The Effects of Changes in Dynamic Risk on Reoffending among Serious Juvenile Offenders Returning from Residential Placement

Abstract

This study examines whether changes in dynamic risk during juvenile justice long-term residential placement affect recidivism. Advancing the work of prior dynamic risk change analyses, we examine a large sample of 11,891 male and 1930 female juvenile offenders while taking methodological steps to ensure successful and non-successful youth (in terms of recidivism) are (statistically) identical at admission. Specifically, we employ propensity score matching on residential placement youth who recidivate post-release and non-recidivists on static risk factors and initial dynamic risk scores assessed at admission to residential placement. Post-matching, changes in dynamic risk factors from initial assessment at admission to exit assessment at release are examined with a focus on whether those changes distinguish recidivists from non-recidivists. Separate analyses are conducted for male and female juveniles. Results indicate that changes in dynamic risk do affect recidivism likelihood, but that different factors matter for males and females. These sex-specific models allow for distinct policy recommendations.

Keywords:

• juvenile offenders
• dynamic risk
• recidivism

Notes

¹ 3590 residential completions did not have an R-PACT assessment, and 704 additional youth only had one R-PACT assessment (prohibiting examination of changes in dynamic risk). Additionally, 6 youth (5 males, 1 female) were classified as “other” race/ethnicity and were excluded, as were 162 whose home county was indicated as outside of Florida. 83.3% of the youth not included in the study completed residential programs during the first fiscal year (1 July 2009–30 June 2010) and were not assessed by the FDJJ risk assessment due to the statewide implementation of that tool during that fiscal year (only new additions were required to be assessed during R-PACT implementation). Included youth did not differ from excluded youth in terms of race (53% Black for included youth, 51% for excluded; χ² = 3.70), ethnicity (11.9% Hispanic for included males, 11.5% for excluded; χ² = 0.059), or age at first offense (t = 0.326). Included youth were a greater proportion male (86% compared to 83.6% for excluded youth, χ² = 14.62, p < .001).

2 The R-PACT is the risk/needs assessment used by the FDJJ for all youth admitted to residential placement. The purpose of the tool is to assess risk and protective factors, and is required by FDJJ at admission, and every 90 days thereafter. As such, reassessments allow for gaging progress during placement in terms of risk reduction and protective factor increases. The validity of the R-PACT has been assessed, revealing strong support (Hay et al., 2016).

3 We note here that length of stay in the residential program was not included as a matching indicator in the PSM model. An ancillary analysis revealed that length of stay was not significantly correlated with recidivism among this sample of juveniles (Spearman’s rank correlation = .0114, p = .1766). For this reason, we chose not to include it in the matching specification.

4 While it is possible that more localized differences in policing and judicial practices exist, as the juvenile justice system in Florida is structured around circuit court judges (in many instances encompassing several counties) who play a prominent role in all juvenile justice dispositions, we believe a matching specification with county-fixed effects does an adequate job parsing out county-level differences in enforcement and judicial practices. Findings which are supportive of this assertion stem from multilevel research that suggests that much less of the variation in recidivism is due to community-level factors (Wang, Hway, Todak, & Bales, 2014). Future work devoted to assessing the impact of differential enforcement on juvenile recidivism should assess the implications of this decision for individual-level studies on juvenile recidivism.

5 Given that nearly 70% of youth released from Florida residential programs are rearrested within one year of completion, matching with replacement was necessary so that a large proportion of recidivists were not lost off support. To avoid matching two youth with propensity scores that differ greatly from each other, we imposed a caliper distance of .001, resulting in high-quality matches for each youth who recidivated. Results of a matching procedure that did not allow for replacement and utilized a caliper of .01 were substantively similar to those presented.

6 The use of two standard deviations above and below the mean were selected due to their computational convenience. It is important to note, however, that the chosen points do indeed represent plausible values given the observed distribution in change scores for this sample of youth.

7 The use of dichotomous variables to represent an especially high-risk group of individuals is fairly commonplace in criminological research. Farrington and Loeber (2000) suggest that the dichotomization of continuous measures allows for substantive differences to be more easily ascertained, leading to meaningful policy implications.

8 In an ancillary analysis not shown, the mean—1SD was used as the cut point for being a top risk reduction performer (rather than the simple 25% method reported in the text). For males 1 standard deviation better than the mean in terms of how much risk was reduced, academic school status, peer relationships, parent/caretaker relationships, attitudes/behaviors/ and skills dealing with difficult emotions decreased recidivism likelihood (identical domains as the 25% method described in the text). For females, the same domains as the 25% best in terms of risk reduction method also significantly reduced recidivism (program supervised tasks and alcohol and drug use). Given the consistency of these findings across various specifications suggests the results are robust to two different cut point selections.

Additional information

Notes on contributors

Michael T. Baglivio

Michael T. Baglivio is Director of Research and Program Development for G4S Youth Services, LLC, examining the effectiveness of juvenile treatment programming on short- and long-term performance measures and outcomes. His research interests include criminological theory, risk assessment, and life-course criminology.

Kevin T. Wolff

Kevin T. Wolff is an Assistant Professor at John Jay College of Criminal Justice in New York City. He earned his PhD from the College of Criminology and Criminal Justice at Florida State University. His research interests include the spatial patterning of crime, juvenile justice, criminological theory, and quantitative methods.

Alex R. Piquero

Alex R. Piquero is Ashbel Smith Professor of Criminology and Associate Dean of Graduate Programs in the School of Economic, Political and Policy Sciences at the University of Texas at Dallas. His research interests include criminal careers, criminological theory, and quantitative research methods. He has received several research, teaching, and service awards and is Fellow of both the American Society of Criminology and the Academy of Criminal Justice Scienes. In 2014, he received The University of Texas System Regents’ Outstanding Teaching Award.

Matt DeLisi

Matt DeLisi is Professor and Coordinator of Criminal Justice Studies and Affiliate with the Center for the Study of Violence at Iowa State University. Professor DeLisi is Editor-in-Chief of the Journal of Criminal Justice and the author of more than 300 scholarly publications mostly in the areas of pathological criminality, psychopathy, self-control/self-regulation, inmate behavior, and the molecular/behavioral genetics of antisocial behavior. In 2012, Dr DeLisi received the prestigious Fellow Award from the Academy of Criminal Justice Sciences.

Michael G. Vaughn

Michael G. Vaughn is Professor and Director of the interdisciplinary PhD program in the School of Social Work at Saint Louis University. He has contributed more than 250 scholarly publications across a wide range of areas. Dr Vaughn’s research interests include temperament and antisocial behavior, biosocial criminology, youth violence, and drug use epidemiology.