American Journal of Community Psychology [ajcp] PP402-368416 March 12, 2002 8:16 Style ﬁle version Nov. 19th, 1999
American Journal of Community Psychology, Vol. 30, No. 2, April 2002 (
The Test of Time: Predictors and Effects of
Duration in Youth Mentoring Relationships
Jean B. Grossman
Jean E. Rhodes
University of Illinois at Urbana-Champaign
The effects and predictors of duration in youth mentor relationships were ex-
amined. The study included 1,138 young, urban adolescents (Mean age =
12.25), all of whom applied to Big Brothers Big Sisters programs. The ado-
lescents were randomly assigned to either the treatment or control group, and
administered questions at baseline and 18 months later. Adolescents in rela-
tionships that lasted a year or longer reported the largest number of improve-
ments, with progressively fewer effects emerging among youth who were in
relationships that terminated earlier. Adolescents who were in relationships
that terminated within a very short period of time reported decrements in
several indicators of functioning. Older adolescents, as well as those who had
been referred for services or had sustained emotional, sexual or physical abuse,
were most likely to be in early terminating relationships, as were married vol-
unteers aged 26–30 and those with lower incomes. Several dyadic factors were
also found to be related to earlier terminations, including race, gender, and
KEY WORDS: mentoring; adolescence; volunteerism.
This study was completed with the assistance of a grant from the William T. Grant Foundation.
The authors also gratefully acknowledge the assistance of Joseph P. Tierney, Nancy Resch,
Sarah Pepper, and the cooperation of Big Brothers Big Sisters of America.
To whom correspondence should be addressed at Public/Private Ventures, One Commerce
Square, 2005 Market Street, Suite 900, Philadelphia, Pennyslvania 19103.
2002 Plenum Publishing Corporation
American Journal of Community Psychology [ajcp] PP402-368416 March 12, 2002 8:16 Style ﬁle version Nov. 19th, 1999
200 Grossman and Rhodes
Interventions that linkadolescents with volunteer mentors havebecome
increasingly common in recent years. An estimated ﬁve million American
youth are currently involved in school- and community-based volunteer
mentoring programs nationwide, including more than 100,000 participants
in Big Brothers Big Sisters of America programs (McLearn, Colasanto,
& Schoen, 1998). Enduring mentoring relationships have been found to be
associated with a range of beneﬁts to youth. But what are the consequences
to adolescents when relationships terminate prematurely? Indeed, approx-
imately half of all youth mentoring relationships dissolve after only a few
months, often as the result of the volunteers feeling overwhelmed, burned
out, or unappreciated (Freedman, 1993; Hamilton & Hamilton, 1990; Styles
& Morrow, 1992). Here we address this issue, and attempt to identify the
predictors of early termination in youth mentoring relationships.
Evaluations of volunteer mentoring programs provide evidence of pos-
itive inﬂuences on adolescent developmental outcomes, including improve-
ments in academic achievement, self-concept, prosocial behavior, and in-
terpersonal relationships (Davidson, Redner, Blakely, Mitchell, & Esmhoff,
1987; DuBois & Neville, 1997; Grossman & Tierney; 1998; LoSciuto, Rajala,
Townsend, & Taylor, 1996). Despite this evidence, very little is known about
how variations in the characteristics of mentor relationships relate to youth
outcomes. For example, while some relationships last for several years, many
volunteer relationships terminate within only a few months. Because the
central component of mentoring is the formation of intensive one-on-one
relationships, terminations may touch on vulnerabilities in youth in ways
that other, less personal interventions do not. This may be particularly true
for youth who are referred to relationship-based interventions. In particular,
many adolescents in mentoring programs come from single-parent homes
(an eligibility requirement for some programs) and may have already sus-
tained the loss of regular contact with their nonresidential parent. Such
youth may feel particularly vulnerable to, and responsible for, problems in
subsequent adult relationships (Wallerstein, 1988). Other youth may have
experienced unsatisfactory or rejecting parental relationships in the past.
Consequently, they may have developed internal representations of rela-
tionships that incorporate fears and doubts about whether others will ac-
cept and support them (Bowlby, 1982; Egeland, Jacobvitz, & Sroufe, 1988).
When such adolescents encounter cues that relationships will not proceed,
American Journal of Community Psychology [ajcp] PP402-368416 March 12, 2002 8:16 Style ﬁle version Nov. 19th, 1999
Predictors and Effects of Duration in Youth Mentoring Relationships 201
however minimal or ambiguous, they may readily perceive intentional re-
jection from their mentors (Downey & Feldman, 1996; Downey, Lebolt,
Rincon, & Freitas, 1998).
Irrespective of their relationship histories, all youth may show certain
vulnerabilities to early terminations. Adolescence is a life stage during which
issues of acceptance and rejection are especially salient (Cauce, Mason,
Gonzales, Hiraga, & Liu, 1994; Lerner & Galambos, 1998). To the extent that
adolescents have identiﬁed with their mentors, and have begun to value the
relationship, they may feel profound disappointment when the relationship
does not progress. Feelings of rejection and disappointment, in turn, may
lead to a host of negative emotional, behavioral, and academic outcomes
(Downey et al., 1998).
