Abstract

The extent to which disproportionate minority contact (DMC) in the juvenile justice system varies across states remains largely unknown. Using a multijurisdictional sample of 146 counties across four states, the present study utilizes multilevel modeling with cross-level interactions to explore whether there is variation in the influence of race and ethnicity among states across four major juvenile justice processing decisions—preadjudication detention, petition of delinquency, adjudication of delinquency, and judicial disposition. The results highlight the existence of some variation in DMC across the four states, with the variation most pronounced at detention and least pronounced at disposition. The possibility of state variation in DMC underscores the need for state-specific analysis of DMC, what contributes to it, and what can be done to reduce it.

Keywords:  disproportionate minority contact; racial and ethnic disparities; juvenile justice; state differences; multiple stages of processing

Introduction

In the juvenile justice system, racial and ethnic disparities at different stages of processing are referred to as disproportionate minority contact (DMC). The most recent national data (2017) indicates that minority youth are overrepresented compared to White youth at every major point of the juvenile justice system except adjudication (Hockenberry & Puzzanchera, 2019). As a result, persistent DMC has become a “national policy issue” in the 21st century (Kempf-Leonard, 2007, p. 72). Although this nation-wide focus helps draw attention to the scope of the problem, it can obscure jurisdictional heterogeneity in racial and ethnic disparities. As Kempf-Leonard (2007, p. 73) emphasizes, national DMC information “does not identify whether DMC is widespread in the United States or the result of acute difficulties in specific states.” There is also the inverse of this problem: A focus on single states limits the generalizability of the findings because of potential idiosyncrasies unique to particular jurisdictions (Britt, 2000).

Given fifty-one independent juvenile justice systems in the United States, an important question is whether there are consistent patterns in racial and ethnic disparities across systems. On the one hand, it is possible that disparities are relatively uniform across states. This often appears to be the presumption in discussions of DMC and how to reduce it (see, e.g., Cabaniss, Frabutt, Kendrick, & Arbuckle, 2007). On the other hand, perhaps DMC is not a problem in all states. Although there have been hundreds of studies of racial and ethnic disparities in the juvenile justice system (see Spinney et al., 2018), studies that focus on multiple states are not as common. In addition, despite a recent increase in studies that examine multiple jurisdictions (see, e.g., Gann, 2019), no studies to date have examined differences in DMC across them. This gap is notable given the mixed evidence of racial and ethnic disparities across multiple stages of juvenile justice processing (see Bishop & Leiber, 2012; Bishop, Leiber, & Johnson, 2010). It is notable, too, because efforts to generate appreciable reductions in DMC nationally require knowledge of both where and at what stages of processing disparities exist (Peck, 2018).

To contribute to efforts to address this research gap, the present study examines variation in racial and ethnic disparities in juvenile justice processing across four states for which complete case processing information was available and that represent different regions of the country. We then discuss implications of the findings and analytic approaches for research, as well as for efforts to identify when and where DMC occurs, what causes it, and what might be done to reduce it.

Background

There is strong prima facie evidence that DMC varies across different juvenile justice systems. For example, in a survey of all 50 states in 1979, Krisberg and colleagues (1987) found that some had large differences in rates of confinement for White versus non-White youth, while others reported similar rates of confinement among the two groups. More recently, descriptive analysis of data from the Office of Juvenile Justice and Delinquency Prevention suggests that states differ in DMC at detention and commitment (Burns Institute, 2019). In the most recent year available (2015), Black-White disproportionalities at preadjudication detention ranged from approximately 70:1 in Wyoming to 2:1 in West Virginia. Similarly, Hispanic-White disproportionalities at detention ranged from 8:1 in Massachusetts to relative rates less than one (i.e., underrepresentation of Hispanic youth) in several states. For judicial disposition, Black-White disproportionalities in commitment ranged from 24:1 in New Jersey to .8:1 in Idaho, while Hispanic-White disproportionalities ranged from 8:1 in Massachusetts to .5:1 in Arkansas.

This state-level variation in DMC is not unique to juvenile justice. Several studies have examined regional and state variation in racial differences in criminal justice processing. Using 1982 correctional data, Crutchfield and colleagues (1994) found that Black-White imprisonment ratios ranged from approximately 3:1 in South Carolina to 21:1 in Minnesota. Regionally, the authors found that disproportionalities were highest in the North Central region and the lowest in the Southern region (see also Bridges & Crutchfield, 1988; Hawkins & Hardy, 1989). Using 1990 data, Blumstein (1993) reported that the national Black-White prison ratio was approximately 7:1, but this ranged from approximately 2:1 in Hawaii to 20:1 in Minnesota. Using 1997 data, Sorensen and colleagues (2003) similarly found that Black-White disproportionality was highest in Minnesota (a ratio of 24:1) and lowest in Arkansas (a ratio of 4:1). Even after accounting for arrest differentials, the lowest disproportionalities were in the South and the highest were in the Midwest. More recently, Beck and Blumstein (2018) reported that while disproportionalities have decreased over time, the highest Black-White imprisonment rate ratio (in 2011) remained in Minnesota (12:1) while the lowest was in Mississippi (3:1). Enders and colleagues (2019) also reported substantial variation in imprisonment disproportionalities, with the Black-to-White incarceration ratio (in 2014) ranging from 2:1 in Hawaii to 14:1 in Minnesota. Consistently across studies, DMC was highest in the Midwest and lowest in the South.[1]

There are, however, limitations in the identification of aggregate state-level variation in DMC. Most state comparisons are largely descriptive and do not address potential confounding that may arise from case- and contextual-level differences across states, such as type of offense, age, community crime, and socioeconomic factors that may be related to juvenile justice processing (Krisberg et al., 1987).[2] This situation is striking given that, as part of the DMC mandate, states must identify DMC, assess causes of it, intervene and evaluate interventions, and monitor the ongoing status of DMC (Leiber & Peck, 2013). Most states have not moved past the first two stages, identification and assessment (Peck, 2018). At the identification stage, states report relative rate indexes (RRIs); these make it possible to track DMC across states and over time. However, they do not control for other factors such as case characteristics and community context (Leiber & Rodriguez, 2011). At the assessment stage, several reviews have found that state-level assessments are of limited validity because they typically do not include key control variables or use robust statistical methods (e.g., Leiber, 2002; Leiber & Rodriguez, 2011; Peck, 2018). There is thus a need for multivariate analyses that identify state-level variation in racial and ethnic disparities in the processing of youth while controlling for differences among states in the youth processed in juvenile court and in the communities in which those courts reside.

While prior research has not assessed variation in DMC across states, a large body of literature has identified variation in racial and ethnic disparities across multiple stages of juvenile justice processing. The most consistent finding involves detention, where research typically finds that minority youth are more likely to be detained than White youth (Guevara, Boyd, Taylor, & Brown, 2011; Leiber, 2013; Leiber, Brubaker, & Fox, 2009; Leiber & Fox, 2005; McCoy, Walker, & Rodney, 2012; Thomas, Moak, & Walker, 2013; but see Rodriguez, 2007; Maggard, Higgins, & Chappell, 2013). Research on intake decisions also has generally reported racial disparities (Leiber & Fox, 2005; Leiber & Peck, 2015; Leiber et al., 2009; Peck & Jennings, 2016). Findings for other stages of processing have been less consistent, however. Although some research has found that non-White youth are significantly more likely to be petitioned than White youth (see, e.g., Ericson & Eckberg, 2016; MacDonald & Chesney-Lind, 2001), other research has found no significant relationship (Freiburger & Jordan, 2011; Leiber et al., 2009). Studies have also consistently found that adjudication of delinquency—the juvenile court equivalent of criminal conviction—occurs less for minority youth than White youth (Leiber & Peck, 2015; Peck & Jennings, 2016; Secret & Johnson, 1997; Thomas et al., 2013). Finally, although some research indicates that minority youth are more likely to receive out-of-home placement at judicial disposition (Leiber et al., 2015, Rodriguez, 2010, 2013, Rodriguez, Smith, & Zatz, 2009; Peck & Jennings, 2016), other research has failed to find a relationship between race and placement (Cauffman et al., 2007, Guevara et al., 2011; Mears et al., 2014), and some studies have found that Black youth receive more lenient dispositions (Bishop et al., 2010; Leiber, 2013; Leiber & Fox, 2005). A central conclusion that flows from prior research, then, is that disparities may arise across one or more stages of juvenile court processing—but not consistently across all stages.[3]

Another conclusion that flows from prior research is that disparities may vary across different juvenile courts as well as stages of processing. As Gann (2019, p. 270) recently noted, “the most promising avenue to gaining a complete understanding of the racial influences on juvenile court decision-making is research that examines multiple decision points in the court process across multiple courts” (see also Crutchfield, Fernandes, & Martinez, 2010). One prominent line of such research suggests that court processing disparities may vary according to whether the court is urban or rural (Feld, 1991). The narrowly construed view of “justice by geography” argues that urban courts tend to be more formalistic and legalistic, which may translate into lower racial and ethnic disparities.[4] To date, empirical support for the urban/rural difference in disparities is limited (see DeJong & Jackson, 1998; but see Bray et al., 2005; Guevara et al., 2011). The more broadly construed view of “justice by geography,” however, argues that disparities may vary across jurisdictions according to various court-level characteristics. An extension of this logic is that differences in racial and ethnic disparities may arise across different states.

