AN ASSESSMENT OF PROTECTIVE FACTORS IN PREDICTING JUVENILE
REOFFENDING
By
Ashlee R. Barnes

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Psychology-Doctor of Philosophy
2017

ABSTRACT
AN ASSESSMENT OF PROTECTIVE FACTORS IN PREDICTING JUVENILE
REOFFENDING
By
Ashlee R. Barnes
Juvenile court practitioners use risk assessment to evaluate level of risk and criminogenic
areas of need to determine the most appropriate consequence for young offenders, and to
determine goals for case planning. Recently, juvenile court practitioners and researchers have
gained interest in evaluating young offenders’ internal and external strengths, or protective
factors. Some scholars assert that incorporating measures of protective factors into the risk
assessment process can increase the accuracy of identifying young offenders’ odds of
recidivating. Relatively few juvenile risk assessment validation studies have evaluated the
predictive validity of protective factor items. Moreover, protective factor items that are included
in many existing risk assessment tools are narrow in scope, particularly within family-, school-,
and community-level protective factor domains.
The current study examined the relationship between protective factors and recidivism
for 278 young probationers from a Midwestern juvenile county court. The study was conducted
in two parts. First, a strengths-based measure of risk of recidivism (Protective Factors for
Reducing Juvenile Reoffending, PFRJR) was created and its factor structure and reliability was
evaluated. Second, the predictive validity, incremental validity, and differential predictive
validity of the PFRJR were examined. In the first study, the author identified two factors,
Individual/Community and Family/Social; both subscales demonstrated strong internal
consistency. In the second study, the author found no significant differences in mean level
composite protective factor scores across gender, however African American offenders had

significantly lower protective factor scores than White offenders. The PFRJR significantly
predicted recidivism and time-to-recidivism, and produced AUC effect sizes that ranged from
small to large for the total sample and across young offender subgroups. The author did not find
evidence of differential predictive validity across gender, however the author found differential
predictive validity by race/ethnicity. Regarding the incremental validity of protective factor
scores, the PFRJR composite scores did not increase the amount of variance explained in
recidivism after accounting for the variance explained by composite risk factor scores (as
measured by the Youth Level of Service/Case Management Inventory). Broadly, the current
study highlights the feasibility of integrating a complementary strengths-based measure into
traditional risk assessment procedures. Findings from the current study also contributed to the
paucity of risk assessment validation studies that emphasized the predictive validity of protective
factor scores.

Copyright by
ASHLEE R. BARNES
2017

To the young people who have lost their lives to senseless violence

v

ACKNOWLEDGMENTS

First, I want to acknowledge God for giving me the strength, discipline, and perseverance
to successfully complete this journey. I recognize that having the opportunity to earn a doctorate
degree is a privilege that I do not take for granted. I also thank God for blessing me with
extraordinary family, friends, colleagues, mentors, and community members that have supported
me in multiple ways.
Mommy, you are the strongest woman I know. You managed to raise four children
practically on your own, with very little help. Thank you for putting us first and emphasizing the
value of education. Without your love and support, none of this would have been possible. Thank
you, mommy.
Gerald and Kamran, there were many nights that I did not cook dinner, or watch our
favorite TV show, or read a bedtime story. I know there were a million other things I probably
forgot or put on the back burner to focus on my work. Thank you both for believing in me and
understanding that this process would only be temporary. Thank you for your love, support, and
encouraging me to keep going when things got hard.
To my family: April, Amber, DJ, Big C, Grandma Shirley, Grandma Ethel & Papa,
Renita, and Dr. Jasmine Lee. Thank you all for words of encouragement, babysitting, hugs, pep
talks, patience, and prayers. I love ya’ll!
To my friends: Jennifer, Ciera, Nkiru, Norida, Christina, Liz, Jasmine Abrams, Marva,
and Latonya. Each of you played a very specific role in this process. Thank you all for being
YOU. Thank you for holding me accountable, brainstorming ideas, being a shoulder to cry on,
making me laugh, offering a listening ear, and saying exactly what I needed to hear at the exact
time I needed to hear it.

vi

To friends and mentors in the MSU community and to my Lansing community partners:
Sean H. Dr. Isis S., Dr. NiCole B., Dr. Zaje H., Dr. Tony N., Dr. Julius J., Steven T., Dr.
Dagbovie, Dr. Kellie M., Dr. Ashley S., Chris S., Ivan W., Sheila B., Kristen M., Jill M., Scott
L., Sue P., Willard W. and to the AGEP learning community. Thank you all for providing social
support, professional resources, and words of wisdom. You all helped make this possible!
To my dissertation committee members: Dr. Ignacio Acevedo, Dr. NiCole Buchanan, and
Dr. Jennifer Cobbina. I humbly appreciate you taking the time to serve on my committee. Your
mentorship, guidance, and support will not be forgotten.
A final (and very special thanks) to my advisor, Dr. William Davidson. Bill, you were
more than just my academic advisor. It was clear to me from the beginning that you were not
only interested in guiding me through graduate school, but also supporting me as a whole person.
Thank you for having a healthy balance of both nurturing and challenging me. Thank you for
showing me what it means to be a community researcher by affording me the opportunity to
“learn on the job.” Thank you for encouraging me to by myself. Thank you for listening to my
candid thoughts and (probably too honest) opinions. Thank you for not only being my advisor,
but for treating me like family.

vii

TABLE OF CONTENTS
LIST OF TABLES ......................................................................................................................x
INTRODUCTION ......................................................................................................................1
Theoretical Framework: Resilience Theory..............................................................................2
STUDY 1: DEVELOPMENT OF THE PROTECTIVE FACTORS FOR REDUCING
JUVENILE REOFFENDING SCALE ........................................................................................5
Risk Assessments with Protective Factor Items ........................................................................5
Limitations of Existing Risk Assessments with Protective Factor Items ...................................6
Summary of Review ................................................................................................................7
Current Study ..........................................................................................................................8
METHOD ................................................................................................................................. 10
Setting and Sample ................................................................................................................ 10
Training and Procedures ........................................................................................................ 10
Construction of the PFRJR .................................................................................................... 11
RESULTS ................................................................................................................................. 14
Investigating Factor Structure ................................................................................................ 15
Assessing Reliability ............................................................................................................. 19
Summary of Study 1 .............................................................................................................. 19
STUDY 2: INVESTIGATING THE VALIDITY OF THE PFRJR ............................................ 21
Predictive Validity of Single Item Protective Factors ............................................................. 22
Individual-Level Single Protective Factor Items .................................................................... 23
Family-Level Single Protective Factor Items ......................................................................... 24
School-Level Single Protective Factor Items.......................................................................... 26
Community-Level Single Protective Factor Items .................................................................. 27
Predictive Validity and Incremental Validity of Composite Protective Factor Scores ............. 29
Predictive Validity by Gender ................................................................................................ 31
Predictive Validity by Race/Ethnicity .................................................................................... 32
Summary of Literature Review Findings................................................................................ 33
Current Study ........................................................................................................................ 34
METHOD ................................................................................................................................. 35
Measures ............................................................................................................................... 35
RESULTS ................................................................................................................................. 38
Descriptives: Mean Level Differences Across Subgroups ...................................................... 38
Evaluation of PFRFR Predictive Validity .............................................................................. 39
Evaluation of PFRJR Incremental Validity ............................................................................ 42
DISCUSSION ........................................................................................................................... 45
Mean Level Differences Across Race/Ethnicity and Gender .................................................. 45

viii

Predictive Validity of Single Protective Factor Items ............................................................. 46
Predictive Validity and Differential Predictive Validity of PFRJR Scores .............................. 47
Predictive and Incremental Validity of PFRJR Scores with Time-To-Recidivism .................. 47
Limitations ............................................................................................................................ 49
APPENDICES .......................................................................................................................... 52
APPENDIX A: STUDY 1 LITERATURE REVIEW TABLE ............................................... 53
APPENDIX B: PFRJR INITIAL ITEM POOL ...................................................................... 70
APPENDIX C: PFRJR ITEMS .............................................................................................. 77
APPENDIX D: YLS/CMI ITEMS ......................................................................................... 78
REFERENCES ......................................................................................................................... 79

ix

LIST OF TABLES

Table 1. Summary of Exploratory Factor Analysis Results for the PFRJR using Maximum
Likelihood Estimation (N = 278) ........................................................................................ 17
Table 2. Item-level Responses (N = 278) ................................................................................... 18
Table 3. PFRJR Composite and Subscale Means and Standard Deviations by Gender and
Race/Ethnicity ................................................................................................................... 39
Table 4. AUC Statistics for Single PFRJR Items ....................................................................... 40
Table 5. Recidivism Rates and PFRJR AUC Statistics For Full Sample and Offender Subgroups
.......................................................................................................................................... 41
Table 6. Hierarchical Cox Regression to Examine Incremental Validity of PFRJR Total Scores
.......................................................................................................................................... 44
Table 7. Summary of Literature Review of Relevant Protective Factors .................................... 54
Table 8. Summary of Protective Factors in Risk Assessment Literature Review ........................ 65
Table 9. Summary of Table 8 .................................................................................................... 68
Table 10. Initial Pool of PFRJR Items ....................................................................................... 70
Table 11. YLS/CMI Items ......................................................................................................... 78

x

INTRODUCTION
Juvenile courts should recognize the potential that strengths-based risk assessment
procedures could have on facilitating positive change (Nissen, 2006). Juvenile justice
practitioners use risk assessment tools to assess risk of recidivism and to evaluate young
offenders’ areas of criminogenic (crime causing) need (e.g., substance abuse, education, and
family circumstances). Moreover, juvenile justice practitioners use the risk and criminogenic
needs information gathered from the risk assessment process to create goals for case
management and to decide the most appropriate sanctions (e.g., probation, community service,
and incarceration) for young offenders. Traditionally, risk assessment tools are exclusively
comprised of criminogenic risk factors. Past research findings indicate that risk assessment tools
have strong reliability and are able to predict risk of recidivism with about 70% accuracy (James,
2015; Latessa & Lovins, 2010). Nevertheless, there is considerable room for improvement as risk
assessment accounts for minimal variation in recidivism predictions (Onifade et al., 2008).
Researchers have suggested that narrowly focusing on criminogenic risk factors while neglecting
protective factors may contribute to erroneous projections of violent reoffending (de Vogel, de
Vries Robbé, Ruiter, & Bouman, 2011; Miller, 2006; Rogers, 2000). Furthermore, given the
multiple disparities that pervade the juvenile justice system it is also important that researchers
understand whether the relationship between protective factors and recidivism varies across
offender subpopulations as past research findings have indicated inconsistencies in the predictive
validity of risk assessment tools for racial minority and female offenders (Shepherd, Luebbers, &
Dolan, 2013),
Incorporating evaluations of young offenders’ strengths could not only improve
recidivism predictions, but might also support justice system effectiveness by encouraging
probation officers to implement interventions that could mitigate risk factors while also

1

enhancing protective factors (Nissen, 2006). A major shortcoming of understanding the role of
protective factors in recidivism prediction models is that few validated risk assessment measures
incorporate protective factors (Loeber & Farrington, 2012). Moreover, researchers that study
these measures, rarely examine the predictive validity of the protective factor domains. It is this
exact topic that is the focus of the current research.
The goal of this dissertation is to build knowledge on the relationship between protective
factors and recidivism. The dissertation is presented in two parts: Study 1 and Study 2. Study 1
details the construction of a measure comprised exclusively of protective factors, Protective
Factors for Reducing Juvenile Reoffending (PFRJR). In this study, the factor structure and
reliability of the tool were evaluated. The purpose of Study 2 was threefold:
1) to evaluate whether single protective factor items and composite protective factor
scores predict recidivism and time-to recidivism,
2) to evaluate whether composite protective factor scores predict recidivism differently
across gender and race/ethnicity and
3) to evaluate whether incorporating protective factors into risk assessment improves
overall predictive validity.
Theoretical Framework: Resilience Theory
Resilience theory was born out of behavioral scientists’ concern that although there was a
clear understanding of the factors that contributed to psychopathology, there was a gap in
knowledge about the factors or processes that facilitated normal development, particularly in
high-risk populations (Masten, 2011). Specifically, resilience theorists are concerned with
studying individuals that thrive despite risky conditions (Bleuler, 1984; Zolkoski & Bullock,
2012). Over the past few decades, theories of resilience have gained popularity across many

2

disciplines however, some believe that the lack of congruence on definitions, central
terminology, and model conceptualization have limited the growth of the field (Fergus &
Zimmerman, 2005; Luthar, Cicchetti, & Becker, 2000). Much of the debate around definitions of
resilience are concerned with identifying the construct as a static, individual characteristic,
versus a process of positive adaptation (Fergus & Zimmerman, 2005; Fougere & Daffern, 2011).
Some scholars believe that defining resilience exclusively as an individual trait fails to capture
the importance of contextual factors that promote individual resilience, in addition to implying
that individuals without a resilient personality do not have the ability to succeed in the face of
adversity (Fergus & Zimmerman, 2005; Luthar et al., 2000).
Despite variations in central terminology (resources, strengths, assets, protective factors,
promotive factors, buffering factors, etc.), there is a consensus around three models of resilience:
compensatory, protective, and challenge (see Fergus & Zimmerman, 2005; Garmezy et al., 1984
for a detailed explication). Every model of resilience includes one or more risk and protective
factor(s), (Fougere & Daffern, 2011; Masten, 2011; Zolkoski & Bullock, 2012). Risk factors are
social and environmental characteristics that increase the probability of a negative outcome
(Fergus & Zimmerman, 2005). Protective factors are attributes of the individual and the
environment that serve to buffer the impact of risk, thereby increasing the probability of a
positive outcome (Rutter, 1985).
A shift in focus from exclusively reducing risk to also promoting prosocial behavior has
recently gained attention in criminology (Chapman et al., 2006; Farrington, 2007; Fougere &
Daffern, 2011; Lodewijks et al., 2010). Resilience theory could enhance criminological theories
of delinquency as it highlights the positive resources of those exposed to criminogenic risk and
aims to understand the mechanisms that delays the onset of delinquency or facilitates desistance

3

(complete termination of engaging in crime/delinquency) (Walker et al., 2013). In the following
section, a brief discussion of risk assessment tools that contain protective factors will be
presented.