Mentor relationships thattake hold, on the otherhand, are likely to grow
progressively more effective with time. Researchers generally agree that
mentors promote positive developmental outcomes through role modeling
and the provision of emotional support and positive feedback. By serving as
supportive models of success, mentors may directly stimulate improvements
in adolescents’ self-perceptions, attitudes, and behaviors (Bandura, 1969;
Hamilton & Hamilton, 1990; Klaw & Rhodes, 1995; Taylor, 1989; Walker
& Freedman, 1996). Additionally, there is some evidence to suggest that
mentors may affect change through their positive inﬂuence on the more
proximal relationships in adolescents’ lives. By helping adolescents cope
with everyday stressors, providing a model for effective conﬂict resolution,
and indirectly reducing parental stress, mentor relationships are thought
to have the capacity to facilitate improvements in parent–child interactions
(Flaxman, Ascher, & Harringon, 1988; Rhodes, Grossman, & Resch, 2000;
Rhodes, Haight, & Briggs, 1999). Additionally, enduring positive relation-
ships may modify adolescents’ general perceptions of relationships (Bowlby,
1982; Belsky & Cassidy, 1994; Sroufe, 1995). Speciﬁcally, mentors can chal-
lenge negative views that adolescents may hold of themselves or of relation-
ships with adults and demonstrate that positive, caring relationships with
adults are possible. The helping relationship can thus become a “corrective
experience” for those adolescents who may have experienced unsatisfactory
relationships with their parents (Olds, Kitzman, Cole, & Robinson, 1997;
Main, Kaplan, & Cassidy, 1985).
Because such processes are complex and, in some instances, may in-
volve changes in internal representations of relationships, it is likely that the
beneﬁts of mentoring emerge over a relatively long period of time (Rhodes
et al., in press). In their qualitative investigation of mentoring relationships,
for example, Styles and Morrow (1992) concluded that youth needed to be
engaged with their mentors for at least 6 months before the relationships
began to take hold. We examine the issue of duration in the current study,
202 Grossman and Rhodes
and attempt to determine whether there is some minimum level of exposure
after which beneﬁts are more likely to emerge.
In light of the potential signiﬁcance of relationship duration, both in
terms of the possible harm associated with early terminations and the ben-
eﬁts of sustained contact, it is also important to identify factors that predict
the length of the relationship. Observations of mentoring programs, as well
as a small but growing body of psychological research on volunteerism, sug-
gest that early terminations of volunteer relationships may occur for a wide
variety of reasons. Graduations, illnesses, or changes in family structure or
residence, for instance, may inﬂuence adolescents’ eligibility or leave dyads
unable to meet on a regular basis (Sipe, 1996). Some volunteers may be dis-
couraged by what they perceive as a lack of appreciation on the part of their
mentee or ﬁnd that the personal investment that is required to work with
troubled adolescents exceeds their expectations, particularly if the volun-
teers’ involvement is drawing them away from social and family obligations
(Freedman, 1993; Omoto & Snyder, 1995). In some instances, adolescents
may terminate relationships in response to what they perceive as unsupport-
ive, disappointing, or overly demanding mentors (Styles & Morrow, 1992).
Still other dyads may lack a basic chemistry and the relationships may grad-
ually give way to other demands. Indeed, Flaxman et al. (1988) has discussed
the social distance that often exists among middle-class mentors and lower-
income mentees, particularly when the mentors and mentees are of different
races. In this study, we will attempt to identify volunteer, adolescent, and
dyadic predictors of sustained involvement in mentoring relationships.
Goals of the Current Paper
To address the issues raised above, we examine the differential effects
and predictors of mentor relationships of varying lengths. It is hypothesized
that the effects of mentoring relationships will intensify with time, and that
relatively short matches will be disruptive to youth. Next we examine the
predictors of relationship duration. At a theoretical level, identifying ef-
fects and predictors of sustained volunteerism touches on questions that
are fundamental to our understanding of helping relationships (i.e., how
does duration affect outcome? what personal and social factors promote
long-term involvement?). Moreover, in light of the sheer number of ado-
lescents who are currently involved in volunteer mentoring interventions,
as well as the lack of empirically based guidelines for the screening and
matching of volunteers, ﬁndings regarding the effects and predictors of re-
lationship duration are likely to have far-reaching implications. This study
makes use of longitudinal data from the largest and arguably most inﬂuential,
Predictors and Effects of Duration in Youth Mentoring Relationships 203
evaluation of mentoring to date (Grossman & Tierney, 1998) to address these
The study included 1,138 youth, all of whom applied to Big Brothers
Big Sisters programs in 1992 and 1993. Applicants were randomly assigned
to either the treatment or control group, and administered questions at base-
line and 18 months later. Eighty-ﬁve percent of the sample (N = 959; 487
treatments and 472 controls) completed both the baseline and the follow-up
interviews. Over half of this analysis sample were boys (62.4%) and ap-
proximately half were members of minority groups (57.5%). Seventy-one
percent of the minority youth were African Americans, 18% were Hispanic,
and the remaining were members of a variety of other racial/ethnic groups.
Participants ranged in age from 10 to 16 (Mean = 12.25), most (69%) of
whom were between the ages of 11 and 13. More than 40% of the youth lived
in households that were receiving food stamps and/or public assistance. The
only systematic difference between the treatment and control group youth
at baseline was that the treatment youth had the opportunity to be matched
Design and Procedure
From the network of over 500 Big Brothers Big Sisters local agencies,
8 agencies were selected to participate in the outcome study. The key selec-
tion criteria for inclusion in the impact study were a large, active caseload
waiting list and geographic diversity. With only a few exceptions, all of the
youth who enrolled in the 8 selected Big Brothers Big Sisters agencies dur-
ing the intake period were encouraged to participate in the research. Once a
youth was informed about the study, determined to be eligible, and assented
to participate (along with parents’ signed, informed consent), he or she was
randomly assigned to either the treatment or control group. Only 2.7% of the
youth refused to participate in the evaluation. The control group was placed
on a waiting list for a poststudy match. All participants were interviewed by
telephone before they knew their experimental status. Follow-up interviews
were conducted 18 months later by telephone with baselined participants.
Agency staff matched particular adult volunteers with particular youth
on the basis of a variety of factors, including shared interest, reasonable
204 Grossman and Rhodes
geographic proximity, and same-race match preference. All volunteers un-
derwent an intensive screening process, followed by agency-based train-
ing and case management. At the conclusion of the study, 378 (78%) of
the treatment youth had been matched.
