Description of original award (Fiscal Year 2013, $297,583)
This program seeks to enhance what is understood about mentoring as a prevention strategy for youth who are at risk of involvement or already involved in the juvenile justice system. While mentoring appears to be a promising intervention for youth, more evaluation work is needed to further highlight the components of a mentoring program that are most effective. Research is also needed to demonstrate the specific components of mentoring programs that have a significant impact in reducing juvenile delinquency and offending. This program funds research studies that will inform the design and delivery of mentoring programs. OJJDP expects that the results of this effort will encourage a more effective utilization of resources as well as enhance the implementation of evidence-based best practices for juvenile mentoring.
Because the short-term effectiveness of community- and school-based youth mentoring for preventing delinquency and dropout is well documented, the field now needs research on those best practices that can amplify program effects. One best practice identified by DuBois and colleagues in 2002 is the use of structured activities, yet little in known about how specific mentoring interactions contribute to program outcomes. Early intervention research has revealed negative long-term outcomes of programs (initially found to be effective) that are linked to specific peer processes occurring in programs. In youth mentoring, we also must differentiate helpful from harmful mentoring interactions and determine their connection to long-term outcomes like criminality in adulthood. The goals of this study are to (a) test theory-driven hypothesizes about the most appropriate mentoring interactions for children versus adolescents and (b) examine whether initial reports of both iatrogenic/harmful and positive programmatic outcomes in two prior studies (Goodman, 1972; Karcher, 2008) have persisted 10 and 40 years later. This study proposes the re-analysis of the original data from Goodman's study and the collection of additional long-term follow up data on adult criminality for both studies (at 40 and 10 years post intervention). These two datasets are unique in three ways. First, neither study used the waitlist comparison group approach so common in large-scale youth mentoring evaluations (which can nullify or preclude longitudinal follow-ups). Both studies employed randomized, experimental no-treatment or alternative treatment control groups. Second, both studies collected weekly activity logs (n > 400 mentors). The data from these logs allow tests of the relationships between specific mentoring interactions and immediate, intermediate, and long-term outcomes. Third, in both studies, the data were collected at four or more points in time (pre, mid, post, and early follow-up) over a two-year period which affords greater statistical power than the common pre-post comparisons. With a combined sample of 725 youth, hypotheses about program effects, moderators, and the most developmentally appropriate mentoring interactions will be tested. Repeated measures analyses of covariance (following FIML missing data imputation), will test main effects and moderator effects of program participation on immediate, intermediate, and longterm outcomes. Regression analyses test whether specific mentoring interactions mediate identified program effects. Latent group-based trajectories (finite mixture modeling) of mentoring interactions over time will be used test hypotheses about which mentoring styles are more effective for older and younger youth and whether these patterns are consistent across school vs. community settings.