Practice Profile: Mentoring for Youth Development

Tolan and colleagues (2008) conducted a comprehensive search to identify studies of mentoring programs that focused on youth at risk for delinquency or engaged in delinquent behavior. Prevention and intervention programs were included if they could be classified as standalone mentoring programs, or they included mentoring as a component. The outcomes of interest included delinquency, aggression, substance use, and academic achievement. The evaluation studies had to have a comparison group (this control condition could be no treatment, waiting list, treatment as usual, or placebo treatment). Studies were limited to programs in predominately English-speaking countries and had to be reported in English. The search covered the years 1970 to 2005.

The search yielded 39 studies that met the criteria for inclusion: 22 were randomized control trials (RCTs), and 17 were quasi-experimental designs (QEDs). Twenty-one studies reported delinquency outcomes, 19 reported academic achievement outcomes, 6 reported drug use outcomes, and 6 reported aggressions as an outcome. Tolan and colleagues reported that the RCTs had larger effect sizes than QEDs (a finding that reverses the more typical finding that QEDs yield larger effect sizes than RCTs). The 39 studies included 9,262 participants (4,732 in the treatment groups, 4,530 in the control groups). A random-effects model was used owing to the heterogeneity across studies.

Raposa and colleagues (2019) conducted a comprehensive meta-analysis to examine the overall impact of youth mentoring with all outcome studies of intergenerational, one-on-one youth mentoring programs written in the English language from 1975 to 2017. A comprehensive search of the literature published before June 2017 was conducted to identify published and unpublished evaluations of mentoring programs. Computerized databases searched were PsycINFO, ERIC, and ProQuest, and manual searches were conducted of prior meta-analyses and qualitative reviews. Studies were included in the sample if they used a randomized controlled trial or quasi-experimental design to evaluate a formal mentoring program, with a nonparental adult or older youth acting in a nonprofessional helping capacity with a specific younger person to promote positive youth outcomes through the relationship. Studies were excluded if they 1) had similar age peer mentoring; 2) only used group mentoring; 3) included mentees over age 18; 4) used insufficient treatment versus control group differentiation; 5) evaluated programs in which mentoring was not a primary or secondary component; 6) included outcome measures outside of the broad categories of psychological, social, school, health, or cognitive; 7) did not include sufficient information to compute an effect size; or 8) were not published in English.

The search process yielded 70 evaluations of mentoring programs with a sample size of 25,286 youths. More than half of the studies (54 percent) used randomized controlled trials, while the remaining ones (46 percent) used quasi-experimental designs. The average age of mentees was 12 years old, and 55 percent of the mentees were male. The sample included youth mentees who were Black (43 percent), white (32 percent), Hispanic (26 percent), Asian (5 percent), multiethnic (4 percent), Hawaiian (1 percent), American Indian (1 percent), or another ethnicity (13 percent). (Some mentees identified as more than one race, hence the totals’ exceeding 100 percent.) The average age of mentors was 36 years, and 58 percent of the mentors were female. The mentors were white (62 percent), Black (31 percent), Hispanic (9 percent), Asian (6 percent), multiethnic (1 percent), American Indian (1 percent), or another ethnicity (10 percent). Mentoring programs were on average 11 months long, meeting four times a week for about 1.7 hours. Program sites were either at a school (63 percent) or in the community (37 percent) and were predominantly urban/suburban (87 percent), in mixed locations (9 percent), or in rural areas (4 percent). Programs were mainly unstructured (62 percent), semistructured (21 percent), or structured (17 percent), and the great majority (88 percent) did not include a curriculum).

The standardized mean difference between the treatment and control groups was calculated as an effect size measure, with a positive value indicating an advantage for the intervention group over the comparison group. This value was transformed into Hedges’ g to adjust for differences in sample size. In most of the studies, more than one effect size was calculated. A three-level approach to meta-analysis was applied to deal with the interdependency of effect sizes. Three sources of variance are modeled in a three-level meta-analysis: the sampling variance of the observed effect sizes (level 1), the variance between effect sizes from the same study (level 2), and the variance between studies (level 3). The three-level meta-analysis was conducted using a multilevel random effects model. The analyses explored the effects of youth mentoring across five outcome categories: school (51 studies), social functioning (43 studies), health (17 studies), cognitive functioning (35 studies), and psychological outcomes (46 studies).