Of the many studies in the scholarly literature that examine disparities in juvenile justice across major stages of processing, however, only a handful examine multiple states in the same study, usually either two (Cauffman et al., 2007; Peck & Beaudry-Cyr, 2016; Peck, Leiber, Beaudry-Cyr, & Toman, 2016; Peck, Leiber, & Brubaker, 2014) or three (Leiber & Peck, 2015; Leiber, Peck, & Rodriguez, 2016; but see Davis & Sorenson, 2012; Sampson & Laub, 1993). Much of this research has begun the important project of seeking more generalizable conclusions about racial and ethnic disparities in juvenile justice across multiple systems. It has not, however, included systematic assessment of differences in juvenile justice disparities across states.[5] Rather, they simply have included multiple states but not examined variation in disparities at the state level or whether disparities in some states surface at different stages. There is, then, a need for studies that examine DMC variation, including disparities across decision-making stages, across states.[6]

Current Study

The current study seeks to contribute to the literature on racial and ethnic disparities in juvenile justice by examining variation across states, and stages of processing within states. We argue that a neglected dimension of thinking about DMC centers on state variation. States can and do differ in their laws and legal cultures, and mixed findings from prior jurisdiction-specific studies suggest the possibility that there may be potential state-level differences in the influence of race and ethnicity on juvenile justice outcomes. A focus on state variation is important for understanding DMC and, from a policy perspective, monitoring and assessing efforts to reduce it. If disparities are relatively uniform across states, then DMC can be properly conceptualized as a truly national problem. If disparities differ dramatically across states, however, then approaching DMC as a national problem may be misguided, and a localized approach may be more appropriate. Moreover, the precise focus might depend on the stage of processing in question. The present study should be viewed as a first step in assessing whether variation in disparities across major stages of processing exists across states. Given prior research and the exploratory nature of this study, then, we anticipate the following:

H₁: For each juvenile court processing outcome, there will be significant variation in racial and ethnic disparities across different state juvenile justice systems (i.e., state context will moderate the influence of race and ethnicity on juvenile justice outcomes).

A corollary to this hypothesis derives from prior work that identifies the potential for disparities to vary across different stages of processing. Such work suggests that a focus on multiple decision-making points is indicated for two reasons. First, examination of multiple processing stage outcomes may provide a more robust test of the hypothesis that DMC may vary across states. Second, disparities may not be uniform across stages of processing (see, e.g., Bishop et al., 2010; Engen et al., 2002). Building on these observations, we therefore anticipate a second hypothesis:

H₂: State-level variation in disparities may not be consistent—for example, although some states may have greater disparities at one stage of processing, they may not have more disparate treatment of minorities at other stages of processing.

Methods

The present study used a combination of juvenile court administrative data and county-level demographic data, where cases are nested within counties in a hierarchical structure. The case-level dataset consists of all juvenile case referrals across four states—Alabama, Connecticut, South Carolina, and Utah—for which final disposition was entered in 2010 (i.e., cases pending removed). The restricted data were provided by the National Juvenile Court Data Archive (NJCDA), which is maintained by the National Center for Juvenile Justice and supported by a grant from the Office of Juvenile Justice and Delinquency Prevention (U.S. Department of Justice).[7] States were selected if they collected data (in 2010) for each of the four outcomes to serve as dependent variables: preadjudication detention, petition of delinquency, adjudication of delinquency, and judicial disposition. According to the NJCDA, only four states provided data for each of these outcomes, along with case-level information for race/ethnicity, age, sex, offense type, and prior record. In short, these four states were the only ones that provided complete case processing data across major decision-making points. An added advantage is that they can be viewed as representing different regions of the country. As we discuss below, a clear limitation is that they may not represent what occurs in other states.

The original data consisted of 106,582 referred delinquency or status offense cases (i.e., dependency cases removed) in 149 counties.[8] This dataset was merged with a level-2 dataset consisting of county-level demographic factors created using decennial Census 2010, the American Community Survey (ACS) five-year estimates for 2010, aggregated official crime statistics from the Uniform Crime Reports (UCR), and county voting data from Dave Leip’s Atlas of U.S. Presidential Elections. Cases were dropped from the final sample based on missing data for included case-level variables (n = 6,224; 5.8 percent) or missing data for included county-level variables (i.e., counties with low quality estimates; n = 53). All counties had a minimum of 10 observations. The resulting two-level dataset consists of 100,358 juvenile court referrals across 146 counties.

Measures

Dependent Variables

The present study examines four main phases of processing in juvenile court: detention at intake, petition, adjudication, and judicial disposition. All outcomes were coded as binary for ease of interpretation in computing marginal effects and second differences for assessing interaction effects (see below). The first outcome, preadjudication detention, was coded as “1” for detention at intake and “0” for release pending adjudication of delinquency. The full sample was available for the detention analyses (N₁ = 100,358; N₂ = 146). Overall, approximately 16 percent of cases were detained prior to petition (n = 15,912). States varied somewhat in the use of detention, with 12 percent of referrals detained in Alabama, 13 percent in Connecticut, 18 percent in South Carolina, and 20 percent in Utah.

The second outcome, petition, was coded as “1” for a formal petition of delinquency and “0” for diversion or release. The sample for these analyses consisted of 99,344 observations across 146 counties. Approximately 59 percent of cases received petition (n = 58,328). States varied in the percent of cases that were formally processed, with 61 percent in Alabama, 55 percent in Connecticut, 41 percent in South Carolina, and 69 percent in Utah.

Adjudication was coded as “1” for an adjudication of delinquency and “0” for dismissal or diversion by the juvenile court judge.[9] This sample consisted of 57,529 cases across the 146 counties. Almost two-thirds of petitioned cases were adjudicated delinquent (n = 36,348). States varied considerably in the percent of petitioned cases that were adjudicated delinquent, however, with 41 percent in Alabama, 46 percent in Connecticut, 94 percent in South Carolina, and 76 percent in Utah.

Finally, judicial disposition was coded as “1” for out-of-home placement (i.e., commitment) and “0” for community supervision (i.e., probation). The sample available for disposition analyses consisted of 33,799 cases across 145 counties, with approximately 15 percent of cases receiving secure placement (n = 5,074). States varied in the percent of delinquent cases that received placement, with 8 percent in Alabama, 12 percent in Connecticut, 23 percent in South Carolina, and 15 percent in Utah.[10] Descriptive statistics for all variables are included in Table 1.