At the time of the follow-up,
matched youth had been meeting with their mentors for approximately
12 months, while 40% of the matches were no longer meeting. Among
closed matches, the pairs met for an average of 9 months. The ongoing
matches had been meeting an average of 12.9 months. Over 70% of the
youth met with their mentor at least three times a month and approximately
45% met one or more times per week. An average meeting lasted 3.6 hr.
Dyads typically engaged in a wide variety of leisure- and goal-oriented dis-
cussions and activities with the overall goal of promoting the youth’s positive
The Inventory of Parent and Peer Attachment (IPPA; Armsden
& Greenberg, 1987) is a 23-item scale containing questions related to a
child or adolescent’s relationship with his/her primary care giver (the cor-
responding peer questions were not administered). Responses are coded
on a 4-point scale, ranging from 1 (hardly ever true)to4(very often true).
The IPPA contains three subscales: communication (e.g., my mother can tell
when I am upset about something), trust (e.g., my father respects my feel-
ings), and alienation (e.g., talking over problems with my mother makes me
feel ashamed or foolish). At pretest, Cronbach’s alpha reliability coefﬁcients
of the subscales were .77, .83, and .76, respectively.
This six-item subscale of the Self-Perception Proﬁle for Children
(Harter, 1986) contains statements describing conﬁdence in school work,
Agency staff reported three major reasons for the failure to match the 109 treatment youth
during the study period. Thirty-three of the unmatched treatment youth became ineligible
during the study period because the parent remarried, the youth was no longer within the
eligible age range, or the youth’s place of residence changed. Thirty-one were not matched
because the youth no longer wanted a Big Brother or Big Sister. Twenty-one were not matched
because a suitable volunteer could not be found during the study period. The 24 remaining
treatment youth were not matched for a variety of reasons, most commonly because the parent
or youth did not follow through with the intake process.
Predictors and Effects of Duration in Youth Mentoring Relationships 205
dividing children into two groups, e.g., “some kids feel that they are very
good at their schoolwork/other kids worry about whether they can do the
schoolwork assigned to them.” Respondents were asked to determine if they
were more like the ﬁrst or second group, and whether the statement was
“really true” or “sort of true” for them (α = .77).
Grades and Attendance
Individual items relating to scholastic behaviors were asked, including
grades, number of unexcused absences from school, visits to college cam-
puses, books read, trips to the library, hours spent on homework, and hours
spent reading. For purposes of this study, we focused on grades and the
number of unexcused absences.
This 18-item measure (Berndt & Miller, 1986) assesses the extent to
which respondents value academic success and the information that they
learn in school, e.g., “do you care about doing your best at school?” Respon-
dents were asked to indicate the frequency with which they felt certain ways
about school, ranging from 1 (hardly ever)to4(pretty often)(α =.86).
This six-item subscale of the Self-Perception Proﬁle for Children
(Harter, 1986) contains statements describing the global self-worth of two
groups, “e.g., some kids are pretty pleased with themselves/other kids are
often unhappy with themselves.” Respondents were asked to determine
whether they were more like the ﬁrst or second group, and whether the
statement was “really true” or “sort of true” for them (α = .75).
Quality of Relationship
Relationship quality was determined by Langhout, Osborne, and
Rhodes’ analysis of scales that characterized youth’s feelings toward their
mentors (Langhout, Osborne, & Rhodes, 1999). The two scales that were
most predictive of outcomes, “youth-centered” or the degree to which the
volunteer took the youth’s desires into consideration and “disappointment”
206 Grossman and Rhodes
or the degree to which youth felt let down or disappointed by their mentors,
were considered indices of relationship quality.
Length of Relationship
Relationship length was assessed in terms of months, and coded as 0 for
all controls and unmatched treatment participants.
Study 1: Effects of Relationship Length
In order to investigate the effects of relationship duration, we began by
categorizing the mentored youth into four groups, depending on how long
their matches had lasted: less than 3 months (6%), 3 to just under 6 months
(13%), 6 to just under 12 months (36%), and 12 months or more (45%). We
then used multivariate regression to estimate the effect of the length of the
match on youth outcomes.
Speciﬁcally, the four length-of-match dummy variables were entered
into a regression equation for each outcome. The equations were estimated
over the full sample—treatment and control youth. All of the length-of-
match dummies were set equal to zero for the controls and unmatched treat-
ment participants. Because we were interested in explaining the changes
during the 18-month period, not the level of the outcomes, we controlled
for baseline levels of variables. Other baseline characteristics were also in-
cluded in the models to reduce the variance unrelated to mentoring.
presents the resulting estimates of how treatment youth fared compared
with similar control group youth who did not have mentors. Youth who
were in matches that terminated within the ﬁrst 3 months suffered sig-
niﬁcant declines in their global self-worth and their perceived scholastic
In addition to the length-of-match variables, the following variables were also included in the
regressions: the baseline value of the outcome; the youth’s age, gender, race; whether the youth
was an academic underachiever; if the youth was learning disabled; if the parent worked full
time; if the family received welfare; if the parent had a GED or high school diploma; if the
youth had repeated a grade; if the youth was an only child; the number of siblings; the number
of moves the youth had made in the 2 years prior to applying for BBBS; whether the youth had
a natural mentor; whether the youth had a Big previously; whether the custodial parent was
male; if the parent had been a teen parent; if the parent had never been married; if the parent
had referred the child; if the youth had experienced emotional, sexual, or physical abuse; if
the other parent was missing due to death, divorce, or illness; if the family had a history of
substance abuse or domestic violence; if the youth live in a rural or urban environment; and
Predictors and Effects of Duration in Youth Mentoring Relationships 207
Table I. Estimated Impacts Using the Observed Length of Match (Standardized Coefﬁcients)
Outcome <3 mo 3–6 mo 6–12 mo 12+ mo
(−.05) 0.30 (.25) 0.08 (.48) 0.76
Perceived social acceptance −0.95 (−.02) 0.19 (.02) 0.28 (.03) 0.83
Perceived scholastic competence −1.83
(.03) 0.58 (.10) 0.53 (.08) 0.93
Skipping school −0.26 (−.09) −0.18 (−.07) −0.65
(−.12) −0.40 (−.08) (p = .06)
Grades 0.07 (.05) 0.10 (.05) 0.08 (.03) 0.26 (.07) (p = .06)
Value of school −1.16 (−.05) 0.58 (.08) −1.15 (−.02) 1.85
Quality of the parental relationship 1.75 (.04) 4.17
(.09) −0.30 (.00) 2.35
Hitting someone −1.28 (−.11) −2.08
(−14) −1.06 (.09) 0.17 (−.02)
Frequency of drug use 0.21 (−.03) 0.39 (.01) −0.40
Frequency of alcohol use 0.29 (.06) 0.18 (.05) −0.12 (.01) −0.57
p ≤ .05.