4

STUDY 1: DEVELOPMENT OF THE PROTECTIVE FACTORS FOR REDUCING
JUVENILE REOFFENDING SCALE
It is well documented that only a small percentage of young offenders are responsible for
committing the majority of new crimes (Cottle, Lee, & Heilbrun, 2001; Milligan, 2010).
Consequently, justice system practitioners aim to identify strategies that enable them to
recognize those most likely to become repeat offenders; risk assessment is one such strategy.
Risk assessment tools are comprised of criminogenic risk factors that assess risk of recidivism,
specify areas of need that provide guidelines for case planning, and standardize decision-making
processes (Schwalbe, 2007). Using risk assessment to estimate level of risk to commit a new
crime can be beneficial for juvenile justice practitioners, as accurate assessments of risk support
the allocation of intensive services (e.g., residential placement) for individuals most likely to
recidivate and who therefore, require increased supervision (Meyers & Schmidt, 2008; Schwalbe
et al., 2004). Similarly, young offenders deemed unlikely to reoffend also receive treatment
options (e.g., community-based programming, diversion) that are aligned with their level of risk
(Krysick & LeCroy, 2002). Historically, risk assessment research has focused on youth at highest
risk of reoffending, with little attention given to youth who either posses high levels of strengths,
or youth who do not engage in future delinquency (Loeber & Farrington, 2012).
Risk Assessments with Protective Factor Items
Measuring young offenders’ strengths has gained attention in the juvenile justice system
only in the past few decades (Loeber & Farrington, 2012); therefore, risk assessment measures
that contain protective factors have rarely been investigated. The author used ProQuest to search
for peer-reviewed risk assessment studies (Key terms: juvenile risk assessment and protective
factors, predictive validity and protective factors, juvenile risk assessment validation and
protective factors, juvenile risk assessment and buffering protective factors, juvenile risk

5

assessment and compensatory factors, juvenile risk assessment and strengths, juvenile risk
assessment and assets, juvenile risk assessment and resources, juvenile risk assessment and
promotive factors, risk assessment and resilience) and found that few (31 out of 608) risk
assessment studies investigated the validity of protective factor scores. Within the 31 studies, the
following juvenile risk assessment tools were represented: the Positive Achivement Change Tool
(PACT; Winokur-Early, Hand, & Blakenship, 2012), the San Diego Risk and Resiliency
Assessment (SDRRC; Turner & Fain, 2007), the Santa Barbara Assets and Risk Assessment
(SBARA; Jimerson, Sharkey, O’Brien, & Furlong, 2003), the Strengths Assessment InventoryYouth (SAI-Y; Rawana & Brownlee, 2010), the Structured Assessment of Violence Risk in
Youth (SAVRY; Borum, Bartel, & Forth, 2000), the Washington State Juvenile Court
Assessment (WSJCA; Washington State Institute for Public Policy, 2000), and the Youth
Assessment and Screening Instrument (YASI; Orbis Partners, 2000). Using resilience theory as a
framework, the author aimed to address some of the limitations of the aforementioned measures
with the development of a strengths-based tool that evaluates risk of recidivism, the Protective
Factors for Reducing Juvenile Reoffending scale.
Limitations of Existing Risk Assessments with Protective Factor Items
The first limitation of existing risk assessment tools is that the most consistent findings
are based on youth who committed violent offenses. The SAVRY was the most investigated
measure in the current review as 14 of 31 studies used this tool to evaluate risk of recidivism.
The SAVRY is also the only published risk assessment that measures risk of violence for young
offenders; therefore, it was not surprising to find that 93% of the research that examined
protective factors with violent offenders used this tool. Conversely, there was much more
variation in the number of tools used to evaluate the relationship between protective factors and

6

recidivism for young offenders who committed non-violent offenses. The SAVRY is a valid and
reliable measure and yielded consistent findings; however, it was designed to exclusively
evaluate risk of violence, leaving a gap in our knowledge concerning the predictive validity of
protective factors for non-violent offenders. The author aimed to fill a gap in the literature by
developing a strengths-based tool, the Protective Factor for Reducing Juvenile Reoffending
(PFRJR), designed to measure general recidivism for youth who do did not commit violent
offenses.
The aforementioned risk assessment measures contain protective factors that span across
multiple ecological domains (i.e., individual, family, school, and community), however the most
notable gap is that single protective factor items being measured in the community domain are
narrow in scope. Specifically, the PACT, SAVRY, and SAI-Y exclusively measure prosocial
involvement in organized activities. The WSJCA and the SDRRC measure both prosocial
involvement and support outside the family. Finally, in addition to adult role models and use of
free time, the SBARA was the only measure that included an item on neighborhood crime. The
current study aimed to build knowledge on the importance of protective factors for young
offenders by developing a tool that encompasses community-level protective factors items that
are rarely examined (e.g., involvement in organized activities, positive adults in the community,
perceived safety, and access to resources).
Summary of Review
Although growing, the topic of protective factors in risk assessment is largely under
researched (Fougere & Daffern, 2011). There are few studies that examined the predictive
validity of protective factors and almost half of these studies used the SAVRY to evaluate risk of
recidivism. The SAVRY is a valid and reliable measure; however, it was designed to evaluate

7

risk of violence, leaving a gap in knowledge concerning the validity of protective factors for
youth who commit non-violent offenses. It is also apparent that risk assessment tools are
measuring protective factors across similar domains (individual, family, school, and community),
yet there is an under emphasis on community-level protective factors. For example, communitylevel protective factors primarily included measures of prosocial involvement and relationships
with adults outside of the family, and fell short in capturing ecological variables found to be
associated with desistance (complete termination of engaging in crime/delinquency) such as
access to resources/services (Molnar, Cerda, Roberts, Buka, 2008) or perceived neighborhood
safety (Reingle et al., 2013).
It is also important to note that existing risk assessment tools do not include protective
factor items that delinquency prevention scholars have found to be associated with recidivism
such as positive relationships with teachers, religiosity, and access to prosocial community
resources (e.g., Hartman, Turner, Daigle, Exum, Cullen, 2009; Krohn et al., 2014; Molnar et al.,
2008). Therefore, the current study sought to build knowledge on the importance of protective
factors for young offenders by developing a tool that encompasses community-level protective
factors items that are rarely examined.
Current Study
Incorporating protective factors into risk assessment has the potential to provide a more
comprehensive evaluation of risk of recidivism and may encourage practitioners to approach
case management from a strength-based perspective (Nissen, 2006). The goal of the current
study was to build knowledge on protective factors that reduce juvenile reoffending, and to meet
the needs of juvenile justice practitioners who desired to respond more effectively to young
offenders’ internal and external strengths. The Protective Factors for Reducing Juvenile

8

Reoffending (PFRJR) measure was developed to contribute to the small number of strengthsbased tools designed to assess risk of recidivism. The current study:
1) described how the PFRJR items were selected; and
2) examined the factor structure and reliability of the measure.

9

METHOD
Setting and Sample
The current study took place in a medium sized, Midwestern juvenile court. The current
study included youth who were involved in the Truancy and Delinquency divisions. Youth in the
Truancy division were referred to the juvenile court for being truant from school. The school
district officials that refer youth to court define truancy as being absent from school for 10 hours
or more per academic semester. Truant youth are assigned a juvenile court officer who is
responsible for holding them accountable for attending school. Youth involved in the
Delinquency division have been adjudicated for committing a criminal act or status offense. In
addition to being assigned a judge and a juvenile court officer, delinquent youth may also be
required to complete one or more court-supervised programs. The data excluded duplicate cases
and had no refusals. The sample (N = 278) is comprised of 75 youth who were supervised in the
Truancy division and 203 youth who were supervised in the Delinquency division. The average
age for the total sample is 15.37 years. Thirty-three percent of the sample is female (n = 93). The
racial/ethnic composition of the sample is as follows: 41.9% African American; 27.4%
Caucasian; 10.1% Hispanic/Latinx; 19.9% Multi-racial.
Training and Procedures
Juvenile court officers administered the risk assessment and protective factor measures
through semi-structured interviews for every offender on their caseloads. Prior to using the
instruments, juvenile courts officers received 40 hours of training on how to administer and score
the assessments. Training activities included providing definitions, clarifying the protocol and
scoring guide, explaining what each item measures, mock interviews, and coding previously
taped cases. Inter-rater reliability for the PFRJR and YLS/CMI was conducted in November

10

2015 and reached 94% exact agreement. Specifically, of the 63 total items there was 100% itemlevel agreement (e.g., agreed on the absence of risk and protective factor items) for 47 items
across 9 raters. For the remaining 16 items, item-level agreement ranged from 56% -89%.
Overall, eight out of nine raters agreed on how to score the youth for seven items. Seven out of
nine raters agreed on how to score the youth for four items. Six out of nine raters agreed on how
to score the youth on three items. Five out of nine raters agreed on how to score the youth for
two items.
Construction of the PFRJR
Items for the Protective Factors for Reducing Juvenile Reoffending (PFRJR) were
derived from multiple sources including a review of the delinquency prevention and juvenile
offending literature, existing resilience and risk assessment measures, and expert knowledge
from juvenile justice practitioners. The author searched electronic databases (ProQuest and
Google Scholar) using search terms that included: protective factors and delinquency, strengths
and delinquency, assets and delinquency, risk assessment and protective factors, buffering
protective factors, promotive factors, compensatory factors, juvenile offenders and resilience,
desistance, protective factors and neighborhoods, female offenders and protective factors,
culture and protective factors. The initial search produced more than 100 peer-reviewed
published manuscripts that were used to inform the selection of items for the PFRJR. Based on
findings from the review, 74 protective factors items were identified. For a list of the studies
reviewed and the initial item pool, please see Appendices A and B, respectively.
Protective factor items from existing risk assessment instruments were also reviewed,
including: the Positive Achivement Change Tool (PACT; Winokur-Early, Hand, & Blakenship,
2012), the San Diego Risk and Resiliency Assessment (SDRRC; Turner & Fain, 2007), the Santa

11

Barbara Assets and Risk Assessment (SBARA; Jimerson, Sharkey, O’Brien, & Furlong, 2003),
the Strengths Assessment Inventory-Youth (SAI-Y; Rawana & Brownlee, 2010), the Structured
Assessment of Violence Risk in Youth (SAVRY; Borum, Bartel, & Forth, 2000), the
Washington State Juvenile Court Assessment (WSJCA; Washington State Institute for Public
Policy, 2000), and the Youth Assessment and Screening Instrument (YASI; Orbis Partners,
2000). Many of the protective factor items from these tools overlapped with the items identified
in literature review of studies that did not employ actuarial measures of recidivism risk.
The juvenile justice practitioners who contributed to the development of the PFRJR were
employed at the juvenile court of interest and included the Deputy Court Administrator (38 years
experience), the Director of Juvenile Services (20 years experience), the Director of Juvenile
Programs (14 years experience), and the Intensive Neglect Services Assistant Administrator (28
years experience). The juvenile justice practitioners were highly involved in the process of
generating items, as well as providing feedback on the practical relevance, phrasing, and
response format of the initial list of items produced based on findings from the literature and
existing measures. For example, the literature review revealed the buffering effects of high
academic achievement in core subjects (e.g., math) on criminogenic risk (for example, see van
Domburgh et al., 2009). The justice practitioners were concerned that very few youth would be
high achievers, restricting the variation in the data; therefore, they suggested an item be included
that would measure whether a youth is passing or failing their core subjects.
The experts also provided feedback on the content and phrasing of the semi-structured
interview protocol questions used to score the PFRJR items. For instance, the literature revealed
the protective effects of having never used drugs/alcohol (for example, see Herrenkohl, Lee, &
Hawkins, 2012). The original question, “Have you ever used drugs/alcohol?” appeared to lack

12

important contextual information about whether a youth is actively abstaining from drugs,
especially given that experimenting with drugs /alcohol at least once was stated to be commonly
reported. Therefore, the experts recommended that the item should measure whether a youth is
abstaining from drug use. They suggested additional probes concerning whether a youth can
identify their triggers and the strategies they use to actively abstain as a way to differentiate those
who are choosing to be drug free from those who are avoiding drugs/alcohol only until they were
released from probation. This collaborative process took place during bi-weekly meetings that
occurred between November 2014 and October 2015 (28 total meetings). Appendix B provides
information on the juvenile justice practitioners’ role in the generation of items.
A senior juvenile court officer (16 years experience) piloted the PFRJR with a newly
adjudicated offender. The author, two graduate students (of criminal justice and psychology), a
juvenile justice researcher (47 years experience), and 19 juvenile court officers (254 years
combined experience, 13.4 average years experience) reviewed a de-identified audio recording
of the semi-structured interview to evaluate the face and content validity of the tool and to
identify areas where more effective probing and adjustments were necessary.
As a result of this extensive collaborative process, the initial item pool of 74 items was
reduced to a final measure consisting of 22 binary items (0 indicating the absence of protection,
1 indicating the presence of protection) that measure internal, interpersonal, and external
characteristics that are theoretically and empirically related to recidivism. The PFRJR is an
additive scale with scores ranging from 0 to 22, in which higher scores indicate higher protective
factor levels.

13

RESULTS
EFA (Exploratory Factor Analysis) was used to identify the number of latent constructs
and the underlying factor structure of the PFRJR. Although (PCA) Principal Components
Analysis (PCA) is the default method of factor analysis for many statistical packages, EFA or
common factor analysis is recommended because it aims to reveal any latent variables that cause
items to covary and unlike PCA, it solely analyzes the shared variance among items to reveal the
underlying factor structure (Costello & Osborne, 2005; Floyd & Widaman, 1995; Netemeyer,
Bearden, & Sharma, 2003). Moreover, EFA is the preferred approach as it produces results that
are more applicable to a CFA (Confirmatory Factor Analysis) (Floyd & Widaman, 1995). The
factors were extracted using the Maximum likelihood method. Maximum likelihood is the best
choice as it allows statistical significance testing of factor loadings, correlations among factors,
and calculation of confidence intervals (Costello & Osborne, 2005; Fabrigar, Wegener,
MacCallum, & Strahan, 1999). A parallel analysis was conducted to help identify the number of
factors that should be extracted in the EFA. Parallel analysis uses adjusted correlation matrices
and random normal data generation to produce principal components eigenvalues (Buja &
Eyuboglu, 1992). In parallel analysis, eigenvalues from the generated data are statistically
compared to the real data eigenvalues to determine whether the eigenvalues from the real data
were yielded beyond chance. Several rotational methods were implemented to produce the
simplest factor structure, facilitating the cleanest interpretation of the extracted factors
(Netemeyer et al., 2003). As recommended, factor loadings above .40 were retained (Costello &
Osborne, 2005; Floyd & Widaman, 1995; Netemeyer, Bearden, & Sharma, 2003).
To evaluate the reliability of the PFRJR, Cronbach’s alpha coefficient, inter-item
correlations, and corrected item correlations were calculated. The goal was to obtain a widely
accepted alpha coefficient of .80 or higher, an average inter-item correlation that falls between

14

.15 and .50 (Clark & Watson, 1995), and corrected item total correlations that are least .50.
(Bearden & Netemeyer, 1998).
Investigating Factor Structure
Prior to running the factor analysis, a number of steps were taken to ensure that the data
was suitable for this statistical procedure. Bartlett’s test of sphericity was used to examine the
observed correlation matrix and to identify whether there is enough redundancy between
variables that the data can be summarized into one or more factors (Bartlett, 1951). The results
demonstrated a significant chi-square (χ2 (231) = 1907.79, p < .05) indicating that the data
correlation matrix was sufficient to factor analyze (Bartlett, 1951). Next, the KMO (KaiserMayer-Olkin) index was used to evaluate sampling adequacy by testing whether the proportion
of variance among the variables in the data might be common variance. KMO values range from
0 to 1 and values above 0.6 indicate the sample is adequate (Cerny & Kaiser, 1977). The KMO
test produced a value of .88. Finally, based on recommendations from the literature, the current
study’s total sample size (e.g., minimum of 100) and participant to item ratio (5 to 10 per items)
was deemed more than appropriate for analysis (Clark & Watson, 1995; Floyd & Widaman,
1995).
Initial eigenvalues indicated a five-factor solution in which the factors explained 21.9%,
10%, 4.6%, 4.3%, and 2.8% of the variance, respectively. In this solution, the first factor
produced an eigenvalue of 6.5, while the remaining factors had eigenvalues slightly larger than
1. Based on the structure of the scree plot, the results of the parallel analysis, and the small
amount of variance explained by each additional factor, a five-factor solution did not seem
appropriate therefore, a two-factor solution was examined using a orthogonal varimax, oblique
oblimin rotation as well as an oblique promax rotation. In the two-factor solution, the factors

15

explained 26.9% and 3.9% of the variance, respectively. The Pattern and Structure matrices for
this solution produced the simplest structure when rotated using the promax rotation, as the
factors were allowed to correlate with each other. A total of seven items (High Achievement,
Passing, Positive Relationships with Teachers, Actively Abstaining from Drugs, Low
Availability of Drugs, Actively Seeking Help and Consistent Supervision) were eliminated
because they failed to meet the minimum criteria of having a primary factor loading of .40 or
above (Costello & Osborne, 2005; Floyd & Widaman, 1995; Netemeyer, Bearden, & Sharma,
2003).
For the final stage, an EFA of the remaining 15 items using promax and varimax
rotations was conducted in a model extracting two factors. In this solution, the two factors
accounted for 43% of the total variance. The promax rotation provided the best-defined factor
structure in which there were no cross loadings and the primary loadings were above .70. The
factor loadings, eigenvalues, communalities, and percentage of total variance for this final
solution is presented in Table 1. Factor 1 represented items that theoretical align with what could
be described as Individual and Community domains (Strong Social Skills, Prosocial Attitudes,
Positive Adults, Access to Resources, Positive Response to Authority, Perceived Safety, and
Low Aggression). Factor 2 represented items that theoretical align with what could be described
as Family and Social domains (Strong Family Management, Consistent Parenting, Positive
Personal Interests, Commitment to School/Education, Strong Adult Bonds, Involvement in
Organized Activities, Close Bonds with Positive Peers, and Religiosity). The Factor Correlation
Matrix demonstrated a strong relationship (r = .66, p < .05) between the two factors. Composite
scores were created for each subscale based on the total number of items within each domain.

