Disproportionate Minority Contact in the Juvenile Justice System: A Systematic Review and Meta-Analysis

Abstract

This paper reports on a systematic review and meta-analysis of disproportionate minority contact in the juvenile justice system. A total of 79 independent samples were analyzed from the 67 studies that met the criteria for inclusion in the meta-analyses. Meta-analytic techniques were used to examine the overall effects of race/ethnicity at the six major decision points in the juvenile justice system: detention, intake, petition, waiver, adjudication, and placement. Moderator analyses were also performed to explore heterogeneity of effects by study design features. Findings indicated small average effects for some outcomes (e.g., detention) and no discernible average effect on others (e.g., petition, waiver, adjudication). There was also substantial heterogeneity across studies, although moderator analyses did not detect any clear patterns. The findings suggest that future research must focus on the specific outcomes in which disparities occur, as well as their practical significance given small effect sizes.

Keywords:

• Racial disparities
• ethnic disparities
• disproportionate minority contact
• juvenile justice
• meta-analysis

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 The terminology of “direct effects” follows the convention to refer to statistical associations between independent and dependent variables in terms of effect sizes. This does not indicate the precise mechanism by which the association emerges, or whether it is causal in nature.

2 Specifically, we performed the following abstract search: (juvenile court OR juvenile justice OR delinquen*) AND (arrest OR referral OR intake OR petition OR diversion OR adjudication OR disposition OR placement OR waiver OR transfer OR disparit* OR discriminat* OR disproportionat* OR race OR black OR Latin* OR Hispanic OR minority).

3 These estimates are similar, though not identical, to other interpretations of odds ratio effect sizes. In an early observation, Rosenthal (1996) suggests that OR = 1.5, 2.4, and 4 be used as indicators of small, medium, and large effects (or weak, moderate, and strong associations). Sullivan and Feinn (2012) propose OR = 1.5, 2.0, and 3.0 as thresholds for designating small, medium, and large effects. Ferguson (2016) argues that OR = 2.0 be adopted as a “recommended minimum effect size” for practical clinical significance, with OR = 3.0 and 4.0 as “moderate” and “strong” effects (p. 302). Elsewhere, Chen and colleagues (2010) recommend OR = 1.68, 3.47, and 6.17 as small, medium, and large effects when the underlying outcome rate is 1%. (When the outcome rate is 10%, the threshold effect for “small” is reduced to OR = 1.46.)

4 For example, we identified 32 studies that met inclusion criteria that were authored by Professor Michael Leiber (and colleagues). These studies were based on 8 independent samples, so only 8 studies with non-duplicate samples were included in the final list (according to the explicit criteria mentioned in the text).

5 For sample size, we separated studies into two groups: above and below the median sample size for studies included in each outcome-specific analysis. For time period, if a study used a sample that included multiple years that crossed decadal boundaries, we placed it in the category with the most overlap. Only three studies did not permit this categorization because their samples included equal number of years in multiple decades; these studies were not included in their respective moderator analyses for time period.

6 There were multiple effect sizes per study due to some studies having multiple samples, multiple outcomes, and/or multiple racial/ethnic comparisons. As noted in the text, there were 79 independent samples across the 67 included studies.

7 One study operationalized race/ethnicity as Black/non-Black, where non-Black included White, Hispanic, Asian, and American Indian (Sridharan et al., 2004). Because 97% of cases were either Black or White, however, this was coded as a Black/White operationalization.

8 To assess whether outlier effect sizes influenced the summary effects, separate meta-analyses were conducted with each individual study removed. In no instance did the direction or significance of the summary effect change, indicating no substantial outliers.

9 For detention, 95% prediction intervals were as follows: Black/White [0.99, 2.25]; Hispanic/White [1.02, 1.80]; non-White/White [0.74, 3.03].

10 Separate meta-analyses were also conducted with each individual study removed. The direction and/or statistical significance of the summary effect only changed for the non-White/White comparison where there were only three effect sizes.

11 For intake, 95% prediction intervals were as follows: Black/White [0.73, 1.94]; Hispanic/White [0.81, 1.43]; non-White/White [0.94, 1.92].

12 Separate meta-analyses were also conducted with each individual study removed. The direction and/or statistical significance of the summary effect only changed for the non-White/White comparison where there were only 3 effect sizes.

13 For petition, 95% prediction intervals were as follows: Black/White [0.70, 1.70]; Hispanic/White [0.88, 1.93]; non-White/White [0.21, 5.05].

14 Separate meta-analyses were also conducted with each individual study removed. The summary effect for the Black/White and Hispanic/White comparisons became significant with the removal of Cheesman et al. (2002), indicating that it represented an outlier.

15 For waiver, 95% prediction intervals were as follows: Black/White [0.21, 7.68]; Hispanic/White [0.24, 12.88]; non-White/White [0.49, 4.64].

16 Separate meta-analyses were also conducted with each individual study removed. The statistical significance of the summary effect only changed for the non-White/White comparison where there were only three effect sizes.

17 For adjudication, the 95% prediction intervals were as follows: Black/White [0.58, 1.58]; Hispanic/White [1.03, 1.19]; non-White/White [0.42, 1.92].

18 Separate meta-analyses were also conducted with each individual study removed. In no instance did the direction or significance of the summary effect change, indicating no substantial outliers.

19 For placement, the 95% prediction intervals were as follows: Black/White [0.81, 1.97]; Hispanic/White [0.93, 1.37]; non-White/White [0.88, 1.63].

20 Rosenthal (1996) suggests that for percentage differences, qualitative assessment of effect size is roughly as follows: 7% for a small effect, 18% for a medium effect, and 30% for a large effect.

21 For other outcomes, these estimated percentage differences are smaller. For petition, the national base rate in 2018 for White referrals was 51.7%. For the Hispanic/White average effect (OR = 1.14), this translates to an expected petition rate of 55%—a 3.3% difference. For adjudication, the national base rate in 2018 for White petitioned cases was 52.4%. For the Hispanic/White average effect (OR = 1.11), this translates to an expected adjudication rate of 55%—a 2.6% difference. Finally, for placement, the national base rate in 2018 for White delinquent cases was 23.3%. For the Black/White average effect (OR = 1.26), this translates to an expected placement rate of 27.7%—a 4.4% difference. For the Hispanic/White average effect (OR = 1.13), this translates to an expected placement rate of 25.6%—a 2.3% difference.