p ≤ .01.
208 Grossman and Rhodes
competence. On the other hand, youth who were in matches that lasted more
than 12 months reported signiﬁcant increases in their self-worth, perceived
social acceptance, perceived scholastic competence, parental relationship
quality, school value, and decreases in both drug and alcohol use.
This pattern of estimated impacts is consistent with the hypotheses that
short-lived matches can have detrimental effects on youth; and that the
impact of mentoring grows as the relationship matures. However, a similar
pattern would also have been observed if the youth who were particularly
well-adjusted were also the youth most likely to be able to sustain mentor
relationships, whereas the less well-adjusted youth were the mostlikely to fail
in establishing a relationship with a mentor. If this were the case, then youth
in longer matches may have improved relative to those in shorter matches,
not because of their greater dosage of mentoring, but simply because length-
of-match sorts youth on the basis of their adjustment status.
Statistically, the selection bias would manifest itself as a correlation be-
tween the length of the mentoring relationship and youth outcomes. One
can correct for this potential selection bias by substituting an instrument for
length of match that is similar to the observed length of match but which
is purged of the unwanted correlation with the error term. This can be ac-
complished using the statistical technique called Two-Stage Least Squares
(2SLS; Berry & Feldman, 1985; James & Singh, 1978). In the ﬁrst-stage re-
gression, the endogenous variable, length-of-match, is regressed on all the
exogenous variables in the model, plus additional variables that are cor-
related with length of match but uncorrelated with the outcome variables.
We use the number of times the youth has moved during the 18-month
follow-up period and whether they live in two-parent families at follow-up,
because if the youth moves out of the BBBS catchment area or lives in
a two-parent household they become ineligible for BBBS. The predicted
value of length-of-match, which is now assumed to be free of correlated
error, is used in place of the original value in the second stage of the re-
gression. By regressing length-of-match onto variables that are unrelated
to youth outcomes, the predicted value is assumed to be unrelated to youth
outcomes. Thus, the estimates produced in the second stage of the regression
are considered consistent estimates of the effect of relationship length on the
A ﬁrst stage regression of length of match on all the system’s exoge-
nous variables plus the identifying variable was performed. Overall, the cor-
relation between the resulting instrumental variable (included in Stage 2)
and the observed length of match is equal to .44. However, the model did
poorly in predicting short matches. The correlation between the instrument
and the observed length of match is equal to .36 for matches that lasted at
least 6 months but only .02 for matches that lasted less than 6 months. The
Predictors and Effects of Duration in Youth Mentoring Relationships 209
Table II. Estimated Impacts Using Two-Stage Least Squares
Outcome <6 mo 6–12 mo 12+ mo
Self-worth 0.02 (.00) 0.20 (.02) 0.48 (.04)
Perceived social acceptance 0.92 (.01) 0.04 (.00) 1.10
Perceived scholastic −3.06 (−.03) 0.69
Skipping school 1.30 (−.02) −0.36
Grades −0.36 (−.01) 0.15 (.04) 0.21 (.05)
Value of school −3.65 (−.02) 0.59 (.35) 0.94 (.05)
Quality of the parental 0.11 (.00) 1.26 (.05) 1.74 (.05)
Hitting someone 1.85 (.01) −1.26
(−.09) −0.14 (−.09)
Frequency of drug use −0.29 (−.09) −0.19 (−.06) ( p = .08) −0.31
Frequency of alcohol use 4.85
(.07) 0.10 (.02) −0.55
p ≤ .05.
p ≤ .01.
ﬁrst-stage regression couldhave done poorly in predictingvery short matches
for one of two reasons. If early terminations were driven primarily by the
mentor, then the estimates in Table I (using the observed length of match)
are unbiased and valid. Alternatively, if the terminations were youth driven,
then we would have a poor model of short matches. This implies that the
standard errors of the estimates in Stage 2 would be large, but the esti-
mates would still be unbiased. For the second-stage regression, we grouped
matches into three categories: less than 6 months, 6 to less than 12 months,
and 12 months or more.
Table II presents the 2SLS estimates of length of match on the out-
comes. Although the point estimates changed, the pattern of impacts still
primarily held. There were no signiﬁcant, positive effects for short matches
lasting less than 6 months and, in fact, the only signiﬁcant ﬁnding for this
group was an increase in alcohol use. There were a few signiﬁcant ﬁndings
in the 6–12-month group—an increase in perceived scholastic competence,
a decrease in days skipped, and a decrease in the number of times the youth
hit someone else. The largest number of signiﬁcant, positive effects emerged
in the 12-month or longer group, an increase in perceived scholastic com-
petence and self-perceived social acceptance, and reductions in truancy and
substance use. In general, the signiﬁcant, positive impacts increased with
relationship duration (see Fig. 1).