16

Table 1. Summary of Exploratory Factor Analysis Results for the PFRJR using Maximum
Likelihood Estimation (N = 278)
Item

Individual/

Family/

Community

Social

Communalities

Strong Social Skills

.79

.53

Prosocial Attitudes

.72

.53

Positive Adults

.63

.44

Access to Resources

.59

.31

Positive Response to Authority

.56

.49

Perceived Safety

.55

.25

Low Aggression

.43

.24

Strong Family Management

.74

.51

Consistent Parenting

.57

.42

Positive Personal Interests

.56

.32

Commitment to School/Education

.53

.37

Strong Adult Bond(s)

.46

.21

Involvement in Organized Activities

.46

.21

Close Bonds with Positive Peer(s)

.44

.24

Religiosity

.43

.12

Eigenvalue

4.99

1.52

% of total variance

33.27

10.17

Table 2 illustrates the proportion of youth who endorsed each perspective item. As
illustrated, item-level endorsement of protective factors ranges from 21.6% to 63.3% for the

17

Individual/Community domain, and 15.5% to 54.3% for the Family/Social domain. A paired
samples t-test yielded significant findings (t(277) = -4.29, p < .05, Cohen’s d = .24), indicating
that youth in the current sample had higher average protective factor scores in the Family/Social
domain (M = 3.07, SD = 2.32) than the Individual/Community domain (M = 2.52, SD = 2.17).
Table 2. Item-level Responses (N = 278)
Item

% of Sample Endorsing the Item

Individual/Community
Strong Social Skills

21.6%

Prosocial Attitudes

30.2%

Positive Adults

48.9%

Access to Resources

57.9%

Positive Response to Authority

37.1%

Perceived Safety

63.3%

Low Aggression

37.4%

Family/Social
Strong Family Management

21.6%

Consistent Parenting

22.3%

Positive Personal Interests

48.9%

Commitment to School/Education

42.4%

Strong Adult Bond(s)

54.3%

Involvement in Organized Activities

27.3%

Close Bonds with Positive Peer(s)

19.8%

Religiosity

15.5%

18

Assessing Reliability
Inter-item correlations for the 15-item tool ranged between .04 and .59 with an average
inter-item correlation of .27. The average inter-item correlation falls within the range of
acceptable correlations for a tool that measures a broadly defined construct (Clark & Watson,
1995). Corrected item correlations were also calculated and ranged between .23 and .63. The
Cronbach’s alpha coefficients for the Individual/Community (7 items, α = .82) and Family/Social
subscales (8 items, α = .76) demonstrated strong internal consistency. Eliminating items did not
result in any substantial increases in the alpha coefficient.
Summary of Study 1
Study 1 described the development of a protective factor assessment, Protective Factors
for Reducing Juvenile Reoffending (PFRJR). Overall, the PFRJR is a multidimensional measure
and the domains within the underlying constructs are correlated. Findings from an Exploratory
Factor Analysis with a promax rotation revealed a simple two-factor structure for 15 out of the
22 items. Both factors were well defined and demonstrated acceptable internal consistency. Items
that measured school- and substance abuse-related protective factors were not included in the
final solution given their insufficient factor loadings. Resilience theory proposes there are four
protective factor domains (individual, family, school, and community); findings from the current
study conveyed that Individual and Community indicators, and Family and Social indicators
shared underlying constructs within desistance (complete termination of engaging in
crime/delinquency) from offending. Other studies that have measured substance abuse- and
education-related protective factors have found that they are related to recidivism (e.g., van der
Put et al., 2014; Shepherd et al., 2016), so it may be worth retaining the eliminated items for case
management purposes. Given the rise in popularity of non-traditional academic institutions (e.g.,

19

online learning, alternative education program, in-person computerized instruction), anecdotal
information from juvenile justice practitioners indicated the challenges of identifying the
academic progress of youth who attend these schools as they typically use a credit system and
not the more traditional grading system described in the interview protocol. Theoretically, school
is a salient institution for adolescents so these items may have contributed to the factor structure
if the items were written in a way that was more relevant to young offender populations.
Therefore, school-related items that were removed (Passing, High Achievement, Positive
Relationship with Teacher) will be rewritten in an attempt to strengthen the factor structure of
the PFRJR. An important limitation of Study 1 is the inability to identify whether the factor
structure is invariant across young offender subpopulations. Future studies will include
conducting a CFA (Confirmatory Factor Analysis) to identify the level of fit of the 2-factor
structure as well as examine measurement invariance across gender, race, and ethnicity.
In addition to describing the theoretical framework for choosing items, identifying the
factor structure of the underlying construct, and establishing reliability, it is important to also
evaluate the validity of a newly developed scale; this was the goal for Study 2.

20

STUDY 2: INVESTIGATING THE VALIDITY OF THE PFRJR
It is important to understand the state of the risk assessment literature regarding the
predictive validity, differential predictive validity (across race/ethnicity and gender), and
incremental validity of protective factors with young offenders. The goal of the literature review
was to identify studies that investigated the predictive validity of juvenile risk assessment tools
that include protective factors. Relevant studies were identified through several avenues.
Electronic databases were searched (ProQuest, Google Scholar, etc.) for peer-reviewed studies
using search terms that included: risk assessment and protective factors, risk assessment and
strengths, risk assessment and assets, risk assessment and buffering factors, risk assessment and
promotive factors, risk assessment and compensatory factors. Studies that used risk assessment
tools specifically designed to measure risk of sexual recidivism were outside of the scope of the
current study and were excluded as there may be offense-specific risk factors for youth who
sexually offend (van der Put & Asscher, 2015b). Studies that examined the predictive validity of
protective factors exclusively with youth that committed sexual offenses were also excluded. The
search returned 608 studies published between 2004 and 2016. Of the 608 studies, 52 studies met
the following search criteria:


Risk assessment validation study



Juvenile risk assessment tool with protective factors



Juvenile offender sample



Exclusion of studies that exclusively investigated sex offenders.
After careful examiniation of these results, the author excluded 29 studies as they did not

specifically examine the predictive validity of the protective factor items included in the risk
assessment measure, resulting in 23 articles. Next, the reference sections of the 23 articles were

21

examined to capture any relevant literature that may have been overlooked. An additional eight
studies were identified, resulting in a final total of 31 published articles in the current review.
In the section that follows, research that examined the predictive validity of single item
protective factors within individual, family, school, and community domains is presented.
Thereafter, research that investigated the predictive validity and incremental validity of
composite protective factor scores is described. The final sections of the literature review
highlights research that examined the differential predictive validity of protective factors across
race/ethnicity and gender. Finally, a summary of the research findings and details about how the
current study filled gaps in the literature is presented.
Predictive Validity of Single Item Protective Factors
Protective factors are defined as “aspects of an individual and their situation that
contribute to a decreased likelihood of criminal behavior by having a direct effect on problem
behaviors or by moderating the relationship between risk factors and criminal behavior”
(Fougere & Daffern, 2011, p. 245). A youth’s “situation” can be interpreted as including every
level of their social environment from family dynamics and peer relationships to neighborhood
safety. Following the lead of resilience pioneers, scholars have defined, understood, and
measured protective factors by grouping them into ecologically-relevant domains: individual,
family, school, and community (Garmezy, 1985; Krohn, Lizotte, Bushway, Schmidt, Phillips,
2014; Losel & Farrington, 2012; Pollard, 1999; Smith, Lizotte, Thornberry, & Krohn, 1995;
Zolkoski & Bullock, 2012). A summary of the predictive validity of single item protective
factors is described using this framework.

22

Individual-Level Single Protective Factor Items
Individual-level protective factors are characteristics associated with an offender’s
internal or personal resources. Researchers have identified multiple internal protective factors
related to non-offending such as emotional control (Stattin, Romelsjo, & Stenbacka, 1997), selfregulation (Gardner, Dishion, & Connell, 2008), social competence (Stepp, Pardini, Loeber, &
Morris, 2011), resilient personality (Fougere, Daffern, & Thomas, 2012; Mowder, Cummings, &
McKinney, 2010; Rennie & Dolan, 2010), and prosocial attitudes (Loeber et al., 1991; Pardini,
Loeber, Farrington, & Stouthamer-Loeber, 2012; Stouthamer-Loeber et al., 2002). Of the studies
reviewed, resilient personality and prosocial attitudes were the most investigated individual-level
protective factors.
Resilient personality has been attributed to overcoming adversity; therefore, researchers
have examined the influence of this personality type (Fougere & Daffern, 2011). At least five
studies have examined the relationship between resilient personality and offending behavior
using a popular juvenile risk assessment, Structured Assessment of Violence Risk in Youth
(SAVRY) (Gamelgård et al., 2015; Lodewijks, Doreleijers, Ruiter, & Borum, 2008; Lodewijks et
al., 2010; Rennie & Dolan, 2010; Shepherd et al., 2016). Researchers assessed resilient
personality as a predictor of violence at different stages of the judicial process including pre-trial,
during residential placement, following release of a residential facility (Lodewijks et al., 2010),
and during incarceration (Gamelgård et al., 2015; Lodewijks et al., 2008; Rennie & Dolan, 2010;
Shepherd et al., 2016). With the exception of one study (Rennie & Dolan, 2010), past research
findings conveyed that resilient personality did not predict recidivism. Resilient personality has
been deemed an important protective factor in the resilience literature (Fougere & Daffern, 2011)

23

however; research that has examined this protective factor with young offender samples does not
support this assertion.
Given that having criminal attitudes is one of the strongest predictors of engaging in
delinquency (Andrews et al., 2012), it is not surprising that researchers have identified prosocial
attitudes and beliefs as a protective factor associated with recidivism (e.g., Gamelgård et al.,
2015). Prosocial attitudes include respecting authority, believing in the legitimacy of having laws
and following rules, and recognizing the importance of seeking help (Ayers et al., 1999; Chui &
Chan, 2012). Of the six studies that investigated the predictive validity of prosocial attitudes
(Gamelgård et al., 2015; Lodewijks et al., 2008; Lodewijks et al., 2010; Rennie & Dolan, 2010;
Shepherd et al., 2016; van der put et al., 2015b), one did not find prosocial attitudes as a
predictor of violence (Rennie & Dolan, 2010). Therefore, based on the existing literature,
prosocial attitudes appear to be an important predictor of recidivism.
Past research using non-offender samples has also identified other internal protective
factors related to delinquency such as impulse control (Williams, Ryan, Davis-Kean, McLoyd, &
Schulenberg, 2014), refusing drugs (Cuervo & Villanueva, 2015; Lee & Villagrana, 2015), low
aggression (van der Put et al., 2015b), and social skills (Jones et al., 2016; Thompson & Pope,
2005; van der Put et al., 2015b). The current study contributed to the risk assessment literature
by evaluating the predictive validity of under researched (or never examined) individual-level
protective factors (e.g., positive personal interests, actively abstaining from drugs, strong social
skills, and low aggression).
Family-Level Single Protective Factor Items
Resilience scholars and criminologists consistently acknowledge that critical protective
factors are promoted within the family environment (for example, Hircshi, 1969; Moffitt, 1993;

24

Hawkins et al., 1985; Sampson & Laub, 1993). Researchers have identified parental discipline
(Jimerson et al., 2004), monitoring (Williams et al, 2014), parenting practices (Cuervo &
Villanueva, 2015), and social/emotional support (Gamelgård et al., 2015; Jimerson et al., 2004;
Lodewijks et al., 2008; 2010; Shepherd et al., 2016) as protective factors related to desistance.
Past research has also identified the important role that parent-child relationships play in
reducing future delinquency in community-based non-offender samples (Chui & Chan, 2012;
Fagan, Van Horn, Hawkins, & Arthur, 2007; Hardaway, McLoyd, & Wood, 2012; Ryan et al.,
2008; Smith et al., 1995; Youngblade et al., 2007); however, there is less support for the
importance of strong parental bonds with young offenders. For example, of the six studies
investigating this relationship (Gamelgård et al., 2015; Jimerson et al., 2004; Lodewijks et al.,
2008; 2010; Shepherd et al., 2016), only two demonstrated that strong attachments and bonds
with family are predictors of recidivism (Lodewijks et al., 2008; 2010).
Few studies from the current review specifically investigated the relationship between
parental monitoring, discipline and recidivism therefore it deserves further investigation. For
example, Jimerson and colleagues (2004a; 2004b) used the Santa Barbara Assets and Risk
Assessment to investigate convergent, construct, and predictive validity with a sample of firsttime offenders. In addition to establishing composite score predictive validity, logistic regression
analyses revealed that fair and consistent discipline was a predictor of recidivism for both male
and female youth. In another study, Williams et al. (2014) used the Washington State Juvenile
Court Assessment (WSJCA) to investigate the relationship between protective factors and
recidivism in a sample of African American probationers. Forty-four percent of the youth
reported good parental supervision; and this protective factor reduced the probability of re-arrest
by 27% (Williams et al., 2014).

25

Collectively, the limited and ambiguous findings within this domain emphasize that
scholars should conduct research that aims to support—or clarify—the theoretical and empirical
importance of familial protective factors for young offenders. The current study contributed to
the risk assessment literature by evaluating the predictive validity of minimally examined familylevel protective factors (e.g., consistent supervision and parenting, parental bonds).
School-Level Single Protective Factor Items
Children and adolescents spend the majority of their daily lives in academic settings
hence, it is not surprising that past research has identified protective factors in the educational
context that contribute to reducing delinquency (Ayers et al., 1999; Baglvio, 2014; Crosnoe,
Erickson, & Dornbusch, 2002; Bernat, Oakes, Pettingell, & Resnick, 2002; Reingle, Jennings,
Lynne-Landsman, Cottler, & Maldonado-Molina, 2013). For example, past research has used
risk assessment to examine the protective functions of high academic achievement (Cuervo and
Villanueva, 2015), relationships with positive peers, and (Cuervo & Villanueva, 2015; van der
Put et al., 2015b; Williams et al., 2014), commitment to education (Gamelgård et al., 2015;
Lodewijks et al., 2008; 2010, Rennie & Dolan, 2010; van der Put, Creemers, & Hoeve, 2014).
One of the most apparent protective factors in the school context is academic
achievement. Although studies using non-offender samples have extensively measured the
relationship between high academic achievement and delinquency (e.g., Baglivio, Jackowski,
Greenwald, & Howell, 2014; Bernat et al., 2012), no studies from the current review investigated
the predictive validity of academic achievement; therefore it is not clear whether this
characteristic serves as a protective factor for this population.
Researchers have suggested that positive peer groups are important predictors of
recidivism (van der Put et al. 2015; Williams et al., 2014). In a recent study, seventy-four percent

26

of the youth reported having pro-social friends; and this protective factor reduced the probability
of re-arrest by 22% (Williams et al., 2014). Another study also investigated the predictive
validity of the protective factors measured in the peer relationships subscale of the WSJCA and
found that the subscale was significantly, however not strongly, related to recidivism (van der
Put and colleagues, 2015b).
Commitment to school/education is another protective factor that has been highly
investigated among community samples (Chui & Chang, 2012; Crosnoe et al., 2002; Fagan et al.,
2007). Again, risk assessment research that investigated commitment to school among young
offenders is scarce and inconclusive. For instance, Rennie and Dolan (2010) and Gamelgård et
al. (2015) both conducted logistic regression analyses to investigate whether strong commitment
to school predicted recidivism. In both studies, commitment to school did not independently
predict recidivism. Conversely, using similar methods as the previously mentioned studies,
Lodewijks et al. (2008, 2010) and van der Put et al. (2014) found commitment to education to be
associated with recidivism. Past research on the relationship between school-level protective
factors and recidivism is promising, yet it is also limited, thus further investigation is warranted.
The current study contributed to the risk assessment literature by evaluating the predictive
validity of educational protective factors that are missing from existing risk assessment measures
(e.g., achievement, relationships with teachers).
Community-Level Single Protective Factor Items
Some research suggests that community-level protective factors are crucial to
understanding the causes and correlates of delinquency and recidivism; (Molnar et al., 2008) yet,
protective factors within this domain were rarely investigated. Community-level protective
factors specifically measured in the context of risk assessment were limited to involvement in

