Am J Crim Just (2013) 38:439–456 DOI 10.1007/s12103-012-9188-9
Exploring the Link between Mentoring Program Structure & Success Rates: Results from a National Survey J. Mitchell Miller & J. C. Barnes & Holly Ventura Miller & Layla McKinnon
Received: 4 October 2012 / Accepted: 4 October 2012 / Published online: 19 October 2012 # Southern Criminal Justice Association 2012
Abstract Though mentoring has emerged as a promising and low-cost intervention for at-risk youth in recent years, the scientific knowledge base on the topic remains under-developed. The current study augments the knowledge base on youth mentoring by analyzing programmatic elements of mentoring programs situated in or adjacent to the juvenile justice system that are predictive of participant success. Poisson regression was utilized to analyze data collected through a national mentoring community saturation survey. Findings indicated that mentoring programs that require more frequent interaction and sustain relationships for longer timeframes realize higher success rates. Similarly, the use of formal mentor training was also observed as indicative of the use of evidence based practices and higher success rates, though likely beyond the logistical and fiscal reach of some local mentoring initiatives. The implications for further research and the mentoring community are discussed. Keywords At-risk youth . Delinquency reduction . Mentoring program Mentoring entails a relationship between an older and more experienced adult and an unrelated younger mentee wherein on-going guidance, instruction, and support from the adult seeks to enhance the character and life skills of the mentee (Rhodes & DuBois, 2008). The appeal and rise of mentoring is understandable as it is a low-cost
This project was supported by Grant #2010-JU-FX-0118 awarded by the Office of Juvenile Justice and Delinquency Prevention, Office of Justice Programs, U.S. Department of Justice. The opinions, findings, and conclusions or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect those of the Department of Justice. The authors wish to thank the grant partners (MENTOR, The National Mentoring Partnership; Global Youth Justice; and the National Partnership for Juvenile Services) for input and assistance on the development of data collection instruments. We would also like to thank Barbara Tatem Kelley of the Office of Juvenile Justice and Delinquency Prevention for project guidance and direction.
J. M. Miller (*) : J. C. Barnes : H. V. Miller : L. McKinnon University of Texas, San Antonio, San Antonio, TX, USA e-mail:
[email protected]
440
Am J Crim Just (2013) 38:439–456
delinquency prevention and intervention option that capitalizes on the resources of local communities and caring individuals. Mentoring services can be customized to a wide range of needs and situations suitable for delivery in multiple forms, ranging from individual, group, peer-to-peer, cross-age, and e-mentoring orientations. Mentoring relationships have dramatically increased in recent years for youth development, generally, and particularly for at-risk youth as an unprecedented amount of federal funding for mentoring initiatives has enabled wide scale implementation of new mentoring programs and initiatives (Office of Justice Programs, 2011). There has been an ongoing commitment by the U.S. Department of Justice (USDOJ) to augment the empirical knowledge base on youth mentoring toward bolstering evidence based practices as a major form of delinquency prevention and reduction. The majority of attention to mentoring has focused on important issues such as preferred processes for successfully matching adult mentors and youth mentees, substantive modality elements, generally, and across mentoring forms, and consideration of which combination of factors and mentoring activities lead to successful outcomes. Academic disciplines such as counseling and education have been examining mentoring for over a decade and generally conclude that it is facilitative of positive youth development, but far less work has based in criminology and the criminal justice sciences. This neglect is curious given the increasing focus of the mentoring community on at-risk youth. The current study focuses on the importance of certain programmatic elements of mentoring activity within and adjacent to the juvenile justice system predictive of program success.