Predictors of Relationship Length
In light of the importance of duration to youth outcomes, the next step
was to identify factors that were associated with longer matches. At follow-
up, 60% of matches that had been made were still intact. Thus, we do not
210 Grossman and Rhodes
Fig. 1. Outcomes as a function of relationship duration.
know how long they ultimately lasted. Given that the completed length of
match is not observed for all sample members (i.e., some of the data are cen-
sured), ordinary least-squares techniques would have produced biased infer-
ences about factors associated with the longevity of matches. An appropri-
ate analytical technique for analyzing censured data is proportional hazard
rate analysis (Cox, 1972; Kalbfreisch & Prentice, 1980). This technique takes
into account the fact that some of the observations are not censured (i.e., the
shorter matches), whereas others are (i.e., the longer matches). Underlying
this approach is the assumption that all matches experience a probability of
breaking up each period. The smaller this break-up or “hazard” rate is, the
longer the match is expected to last. Before presenting information about
which factors increase or decrease the likelihood that a match breaks up, we
Predictors and Effects of Duration in Youth Mentoring Relationships 211
Fig. 2. Kaplan–Meier empirical hazard rates.
present information on the observed monthly hazard rates to help put the
estimated parameters into perspective.
Figure 2 plots the Kaplan–Meier empirical hazard rate. Each rate is
calculated as the number of matches that close in a given month relative to
the number of matches that survived that month. The average hazard rate (h)
is .06, which implies that the expected length of a match is 1/h or 16.6 months.
A 25% increase in this average hazard rate would decrease the length of the
match to 13.3 months, whereas a 50% increase would decrease the length of
the match to 11.1 months. A 25% decrease in the average hazard rate would
increase the length of the match to 22 months.
Four sets of factors were examined as possible predictors of relation-
ship duration. These included the baseline characteristics of the youth, the
baseline characteristics of the adult, the characteristics of the match, such as
whether the pair was matched primarily because of similar interests or race,
and the quality of the relationship. We examined how the length of the match
was related to the latter two characteristics and whether the inﬂuences of
the other factors changed when these dimensions were taken into account.
Matches with adolescents who were referred for psychological or ed-
ucational programs, or had sustained emotional, sexual, or physical abuse,
were more likely to break up. Additionally, matches involving 13–16 year
olds were 65% more likely to break up in each period than matches with
212 Grossman and Rhodes
10–12 year olds. Using the average hazard rate, this would imply that if the
match of a younger adolescent lasted for two years, then the match of a
similar, but older, adolescent would last for a year and a quarter.
Matches involving higher income volunteers lasted longer than those
involving lower income volunteers. Volunteers’ age appeared to interact
with marital status in its effects on match duration. Relative to matches
with 18- to 25-year-old volunteers, unmarried volunteers aged 26–30 were
65% less likely to terminate each month, but married volunteers aged 26–
30 years were 86% (exp[1.05 − .43]) more likely to terminate each month.
The volunteers’ and adolescents’ age did not interact with each other in their
prediction of relationship duration.
Next, the effects of the characteristics of the match on duration were
examined, including the role of gender, race, and assignment considerations.
BBBS makes no cross-gender matches and so the effects of the adults’ gender
could not be separated from that of the youth. Nonetheless, female matches
were marginally more likely to terminate than those of males (p <.08).
Additionally, although same-race minority matches terminated marginally
more often than same-race white matches (p = .08), this ﬁnding did not hold
with respect to minority dyads in which race was an explicit matching criteria.
Similarly, although cross-race minority matches terminated more often than
same-race white ( p <.05), this ﬁnding did not hold with respect to dyads
in which the interests of the youth and volunteer were a primary matching
criteria (see Table III).
Table III. Hazard Rate Analysis of Length of Match
Variable Coefﬁcient Risk ratio p
Baseline values of
Volunteer is 26–30 −.43 .65 .08
Volunteer is 31 or older −.18 .84 .44
Youth is 13–16 .50 1.65 .001
Volunteer is 26–30 and married 1.05 2.87 .01
Female .36 1.40 .08
Same-race minority match .42 1.53 .09
Cross-race match .40 1.49 .05
Reason for match—race of mentor −.50 .61 .24
Reason for match—interest of mentor −.33 .72 .25
Referred as a school underachiever .35 1.42 .05
Referred for being overly dependent on adults .67 1.97 .002
Referred after intake for psychological testing 2.63 13.94 .0001
Referred after intake to an educational program .81 2.25 .04
Volunteer’s household income ($000s) −.23 .79 .02
Number of moves 2 years prior to baseline .19 1.21 .03
Youth had experienced abuse .42 1.53 .03
(emotional, sexual, or physical)
Note. The sample consists of 376 observations, 229 are censored. −2(Log Likelihood) is 1146
without the covariates and 1076 with them. The global null hypothesis is rejected at p = .0001.
Predictors and Effects of Duration in Youth Mentoring Relationships 213
Finally, we examined the potential mediating role of relationship qual-
ity (as measured by “youth-centered” and “disappointment” domains) on
the inﬂuence of the factors cited above. If all the factors became insigniﬁ-
cant once the quality of the relationship was held constant, then we could
conclude that the quality of the relationship fully mediated the inﬂuence
of the factors on the length of the match. If some of the factors remained
signiﬁcant but their coefﬁcients changed, then we could conclude that they
exerted some independence but are partially mediated through the quality of
the relationship. The relationship scales signiﬁcantly increased the explana-
tory power of the model (model chi-square 69.6 vs. 109.30) and attenuated
the negative effects of being a married volunteer 26–30 years old and being
of lower income. All of the other factors remained signiﬁcant, even after
taking into account the inﬂuence of relationship quality.
