−.46–4.95	−2.34–3.27	1–3	1–4	1–3	0–12	1–4	1–3	0–1	0–5	0–4	0–4

Note: Higher scores indicate more of the construct. Number after variable name indicates wave of data.

^#= Standardized. PRD = Perceived Racial Discrimination, excluding Police question, PRD 234 = PRD averaged over waves 2, 3, and 4, PRD-POL = Police “Hassle,” RP = Racial Pride, GENDER = 1 for Males, 0 for Females, MONIT = Parental Monitoring, NEIPOV = Neighborhood Poverty, VICTIM = Victimization, PUBTY = Puberty, RISK = Risk Taking Behaviors, ACAD = School Orientation, POORSC = Poor Self Control, DELBEH = Delinquent Behavior, DEVVAL = Deviant Values, DEVAFF = Deviant Affiliation, SUBUSE = Substance Use, IL BEH = Illegal Behavior severity averaged over waves 5–6

^*p < .05,

^**p < .01,

^***p < .005

At W1, some PRD was reported by 90% of the sample. In contrast, the figures for police hassle at W1/W2 were only 4%/8% for once or twice, and 2%/3% for a few experiences or more. By W3, the hassle numbers had risen to 15% once or twice and 11% a few times or “frequently.” Percentages reporting delinquent behavior (a control) by W3 were: None = 64%, 1 or 2 incidents = 25%, and > 2 = 11%. Percentages reporting illegal behaviors in the last year at W5/W6 were: None = 65%/76%, 1 or 2 = 16%/12%, and > 2 = 19%/12% (decline in these behaviors from age 21 to 24 is typical; Snyder & Mulako-Wangota, 2018). For arrest, the W6 lifetime figures were: None = 54%, once = 16%, and > once = 30%. The W6 lifetime jail figures were: none = 61%, once = 14%, and > once = 26%. At W5, participants were also asked how old they were the first time they were arrested. Four (of 687 who answered) reported being arrested by age 11; the numbers for other early ages were: age 12 = 6, 13 = 14, 14 = 21, 15 = 32, 16 = 43, 17 = 77. By W6, the total number arrested was 320 or 46% of the responding sample.⁶

Several correlations among the covariates and the focal variables are worth noting. All three PRD measures plus cumulative PRD234 (the mean of PRD at W2, 3 and 4) were correlated with W5/6 illegal behaviors, both number and severity (all rs > .14, ps < .001). RP was correlated positively with academic orientation (Miller-Cotto & Byrnes, 2016) and with PRD234 (both ps < .001), and negatively with illegal behavior (p = .03), poor self-control, and deviant values (both ps < .003). Puberty (early maturation) was positively correlated with PRD1 and PRD2 (both ps < .007; cf. Seaton & Carter, 2019); but, it was not correlated with illegal behavior, deviant values, or arrest or jail (all rs < .02, NS). W3 delinquent behavior was positively correlated with PRD1, PRD2, and PRD234 (all ps < .05), as well as W3 police hassle, W5/W6 illegal behavior, arrest, and jail (all ps < .001). Finally, all of the control variables were correlated with PRD, and most of them were correlated with the outcomes (arrest and jail).

B. Prospective Relations of PRD and Other Stressors with Jail (see Figure 1)

Figure 1:

Path Analysis of Wl Stressors Predicting Incarceration

Note: W1 PRD = Perceived Racial Discrimination, excluding Police Hassle, reported at W1 Jail = number of times incarcerated.

^aW1 PRD path to Jail without covariates or controls had coefficient .13**(3.19) and with covariates and controls had coefficient .10*(2.19). See text for explanation.

W1 stressors that are significantly correlated with each other are depicted in the Figure. Controls are correlated with each other and exogenous variables in the model. Controls are also predictors of jail.

See text for details on the indicators that make up all latent variables.

All coefficients are standardized with z values in the parentheses.

* p<.05; ** p<.01; ***p<.001.

