Abstract

Objectives. To test the cumulative disadvantage hypothesis—that system-level racial and ethnic disparities accumulate from intake to final disposition—by investigating relative and absolute disparities across different pathways through the juvenile justice system. Methods. Using a sample of 95,670 juvenile court referrals across 140 counties in four states, the present study employed multinomial logistic regression to examine racial and ethnic disparities across 14 possible combinations of juvenile justice outcomes (i.e., pathways), ranked from least to most punitive. We then estimated predicted probabilities and marginal effects of race and ethnicity for each pathway. Results. We found limited support for the cumulative disadvantage hypothesis. Racial and ethnic disparities were greatest for the most punitive pathways, but the findings do not point to extensive evidence of cumulative disadvantage. Specifically, neither relative nor absolute disparities accumulated from least to most punitive pathways, and some of the least punitive pathways were actually more likely for minority defendants. Conclusions. The results underscore the need for more careful measurement and analysis of disadvantage and disparities in the criminal and juvenile justice systems. In particular, more attention should be paid to early outcomes such as detention, where large differences between racial and ethnic groups were observed, as well as to relative and absolute differences in processing outcomes.

Introduction

Racial and ethnic disparities in case processing outcomes in the juvenile and criminal justice systems constitute a persistent social problem (Enders, Pecorino, and Souto 2019). As Sampson and Lauritsen (1997: 332) have observed, the overrepresentation of minority defendants is at odds with “the symbolism of equality before the law” in the United States, posing a direct threat to the perceived legitimacy and fairness of the legal system. Given these high stakes, racial disparities in the criminal and juvenile justice system have been the focus of much scholarly inquiry over the years (see, for example, Beck and Blumstein 2018; Blumstein 1982, 1993; Crutchfield, Bridges, and Pitchford 1994; Piquero 2008; Spohn 2015). A general limitation of much of this research, however, is that it exhibits “a narrow focus on episodic disparity in individual outcomes at singular stages of criminal case processing” and thus may fail to capture the full impact of race over multiple stages of processing (Kurlychek and Johnson 2019: 307). This is problematic on several grounds. First, focusing only on final outcomes introduces sample selection issues, creating the possibility of over- or under-estimating the true effects of race on processing decisions (Baumer 2013). Second, a focus on final outcomes can obscure more subtle ways in which inequalities can accumulate.

A complementary perspective on racial disparities might instead focus on system-level disparities in terms of combinations of outcomes experienced from initial referral to final disposition, what has been referred to as pathways through the justice system. In an early recognition of this alternative approach, Hagan (1974: 379) argued that research needed to focus on the “transit through the criminal justice system” and the extent to which this dynamic process works “cumulatively to the disadvantage of minority group defendants” (see also Crutchfield et al. 1994). In recent years, scholars in sentencing research have begun to echo this observation. Baumer (2013: 240), for instance, calls for future research to examine the impact of race and ethnicity “across multiple stages of the criminal justice process” (see also Crutchfield et al. 2010: 909). As Piquero (2008: 69) emphasizes, “it may be that racial and ethnic disparities begin at the very earliest stage and that effects accumulate as youth proceed through the system.” Extending that idea, Kutateladze and colleagues (2014: 515) point out that “to understand better the locus and magnitude of racial differences in punishment, it is useful to conceptualize the punishment process as a dynamic set of interrelated decisionmaking points.” For these reasons, and echoing Spohn’s (2015) call for a “fifth wave” of research on race and sentencing, Johnson (2015: 183) encourages future research to move “from single-stage analyses that present only a snapshot of individual outcomes toward more dynamic, multistage investigations that account for both the direct and the indirect causal pathways that can contribute to racial disadvantage.”

Scholars have called, at the same time, for examining cumulative disadvantage in justice system processing. Cumulative disadvantage involves small disparities that accumulate over time or from one decision-making point to another, resulting in growing inequality (DiPrete and Eirich 2006). In the criminal justice context, cumulative disadvantage “specifically refers to the potentially discriminatory effects that accrue across criminal justice domains and over time, where the emphasis is on dynamic and systemic processes that accentuate unequal treatment and disparate outcomes in criminal punishments” (Kurlychek and Johnson 2019: 296). There are at least two ways of conceiving of cumulative disadvantage in the criminal justice context. First, we might envision path dependency for individual referrals, where early decisions influence later ones. Research on the indirect effects of race across multiple outcomes can be conceived as one way to measure cumulative disadvantage. For example, Wooldredge and colleagues (2015) recently examined how early racial disparities in pretrial detention influenced later disparities at sentencing, possibly compounding the effects of race.

Another approach to thinking about cumulative disadvantage takes a more holistic view of racial disparities in overall system processing (Kurlychek and Johnson 2019). We can envision possible pathways through the justice system—combinations of outcomes from initial contact through final disposition—and then ask whether certain racial and ethnic groups are more likely to experience the most punitive pathways. Most prior research examines discrete decision-making points. By contrast, a system-level approach emphasizes that some groups may experience accumulated disadvantage relative to other groups. That is, small disparities at sequential stages of processing may contribute to overall group disadvantage that is greater than can be observed at any individual point. This represents an alternative (and complementary) way of conceiving of disparity and inequality in the justice system, one that allows for contributions to the large research literature on the direct effects of race on individual outcomes (see Bishop and Leiber 2012), and the smaller but growing literature on indirect effects (e.g., Wooldredge et al. 2015). Few studies have attempted to assess cumulative disadvantage in the criminal justice system across multiple stages, and only two have attempted to estimate disparities across system pathways (Kutateladze, Andiloro, Johnson, and Spohn 2014; Sutton 2013). In addition, no study has done so in the juvenile justice system. This gap is notable because cumulative disadvantage may be especially pronounced in the juvenile court, where there is more discretion given the court’s paternalistic mission and the dual emphasis both on punishing youth and on acting, per the doctrine of parens patriae, in their best interests (Feld 2017).

To investigate cumulative disadvantage across major stages of juvenile justice processing, the present study assesses whether racial and ethnic disparities are greater for more punitive combinations of outcomes (i.e., pathways). This is accomplished through analysis of almost 100,000 referrals across 140 counties in four states, estimating the relative probabilities that different groups experience at each of 14 possible case processing pathways. We begin by situating the study’s focus and approach in the context of prior research.

Background

Racial and Ethnic Disparities in Juvenile Justice Processing

Pathways through the juvenile justice system consist of a series of interrelated decision points from initial referral to final disposition. One of the first and most consequential decisions is preadjudication detention. An intake department official will usually make a recommendation regarding detention or release (pending further processing), and the juvenile court judge must decide whether detention is warranted. In juvenile court, this is not determined by financial considerations (as in criminal court), but by considerations of community safety, the juvenile’s safety, and extralegal factors such as parental availability and cooperation (Barton 2012). A large body of research has investigated disparities at detention. It generally finds that Black and Hispanic defendants are more likely than White youth to be detained (e.g., Leiber 2013; Leiber and Fox 2005; Rodriguez 2010; but see Rodriguez 2007).

Another early decision point involves the formal petition of delinquency. In most jurisdictions, the intake department will recommend whether a case should be handled formally and, if so, a petition will be filed by the prosecutor and the case will be placed on the juvenile court calendar for an adjudicatory hearing (notably, jurisdictions vary considerably in how the intake officer and prosecutor interact at this stage; see Fairchild, Gupta-Kagan, and Andersen 2019). Otherwise, the prosecutor and intake department may choose to divert the case to informal processing (such as a suspended sentence or consent decree) or dismiss the case outright (Mears 2012). Research at the intake and petition stages has also found some evidence of racial and ethnic disparities, albeit smaller than at detention (e.g., Bishop, Leiber, and Johnson 2010; Leiber and Peck 2015). Although less studied, research also finds that Black and Hispanic youth are less likely to be diverted from formal processing than White youth and may even be more likely to have their cases dismissed rather than diverted (e.g., Cochran and Mears 2015).

At the adjudicatory hearing, which functions as the juvenile court equivalent of a criminal trial, the judge determines whether the juvenile defendant committed the charged offense. The outcome is either an adjudication of delinquency or dismissal of the case. Studies that have examined disparities at adjudication have largely found that minority youth are in fact less likely to be adjudicated delinquent than similarly situated White youth (Bishop and Leiber 2012). One possible explanation for this is a correction effect by judges who see minority overrepresentation at earlier stages of processing that may arise from referral of weaker cases (see Rodriguez 2007).

Prior to adjudication, the intake department or prosecutor may also recommend that a juvenile case be removed from the juvenile court’s original jurisdiction and transferred to criminal court. While juvenile waiver usually occurs before the adjudication stage of processing, it is considered the “capital punishment of juvenile justice” since it transfers adolescent offenders to the criminal court where more severe punishments are available (Zimring 1981: 193). At a waiver hearing, the juvenile court judge decides whether the case merits criminal prosecution. Evidence of racial and ethnic disparities at waiver is mixed (Zane, Welsh, and Drakulich 2016).

