International Journal of Educational Research 74 (2015) 1–12

Contents lists available at ScienceDirect

International Journal of Educational Research

journal homepage: www.elsevier.com/locate/ijedures

The role of resilience in assisting the educational

connectedness of at-risk youth: A study of service users and
non-users
Andrew J. Martina,* , Dorothy Bottrellb , Derrick Armstrongc ,
Marianne Mansoura , Michael Ungard, Linda Liebenbergd , Rebecca J. Colliea
a
School of Education, University of New South Wales, Australia
b
College of Education, Victoria University, Australia
c
[156_TD$IF]The University of the South Pacific
d
Dalhousie University, Canada

A R T I C L E I N F O A B S T R A C T

Article history: Many at-risk youth utilize support services, including educational, health, correctional, and
Received 27 January 2015 community/family/youth services. This study investigated young service users and non-
Received in revised form 11 September 2015 users, resilience, and ‘educational connectedness’ (academic engagement, academic
Accepted 15 September 2015
achievement, academic difficulty). Structural equation modeling with 249 young people
Available online xxx
(M = 16.5 years) showed that service users were lower in educational connectedness, while
resilience was positively associated with educational connectedness. There was also
Keywords:
evidence that resilience played a moderating or buffering role for service users. Specifically,
At-risk youth
Resilience
service users’ negative educational outcomes were attenuated once resilience was entered
Service use into modeling and there were significant interactions between service use and resilience
Engagement such that resilience explained more variance in the educational connectedness of service
Achievement users than non-service users.
ã 2015 Elsevier Ltd. All rights reserved.

1. Introduction

There are many young people who are deemed ‘at-risk’ on a number of fronts. They access support, assistance, and
advocacy from one or more services, including youth/family and community services (e.g., social worker, youth refuge),
correctional services (e.g., probation, detention), educational support services (e.g., school counselor, school-based
psychologist), and health services (e.g., pediatric support, addictions program, therapist). Relative to youth who do not
require these services, young service users are at greater risk of mental and physical health problems, substance abuse, risky
and suicidal behaviors, criminality, academic underachievement, and disengagement from the educational process
(Fergusson & Horwood, 2003). The present study focuses on the ‘educational connectedness’ of at-risk young people who use
various educational and other support services. It examines service users’ academic engagement, academic achievement and
academic difficulties (collectively referred to as ‘educational connectedness’) —and the accompanying association between
resilience and these outcomes and [157_TD$IF]the role of resilience in moderating the potentially negative effects of service use on these
outcomes. Fig. 1 demonstrates.

* Corresponding author at: School of Education, University of New South Wales, Sydney NSW 2052, Australia. Fax: +61 2 9385 1946.
E-mail address: andrew.martin@unsw.edu.au (A.J. Martin).

http://dx.doi.org/10.1016/j.ijer.2015.09.004
0883-0355/ ã 2015 Elsevier Ltd. All rights reserved.
2 A.J. Martin et al. / International Journal of Educational Research 74 (2015) 1–12
[(Fig._1)TD$IG]
Planning Manage Persist

Individual Family School Community

Academic
Engagement
Numeracy

Resilience Academic
Ecology Achievement

Literacy

Service Use Variables:

- Service User
- Sum of Services Used Academic Difficules
- Type of Service Used

Interacon Variables:
- Service User x Resilience
- Sum of Services Used x Resilience
- Type of Service Used x Resilience

Fig. 1. Service use, resilience, and educational connectedness.

Note: Models control for covariates (gender, age, ethnicity, SES).