A path analysis examined the prospective relations between the four stressors plus PRD1 and W6 reports of jail. All five stressors and the three controls were specified as manifest constructs. The coefficient for just the PRD1 → jail path without the other stressors or any of the controls can be seen on the left side of that path (β = .13; z = 3.19, p = .001). Adding in the controls and the other stressors reduced the strength of that relation, but it was still significant (β = .10; z = 2.19, p < .03). Regarding the other stressors, numerous studies have linked low SES and jail (e.g., Ou & Reynolds, 2010; cf. Moffitt, 2006), and that was the case here (β = −.09; z = −2.76, p = .006). In contrast, monitoring, neighborhood poverty, and parent and adolescent victimization did not predict jail significantly (all βs < .05, ps > .35). These measures were all correlated with PRD, however, so they were retained, along with SES and other constructs, as controls in the SEMs. In short, consistent with H₁, early PRD (experienced by age 10 or 11) predicted jail reported at age 24 or 25, and did so controlling for multiple other factors.

C. SEM: Model Fit

The SEMs included up to four latent constructs: PRD (W1 plus others), each with three randomly-assigned parcels comprising the same 12 items; W3 deviant affiliation (three parcels with four randomly-assigned items); and W4 substance use (also three parcels, with two, one and one items, as described earlier). When multiple waves of PRD were included in a model, the indicators for the parcels were constrained to be equal and the errors of the parcels from different waves were allowed to correlate. Because of their composition, the other constructs were manifest (e.g., arrest and jail were counts, police hassle and illegal behavior were single items). All measures in the measurement models were allowed to correlate. For the primary PRD SEM (i.e., PRD1/PRD2), the model was identified with more observations (405) than free parameters (238) and the model fit was good: χ²: df ratio < 1.63, CFI and TLI ≥ .96, RMSEA < .03, 90% CI = [.02, .03], SRMR < .03. The other models with different waves of PRD were also identified with more observations than free parameters and had good or acceptable fit.

C. SEM: PRD and Police Hassle (see Figure 2)

Figure 2:

SEM: Mediation and Moderation of Childhood and Early Adolescent Discrimination Effects on Arrest and Incarceration

Note.

W1 PRD = Perceived Racial Discrimination, excluding Police Hassle question, at W1

Dev. Aff = Deviant Affiliation, Sub Use = Substance Use, W5/6 Illegal Beh. = Illegal Behavior severity averaged over waves 5–6, Jail = number of times incarcerated For Multigroup analysis, coefficients indicate significant moderation: Racial Pride (RP): Above the line =low RP, Below the line =high RP Controls are correlated with each other and exogenous variables in the model. Controls are predictors of all endogenous model variables.

See text for details on the indicators that make up all latent variables. All coefficients are standardized with z values in the parentheses. * p< .05; ** p < .01; *** p < .001.

Early PRD.

PRD1 and PRD2 were correlated: r [698] = .47 (p < .001). However, the SEM indicated that the predictive power for childhood PRD (PRD1) exceeded that of adolescent PRD. The total effect, direct and indirect (mediated) of childhood PRD on illegal behavior was significant: β = .19, 95% CI = [.10, .28], z = 4.15 (p < .001), as was the indirect effect through deviant affiliation: β = .03, 95% CI = [.004, .048], z = 2.49 (p = .01). The indirect effect of PRD1 on arrest was also significant: β = .06, 95% CI = [.03, .09], z = 3.78 (p < .001). The PRD1 social path to arrest (i.e., through W3 deviant affiliation and then illegal behavior) was significant: β = .01, 95% CI = [>.000, .009], z = 2.14 (p = .03). The same path for PRD2 to deviant affiliation was not significant (p > .27). In addition, there were two direct PRD1 effects. First, the PRD1 to W5/6 illegal behavior direct path was significant: β = .16, 95% CI = [.07, .25], z = 3.57 (p < .001). Because self-reported illegal behavior at W3 (which correlated with W5/6 illegal behavior; r [698] = .20, p < .001) was controlled, this effect shows that childhood PRD predicted change in illegal behavior from age 15 up to age 24 or 25.