For cases that remain in juvenile court and are adjudicated delinquent, the juvenile court judge also determines the defendant’s disposition. Most often, this involves a decision between out-of-home placement and community supervision. Here, research is also mixed. While many studies show that minority youth are more likely to receive secure placement (e.g., Rodriguez 2010, 2013; Rodriguez, Smith, and Zatz 2009), others find no direct race effects (see, for example, Cauffman et al. 2007), and some have even found that Black youth receive more lenient dispositions (e.g., Bishop et al. 2010; Leiber 2013).

Kurlychek and Johnson (2019: 307) have observed that in the criminal justice system, “the overwhelming majority of research on inequality in punishment is limited to singular outcomes of interest, usually at a specific stage of criminal case processing.” The same holds true of the research in juvenile justice. This is problematic, given that a focus on single decision points may mask disparities experienced over multiple stages of processing. Moreover, as others have pointed out, endogenous selection bias threatens research into processes with non-random reductions in sample size (see Elwert and Winship, 2014). If, for example, Black juvenile defendants are more likely to be petitioned than similarly situated White defendants due to racial bias, the sample of petitioned referrals is now different from the sample of total referrals. If there are no differences in dispositional outcomes between Black and White juveniles, this might signal that there is no bias—but this does not account for the bias at petition.[1]

For discrete outcomes such as judicial disposition (or its criminal justice equivalent, sentencing), many researchers have employed the Heckman two-step procedure to account for potential selection effects. As Bushway and colleagues (2007) have pointed out, however, this adjustment requires an exclusion restriction (i.e., an instrumental variable); valid exclusion restrictions are typically not available in the data used in criminal justice research, unfortunately.[2] Another approach is to examine all possible combinations of outcomes among the total sample of referrals—pathways through the juvenile justice system. This allows for an examination of growing system-level inequalities that can be missed by examining isolated disparities at discrete decision points where selection effects are present.

Cumulative Disadvantage in Justice System Processing

There have been calls for increased attention to cumulative disadvantage in the criminal justice system (Johnson 2015; Kurlychek and Johnson 2019; Spohn 2015). Cumulative disadvantage can be conceived broadly in terms of compounding disadvantages that certain persons or groups experience over time. In this sense, cumulative disadvantage refers to the manner in which “small differences over time . . . makes it difficult for an individual or group that is behind at a point in time . . . to catch up” (DiPrete and Eirich 2006: 272). In fact, though, the term has been employed in myriad ways, leading to some confusion in the literature.

There are two distinct ways to conceive of cumulative racial disadvantage in the criminal justice context (Kurlychek and Johnson 2019). The first is to think about indirect effects of race and ethnicity on final outcomes as mediated by earlier outcomes, such as pretrial detention (e.g., Rodriguez 2007) or charging decisions (e.g., Ward, Hartley, and Tillyer 2016). Here, the focus is on cumulative disadvantage for an individual case: early negative outcomes make later negative outcomes more likely by signaling certain attributes, such as culpability or dangerousness, to later decision makers.[3] For example, if pretrial detention acts as a signal of dangerousness to judges, this might explain why detention is consistently associated with more punitive sentencing outcomes (e.g., Wooldredge et al. 2015). Another example is that criminal court judges may view youth transferred from the juvenile justice system as especially blameworthy and treat them more punitively than similarly situated adult defendants (Johnson and Kurlychek 2012; Kurlychek and Johnson 2010; Steiner 2009). This approach sees cumulative disadvantage as an “additive and interactional process, comprising the persistent direct and indirect effects that status variables exert over the life course” (Kurlychek and Johnson 2019: 291).[4]

A second way to conceptualize cumulative disadvantage in the criminal justice system is in terms of increasingly adverse effects that some groups face relative to others, such that minority defendants experience ever greater disadvantage for more punitive combinations of system outcomes. As Sutton (2013: 1219) observes, a “strict definition of cumulative disadvantage . . . requires evidence of accelerating status differences across life events—in this case, widening disparities between Anglos and minorities.” The logic reduces to the following: if differential treatment results in disadvantaged outcomes at each decision point, then overall disadvantage for some groups relative to others will accumulate from early to later case processing, resulting in greater overall disadvantage than is apparent at any one decision point. Fortunately, this possibility can be investigated in terms of pathways of particular cases from early to later case processing. Cumulative disadvantage will exist where “minority defendants experience enhanced probabilities of certain combinations of less favorable case-processing outcomes” (Kutateladze et al. 2014: 520). It is this second view of group-level cumulative disadvantage within the juvenile justice system that is the focus of the current study.

Prior Attempts to Measure System-Level Cumulative Disadvantage

The present study responds to a critique that Kurlychek and Johnson (2019: 308) recently made in a review of the literature: “Although countless research studies have investigated inequality in specific stages of the punishment process, few attend to the accumulation of disadvantage across the multiple, interrelated stages of criminal case processing.” Attempts to measure cumulative disadvantage in the juvenile justice system to date have been largely descriptive, comparing the relative proportions of minority defendants at different stages of processing. Based on such descriptive analyses, for example, McGuire (2002) concluded that cumulative disadvantage was not present in the Missouri juvenile justice system because the greatest disparities were observed at detention, early on in case processing.

A more recent approach to measuring cumulative disadvantage is to estimate cumulative effects by calculating predicted possibilities for different pathways through the justice system. To date only two studies have estimated the cumulative effects of race using this approach, both in the criminal justice context (Kutateladze et al. 2014; Sutton 2013). In a sample of 11,505 male defendants nested in 40 counties, Sutton (2013) found that Black and Latino defendants had an approximately 26 percent higher probability of ending up in prison compared to White defendants. To estimate cumulative disadvantage more precisely, Sutton (2013) computed conditional probabilities (based on multilevel logistic and ordered logistic regression) for four possible pathways that could lead to a prison term: (1) no detention and guilty plea; (2) no detention and guilty trial verdict; (3) detention and guilty plea; and (4) detention and guilty trial verdict. Of these paths, one stood out: minority defendants were overrepresented by more than 200 percent among those who were detained, received a guilty verdict at trial, and then were imprisoned. For the two non-detention pathways, however, racial and ethnic disparities largely disappeared. Sutton (2013: 1219) concluded that while minority defendants were more likely to end up in prison, evidence of cumulative disadvantage—defined as “widening disparities” among racial and ethnic groups—was less clear since the greatest disparity occurred at detention, and sentencing decisions did not exacerbate overall disparities.

In the only other study to estimate cumulative disadvantage via possible pathways through the criminal justice system, Kutateladze and colleagues (2014) examined a large sample (N = 164,783) of New York City criminal cases from 2010 to 2011. The authors looked at racial disparities across five discrete stages of criminal justice processing: case acceptance, pretrial detention, case dismissal, custodial plea offer, and incarceration. They then examined cumulative disadvantage by estimating predicted probabilities for six pathways based on these outcomes—ranging from detained and later incarcerated (the most disadvantaged) to dismissed without pretrial detention (the least disadvantaged). Although Black and Hispanic defendants were more likely to experience the most disadvantaged of the six pathways, no other clear patterns emerged. The authors concluded that “research on cumulative disadvantages in the justice system remains in its infancy,” and called for future work to “account for the multiple and interrelated stages of criminal cases processing” (Kutateladze et al. 2014: 539).

Current Study

This study responds to calls by scholars to understand better the ways in which racial and ethnic disparities may arise from dynamic decision-making that unfolds across multiple processing decisions. It responds, too, to calls to investigate cumulative disadvantage in juvenile justice, a setting where the emphasis both on punishment and on advancing the best interests of youth may provide a setting for discretion to result in greater disparities in the punishment of minority youth (Cochran and Mears 2015).[5]

Specifically, this study investigates the hypothesis that racial and ethnic disparities accumulate from early to later processing, resulting in a system-level pattern of cumulative disadvantage for minority defendants. It adopts a pathways perspective and examines juvenile court referrals that could experience any of 14 possible combinations of outcomes from referral to final outcomes. These pathways, as discussed below, can be ranked from least to most punitive and give rise to the following hypothesis:

H: Compared to White defendants, Black and Hispanic defendants will have an increased probability of experiencing more punitive pathways and a decreased probability of experiencing less punitive pathways.

For example, “pathway 14” (the most punitive pathway) will be more likely for Black defendants while “pathway 1” (the least punitive pathway) will be more likely for White defendants. As pathways become more punitive (i.e., from pathway 1 to pathway 14), the differential probability should increase for Black and Hispanic defendants relative to White defendants. That is, it is predicted that relative Black-White and Hispanic-White disparities will steadily increase from negative disparities at pathway 1 to positive disparities at pathway 14.