[Table 1 about here]

Case-Level Variables

The main independent variable of interest was the race and ethnicity of the defendant, coded as a factor variable with four categories: Black, Hispanic (non-White), Other (Asian, American Indian, or Other), and White (reference group). (For short-hand, this variable is referred to as defendant race.) The full sample of juvenile referrals was approximately 50 percent White (n = 50,479), 33 percent Black (n = 32,768), 13 percent Hispanic (n = 13,279), and 4 percent Other (n = 3,832). States varied in the racial makeup of their juvenile referrals. Black youth made up approximately 45 percent of referrals in Alabama, 23 percent in Connecticut, 57 percent in South Carolina, and 3 percent in Utah. Hispanic youth, on the other hand, made up approximately 3 percent of referrals in Alabama, 23 percent in Connecticut, 3 percent in South Carolina, and 28 percent in Utah. (“Other race” youth were largely from Utah, where they made up 9 percent of referrals due to the American Indian population in that state.)[11]

Defendant age, sex, offense type, and prior record were included as controls. Age at referral was measured as a continuous variable ranging from 7 to 20, with a mean age (SD) of approximately 15 (1.9). Sex was coded as “1” for male (67.9 percent; n = 68,176) and “0” for female. Offense type was measured as a series of six dummy variables for each type of offense, based on initial charges: violent (13.1 percent; n = 13,176), property (25.5 percent; n = 25,550), drug/alcohol crimes (10.4 percent; n = 10,473), non-violent weapon offense (1.4 percent; n = 1,385), probation violation (7.5 percent; n =  7,498), status offense (16.7 percent; n = 16,753), and other offenses, such as public order and traffic crimes (23.2 percent; n = 23,271).[12] Property offense was the reference category in all analyses since it is the most common type of delinquency. Prior record was measured as “1” for any prior referrals and “0” for none; the majority of cases involved at least one prior referral (60.8 percent; n = 60,022).[13]

Contextual Variables

Due to the hierarchical structure of the data and the multijurisdictional nature of the analyses, several contextual controls were also included. In addition to the relevance of urbanism (as predicted by “justice by geography”), differences in juvenile justice processing could involve variation in racial and ethnic contexts, economic conditions, surrounding crimes rates, case processing pressures, political climate, as well as demographic context (see McGarrell, 1991; Mears, 2006). As such, the present study controls for urbanism, racial composition, economic inequality, crime rates, juvenile arrest rates, and political conservatism.

First, population density based on Census 2010 (recoded as 100 persons per square mile) was used to measure urbanism. This ranged from .01 (i.e., 1 person per square mile) to 14.67 (i.e., 1,467 persons per square mile), with a mean (SD) value of 4.89 (5.05). Second, a measure of percent non-White was constructed from Census 2010 data. Counties ranged from 2.8 to 84.3 percent non-White, with a mean average (SD) of 24.6 (13.6). Third, to capture socio-economic context, the traditional measure of income inequality, the Gini coefficient, was also included. The Gini coefficient measures relative incomes within a group (such as a county), where “0” represents perfect equality and “1” represents perfect inequality. Counties ranged from Gini scores of .34 to .58, with a mean average (SD) of .44 (.04). To control for the surrounding crime context, crime rates were also included based on UCR reports, measured as an average for total index (i.e., felony) crime rates for the years 2009–2011 (per 1,000 residents). Crime rates ranged from 5.7 to 78.6 crimes per 1,000 residents, with a mean average (SD) of 37.5 (13.6). Additionally, we controlled for juvenile arrests (per 1,000 residents) to capture juvenile court case processing pressures at the county level. Juvenile arrest rates ranged from .26 to 14.81, with a mean average (SD) of 5.35 (3.26). Last, we included a measure of political conservatism at the county level, based on the difference in percentage voting Republican versus Democrat in the 2010 Gubernatorial election (Dave Leip’s, 2018). Political conservatism ranged from -75.6 (indicating a 76 percent Democratic majority) to 68.8 (indicating a 69 percent Republican majority), with a mean average (SD) of 18.1 (24.8).

Finally, since the hypotheses involve cross-level interactions between race and state, dummy variables for states were included. Alabama made up 33.5 percent of the cases (n = 33,568), Connecticut 12.2 percent (n = 12,265), South Carolina 21.2 percent (n = 21,314), and Utah 33.1 percent (n = 33,211). Alabama served as the reference category.

Analytic Strategy

As Frase (2013, p. 266) observes, race effects can be confused with jurisdictional effects due to differences in racial composition across different locations. As an example, he points out that if the majority of Black residents within a state live in urban areas where sentences are typically shorter than rural areas, it may conceal racial disparities that would be detected if we controlled for geographic differences in punitiveness. It is possible that some or all of the variation in state-level disparities arises from such compositional differences. A modeling strategy is needed that controls for such contextual factors. Accordingly, a multilevel approach is used to estimate the effects of race and ethnicity on detention, petition, adjudication, and placement outcomes in juvenile court across 146 counties in four states, with cases nested within counties. Although it has become commonplace to use the Heckman two-step procedure to account for potential selection effects when looking at sequential outcomes in criminal justice research, doing so is unwarranted without including exclusion restrictions (Bushway, Johnson, & Slocum, 2007). Here, no exclusion restrictions could be identified. (In addition, there is no straightforward application of the Heckman selection model in a multilevel context.) All outcomes are binary for ease of interpretation of cross-level interactions (see below). Continuous control variables (i.e., age, population density, percent non-White, crime rate, juvenile arrest rate, conservatism) were grand mean centered. Gini scores were standardized for ease of interpretation.[14]

For all four outcomes, a parallel analytical strategy was adopted. Unconditional models were first estimated, followed by random intercept models, random coefficient models, and random coefficient models with interaction terms between race and states (see Johnson, 2010). In the random coefficient models, race (treated as a factor variable) was designated as a random effect, and cross-level interactions between race (whose slope was allowed to vary across counties) and state were estimated. In the context of a nonlinear dependent variable, however, interaction effects become more complicated and the product term in regression output does not represent a sufficient test of the interaction (see Mustillo, Lizardo, & McVeigh, 2018). Therefore, following Mize (2019), we calculated predicated probabilities in post-estimation for each of the random coefficient models (using the margins command in Stata 15). Marginal effects were estimated to assess the differences in probability of each outcome by race and ethnicity (for each state). To test interactions between race and state, we then estimated second differences in the marginal effects of race across states (i.e., differences-in-differences). All analyses used 2-sided hypothesis testing with an a of .05 and were performed using Stata 15.

Results

Detention

To assess whether detention outcomes varied across counties, an unconditional multilevel logistic regression model was estimated. The analysis (not shown) indicated significant variation in odds of detention across counties, with a level-2 variance component (ψ) of .75 and intraclass correlation (ρ) of .19. (Since multilevel logit models do not include a level-1 variance term, intraclass correlation (ρ) can be estimated using p²/3 as the level-1 variance, assuming that the level-1 variance follows the logistic distribution; see Johnson, 2010). This indicates that before any predictors are added to the model, 19 percent of the variation in detention outcomes is attributable to between-county differences. A multilevel modeling strategy is thus warranted.[15]

Before adding cross-level interactions (results not shown), Black defendants were 21 percent more likely to be detained than White defendants (OR = 1.21, p < .001), while Hispanic defendants were 23 percent more likely to be detained (OR = 1.23, p < .001). In the full model with interactions (see Table 2), Black defendants remained more likely to be detained (OR = 1.15), while the Hispanic-White difference in odds of detention was no longer significant. Among states, odds of detention were highest in South Carolina (OR = 1.84).

[Table 2 about here]

To investigate cross-level interactions, we estimated predicted probabilities of detention by race and state (Tables 3a, 3b), followed by marginal effects of race for each state and for comparisons across states (i.e., tests of second differences). For the Black-White comparison, marginal effects were statistically significant in Alabama (1.0 percent difference), Connecticut (4.1 percent difference), and Utah (6.6 percent difference), indicating that Black defendants had a higher probability of detention (than White defendants) in those states. For the Hispanic-White comparison, marginal effects were significant for Connecticut (3.6 percent difference), South Carolina (6.6 percent difference), and Utah (2.8 percent difference), indicating that Hispanic defendants had a higher probability of detention (than White defendants) in those states.

Tests of second differences revealed that Black-White differences were higher in Connecticut and Utah than in Alabama and South Carolina, with no significant differences emerging between Connecticut and Utah or Alabama and South Carolina (see Table 3a, contrast column). Tests of second differences also revealed that the marginal effect of Hispanic defendants was higher in South Carolina than Alabama or Utah, with no differences between South Carolina and Connecticut (see Table 3b, contrast column).

[Tables 3a and 3b about here]

Petition

To assess whether petition outcomes varied across counties, an unconditional model was again estimated. The analysis (not shown) indicates significant variation in petition odds across counties, with a level-2 variance component (ψ) of 1.74 and intraclass correlation (ρ) of .35. This indicates that before any predictors are added to the model, approximately 35 percent of the variation in petition outcomes is attributable to between-county differences.[16]

Before adding cross-level interactions, Black defendants were 20 percent less likely to be petitioned than White defendants (OR = .80, p < .01), while Hispanic defendants were 21 percent more likely to be petitioned than White defendants (OR = 1.21, p < .001). In the full model with interactions (Table 2), Black defendants were 12 percent less likely to be petitioned than White defendants (OR = .88), while the Hispanic-White difference was no longer significant. Odds of petition were lowest in South Carolina (OR = .13) and Connecticut (OR = .16).