27

prosocial activities (Cuervo & Villanueva, 2015; Gamelgård et al., 2015; Lodewijks et al., 2008;
2010, Rennie &Dolan, 2010; Shepherd et al., 2016) and support from adults in the community
(Thompson & Pope, 2005; Williams et al., 2014). Researchers did not find a significant
relationship between recidivism and support from adult community members (Thompson &
Pope, 2005; Williams et al., 2014); however, findings supported the importance of involvement
in organized positive activities for reducing recidivism among young offenders (Cuervo &
Villanueva, 2015; Gamelgård et al., 2015; Shepherd et al., 2016). For example, Shepherd et al.
(2016) identified prosocial involvement as the strongest predictor of general and violent
recidivism (Shepherd et al., 2016). The current study aimed to contribute to this growing body of
literature by examining the predictive validity of community-level protective factors that are not
included in current risk assessment measures (i.e., perceived safety, access to resources).
To summarize, the theoretical and empirical value of individual, family, school, and
community protective factors has been extensively investigated with non-offender adolescents,
yet rarely with juvenile offenders, a major limitation. Collectively, researchers suggested that the
predictive value of protective factors in the individual domain are: prosocial attitudes
(Gammelgård et al., 2015; Jones et al., 2016; Lodewijks et al., 2008; 2010; Shepherd et al., 2016;
van der Put et al., 2012; 2015a; 2015b), impulse control (Williams et al., 2014), social skills
(Jones et al., Thompson et al., 2005), and resilient personality (Rennie & Dolan, 2008).
Researchers also suggested the predictive validity of protective factors within the family domain
are: parental supervision (Williams et al., 2014), discipline (Jimerson et al., 2004a), social
support (Gammelgård et al., 2015; Lodewijks et al., 2008; 2010), and strong attachment and
bonds (Lodewijks et al., 2008; 2010). Moreover, scholars have found protective factors within
the school domain to be associated with recidivism including: positive peers (Jimerson et al.,

28

2004b; Turner et al., 2006; Williams et al., 2014), commitment to education (Lodewijks et al.,
2008; 2010; Shepherd et al., 2016), and progress towards graduation (Jimerson et al., 2004a;
Jimerson et al., 2004b). Finally, community-level protective factors such as low neighborhood
crime (Jimerson et al., 2004a) and prosocial involvement (Gammelgård et al., 2015; Shepherd et
al., 2016) have also been found to decrease the odds of reoffending. Based on the number of
studies reporting significant findings, the literature most strongly supports the predictive validity
of prosocial attitudes, social support, positive peers, and commitment to education. The current
study examined the predictive validity of single protective factor items to build knowledge on the
specific protective factors most related to recidivism.
Predictive Validity and Incremental Validity of Composite Protective Factor Scores
Many researchers that examined the predictive validity of composite protective factor
scores found they were significant predictors of recidivism, indicating that evaluating offenders’
protective factors shows promise in effectively reducing odds of recidivism (e.g., Hilterman,
Bongers, Nicholls, & van Nieuwhenhuizen, 2016). Significant findings were consistent across
multiple recidivism outcomes and samples including: violent recidivism (Hilterman et al., 2016;
Rennie & Dolan 2010; Shepherd et al., 2016), general recidivism (Khanna, Shaw, Dolan, &
Lennox, 2014; Penney, Lee, & Moretti, 2010), youth in detention (Gamelgård et al., 2015;
Rennie & Dolan 2010; Vincent, Chapman, Cook, 2011), and probationers (Lee & Villagrana,
2015). Moreover, protective factors were found to be valid predictors of recidivism across
multiple international contexts including China (Chu et al., 2016), Australia (Penney et al., 2010;
Shepherd et al., 2016), the United Kingdom (Dolan & Rennie, 2008), Finland (Gamelgård et al.,
2015), The Netherlands (Lodewijks et al., 2008; 2010), Canada (Jones et al., 2016), Spain

29

(Cuervo & Villanueva, 2015) and the United States (Turner & Fain, 2006; Williams, Ryan,
Davis-Kean, McLoyd, & Schulenberg, 2014).
In addition to investigating overall predictive validity, researchers have also examined
whether protective factors add incremental value (beyond of risk factors) in recidivism prediction
models (Chu et al., 2016; Cuervo & Villanueva, 2015; Dolan & Rennie, 2008; Hilterman et al.,
2016; Jones et al., 2016; Lodewijks et al., 2008; 2010; Penney et al., 2010; van der Put et al.,
2015b). Of the nine studies in the current review that examined the incremental predictive
validity of protective factors, four produced significant findings (Cuervo & Villanueva, 2015;
Dolan & Rennie, 2008; Lodewijks et al., 2008; 2010). For instance, Dolan and Rennie (2008)
examined protective factors with an all-male sample of recently released young offenders and
found that protective factors added incremental validity to risk factors in a model predicting
general recidivism. In another study, Cuervo & Villanueva (2016) found that measures of
youth’s strengths increased the validity of general recidivism predictions. Similarly, Lodewijks
and colleagues (2008; 2010) found that protective factor scores predicted violence beyond risk
factors for young offenders awaiting trial, detained offenders, as well as recently released
offenders.
Overall, findings from the current review affirmed that composite protective factor scores
were significant predictors of recidivism across multiple samples and offense types (e.g.,
Gamelgård et al., 2015). Conversely, research studies that highlighted the investigation of the
incremental validity of protective factor scores were scarce and inconclusive as less than half of
the studies that examined this relationship produced significant findings. The current study
contributed to the risk assessment literature by examining the predictive and incremental validity
of composite protective factor scores.

30

Predictive Validity by Gender
The differential predictive validity of protective factors across gender has been rarely
investigated, however it is potentially important to support gender-specific rehabilitation as well
as the improvement of recidivism predictions for female offenders. Jones and colleagues (2016)
published the first investigation of the Youth Assessment and Screening Instrument (YASI).
Although the YASI developers (YASI; Orbis Partners, 2000) consider the tool to be genderresponsive, Jones and colleagues (2016) found that the YASI did not predict recidivism equally
across gender. Area under the curve (AUC) effect sizes revealed statistically stronger predictive
validity for boys than girls (AUC = .82; .68, respectively). Furthermore, while level of strengths
(low, moderate, and high) was associated with recidivism for boys, the same relationship did not
emerge for girls.
In another study, Craig and colleagues (2016) used components of social control theory
to examine how strong or weak attachments to conventional others (e.g., prosocial nonfamily
adults or friends) impacted exposure to adverse childhood experiences. Results indicated that
social bonds decreased the odds of re-arrest for female, but not male offenders (Craig et al.,
2016). Jimerson et al. (2004b) found that emotional support and low drug use were significant
predictors for female youth, but not for males. Further demonstrating the ambiguity of predictive
validity across gender, Williams et al. (2014) found that impulse control significantly predicted
recidivism for female offenders, but not for their male counterparts.
Overall, few studies consistently reported whether there were differences in the predictive
validity of protective factors for male and female young offenders (Craig et al., 2016; Jimerson
et al., 2004; Penney et al., 2010; Turner & Fain, 2006; Williams et al., 2014). To build

31

knowledge in this area, the current study explored whether the predictive validity of composite
protective factor scores varied by gender.
Predictive Validity by Race/Ethnicity
Researchers have investigated differential predictive validity of protective factors for
racial/ethnic minorities in the United States, as well as international contexts (Craig et al., 2016;
Jones et al. 2016; Pearl, Ashcraft, & Geis, 2009; Shepherd et al., 2014; Turner & Fain, 2006;
Vincent et al., 2011). For example, correlational analyses revealed the relationship between
protective factors and recidivism was stronger for African American (r = .37) and Hispanic (r =
.36) offenders than for White offenders (r = .15) (Pearl et al., 2009).
In another study, researchers investigated whether bonds with prosocial others moderated
the relationship between adverse childhood experiences and recidivism (Craig et al., 2016). The
authors found that Hispanic youth who had higher exposure to adverse experiences and stronger
social bonds were (marginally) less likely to reoffend with odds of re-arrest at 6% compared to
9% and 11% odds of re-arrest for African American and White offenders, respectively (Craig et
al., 2016).
Researchers in Australia examined the predictive validity of the SAVRY for youth who
identified as ESB (English-Speaking Background), CALD (Culturally and Linguistically
Diverse), or IND (Indigenous) (Shepherd et al., 2014). The researchers found that the protective
factor domain significantly predicted violent (AUC = .77) and general (AUC = .80) recidivism
for youth that identified as ESB. Conversely, protective factor scores did not predict violent
(AUC = .57, n.s.) or general (AUC = .57, n.s.) recidivism for CALD offenders. As for Indigenous
youth, protective factors demonstrated the strongest prediction of general recidivism (AUC =

32

.91), however composite scores did not predict violent recidivism for this offender group (AUC =
.67, n.s.).
Investigating whether there is a differential relationship between total protective factor
scores and recidivism across groups may be critical to seeking ways to improving predictive
validity for these vulnerable offenders. Although research in this area is growing, more research
is necessary to better understand this relationship. To build knowledge in this area, the current
study examined the differential predictive validity of composite protective factor scores by
race/ethnicity.
Summary of Literature Review Findings
Scholars have proposed that protective factors are an important but often overlooked
component of risk assessment (Rennie & Dolan, 2010). Relative to the risk assessment research
that focuses on the predictive validity risk factors, the state of the literature on the predictive
validity of protective factors is scarce, yet developing. The current review identified research
studies that examined the predictive validity of single protective factor items within individual,
family, school, and community domains. Based on studies reporting significant findings, the
literature most strongly supports the predictive validity of prosocial attitudes, social support,
positive peers, and commitment to education. The most consistent finding that emerged from the
current review is that composite protective factor scores are valid predictors of recidivism,
supporting the further exploration of the role of protective factors in reducing the likelihood of
reoffending. Moreover, there was a paucity of studies that investigated the incremental validity
of protective factors; collective findings were inconclusive as less than half of these studies (four
out of nine) produced significant findings. Finally, findings from past studies also indicated the
importance of exploring the role of gender and race/ethnicity as many researchers identified

33

evidence of differential predictive validity of protective factors across offender subgroups. The
goal of the current study was to contribute to the risk assessment literature by addressing the
aforementioned gaps in knowledge about protective factors.
Current Study
The current state of the risk assessment literature indicates that the relationship between
protective factors and recidivism is in need of further exploration in order to better understand
whether incorporating protective factors into recidivism prediction models can improve
predictive validity. The purpose of the current study was to identify whether the Protective
Factors for Reducing Juvenile Reoffending (PFRJR) is a valid tool.
The following research questions were investigated:
1. Are there significant differences in composite and subscale protective factors scores
across gender? Are there significant differences in composite and subscale protective
factors scores across race/ethnicity?
2. Do single protective factor items and composite protective factor scores predict
recidivism?
3. Is there differential predictive validity across race/ethnicity and gender?
4. Do composite protective factor scores predict time-to-recidivism? Do composite
protective factor scores add incremental validity to risk factor scores in predicting timeto-recidivism?

34

METHOD
See Study 1 for information on the Setting and Sample, and Training and Procedures as
they are identical for Study 2.
Measures
The YLS/CMI is a widely investigated third-generation risk assessment tool that was
designed to predict general recidivism for young offenders ages 12-18 (Hoge & Andrews, 2002).
The YLS/CMI was adapted from the Level of Service Inventory-Revised (LSI-R), an adult risk
assessment tool used in correctional systems (Andrews, Hoge, & Leschied, 2002). Andrews,
Robinson, and Hoge (1984) created the Youth Level of Service Inventory (YLSI) by conducting
a literature review and identifying 112 risk factors that demonstrated the strongest relationships
with initial delinquency. Later, scholars investigated (e.g., Andrews, Robinson, & Balla, 1996)
the psychometric properties of the YLSI and determined that 42 of the 112 items were
consistently associated with juvenile recidivism. Simourd, Hoge, and Andrews (1991) conducted
a factor analysis and the 42 items were grouped into 8 domains, resulting in the current study’s
measure, the YLS/CMI (See Appendix D for a list of items and alpha coefficients).
The YLS/CMI is an additive scale administered via semi-structured interview and each
item is scored dichotomously indicating whether or not risk is present (i.e., the presence of risk
represents a score of one for each item). The items are totaled and the composite score is
translated into a level of risk; low (0-8), moderate (9-22), and high (23-42). The instrument has
consistently produced acceptable reliability (α = .91) and predictive validity (AUC = .64) across
multiple samples (McGrath & Thompson, 2012; Onifade et al., 2008; Schmidt, Campbell, &
Houlding, 2011; Schwalbe, 2007).

35

The PFRJR is a new measure that was developed by the author in order to fill a gap in the
risk assessment literature, build knowledge on protective factors that reduce juvenile
reoffending, and to meet the needs of juvenile justice practitioners who desired to respond more
effectively to young offenders’ responsivity factors (i.e., non-criminogenic needs). The PFRJR is
a multidimensional, additive scale comprised of 15 dichotomous items, and is administered using
a semi-structured interview format (see list of items in Appendix C). Findings from Study 1
revealed that the instrument has acceptable reliability (α = .88) and overall predictive validity
(AUC = .64).
The dependent variable for assessing the predictive validity of the newly developed
measure (PFRJR) is recidivism. Recidivism was defined as any new court petition(s) received
within seven months subsequent to the date of the risk assessment and protective factor tools’
administration. Recidivism was coded as either 1 (new petition) or 0 (no new petitions). A
petition is a legal document in which a prosecutor officially files charges against a young
offender based on an allegation of committing a delinquent act (Snyder & Sickmund, 2006).
Although there are multiple methods of measuring recidivism (e.g., re-arrest, re-adjudication, reincarceration), evaluation of petitions is among one of the most common approaches (Harris,
Lockwood, & Mengers, 2009). Given that not all arrests result in a petition to juvenile court,
petitions are also considered an appropriate measure of official recidivism, as it requires
screening by the police, a prosecutor, or court intake officer (Harris et al., 2009; Maltz, 1984).
Recidivism data was collected archivally through the court data management system and was
provided to the author in de-identified form. In the event that a young offender aged out of the
juvenile justice system during the current study, both juvenile and adult records were examined
to track any new petitions acquired.

36

The second dependent variable for the current study was time-to-recidivism. This
measure reflected the number of days until the occurrence of a new juvenile or adult petition
following the administration of the YLS/CMI and PFRJR. For young offenders that did not
recidivate, this measure reflected the number of days until the end of the study period
(12/31/2016). Follow-up data were collected for at least seven months for the entire sample. The
maximum follow-up time for this sample was 13 months.

37

RESULTS
Descriptives: Mean Level Differences Across Subgroups
Both an ANOVA and a MANOVA (multivariate analysis of variance) were implemented
in order to examine whether there were significant differences across composite and protective
factor subscale scores by gender and race/ethnicity. A MANOVA tests the effect of multiple
independent variables on multiple dependent variables. MANOVAs can protect against Type I
errors that could occur as a result of running multiple ANOVAs or t-tests as it takes into account
colinearity among dependent variables. In other words, a MANOVA acts as a Bonferonni
correction as it keeps the probability of making a Type I error less than 5% (Tweedy &
Lunardelli, 2014).
Descriptive statistics are displayed in Table 3. The multivariate and univariate results
showed no significant differences in composite or subscale PFRJR scores for male and female
young offenders. Regarding race/ethnicity, an ANOVA with a planned contrast (White versus
African American youth) revealed that African American youth had significantly lower
composite PFRJR scores (F(3,272) = 1.85, p > .05; t(272) = 2.33, p < .05, d = .33) however,
composite PFRJR scores did not significantly differ between African American, Hispanic, and
Multi-racial young offenders. Moreover, composite PFRJR scores did not significantly differ
between White, Hispanic, and Multi-racial young offenders. As for PFRJR subscales, a post hoc
Tukey test revealed that White youth scored significantly higher than African American youth in
the Individual/Community subscale, Pillai’ Trace = .05, (F(6,544) = 2.31, p < .05), yet PFRJR
subscale scores did not statistically differ in the Individual/Community or Family/Social
subscales for any of the remaining subgroups.