Methods

Data 

The present study used a combination of juvenile court data and county-level data. 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. The 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.[6] This juvenile court data is restricted. To access the data, a proposal was submitted to and approved by NJCDA. States were selected if they collected data for each of five outcomes to serve as dependent variables: preadjudication detention, petition of delinquency, adjudication of delinquency, judicial disposition, and judicial waiver to adult court. Only four states provided data for each of these five outcomes to the NJCDA, along with case-level information for race/ethnicity, gender, age, offense type, and prior record. The juvenile court data were then merged with a level-2 dataset consisting of county demographic factors created using decennial Census 2010, the American Community Survey (ACS) five-year estimates for 2010, and aggregated official crime statistics from the Uniform Crime Reports (UCR) for 2009–2011.[7] The original dataset consisted of 99,857 referred non-dependency cases with missing information on the outcomes of interest (i.e., pending cases were dropped as missing completely at random). Since the unit of analysis is referrals, some defendants had multiple referrals disposed in 2010. Approximately four percent (n = 4,187) were missing data on one or more independent variables and were dropped from the analysis.[8] The final sample consists of 95,670 cases distributed across 140 counties.

Measures

Dependent variables

The outcome variables in the present study are 14 mutually exclusive pathways through the juvenile justice system, based on outcomes at detention, petition, adjudication, judicial waiver,[9] and judicial disposition. The descriptive statistics for these underlying five outcomes and the rest of the variables included in subsequent models are included in Table I. Preadjudication detention was coded as “1” for detention and “0” for release pending the petition decision. Approximately 16.3 percent of cases (n = 15,564) were detained prior to petition. Formal petition was coded as release, diversion (i.e., informal processing), or petition. The majority of cases were petitioned (57.5 percent; n = 54,987), while 19.69 percent were released (n = 18,837) and 22.83 percent were diverted (n = 21,846).

The next decision point, waiver to adult court, was coded as binary and accounted for just under one percent of all petitioned cases (.8 percent; n = 383). For states that utilize blended sentencing, an adult disposition by a juvenile court judge is treated as waiver to adult court.[10] Next, adjudication of delinquency was coded as a binary variable indicating whether a petitioned (non-waived) case was adjudicated delinquent or dismissed. The majority of these cases were adjudicated delinquent (64.81 percent; n = 35,387), moving on to the judicial disposition stage. Disposition was coded as judicial discharge, community supervision (e.g., probation), or secure out-of-home placement. Most adjudicated delinquents were placed on community supervision (78.8 percent; n = 27,877), while 14.2 percent received secure placement (n = 5,040) and 7.0 percent were discharged (i.e., suspended sentence) by the judge (n = 2,471).

[Table I about here]

Based on the five outcomes described above, 14 dichotomous variables were created to represent possible pathways through the juvenile justice system for the sample of referrals. These mutually exclusive and collectively exhaustive pathways are displayed in Figure 1.

[Fig. 1 about here]

Table II lists the pathways from least to most punitive. As can be seen, some paths were more traveled than others. The least punitive pathway (i.e., path 1, release with no detention) accounted for 18.5 percent of cases (n = 17,687). By contrast, the most punitive pathways, waiver to adult court with preadjudication detention (path 14; n = 143) and without detention (path 13; n = 240), together accounted for .4 percent of cases. In addition to path 1, the most common pathways were as follows: path 3, non-detained cases that were diverted from formal processing (22.5 percent; n = 21,502); path 5, non-detained cases that were formally petitioned but not adjudicated delinquent (17 percent; n = 16,229); and path 9, non-detained cases that were adjudicated delinquent and received community supervision (20.8 percent; n = 19,914). Thus, four of 14 pathways through the juvenile justice system—paths 1, 3, 5, and 9—together accounted for approximately 79.0 percent of cases.

[Table II about here]

Because only 16.3 percent of cases received preadjudication detention (see Table I), pathways including detention (i.e., even-numbered) were less common. Interestingly, most juvenile referrals were not adjudicated delinquent. Instead, they were dismissed, diverted, or found not delinquent (62.6 percent; n = 59,900). These first six pathways represent the less punitive side of the spectrum, suggesting that most juvenile referrals did not experience the more punitive side of the juvenile court. Only 5.6 percent of cases experienced the most punitive pathways (i.e., 11–14) involving secure placement or waiver to criminal court (n = 5,483).

Ranking these outcomes from least to most punitive is imperfect but reflects the general view in prior research that a continuum of severity in processing and sanctioning exists. It assumes that, ceteris paribus, it is worse to be detained (prior to petition) than released, worse to be formally petitioned than diverted to informal processing or released, worse to be diverted to informal processing than dismissed outright, worse to receive secure placement than community supervision, and worse to be waived to criminal court than receive any other juvenile court outcome. Paths that are identical but for detention-versus-release decisions are adjacent, reflecting that these paths are the same in terms of final disposition—but that it is worse to be detained (even if only for a few days) than to be released. For purposes of assessing system-level cumulative disadvantage, a pattern of accelerating disparities as we move from less to more punitive pathways would constitute evidence of cumulative disadvantage.

Independent variables

The main independent variable of interest at the case-level was the race and ethnicity of the defendant, coded as Black (non-Hispanic), Hispanic (non-White), Other, and White (reference category). (Other consisted of Asian/Pacific Islander, American Indian/Native American, and Other.) The final sample of juvenile referrals included approximately 50.2 percent White (n = 48,004), 32.4 percent Black (n = 30,959), 13.7 percent Hispanic (n = 13,096), and 3.8 percent Other (n = 3,611). Defendant age, sex, offense type, prior record, and multiple charges were included as controls. Age at referral was measured as a continuous variable ranging from 7 to 20 years, with a mean age of approximately 15 years (SD = 1.8). Approximately 96 percent of referrals were between 10 and 17 years old. Sex was measured as a binary variable, with a majority of defendants identified as male (68.2 percent; n = 65,196). Offense type was measured as a series of seven dummy variables for each type of offense, based on initial charge: violent (13.5 percent; n = 12,894), property (26.2 percent; n = 25,077), drug/alcohol crimes (10.7 percent; n = 10,233), non-violent weapon offense (1.4 percent; n = 1,358), probation violation (7.6 percent; n = 7,220), status offense[11] (17.1 percent; n = 16,347), and other offenses such as public order crimes (23.6 percent; n = 22,549). Property offense served as the reference category. Prior record was coded as a binary variable indicating whether the juvenile defendant had any prior referrals (60.5 percent; n = 57,904). Lastly, approximately 20.3 percent of referrals involved multiple charges (n = 19,384), which was also coded as a binary variable.

Since cases are nested within 140 counties, several contextual variables were also included that might be associated with juvenile court outcomes: percent non-White, crime rates, concentrated disadvantage, and urbanism. First, racial and ethnic demographic context was measured using percent non-White based on decennial Census 2010. At the county level, percent non-White ranged from 3.3 to 84.3 percent, with a mean value of approximately 30.8 percent (SD = 19.9). Using UCR data, crime rates were measured as an average for total index (i.e., felony) crime rates for the years 2009–2011, to provide an estimate of perceived crime level in 2010 (per 1000 residents). Felony crime rates in 2010 ranged from 5.7 (per 1000) to 78.6 (per 1000), with a mean value of 33.7 (SD = 13.6). Concentrated disadvantage was calculated as a weighted index score based on four indicators from the ACS: percent unemployed, percent households below poverty, percent families using supplemental nutrition assistance program, and percent female-headed households. Principal components analysis suggested a one-factor solution with all factor loadings above .48 and a first eigenvalue of 3.47 (all others below .26); approximately 87 percent of the variance among the items was accounted for by the first factor. Prior to standardization, concentrated disadvantage scores ranged from -3.2 (indicating low disadvantage) to 5.8 (representing high disadvantage), with a mean value of 0 (SD = 1.9). Urbanism was measured as population density based on Census 2010 (recoded as 100 persons per square mile). This ranged from .02 (i.e., 2 persons per square mile) to 14.7 (i.e., 1,470 persons per square mile), with a mean value of approximately 1.6 (SD = 5).

Finally, to control for state-level variation a series of dummy variables were created for the four states. Two states are southern (Alabama, South Carolina), and two are not (Connecticut, Utah), such that approximately half of the juvenile case referrals were from southern states (54.4 percent; n = 53,344). Alabama accounted for the most cases (33.21 percent; n = 32,559), while Connecticut accounted for the fewest (12.51 percent; n = 12,265). Alabama served as the reference category. Correlations between variables were weak (< .4) and tests for multicollinearity (variance inflation factor [VIF]) indicated no serious concerns (VIF < 2).[12]

Analytical Strategy

We first estimated a series of logistic and multinomial regression models for each juvenile court outcome of interest: detention, petition, adjudication, judicial disposition, and waiver. This provided baseline results for outcome-specific disparities before turning to our pathways approach. Following Kutateladze and colleagues (2014: 527), who calculated “predicted probabilities . . . from multivariate regression models predicting membership in different combinations of outcomes,” we employed a multinomial model with 14 categories and cluster-adjusted standard errors.[13] Pathway 5 serves as the reference category because it is a common combination of outcomes (17 percent) and an approximately similar number of referrals experience less punitive pathways (43 percent) and more punitive pathways (40 percent) compared to path 5. Regression coefficients thus reflect the risk of less punitive and more punitive pathways, relative to pathway 5. (The more “middle” categories, 6–8, were too uncommon to serve as a meaningful reference category.)[14] The measure of concentrated disadvantage was standardized at the county-level for interpretation purposes, and all other continuous variables were grand-mean centered. Although some path-outcomes are rare events (<1 percent; see Table II), the analyses were still possible due to the large sample sizes at level 1 (N = 95,670) and level 2 (N = 140).