To investigate cross-level interactions, we estimated predicted probabilities of petition by race and state (Tables 4a, 4b), followed by marginal effects of race for each state and comparisons across states. For the Black-White comparison, marginal effects were only statistically significant in South Carolina (5.7 percent difference), indicating that Black defendants had a lower predicted probability of petition in that state. Across states, tests of second differences revealed that Black-White difference in petition was significantly lower in South Carolina compared to Connecticut and Utah (see Table 4a, contrast column). For the Hispanic-White comparison, marginal effects were not significant for any states.

[Tables 4a and 4b about here]

Adjudication

To assess whether adjudication outcomes vary significantly across counties, estimating an unconditional model was again the first step. The analysis (not shown) indicates that there is significant variation in odds of adjudication across counties, with a level-2 variance component (ψ) of 3.31 and intraclass correlation (ρ) of .50. This indicates that before any predictors are added to the model, approximately 50 percent of the county-level variation in adjudication outcomes is attributable to between-county differences.[17]

Before adding cross-level interactions, Black defendants were 10 percent more likely to be adjudicated delinquent than White defendants (OR = 1.10, p = .05), while Hispanic defendants were 8 percent more likely to be adjudicated delinquent (OR = 1.08, p = .05). In the full model with interactions (see Table 2), the odds of Black defendants being adjudicated delinquent were 22 percent greater (OR = 1.22); by contrast, there was no significant difference in the odds of adjudication when comparing Hispanic and White defendants. Among states, adjudication was most likely in South Carolina (OR = 42.33) and Utah (OR = 8.38).

To investigate cross-level interactions, we estimated predicted probabilities of adjudication by race and state (Tables 5a, 5b), followed by marginal effects of race for each state and comparison of marginal effects across states. For the Black-White comparison, marginal effects were statistically significant in Alabama (3.7 percent difference)—indicating that Black defendants had a higher probability of adjudication of delinquency than White defendants—and approached significance in South Carolina (1.4 percent difference)—indicating that Black defendants had a lower probability of adjudication of delinquency. For the Hispanic-White comparison, no marginal effects were significant. Across states, tests of second differences revealed that the Black-White difference was higher in Alabama than in South Carolina and Utah (see table 5a, contrast column). For Hispanic defendants, tests of second differences revealed no significant differences across marginal effects of race by state.

[Tables 5a and 5b about here]

Disposition

To assess whether disposition outcomes vary by county, an unconditional model was estimated. The analysis (not shown) indicates that there is significant variation in odds of placement across counties, with a level-2 variance component (ψ) of 1.20 and intraclass correlation (ρ) of .27. This indicates that before predictors are added to the model, 27 percent of the variation in disposition outcomes is attributable to between-county differences.[18]

Before adding cross-level interactions, the odds of Black defendants receiving secure placement were 19 percent greater than that for White defendants (OR = 1.19, p < .05), while there was no significant difference between Hispanic and White defendants. In the full model with interactions (see Table 2), there were no significant differences in odds of placement among Black, Hispanic, and White defendants. Among states, odds of placement were highest in South Carolina (OR = 5.58), and Utah (OR = 4.23).

To investigate cross-level interactions, we estimated predicted probabilities of placement by race and state (Tables 6a, 6b), followed by marginal effects of race for each state and comparisons across states (i.e. tests of second differences). For the Black-White comparison, marginal effects were only significant in Utah (4.6 percent difference), indicating that Black defendants had a higher probability of placement than White defendants. Across states, tests of second differences revealed that Black-White differences were significantly higher in Utah than in Alabama (see table 6a, contrast column). For Hispanic defendants, tests of second differences revealed no significant differences across marginal effects.

[Tables 6a and 6b about here]

Discussion

The present study examined whether DMC varied significantly across state juvenile justice systems. For each stage of processing under examination, our main hypothesis was that racial and ethnic disparities would be significantly greater in some states than others (H₁). We also explored the possibility that this state-level variation itself may vary across stages of processing, given that prior research indicates that disparities are not uniform across stages (H₂).

The main hypothesis (H₁) received general support: there was significant variation in Black-White differences across states for all four stages of processing, and significant variation in Hispanic-White differences across states for detention only. The differences were as follows. First, Black-White disparities in detention were greater in Connecticut and Utah than in Alabama and South Carolina. Hispanic-White disparities were greater in South Carolina than Alabama and Utah. Second, and similar to detention, Black-White disparities at petition were greater in Connecticut and Utah than in South Carolina. Third, Black-White disparities at adjudication were greater in Alabama than in South Carolina or Utah. Fourth, Black-White disparities at placement were greater in Utah than Alabama.

The corollary hypothesis (H₂) was also supported. We expected that state-level variation in disparities would be more pronounced for some stages of processing than others. While we did find some state-level variation in disparities across all four outcomes, some stages exhibited more variation than others. Specifically, the most variation in effects was observed at detention, where racial and ethnic disparities varied significantly across all states. For other stages of processing, we also observed some variation in racial disparities, but did not observe any variation in ethnic disparities. Moreover, the observed variation in racial disparities for these other stages indicated that only one state evidenced significant disparities for each stage of processing: Black-White disparities in petition occurred only in South Carolina; Black-White disparities in adjudication occurred only in Alabama; and Black-White disparities in placement occurred only in Utah.

In short, we found that DMC did vary significantly across states. Moreover, while we found some state-level variation in racial and ethnic disparities across all four stages of processing, this was most pronounced at detention.

Interpretation of Findings

The central finding of the present study is that racial disparities were greater in some states than others. The clearest pattern in the findings is that Connecticut and Utah exhibited greater racial disparities than Alabama and South Carolina. Others have observed similar regional patterns in racial disparities in the criminal justice system. Enders and colleagues (2019) found, for example, that the largest racial disparities in imprisonment were in Northeast and Midwest states, while the smallest disparities were in the South (see also Beck & Blumstein, 2018; Blumstein, 1993; Bridges & Crutchfield, 1988; Crutchfield et al., 1994; Sorensen et al., 2003). As Blumstein (1993, pp. 755–776) earlier observed, this is a “surprising result” that is not “consistent with the commonly-held stereotype that it is southern states that practice racial discrimination.” Instead, this finding suggests that systems with fewer minority referrals overall may have greater disparities. (In the present study, the percentage of Black referrals was 45 percent in Alabama and 57 percent in South Carolina, compared to 23 percent in Connecticut and 3 percent in Utah.) One possible explanation is intergroup contact theory, which posits that more contact among racial groups diminishes stereotypes and biases, leading to more equal treatment (Hughes, Warren, Stewart, Tomaskovic-Devey, & Mears, 2017). According to this theory, we should expect lower disparities in states with larger minority populations. While we could not test for this hypothesis directly given the number of states in our sample, it is consistent with the size of the Black youth population (in 2010) in Alabama (31 percent) and South Carolina (34 percent) compared to Connecticut (12 percent) and Utah (2 percent).           

Another possible explanation for the observed variation in disparities is that states that are more punitive overall may also, paradoxically, have lower racial disparities. Blumstein (1993) considers three possible relationships between overall punitive orientation and racial disparities in punishment. First, differences in punitive orientation may be unrelated to racial and ethnic disparities. Second, and perhaps most intuitively, more punitive states may have greater disparities. Here, we might imagine that the same forces that give rise to greater punishment (such as political conservatism) also give rise to increased racial discrimination. A third possibility is that less punitive states will exhibit greater disparities. As Zimring (2014, p. 173) has stated, more punitive states may “equalize disadvantage” for all defendants. To put the point provocatively, the simplest way to reduce overrepresentation of minorities would be to increase the punitive treatment of White defendants. This would produce a more relatively proportionate—and more absolutely punitive—result (Zimring, 2014). While it is the most “surprising” of the three, Blumstein (1993) pointed out that this third possibility—greater disparities in less punitive states—seemed most consistent with state-by-state incarceration rates.

It bears emphasizing, however, that our findings do not provide clear support for this inverse relationship between punitiveness and disparities. While we did find a regional pattern—Southern states exhibited lower disparities—overall punitiveness did not correspond to region. For example, both South Carolina and Utah appeared to be the most “punitive” states in terms of higher odds of detention, adjudication, and placement for youth (see Table 2). Whether differences in intergroup contact, differences in overall punitiveness, or other state-level factors, such as laws, court structures, or the like are responsible for the variation in disparities we observed remains to be investigated.[19] The present findings should provide impetus for further exploring how and why racial disparities vary across state juvenile justice systems. 