The direct path from PRD1 to jail was also significant: β = .05, 95% CI = [.002, .092], z = 2.12 (p = .03). Thus, controlling for several covariates, including prior self-reports of deviant values and delinquency (at W3), arrest (W6), and illegal behavior (at W5 and W6), childhood PRD predicted jail directly. It also directly predicted severity, as well as yes/no for (any) arrest (rs > .12, all ps ≤ .002). Finally, an additional model with just PRD1 was also run. This model looked almost identical to that with PRD1 and PRD2. In sum, the effect of early PRD on arrest / jail followed three paths: a) an indirect (“social”) path to arrest, mediated by affiliation with risky peers, and then illegal behavior; another indirect path mediated just by illegal behavior; and b) a direct path to jail (controlling for arrest).

Sensitivity analyses (focusing on relevant subsamples) provided more evidence of the predictive effect of early PRD. The direct PRD1 to jail path was still significant (β = .05, z = 2.09, p < .04) when the sample was limited to those who had not experienced any police hassle by W3 (age 15/16), in other words, those whose initial unpleasant encounters with the police came at least five years after their early experiences with discrimination. Among those who had been arrested and sent to jail, PRD1 also predicted severity of the crime for which they were sent to jail (r [196] = .14, p = .05).

Early vs. later waves of PRD.

To gauge the impact of early PRD (PRD1) relative to later PRD (after W2), three more models were run that included PRD1 along with: PRD3, PRD4, and cumulative PRD234 (models with PRD4 used W4 deviant affiliation rather than W3). In each case, the direct paths from PRD1 to illegal behavior (all βs ≥ .14, zs > 3.07, ps ≤ .002) and to jail (all βs = .05, zs > 2.36, ps < .02) remained significant. In no case was the direct path from the later PRD (including cumulative PRD) to jail or to illegal behavior significant (all βs < .05, zs < 1.62, ps > .10).⁷ Finally, a follow-up regression was conducted in which jail was regressed on all four PRD reports (PRD1 through PRD4) along with arrest. Consistent with the SEMs, PRD1 was the only significant PRD predictor of jail before and after arrest was entered (βs = .26 and .11, both ps < .03; all other absolute values of βs < .06, ps > .31). The same was true for a regression predicting severity of the offense for which participants were sent to jail (β = .25 and .15, both ps < .04; all other l βs l < .07, ps > .44).

Police hassle.

Controlling for the strong relation between illegal behavior and arrest, W3 police hassle directly predicted arrest in all of the PRD models (all βs ≥ .14, all zs > 3.08, all ps ≤ .002). However, because an arrest reported at W6 and a hassle reported at W3 or W4 could have involved the same incident (meaning the hassle did not predict later arrest), the models were redone, first restricting the sample to those reporting no arrests by W3 and no hassle until W3 (for the W3 hassle model), and then only those with no arrest by W4 (for the W4 hassle model). Both direct paths remained significant (βs = .11, zs > 2.24, ps ≤ .02). In short, self-reports at ages 15/16 and 18/19 of unpleasant experiences with police--that participants attributed to their race-- predicted arrests reported at age 24 or 25, controlling for the 11 covariates, plus W4 substance use and W5/6 illegal behavior, both of which predicted arrest in all analyses (all βs > .11, zs > 2.43, ps ≤ .02).⁸ Importantly, however, some of the relations with police hassle and PRD were moderated by RP.

D. Moderation: Racial Pride

Risk.