Predicted probabilities by race and ethnicity were calculated after estimating the multinomial model. These probabilities allow for interpretation of the differences in pathway by race and ethnicity while taking into account the overall probability of each pathway. Using the margins command in Stata 15, we estimated the predicted probability for each pathway separately for Black, Hispanic, White, and Other defendants. Marginal effects were then estimated to assess the differences in probability by race/ethnicity (including testing for significance). Since most of the control variables in the model are dichotomous, we employed marginal standardization rather than conditional probabilities (see Muller and MacLehose 2014).

Results

Outcome-Specific Relative Racial and Ethnic Disparities

Before turning to an examination of disparities in pathways through the juvenile justice system, we first present results from five outcome-specific models. First, among Black referrals, the odds of being detained were 21 percent greater than were the odds among White referrals (OR = 1.21, p < .001), and Hispanic referrals had 28 percent higher odds of being detained than White referrals (OR = 1.28, p < .001). Second, there were no significant racial or ethnic differences in odds of diversion or petition, compared to dismissal. Third, among Black petitioned referrals, the odds of adjudication of delinquency were 11 percent higher than the odds among White petitioned referrals (OR = 1.11, p = .04); there were no significant Hispanic-White differences in adjudication. Fourth, the odds of judicial discharge (versus community supervision) among Black delinquent referrals were 18 percent lower than the odds among White delinquent referrals (OR = .82, p = .02), with no differences in odds of secure placement (versus community supervision). Hispanic delinquent referrals, however, had 21 percent higher odds of placement than White delinquent referrals (OR = 1.21, p < .001), but no differences in odds of judicial discharge. Fifth, no significant differences in odds of waiver to criminal court were observed.

[Table III about here]

In short, racial disparities were observed at preadjudication detention, adjudication of delinquency, and judicial discharge, while ethnic disparities were observed at detention and judicial placement. As noted, however, the above treats each juvenile court outcome as a discrete point with its own sample corresponding to eligibility (e.g., cases that are dismissed or diverted at intake are not eligible for waiver, adjudication, or disposition); put differently, it ignores selection biases that occur from one stage to the next. The following pathways analysis represents a system-level approach that avoids selection bias and provides a more holistic assessment of racial and ethnic disparities in juvenile justice processing.

Relative Disparities in Pathways

We begin by assessing whether the relative disparities increase from path 1 (least punitive) to path 14 (most punitive) for Black and Hispanic referrals compared to White referrals. Table IV reports the findings from a multinomial logistic regression with the 14 pathways as categories (path 5 as reference), controlling for age, gender, type of offense, prior record, multiple charges, and contextual and jurisdictional characteristics. For interpretation purposes, relative risk ratios (RRR) are displayed for Black and Hispanic variables only in Table IV (for full results, see Supplemental Table A). Support for the cumulative disadvantage hypothesis would require risk ratios < 1 for the least punitive outcomes (i.e., lower-numbered paths) and risk ratios > 1 for the more punitive outcomes (i.e., higher-numbered paths). We expect that relative disparities will accelerate from lower- to higher-numbered paths, relative to middle pathway 5. For presentation purposes, in Table IV we have divided the 14 paths into six panels, a through f. In these panels, a includes the least punitive paths (where we expect RRR < 1), f includes the most punitive paths (where we expect RRR > 1), and c includes prototypical “middle paths” (where we do not expect significant disparities to be observed).

[Table IV about here]

Starting with the most punitive end of the spectrum, panel f shows that the largest Black-White disparity was for the most punitive combination of outcomes: detention and waiver to criminal court (i.e., pathway 14). This indicates that, compared to White referrals, Black referrals were approximately twice as likely to experience path 14 than pathway 5 (RRR = 2.02, p = .008). Similarly, this pathway exhibits the greatest Hispanic-White disparity (RRR = 1.98, p = .01). At the same time, there were no significant differences among White, Black, and Hispanic referrals for pathway 13 (versus pathway 5): waiver to adult court with no prior detention.

The next most punitive pathways, as seen in panel e, involved final dispositions with secure placement outside the home. At path 12 we see a notable (but smaller) disparity for Black referrals (RRR = 1.49, p < .001) as well as Hispanic referrals (RRR = 1.48, p < .001), while no statistically significant differences emerged for path 11 compared to path 5. Moving up to panel d, we again see this pattern continue in formal community supervision outcomes, with smaller disparities for Black (RRR = 1.30, p < .001) and Hispanic (RRR = 1.26, p = .001) referrals at path 10. At pathway 9, similar small disparities were also observed for Black (RRR = 1.20, p < .001) and Hispanic referrals (RRR = 1.09, p = .04) compared to White defendants. At the most punitive end of the spectrum (i.e., panels d–f), then, we can observe some evidence of accumulating disparities relative to path 5. The disparities are more pronounced, or evident, for pathways like 12 and 14 that begin with detention.

Less consistent with the cumulative disadvantage hypothesis are the differences between path 5 and other “middle” outcomes. This can be seen by inspection of panel c, which presents pathways 5–8, that is, formal petitions that were ultimately discharged (either prior to or following an adjudication of delinquency). For this class of pathways (which accounts for approximately one quarter of cases), findings indicate no significant racial or ethnic disparities for paths 7 and 8 compared to path 5. However, Black (RRR = 1.37, p < .001) and Hispanic referrals (RRR = 1.33, p < .001) were more likely than White referrals to experience path 6 than path 5, which is broadly consistent with the hypothesis since path 6 involves preadjudication detention while path 5 does not.

Least consistent with the cumulative disadvantage hypothesis are the pathways in panels a and b. Here, we would expect to find Black and Hispanic referrals under-represented—but this is not borne out in the data. Instead, we see that the least punitive outcomes—involving outright release or diversion rather than formal petition—are not significantly less likely for Black and Hispanic youth. First, in panel b, there are no significant racial or ethnic differences in diversion outcomes compared to path 5. Second, Black (RRR = 1.56, p < .001) and Hispanic referrals (RRR = 1.66, p = .02) are actually more likely to receive pathway 2 (than pathway 5), release following detention. Black referrals are also more likely than White referrals to receive punitive pathway 1, release with no detention (RRR = 1.17, p = .047), while Hispanic referrals are not. In other words, minority defendants were more likely to experience some of the least punitive pathways as we have defined them.

In sum, there is some evidence of increasing disparities as cases progress deeper into the juvenile justice system from path 6 to path 14. The increases, however, are not substantial and do not arise in a clear linear manner. Specifically, disparities appear largely driven by whether the path involved preadjudication detention or not (i.e., odd- vs. even-numbered paths). Moreover, the least punitive pathways are not less likely for Black and Hispanic referrals.

Predicted Probabilities by Pathway

We next present the predicted probabilities of each pathway through the juvenile justice system for defendants in different racial and ethnic groups. This was motivated by the need to move beyond relative disparities and consider the practical significance of the disparities experienced by some groups relative to others, that is, the overall impact of these racial and ethnic differences in terms of absolute probability of case combinations.[15] Table V shows the overall probability of experiencing each pathway for White, Black, and Hispanic defendants, as well as the difference between probabilities (i.e., marginal effects) for Black-White and Hispanic-White comparisons. Consistent with Table IV, the marginal effects do not evidence consistently growing disparities from path 1 to path 14.

[Table V about here]

For Black-White differences in predicted probabilities of each pathway, statistically significant differences were observed at pathways 2, 5–7, 10, 12, and 14. The probability for path 2 was 1.32 percent for Black defendants, compared to 1 percent for White defendants—a .32 percent difference (p <.001). The probability for path 5 was 15.61 percent for Black defendants, compared to 17.84 percent for White defendants—a 2.22 percent absolute difference (p <.001). For path 6, the probability was 3.33 percent for Black referrals and 2.84 percent for White referrals—a .49 percent difference (p = .02). For path 7, the probability was 1.88 percent for Black referrals and 2.43 percent for White referrals—a .55 percent difference (p = .006). The probability for path 10 was 8.61 percent for Black referrals and 7.88 percent for White referrals—a .74 percent difference (p = .04). For path 12, the probability was 2.95 percent for Black referrals and 2.38 percent for White referrals—a .57 percent difference (p = .03). Finally, the probability for path 14 was .18 percent for Black referrals and .1 percent for White referrals—a .07 percent difference (p = .02).