Limitations

Several study limitations warrant discussion. First, our conclusions are limited to the four states we examined and cannot be generalized to all juvenile justice systems. The states were selected based on data availability from the National Juvenile Court Data Archive. Specifically, only four states were able to provide data on all major outcomes of interest, in addition to the case-level variables included in the analysis. These states have the advantage of representing several U.S. regions. Going forward, however, there is a need to establish whether similar patterns hold across all states. Ideally, studies could rely on a larger sample of states or even all 50 states plus the District of Columbia. Doing so would allow for more systematic assessment of state and regional variation. While we did control for several contextual influences at the county-level, reliable hypothesis testing of state-level factors would require a larger sample of states since states would serve as the unit of analysis.[20]

Another limitation is that while we were able to control for prior record, the dichotomous measure did not include information on the nature or number of such referrals. Others have noted that there is a strong positive association between prior record and race, such that failing to control for prior record likely overestimates the true race effect (see, e.g., Franklin & Henry, 2020). Interestingly, some have also suggested that controlling for prior record may underestimate race effects (Baumer, 2013). Namely, if the influence of race on court outcomes is mediated by prior record, as has been suggested elsewhere (see Frase, 2013), then controlling for prior record may in fact control away some of the indirect race effect, what has been called “overcontrol” (Elwert and Winship, 2014, p. 32). It is possible, then, that a different measure of prior record, such as a count measure, might have increased or decreased the race effects in our models. To ensure reliable comparisons in the effects of race and ethnicity across states, we used a simple binary measure of prior record.

A third limitation involves omitted variable bias (Baumer, 2013). There are myriad unmeasured case characteristics, such as developmental maturity, substance abuse, family situation, attitudinal factors (e.g., demeanor), victim characteristics, and strength of evidence, that may be associated with outcomes as well as racial disparities in outcomes (see, e.g., van Wingerden, Wilsem, & Johnson, 2016). For example, the present study found consistent disparities at preadjudication detention. One possibility is that the detention decision is made with little oversight and has high potential for racial bias that may produce cumulative effects on later processing (see Rodriguez, 2010). Another possibility is that unmeasured factors play a larger role at detention than other stages. From this perspective, detention does not produce greater disparities via higher discretion and racial bias, but rather via some unmeasured factors, such as perception of risk based on situational and attitudinal factors not easily captured by officially recoded measures. This challenge of course confronts all research inquiries into racial disparities whose causes involve perceptual processes and decision-making that cannot be easily observed (Lynch, 2019; Ulmer, 2012). As Peck (2018, p. 312) recently argued, “Moving forward, empirical examinations of DMC need to be able to clearly measure the attitudes, beliefs, and perceptions of juvenile court workers instead of implying that racial biases account for the differential treatment of minority youth.”

Implications for Research, Policy, and Practice

Several implications for research, policy, and practice flow from these analyses. First, the fact that in all four states, Black youth were overrepresented at certain stages compared to White youth, provides some support for the notion that DMC is a national rather than only a state- or jurisdiction-specific problem. This was especially true of detention, an early stage of processing where significant racial and ethnic disparities emerged in most states.

Nevertheless, significant variation in disparities across states may indicate that DMC should be reconceived as a state rather than a national problem. This is contrary to the current focus of the DMC mandate, which takes a more aggregate approach to national disparities in juvenile justice processing (Leiber & Rodriguez, 2011). As Sampson and Lauritsen (1997, p. 855) noted over two decades ago, “because there may be considerable variation among states in the processing of white and black offenders, arrestees, defendants, and sentenced inmates, studies that simply aggregate across states and jurisdictions may mask significantly different patterns of treatment by the police, prosecutors, and courts.” One possibility is that levels of racial disparities within a state may reflect organizational features of the court, such as whether the court has a “traditional,” “transitional,” or “due process” orientation (Stapleton, Aday, & Ito, 1982). While organizational differences between courts are often conceived in terms of differences within states, such as the urban/rural distinction emphasized by “justice by geography” (Feld, 1991), another possibility is that state justice systems differ in organizational characteristics, including factors such as court culture (see Eisenstein, Flemming, & Nardulli, 1988; Hester, 2017) or even state political culture (see Barker, 2006; Garland, 2013).

There is, then, a second major implication of the present study: more comparative research is needed on juvenile justice systems and DMC across states. Such research can be accomplished through quantitative studies that test state-level hypotheses about variation in racial and ethnic disparities in juvenile justice processing. As noted above, the present study was limited to four states and could only test whether racial disparities varied across state systems but not why. There is a need for research that identifies and examines the characteristics of juvenile justice systems that might exacerbate or mitigate racial disparities. Some hypotheses, like differences in state law and policy, could be most easily tested. Other differences, such as how decision-making is allocated among intake officer, prosecutor, and judge, might also be studied with more primary data collection (Ulmer, 2012). Still other differences, such as court cultures or general philosophies animating the juvenile court, may prove more elusive. As Garland (2013) observes, quantitative analyses comparing states is limited by the number of macro-level variables that can be examined.

For this reason, there is also a need for in-depth qualitative research comparing different state justice systems. In his presidential address for the American Society of Criminology, Garland (2013, p. 489) urged more “small-n comparative studies that are in-depth, qualitative and quantitative analyses, focused on penal state processes—a form of analysis that is still surprisingly rare.” Many state-level differences will elude aggregate measures, and only in-depth comparative historical research will be able to provide greater understanding of how these differences may contribute to criminal justice policies and outcomes (see, e.g., Barker, 2006). With 51 different criminal justice systems operating like “fifty-one different countries” (Frost, 2009, p. 277), the opportunities for greater understanding through comparative qualitative research are substantial (see also Peck, 2018).

Third, the uncertainties surrounding the true nature and extent of DMC limits our ability to make specific policy recommendations applicable to all juvenile justice systems (see Nellis & Richardson, 2011). If there is state-level variation in such disparities—as suggested by the present findings—this introduces complexity for efforts to identify, understand, and address DMC (see Leiber & Fix, 2019). Krisberg and colleagues (1987, p. 201, emphasis added) observed more than three decades ago that “we know that some jurisdictions have less of a disproportionately incarcerated minority population than others. We need to study these jurisdictions more closely to understand the ameliorative factors that may be operating.” For researchers, this situation affords opportunities to build on sentencing scholarship that points to the salience of local context, courtroom workgroups, and the organizational dynamics specific to different stages of processing (see, e.g., Britt, 2000; Bishop et al., 2010; Hester, 2017; Ulmer, 2019). Advances along these lines may help to identify state-, local-, and stage-specific disparities, causal processes at these levels and stages, and policies that might address them.

Finally, for policymakers and practitioners, the present findings suggest the need to approach DMC in a more systematic manner. As Ulmer (2019, p. 483) has recently argued, “variation between courts in sentencing practices should be understood not as a nuisance in top-down imposition of sentencing policies, but as a valuable but underappreciated source of policy feedback and learning.” In other words, much policy-oriented research takes for granted that “top-down” (e.g., “silver bullet”) solutions will be most efficient once we have identified the source of a problem such as DMC. But perhaps top-down solutions are limited given the built-in variations in outcomes across courts when viewed as complex organizational systems inhabited by different actors (see Mears, 2017). Solutions that focus only at one level of analysis, for example, likely will have little appreciable impact on disparities that arise at other levels of analysis or occur more so at certain stages of processing. One possible policy response at the national level would involve requiring states to implement racial impact statements as part of their compliance with the DMC mandate (Leiber & Rodriguez, 2011). Unlike prior identification requirements (i.e., state-reported RRIs), racial impact statements would have the benefit of localizing the DMC problem and incentivizing state policymakers to consider what factors contribute to disparities within their specific jurisdictions. Along with a robust comparative research agenda, these racial impact statements might enhance our understanding of why DMC varies across juvenile justice systems and how policymakers and practitioners can adopt appropriate efforts to reduce DMC within their jurisdictions.

References

Barker, V. (2006). The politics of punishing: Building a state governance theory of American imprisonment variation. Punishment & Society, 8(1), 5–32.

Beck, A. J., & Blumstein, A. (2018). Racial disproportionality in U.S. state prisons: Accounting for the effects of racial and ethnic differences in criminal involvement, arrests, sentencing, and time served. Journal of Quantitative Criminology, 34, 853–883.