2. At-risk youth and service use

Services and service providers may serve as facilitative environments to improve the outcomes of young service users.
Research shows that when at-risk youth are provided with resources in the form of psychosocial services, many do well in
spite of their personal characteristics such as low motivation, self-esteem, or sense of efficacy (Ungar, Liebenberg,
Dudding, Armstrong, & van de Vijver, 2013). However, relative to the empirical emphasis on personal and intra-individual
factors, there is relatively little attention given to the accessibility or availability of formal services that facilitate the
positive psychosocial development of at-risk young people. Ungar et al. (2013) report that the social welfare system and
services are largely “invisible” in research on the factors that enhance youth outcomes.
Approaching youth outcomes in this way goes beyond studying how an individual child overcomes problem behaviors
and psychopathology. Instead, they are considered in terms of the ecologies of service providers and children’s
communities (Armstrong, Stroul, & Boothroyd, 2005[158_TD$IF][2). What resources are provided to at-risk children, children’s access to
these resources, and how well these resources address any problem behaviors and/or psychopathology, are all factors that
influence children’s positive development when faced with adversity (see, for example, Dupree, Spencer & Bell, 1997).
Taken together, young people’s outcomes are as dependent on what is built inside them as they are on what is built around
them (Ungar, 2005). Educational and other services for at-risk youth may be considered part of these processes. For this
reason, the present study seeks to examine the educational outcomes of young service users and non-users.
To fully explore the role of service use in young people’s educational connectedness, service use is investigated in three
ways. The first is by examining the educational connectedness of service users and non-users; that is, considering youth who
use any type of educational, community, correctional, and health service. The second is by examining educational
connectedness as a function of the sum of services used; that is, considering whether the accumulation of multiple services
use impacts educational connectedness. The third is by examining the role of service use type in educational connectedness;
that is, for example, juxtaposing the educational connectedness of youth who use educational services with those using
other services.

3. Resilience and educational connectedness

The role of resilience in at-risk young people’s academic outcomes is well documented. Resilient youth tend to “sustain
high levels of achievement motivation and performance despite the presence of stressful events and conditions that place
them at risk of doing poorly in school and ultimately dropping out of school” (Alva, 1991, p. 19). Other factors positively
related to resilience are school engagement, pro-social behavior, mastery, autonomy, and problem solving skills (see
Fergusson & Horwood, 2003; Gore & Eckenrode, 1994; Masten, Best, & Garmezy, 1990; Rutter, 1985; Werner & Smith, 1988).
The present study continues this line of research by examining the role of resilience in the educational connectedness of
at-risk young service users.
Consistent with Ungar and Liebenberg (2008), resilience is defined ecologically as “the capacity of individuals to navigate
their way to resources that sustain well-being [and] the capacity of individuals’ physical and social ecologies to provide those
resources” (p. 12). Taking into account the individual attributes and social supports embedded within this definition, in the
 A.J. Martin et al. / International Journal of Educational Research 74 (2015) 1–12 3

context of at-risk service users we operationalize resilience in terms of young people’s developmental assets relevant to the
individual, family, school, and community (Ungar & Liebenberg, 2011). This view of resilience and educational
connectedness moves beyond student and school factors, to wider ecological settings that serve at-risk young people.
In seeking to understand young service users, resilience and educational connectedness, we consider three academic
dimensions that we believe reflect students’ educational connectedness: academic engagement, academic achievement, and
experience of academic difficulties. Collectively, these denote the extent to which students are immersed in the psycho-
educational processes relevant to school and schoolwork (academic engagement; Reschly & Christenson, 2012), the
acquisition and execution of domain-relevant content and skill (academic achievement; Hattie, 2009), and the experience of
any significant academic challenges, disruptions, and barriers (academic difficulty; Coleman & Hagell, 2007; Martin, 2014a,
2014b; Martin & Marsh, 2009). These three factors are important as ends in themselves (Reschly & Christenson, 2012;
Coleman & Hagell, 2007; Marsh, 2007) and also as means to important ends. For example, engagement and achievement are
predictive of school enjoyment, interest, post-school outcomes (Reschly & Christenson, 2012; Schmitt, 2012; Schunk,
Pintrich, & Meece, 2002; Stemler, 2012) and personal well-being outcomes such as life satisfaction and self-esteem (Chang,
McBride-Chang, Stewart, & Au, 2003; Marsh, 2007). Academic difficulties are predictive of social exclusion (MacDonald,
2007), persistent truancy (Pavis & Cunningham-Burley, 1999), and increased substance misuse, unemployment, and crime
(MacDonald, 2007; Roberts, 1995). To our knowledge, an encompassing approach to academic life that includes
multidimensional educational connectedness (academic engagement, achievement, difficulty) has not been investigated
among service users, nor has the relative salience of resilience in this process been the subject of investigation.

4. The potentially moderating (or buffering) role of resilience

Recent advances in the theory of resilience as an ecological construct (Ungar, 2012) emphasize the processes by which
individuals who face significant challenges interact with their environments to optimize personal outcomes (Ungar &
Liebenberg, 2011). This brings into consideration the potential role of factors that may moderate (or buffer) the effects of
service use on educational connectedness. There may be some factors that mitigate negative outcomes that service users
may exhibit. In this study, we explore the potentially moderating role of resilience. We do so by examining for any changes in
service users’ educational connectedness as a result of including resilience in our modeling and also by examining the
interaction between service use and resilience (e.g., whether there is a salient role of resilience for service users more than
non-users). Indeed, through testing the interaction, we are also able to represent the ‘classic’ resilience profile by way of
exploring the educational connectedness of service users (risk) who are also high in individual and social assets as assessed
by our resilience constructs. To the extent that there is a moderating or buffering role for resilience provides direction to
better assist at-risk youth and also extends the knowledge base in this area.