Similar patterns emerged for Hispanic-White comparisons, with significant differences observed at pathways 2, 6, 10, 12, and 14. The probability for path 2 was 1.56 percent for Hispanic defendants and 1.0 percent for White defendants—a .56 percent difference (p = .049). The probability for path 6 was 3.49 percent for Hispanic referrals and 2.84 percent for White referrals—a .65 percent difference (p = .001). For path 10, the probability was 8.99 percent for Hispanic referrals and 7.88 percent for White referrals—a 1.11 percent difference (p = .006). For path 12, the probability was 3.12 percent for Hispanic referrals and 2.38 percent for White referrals—a .74 percent difference (p <.001). Finally, the probability for path 14 was .19 percent for Hispanic referrals and .1 percent for White referrals—a .08 percent difference (p = .02).[16]

Finally, using the predicted probabilities in Table V, we can estimate the expected difference in number of Black and Hispanic referrals receiving each path had they been White.[17] At path 14, the most severe combination of outcomes, our findings suggest that 22 more Black defendants and 11 more Hispanic referrals received detention followed by waiver to criminal court than would have been expected if they had been White. Several other pathway probabilities also stand out. For path 12, 177 more Black referrals and 97 more Hispanic referrals received preadjudication detention and secure placement than would have been expected if they had been White. Similarly, 229 more Black referrals and 145 more Hispanic referrals received preadjudication detention and community supervision (path 10). The largest expected difference can be observed at path 5, petition followed by no adjudication of delinquency (with no detention): 687 fewer Black defendants received this combination of outcomes than would have been expected had they been White. This served as our reference category in the analysis of relative disparities, and helps explain why we did not observe disparities that grew in linear fashion for Black referrals: pathway 5, a “middle” path, constitutes the most practically substantive difference between Black and White referrals. On the other hand, Black and Hispanic referrals were more likely to receive no adjudication of delinquency after being detained: 152 more Black referrals and 85 more Hispanic referrals received this combination of outcomes than would have been expected had they been White. Finally, path 2 is also notable: 99 more Black defendants and 72 more Hispanic defendants were detained but then dismissed (before petition) than would have been expected had they been White.

State-Specific Analyses

Finally, because the main analysis involved a large sample of referrals from 140 counties across four heterogeneous states, we also examined predicted probabilities and marginal effects by race/ethnicity across states. This sensitivity analysis was performed to determine whether the same patterns observed in the multi-state data were observed for each state. This is especially important in light of prior research showing that racial and ethnic disparities vary across justice systems (see Bridges and Crutchfield 1988; Blumstein 1993; Crutchfield et al. 1994; Enders et al. 2019; Hawkins and Hardy 1989; Krisberg et al. 1987; Zane, Mears, and Welsh 2020).

While the probabilities of each pathway varied across states, we observed remarkable consistency in the racial and ethnic differences in predicted probabilities (i.e., marginal effects) for each pathway. There were several differences, however. First, Black and Hispanic referrals were significantly more likely to experience pathway 2 in the full model, but this did not hold for Connecticut. Additionally, the positive association between Hispanic referrals and pathway 2 only approached significance for Alabama (p = .11) and Utah (p = .16). Second, while there was no significant association between race/ethnicity and path 5 for Hispanic referrals, Hispanic referrals were significantly less likely to receive path 5 in Utah. (All four states exhibited a negative association between Black referrals and pathway 5.) Third, the positive association between path 6 and race/ethnicity did not hold for South Carolina. Fourth, while Black referrals were more likely to experience pathway 10 in the full model, this did not hold for South Carolina or Utah. Finally, for path 14, the positive association with race/ethnicity did not hold for any individual states—although it did approach significance in each (i.e., p < .15). This is likely due to path 14 being a rare event, such that there was low power to detect effects at the state level.

Discussion

Building upon prior research on racial and ethnic disparities at individual decision points in the juvenile justice system, this study analyzed pathways as a means to assess whether disparities accumulate as referrals travel deeper into the system and encounter multiple decision points. In short, the cumulative disadvantage hypothesis predicts that such accumulation will be observed, evidenced by greater disparities for more punitive combinations of outcomes.

Our main finding is that cumulative disadvantage in the juvenile justice system was not clearly observed. On the one hand, racial and ethnic disparities were greatest for the most punitive pathways; indeed, the most punitive combination of outcomes (i.e., pathway 14) exhibited the greatest relative disparities among all pathways.[18] But less punitive pathways did not tend to exhibit underrepresentation of Black and Hispanic defendants, as the cumulative disadvantage hypothesis would anticipate. Contrary to the hypothesis, pathway 2 was in fact more likely for Black and Hispanic than White defendants. Moreover, there was no clear pattern of increasing disparities from less to more punitive pathways, as a strict reading of the cumulative disadvantage hypothesis predicts.

The picture that emerges is thus not one of cumulative disadvantage; rather, two patterns stand out. First, for minority defendants, as compared to White defendants, detention paths (i.e., even-numbered paths) were significantly more likely for both less punitive paths (e.g., path 2, release) and more punitive paths (e.g., path 14, waiver to criminal court). Second, among non-detained referrals that were petitioned, those involving Black youth were significantly less likely to be released, whether via no adjudication of delinquency (path 5) or via judicial discharge following an adjudication (path 7).

These results have implications for future research and policy, ones that might be missed by relying solely on a traditional stage-specific analysis. First, we can think about racial and ethnic disparities in terms of relative or absolute differences in outcomes. The pathways approach illustrates this point. As noted above, the greatest relative disparity occurred at the most punitive pathway—detention followed by waiver to criminal court. This finding, while consistent with the cumulative disadvantage hypothesis, also involves the least common pathway through the juvenile justice system. In our sample of 95,670 referrals, only 143 defendants (.15 percent) experienced this path. As such, the largest relative disparities only amount to an expected Black-White difference of 22 cases and an expected Hispanic-White difference of 11 cases.

This small number of cases does not mean, of course, that the relative disparities are trivial. Consider racial disparities in capital punishment (e.g., Paternoster and Brame 2008), which are substantively important and receive much attention despite the application of the death penalty being quite rare (see Bedau and Cassell 2004). At the same time, a focus on the worst and rarest outcomes can be misleading if generalized to other less severe and far more common outcomes. It may be that smaller relative disparities observed for less punitive pathways are of much more practical significance given the larger number of juvenile defendants affected. For example, our findings indicate that 687 fewer Black referrals experience path 5: non-detained cases that are petitioned but not adjudicated delinquent. This might indicate that among less serious or first-time offenders who are not initially detained, judges are more forgiving for White youth. This interesting possibility is illustrated by the pathways approach; by contrast, a traditional stage-specific analysis, such as examining detention, petition, and adjudication decisions in isolation (see Table III), would not reveal this pattern. That does not mean that a pathways approach is better. Rather, we submit that employing both approaches provides a more comprehensive understanding of racial disparities—relative and absolute.

A second related implication is that disparities in beneficial outcomes in the juvenile justice system may be as important as disparities in more punitive outcomes. Contrary to the cumulative disadvantage hypothesis—and similar to prior findings (Kutateladze et al. 2014)—we found that Black and Hispanic youth were more likely than White youth to experience the two least punitive pathways, both involving case dismissal prior to petition. (Again, this was especially pronounced for detained cases.) One explanation is that there may be a “correction effect” by intake officers cognizant of the possibility that minority youth are more likely to be arrested than similarly situated White youth (Rodriguez 2010). Another possibility is that due to bias at arrest, evidence against minority defendants tends to be weaker. There is, too, the fact that informal processing often arises out of plea bargaining. It is possible that minority youth are less likely to admit guilt in exchange for informal treatment, which could lead both to more formal petitions and to more dismissals. (In the criminal justice system, research consistently finds that Black defendants are less likely to plead guilty; see Johnson 2019.) Another possibility is that minority youth, and especially Black youth, are seen as less deserving of the treatment provided by juvenile court, specifically by diversion to informal processing, so they are released when formal processing is not warranted (see, generally, Cochran and Mears 2015). This raises the possibility that paths 1 and 2 do not reflect the least punitive pathways at all: rather, they might be seen as unnecessary contact with the juvenile justice system. That possibility can best be seen in pathway 2, where minority youth are more likely to be detained—only later to be released.

A third implication involves the role of detention. The pathways approach illustrates how detention functions in the overall processing of juvenile defendants through the justice system. Part of the reason why the pattern of cumulative disadvantage was not clearly observed in the present study is that detention (i.e., even-numbered) paths tended to exhibit greater disparities than non-detention (i.e., odd-numbered) paths. In general, Black and Hispanic defendants were more likely to experience pathways that included detention, from less punitive outcomes like release (path 2) to more punitive outcomes like community supervision (path 10), secure placement (path 12), and waiver (path 14). When we look only at pathways without detention (i.e., odd-numbered), however, racial and ethnic disparities largely disappear.[19] Sutton (2013) found the same pattern in the criminal justice system. In short, there appear to be two pathways through the justice system, those that begin with detention and those that do not. Among those that do not, disparities are not as apparent.[20]

This last point suggests support for cumulative disadvantage in another sense: early disadvantage may lead to later disadvantage (Sacks and Ackerman 2014). Prior work consistently shows that pretrial detention is associated with harsher sentencing outcomes, and that race effects are mediated by detention (Rodriguez 2007; Spohn 2009, 2013; Wooldredge et al. 2015).[21] Under this conceptualization, cumulative disadvantage for individual defendants begins at referral, where minority youth (especially Black youth) are more likely to enter the system in the first place. Once in the system, minority defendants are more likely to be detained prior to adjudication, and this detention is strongly associated with later punitive outcomes—possibly by signaling dangerousness or culpability. Alternatively, it is possible that detention is associated with more severe dispositions for legitimate reasons rather than motivated by bias. For one, there could be other factors, such as gang involvement, not captured by control variables such as prior record and offense type. Detention may also reflect an assessment that a juvenile has a less promising prospect of returning safely to the community (Clair and Winter 2016). Given that disparities were more pronounced for detention paths than non-detention paths, one possible implication is that racial disparities are largely due to uncaptured differences rather than racial bias. In either case—racial bias or unmeasured case characteristics—the strong association between detention and other negative outcomes points to the importance of this early processing outcome for assessing cumulative disadvantage in the juvenile justice system.