The History and Evolution of Youth Mentoring The proliferation of youth mentoring programs in recent years has been the subject of considerable research and discussion (DuBois, Portillo, Rhodes, Silverthorn, & Valentine, 2011). Estimates put the current number of programs and youth population served at more than 5,000 and approximately 3 million, respectively (MENTOR/ National Mentoring Partnership, 2006). Despite the widespread proliferation of these programs, there is no officially recognized definition of what constitutes a mentoring relationship. Mentoring, however, is generally characterized as a relationship wherein the growth and development of a younger protégé is fostered through instruction and support provided by an older, more experienced individual (DuBois & Karcher, 2005). Relationships can be formal and arranged through an organization that matches youth and adults or informal, naturally occurring connections such as those that develop between teacher and student. The former classification represents an estimated 30 % of all mentoring relationships (Wood & Mayo-Wilson, 2012) and is a focus of this review. While the dimensions and attributes of mentoring relationships can vary across programs and settings, the common focus or purpose is to provide positive or pro-social influence on youth development in areas where it may be lacking. This theme of youth development is evident in the developmental stages of mentoring identified by Baker and McGuire (2005) that illustrate its growth and evolution in the United States. The noticeable increase and prevalence of
Am J Crim Just (2013) 38:439–456
441
delinquent behavior that accompanied the industrialization and urbanization boom of the early 20th Century prompted creation of juvenile courts and demand for prevention and intervention efforts. Part of mentoring’s attractiveness lies in the fact that it provides a seemingly simplistic and inexpensive remedy to the problem of diverting socially and economically disadvantaged youth away from risky or delinquent behaviors (Cavell & Smith, 2005; DuBois & Karcher, 2005; Keller, 2005; McCluskey, Noller, Lamoureux, & McCluskey, 2004; Smith & Stormont, 2011). Matching of disaffected children and adolescents with caring adults who can offer emotional and social support that may be lacking at home or school is expected to counterbalance negative influences, help youth to overcome hardships, and avoid criminal involvement. Such assumptions and expectations are grounded more in faith than theory and do not consider the potential for participant characteristics, program structure and delivery, and fidelity to affect intermediate and longterm outcomes (Bogat & Liang, 2005; Newburn & Shiner, 2006; Rhodes, 2005; Rhodes & DuBois, 2008). Furthermore, the significance of adequate training, quality relationships, specified goals, and linking program processes and activities with desired outcomes can be overlooked or ignored (Bouffard & Bergseth, 2008; Keller, 2005; Nakkula & Harris, 2005; Pryce & Keller, 2011; Spencer, 2006; Spencer, 2007; Thompson & Zand, 2010; Tolan, Henry, Schoeny, & Bass, 2008). It is this absence of theoretical foundation and inattention to processes and practices that likely explains the mixed findings and positive, but limited, degree of impact documented in the evaluation literature (Coyne, Duffy, & Wandersman, 2005; DuBois & Silverthorn, 2005; DuBois et al., 2011; Grossman, Chan, Schwartz, & Rhodes, 2011; Newburn & Shiner, 2006; Rhodes & DuBois, 2008; Tolan et al., 2008; Wood & Mayo-Wilson, 2012).