5. Aims of the present study

It has been established that young people accessing support services such as educational, youth/family, community,
correctional, and health services are more likely to demonstrate disrupted and less successful educational pathways and
outcomes (broadly conceived as a lack of ‘educational connectedness’). These at-risk youth are, therefore, an important focus
for educational, psychological, and social researchers and practitioners seeking to optimize academic trajectories and
connectedness. With particular focus on service users, the present study explores young people’s resilience from an
ecological perspective (encompassing developmental assets relevant to the individual, family, school, and community) and
its association with service users’ educational connectedness (academic engagement, achievement, and difficulties).

6. Methods

6.1. Sample and procedure

The sample comprises 249 young people drawn from three schools (two single-sex boys’ schools; one single-sex girls’
school from the systemic Catholic sector) and two community/youth service providers in high-needs areas in Sydney,
Australia. Schools were identified by the central education office as comprising students who were high support service
users. Although these schools were ‘mainstream’ district schools (not specialized schools for at-risk youth), they comprised
larger numbers of students using support services relative to other schools in the educational jurisdiction (see below for
service use statistics). Sampling from such schools thus provided sufficient numbers of service users, whilst also comprising
sufficient numbers of non-users to enable meaningful tests of service use/non-use effects. Community/youth centers were
located in inner city areas, specifically supporting at-risk youth—often those not attending school or other education and
training. Thus, in contrast to the school sample that comprised service users and non-users, all participants at the
community centers were service users and can be considered at some form of risk.
Half (51%) of the sample were service users (indicating ‘Yes’ in the previous six months to accessing: youth/family and
community services, correctional services, educational support services, and/or mental and other health services). For these
participants, service use was as follows: educational service use, n = 85 (34% of total sample); community service use, n = 50
(20%), child/youth service use, n = 12 (5%), mental health service use, n = 12 (5%), physical health service use, n = 11 (4%), and
4 A.J. Martin et al. / International Journal of Educational Research 74 (2015) 1–12

correctional service use, n = 11 (4%). The participants in schools typically utilize services if mandated to do so (schools often
take a role in organizing and/or supporting a student to access a service) and the participants recruited at the community
centers were already accessing a service (the community center). Thus, we can infer that non-users (49% of sample) are
participants who did not require or use services in the previous six months.
The mean age of participants was 16.5 years (SD = .84 years). School participants were aged 16–18 years (M = 16.30,
SD = .52) and community center participants were aged 16–20 years (M = 17.22, SD = 1.27). This is a time when young people
would typically participate in school or some other form of education and training (e.g., apprenticeship, vocational college,
technical training). Just over half (57%) of the respondents were male (school participants = 57%; community center
participants = 57%); 43% were female (school participants = 43%; community center participants = 43%). A total of 39% of the
sample was from a non-English speaking background (school participants = 42%; community center participants = 23%).
Three of the school participants and seven of the community center participants reported having a learning difficulty/
disability. On a 6-point scale (1 = no formal education to 6 = university education), the average parental education was 3.22
(SD = 1.38; school participants M = 3.38, SD = 1.41; community center participants M = 2.63, SD = 1.08).
For school surveys, the teacher or school counselor administered the instrument to students during class or counseling
session. Students were then asked to complete the instrument on their own and to return it at the end of the class or session.
Teachers and counselors were asked to provide assistance for any students with literacy needs. For surveys administered at
community/youth centers, respondents completed the survey in small groups with guided assistance from a research
assistant, or one-on-one with the research assistant if requiring more intensive literacy (or other) support. The data in this
(variable-centered) study are shared with another investigation that conducts person-centered cluster analysis exploring
different youth profiles in academic buoyancy, social support, and academic adversity (Collie et al., 2014).