The ﬁrst goal of this study was to test the hypothesis that the effects of
mentoring relationships grow stronger over time, and that relatively short
matches can lead to negative outcomes. In support of this prediction we
found that youth who were in relationships that lasted a year or longer re-
ported improvements in academic, psychosocial, and behavioral outcomes;
and progressively fewer effects emerged among youth who were in rela-
tionships that terminated between 6 months and 1 year or between 3 and
6 months. Additionally, youth who were in relationships that terminated
within 3 months reported drops in self-worth and perceived scholastic com-
petence. When potential self-selection biases were taken into account, the
basic pattern of effects remained. Speciﬁcally, youth in relationships that
lasted for a year or more reported the largest number of improvements,
with fewer effects emerging among youth in relationships that lasted from
6 to 12 months. Those in relationships that terminated within 6 months re-
ported decrements in several indicators of functioning, including signiﬁcant
increases in alcohol use.
Taken together, this pattern of ﬁndings underscores the importance of
considering relationship duration in determining the effects of mentoring
programs. Consistent with previous research regarding the complexities of
mentoring relationships (Rhodes et al., 1999), most of the positive effects
emerged in relationships that persisted for a year or longer. This lag may
help to explain the relatively modest effects that have been reported in
mentoring program evaluations that occur before matches have been meet-
ing for at least a year (see Abbott, Meredith, Self-Kelly, & Davis, 1997;
Freedman, 1993). Modest effects sizes may also be an artifact of evaluation
214 Grossman and Rhodes
designs that combinerelationships of varying duration intoa single treatment
The ﬁndings regarding early terminations are consistent with previ-
ous work which has demonstrated the particular vulnerabilities of youth
to relationship disruption (Downey et al., 1998). Still, it is unclear whether
these negative effects stemmed from youth’s feelings of rejection and disap-
pointment or from other processes or contextual inﬂuences. Future studies
should include measures of adolescents’ sensitivitytorejection (e.g., Downey
et al., 1998), their attributions of regarding their mentors’ intent (e.g., Dodge,
1980), and other potential mediators of this link between early termination
and poor outcomes.
Of course, it remains possible that relationship duration is simply a
proxy for poorer underlying adjustment in youth. Speciﬁcally, the observed
negative effects of early terminations may reﬂect unmeasured factors such
as poor social skills or underlying psychopathology. It should be noted,
however, that the basic pattern of ﬁndings held even after controlling for
potential self-selection biases. Moreover, there were no baseline differences
between the treatment group and the controls on any measures, including
indices of psychosocial adjustment.
Proportional hazard rate analyses revealed several youth, volunteer,
and dyadic characteristics that were associated with higher termination rates.
In particular, older adolescents tended to have shorter relationships than
younger adolescents. In light of developmental changes that occur through-
out adolescence, this is not particularly surprising. For example, older ado-
lescents’ desires for autonomy and independence from adults may result
in less compliance and emotional accessibility. Similarly, peer and roman-
tic relationships may compete increasingly for adolescents’ attention and
Beyond age, adolescents who had sustained emotional, sexual, or physi-
cal abuse were also more likely to have shorter relationships. The challenges
associated with working with maltreated adolescents are likely to be sub-
stantial and, at least in the early stages of the relationships, accompanied by
fewer rewards. Indeed, maltreated youth frequently manifest highly prob-
lematic attachment relationships with their parents and other adults (e.g.,
Carlson, Cicchetti, Barnett, & Braunwald, 1989) and may ﬁnd it relatively
difﬁcult to establish close, supportive relationship with mentors. Unfortu-
nately, such youth are most likely to harbor expectations of rejection and
to experience negative consequence following early terminations (Downey,
Khouri, & Feldman, 1997). Given the potential of supportive relationships
for helping adolescents to transcend severe childhood rejection (Egeland
et al., 1988; Rhodes et al., in press), caseworkers should work closely with
Predictors and Effects of Duration in Youth Mentoring Relationships 215
such dyads to move them beyond the initial, challenging stages of the rela-
tionship. Along similar lines, the mentor relationships of adolescents who
had been referred for psychological treatment or educational remediation
were less likely to remain intact. Again, such youth may present challenges
that overwhelm the mentors’ capacity or willingness to help.
Several factors associated with the mentors’ characteristics were also
predictive of relationship duration. For example, volunteers with higher in-
comes tended to be in matches that lasted longer than lower income vol-
unteers. Although the positive impact of income on initial levels of vol-
untary participation has been conﬁrmed in many previous studies (Wilson
& Musick, 1997), no other studies have identiﬁed this variable as a predictor
of relationship duration. Nonetheless, mentors with higher incomes probably
have greater ﬂexibility in their work schedules and can more readily afford
amenities, such as child care and personal transportation, that increase the
convenience of sustained contact (Miller, Powell, & Seltzer, 1990).
Interestingly, married volunteers aged 26–30 were at greatest risk for
early termination. Although not speciﬁcally measured, this cohort may be
coping with the competing demands of their small children and have neither
the time nor ﬂexibility to sustain contact with potentially troubled youth.
As a corollary, unmarried adults in their late 20s may have approached the
volunteering activity as an opportunity to meet people, enrich their lives, and
contribute to the community, all of which have been identiﬁed as motivations
associated with volunteer relationship longevity (Omoto & Snyder, 1995;
Penner & Finkelstein, 1998). It should be noted that the risks associated
with being married and 26–30 years old were attenuated when relationship
quality was taken into account. In other words, if volunteers were able to
form good relationships with their youth, their marital status had little effect
on the ultimate length of the match. This highlights the need for more careful
screening and supervision of volunteers.
Several dyadic factors were related to somewhat higher termination
rates, including gender (matches with females) and race (matches with same-
and cross-race minorities), but these effects were only marginally signiﬁcant
and did not remain when speciﬁcations regarding the mentors’ race or inter-
ests were considered in the analyses. Still, these trends are worth noting as
they may provide insights into factors that may precipitate termination. Be-
ing female and/or of minority status tends to be associated with higher levels
of stress (Reid, 1988), which may increase the likelihood of early termination
in the relationship (Wilson & Musick, 1997). It appears, however, that this
risk can be overcome through the exercise of greater matching precision.