Mentoring as a Delinquency Prevention/Reduction Tool Although no formal mentoring typology exists and some variation and overlap occur, interventions can be classified as site-based or community-based according to where services are delivered (Sipe, 2005). Site-based programs generally operate out of schools, faith-based organizations, and local service clubs (Dappen & Isernhagen, 2006; DuBois & Karcher, 2005) and use either paid or volunteer mentors. Activities are highly structured, may be group oriented, involve little or no interaction outside program functions, and relationships are often short-lived (Portwood & Ayers, 2005; Pryce & Keller, 2011; Smith & Stormont, 2011). Community-based national organizations such as Big Brothers Big Sisters, Boys and Girls Clubs, and United Way represent a slight majority of programs, are characterized by one-on-one mentorprotégé matches, and involve less structured off-site activities (DuBois & Karcher, 2005). Participants have voice in scheduling and the selection of activities in these relationships that tend to be longer as a minimum one-year commitment is usually recommended (DuBois et al., 2011; Portwood & Ayers, 2005). Since gaining wide acceptance as an intervention for socially and emotionally vulnerable youth, mentoring has also been enthusiastically embraced as a remedy for misconduct and delinquency among at-risk youth. At-risk is a broad classification but typically encompasses youth who, due to personal or environmental disadvantages, are more susceptible to negative life outcomes (Bouffard & Bergseth, 2008) but have
442
Am J Crim Just (2013) 38:439–456
not yet been labeled as delinquent or as an offender. Mentoring for this group is expected to function as a primary prevention or early intervention strategy to divert participants from the justice system. The term ‘system-involved’ refers to youthful offenders who may be incarcerated or under community supervision. In these settings, mentoring is utilized as a reentry or aftercare approach to reduce or prevent recidivism (Bazron, Brock, Read, & Segal, 2006; Bouffard & Bergseth, 2008; Blechman & Bopp, 2005; Enriquez, 2011). While mentoring interventions have been the focus of numerous studies and positive findings have fueled youth service, research interest, and funding of corresponding juvenile offender programs, very little research has focused on mentoring for system involved youth, specifically (DuBois et al., 2011). Mirroring outcomes with other populations, mentoring for at-risk and systeminvolved youth has generally positive but mixed effects (Bouffard & Bergseth, 2008; Dallos & Comley-Ross, 2005; Dappen & Isernhagen, 2006; DuBois et al., 2011; Enriquez, 2011; Keating, Tomishima, Foster, & Alessandri, 2002; Laakso & Nygaard, 2007; Langhout, Rhodes, & Osborne, 2004; LoSciuto, Rajala, Townsend, & Taylor, 1996; Newburn & Shiner, 2006; Thomas, Lorenzetti, & Spragins, 2011; Tolan et al., 2008; Wood & Mayo-Wilson, 2012), a fairly consistent observation even in quasi-experimental designs. Participants report overall positive experiences and benefits (Dallos & Comley-Ross, 2005; Laakso & Nygaard, 2007; Thompson & Zand, 2010) and findings indicate improved behavior and attitudes are associated with mentoring interventions (Dappen & Isernhagen, 2006; DuBois et al., 2011; Keating et al., 2002; LoSciuto et al., 1996; Thomas et al., 2011; Tolan et al., 2008; Wood & Mayo-Wilson, 2012. Single studies and meta-analyses, however, reveal a consistently muted effect size and that outcomes related to delinquency prevention and reduction vary or are rarely evaluated (DuBois, Holloway, Valentine & Cooper, 2002; DuBois et al., 2011; Enriquez Jr, 2011; Keating et al., 2002; Newburn & Shiner, 2006; Thomas et al., 2011; Tolan et al., 2008; Wood & Mayo-Wilson, 2012). Furthermore, program practices and relationships – the change agents of mentoring – receive far less scrutiny and assessment than outcomes in determining effectiveness (Dallos & Comley-Ross, 2005; Keller, 2005; Rhodes & DuBois, 2008; Spencer, 2006). Mentoring is linked with modest reductions in initiation to alcohol and illicit substances and additional use, violence, and delinquency in general (Bouffard & Bergseth, 2008; DuBois et al., 2002; DuBois et al., 2011; LoSciuto et al., 1996; Thomas et al., 2011; Tolan et al., 2008). Infrequent review of youth offending outcomes and low baseline substance use among younger adolescents, however, make accurate assessment of effectiveness difficult (DuBois et al., 2011; Thomas et al., 2011). Improved school attendance, academic performance and achievement, and the development of vocational skills provide additional examples of behavioral outcomes reflecting mentoring effectiveness (Dappen & Isernhagen, 2006; DuBois et al., 2011; Laakso & Nygaard, 2007; Langhout et al., 2004; Newburn & Shiner, 2006; Wood & Mayo-Wilson, 2012). Positive attitudinal, social, and emotional change are also associated with delinquency-focused mentoring (Bazron et al., 2006; DuBois et al., 2011; Laakso & Nygaard, 2007). Increased levels of confidence, positive outlook, and self-image have been consistently observed across multiple studies (DuBois et al., 2011; Laakso & Nygaard, 2007; Keating et al., 2002; LoSciuto