6.2. Materials

6.2.1. Resilience
Using the Child and Youth Resilience Measure (CYRM; Liebenberg, Ungar, & Van de Vijver, 2012; Ungar & Liebenberg,
2011) and taking into account the individual attributes and social supports embedded within Ungar and Liebenberg’s (2008)
definition (see Introduction above), resilience was represented as a latent construct comprising: individual attributes and
developmental assets (9 items; e.g., “I am aware of my own strengths”; “I cooperate with people around me”), family support
(7 items; e.g., “My caregiver(s) watch me closely”; “My caregiver(s) stand(s) by me during difficult times”), school support (3
items; e.g., “I feel I belong at my school”; “Teachers at my school who see students hurting each other will do something to
stop them”), and community support (9 items; “e.g., “I have people I look up to”; “I am treated fairly in my community”).
Items were rated on a 1 (‘Does not describe me at all’) to 5 (‘Describes me a lot’) scale. Descriptive statistics, reliability, and
factor loadings are presented in Table 1.

6.2.2. Educational connectedness

Educational connectedness was operationalized by academic engagement, academic achievement, and academic
difficulties (the latter, inversely). Academic engagement was represented by three single-item indicators from the
engagement component of the Motivation and Engagement Scale—High School (MES-HS; Martin, 2010). These were:
planning (“In my schoolwork/studies I plan how I will do my assignments and study”), task management (“In my
schoolwork/studies I use my time well and try to study under conditions that bring out my best”), and persistence (“I persist
at my schoolwork/studies even when it is challenging or difficult”). Items were rated on a 1 (‘Disagree strongly’) to 7 (‘Agree
strongly’) scale. Academic achievement was based on participants’ reports of their results in nation-wide assessment of
literacy and numeracy (National Assessment Program in Literacy and Numeracy, NAPLAN) annually administered by the
Australian Curriculum and Assessment and Reporting Authority (ACARA). In the present study, a latent achievement factor
was formed by the two literacy and numeracy scores. Academic difficulties were assessed using the Academic Risk and
Resilience Scale (ARRS; Martin, 2013). Respondents were asked to indicate ‘Yes’ or ‘No’ to a set of academic adversity items.
These adversities include suspension, expulsion, repeating a grade, failing a subject, and the like. A total academic adversity
score was formed by tallying the number of ‘Yes’ responses. Descriptive statistics, reliability and factor loadings for these
educational connectedness measures are presented in Table 1.

Table 1
Descriptive statistics, reliability, and mean CFA loadings.

Mean (SD) Skew Kurtosis Cronbach’s a Mean CFA loading

Resilience 3.90 (0.69)
.89 .60 .85 .79
Academic engagement 4.43 (1.33)
.21
.46 .78 .83
Academic achievement 6.68 (2.11)
.46
.54 .93 .94
Academic difficulty 1.31 (1.77) 1.81 3.69 .71 1.00

Notes: Academic difficulty is a tally and thus, estimated as a single indicator; Skew and kurtosis are indicators of distribution; Cronbach’s alpha is the
measure of internal consistency (reliability) used in the present study; Mean CFA loadings are based on the CFAs reported in Section 7.
 A.J. Martin et al. / International Journal of Educational Research 74 (2015) 1–12 5

6.2.3. Service use

To fully explore service use we employed three operationalizations: service users (No/Yes), sum of service use, and types
of service use.

6.2.3.1. Service users. Service users were identified by indicating ‘Yes’ (thus, 0 = No, 1 = Yes) to accessing in the six months
before the survey one or more of the following services: youth/family and community services (e.g., social worker, foster
placement, youth refuge), correctional services (e.g., probation, community service work, conferencing or juvenile justice
supervision, detention), educational support services (e.g., school counselor, individual or personalized program, reading
recovery tutoring, school-based social worker, school-based psychologist), and health services (e.g., counselor, therapist,
psychologist or psychiatrist, addictions program, eating disorders program).

6.2.3.2. Sum of service use. Alongside the occurrence of service use (i.e., No/Yes, as above), it may be that the number of
services used has a bearing on educational connectedness. Accordingly, a variable was created that was the sum of ‘Yes’
responses to accessing in the six months before the survey one or more of the above-mentioned services: youth/family and
community services, correctional services, educational support services, and health services.

6.2.3.3. Service use type. Beyond the presence and count of service use, it was deemed important to consider service types.
Two variables were included in analyses for service use type: (a) educational support services (as described in Section 6.2.3.1
above) and (b) other services (comprising youth/family and community services, correctional services, and health services
(as described in Section 6.2.3.1)). For two reasons service type was operationalized in these ways. First, as the study is
educationally-focused, it was of interest to determine the role of education service use juxtaposed with other service use.
Second, given cell sizes for some non-educational services were too low to enable valid follow-up tests of interaction effects
(or, indeed, valid main effects), some data aggregation was necessary. Type of service use was indicated by respondents
answering ‘Yes’ or ‘No’ (0 = No, 1 = Yes) to using each of the two service groupings any time in the previous 6 months.

