Looking Inside the Black Box of Drug Courts: A Meta‐Analytic Review

Abstract

There has been a rapid proliferation of drug courts over the past two decades. Empirical research examining the effectiveness of the model has generally demonstrated reduced rates of recidivism among program participants. However, relatively little is known about the structure and processes associated with effective drug courts. The current study seeks to address the issues by exploring the moderating influence of programmatic and non‐programmatic characteristics on effectiveness. The methodology goes beyond previous meta‐analyses by supplementing published (and unpublished) findings with a survey of drug court administrators. Consistent with previous research, the results revealed drug courts reduce recidivism by 9% on average. Further analyses indicated target population, program leverage and intensity, and staff characteristics explain the most variability in drug court effectiveness. These findings are discussed within the context of therapeutic jurisprudence and effective interventions.

Keywords:

• drug court
• meta‐analysis
• therapeutic jurisprudence
• effective interventions

Notes

1. Drug courts evaluated in conjunction with other courts were excluded from this portion of the data collection process unless outcome data could be disaggregated.

2. Administrators, probation officers, and others filled the role of “coordinator” for programs that did not have coordinators.

4. Statistical power was a dichotomous measure and calculated using the GPOWER statistical power analysis (Faul & Erdfelder, 1992). Studies that had 80% or greater power were coded as having sufficient power.

3. A coding guide is available upon request.

5. Alpha was set to 0.05 if not reported in the study.

6. Analyses relating to inter‐rater reliability were limited to the study features, methodological characteristics, general drug court characteristics, and outcome data.

7. Base rates were unavailable for nine studies and an assumption was made that extreme base rates was not an issue.

8. Macros written by Wilson (2002) were used for the calculations of the mean effect size and subsequent analyses. These macros are detailed in Lipsey and Wilson (2001) and are available online at http://mason.gmu.edu/~dwilsonb/ma.html.

9. Bivariate correlations are available upon request.

10. The variance inflation factor (VIF) was also examined for each model as an additional test of multi‐collinearity.

11. A fail‐safe estimate was calculated using Orwin’s formula (1983) as presented by Lipsey and Wilson (2001): k ₀ = k[(ĒS _k/ĒS _c)−1] where k ₀ equals the number of studies needed to reduce the mean effect size to the ES_c or the alternative mean effect size. Using this formula, k is the number of studies included in the calculation of the weighted mean effect size and ES _k is the estimated weighted mean effect size. The alternative effect size, ES _c was set to (ES _k /2). The number of studies necessary to reduce the mean effect size in half was 82.