Am J Crim Just (2013) 38:439–456
443
et al., 1996; Wood & Mayo-Wilson, 2012) and qualitative findings suggest participants’ interpersonal skills and relations with family and peers are also enhanced through mentoring (Dallos & Comley-Ross, 2005; Langhout et al., 2004; Thompson & Zand, 2010). These findings demonstrate the ability of mentoring to generate positive results across multiple dimensions, including “hard” (behavioral) and “soft” (developmental) outcomes (DuBois et al., 2011). Yet, reviews indicate improvements and benefits may not be sustained long term (Rhodes & DuBois, 2008), particularly if the duration of mentoring relationships are brief (DuBois et al., 2011; Wood & Mayo-Wilson, 2012). Additionally, effect sizes tend to be small across all outcomes (DuBois et al., 2002; DuBois et al., 2011) and fail to reach significance in some cases (Wood & Mayo-Wilson, 2012). While results are largely positive regarding mentoring as a delinquency prevention strategy, some studies conclude there may have limited potential for success with certain individuals with special needs and/or co-occurring conditions, particular groups, and some contexts (DuBois et al., 2011; Enriquez, 2011; Jones-Brown & Henriquez, 1997; Keating et al., 2002; Langhout et al., 2004; Pryce & Keller, 2011; Spencer, 2007). Findings from a program targeting juvenile probationers, for example, suggest mentoring may not produce desired effects with chronic offenders, as re-arrest was three times higher for participants compared to a control group (Enriquez, 2011). Results here also signal that one-on-one mentoring may not offer any advantage over group mentoring because recidivism likelihood appears to be the same regardless of method used, an observation with strong fiscal and program design implications. Also worth noting is the fact that this study reinforces earlier findings that mentoring alone is not as successful as when supplemented with other treatments (Bouffard & Bergseth, 2008). When considered with other study outcomes highlighting lengthier multidimensional programs and more precise targeting of participants (Bouffard & Bergseth, 2008; Keating et al., 2002; LoSciuto et al., 1996), mentoring appears to produce more positive results when used as a delinquency prevention or part of a comprehensive approach rather than as a reduction strategy. Although, Jones-Brown and Henriquez (1997) and Blechman and Bopp (2005) make the observation that at-risk youth fare better with mentoring than their counterparts subjected to more punitive responses such as boot camp, or waiver to adult court, and probation. Similar to outcomes with at-risk groups, mentoring results for reentry and aftercare participants have also been inconsistent (Blechman & Bopp, 2005; Bouffard & Bergseth, 2008; Enriquez Jr, 2011) and the lack of rigorous testing has slowed evidence based practice specificiation, especially across system settings and participant populations. In a recent meta-analysis by DuBois et al. (2011), programs aimed at reducing juvenile offending were omitted due to underrepresentation and the potential for unreliable findings in this area. Mentoring also shows promise when used as a therapeutic approach in lieu of harsher treatment (Jones-Brown & Henriquez, 1997) as findings suggest positive effects are stronger and more likely as part of a comprehensive reentry program (Dubberley, 2006), but less effective for chronic offenders (Enriquez Jr, 2011). It remains unclear whether and exactly how much system-involved and high-risk youth can benefit from mentoring (Bouffard & Bergseth, 2008; Enriquez, 2011). Understanding the influence of mentoring setting,
444
Am J Crim Just (2013) 38:439–456
site location (school, community, justice system), population characteristics (risk level, needs), and the referral process (social, legal) can help contextualize findings and provide direction for future research.