Overall, RP was associated with less illegal behavior: r [683] = −.08 (p = .03); however, this relation varied considerably as a function of police hassle experience. Among those reporting no police hassle, RP was correlated more strongly (negatively) with illegal behavior: r [475] = −.14 (p = .002). For those reporting some police hassle, however, the correlation was marginally positive: r [161] = .13 (p = .09), suggesting, as expected (see H₃), the W3 police hassle → illegal behavior path was moderated by RP. To assess this moderation, a multigroup model was run in which the sample was divided at the RP median.⁹ The hassle → illegal behavior paths were first constrained to be equal for the two (high/low) RP groups, and then allowed to vary by group, and then the change in χ² from the constrained to the free model was assessed. An improvement in fit (Δχ²[1] = 8.33, p = .004) indicated the paths differed significantly: the path was significant for high, but not for low RP persons (β = .40, z = 3.08, p = .002 vs. β = −.12, z = −.91, p = .36). Similarly, the indirect effect of police hassle on arrest through illegal behavior was significant for high RP persons (β = .07, 95% CI = [.02, .12], z = 2.36, p < .02), but not low (β = −.02, 95% CI = [−.07, .02], z = −.90, p = .37).

Sensitivity analyses.

Narrowing the sample further, the hassle → illegal behavior path was examined separately for the subgroup previously shown to be most responsive to PRD (Lee & Ahn, 2013): high RP males vs. the rest of the sample. As expected (H₃), the path was significantly stronger for the high RP males (Δχ²[1] = 3.96, p < .05). To further illustrate this latter gender-based effect, a follow-up 2 × 2 × 2 ANOVA on illegal behavior was conducted with independent variables of W3 police hassle (yes/no), gender, and RP (high/low). The 3-way interaction was not significant, but planned comparisons of the high RP / hassled males with the other seven groups indicated the males’ mean (2.04) was higher than each of the other groups (all other Ms < 1.44; all comparison ps < .01), as well as the overall mean of the other seven combined (M = .90; t = 4.07, p < .001).

Protection.

The RP measure was not introduced until W3 (after early PRD was assessed).¹⁰ However, consistent with H₃, RP moderation of reactions to PRD1 was examined. Evidence of a buffering effect emerged: Δχ²[1] = 4.68, p = .03. The relation between PRD1 and illegal behavior was positive and strong for those low in RP (β = 1.84, z = 3.21, p = .001), but it was nonsignificant for those high in RP (β = .40, z = 1.07, p > .28). Thus, RP promoted the hassle → illegal behavior relation, but it buffered against the PRD → illegal behavior relation.

Discussion

Rates of both arrests and incarceration are high among African Americans (Carson, 2016), and that was true for the current sample—fully 46% reported being arrested by age 24 or 25. The results of the current study provide evidence of one reason why that is the case. Reports of early experiences with discrimination indirectly predicted arrest, and directly predicted both illegal behavior and jail. This direct relation of PRD with jail was not stronger than the same relation with SES, but it was stronger than the relations with three other stressors previously shown to be predictive of involvement with the criminal justice system: low parental monitoring, and neighborhood risk, which included both victimization and poverty. Moreover, the early PRD relation with jail maintained controlling for a number of variables previously shown to be related to both arrest and jail, including self-reports of “deviant” values and illegal behavior, the latter at three different time periods. Importantly, these relations were mediated and moderated by several factors, some with implications for intervention, prevention, and/or policy.

Mediation

Based on research linking PRD to risky behaviors (Burt et al., 2012; Gibbons & Stock, 2018), pathways from PRD through various mediators to arrest and jail were specified in these analyses. One was the social path. Early discriminatory experiences were predictive of subsequent affiliation with peers engaging in risky and/or illegal behaviors. Previous studies have identified this stress / affiliation relation, suggesting it often reflects selection more than socialization (Dodge et al., 2006); i.e., adolescents experiencing stress are more likely to seek the company of peers engaging in risky behaviors than to be influenced by current friends who have already taken up these behaviors. That includes affiliation after PRD stress-- perhaps with others who have also experienced discrimination (Gibbons et al., 2007; Whitbeck et al., 2001). This affiliation is then associated with an increase in substance use and other illegal behaviors (Gibbons et al., 2012a), which are all strong predictors of arrest.