The strengths and limitations of the research deserve comment. Our
collection of data from a large, national sample of adolescents in naturalistic
216 Grossman and Rhodes
settings, over time (1.5 years) confers conﬁdence in the precision and gen-
eralizability of the ﬁndings. Nonetheless, the mentor relationships were all
situated within the context of a single youth mentoring program and, as
such, the pattern of ﬁndings may not apply as well to other, short-term or
less formal mentoring interventions. For example, some mentoring programs
may coincide with school calendars and, as such, have predetermined rela-
tionship durations of 9 months or shorter. Since students may enter such
interventions with different expectations, they may be less negatively af-
fected by terminations. Ideally, this study should be replicated with other
samples of adolescents and volunteers in other types of mentoring interven-
tions. It is also worth noting that the assessments were based solely on the
adolescents’ perceptions. Participants in this study may have been limited
in their ability to engage in assessments of their relationships and inhibited
in their willingness to report personal problems or relationship difﬁculties.
Future evaluations should move beyond adolescent self-reports to include
data from school records, teachers, and case managers.
Despite these limitations, this research has both basic and applied im-
plications. The ﬁndings shed light onto adolescents’ relationships with non-
parental adults and address fundamental issues regarding helping behavior.
The pattern of effects should stimulate additional research on adolescents’
attributions and attachment relationships, including variations in rejection
sensitivity and the underlying processes by which mentors effect positive
change. Additional research is also needed regardingthe factors that mediate
sustained mentoring, including the dispositional attributes and motivations
of volunteers in long-term relationships.
These ﬁndings also have implications for the reﬁnement of mentoring
interventions. By all accounts, the number of mentor volunteer programs
will only increase in the years ahead and it is very likely that this expansion
will include poorly funded efforts that are neither as intensive nor lasting
as Big Brothers Big Sisters (Sipe, 1996). Freedman (1993) has referred to
this enthusiasm for the rapid expansion of mentoring programs as “fervor
without infrastructure,” a view that amounts to the belief in simple solutions
to complex problems. He warns that
Fervor without infrastructure is dangerous at the program level because it leads
to disappointed mentors and youth. It is dangerous at the policy level because it
plays into the unfortunate tendency to lunge at new and glossy strategies, glorify
them over the short term, and discard them as they tarnish. More disturbing is the
way that fervor without infrastructure feeds the recurring appetite for voluntaristic
panaceas, idealized in isolation from institutions, and proposed as quick, cheap, and
easy. (p. 93)
Our research has the potential to contribute to a theoretically informed
and practically applicable understanding of mentoring relationships. The
Predictors and Effects of Duration in Youth Mentoring Relationships 217
ﬁndings serve as an acknowledgment of the potential beneﬁts of enduring
mentoring relationships and as a mandate for sufﬁcient program resources
to ensure reasonable levels of screening, training, and postmatch mentor
Abbott, D. A., Meredith, W. H., Self-Kelly, R., & Davis, M. E. (1997). The inﬂuence of a
Big Brothers program on the adjustment of boys in single-parent families. The Journal of
Psychology, 131, 143156.
Armsden, G., & Greenberg, M. T. (1987). Inventory of parent and peer attachment: Individual
differences in their relationship to psychological well-being in adolescence. Journal of
Youth and Adolescence, 16, 427–453.
Bandura, A. (1969). Social learning theory of identiﬁcation processes. In D. A. Goslin (Ed.),
Handbook of socialization theory and research. Chicago: Rand-McNally.
Belsky, J., & Cassidy, J. (1994). Attachment and close relationships: An individual difference
perspective. Psychological Inquiry, 5, 27–30.
Berndt, T., & Miller, K. (1986). Expectancies, values, and achievement in junior high school.
Journal of Educational Psychology, 82, 319–326.
Berry, W. D., & Feldman, S. (1985). Multiple Regression in Practice, Quantitative Applications
in the Social Sciences Series, No. 50, Beverly Hills: Sage.
Bowlby, J. (1982). Attachment and loss: Retrospect and prospect. American Journal of
Orthopsychiatry, 52, 664–676.
Carlson, V., Cicchetti, D., Barnett, D., & Braunwald, K. (1989). Finding order in disorganiza-
tion: Lessons from research on maltreated infants’ attachments to their care givers. In
D. Cicchetti & V. Carlson (Eds.), Child maltreatment: Theory and research on the
causes and consequences of child abuse and neglect. New York: Cambridge University
Cauce, A. M., Mason, C., Gonzales, N., Hiraga, Y., & Liu, G. (1994). Social support during ado-
lescence: Methodological and theoretical considerations. In F. Nestmann & K. Hurrelmann
(Eds.), Social networks and social support in childhood and adolescence. New York: Walter
Cox, D. R. (1972). Regression modes and life-tables (with Discussion), Journal of the Royal
Statistical Society, 34, 187–220.
Davidson, W. S., II, Redner, R.,Blakely, C. H., Mitchell, C.M., & Esmhoff, J. G. (1987). Diversion
of juvenile offenders: An experimental comparison. Journal of Consulting and Clinical
Psychology, 55, 68–75.
Dodge, K. A. (1980). Social cognition and children’s aggressive behavior. Child Development,
Downey, G., & Feldman, S. I. (1996). The implications of rejection sensitivity for intimate
relationships. Journal of Personality and Social Psychology, 70, 1327–1343.
Downey, G., Khouri, H.,& Feldman, S. (1997). Early interpersonal trauma andadult adjustment:
The mediational role of rejection sensitivity. In D. Cicchette & S. Toth (Eds.), Rochester
Symposium on Developmental Psychopathology: Vol. 8. The effects of trauma on the devel-
opmental process (pp. 85–114). Rochester, NY: University of Rochester Press.