Current Study The overall impact of mentoring on youth development has been positive with regards to outcomes such as improved attitudes, self-perception (LoSciuto et al., 1996), interpersonal relations, reduced truancy, dropout rates (Dondero, 1997; JonesBrown & Henriquez, 1997), and reduced levels of substance abuse (Bouffard & Bergseth, 2008; LoSciuto et al., 1996; Thomas et al., 2011; Tolan et al., 2008). These and other benefits, unfortunately, have been quite small in scale and have been shown to vary with program structure and relationship duration (DuBois et al., 2011; Enriquez, 2011; Keller, 2005; LoSciuto et al., 1996; Wood & Mayo-Wilson, 2012) as longer and better designed programs enhance positive effects of mentoring while shorter or prematurely terminated matches can have adverse consequences (Rhodes & DuBois, 2008). Given the lack of robustness and variability of mentoring effects, several concerns and limitations have emerged in the research literature. Outcomes and effectiveness differ for certain populations (DuBois et al., 2011; Enriquez, 2011; Keating et al., 2002; Smith & Stormont, 2011; Spencer, 2007; Tolan et al., 2008) and across different locations and settings (Bouffard & Bergseth, 2008; Dallos & Comley-Ross, 2005; Dappen & Isernhagen, 2006; Langhout et al., 2004; Portwood & Ayers, 2005). Additionally, research notes several problem areas practitioners and evaluators have neglected: modeling or structuring of programs (DuBois et al., 2011), delivery and implementation (Rhodes & DuBois, 2008), mentoring relationship preferred substantive activity and quality (Keller, 2005), and targeting of at-risk populations (Smith & Stormont, 2011). The following analysis of data drawn from a national survey of mentoring programs examines if and how between-program differences impact success rates for mentored youth referred from the juvenile justice system.
Methods Sampling Strategy Mentoring is a relatively new juvenile justice intervention strategy and, as such, posed several unique challenges to conducting a national survey. In order to carry out a probability sample such as a simple random or stratified random sample, it is necessary to have access to a sample frame - a list of known eligible respondents/ participants (Groves et al., 2009). In the absence of a meaningful sample frame, only cluster sampling remains as a viable probability sample option. A cluster sample entails several general guidelines, the most basic being that the researcher starts with a higher level of aggregation than realized from final survey participation. Most researchers conducting a cluster sample will begin with a list of the 50 United States and work “down” from there (i.e., randomly choosing counties within those states,
Am J Crim Just (2013) 38:439–456
445
then choosing cities within the chosen counties, and then selecting participants from those locales). Although cluster sampling addresses the ostensible problem of having no sample frame, it does rest on several assumptions that may or may not hold when the survey is for the mentoring community. Most importantly, the cluster sample assumes that all states (if that is the beginning level of aggregation) have mentoring programs at equivalent (or, at least, proportional) rates. While this assumption may hold, there is no known data source that can be referenced for confirmation. Put differently, a cluster sample poses a risk that states, counties, and cities may be selected that actually have no mentoring programs available to be studied which would lead to an increase in sampling error. In light of these issues, a targeted saturation sampling approach guided the current study. While the sampling strategy used is a non-probability sample, there were several features of the chosen design that made it the most attractive option. Primarily, the targeted saturation design ensured that mentoring programs would be contacted, that eligible participants would have the opportunity to respond, and that wide coverage of mentoring programs would be achieved. The targeted saturation sampling strategy utilized the networking resources of four agencies: Global Youth Justice (GYJ), The National Partnership for Juvenile Services (NPJS), MENTOR, and The Office of Juvenile Justice and Delinquency Prevention (OJJDP). GYJ used its organizational membership database to reach juvenile justice professionals within several primary settings, including Teen/Youth Court Diversion Programs, Delinquency and Dependency Courts, and Juvenile Probation Departments. The database included contact information for approximately 3,100 individuals in those settings. NPJS used its organizational membership database to reach individuals and facilities that fall within the juvenile detention, juvenile corrections, or juvenile probation settings. The database included contact information for approximately 1,000 individuals in those settings. MENTOR’s distribution list covers a broad list of programs and mentoring practitioners. There are close to 12,000 contacts in the list but the survey completion instructions regarding eligible respondents effected considerable attrition. Finally, OJJDP posted a call for participants to its national JuvJust listserv with thousands of practitioners representing a large number of potential eligible respondents. The final sample included 1,197 respondents. It is important to note, however, that the analytic sample sizes varied from question to question due to built in skip patterns in the survey. These analytic sample sizes varied primarily as a function of the type of program the respondent represented and as a function of missing data (i.e., item nonresponse). Because mentoring programs are not specific to one location, one region of the U.S., or one culture, a primary aim of the current study was to draw information from mentoring programs located across the U.S. and in different cultural settings. Table 1 presents statistics on the “spread” of the final sample across the 50 States. As can be seen, all 50 States were represented, as was Washington D.C. As expected, more populated states expectedly provided more respondents compared to less populated states (e.g., California provided 68 respondents while Rhode Island only provided 3). As reflected in Table 1, no region of the country was overlooked which minimizes concern that the results from any quantitative analysis will be biased toward specific regions of the country.