Bishop, D. M., & Leiber, M. J. (2012). Racial and ethnic differences in delinquency and justice system responses. In B. C. Feld & D. M. Bishop (eds.), The Oxford handbook of juvenile crime and juvenile justice (636–663). Oxford, UK: Oxford University Press.

Bishop, D. M., Leiber, M. J., & Johnson, J. (2010). Contexts of decision making in the juvenile justice system: An organizational approach to understanding minority overrepresentation. Youth Violence and Juvenile Justice, 8, 213–233.

Bridges, G. S., & Crutchfield, R. D. (1988). Law, social standing, and racial disparities in imprisonment. Social Forces, 66, 699–724.

Blumstein, A. (1993). Racial disproportionality of U.S. prison populations revisited. University of Colorado Law Review 64:743–1223.

Bray, T. M., Sample, L. L., & Kempf-Leonard, K. (2005). Justice by geography: Racial disparity and juvenile courts. In D. F. Hawkins & K. Kempf-Leonard (eds.), Our children, their children: Confronting racial and ethnic differences in American juvenile justice (270–299). Chicago, IL: University of Chicago Press.

Britt, C. L. (2000). Social context and racial disparities in punishment decisions. Justice Quarterly, 17, 707–732.

Bushway, S. D., Johnson, B. D., & Slocum, L. A. (2007). Is the magic still there? The use of the Heckman two-step correction for selection bias in criminology. Journal of Quantitative Criminology, 23, 151–178.

The W. Haywood Burns Institute (2019). Unbalanced youth justice. Available at: http://data.burnsinstitute.org/#comparison=3&placement=1&races=1,2&offenses=5,2,8,1,9,11,10&year=2010&view=map. Last accessed October 19, 2019.

Cabaniss, E. R., Frabutt, J., Kendrick, M. H., & Arbuckle, M. B. (2007). Reducing disproportionate minority contact in the juvenile justice system: Promising practices. Aggression and Violent Behavior, 12, 393–401.

Cauffman, E., Piquero, A. R., Kimonis, E., Steinberg, L., Chassin, L., & Fagan, J. (2007). Legal, individual, and environmental predictors of court disposition in a sample of serious adolescent offenders. Law and Human Behavior, 31, 519–535

Coalition for Juvenile Justice (2014). Deinstitutionalization of status offenders (DSO): Facts and resources. Washington, D.C. Available at: http://www.juvjustice.org/sites/default/files/resource- files/DSO%20Fact%20Sheet%202014.pdf

Crutchfield, R. D., Bridges, G. S., & Pitchford, S. R. (1994). Analytical and aggregation biases in analyses of imprisonment: Reconciling discrepancies in studies of racial disparity. Journal of Research in Crime and Delinquency, 31, 166–182.

Crutchfield, R. D., Fernandes, A., & Martinez, M. (2010). Racial and ethnic disparity and criminal justice: How much is too much? Journal of Criminal Law and Criminology, 100, 903–932.

Eisenstein, J., Flemming, R. B., & Nardulli, P. F. (1988). The contours of justice: Communities and their courts. Boston: Little, Brown.

Davis, J., & Sorensen, J. R. (2012). Disproportionate juvenile minority confinement: a state-level assessment of racial threat. Youth Violence and Juvenile Justice, 11, 296–312.

DeJong, C., & Jackson, K. C. (1998). Putting race into context: Race, juvenile justice processing, and urbanization. Justice Quarterly, 15, 487–504.

Durante, K. A. (2020). Racial and ethnic disparities in prison admissions across counties: an evaluation of racial/ethnic threat, socioeconomic inequality, and political climate explanations. Race and Justice, 10, 176–202.

Elwert, F., & Winship, C. (2014). Endogenous selection bias: The problem of conditioning on a collider variable. Annual Review of Sociology, 40, 31–53.

Enders, W., Pecorino, P., & Souto, A. C. (2019). Racial disparity in U.S. imprisonment across states and over time. Journal of Quantitative Criminology, 35, 365–392.

Engen, R. L., Steen, S., & Bridges, G. S. (2002). Racial disparities in the punishment of youth: a theoretical and empirical assessment of the literature. Social Problems, 49, 194–220.

Ericson, R. D., & Eckberg, D. A. (2016). Racial disparity in juvenile diversion: The impact of focal concerns and organizational coupling. Race and Justice, 6, 35–56.

Feld, B. C. (1991). Justice by geography: Urban, suburban, and rural variations in juvenile justice administration. The Journal of Criminal Law and Criminology, 82, 156–210.

Feld, B. C. (2017). The evolution of the juvenile court: Race, politics, and the criminalizing of juvenile justice. New York: New York University Press.

Franklin, T. W., & Henry, T. K. S. (2020). Racial disparities in federal sentencing outcomes: Clarifying the role of criminal history. Crime & Delinquency, 66, 3–32.

Frase, R. S. (2013). Research on race and sentencing: Goals, methods, and topics. Justice Quarterly, 30, 262–269.

Freiburger, T. L., & Jordan, K. L. (2011). A multilevel analysis of race on the decision to petition a case in the juvenile court. Race and Justice, 1, 185–201.

Frost, N. A. (2009). The mismeasure of punishment: Alternative measures of punitiveness and their (substantial) consequences. Punishment & Society, 10(3), 277–300.

Gann, S. M. (2019). Examining the relationship between race and juvenile court decision-making: A counterfactual approach. Youth Violence and Juvenile Justice, 17, 269–287.

Garland, D. (2013). Penality and the penal state. Criminology, 51(3), 475–517.

Guevara, L., Boyd, L. M., Taylor, A. P., & Brown, R. A. (2011). Racial disparities in juvenile court outcomes: A test of the liberation hypothesis. Journal of Ethnicity in Criminal Justice, 9, 200–217.

Hawkins, D. F., & Hardy, K. A. (1989). Black-White imprisonment rates: a state-by-state analysis. Social Justice, 16, 75–94.

Hester, R. (2017). Judicial rotation as centripetal force: Sentencing in the court communities of South Carolina. Criminology, 55, 205–235.

Hockenberry, S. & Puzzanchera, C. (2019). Juvenile court statistics 2017. Pittsburgh, PA: National Center for Juvenile Justice.

Hughes, C., Warren, P. Y., Stewart, E. A., Tomaskovic-Devey, D., & Mears, D. P. (2017). Racial threat, intergroup contact, and school punishment. Journal of Research in Crime and Delinquency, 54, 583–616.

Kempf-Leonard, K. (2007). Minority youths and juvenile justice: Disproportionate minority contact after nearly 20 years of reform efforts. Youth Violence and Juvenile Justice, 5, 71–87.

Krisberg, B., Schwartz, I., Fishman, G., Eisikovits, Z., Guttman, E., & Joe, K. (1987). The incarceration of minority youth. Crime & Delinquency, 33, 173–205.

Leiber, M. J. (2002). Disproportionate minority confinement (DMC) of youth: An analysis of state and federal efforts to address the issue. Crime & Delinquency, 48, 3–45.

Leiber, M. J. (2013). Race, pre- and postdetention, and juvenile justice decision making. Crime & Delinquency, 59, 396–418.

Leiber, M. J., & Fix, R. (2019). Reflections on the impact of race and ethnicity on juvenile court outcomes and efforts to enact change. American Journal of Criminal Justice, 44, 581–608.

Leiber, M. J., & Fox, K. C. (2005). Race and the impact of detention on juvenile justice decision making. Crime & Delinquency, 51, 470–497.

Leiber, M. J., & Peck, J. H. (2013). Race in juvenile justice and sentencing policy: An overview of research and policy recommendations. Law & Inequality: A Journal of Theory and Practice, 31, 331–368.

Leiber, M. J., & Peck, J. H. (2015). Race, gender, crime severity, and decision making in the juvenile justice system. Crime & Delinquency, 61, 771–797.

Leiber, M. J., & Peck, J. H. (2019). Clarifying the theoretical tenants of the symbolic threat perspective. Justice Quarterly. doi: 10.1080/07418825.2019.1593488.

Leiber, M., Brubaker, S. J., & Fox, K. (2009). A closer look at the individual and joint effects of gender and race on juvenile justice decision making. Feminist Criminology, 4, 333–358.

Leiber, M. J., Peck, J. H., & Rodriguez, N. (2016). Minority threat and juvenile court outcomes. Crime & Delinquency, 62, 54–80.

Lynch, M. (2019). Focally concerned about focal concerns: A conceptual and methodollgicla critique of sentencing disparities research. Justice Quarterly, 36, 1148–1175.