38

Table 3. PFRJR Composite and Subscale Means and Standard Deviations by Gender and
Race/Ethnicity
Group

N

Full Sample

278

Composite
M(SD)
5.57(3.94)

Family/Social
M(SD)
2.50(2.16)

Individual/Community
M(SD)
3.06(2.32)

Female

93

5.61(3.76)

2.38(2.01)

3.32(2.34)

Male

185

5.54(4.03)

2.58(2.25)

2.95(2.31)

White

76

6.41(4.22)*

2.75(2.31)

3.65(2.44)*

Hispanic

28

5.32(3.44)

2.35(1.80)

2.96(2.28)

African American

117

5.05(3.93)*

2.47(2.28)

2.58(2.22)*

Multi-Racial

55

5.62(3.70)

2.30(1.88)

3.30(2.32)

Note. *Means are significantly different from each other across groups. (p < .05).
Evaluation of PFRFR Predictive Validity
To evaluate the predictive validity of the measure for the total sample and across offender
subgroups; a Receiver Operating Characteristic/Area Under the Curve (ROC/AUC) analysis was
implemented. This test specifies the proportion of true positives, or the number of offenders
predicted to reoffend that did in fact commit a future offense, to the number of true negatives, or
the number of youth predicted to not reoffend that indeed did not commit a future crime (Singh,
2013). The AUC can range from 0.0−indicating that a measure cannot discriminate between
reoffenders and non-reoffenders−to 1.0, demonstrating perfect discrimination. This statistic
caluclates the probability that a randomly selected recidivist would score higher on the PFRJR
than a randomly selected non-recidivist. In other words, an AUC above .50 indicates that the
predictive validity of the measure is better than chance (Rice & Harris, 2005). Rice & Harris
(2005) described AUC vales of .556 as small, .639 as moderate, and .714 as large predictive
validity effect sizes. This statistic is useful when comparing the predictive validity across
samples because it controls for base rates of the criterion variable (Rice & Harris, 2005; Singh,
2013).

39

ROC/AUC analyses were conducted to evaluate the predictive validity of each protective
factor item, the protective factor composite score for the total sample, as well as the protective
factor total score across young offender subgroups. Results of the ROC/AUC analysis of single
protective factor items are displayed in Table 4. As illustrated, the protective factor items that
significantly predicted recidivism produced small to moderate AUC statistics (.59 - .63),
including Strong Adult Bonds, Positive Response to Authority, Prosocial Attitudes, Access to
Resources, and Positive Adults. It is also worth mentioning that Strong Family Management,
Positive Personal Interests, Low Aggression, and Strong Social Skills produced small AUC
statistics that ranged from .58-.59 and yielded marginal statistical significance (p < .10).
Table 4. AUC Statistics for Single PFRJR Items
PFRJR Items
1. Family Management

AUC
.59

Std. Error
.04

p value
.06

CI
.51-.67

2. Commitment to School

.53

.04

.58

.44-.61

3. Organized Activities

.53

.04

.56

.44-.61

4. Personal Interests

.58

.04

.09

.49-.66

5. Religiosity

.47

.05

.43

.38-.55

6. Positive Peers

.55

.04

.25

.47-.63

7. Consistent Parenting

.57

.04

.14

.49-.65

8. Strong Adult Bonds

.59

.04

.05*

.50-.67

9. Response to Authority

.59

.04

.05*

.50-.67

10. Prosocial Attitudes

.59

.04

.04*

.51-.67

11. Low Aggression

.58

.04

.09

.49-.66

12. Social Skills

.59

.04

.06

.50-.67

13. Perceived Safety

.55

.05

.30

.46-.63

14. Access to Resources

.63

.04

.00*

.55-.72

15. Positive Adults

.59

.04

.05*

.50-.67

Note. * = (p < .05). AUC statistic reverse coded for interpretability.

40

Table 5 shows the recidivism rates and PFRJR total score AUC statistics for the full
sample and offender subgroups. Recidivism rates for male, female, White, and racial minority
youth ranged from 7.9% to 22.8%. Chi-square tests revealed no statistical differences in male
and female youths recidivism rates, however racial minority youth had significantly higher
recidivism rates than their White counterparts (x2 (1) = 8.04, p < .05).
The PFRJR total score significantly predicted recidivism for the full sample, male youth,
and White youth. The AUC for youth who identified as racial minority reached marginal
statistical significance and the AUC for female youth did not yield significant findings. To
identify whether the PFRJR predicted recidivism differently across gender and race/ethnicity,
post hoc significance tests were conducted. The results revealed that PFRJR scores were not
differentially valid predictors of recidivism for male and female offenders (p > .05), yet results
revealed that the PFRJR is a differentially valid predictor of recidivism for White and racial
minority youth. In other words, the PFRJR predicts recidivism comparably well across gender,
however the tool predicts recidivism better for White youth.
Table 5. Recidivism Rates and PFRJR AUC Statistics For Full Sample and Offender Subgroups
Group

Recidivism Rate (%)

AUC

Std. Error

p value

CI

Full Sample

18.7

.64

.04

.00

.56-.72

Female youth

18.3

.64

.07

.07

.49-.79

Male youth

18.9

.65

.05

.00

.55-.74

White youth

7.9*

.90*

.04

.00

.83-.98

.07

.49- .67

Racial minority
22.8*
.59*
.05
youth
Note. * = (p < .05). AUC statistic reverse coded for interpretability.

41

Evaluation of PFRJR Incremental Validity
A survival analysis was used to investigate the predictive validity and incremental
validity of protective factor scores. Survival analysis is a collection of statistical procedures that
examine time-to-event outcome data. Cox Proportional Hazards Model, or Cox regression--a
popular model in survival analyses--is a semiparametric test that models the relationship between
predictor variables (i.e., risk and protective factors) and an event (i.e., recidivism), while
accounting for differences in time to the occurrence of the event (Blakely & Cox, 1972;
Shepherd et al., 2016; Singer & Willet, 2003; Vincent et al, 2011). Given that every young
offender does not recidivate during the study period (i.e., censored cases), the Cox regression is
able to estimate time to a hypothetical date of recidivism for censored cases, based on the
survival times of young offenders who recidivated (Hosmer & Lemeshow, 1999).
The Cox regression is particularly popular as its mathematical modeling approach is
similar to a logistic regression, yet allows for the estimation of survival curves while accounting
for multiple explanatory variables (Kleinbaum & Klein, 2005). Predictor variables that produce
positive regression coefficients (β) decrease survival times, and variables with negative
regression coefficients increase survival times. Put simply, the author anticipates that higher
levels of protective factors are associated with longer time periods before recidivism occurs. In
the Cox regression, the preferred indices for interpretation are Hazard ratios (Exp[B]). Hazard
ratios represent the ratio of the likelihood that a youth will recidivate relative to a specified
survival time. For instance, an (Exp[B]) of 1.23 denotes that a one-unit increase in total score
would lead to a 23% increased likelihood of recidivism, given the specified survival time.
The average survival time (number of days until recidivism or censoring) for the current
sample was M = 412.33, SD = 137.96. There were no significant differences in average survival

42

times across gender, however results of a Tarone-Ware chi-square test (x2(1) = 9.61, p < .05)
revealed that White youth had significantly longer survival times (M = 452.76, SD = 83.17) than
African American youth (M = 405.70, SD = 139.77) and Multi-racial youth (M = 376.11, SD =
170.31).
A Cox regression was implemented to examine the predictive strength of PFRJR total
scores; the protective factor scores were negatively related to time-to-recidivism (β = -.13, SE =
.04, Exp[B] = .88, p < .05), x2(1) = 10.14, p < .05. In other words, at any time point during the
study period, for every one-point increase in composite protective factor score, youth were 12%
less likely to recidivate.
In order to examine the incremental predictive validity of composite protective factor
scores relative to composite risk factor scores, a hierarchical Cox regression was performed.
YLS/CMI total risk scores were entered in the first block, and the PFRJR total protective factor
scores were added to the second block. In the first model, YLS/CMI risk scores significantly
predicted time-to-recidivism (β = .06, SE = .02, Exp[B] = 1.06 , p < .05), x2(1) = 9.72, p < .05.
Put simply, at any time point during the study period, for every one-point increase in composite
risk scores, youth were 6% more likely to recidivate. Results for the final model are shown in
Table 6. As illustrated, YLS/CMI risk scores did not remain a significant predictor once PFRJR
protective factor scores were entered in the model. Moreover, entering PFRJR scores in Block 2
did not add incremental value to the prediction model. Cox regressions were also conducted by
entering PFRJR scores in the first block and YLS/CMI scores in the second block (not shown in
Table). Protective factor scores significantly predicted recidivism and risk scores did not add
incremental value to the model.

43

Table 6. Hierarchical Cox Regression to Examine Incremental Validity of PFRJR Total Scores
β (SE)

Exp[B]

Overall x2 (df)

p value

Exp[B] CI

Block 1: YLS/CMI scores

.03 (.03)

1.03

9.82 (1)

.21

.98 – .1.09

Block 2: PFRJR scores

-.09 (.05)

.91

11.86 (2)

.09

.82 – 1.01

Notes. YLS/CMI scores = Youth Level of Service/Case Management Inventory; PFRJR =
Protective Factors for Reducing Juvenile Reoffending. x2 values are for the overall model after
each block.

44

DISCUSSION
Scholars have criticized risk assessment processes that exclusively evaluate risk factors.
(de Vogel et al., 2011; Loeber & Farrington, 2012; Miller, 2006; Rogers, 2000). Although the
field is slowly shifting its focus away from exclusively measuring risk with the emergence of
studies exploring both risk and protective factors, there is still much work to be done to better
understand the individual and ecological characteristics that contribute to desistance (Walker et
al., 2013). The current study described the development of a strength-based assessment tool
comprised solely of protective factors, the Protective Factors for Reducing Juvenile Reoffending
(PFRJR). The current study also described the tool’s factor structure, reliability, predictive
validity, differential predictive validity, and incremental predictive validity.
Mean Level Differences Across Race/Ethnicity and Gender
Male and female offenders reported mean PFRJR scores that were not statistically
different. Conversely, White offenders also scored significantly higher than African American
offenders on composite PFRJR scores. It is important to note that differences in predictive
validity and mean composite scores across race/ethnicity could be attributed to the assessment
process itself. For instance, Chapman et al. (2006) found that African American youth were rated
as being significantly lower in risk of violence, and higher in prosocial involvement, strong
attachment and bonds, and resilient personality than their White counterparts. The researchers
noted that in their sample, young offenders were often assigned to juvenile court officers who
identified as having a similar race/ethnicity. This is not the common practice in the current
study’s court of interest as the racial/ethnic composition of the court officers (approximately
10% racial/ethnic minority) is not commensurate with the racial/ethnic composition of the young
offenders (approximately 60% racial/ethnic minority). Therefore, it is possible that the current
study’s findings reflect bias in the assessment process in which racial minority youth are being

45

perceived as having fewer assets and strengths than White youth. Future studies should aim to
integrate the PFRJR into risk assessment procedures in juvenile courts with more balanced court
officer to young offender racial/ethnic compositions, as well as qualitatively examining the
potential for implicit bias to enter the risk assessment process.
Predictive Validity of Single Protective Factor Items
The author investigated the predictive validity of single protective factor items to
understand which items were most related to recidivism. Of the 15 items retained on the PFRJR,
5 were significantly related to recidivism: Strong Adult Bonds, Positive Response to Authority,
Prosocial Attitudes, Access to Resources, and Positive Adults. All but one of these items belong
to the Individual/Community domain, as they strongly loaded onto the first factor in the EFA.
Past research findings supported the predictive validity of these individual-level items as having
a positive response to authority (Lodewijks et al., 2010), prosocial attitudes (Gamelgård et al.,
2015), and strong bonds with prosocial adults (Lodewijks et al., 2008) has emerged in the
literature as predictors of recidivism. Based on the results of the literature review conducted by
the author, community-level items are far less examined with young offender samples.
Nevertheless, research that examined the relationship between delinquency and protective factors
with non-offender samples has indicated that access to resources in one’s community as well as
living in a community with prosocial adults is associated with decreased odds of engaging in
delinquency (Molnar et al., 2008). Overall, these results supported past empirical findings as
well as theoretical arguments that highlight the contributions of both individual and ecological
factors in predicting delinquency and recidivism.

46

Predictive Validity and Differential Predictive Validity of PFRJR Scores
Results of the ROC/AUC analyses revealed that composite protective factor scores were
significant predictors of recidivism (AUC = .64, p < .05). These findings are comparable with
AUC statistics produced by SAVRY protective factor scores (AUC = .72; Lodewijks et al.,
2010). Regarding predictive validity by gender, there were no significant differences for male
and female offenders. Regarding the predictive validity by race/ethnicity, composite protective
factor scores demonstrated a stronger relationship with recidivism for White youth compared
(AUC = .90, p < .05) to racial minority youth (AUC = .59, p < .05), suggesting that the tool may
lack cultural relevance. Many of the PFRJR items were derived from risk assessment validation
studies and community-based longitudinal studies that employed largely White samples. In other
words, the PFRJR may be missing items that are specifically salient in reducing the odds of
recidivism for racial minority offenders. Scholars have suggested that coping abilities and family
ethnic socialization may be important protective factors for racial minority youth (Caldwell,
Kohn-Wood, Schmeelk-Cone, Chavous, & Zimmerman, 2004; Sexias & Wade, 2014; Wright &
Younts, 2009). In the future, the PFRJR will be modified to include theoretically based culturally
specific items that could potentially improve predictive validity for racial minority young
offenders.
Predictive and Incremental Validity of PFRJR Scores with Time-To-Recidivism
As expected, the author found that composite PFRJR scores predicted time-to-recidivism.
In other words, higher levels of protection were related to longer periods of time to committing a
new offense. This finding is consistent with past research that has examined the relationship
between time-to-recidivism and protective factors using similar methods and analyses (Chu et
al., 2015; Shepherd et al., 2016; Vincent, Chapman & Cook, 2011). It is worth noting that

47

Vincent and colleagues (2011) found that protective factor scores predicted time to recidivism
solely for White youth, illuminating the need to cautiously acknowledge findings that have not
considered the potential influences of race/ethnicity or gender. In the future, the author will
investigate whether race/ethnicity (as well as intersections of race/ethnicity and gender)
moderates the relationship between PFRJR scores and time-to-recidivism, an important
contribution.
Composite PFRJR protective factor scores did not add incremental validity to YLS/CMI
risk scores. There was also interest in whether YLS/CMI scores added incremental validity to
PFRJR scores and similar (non-significant) results were produced. Interestingly, both scores
predicted time-to-recidivism when in the model alone, however when the respective composite
score was entered into the second block of the model, both composite scores failed to remain a
significant predictor of the outcome. Given the correlation between PFRJR and YLS/CMI scores
(r = -.67), the author speculated that there may be an issue of multicollinearity, however results
of Tolerance and Variance Inflation Factor tests did not confirm this conjecture.
Although scholars have proposed that the addition of protective factors may improve risk
assessment prediction models (de Vogel et al., 2011; Rogers, 2000), findings from the literature
review described earlier suggested that this notion is inconclusive, as less than half of the studies
exploring this hypothesis produced significant results. It is worth noting that de Vogel and
colleagues (2011) and Rogers (2000) specifically emphasized the importance of protective
factors for predicting violent reoffending. It seems that the most apparent distinction between the
studies that yielded significant findings and those that did not, was the use of detained offenders
(primarily violent offenders) versus probationers. Compared to youth on probation, youth who
are detained, in theory, likely have higher risk scores and likely commit more serious offenses.