446
Table 1 Sample Coverage
Am J Crim Just (2013) 38:439–456
State
State
Number of respondents
Alabama
15
Nevada
Alaska
10
New Hampshire
Arizona
13
New Jersey
Arkansas
4
New Mexico
California
68
New York
58
Colorado
28
North Carolina
31
Connecticut
15
North Dakota
Delaware
4
Ohio
9 9 20 9
8 28
Florida
67
Oklahoma
Georgia
30
Oregon
17
8 26
Hawaii
9
Pennsylvania
Idaho
9
Rhode Island
3
Illinois
38
South Carolina
7
Iowa
27
South Dakota
9
Indiana
26
Tennessee
18
Kansas
13
Texas
39
Kentucky
10
Utah
Louisiana
12
Virginia
Maine Massachusetts
Although it is possible that these respondents were from Guam and Puerto Rico, it is more likely that the respondent from Guam was intending to select Georgia and the two respondents from Puerto Rico intended to select Pennsylvania
Number of respondents
2 23
Vermont
4 21 6
Washington
13 22
Maryland
16
Wisconsin
Michigan
42
West Virginia
5
Minnesota
20
Wyoming
4
Mississippi
7
Missouri
13
Montana
5
Nebraska
7
Wash. D.C.
8
Guama
1
Puerto Ricoa
2
Another, perhaps more important, indicator of sample coverage gauges the types of communities from which the respondents hailed. The majority of the respondents indicated that their program was located in an urban (54.26 %) or a suburban (18.72 %) setting. Little more than 1 % of the sampled programs were located in a Tribal setting and 25.64 % of the sampled programs were located in a rural setting. Based on these results and in light of generalizability concerns, it may be important to control for a program’s community setting when conducting multivariate analysis. Response Rate Due to the sampling strategy outlined above, typical response rates would not provide an appropriate overview of the sample’s coverage or an indication of sampling efficacy. As there is no national register of mentoring programs to constitute a sampling frame from which a probability sample could be drawn, the most
Am J Crim Just (2013) 38:439–456
447
appropriate available sampling option was to realize wide coverage of the United States mentoring community. While these limiting features preclude calculation of a conventional response rate, the completion rate is observable. The completion rate, an indicator of the success of the survey implementation strategy, is calculated by carrying out the following basic formula: Number of Respondents Who Completed Survey Completion Rate ¼ 100 Number of Respondents Who Started Survey The completion rate for this survey was 64.22 %. Given that a little more than one third of respondents did not complete started surveys, it is likely that the instrument may have been too lengthy or overly complex. Dependent Variable Mentees Meeting/Exceeding Goals Mentoring program respondents were asked the following question: “On average, what percentage of the mentees referred to your program from juvenile justice settings meet or exceed the goals set for them?” Responses were given on a five-point scale where 00fewer than 10 %, 1011 - 25 %, 2026 - 50 %, 3051 - 75 %, and 4076 - 100 %. Key Covariates Meeting Frequency The frequency with which mentors and mentees meet was gauged by one question asked to the respondent: “On average, how frequently do mentors and juvenile justice involved mentees meet?” Responses were coded so that 101–2 times a month, 203–4 times a month, and 30more than 4 times a month. Meeting Length All respondents from mentoring programs were asked to report on the length of the typical mentor-mentee meeting. Responses were coded such that 00less than one hour, 10one hour to less than two hours, 20two hours to less than three hours, and 30three hours or more. Mentor Training Respondents were asked whether their mentoring program provided special training or guidance to mentors working with youth from juvenile justice settings. Answers were coded so that 00never, 10rarely, 20sometimes, and 30always. Control Variables Background Checks Respondents were asked about the frequency with which their program performs background checks on mentors. Answers were given on a fourpoint scale where 00never, 10rarely, 20sometimes, and 30always. Individualized Mentoring Respondents from mentoring programs were given the following question: “What is the typical approach to mentoring for juvenile justice involved youth in your program? (click all that apply)” Participants were then given