Early Discrimination

The relations between PRD and arrest / jail were anticipated. However, given the time lags, we did not expect that PRD reported prior to age 12--before any run-ins with the law for > 90% of the participants--would predict arrest and jail as strongly as it did. In fact, some of those arrests did not occur until 14 years after PRD1 was assessed. These types of early experiences appear to be especially aversive for Black children (Benner et al., 2018), and they can have a long-lasting and potentially life-changing impact on them. Previous studies have shown this impact includes effects on mental health (Schmitt et al., 2014), as well as physical health (Pascoe & Richman, 2009), the latter partly due to PRD’s relation with unhealthy behavior (Gibbons et al., 2010). In general, stress experienced early in life has been shown to increase the likelihood of later emotional problems (Pechtel & Pizzagalli, 2010) and trouble with the law (Reavis et al., 2013). The current data add to this literature by showing these effects are also associated with the stress produced by PRD.

Why? Late childhood is a particularly vulnerable developmental period during which self- and group-identity are being formed (Sellers et al., 2006). Slights to either can have long-term consequences. In this sense, we agree with the perspective expressed by Sanders-Phillips et al. (2009): “Younger children may be especially vulnerable to the effects of racial discrimination, because they may not understand the source of harsh or negative behaviors from others or the effects on their parents. Such experiences can reinforce the feelings of injustice, powerlessness, and victimization that lead to violent behaviors in older children.” (p. 181). As a result, some of these Black children may have started down a path of unwanted involvement with the criminal justice system. In other words, early PRD may act in a similar manner to the combination of reactive temperament and harsh environments that triggers antisocial behavior as described in Moffitt’s (2006) “life-course persistent” model. Moreover, early PRD is also likely to lead to early deviant affiliation, and that is a strong predictor of illegal behavior—apparently stronger than similar affiliations that begin later in life (Evans et al., 2016).

Early PRD and jail.

Early PRD predicted jail indirectly, through illegal behavior. That relation is partly due to the fact that those who reported early PRD also reported engaging in (and being convicted of) more serious criminal behavior; so, once arrested, they were more likely to be sentenced. Importantly, there was also a direct relation between childhood PRD and jail, controlling for both illegal behavior and arrest. Perhaps these high-PRD individuals were acting in a manner after arrest that increased their likelihood of being sentenced to jail. Information about what they were doing after being arrested is not available, so those possible actions cannot be determined. However, analyses conducted with multiple possible mediators measured since W1 (see Supplemental Materials) produced no evidence of this kind of mediating behavior or attitude. Alternatively, this relation may reflect the fact that some individuals are simply more likely to be discriminated against in different situations (e.g., police encounters, sentencing)—independent of either their personality or their behavior¹¹ --and this vulnerability increases the likelihood of involvement in the criminal justice system. That also cannot be determined from these data; but, it seems clear that this direct path is worthy of future investigation.

Moderation: Racial Pride

As in previous research, RP had mixed, but very interesting effects. It was positively correlated with academic orientation, and negatively correlated with deviant values and poor self-control, and also with illegal behavior--among those who had not experienced police discrimination. In fact, it acted as a buffer against the tendency for early PRD to promote illegal behavior. Moreover, there was no evidence that high RP persons (including high RP males) perceived more PRD from the police, suggesting at least the possibility that they did not trigger the hassling that they received. On the other hand, as expected, RP was positively correlated with cumulative PRD, suggesting some vigilance (for discrimination). Finally, RP did appear to amplify the iatrogenic effects of early aversive experiences with the police: the combination of high RP and police hassle predicted more illegal behavior over the six- to nine-year period following the aversive police experience(s), and that relation was especially strong for males.

Overall, these results suggest that RP is protective for young Black adolescents, including its buffering effects on PRD. However, if individuals with a strong sense of identity and pride in their racial/ethnic group experience an unpleasant interaction with a police officer--which they believe is at least partly attributable to their race--that interaction appears to cross a line, evoking a negative reaction that increases the likelihood that they will engage in future illegal behaviors. It also increases the likelihood they will have continued involvement with the criminal justice system—involvement that appears to be only partly a reflection of that behavioral reaction. Given the potential translational implications of this construct (see below), further examination of these relations also seems warranted.