MacDonald, J. M., & Chesney-Lind, M. (2001). Gender bias and juvenile justice revisited: a multiyear analysis. Crime & Delinquency, 47, 173–195.

Maggard, S. R., Higgins, J. L., & Chappell, A. T. (2013). Pre-dispositional juvenile detention: an analysis of race, gender and intersectionality. Journal of Crime and Justice, 36, 67–86.

McCoy, T., Walker, J. T., & Rodney, H. E. (2012). Predicting preadjudication detention decisions: an examination of family status and race. Journal of Ethnicity in Criminal Justice, 10, 87–107.

McGarrell, E. F. (1991.) Differential effects of juvenile justice reform on incarceration rates of the states. Crime & Delinquency, 37, 262–280.

Mears, D. P. (2006). Exploring state-level variation in juvenile incarceration rates: Symbolic threats and competiting explanations. The Prison Journal, 86, 470–490.

Mears, D. P. (2017). Out-of-control criminal justice: The systems improvement solution for more safety, justice, accountability, and efficiency. Cambridge: Cambridge University Press.

Mears, D. P., Cochran, J. C., & Lindsey, A. M. (2016). Offending and racial and ethnic disparities in criminal justice: a conceptual framework for guiding theory and research and informing policy. Journal of Contemporary Criminal Justice, 32, 78–103.

Mears, D. P., Cochran, J. C., Stults, B. J., Greenman, S. J., Bhati, A. S., & Greenwald, M. A. (2014). The “true” juvenile offender: Age effects and juvenile court sanctioning. Criminology, 52, 169–194.

Mize, T. D. (2019). Best practices for estimating, interpreting, and presenting nonlinear interaction effects. Sociological Science, 6, 81–117.

Mustillo, S. A., Lizardo, O. A., & McVeigh, R. M. (2018). Editors’ comments: A few guidelines for quantitative submissions. American Sociological Review, 83, 1281–1283.

Nellis, A., & Richardson, B. (2010). Getting beyond failure: Promising approaches for reducing DMC. Youth Violence and Juvenile Justice, 8, 266–276.

Pasko, L. (2002). Villain or victim: Regional variation and ethnic disparity in federal drug offense sentencing. Criminal Justice Policy Review, 13, 307–328.

Peck, J. H., & Beaudry-Cyr, M. (2016). Does who appears before the juvenile court matter on adjudication and disposition outcomes? The interaction between client race and lawyer type. Journal of Crime and Justice, 39, 131–152.

Peck, J. H., & Jennings, W. G. (2016). A critical examination of “being Black” in the juvenile justice system. Law and Human Behavior, 40, 219–232.

Peck, J. H., Leiber, M. J., & Beaudry-Cyr, M. (2019). Expanding the applicability of Sampson and Laub’s theory of inequality and social control: A multilevel examination. Criminal Justice and Behavior, 46, 902–919.

Peck, J. H., Leiber, M. J., Beaudry-Cyr, M., & Toman, E. L. (2016). The conditioning effects of race and gender on the juvenile court outcomes of delinquent and “neglected” types of offenders. Justice Quarterly, 33, 1210–1236.

Puzzanchera, C., Sladky, A., & Kang, W. (2019). Easy access to juvenile populations. Available at: https://www.ojjdp.gov/ojstatbb/ezapop/.

Rodriguez, N. (2007). Juvenile court context and detention decisions: Reconsidering the role of race, ethnicity, and community characteristics in juvenile court processes. Justice Quarterly, 24, 629–656.

Rodriguez, N. (2010). The cumulative effect of race and ethnicity in juvenile court outcomes and why preadjudication detention matters. Journal of Research in Crime and Delinquency, 47, 391–413.

Rodriguez, N. (2013). Concentrated disadvantage and the incarceration of youth: Examining how context affects juvenile justice. Journal of Research in Crime and Delinquency, 50, 189–215.

Rodriguez, N., Smith, H., & Zatz, M. S. (2009). Youth is enmeshed in a highly dysfunctional family system: Exploring the relationship among dysfunctional families, parental incarceration, and juvenile court decision making. Criminology, 47, 177–208.

Sampson, R. J., & Laub, J. H. (1993). Structural variations in juvenile court processing: Inequality, the underclass, and social control. Law & Society Review, 27, 285–311.

Sampson, R. J., & Lauritsen, J. L. (1997). Racial and ethnic disparities in crime and criminal justice in the United States. Crime and Justice, 21, 311–374.

Secret, P. E., & Johnson, J. B. (1997). The effect of race on juvenile justice decision making in Nebraska: Detention, adjudication, and disposition, 1988–1993. Justice Quarterly, 14, 445–478.

Sorensen, J., Hope, R., & Stemen, D. (2003). Racial disproportionality in state prison admissions: Can regional variation be explained by differential arrest rates? Journal of Criminal Justice, 31, 73–84.

Spinney, E., Cohen, M., Feyerherm, W., Stephenson, R., Yeide, M., & Shreve, T. (2018). Disproportionate minority contact in the U.S. juvenile justice system: A review of the DMC literature, 2001–2014, part 1. Journal of Crime and Justice, 41, 573–595.

Stapleton, V., Aday, D. P., & Ito, J. A. (1982). An empirical typology of American metropolitan juvenile courts. American Journal of Sociology, 88, 549–564.

Stringer, R. J., & Holland, M. M. (2016). It’s not all black and white: A propensity score matched, multilevel examination of racial drug sentencing disparities. Journal of Ethnicity in Criminal Justice, 14, 327–347.

Thomas, S. A., Moak, S. C., & Walker, J. T. (2013). The contingent effect of race in juvenile court detention decisions: The role of racial and symbolic threat. Race and Justice, 3, 239–265.

Ulmer, J. T. (2019). Criminal courts as inhabited institutions: Making sense of difference and similarity in sentencing. Crime and Justice, 48, 483–522.

van Wingerden, S., Wilsem, J. & Johnson, B. D. 2016. Offender’s personal circumstances and punishment: Toward a more refined model for the explanation of sentencing disparities. Justice Quarterly, 33, 100–133.

Zimring, F. E. (2014). Minority overrepresentation: On causes and partial cures. In F. E. Zimring & D. S. Tanenhaus (eds.), Choosing the future for American juvenile justice (169–186). New York: New York University Press.

Tables

Table 1. Descriptive statistics

^a N₁ = 100,358; N₂ = 146

^b N₁ = 99,344; N₂ = 146

^c N₁ = 57,529; N₂ = 146

^d N₁ = 33,799; N₂ = 145

Table 2. Hierarchical logistic regression models

^a N₁ = 100,358; N₂ = 146

^b N₁ = 99,344; N₂ = 146

^c N₁ = 57,529; N₂ = 146

^d N₁ = 33,799; N₂ = 145

^e Reference category: White

^f Reference category: Property offense

^g Reference category: Alabama  

^h Reference categories: Alabama; White

^† < .10 * < .05 ** < .01 *** < .001 (two-tailed)

Table 3a. Probability of detention by Black-White and state: marginal effects across states

Notes: The “contrasts” column reports which marginal effects are significantly different across states (second differences) at p < .05

^† < .10 * < .05 ** < .01 *** < .001 (two-tailed)

Table 3b. Probability of detention by Hispanic-White and state: marginal effects across states

Notes: The “contrasts” column reports which marginal effects are significantly different across states (second differences) at p < .05

^† < .10 * < .05 ** < .01 *** < .001 (two-tailed)

Table 4a. Probability of petition by Black-White and state: marginal effects across states

Notes: The “contrasts” column reports which marginal effects are significantly different across states (second differences) at p < .05

^† < .10 * < .05 ** < .01 *** < .001 (two-tailed)

Table 4b. Probability of petition by Hispanic-White and state: marginal effects across states

Notes: The “contrasts” column reports which marginal effects are significantly different across states (second differences) at p < .05

^† < .10 * < .05 ** < .01 *** < .001 (two-tailed)

Table 5a. Probability of adjudication of delinquency by Black-White and state: marginal effects across states

Notes: The “contrasts” column reports which marginal effects are significantly different across states (second differences) at p < .05

^† < .10 * < .05 ** < .01 *** < .001 (two-tailed)

Table 5b. Probability of adjudication of delinquency by Hispanic-White and state: marginal effects across states

Notes: The “contrasts” column reports which marginal effects are significantly different across states (second differences) at p < .05

^† < .10 * < .05 ** < .01 *** < .001 (two-tailed)

Table 6a. Probability of secure placement by Black-White and state: marginal effects across states

Notes: The “contrasts” column reports which marginal effects are significantly different across states (second differences) at p < .05

^† < .10 * < .05 ** < .01 *** < .001 (two-tailed)

Table 6b. Probability of secure placement by Hispanic-White and state: marginal effects across states

Notes: The “contrasts” column reports which marginal effects are significantly different across states (second differences) at p < .05

^† < .10 * < .05 ** < .01 *** < .001 (two-tailed)

[1] A few studies have also used regression analysis to examine regional variation in disparities. Pasko (2002) assessed racial disparities in sentence length for drug offenders across federal courts in the East, Midwest, West, and South, and found that race was positively associated with sentence length in the West only, while Hispanic ethnicity was positively associated with sentence length in the East only. Elsewhere, using correctional data for 38 states, Durante (2020) regressed Black-White and Hispanic-White imprisonment ratios on contextual predictors, including region. The author found that the South was negatively associated with Black-White disproportionality but not associated with Hispanic-White disproportionality. In the juvenile justice context, Davis and Sorenson (2012) examined Black-White placement ratios across 38 states over five time periods; they found no significant association between racial disproportionality and Southern region.