Limitations

While the present study’s approach—taking a more system-level view of disparities by examining pathways through the juvenile justice system—illustrates many of the important observations above, there are several limitations. First, this study utilizes a large, multi-jurisdictional sample both to increase the external validity of the study and to obtain sufficient statistical power to estimate relative odds and predicted probabilities for every possible pathway through the juvenile justice system, some of which are quite uncommon (e.g., paths 4, 8, 13–14). However, a multi-jurisdictional sample raises the possible concern that different state justice systems are not comparable and that effects across systems may even cancel each other out (Myers and Talarico 1987). The present study found that racial and ethnic disparities in juvenile justice outcomes were significant but not large, perhaps lending some credence to this concern. There are tradeoffs here between precision and generalizability. Research on a single jurisdiction may be better positioned to interpret its particular findings—but less well positioned to form conclusions about juvenile justice more generally. Unfortunately, limiting focus to single jurisdictions means never moving beyond 51 different, possibly conflicting, conclusions about justice system disparities (Britt 2000). These concerns may be offset by the fact that the results are largely concordant with those from prior work: compared to White defendants, Black and Hispanic defendants were more likely to have their cases dismissed, more likely to be detained, and more likely to receive secure placement (see Bishop and Leiber 2012). Moreover, sensitivity analyses revealed the same basic patterns emerged within each state.

Additionally, while the present research sought to take a holistic view of juvenile justice system processing and the racial and ethnic disparities therein, it did not take into account how juveniles were referred to the system. Police contact represents the first stage of juvenile justice processing, but police decisions are not included in most administrative datasets (Piquero, 2008). Clearly, though, arrests represent the main contribution to racial and ethnic disparities in juvenile justice. National RRIs indicate that referral rates in 2019 (latest available) were 2.4 times greater for Black youth than White youth (OJJDP, 2021). Similarly, in the present study, Black and Hispanic youth made up 32.4 and 13.7 percent of total referrals, respectively, compared to 19.5 and 8.0 percent of the youth population (age 10–17) in the same states in 2010 (see Puzzanchera et al. 2020). Arrest differentials are thus major drivers of system disparities, but the causes of racial differences in arrests are difficult to study due to unique methodological challenges (see Neil and Winship 2019). More research is needed on this key, but under-studied, decision point (Dillard, 2013; Piquero, 2008).[22]

Another limitation is that the regression models did not include every relevant case-level predictor. While the most important controls were included (i.e., type of offense, prior record, age), other relevant case-level variables such as crime severity were not available. As others have observed, some degree of omitted variable bias is inevitable in research on criminal and juvenile justice processing that relies on administrative data (Baumer 2013). In the present case, this means that racial and ethnic disparities cannot be interpreted as necessarily reflecting racial bias. Relatedly, although our attempt to measure system-level cumulative disadvantage seeks to improve upon prior attempts by investigating all possible pathways through the juvenile justice system, it is nonetheless possible that alternative conceptualizations and measurements might lead to different conclusions (see, generally, Kurlychek and Johnson 2019).

Conclusion

Our system-level analysis of pathways through the juvenile justice system shows that minority youth are more likely to experience most combinations of outcomes than White youth (including dismissal), despite minimal evidence of accumulating disparities. It may be that the overall disproportionate number of minority defendants in the juvenile justice system is largely a function of differential referrals rather than cumulative disadvantage within the system. Indeed, the latest national statistics appear to bear this out: the greatest disparity is at arrest, where the rate of Black youth referrals (relative to population) is more than twice the rate of White youth (OJJDP, 2021). No other system disparity even approaches this order of magnitude (see Puzzanchera and Hockenberry 2018). This could be due to differences in offending, differences in treatment of similarly situated youth by police and schools, disparate impact of racially neutral laws or policies, or some combination of these factors (see Zane et al. 2016).

Cumulative disadvantage remains a central concern for policymakers and practitioners who seek to create a more equitable juvenile or criminal justice system. Addressing it requires better information about the nature of disadvantage and its causes. Bishop and Leiber (2012: 475) have written: “To be sure, there are some improvements that can be made within the system, but these are ancillary to the social and economic restructuring that will be required to produce lasting change” (see also Mears, Cochran, and Lindsey 2016). In the present context, it may be that the most important kind of cumulative disadvantage goes beyond the justice system. It may entail social disadvantages that contribute to juvenile and criminal justice system involvement. These adverse life events can lead to a greater likelihood of further criminal justice system involvement, which in turn further contributes to negative life outcomes, producing a vicious cycle of “cascading effects” (Kurlychek and Johnson 2019: 292). In the broadest sense, cumulative disadvantage thus can be both a cause and a consequence of disparities in the criminal justice system. To understand and address inequality in the juvenile justice system therefore may require not only a focus on disparities that exist within the system, but also a focus on the accumulation of negative life events that occur outside the system and that likely contribute to the over-representation of racial and ethnic minorities in it.

References

Barton, W. H. (2012). Detention. In Feld, B. C., and Bishop, D. M. (eds.), The Oxford handbook of juvenile crime and juvenile justice. Oxford University Press, New York, pp. 636–663.

Baumer, E. P. (2013). Reassessing and redirecting research on race and sentencing. Justice Q 30(2):231–261.

Beck, A. J., and 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.

Bedau, H. A. and. Cassell, P. G. (2004). Debating the death penalty: Should America have capital punishment? The experts on both sides make their case. Oxford, U.K.: Oxford University Press.

Bishop, D. M. and Leiber, M. J. (2012). Racial and ethnic differences in delinquency and justice system responses. In Feld, B. C., and Bishop, D. M. (eds.), The Oxford handbook of juvenile crime and juvenile justice. Oxford University Press, Oxford, pp. 445–484.

Bishop, D. M., Leiber, M. J., and 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(3):213–233.

Blumstein, A. (1982). Of the racial disproportionality of United States’ prison populations. Journal of Criminal Law and Criminology 73:1259–1281.

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

Brame, R., Turner, M. G., and Paternoster, R. (2010). Missing data problems in criminological research. In Piquero, A. R. and Weisburd, D. (eds.), Handbook of quantitative criminology. Springer, New York, pp. 273–288.

Bridges, G. S., and Crutchfield, R. D. (1988). Law, social standing and racial disparities in imprisonment. Social Forces 66(3):699–724.

Britt, C. L. (2000). Social context and racial disparities in punishment decisions. Justice Q 17(4):707–732.

Bushway, S., Johnson, B. D., and 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.

Cauffman, E., Piquero, A. R., Kimonis, E., Steinberg, L., Chassin, L., and Fagan, J. (2007). Legal, individual, and environmental predictors of court disposition in a sample of serious adolescent offenders. Law and Human Behavior 31(6):519–535.

Clair, M., and Winter, A. S. (2016). How judges think about racial disparities: Situational decision-making in the criminal justice system. Criminology 54(2):332–359.

Cochran, J. C., and Mears, D. P. (2015). Race, ethnic, and gender divides in juvenile court sanctioning and rehabilitative intervention. Journal of Research in Crime and Delinquency 52(2):181–212.

Crutchfield, R. D., Bridges, G. S., and 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(2): 166–182.

Crutchfield, R. D., Fernandes, A., and Martinez, M. (2010). Racial and ethnic disparity and criminal justice: How much is too much? Journal of Criminal Law and Criminology 100(3): 903–932.

Dillard, D. (2013). Limited disproportionate minority contact discourse may explain limited progress in reducing minority over-representation in the US juvenile justice system. Youth Justice 13(3):207–217.

DiPrete, T. A., and Eirich, G. M. (2006). Cumulative advantage as a mechanism for inequality: a review of theoretical and empirical developments. Annual Review of Sociology 32:271–297.

Elwert, F., and Winship, C. (2014). Endogenous selection bias: the problem of conditioning on a collider variable. Ann Rev of Sociology 40:31–53.

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

Fairchild, A. J., Gupta-Kagan, J., and Andersen, T. S. (2019). Operationalizing intake: Variations in juvenile court intake procedures and their implications. Children and Youth Services Review 102:91–101.

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

Franklin, T. W., and Henry, T. K. S. (2020). Racial disparities in federal sentencing outcomes: Clarifying the role of criminal history. Crime & Delinquency 66(1):3–32.

Hagan, J. (1974). Extra-legal attributes and criminal sentencing: an assessment of a sociological viewpoint. Law & Society Review 8:357–384.

Hawkins, D. F., and Hardy, K. A. (1989). Black-White imprisonment rates: a state-by-state analysis. Social Justice 16(4):75–94.