48

Future research should separately examine the predictive validity of the PFRJR using both
violent and non-violent recidivism outcomes.
Moreover, if models of resilience as outlined in resilience theory are also considered, one
could speculate that protective factors scores improve risk assessment prediction models
specifically for more serious offenders. In other words, given that resilience (in this case,
desistance from delinquency or non-recidivating) is conceptualized as having a successful
outcome despite risky conditions, youth at higher risk of recidivism may reap greater benefits of
the presence of protective factors than low risk offenders (see Lodewijks et al., 2010). To build
knowledge on the role of protective factor models of resilience in risk assessment research,
future studies should examine the buffering effect of protective factor scores at high and low
levels of risk.
Limitations
It is important to mention the methodological shortcomings of the current study. The
most salient drawbacks of this study are connected to data collection time restraints. First, the
author evaluated the reliability and validity of the PFRJR with the same sample in which the
measure was developed. Scholars admonish this practice and encourage researchers to crossvalidate new measures as any results obtained are likely specific to the sample, therefore, lacking
generalizability (Dawis, 2000). In future studies, the author will obtain additional data from the
court of interest and as well as initiate data collection in a different juvenile court to more
rigorously evaluate the psychometric properties of the PFRJR. The current study’s sample size
was also a limitation as psychometricians recommend the use of large samples (N ≥ 400) to
optimize the results of factor analysis and to ensure the validity of the selected items (Dawis,
2000). Moreover, the sample size was a limitation given the authors’ interest in exploring the

49

moderating effect of race/ethnicity by gender interactions. Theories and methods of
intersectionality are rarely emphasized in criminological research (Potter, 2013), however it is
important given that individuals’ experiences with the justice system can vary by their
intersecting identities (e.g., race, ethnicity, class, gender, etc.). Finally, although survival analysis
controlled for various times at risk, over 80% of the sample was censored, which was likely due
to the minimum follow up period (i.e., seven months). Future studies should follow the total
sample of youth for at least 12-months to capture a fuller trajectory of juvenile recidivism.
Findings from the current study are promising and aimed to make important contributions
to the juvenile risk assessment literature. These findings also have critical practical implications
for juvenile justice practitioners. Training, implementing and evaluating any evidence-based
intervention takes time, money, and buy-in from juvenile justice practitioners (Vincent, PaivaSalisbury, Cook, Guy, & Perrault, 2012). Courts that are employing validated risk assessments
that are integrated into standing policies and procedures may be less resistant to adopting
strengths-based strategies if they are presented with a supplemental measure of protective factors
that can be adapted to their existing practices, like the PFRJR. The dominant model of offender
rehabilitation, Risk-Need-Responsivity, emphasizes the importance of using interventions that
target offenders’ specific criminogenic needs (Andrews et al., 2006). Proponents of RNR assert
that juvenile courts should assess risk factors (e.g., poor academic achievement) and implement
strategies (e.g., tutoring) intended to reduce risk of reoffending (James, 2015). Targeting risk
factors is critical; however working towards enhancing an offender’s existing assets could also
contribute to decreasing risk of recidivism (Baglivio et al., 2014; Lodewijks et al., 2008). In
addition, promoting protective factors could facilitate goal setting in case management as well as
improve service delivery because practitioners would have a thorough evaluation of the

50

offenders’ criminogenic and non-criminogenic profiles. Researchers should continue to
investigate the role of protective factors in the risk assessment process as they have the potential
to improve recidivism predictions while promoting positive youth development.

51

APPENDICES

52

APPENDIX A: STUDY 1 LITERATURE REVIEW TABLE
The following tables provide a summary of the information found from the literature
review that was conducted to identify protective factors associated with desistance. The table
provides a visual representation of the studies that were used to write the initial list of items for
the newly developed measure, Protective Factors for Reducing Juvenile Reoffending. Each
column illustrates the following information: 1) Author: first author and year of publication; 2)
Protective Factor Measure: the scale or example of items used to measure protective factors; 3)
Sample: the population (community, offenders, etc.) that the study’s sample was drawn from; 4)
Percent female: the percentage of female offenders reported in each study; 5) Examined gender:
indicates whether the authors tested for differential effects across gender 4) Percent minority: the
percentage of minority offenders reported in each study; 5) Examined culture: indicates whether
the authors tested for differential effects across race/ethnicity or cultural heritage; 8) Domain:
indicates which domains (community, individual, family, and school) the protective factors
examined belonged to and 9) significant findings: indicates whether the study found any one
protective factor to be negatively associated with delinquency. Detailed descriptions of the key
findings from exemplar studies are available upon request.

53

Table 7. Summary of Literature Review of Relevant Protective Factors
Author

Protective factor
measure

Sample

Larzelere,
1990

22 items
(monitoring)

Community

Loeber, 1991

188 items

Community

Percent
Female

Examined
Gender

Percent

0%

Yes ☐

NR

Minority

Examined
culture

Domain

Yes ☐

F
No ☐

0%

Yes ☐

50%

Yes ☐

FIS
No ☐

(school
functioning)
StouthamerLoeber, 1993

188 items

Smith, 1995

18 items
(attachment)

Community

4 items

Offenders

Hoge, 1996

Community

0%

Yes ☐

50%

Yes ☐

FIS
No ☐

(peer support)
26%

Yes ☐

85%

Yes ☐

CFIS
No ☐

20%

Yes ☐

NR

Yes ☐

IS
No ☐

(respect for
authority)
Born, 1997

Desistance from
offending

Significant
Findings

Offenders

20%

Yes ☐

NR

Yes ☐

FIS
No ☐

54

Table 7 (cont’d)
Stattin, 1997

5 items (IQ)

Community

0%

Yes ☐

NA

NS

I
No ☐

Ayers, 1999

Pollard, 1999

Community
Communities that
Care Youth Survey

48%

8 items

NR

Community

50%

Yes ☐

CFIS

No ☐
Yes ☐

No ☐
28%

Yes ☐

FIS
No ☐

(prosocial
attitudes)
Duncan,
2000

6 items

Wikström,
2000

42 items (ADHD)

Carr, 2001

23 items

Community

47%

19%

Yes ☐

CFIS

No ☐

(family support)
Community

0%

Yes ☐

No ☐
57%

Yes ☐

FI
No ☐

Offenders

43%

Yes ☐

87%

Yes ☐

FIS
No ☐

(self-esteem)
Arthur, 2002

Community
Communities that
Care Youth Survey

49.1%

Yes ☐

NR

Yes ☐

CFIS
No ☐

55

Table 7 (cont’d)
Calvert, 2002 6 items

Community

16%

Yes ☐

47%

Yes ☐

FIS
No ☐

(temperament)
Chung, 2002

Community
Communities that
Care Youth Survey

49%

Crosnoe,
2002

6 items

56%

Hogue, 2002

25 items

Community

Yes ☐

54%

Yes ☐

CFIS
No ☐

35%

Yes ☐

FS

No ☐

(teacher bonding)
Community

56%

Yes ☐

No ☐
97%

FIS
No ☐

(school
involvement)
McKnight,
2002

16 items
(religiosity)

Community

StouthamerLoeber, 2002

188 items (school
functioning)

Community

Clingempeel,
2003

10 items

Offenders

100%

40%

Yes ☐

No ☐

FIS

No ☐
0%

Yes ☐

No ☐
50%

Yes ☐

FIS
No ☐

22%

Yes ☐

51%

Yes ☐

FIS
No ☐

(social
competence)

56

Table 7 (cont’d)
Deković,
2003

Relationship
quality with peers
and parents

Herrenkohl,
2003

Community
Communities that
Care Youth Survey

49%

Aspy, 2004

Youth Assets

51%

Community

51%

NR

F

No ☐

Community

No ☐
64%

No ☐

Caldwell,
2004

Multidimensional
Inventory of Black
Identity

Community

Resnick,
2004

30 items
(neighborhood
safety)

Community

StouthamerLoeber, 2004

188 items (positive
peers)

Community

Sprott, 2005

12 items

Community

CFIS
No ☐

52%
No ☐

(school bonding)

Yes ☐

55%

CFIS
No ☐

100%
No ☐

52%

No ☐

No ☐
I

No ☐
NR

Yes ☐

No ☐
CFIS

No ☐
0%

Yes ☐

No ☐
50%

Yes ☐

FIS
No ☐

50%

NR
No ☐

57

Yes ☐

S
No ☐

Table 7 (cont’d)
Cantillon,
2006

Neighborhood
advantage

Community

Chapman,
2006

SAVRY

Offenders

55%

Yes ☐

0

Yes ☐

C
No ☐

30%

Yes ☐

64%

FIS
No ☐

*Miller, 2006 IORNS

Offenders

0%

Yes ☐

83%

No ☐
IS

No ☐
Barnes, 2007

Fagan, 2007

Hart, 2007

Liau, 2007

Time spent in
prosocial activities

Community

55%

30%
No ☐

Community
Communities that
Care Youth Survey

52%

5 scales (Healthy
Kids Survey)

Community

47%

67 items

Offenders

No ☐
FI

No ☐
20%

Yes ☐

No ☐
CFIS

No ☐

No ☐
90%

Yes ☐

FIS

No ☐
29%

Yes ☐

No ☐
NA

NA

FI
No ☐

(social skills)
Youngblade,
2007

17 items
(family closeness)

Community

51%

16%
No ☐

58

CF
No ☐

No ☐

Table 7 (cont’d)
Fredricks,
2007

Organized
activities

Community

Benhorin,
2008

Social Support
Scale for Children

Community

Gardner,
2008

Self-regulation

Community

51%
No ☐
66%

No ☐
100%

No ☐
49%

I

Yes ☐

No ☐
FS

No ☐
56%

Yes ☐

No ☐
I
No ☐

Molnar, 2008 74 items

Community

50%

13 items
(commitment)

87%

Yes ☐

CFI

No ☐

(neighborhood
resources)
Ryan, 2008

67%

Foster care

0%

Yes ☐

No ☐

100%

FIS
No ☐

VanderbiltAdriance,
2008

32 items (nurturing Community
parenting)

0%

Chilenski,
2009

5 items

50%

Community

Yes ☐

50%

Yes ☐

No ☐
FI
No ☐

Yes ☐

15%

Yes ☐

CFIS
No ☐

(school leadership)

59

Table 7 (cont’d)
Domburgh,
2009

40 items
(academic
achievement)

Community

Hartman,
2009

55 items

Community

Petts, 2009

24 Items (parental
resources)

Community

139 items
(academics)

Community

Lodewijks,
2010

SAVRY

Offenders

Mowder,
2010

Resiliency Scales
for Children &
Adolescents

Offenders

Rennie, 2010

SAVRY

Offenders

Laan, 2010

0%

Yes ☐

62%

Yes ☐

CFIS
No ☐

50%

67%

Yes ☐

FIS

No ☐

(religiosity)
NR

No ☐
NR

No ☐
50%

Yes ☐

FI
No ☐

NA

NA

No ☐
FIS
No ☐

0%

Yes ☐

NA

NA

FIS
No ☐

76%

Yes ☐

44%

Yes ☐

I
No ☐

0%

Yes ☐

15%

Yes ☐

FIS
No ☐

Whitney,
2010

47 items
(future orientation)

Community

45%

Yes
No ☐

60

20%

Yes ☐

FI
No ☐

Table 7 (cont’d)
Hirshfield,
2011

15 items
(academic
engagement

Community

Stepp, 2011

8 items

Community

52%

93%

Yes ☐

S

No ☐
0%

Yes ☐

No ☐
58%

Yes ☐

I
No ☐

(social
competence)
Stevens,
2011

12 items

Bernat, 2012

10 items (activities
with parents)

Briney, 2012

Chui, 2012

Community

100%

28%

Yes ☐

FIS

No ☐

(support)
Community

55%

Yes ☐

No ☐
33%

Yes ☐

CIFS
No ☐

Community
Communities that
Care Youth Survey

49%

Social Bond Scale

52%

Community

Yes ☐

30%

Yes ☐

CIFS
No ☐

NA

NA

IF

No ☐
Coster, 2012

Maternal
employment

Community

NR

Yes ☐

No ☐
NR

Yes ☐

F
No ☐

61

Table 7 (cont’d)
Fougere,
2012

Resilience Scale

Hardaway,
2012

25 items

Henry, 2012

36 items

Offenders

15%

36%

Yes ☐

I
No ☐

Community

50%

Yes ☐

89%

Yes ☐

IFS
No ☐

(school climate)
Community

51%

74%
No ☐

(attitude towards
school)
Herrenkohl,
2012

25 items

Kurlycheck,
2012

22 items
(collective
efficacy)

Community

Pardini, 2012

141 items
(prosocial
attitudes)

Community

12 items
(neighborhood
safety)

Community

Reingle,
2013

Yes ☐

Community

49%

Yes ☐

IFS
No ☐

53%

Yes ☐

No ☐

CFIS
No ☐

(school
attachment)
27%

Yes ☐

85%

Yes ☐

C
No ☐

0%

Yes ☐

60%

Yes ☐

CFIS
No ☐

0%

Yes ☐

35%

CFIS
No ☐

62

No ☐

Table 7 (cont’d)
Robbé, 2013

SAPROF

Clinical
patience

0%

27%

Baglivio,
2014

Positive
Achievement
Change Tool

Offenders

Krohn, 2014

19 items
(attachment to
teacher)

Community

Van der Put,
2014

WSJCA

Offenders

Burt, 2015

Ethnic-racial
socialization scale

Community

Hardaway,
2015

Kinship Support
Scale; Parenting
Style Index

Community

Henry, 2015

Teenager
Experience of
Racial
Socialization Scale

Community

Yes ☐

NA

NA

I
No ☐

60%
No ☐

25%

Yes ☐

IFS
No ☐

85%

Yes ☐

No ☐
IFS
No ☐

37%

Yes ☐

40%

Yes ☐

IFS
No ☐

100%

100%
No ☐

50%

Yes ☐

F
No ☐

71%

Yes ☐

No ☐
F
No ☐

57%

Yes ☐

100%

FI
No ☐

63

No ☐

Table 7 (cont’d)
Summary

9% of studies used
risk assessments

18% used
offender
samples

32% of
studies
had less
than
25%
female
youth

37%
examined
gender

14% did
not report
percent
minority;
24%
reported
less than
25%
minority

24%
examined
culture
effects

I-83%
F-82%
S-65%
C-28%

Note: I-Individual, F-Family, S-School, C- Community; NR-not reported
The following table displays a summary of the information found from the literature review that was conducted to identify
juvenile risk assessment validation studies that investigated protective factors. Table 2 presents the first author and year of publication
(First author); the actuarial risk assessment used to measure protective factors (Measure); protective factor domains represented in the
measure (Domains); whether the study examined the predictive validity of single protective factor items (Single items examined);
whether the study examined the predictive validity of protective factor composite scores (Composite score examined); whether the
study examined the incremental validity of protective factors (Incremental validity examined); whether the study tested the protective
factor model of resilience (Resilience model examined); whether the authors explored the relationship between protective factors and
gender (Gender examined); and whether the authors explored the relationship between protective factors and race/ethnicity
(Race/ethnicity) examined.