[2] As Mears and colleagues (2016, pp. 85, 87) point out, “disproportionality” refers to “racial or ethnic differences that are greater than what would be expected given the group population sizes among those for whom a given outcome is possible,” while “disparity” is often taken to indicate “any disproportionality attributable to overt or covert, or intended or unintended, discrimination against minorities.” While aggregate analyses can illustrate disproportionalities, they are limited in identifying disparities, which requires controlling for possibly confounding case-level and contextual factors. 

[3] Scholars have suggested several possibilities for why DMC varies across stages of processing (see Engen et al., 2002; Peck & Jennings, 2016). For example, Bishop and colleagues (2010) suggest that each stage of processing will have distinct focal concerns, with some stages focused more on punishment and others more focused on needs assessment and treatment, and this may account for variation in racial and ethnic disparities. 

[4] Feld (1991, 2017) argues that in the wake of In re Gault (1967), the Supreme Court decision that granted juvenile defendants many due process rights, urban courts responded by “criminalizing” juvenile justice. According to this argument, while rural jurisdictions largely retained the traditional rehabilitative approach—the “pre-Gault” orientation—urban courts shifted to a greater emphasis on due process—the “post-Gault” orientation. Feld (2017) suggests that this more legalistic orientation is less constrained by the original court’s rehabilitative mission and is, by extension, more punitive. This legalism, however, may also contribute to less individualized discretion and hence fewer disparities. As a result, in urban jurisdictions there might be a more equitable but also more punitive court, while in rural jurisdictions there may be a less punitive but also less equitable one (see Zimring, 2014).

[5] Others have examined variation in the severity of juvenile court outcomes across states for all defendants rather than variation in racial disparities (see, e.g., Mears, 2006).

[6] Using 2000–2012 data from the National Corrections Reporting Program, Stringer and Holland (2016) employed hierarchical linear modeling to examine regional variation in sentence length for drug offenders. The authors computed cross-level interactions between race and region, finding that Black-White sentence length disparities were significantly lower in the South. To our knowledge, this is the only prior study to examine regional variation in racial disparities using multilevel modeling with cross-level interactions.

[7] These data were originally collected by Administrative Office of the Courts, Alabama; Judicial Branch, Connecticut; South Carolina Department of Juvenile Justice, South Carolina; and Administrative Office of the Courts, Utah. These agencies and the National Center for Juvenile Justice bear no responsibility for the analyses or interpretations presented herein.

[8] Our unit of analysis is referral, not individual youth. Some youths had multiple referrals disposed in 2010. There were approximately 83,651 unique youth among the final sample of referrals (N = 100,358). Since not all states reported case numbers, we identified unique youth based on county, date of birth, race, and sex.

[9] The data do not indicate whether an adjudication of delinquency was the result of plea bargaining or a full hearing.

[10] It is noteworthy that South Carolina exhibited the lowest rate of petition (41 percent) and the highest rates of adjudication (93 percent) and commitment (23 percent). This likely reflects a difference in institutionalized case processing where only the most serious cases are petitioned, most of which are then adjudicated delinquent (and many of which are committed).

[11] In 2010, Black youth made up approximately 31 percent of the youth population (0–17) in Alabama, 12 percent in Connecticut, 34 percent in South Carolina, and 2 percent in Utah, while Hispanic youth made up approximately 6 percent of the youth population in Alabama, 20 percent in Connecticut, 8 percent in South Carolina, and 17 percent in Utah (Puzzanchera, Sladky, & Kang, 2019). This indicates that at referral, Black youth were over-represented by 45 percent in Alabama, 92 percent in Connecticut, 68 percent in South Carolina, and 50 percent in Utah. Hispanic youth were over-represented at referral by 15 percent in Connecticut, while they were actually under-represented in Alabama, South Carolina, and Utah.

[12] The Deinstitutionalization of Status Offenders (DSO) requirement of the Juvenile Justice and Delinquency Prevention Act (JJDPA) provides that states that accept federal funding under the JJDPA must not place status offenders in secure detention or locked confinement. However, many states—including Alabama, South Carolina, and Utah—exercise the valid court order (VSO) exception to the DSO requirements. As a result, status offenders were still eligible for detention and placement in these states. Moreover, even though Connecticut moved to completely deinstitutionalize status offenders in 2007, exceptions could still be made for repeat offenders (see Coalition for Juvenile Justice, 2014). This was confirmed in our data, with status offenders detained and placed in all four states (ranging from 1 percent detained in Connecticut to 15 percent detained in South Carolina).

[13] A count measure or description of the type of prior referrals would have been preferable; however, not all states provided such data. Additionally, data on prior petitions or adjudications was not available. 

[14] Correlations between variables were generally weak to moderate (< .5) and tests for multicollinearity (variance inflation factor [VIF]) indicated no serious concerns across all models (VIF < 5). (The exceptions were several high correlations between contextual control variables: percent non-White was strongly correlated with income inequality (.63), and crime rates (.59), and conservatism (-.75), while conservatism was strongly correlated with income inequality (-.51) and crime rates (-.57); the largest VIF values were 4.28 for percent non-Whiten and 3.27 for conservatism, while other values were < 2.5).

[15] We also ran a 3-level unconditional model with cases nested within counties (level-2) nested within states (level-3). Here, intraclass correlation is calculated for each level of the model (Johnson, 2010). The formula for level-3 intraclass correlation is ρ = (ψ³) / (ψ² + ψ³ + p²/3); the formula for level-2 intraclass correlation is ρ = (ψ² ) / (ψ² + ψ³ + p²/3). The results indicated significant variation in detention across states as well as counties, with a level-3 variance component of .12 (ψ³) and a level-2 variance component (ψ²) of .62. This indicates that before any predictors are added to the model, approximately 3 percent of the variation in detention outcomes is attributable to between-state differences, while 15 percent is due to county-level differences.

[16] We also ran a 3-level unconditional model. The results indicated significant variation in petition across states as well as counties, with ψ³ =.43 and ψ² = 1.22. This indicates that before any predictors are added to the model, approximately 9 percent of the variation in petition outcomes is attributable to between-state differences, while 25 percent is attributable to county-level differences.

[17] We also ran a 3-level unconditional model. The results indicated significant variation in adjudication across states as well as counties, with ψ³ = 1.72 and ψ² = 1.13. This indicates that before any predictors are added to the model, approximately 28 percent of the variation in adjudication outcomes is attributable to between-state differences, while 18 percent is due to county-level differences.

[18] We also ran a 3-level unconditional model. The results indicated significant variation in placement across states as well as counties, with ψ³ =.35 and ψ² = .70. This indicates that before any predictors are added to the model, approximately 8 percent of the variation in placement outcomes is attributable to between-state differences, while 16 percent is due to county-level differences.

[19] Among our states, there are several notable differences in how juvenile justice systems are organized. For example, only South Carolina involved the prosecutor as the primary decision-maker at intake, only Alabama involved a largely decentralized juvenile justice system, and a state parole board determines release in South Carolina and Utah (compared to a court agency in Alabama and Connecticut).

[20] Given the standard rule of 10 level-2 observations per level-2 variable (Johnson, 2010), this would require 10 states to test one state-level hypothesis—and an analysis of the full population of 51 juvenile justice systems would still be constrained to approximately five hypotheses.