Limitations

There are limitations of the study that should be considered. First, wording of some of the items evolved over the course of the data collection. This is not uncommon for longitudinal research, especially studies like this one that included six waves of data collected from ages 10 to 25--a period in which there is significant change in behavior, contexts, reading skills, etc. It does, however, impose some limits on the ability to assess change in the variables. Second, internal reliability of some constructs was low (more so for the adolescents than the parents). Third, also typical of longitudinal research, the time lag between the assessments most likely attenuated some of the relations. The lag between waves (after W2) was at, or close to, three years, which is a long time in adolescence.

A fourth limitation is reliance on self-reports for most of the measures. This is also not uncommon (in studies of illegal behavior), and there is evidence of validity for self-reports of illegal behavior (Roberts & Wells, 2010), including delinquency (Thornberry & Krohn, 2003) and arrest (Morris & Slocum, 2010). More specifically, there are several reasons to believe that responses in the current study were valid and reliable for most respondents: a) A realization by participants that no negative consequences had arisen from acknowledgements, often in multiple waves, of socially undesirable or illegal behaviors (in fact, their self-reports of illegal drug use were consistently above national norms; Gibbons et al., 2012b); b) High internal consistency within waves, and from wave to wave, in most measures; and c) Numerous individual difference measures likely to be related to misreporting (risk-taking, deviant values, low school orientation) were controlled in the analyses. Nonetheless, these behaviors are illegal, and this may have had an impact on some of the respondents’ willingness to acknowledge engaging in them.

A fifth limitation has to do with the sample. MPlus allowed for use of the full data set, even if the participant did not respond at W6, when arrest and jail were assessed. At least 25 participants (out of the 889) were in jail at W6. Given the possibility that they had committed more serious crimes (14 of them were also in jail at W5), the analyses may have underestimated the amount and severity of criminal behavior (although not by much, given the small number of individuals involved). On the other hand, as mentioned earlier, evidence of racial discrimination in “accountability”—the relation between arrests and jail—is higher for arrests due to low-level crimes, which occur more often among Blacks (Beck & Blumstein, 2017). The vast majority of crimes reported in this sample were low-level; perhaps the role of PRD in the overall process would be reduced for more serious crimes. Sixth, the sample comprised only African Americans. Determining whether similar patterns occur for other minority groups (e.g., Latinx) should be a priority for future studies. Similarly, few gender effects were found in these analyses, but such effects are logical targets of future research. Also, although we examined many possible mediators and moderators (see Supplemental Materials), it is certainly possible that there is another unmeasured “third variable” that could be underlying these effects.

Finally, at the start of this study (1997), the issue of police hassling of Black adolescents was not widely discussed outside of Black communities. That has changed dramatically in the last 6 or 7 years. Scenes on the nightly news of police violence toward unarmed Black men or boys can only lead to more vigilance and more anger in Black communities, which suggests that if this study had been conducted starting (with 10-year olds) after the shootings of Trayvon Martin (in 2012) or Michael Brown (2014), the negative impact of both police hassling and discrimination on Black adolescents might be even greater.

Intervention Implications and Other Future Directions

The police.

Ultimately, the source of the problem identified in this study is the racism that permeates American society as a whole (not just police departments). Unfortunately, little progress has been made on this dimension, and there is evidence that it may actually be getting worse--in general (Pew Research Center, 2019) and in the criminal justice system (Balko, 2019). Reducing discrimination at the national level is, to say the least, a daunting task. There are multiple programs intended to educate police about racial discrepancies in law enforcement (LaMotte et al., 2010), but, to date, relatively few have led to evidence-based interventions (Dunham & Petersen, 2017). In fact, interventions within police departments have not proven very effective at reducing discriminatory attitudes (e.g., implicit bias; Schlosser, 2013); but those aimed at changing the way police interact with citizens have shown some promise (e.g., the “Out-of-car Experience;” Johnson, 2017). These behavior-focused efforts should include educating officers about the pronounced negative impact that interactions that could be perceived as (and may well be) discrimination-based can have on Black adolescents (Hall, Hall, & Perry, 2016), and potentially also on police officers. This is especially true when those interactions occur with Black children and adolescents. Such interventions aimed at reducing the volatility and hostility that underlies many police / Black adolescent interactions should be encouraged.