448
Am J Crim Just (2013) 38:439–456
the choice between “individually based mentoring (i.e., one-to-one)”, “group-based mentoring (one mentor/multiple youth)”, “team-based mentoring (multiple mentors/ multiple youth)”, “e-mentoring (i.e., over the internet or via email)”, or “other”. Responses to this question were dichotomized where 00not individually based mentoring and 10individually based mentoring. Years in Operation Respondents were asked to report on the number of years their program had been in operation. Responses were coded in whole numbers as years ranging from 1 (1 year or less) to 21 (more than 20). Percent of Youth Who are Male Participants reported the percentage of youth referred to the mentoring program from a juvenile justice setting that were male. Responses ranged from 1 (0–15 %) to 6 (76 – 100 %). Percent of Youth Who are African-American Respondents also reported on the percentage of youth referred to mentoring from a juvenile justice setting that were black. Answers were coded on a scale ranging from 100–15 % to 6076 – 100 %. Community Type As noted above, respondents reported on the community setting of their mentoring program. The majority of all respondents’ programs operated in an urban setting. Thus, all responses were dichotomized so that 00non-urban and 10urban. Mentoring Facility Respondents were asked the following: “Where does mentoring typically take place?” This variable was dichotomized so that 00somewhere other than a designated “mentoring” facility and 10in a designated “mentoring” facility.
Analysis Plan Quantitative analysis involved two steps. The first step to the analysis utilized various descriptive statistical techniques in order to provide an overview of the sample and of certain features of mentoring that are frequently encountered. The second step to the analysis utilized two inferential statistical techniques to unpack the correlation between certain programmatic elements of mentoring programs and mentee success rates. As can be seen in Fig. 1, the majority of respondents (~60 %) represented mentoring programs. This result was expected due to the sampling strategy utilized (i.e., contact list from MENTOR was one of the primary sampling frames) and is reassuring that the sampling strategy netted information from eligible and appropriate respondents. The remainder of the respondents (~40 %) represented a juvenile justice setting (i.e., juvenile probation, juvenile detention, juvenile corrections, delinquency court, youth court/teen court diversion program, and dependency court). Because of the structure of the survey and the types of questions asked to the different respondents, the current analysis was restricted to respondents from mentoring programs (N ranged between 491 and 591 for the multivariate models). There were no differences in the representativeness of the
Am J Crim Just (2013) 38:439–456
449
Fig. 1 Respondents who completed the survey, by program type
700 630
Number of Respondents
600 500 400 328
300 200 100 0 Mentoring Program
Juvenile Justice Setting Program Type
sample when restricted to mentoring respondents (i.e., all 50 states and Washington D.C. were represented and the community context of the respondents mirrored that of the full sample).
Findings Presented in Table 2 are descriptive statistics and a series of bivariate zero-order correlation coefficients between the key variables used in the analysis. All correlations statistically significant at the p