Holmes, B., and Feldmeyer, B. (2019). Reassessing the influence of criminal history in federal criminal courts. Justice Quarterly 36(7):1206–1228.

Johnson, B. D. (2015). Examining the ‘life course’ of criminal cases: a new frontier in sentencing research.” Criminology & Public Policy 14:183-186.

Johnson, B. D. (2019). Trials and tribulations: The trial tax and the process of punishment. Crime and Justice 48(1):313–363.

Johnson, B. D. and Kurlychek, M. C. (2012). Transferred juveniles in the era of sentencing guidelines: Examining judicial departures for juvenile offenders in adult criminal court. Criminology 50:525–564.

Kramer, L., and Wang, X. (2019). Assessing cumulative disadvantage against minority female defendants Justice Quarterly 36(7):1284–1313.

Krisberg, B., Schwartz, I., Fishman, G., Eisikovits, Z., Guttman, E., and Joe, K. (1987). The incarceration of minority youth. Crime & Delinquency 33(2):173–205.

Kurlychek, M. C. and Johnson, B. D. (2010). Juvenility and punishment: Sentencing juveniles in adult criminal court. Criminology 48:725–758.

Kurlychek, M. C. and Johnson, B. D. (2019). Cumulative disadvantage in the American criminal justice system. Ann Rev Criminol 2:291–319.

Kutateladze, B. L., Andiloro, N. R., Johnson, B. D., and Spohn, C. C. (2014). Cumulative disadvantage: Examining racial and ethnic disparity in prosecution and sentencing. Criminology 52(3):514–551.

Leiber, M. J. (2013). Race, pre- and postdetention, and juvenile justice decision making. Crime & Delinquency 59(3):396–418.

Leiber, M. J. (2016). Race, prior offending, and juvenile court outcomes. Journal of Crime and Justice 39(1):88–106.

Leiber, M. J., and Fox, K. C. (2005). Race and the impact of detention on juvenile justice decision making. Crime & Delinquency 51(4):470–497.

Leiber, M. J., and Peck, J. H. (2015). Race, gender, crime severity, and decision making in the juvenile justice system. Crime & Delinquency 61(6):771–797.

McGuire, M. D. (2002). Cumulative disadvantage as an explanation for observed disproportionality within the juvenile justice system: an empirical test. Juvenile and Family Court Journal 53(1):1–17.

Mears, D. P. (2012). The front-end of the juvenile court: intake and informal versus formal processing. In Feld, B. C., and Bishop, D. M. (eds.), The Oxford handbook of juvenile crime and juvenile justice. Oxford University Press, Oxford, pp. 573–605.

Mears, D. P., Cochran, J. C., and Lindsey, A. M. (2016). Offending and racial and ethnic disparities in criminal justice: a conceptual frameowork for guiding research and informing policy. Journal of Contemporary Criminal Justice 32(1):78–103.

Mood, C. (2010). Logistic regression: Why we cannot do what we think we can do, and what we can do about it. European Sociological Review 26(1):67–82.

Muller, C. J., and MacLehose, R. F. (2014). Estimating predicted probabilities from logistic regression: Different methods correspond to different target populations. International Journal of Epidemiology 43(3):962–97.

Myers, M. A. and Talarico, S. M. (1987). The social contexts of criminal sentencing. New York: Springer-Verlag.

Neil, R., & Winship, C. (2019). Methodological challenges and opportunities in testing for racial discrimination in policing. Annual Review of Criminology, 2, 73–98.

Norton, E. C., and Dowd, B. E. (2018). Log odds and the interpretation of logit models. Health Services Research 53(2):859–878.

Office of Juvenile Justice and Delinquency Prevention (2021, February 9). OJJDP statistical briefing book. https://www.ojjdp.gov/ojstatbb/special_topics/qa11501.asp?qaDate=2019.

Paternoster, R., and Brame, R. (2008). Reassessing race disparities in Maryland capital cases. Criminology 46(4):971–1008.

Piquero, A. R. (2008). Disproportionate minority contact. The Future of Children 18(2):59–79.

Puzzanchera, C., and Hockenberry, S. (2018). National disproportionate minority

contact databook. National Center for Juvenile Justice for the Office of Juvenile Justice and Delinquency Prevention. Archived at: https://web.archive.org/web/20180907071657/https://www.ojjdp.gov/ojstatbb/dmcdb/asp/display.asp?display_in=1.

Puzzanchera, C., Sladky, A. & Kang, W. (2020). Easy access to juvenile populations: 1990-2019. 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 Q 24(4):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(3):391–413.

Rodriguez, N. (2013). Concentrated disadvantage and the incarceration of youth: examining how context affects juvenile justice. J Res Crime Delinq 50(2):189–215.

Rodriguez, N., Smith, H., and 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(1):177–208.

Sacks, M., and Ackerman, A. R. (2014). Bail and sentencing: Does pretrial detention lead to harsher punishment? Crim Justice Policy Rev 25(1):59–77.

Sampson, R. J., and Laub, J. H. (1993). Structural variations in juvenile court processing: inequality, the underclass, and social control. Law & Society Rev 27(2)285–312.

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

Spielman, S. E., Folch, D., and Nagle, N. (2014). Patterns and causes of uncertainty in the American Community Survey. Applied Geography 46:147–157.

Spohn, C. (2009). Race, sex, and pretrial detention in federal court: Indirect effects and cumulative disadvantage. University of Kansas Law Review 57:879–901.

Spohn, C. (2013). The effects of the offender’s race, ethnicity, and sex on federal sentencing outcomes in the guidelines era. Law and Contemporary Problems 76(1):75–104.

Spohn, C. (2015). Race, crime, and punishment in the twentieth and twenty-first centuries. Crime and Justice 44(1):49–97.

Spohn, C., and Belenko, S. (2013). Do the drugs, do the time? The effect of drug abuse on sentences imposed on drug offenders in three U.S. district courts. Criminal Justice and Behavior 40(6):646–670.

Steiner, B. (2009). The effect of juvenile transfer to criminal court on incarceration decisions. Justice Q 26:77–106.

Stolzenberg, R. M., and Relles, D. A. (1997). Tools for intuition about sample selection bias and its correction. American Sociological Review 62(3):494–507.

Sutton, J. R. (2013). Structural bias in the sentencing of felony defendants. Social Science Research 42(5):1207–1221.

Ward, J. T., Hartley, R. D., and Tillyer, R. (2016). Unpacking gender and racial/ethnic biases in the federal sentencing of drug offenders: a causal mediation approach. Journal of Criminal Justice 46:196–206.

Wooldredge, J., Frank, J., Goulette, N., & Travis, L. (2015). Is the impact of cumulative disadvantage on sentencing greater for black defendants? Criminology & Public Policy 14(2):187–223.

Zane, S. N. (2017). Do criminal court outcomes vary by juvenile transfer mechanism? A multi-jurisdictional, multilevel analysis. Justice Quarterly 34(3):542–569.

Zane, S. N., Mears, D. P., and Welsh, B. C. (2020). How universal is disproportionate minority contact? An examination of racial and ethnic disparities in juvenile justice processing across four states. Justice Quarterly 37(5):817–841.

Zane, S. N., Welsh, B. C., & Drakulich, K. M. (2016). Assessing the impact of race on the juvenile waiver decision: A systematic review and meta-analysis. Journal of Criminal Justice 46:106–117.

Zimring, F. E. (1981). Notes toward a jurisprudence of waiver. In J. C. Hall, D. M. Hamparian, J. M. Pettibone, and J. L. White (eds.), Major issues in juvenile justice information and training: Readings in public policy, Academy for Contemporary Problems, Columbus, Ohio, pp. 193–205.

Figure & Tables

Table I. Descriptive statistics

^a Reduced sample (N = 47,765)

^b Reduced sample (N = 54,604)

^c Reduced sample (N = 35,388)

Table II. Combinations of juvenile court outcomes (14 pathways, least to most punitive)

Notes: – (en dash) indicates not applicable. Preadjudication detention, 0 = released, 1 = detained in secure facility; petition of delinquency, 0 = released, 1 = informal processing (i.e., diversion), 2 = formal petition; adjudication of delinquency, 0 = not delinquent, 1 = delinquent; judicial disposition, 0 = discharge, 1 = community supervision, 2 = secure placement; waiver to adult court, 0 = not waived, 1 = waiver

Table III. Outcome-specific logistic regression models (with cluster-adjusted standard errors)

^a N₁ = 95,670; N₂ = 140

^b N₁ = 54,604; N₂ = 140

^c N₁ = 35,388; N₂ = 140

^d N₁ = 47,765; N₂ = 140

^e Reference category: Dismissal

^f Reference category: Community supervision

^g Reference category: White

^h Reference category: Property offense

ⁱ Reference category: Alabama

^j Variable omitted from model due to collinearity with outcome variable

^* <.05 ^** <.01 ^*** <.001

Table IV. Relative racial/ethnic disparities by pathway

 Note: Included relative risk ratios are based on multinomial logistic regression results (see Supplemental Table A for full results)

^a Reference category: White

^* <.05 ^** <.01 ^*** <.001

Table V. Predicted probabilities and marginal effects by pathway and racial/ethnic group  

^a Predicted probabilities, marginal effects, and standard errors are multiplied by 100 for interpretation

^* <.05 ^** <.01 ^*** <.001

Supplemental Table A. Multinomial logistic regression findings, with cluster-adjusted standard errors (14 pathways)

Note: Level-1 N = 98,030; Level-2 N = 140. Reference category: Path 5.