64

Table 8. Summary of Protective Factors in Risk Assessment Literature Review
First
author

Measure

Domains

Single
items
examined

Composite
score
examined

Incremental
validity
examined

Gender
examined

Race/ethnicity
examined

Jimerson,
2004

SBARA

CFIS

CFIS

No

No

Yes

No

Jimerson,
2004

SBARA

CFIS

CFIS

Yes

No

Yes

No

Thompson,
2005

YLS-AA

CFI

CFI

Yes

No

No

No

Turner,
2006

SDRRC

CFIS

CFIS

Yes

No

Yes

Yes

Upperton,
2007

YLS-AA

CFIS

No

Yes

No

No

No

Lodewijks,
2008a

SAVRY

CFIS

CFIS

Yes

No

No

No

Lodewijks,
2008b

SAVRY

CFIS

No

Yes

Yes

No

No

Dolan,
2008

SAVRY

CFIS

No

Yes

Yes

No

No

Pearl, 2009

SDRRC

CFIS

No

Yes

No

Yes

Yes

65

Table 8 (cont’d)
Penney,
2010

SAVRY

CFIS

No

Yes

Yes

Yes

No

Rennie,
2010

SAVRY

CFIS

CFIS

Yes

No

No

No

Lodewijks,
2010

SAVRY

CFIS

CFIS

Yes

Yes

No

No

Vincent,
2011

SAVRY

CFIS

No

Yes

No

No

Yes

van der
Put, 2012

WSJCA

CFIS

CFIS

Yes

No

No

No

Khanna,
2014

YLS &
SAVRY

CFIS

No

Yes

No

No

No

Shepherd,
2014a

YLS &
SAVRY

CFIS

No

Yes

No

No

No

Hilterman,
2014

YLS &
SAVRY

CFIS

No

Yes

Yes

No

No

Baglivio,
2014

PACT

CFIS

CFIS

No

No

No

No

Shepherd,
2014b

SAVRY

CFIS

No

Yes

No

No

Yes

66

Table 8 (cont’d)
Williams,
2014

WSJCA

CFIS

CFIS

No

No

Yes

Yes

van der
Put, 2015a

WSJCA

CFIS

CFIS

No

No

No

No

van der
Put, 2015b

WSJCA

CFIS

CFIS

Yes

Yes

Yes

No

Chu, 2015

YLS/CMI
2.0

CFIS

No

No

No

Yes

No

Gamelgård,
2015

SAVRY

CFIS

CFIS

Yes

No

No

No

Cuervo,
2015

YLS/CMI

CFIS

No

Yes

Yes

No

No

Lee, 2015

SDRRC

CFIS

CFIS

Yes

No

No

No

Shepherd,
2016

SAVRY

CFIS

CFIS

No

No

No

No

Chu, 2016

YLS &
SAVRY

CFIS

No

Yes

Yes

No

No

RoyerGagnier,
2016

SAI-Y

CFIS

No

No

No

No

No

67

Table 8 (cont’d)
Craig, 2016 PACT (social CFIS
bonds)

F

No

No

Yes

Yes

Jones, 2016 YASI

CFIS

Yes

Yes

Yes

Yes

FIS

Note. Positive Achivement Change Tool = PACT; San Diego Risk and Resiliency Assessment = SDRRC; Santa Barbara Assets and
Risk Assessment = SBARA; Strengths Assessment Inventory-Youth = SAI-Y; Structured Assessment of Violence Risk in Youth =
SAVRY; Washington State Juvenile Court Assessment = WSJCA; Youth Assessment and Screening Instrument = YASI; I-Individual,
F-Family, S-School, C- Community
Table 9. Summary of Table 8
Measure

45% percent of the studies investigated the predictive validity of the SAVRY.
26% percent of the studies investigated the predictive validity of versions of the YLS.
13% percent of the studies investigated the predictive validity of the WSJCA.
10% percent of the studies investigated the predictive validity of the SDRRC.
6% percent of the studies investigated the predictive validity of the PACT and SBARA.
3% percent of the studies investigated the predictive validity of the YASI and SAI-Y.

Domain

93% of the studies used measures that were comprised of protective factors across each domain. Each risk
assessment tool measured community-level factors (i.e., prosocial involvement), however only four
measures captured more ecological factors (e.g., neighborhood crime).

Single Items
Examined
Composite
Score
Incremental
Validity

55% of studies examined the independent predictive validity of single protective factor items.
74% of studies examined the predictive validity of the protective factor cumulative score.
29% of studies examined the incremental validity of the protective factor cumulative score.

68

Table 9 (cont’d)
Gender

32% of studies examined the differential predictive validity of protective factors by gender.

Race/ethnicity 26% of studies examined the differential predictive validity of protective factors by race and/or ethnicity.

69

APPENDIX B: PFRJR INITIAL ITEM POOL
Table 10. Initial Pool of PFRJR Items
Scale Items

Item adopted from

Domain

Decision for final scale

Passing core subjects

Justice experts

Education

Retained

Grades in core subjects

Ayers, 1999

Education

Retained

Positive school history

Baglivio, 2014

Education

Limited evidence

No special education needs

Chilenski, 2009

Education

Limited evidence

Commitment to school

Chui, 2012

Education

Retained

Mostly A’s

Crosnoe, 2002

Education

Limited practical relevance

Attitudes towards school

Crosnoe, 2002

Education

Adapted as an interview question

Grades in math and science

Bernat, 2006

Education

Redundant

Standardized test scores

Hartman, 2009

Education

Difficult to consistently measure

Positive school environment

Hartman, 2009

Education

Limited evidence

Teacher reported study skills

Henry, 2012

Education

Difficult to consistently measure

Bonding to school

Herrenkohl, 2012

Education

Redundant

70

Table 10 (cont’d)
Emotional engagement

Hirschfield, 2011

Education

Limited evidence

Good school achievement

Hoge, 1996

Education

Redundant

Academic competence

Kelly, 2012

Education

Redundant

Aspirations

Krohn, 2014

Education

Limited relevance

Teacher reported performance

Laan, 2010

Education

Difficult to measure consistently

Motivation to succeed

Lodewijks, 2010

Education

Redundant

Parent attitude toward education

Losel, 2012

Education

Limited evidence

Fair teachers

Mcknight, 2002

Education

Retained

Academic Achievement

Pardini, 2012

Education

Redundant

Grade point average

Reingle, 2013

Education

Redundant

Positive school relationships

Simoes, 2008

Education

Redundant

Attachment to teachers

Smith, 1995

Education

Adapted as an interview question

Teacher report of hard working

Stouthamer-Loeber, 2002

Education

Difficult to measure consistently

Few opportunities to get weed

Ayers, 1999

Individual

Adapted as an interview question

71

Table 10 (cont’d)
Norms against drug use

Ayers, 1999

Individual

Adapted as an interview question

Infrequent past use

Baglivio, 2014

Individual

Limited relevance

Availability of drugs

Chung, 2002

Individual

Retained

Peer use of drugs

Fagan, 2007

Individual

Limited evidence

Actively abstaining

Justice experts

Individual

Retained

Low access to drugs

Herrenkohl, 2012

Individual

Redundant

Monitoring

Cantillon,

Family

Retained

Support

Carr, 2001

Family

Adapted as an interview question

Parental bonding

Chui, 2012

Family

Retained

Attachment to mother

Coster, 2012

Family

Redundant

Household organization

Crosnoe, 2002

Family

Retained

Consistent discipline

Domburgh 2009

Family

Retained

Family functioning

Domburgh 2009

Family

Redundant

Relationship with parents

Fagan, 2007

Family

Retained

72

Table 10 (cont’d)
Demanding/responsive parents

Hart, 2007

Family

Redundant

Family management

Herrenkohl, 2003

Family

Retained

Family cohesion

Hjemdal, 2006

Family

Not clearly defined

Attachment and monitoring

Krohn, 2014

Family

Redundant

Low parental stress

Laan, 2010

Family

Limited evidence

Parents marital status

Liau, 2007

Family

Limited evidence

Parental support

Lodewijks, 2010

Family

Redundant

Positive relationships

Loeber, 1991

Family

Redundant

Strict discipline

Loeber, 1991

Family

Redundant

Feeling loved

Mcknight, 2002

Family

Limited evidence

Having many friends

Carr, 2001

Peers

Limited relevance

Low peer delinquency

Bernat, 2006

Peers

Adapted as an interview question

Quality relationships

Clingepeel, 2003

Peers

Adapted as an interview question

Peers obey rules

Molnar, 2008

Peers

Adapted as an interview question

73

Table 10 (cont’d)
Low peer deviance

Pardini, 2012

Peers

Redundant

Prosocial attitudes

Baglivio, 2014

Individual

Retained

Belief in legal system

Chui, 2012

Individual

Adapted as an interview question

Positive response to

Lodewijks, 201

Individual

Retained

Pardini, 2012

Individual

Adapted as an interview question

Guilt feelings

Stouthamer-Loeber, 2002

Individual

Limited evidence

Low aggression

Born, 1997

Individual

Retained

Positive personality

Carr, 2002

Individual

Not clearly defined

Resilient personality

Chapman, 2006

Individual

Limited evidence

Elicits positive attention

Hawkins, 1995

Individual

Adapted as an interview question

Low ADHD

Loeber, 1991

Individual

Limited evidence

Healthy self-esteem

Chapman, 2006

Individual

Adapted as an interview question

authority/help
Negative attitudes towards
delinquency

74

Table 10 (cont’d)
Positive personal interests

Justice experts

Individual

Retained

Neighborhood resources

Molnar, 2008

Community

Retained

Prosocial involvement

Chapman, 2006

Community

Retained

Organization involvement

Chui, 2012

Community

Redundant

Involved in activities

Fredericks, 2008

Community

Redundant

Good use of time

Hoge, 1996

Community

Adapted as an interview question

Use of time

Van der put, 2014

Community

Redundant

Social ties

Chung, 2006

Community

Limited evidence

Non-parent mentor

Molnar, 2008

Community

Redundant

Perceived safety

Reingle, 2013

Community

Retained

Advantaged neighborhood

Stouthamer-Loeber, 2002

Community

Lacked practical relevance

Advantaged neighborhood

Wikström, 2000

Community

Redundant

Caring adults

Hart, 2007

Community

Retained

Religiosity

Hartman, 2009

Community/Indv.

Retained

75

Table 10 (cont’d)
Religious attendance frequency

Herronkohl, 2012

Community/Indv.

Adapted as an interview question

Religious commitment

Ryan, 2008

Community/Indv.

Adapted as an interview question

Religious attendance

Stevens, 2011

Community/Indv.

Adapted as an interview question

Social competence

Stepp, 2011

Individual

Redundant

Social skills

Van der put, 2014

Individual

Retained

Skills for interacting

Ayers, 1999

Individual

Redundant

Personal competence

Hjemdal, 2006

Individual

Redundant

Summary of final decisions

Adapted as an interview question (n = 16)
Retained (n = 22)
Redundant (n = 25)
Limited evidence or practical relevance (n = 14)
Total items (N = 77)

76

APPENDIX C: PFRJR ITEMS
Commitment to School/Education
Strong Family Management
Consistent Parenting
Strong Adult Bond(s)
Involvement in Organized Activities
Positive Personal Interest(s)
Religiosity
Positive Response to Authority
Prosocial Attitudes
Close Bonds with Positive Peer(s)
Low Aggression
Strong Social Skills
Perceived Safety
Access to Resources
Positive Adults

77

APPENDIX D: YLS/CMI ITEMS
Table 11. YLS/CMI Items
Prior/Current Offenses (α = .67)
1.
2.
3.
4.
5.

Substance Abuse (α = .79)

Three or More Prior Convictions
Two or more failures to comply
Prior Probation
Prior Custody
Three or More Current Convictions

Education (α = .69)

19.
20.
21.
22.
23.

Family Circumstances (α = .67)

6. Low Achievement
7. Problems with Teachers
8. Problems with Peers
9. Disruptive Classroom Behavior
10. Disruptive Behavior on School
Property
11. Truancy
Leisure & Recreation (α = .68)

24.
25.
26.
27.
28.
29.

Inadequate Supervision
Difficultly in Controlling Behavior
Inappropriate Discipline
Inconsistent Parenting
Poor Relations (Father-Youth)
Poor Relations (Mother-Youth)

Attitudes & Orientation (α = .75)

12. Lack of Organized Activities
13. Could Make Better Use of Time
14. No Personal Interests

Peer Relations (α = .75)

Occasional Drug Use
Chronic Drug Use
Chronic Alcohol Use
Substance Abuse Interferes with Life
Substance Use Linked to Offense(s)

30.
31.
32.
33.
34.

Not Seeking Help
Actively Rejecting Help
Defies Authority
Antisocial/Procriminal Attitudes
Callous, Little Concern for Others

Personality & Behavior (α = .67)

15. Lack of Positive Peer Acquaintances
16. Lack of Positive Friends
17. Some Delinquent Peer Acquaintances
18. Some Delinquent Friends

35.
36.
37.

Short Attention Span
Poor Frustration Tolerance
Verbally Aggressive/Verbally
Intimidating
38. Explosive Episodes
39. Physically Aggressive
40. Inadequate Guilt Feelings
41. Inflated Self –Esteem
42. *Unemployment/Not Looking for
Work
*Note: The variable Unemployment/Not looking for Work was omitted from the measure. This
item was not relevant to this sample due to average age and had no variation.

78

REFERENCES

79

REFERENCES

Andrews, D., Bonta, J., & Wormith, J. (2006). The recent past and near future of risk and/or need
assessment. Crime and Delinquency, 52 (1), 7-27.
Andrews, D., Guzzo, L., Raynor, P., Rowe, R., Rettinger, L., Brews, A., et al. (2012). Are the
major risk/need factors predictive of both female and male reoffending? A test with the
eight domains of the Level of Service/Case Management Inventory. International
Journal of Offender Therapy and Comparative Criminology, 56 (1), 113-133.
Andrews, D. A., Hoge, R. D., & Leschied, A. W. (2002). Youth level of service and case
management inventory. Toronto, Canada: Multi-Health Systems.
Andrews, D. A., Robinson, D., & Balla, M. (1996). Risk principle of case classification and the
prevention of residential placements: An outcome evaluation of the share the parenting
program. Journal of Consulting and Clinical Psychology, 54, 203–207.
Andrews, D. A., Robinson, D., & Hoge, R. D. (1984). Manual for the youth level of service
inventory. Ottawa, Ontario: Department of Psychology, Carleton University.
Baglivio, M. T., Jackowski, K., Greenwald, M. A., & Howell, J. C. (2014). Serious, violent, and
chronic juvenile offenders. Criminology & Public Policy, 13(1), 83-116.
Bernat, D. H., Oakes, J. M., Pettingell, S. L., & Resnick, M. (2012). Risk and direct protective
factors for youth violence: results from the National Longitudinal Study of Adolescent
Health. American Journal of Preventive Medicine, 43(2).
Blakely, R. J., & Cox, A. (1972). Evidence for short geomagnetic polarity intervals in the early
Cenozoic. Journal of Geophysical Research, 77(35), 7065-7072.
Borum, R., Bartel, P., & Forth, A. (2000). Manual for the Structured Assessment of Violence
Risk in Youth (SAVRY). Tampa: University of South Florida.
Chapman, J. F., Desai, R. A., Falzer, P. R., & Borum, R. (2006). Violence risk and race in a
sample of youth in juvenile detention: The potential to reduce disproportionate minority
confinement. Youth Violence and Juvenile Justice, 4(2), 170-184.
Chu, C. M., Goh, M. L., & Chong, D. (2016). The predictive validity of SAVRY ratings for
assessing youth offenders in Singapore: a comparison with YLS/CMI ratings. Criminal
Justice and Behavior, 43(6), 793-810.
Chu, C. M., Lee, Y., Zeng, G., Yim, G., Tan, C. Y., Ang, Y., ... & Ruby, K. (2015). Assessing
youth offenders in a non-Western context: The predictive validity of the YLS/CMI
ratings. Psychological Assessment, 27(3), 1013.