Policing.

One specific policy addressed by the current data has to do with what has been termed “proactive policing,” which includes the targeting of high-risk neighborhoods and high-risk individuals (i.e., those thought to be most likely to commit crimes). Although there is some evidence that this approach can be effective (National Academy of Science, 2017), there is also recent evidence that it can backfire—police stopping Black adolescents can increase the likelihood of their future delinquent behavior (Del Toro et al., 2019). The current results add to this evidence, suggesting that the risk of iatrogenic effects of police encounters with Black adolescents may be too high to justify the policy—unless demonstrably effective intervention / education programs are implemented.

Adolescents.

Given the developmental range of the six waves of data (age 10 to 25), including repeated assessments of PRD and its various consequences, the results of these analyses suggest some targets for intervention with Black adolescents. One focus could be addressing reactions to PRD. It is easy to understand why PRD would anger anyone, especially Black adolescents--who experience more of it than adolescents of other races and ethnicities (Greene, Way, & Pahl, 2006)--and it is also a facile suggestion that other reactions might be more “constructive” (see Culp, 1992; Russell-Brown, 2009). Nonetheless, the fact that high RP persons were less likely to engage in criminal activity if they had not experienced hassling from the police early in their lives illustrates the significant potential for interventions that involve RP. Supporting evidence for this comes from studies that have shown positive effects of the use of racial affirmation to promote healthy behavior (e.g., Stock et al., 2018); or interventions and preventive-interventions that promote RP, such as the Strong African American Families (SAAF) Program, which has been shown to reduce risky behavior among Black adolescents (Gerrard et al., 2006; Murry, Berkel, Brody, Gerrard & Gibbons, 2007). These approaches (see Jones & Neblett, 2016; and Loyd & Williams, 2017, for reviews) are likely to be more effective if they take advantage of the tendency for high RP individuals to focus on the value of achievement—e.g., academic involvement —perhaps as a way to channel PRD-based negative reactions into commitment to alternative paths to success. Importantly, however, the fact that high RP individuals (especially males) responded more negatively to police hassling means RP-based interventions are not likely to work unless accompanied by meaningful changes in police behavior. Finally, the current study provides some evidence that preparing Black adolescents for possible unpleasant encounters with the police (i.e., “the talk”) might alter the eventual outcome. Again, that will be easier said than done, but the data suggest it may be worth trying.

Conclusion

Previous studies have shown that PRD is predictive of involvement with the criminal justice system for Black adolescents. The current results suggest this effect is especially pronounced for discriminatory experiences that occur during childhood. This relation is mediated by affiliation choices and delinquent behavior. There is also a direct relation between childhood PRD and incarceration that was not mediated by these factors--or any of the multiple other factors assessed in this study (see Supplemental Materials and Footnote 3)-- suggesting that it may not be attributable to anything that these adolescents have done, thought or felt. Hassling by the police that is perceived as being racist also appears to promote delinquent / illegal behavior for some Black adolescents, especially those who are high in RP. Importantly, however, in the absence of police discrimination, RP is associated with less illegal behavior. These observed relations can inform the development of more effective interventions targeting both Black adolescents and the police that interact with them. The need for such interventions seems obvious. The stakes are high.

Public Significance Statement:

Results of this study suggest that early perceived racial discrimination (by age 10 or 11), especially when it involves the police (i.e., “hassling”), increases the likelihood of African American adolescents engaging in illegal behavior and being arrested and incarcerated. Racial pride moderates these effects: it reduces the likelihood of illegal behavior in the absence of police “hassling,” but increases illegal behavior and arrest when the discrimination comes from the police.

Biography