Prevention Science, Vol. 1, No. 3, 2000
Can Mentoring or Skill Training Reduce Recidivism?
Observational Study with Propensity Analysis
Elaine A. Blechman,
and Clay Helberg
We compared juvenile offenders’ recidivism following nonrandom assignment to juvenile
diversion (JD, n ϭ 137), JD plus skill training (ST, n ϭ 55), or JD plus mentoring (MEN,
n ϭ 45). Intake characteristics that distinguished intervention groups were used to calculate
assignment propensity scores. After propensity score blocking balanced intake characteristics,
ST proved more cost effective than MEN, achieving a 14% relative reduction in recidivism
at a savings of $33,600 per hundred youths. In ST, 37% were rearrested 2 years or more
after intake, compared to 51% in MEN and 46% in JD. In two of ﬁve propensity subclasses,
time to ﬁrst rearrest was longer in ST (M ϭ 767 days) than in MEN (M ϭ 638 days) or JD
(M ϭ 619 days). These results argue for an experimental comparison of ST and MEN and
for observational studies with propensity analysis when randomization to juvenile justice
interventions is infeasible.
KEY WORDS: prevention; recidivism; juvenile offender; skill training; mentoring.
Each delinquent youth who matures into a career
criminal costs taxpayers from $1.3–$1.5 million (Co-
hen, 1998). To achieve clinical signiﬁcance, an inter-
vention for juvenile offenders must reduce recidivism
signiﬁcantly below the 50% base rate (Dembo et al.,
1995; U.S. Department of Justice, 1987; Visher, Latti-
more, & Linster, 1991). As evidence of cost effective-
ness, a juvenile justice intervention must reduce recid-
ivism below base rate at a lower cost than a viable
alternative intervention (Yates, 1995). For ready dis-
semination, an intervention must win the conﬁdence
of juvenile justice practitioners and suit the broad
spectrumofjuvenile offenders including youths whose
parents will not or cannot participate.
Problem-solving skill training and mentoring re-
quire no parental involvement and are popular with
juvenile justice practitioners (Montgomery et al.,
1994). Experimental studies support the promise of
skill-training interventions for reduction of recidivism
Department of Psychology, University of Colorado at Boulder.
Correspondence should be directed to Elaine A. Blechman, De-
partment of Psychology, Campus Box 345; U. of Colorado-
Boulder, Boulder, CO 80309-0345. e-mail: eblechman@worldnet.
1389-4986/00/0900-0139$18.00/1 2000 Society for Prevention Research
among delinquent youths (Hawkins, Jenson, Cata-
lano, & Wells, 1991; Kazdin, 1990; Ollendick, 1996;
Tate, Reppucci, & Mulvey, 1995). The Cambridge-
Somerville Youth Study (CSYS) provides the most
credible information, to date, about mentoring. Be-
tween 1935 and 1939, the CSYS recruited boys below
age 12 from a low-income, high-crime area. Boys were
matched on age, family environment, and delin-
quency-prone histories and then randomized to either
an untreated control or to an experimental condition.
In the experimental condition, ‘‘a social worker . . .
tried to build a close personal relationship with the boy
and assist both the boy and his family in a number of
ways’’ (McCord, 1992, p. 198). At the program’s end
in 1945, boys in the experimental condition had been
visited, on average, two times a month for 5.5 years.
Some might argue that mentoring in the experimental condition
was not representative of usual mentoring procedures. However,
an equally well-conducted study of the long-term impact of typical
agency-arranged mentoring is unavailable. Reports from commu-
nityagenciesthatoperate mentoring programs suggest that the lack
of long-term commitment of mentors impedes program success.
The CSYS ended when mentors were drafted into the military at
the onset of World War II and proteges were 14 or 15 years old.
Some have suggested that the experimental condition fared worse
owing to abrupt termination of the mentoring relationship.