Downey, G., Lebolt, A., Rincon, C., & Freitas, A. L. (1998). Rejection sensitivity and children’s
interpersonal difﬁculties. Child Development, 69, 1074–1091.
DuBois, D. L., & Neville, H. A. (1997). Youth mentoring: Investigation of relationship charac-
teristics and perceived beneﬁts. Journal of Community Psychology, 25, 227–234.
Egeland, B., Jacobvitz, D., & Sroufe, L. A. (1988). Breaking the cycle of abuse. Child Develop-
ment, 59, 1080–1088.
Flaxman, E., Ascher, C., & Harringon, C. (1988, December). Youth mentoring: Programs and
practices. New York: Columbia University, Teachers College. (Available from the ERIC
218 Grossman and Rhodes
Clearinghouse on Urban Education, Institute for Urban Minority Education, Box 40,
Teachers College, Columbia University, New York, NY 10027)
Freedman, M. (1993). The kindness of strangers: Adult mentors, urban youth, and the new
volunteerism. San Francisco: Jossey-Bass.
Grossman, J. B., & Tierney, J. P. (1998). Does mentoring work? An impact study of the Big
Brothers/Big Sisters program. Evaluation Review, 22, 403–426.
Hamilton, S. F., & Hamilton, M. A. (1990, June). Linking up: Final report on a mentoring
program for youth. New York: Cornell University, College of Human Ecology, Department
of Human Development & Family Studies.
Harter, S. (1986). The self-perception proﬁle for children. Unpublished manual, University of
James, L. R., & Singh, B. K. (1978). An introduction to the logic, assumptions, and basic
analytical procedures of two-stage least squares. Psychological Bulletin, 85, 1104–
Kalbfreisch, J. & Prentice, R. (1980). The Statistical Analysis of Failure time Data. New York:
Klaw, E. L., & Rhodes, J. E. (1995). Mentor relationships and the career development of preg-
nant and parenting African-American teenagers. Psychology of Women Quarterly, 19,
Langhout, R. D., Osborne, L. & Rhodes, J. E. (1999). Volunteer mentoring with at-risk youth:
Toward a typology of relationships. Manuscript submitted for publication.
Lerner, R. M., & Galambos, N. L. (1998). Adolescent development: Challenges and opportu-
nities for research, programs, and policies. Annual Review of Psychology, 49, 413–446.
LoSciuto, L., Rajala, A. K., Townsend, T. N., & Taylor, A. S. (1996). An outcome evaluation
of Across Ages: An intergenerational mentoring approach to drug prevention. Journal of
Adolescent Research, 11, 116–129.
Main, M., Kaplan, K., & Cassidy, J. (1985). Security in infancy, childhood, and adulthood: A
move to the level of representation. In I. Bretherton & E. Waters (Eds.), Growing points
of attachment theory and research (pp. 66–104). Monographs of the Society for Research
in Child Development, 50(1–2, Serial No. 209).
McLearn, K. T., Colasanto, D., & Schoen, C. (1998, June). Mentoring makes a difference: Find-
ings from The commonwealth Fund 1998 Survey of Adults Mentoring Young People. Paper
presented at the State and Future of Mentoring Symposium, Washington, DC.
Miller, L. E., Powell, G. N., & Seltzer, J. (1990). Determinants of turnover among volunteers.
Human Relations, 43, 901–917.
Olds, D., Kitzman, H., Cole, R., & Robinson, J. (1997). Theoretical formulations of a program of
home visitation for pregnant women and parents of young children. Journal of Community
Psychology, 25, 9–26.
Omoto, A.M., & Snyder,M. (1995). Sustained helping withoutobligation: Motivation, longevity
of service, and perceived attitude change among AIDS volunteers. Journal of Personality
and Social Psychology, 68, 671–686.
Penner, L. A., & Finkelstein, M. A. (1998). Dispositional and structural determinants of vol-
unteerism. Journal of Personality and Social Psychology, 74, 525–537.
Reid, P. T. (1988). Racism and sexism: Comparisons and conﬂicts. In P. A. Katz & D. A.
Taylor (Eds.), Eliminating racism (pp. 203–221). New York: Plenum.
Rhodes, J. E., Grossman, J. B., & Resch, N. L. (2000). Agents of change: Pathways through which
mentoring relationships inﬂuence adolescents academic adjustment. Child Development,
Rhodes, J. E., Haight, W. L., & Briggs, E. C. (1999). The inﬂuence of mentoring on the peer
relationships of foster youth in relative and non-relative care. Journal of Research on
Adolescence, 2, 185–202.
Sipe, C. L. (1996). Mentoring: A synthesis of P/PV’s Research: 1988–1995. Philadelphia, PA:
Sroufe, A. L. (1995). Contribution of attachment theory to developmental psychopathology. In
E. A. Carlson & A. L. Sroufe (Eds.), Developmental psychopathology: Vol. 1. Theory and
methods. New York: Plenum.
Predictors and Effects of Duration in Youth Mentoring Relationships 219
Styles, M. B., & Morrow, K. V. (1992, June). Understanding how youth and elders form relation-
ships: A study of four Linking Lifetimes programs (Research report). Philadelphia, PA:
Taylor, R. L. (1989). Black youth, role models and the social construction of identity. In
R. L. Jones (Ed.), Black adolescents. Berkeley, CA: Cobb & Henry.
Walker, G., & Freeman, M. (1996). Social change one on one: The new mentoring movement.
The American Prospect, 27, 75–81.
Wallerstein, J. S. (1988). Children of divorce: Stress and developmental tasks. In N. Garmezy &
M. Rutter (Eds.), Center for advanced study in the behavioral sciences. Inc.: Stress, coping
and development in children (pp. 265–302). Baltimore, MD: Johns Hopkins University
Wilson, J., & Musick, M. (1997). Who cares? Toward an integrated theory of volunteer work.
American Sociological Review, 62, 694–713.