80

Chui, W. H., & Chan, H. C. O. (2012, August). An empirical investigation of social bonds and
juvenile delinquency in Hong Kong. In Child & Youth Care Forum (Vol. 41, No. 4, pp.
371-386). Springer US.
Costello, A. B., & Osborne, J. W. (2011). Best practices in exploratory factor analysis: four
recommendations for getting the most from your analysis. Pract Assess Res Eval 2005;
10. URL http://pareonline. net/getvn. asp, 10(7).
Cottle, C. C., Lee, R. J., & Heilbrun, K. (2001). The prediction of criminal recidivism in
juveniles: A meta-analysis. Criminal Justice and Behavior, 28, 367-294.
Craig, J. M., Baglivio, M. T., Wolff, K. T., Piquero, A. R., & Epps, N. (2017). Do social bonds
buffer the impact of adverse childhood experiences on reoffending? Youth Violence and
Juvenile Justice, 15(1), 3-20.
Crosnoe, R., Erickson, K. G., & Dornbusch, S. M. (2002). Protective functions of family
relationships and school factors on the deviant behavior of adolescent boys and girls
reducing the impact of risky friendships. Youth & Society, 33(4), 515-544.
Cuervo, K., & Villanueva, L. (2015). Analysis of risk and protective factors for recidivism in
Spanish youth offenders. International Journal of Offender Therapy and Comparative
Criminology, 59(11), 1149-1165.
Dawis, R. V. (2000). Scale construction and psychometric considerations. Handbook of applied
multivariate statistics and mathematical modeling, 65-94.
de Vogel, V., de Vries Robbé, M., de Ruiter, C., & Bouman, Y. H. (2011). Assessing protective
factors in forensic psychiatric practice: Introducing the SAPROF. International Journal
of Forensic Mental Health, 10(3), 171-177.
Dolan, M. C., & Rennie, C. E. (2008). The Structured Assessment of Violence Risk in Youth as
a predictor of recidivism in a United Kingdom cohort of adolescent offenders with
conduct disorder. Psychological Assessment, 20(1), 35.
Fabrigar, L. R., Wegener, D. T., MacCallum, R. C., & Strahan, E. J. (1999). Evaluating the use
of exploratory factor analysis in psychological research. Psychological Methods, 4(3),
272.
Fagan, A. A., Van Horn, M. L., Hawkins, J. D., & Arthur, M. W. (2007). Gender similarities and
differences in the association between risk and protective factors and self-reported
serious delinquency. Prevention Science, 8(2), 115-124.
Farrington, D. P. (2007). Advancing knowledge about desistance. Journal of Contemporary
Criminal Justice, 23(1), 125-134.

81

Fergus, S., & Zimmerman, M. A. (2005). Adolescent resilience: A framework for understanding
healthy development in the face of risk. Annu. Rev. Public Health, 26, 399-419.
Floyd, F. J., & Widaman, K. F. (1995). Factor analysis in the development and refinement of
clinical assessment instruments. Psychological Assessment, 7(3), 286.
Fougere, A., & Daffern, M. (2011). Resilience in young offenders. International Journal of
Forensic Mental Health, 10(3), 244-253.
Gammelgård, M., Koivisto, A. M., Eronen, M., & Kaltiala‐Heino, R. (2015). Predictive
validity of the structured assessment of violence risk in youth: A 4‐year follow‐up.
Criminal Behaviour and Mental Health, 25(3), 192-206.
Garmezy, N. (1985). Stress-resistant children: The search for protective factors. In
J. E. Stevenson (Ed.), Recent research in developmental psychopathology
(pp. 213-233). New York, NY: Elsevier Science.
Garmezy, N., Masten, A. S., & Tellegen, A. (1984). The study of stress and competence in
children: A building block for developmental psychopathology. Child Development, 97111.
Hardaway, C. R., McLoyd, V. C., & Wood, D. (2012). Exposure to Violence and
Socioemotional Adjustment in Low‐Income Youth: An Examination of Protective
Factors. American Journal of Community Psychology, 49(1-2), 112-126.
Hartman, J. L., Turner, M. G., Daigle, L. E., Exum, M. L., & Cullen, F. T. (2009). Exploring the
gender differences in protective factors: Implications for understanding resiliency.
International Journal of Offender Therapy and Comparative Criminology, 53(3), 249277.
Hawkins, J. D., & Weis, J. G. (1985). The social development model: An integrated approach to
delinquency prevention. Journal of Primary Prevention, 6(2), 73-97.
Hilterman, E. L., Bongers, I., Nicholls, T. L., & Van Nieuwenhuizen, C. (2016). Identifying
gender specific risk/need areas for male and female juvenile offenders: Factor analyses
with the Structured Assessment of Violence Risk in Youth (SAVRY). Law and Human
Behavior, 40(1), 82.
Hirschi, T. (1986). On the compatibility of rational choice and social control theories of crime.
The reasoning criminal: Rational choice perspectives on offending, 105-118.
Hoge, R. D., & Andrews, D. A. (2002). Youth Level of Service. Case Management
Inventory™(YLS/CMI™): User’s Manual. Toronto, Canada: MultiHealthServices.

82

Jimerson, S. R., Sharkey, J. D., Furlong, M. J., & O’Brien, K. M. (2004). Using the Santa
Barbara Assets and Risks Assessment to examine the ecology of youths experiencing
behavior problems. The California School Psychologist, 9(1), 99-113.
Jimerson, S. R., Sharkey, J. D., O'Brien, K. M., & Furlong, M. J. (2004). The Santa Barbara
Assets and Risks Assessment to predict recidivism among male and female juveniles: An
investigation of inter-rater reliability and predictive validity. Education and Treatment of
Children, 353-373.
Jones, N. J., Brown, S. L., Robinson, D., & Frey, D. (2016). Incorporating strengths into
quantitative assessments of criminal risk for adult offenders: The Service Planning
Instrument. Criminal Justice and Behavior, 42(3), 321-338.
Khanna, D., Shaw, J., Dolan, M., & Lennox, C. (2014). Does diagnosis affect the predictive
accuracy of risk assessment tools for juvenile offenders: Conduct Disorder and Attention
Deficit Hyperactivity Disorder. Journal of Adolescence, 37(7), 1171-1179.
Krohn, M. D., Lizotte, A. J., Bushway, S. D., Schmidt, N. M., & Phillips, M. D. (2014). Shelter
during the storm: A search for factors that protect at-risk adolescents from violence.
Crime & Delinquency, 60(3), 379-401.
Latessa, E. J., & Lovins, B. (2010). The role of offender risk assessment: A policy maker guide.
Victims and Offenders, 5(3), 203-219.
Lee, S. Y., & Villagrana, M. (2015). Differences in risk and protective factors between crossover
and non-crossover youth in juvenile justice. Children and Youth Services Review, 58, 1827.
Lipsey, M., Howell, J., Kelly, M., Chapman, G., & Carver, D. (2010). Improving the
effectiveness of juvenile justice programs: A new perspective on evidence-based practice.
Georgetown University. Washington DC: Center for Juvenile Justice Reform.
Lodewijks, H. P., Doreleijers, T. A., De Ruiter, C., & Borum, R. (2008). Predictive validity of
the Structured Assessment of Violence Risk in Youth (SAVRY) during residential
treatment. International Journal of Law and Psychiatry, 31(3), 263-271.
Lodewijks, H. P., de Ruiter, C., & Doreleijers, T. A. (2010). The impact of protective factors in
desistance from violent reoffending a study in three samples of adolescent offenders.
Journal of Interpersonal Violence, 25(3), 568-587.
Loeber, R., & Farrington, D. P. (2012). Advancing knowledge about direct protective factors that
may reduce youth violence. American Journal of Preventive Medicine,
Lösel, F., & Farrington, D. P. (2012). Direct protective and buffering protective factors in the
development of youth violence. American Journal of Preventive Medicine, 43(2), S8S23.

83

Luthar, S. S., Cicchetti, D., & Becker, B. (2000). The construct of resilience: A critical
evaluation and guidelines for future work. Child Development, 71(3), 543-562.
Masten, A. S. (2011). Resilience in children threatened by extreme adversity: Frameworks for
research, practice, and translational synergy. Development and Psychopathology, 23(02),
493-506.
McGrath, A., & Thompson, A. (2012). The relative predictive validity of the static and dynamic
domain scores in risk-need assessment of juvenile offenders. Criminal Justice and
Behavior, 39(3), 250-263.
Meyers, J., & Schmidt, F. (2008). Predictive validity of the Structured Assessment for Violence
Risk in Youth (SAVRY) with juvenile offenders. Criminal Justice and Behavior, 35(3),
344-355.
Miller, H. A. (2006). A dynamic assessment of offender risk, needs, and strengths in a sample of
pre‐release general offenders. Behavioral Sciences & the Law, 24(6), 767-782.
Milligan, S. (2010). Youth court statistics 2008/2009. Juristat: Canadian Centre for Justice
Statistics, 30(2), 1F.
Moffitt, T. E. (1993). Adolescence-limited and life-course-persistent antisocial behavior: a
developmental taxonomy. Psychological Review, 100(4), 674.
Molnar, B. E., Cerda, M., Roberts, A. L., & Buka, S. L. (2008). Effects of neighborhood
resources on aggressive and delinquent behaviors among urban youths. American Journal
of Public Health, 98(6), 1086-1093.
Netemeyer, R. G., Bearden, W. O., & Sharma, S. (2003). Scaling procedures: Issues and
applications. Sage Publications.
Nissen, L. (2006). Bringing strength-based philosophy to life in juvenile justice. Reclaiming
Children and Youth, 15(1), 40.
Onifade, E., Davidson, W., Campbell, C., Turke, G., Malinowski, J., & Turner, K. (2008).
Predicting recidivism in probationers with the Youth Level of Service Case Management
Inventory (YLS/CMI). Criminal Justice and Behavior, 35(3), 474-483.
Pardini, D. A., Loeber, R., Farrington, D. P., & Stouthamer–Loeber, M. (2012). Identifying
direct protective factors for nonviolence. American Journal of Preventive Medicine,
43(2), S28-S40.
Partners, O. (2000). Youth Assessment Screening Inventory (YASI). Ottawa, Ontario, Canada:
Author.

84

Pearl, N., Ashcraft, R. G., & Geis, K. A. (2009). Predicting juvenile recidivism using the San
Diego regional resiliency check-up. Federal Probation, 73, 46.
Penney, S. R., Lee, Z., & Moretti, M. M. (2010). Gender differences in risk factors for violence:
An examination of the predictive validity of the Structured Assessment of Violence Risk
in Youth. Aggressive Behavior, 36(6), 390-404.
Pollard, J. A., Hawkins, J. D., & Arthur, M. W. (1999). Risk and protection: Are both necessary
to understand diverse behavioral outcomes in adolescence?. Social Work Research, 23(3),
145-158.
Reingle, J. M., Jennings, W. G., Lynne-Landsman, S. D., Cottler, L. B., & Maldonado-Molina,
M. M. (2013). Toward an understanding of risk and protective factors for violence among
adolescent boys and men: A longitudinal analysis. Journal of Adolescent Health, 52(4),
493-498.
Rennie, C. E., & Dolan, M. C. (2010). The significance of protective factors in the assessment of
risk. Criminal Behaviour and Mental Health, 20(1), 8-22.
Rice, M., & Harris, G. (2005). Violent recidivism: Assessing predictive validity. Journal of
Consulting and Clinical Psychology, 63(5), 737-748.
Rogers, R. (2000). The uncritical acceptance of risk assessment in forensic practice. Law and
Human Behavior, 24(5), 595.
Royer-Gagnier, K., Skilling, T. A., Brown, S. L., Moore, T. E., & Rawana, J. S. (2016). The
Strengths Assessment Inventory–Youth Version: An evaluation of the psychometric
properties with male and female justice-involved youth. Psychological Assessment, 28(5),
563.
Ryan, J. P., Testa, M. F., & Zhai, F. (2008). African American males in foster care and the risk
of delinquency: The value of social bonds and permanence. Child Welfare, 87(1), 115.
Rutter, M. (1985). Resilience in the face of adversity. Protective factors and resistance to
psychiatric disorder. The British Journal of Psychiatry, 147(6), 598-611.
Sampson, R., & Laub, J. (1993). Structural variations in juvenile court processing: Inequality, the
underclass, and social control. Law and Society Review, 27(2), 285-311.
Schmidt, F., Campbell, M., & Houlding, C. (2011). Comparative analyses of the YLS/CMI,
SAVRY, and PCL:YV in adolescent offenders: A 10-year follow-up into adulthood.
Youth Violence and Juvenile Justice, 9(1), 23-42.
Schwalbe, C. (2007). Risk assessment for juvenile justice: A meta-analysis. Law and Human
Behavior, 31, 449-462.

85

Schwalbe, C., Fraser, M., Day, S., & Arnold, E. (2004). North Carolina Assessment of Risk
(NCAR): Reliability and predictive validity with juvenile offenders. Journal of Offender
Rehabiliation, 40(1/2), 1-22.
Shepherd, S., Luebbers, S., & Dolan, M. (2013). Gender and ethnicity in juvenile risk
assessment. Criminal Justice and Behavior, 40(4), 388-408.
Shepherd, S. M., Luebbers, S., & Ogloff, J. R. (2016). The role of protective factors and the
relationship with recidivism for high-risk young people in detention. Criminal Justice
and Behavior, 43(7), 863-878.
Shepherd, S. M., Luebbers, S., Ogloff, J. R., Fullam, R., & Dolan, M. (2014). The predictive
validity of risk assessment approaches for young Australian offenders. Psychiatry,
Psychology and Law, 21(5), 801-817.
Simourd, L., Hoge, R. D., & Andrews, D. A. (1991). The Youth Level of Service Inventory: An
examination of the development of a risk/needs instrument. In Presentation at the annual
conference of the Canadian Psychological Association, Calgary.
Singh, J. P. (2013). Predictive validity performance indicators in violence risk assessment: A
methodological primer. Behavioral Sciences & the Law, 31(1), 8-22.
Smith, C., Lizotte, A. J., Thornberry, T. P., & Krohn, M. D. (1995). Resilient youth: Identifying
factors that prevent high-risk youth from engaging in delinquency and drug use. Current
Perspectives on Aging and the Life Course, 4, 217-247.
Stouthamer-Loeber, M., Loeber, R., Wei, E., Farrington, D. P., & Wikström, P. O. H. (2002).
Risk and promotive effects in the explanation of persistent serious delinquency in boys.
Journal of Consulting and Clinical Psychology, 70(1), 111.
Thompson, A. P., & Pope, Z. (2005). Assessing juvenile offenders: Preliminary data for the
Australian adaptation of the youth level of service/case management inventory (Hoge &
Andrews, 1995). Australian Psychologist, 40(3), 207-214.
Turner, S., & Fain, T. (2006). Validation of the risk and resiliency assessment tool for juveniles
in the Los Angeles county probation system. Federal Probation, 70, 49.
van der Put, C. E (2015). Protective factors in male adolescents with a history of sexual and/or
violent offending: a comparison between three subgroups. Sexual Abuse, 27(1), 109-126.
van der Put, C. E., & Asscher, J. J. (2015). Female Adolescent sexual and nonsexual violent
offenders: A comparison of the prevalence and impact of risk and protective factors for
general recidivsim. Sexual Abuse, 27(1), 109-126.

86

van der Put, C. E., Creemers, H. E., & Hoeve, M. (2014). Differences between juvenile offenders
with and without substance use problems in the prevalence and impact of risk and
protective factors for criminal recidivism. Drug and Alcohol Dependence, 134, 267-274.
Vincent, G., Chapman, J., & Cook, N. (2011). Risk-needs assessment in juvenile justice:
Predictive validity of the SAVRY, racial differences, and the contribution of needs
factors. Criminal Justice and Behavior, 38(1), 42-62.
Walker, K., Bowen, E., & Brown, S. (2013). Psychological and criminological factors associated
with desistance from violence: A review of the literature. Aggression and Violent
Behavior, 18(2), 286-299.
Williams, A. B., Ryan, J. P., Davis-Kean, P. E., McLoyd, V. C., & Schulenberg, J. E. (2014).
The Discontinuity of Offending Among African American Youth in the Juvenile Justice
System. Youth & Society, 0044118X14551322.
Youngblade, L. M., Theokas, C., Schulenberg, J., Curry, L., Huang, I. C., & Novak, M. (2007).
Risk and promotive factors in families, schools, and communities: A contextual model of
positive youth development in adolescence. Pediatrics, 119(Supplement 1), S47-S53.
Zolkoski, S. M., & Bullock, L. M. (2012). Resilience in children and youth: A review. Children
and Youth Services Review, 34(12), 2295-2303.

87