^a Reference category: White

^b Reference category: Property offense

^c Reference category: Alabama

^* <.05 ^** <.01 ^*** <.001

[1] As Crutchfield and colleagues (2010: 929) observe, “For example, if whites are differentially sorted out prior to sentencing, and then a sentencing study reports no racial difference, we cannot know if there really is no difference between comparable cases, or if minority defendants, of varying types, are being sentenced similarly compared to only the worst of white defendants” (see also Crutchfield et al. 1994).

[2] Stolzenberg and Relles (1997: 504) observe that there is “no automatic way to diagnose and correct for sample selection bias” because “methods cannot make imperfect data perfect.” In other words, statistical methods cannot address what is fundamentally a data problem (i.e., non-random reduction in sample size across stages of processing). Instead, examining all possible outcome combinations addresses selection bias by utilizing the full data sample to assess a system-level (rather than outcome-specific) research question.

[3] Similarly, prior contact with the system may be viewed as a form of individual cumulative disadvantage, where minority defendants are more likely to have prior records, and prior records are associated with more punitive outcomes (see, for example, Franklin and Henry 2020; Holmes and Feldmeyer 2019; Leiber 2016).

[4] To date, the most common approach to estimating cumulative racial disadvantage in criminal justice has involved assessing the indirect effects of race via pretrial detention (e.g., Spohn 2009, 2013). As Wooldredge and colleagues point out (2015: 109), many studies have “implied cumulative disadvantage from significant effects of race on pretrial detention and, in turn, between pretrial detention and sentencing” (e.g., Spohn 2009, 2013; Spohn and Belenko 2013), but few have used mediation analysis to unpack the direct and indirect effects of race. Those that have employed mediation analysis have found indirect effects to be greater than direct effects (see Kramer and Wang 2019; Wooldredge et al. 2015).

[5] After finding limited evidence of cumulative disadvantage in criminal justice processing, Sutton (2013: 1219) speculated that, “It may be, as Sampson and Laub (1993b) suggest, that effects of racial oppression operate only in the juvenile justice context, where procedures are typically more informal than in the criminal courts.”

[6] 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.

[7] The ACS provides more detailed information than long-form Census questionnaire, but does so for a smaller sample. Each estimate is accompanied by a margin of error (90 % confidence interval) that was transformed into a coefficient of variation (CV) to determine whether a county should be dropped. The CV is the ratio of the standard error to the value of the estimate (Spielman, Folch, and Nagle 2014). The choice was made to drop counties with low-quality estimates (CV > .4), resulting in 454 observations from six counties dropped from the final sample.

[8] Two independent variables were largely responsible for missing data: race/ethnicity (n = 1,918) and age (n = 1,515). Missing cases were not imputed in the present study since a large portion involved the main variable of interest (i.e., race/ethnicity) and it could not be assumed that data were “missing at random”—and thus could be ignored (see Brame, Turner, and Paternoster 2010). Listwise deletion of cases with missing data “is the method that is the most robust to violation of the MAR assumption among the independent variables in a regression analysis” (Brame et al. 2010: 283).

[9] Not all referrals were waiver-eligible, which varied by state. However, the present analysis is focused on presenting all possible pathways through the juvenile justice system, even though some pathways (i.e., those involving waiver) may be statutorily eliminated for some referrals. In order to present this holistic picture of juvenile justice processing, the population of interest is all referrals across four states.

[10] Juvenile cases waived by prosecutor (i.e., direct-file) or legislation (i.e., statutory exclusion) are not included in the data, since those cases circumvent the juvenile justice system entirely. Importantly, this accounts for the majority of juvenile defendants in criminal court today (see Feld 2017; Zane 2017). 

[11] Status offenses are included in the analysis since these cases are still processed by the juvenile court and can receive most of the possible pathways, although certain pathways are quite rare for status offenders (e.g., detention, placement). Status offenders are not waiver-eligible (i.e., no criminal conduct), so no status offenders received pathways 13 or 14. When we examined individual states, we also found that no status offenders received pathway 4 in Connecticut or pathway 8 in Alabama. We examined pathways for non-status offenders in sensitivity analysis (see footnote 16).

[12] Two contextual control variables stood out as strongly correlated: percent non-white and concentrated disadvantage (r = .76). The VIFs associated with these variables were 3.5 and 3.6, respectively.

[13] Two-level multinomial logistic regression models with random effects could not be estimated due to the number of categories (i.e., 14) for which random effects (either shared or separate but correlated) could not be computed without error. We thus employed cluster-adjusted standard errors to account for the nested nature of the data.

[14] Another option would be an ordered logistic regression model, as employed by Sutton (2013). However, ordered logistic regression makes the “proportional odds” or “parallel regressions” assumption, that the slopes between each pair of outcomes are identical. Pathways through juvenile court do not intuitively satisfy this assumption, and a formal test of the assumption (i.e., the Brant test) revealed that it was not valid.

[15] One interpretative difficulty with multinomial logistic regression is that effect sizes (i.e., RRR) are relative to some reference category, in our case pathway 5. Predicted probabilities and marginal effects do not suffer from this problem, and are identical regardless of the reference category. For this reason, some authors have recommended moving beyond reporting of odds ratios and risk ratios and instead report marginal effects (see, for example, Mood, 2010; Norton and Dowd, 2018).

[16] These findings were quite robust to model specification. We ran several sensitivity analyses, including looking at the interaction of prior record and race as well as looking at a sub-sample with only delinquent referrals (i.e., dropping status offenders). For prior record, findings for Black-White and Hispanic-White differences were identical in statistical significance and direction for all pathways with only one exception: among referrals with a prior record, Hispanic defendants had a significantly lower probability of path 5 than White defendants. For the delinquency-only sub-sample (N₁ = 79,323, N₂ = 140), several differences from the full model emerged. For Black-White comparisons, no significant differences were observed at path 7, while differences only approached significance for paths 10 (p = .09) and 12 (p = .06). For Hispanic-White comparisons, path 2 only approached significance (p = .08), while a significantly greater probability of path 3 for Hispanic defendants was observed. These deviations from the main findings do not alter the overall conclusion that cumulative disadvantage, conceived as accumulating disparities from less to more punitive pathways, was not observed.

[17] This can be accomplished by multiplying the number of Black defendants in the sample (n = 30,959) by the marginal difference. For example, for path 2 this would be calculated as follows: 30,959´ .0032 = 99. The same calculation can be performed for Hispanic defendants (n = 13,096) using the Hispanic-White marginal difference.

[18] We performed an additional sensitivity analysis to examine what variables predicted full case processing from referral to disposition. We found that Hispanic defendants were more likely to receive a judicial disposition than White defendants (OR = 1.11, p < .01), but observed no Black-White differences. (In addition, males were more likely to reach judicial disposition, as were defendants with a prior referral or multiple charges.)

[19] We performed sensitivity analyses that repeated the pathways approach without including the detention decision, producing 7 paths instead of 14. We found that Black-White disparities were only observed for petitions that were not adjudicated delinquent or discharged, while Hispanic-White disparities were only observed for referrals that received residential placement. This indicates that most disparities disappeared when detention was not included in the definition of pathways.

[20] In terms of absolute versus relative disparities, it bears repeating that most referrals do not receive detention—and thus travel through justice system pathways where disparities were not observed.

[21] In the present study, outcome-specific models (see Table III) indicated that detained referrals were significantly more likely to be petitioned (OR = 1.33), less likely to be diverted (OR = .22), more likely to be adjudicated (OR = 1.92), more likely to be securely placed (OR = 2.14), and more likely to be waived to criminal court (OR = 1.91). The impact of detention on subsequent processing is thus greater than the impact of race/ethnicity (see Table III).

[22] One indirect way of assessing police decisions is to examine status offenders. To that end, we ran several additional sensitivity analyses. First, there were significant differences in the proportion of juveniles charged with status offenses among White (20.2 percent) versus Black (14.3 percent) and Hispanic referrals (13.3 percent). Since detention was the primary driver of disparities in the present study, we then looked at whether race and detention were associated for status offenders only. We found that, among status offenders (N₁ 16,347; N₂ = 135), Black referrals were still more likely to be detained than White referrals (OR = 1.53, p <.01), but there were no significant differences in odds of detention for Hispanic and White referrals. Among non-status (i.e., delinquent) referrals (N₁ 79,323; N₂ = 140), Black (OR = 1.22, p < .001) and Hispanic referrals (OR = 1.32, p < .001) were more likely to be detained than White referrals—as in the full model. This would seem to indicate that differences in status offending did not drive Black-White disparities, but may be related to Hispanic-White disparities.