6.2.3.4. Covariates. It is important to understand the role of resilience in service users’ educational connectedness
controlling for socio-demographic and achievement covariates that may share variance with service use, resilience, and/or
academic outcome factors. Accordingly, socio-demographic covariates were included to partial out their potential influence
and better establish unique effects of service use and resilience on educational connectedness. Socio-demographic variables
were age, language background, gender, and socio-economic status (SES). Age was implemented as a continuous variable.
Gender was coded as 1 for boys and 0 for girls. Language background was coded by asking students if they spoke English (0)
or another language (1–non-English speaking background, NESB) at home. Students’ home postcode assessed SES and was
based on the Australian Bureau of Statistics (ABS) relative advantage/disadvantage index (higher scores reflected higher SES).
Gender (Martin & Marsh, 2008), age, language background, and SES factors (Martin, 2007, 2009; Rutter, 2006) are
significantly associated with academic resilience (and cognate constructs), and/or academic engagement and achievement
and were therefore considered important to include as covariates for this study.

6.3. Analyses

Psychometric properties were assessed using confirmatory factor analysis (CFA) to ascertain factor structure, reliability
(Cronbach’s alpha) to examine internal consistency, and skewness and kurtosis for distributional properties. The central
research issues were examined using structural equation modeling (SEM). Both CFA and SEM were conducted using Mplus
version 7.2 (Muthén & Muthén, 2013). For goodness of fit, the root mean square error of approximation (RMSEA) and the
comparative fit index (CFI) are used. For RMSEA, coefficients at or less than .08 and .05 denote good and excellent fits
respectively (see Schumacker & Lomax, 1996). For CFI, values at or greater than .90 and .95 denote good and excellent fits
respectively (McDonald & Marsh, 1990).
For most studies, missing data is a potentially important problem, particularly if the missing data exceeds 5% (e.g.,
Graham & Hoffer, 2000). A good deal of research has emphasized the limitations of traditional pairwise, listwise, and mean
substitution missing data approaches (e.g., Graham & Hoffer, 2000). This has led to suggestions to implement approaches
such as the Expectation Maximization (EM) Algorithm, a widely recommended approach to imputation for missing data that
are missing at random. In our study, less than 5% of data were missing and we imputed it using the EM Algorithm (with
LISREL 8.80; Jöreskog & Sörbom, 2006). We also explored alternative approaches which showed that the EM results were
very similar to those based on the traditional pairwise deletion methods for missing data—as we would expect when there
was so little missing data. Given the comparable solutions and the fact our sample was not large, we decided to analyze the
imputed dataset in order to retain as many cases as possible for the study.
For the central modeling, three sets of analyses were conducted. First, we examined service use, resilience, and their
interaction as the focal independent variables, alongside gender, age, language background, and SES as covariates—all
associated with the outcome variables (academic engagement, achievement, and difficulties). Second, we examined sum of
services used, resilience, and their interaction as the focal independent variables, alongside gender, age, language background,
and SES as covariates—all associated with the outcome variables (academic engagement, achievement, and difficulties).
Third, we examined educational services used, resilience, and their interaction as the focal independent variables, alongside
6 A.J. Martin et al. / International Journal of Educational Research 74 (2015) 1–12

gender, age, language background, and SES as covariates—all associated with the outcome variables (academic engagement,
achievement, and difficulties). For each interaction, we used the ‘xwith’ procedure in Mplus. This also required ‘random’
estimation and the ‘integration’ algorithm, which was used for all models for consistency. These specifications enable an
estimate of an interaction incorporating a latent variable (in this case, resilience). In each of the three sets of analyses,
modeling proceeded in a hierarchical fashion, with Model 1 (M1) including covariates and service use, Model 2 (M2) adding
resilience to M1, and Model 3 (M3) adding the interactions to M2. In Models 2 and 3, we therefore gain a sense of how
resilience may moderate the effects of service use. For Model 2, we can determine if there is any reduction in the effects of
service use following inclusion of resilience. For Model 3, we can see if there are distinct effects of resilience for students
using services compared to students not using services (i.e., the interaction effect). Given we are modeling interactions,
unstandardized coefficients are reported consistent with Mplus conventions.

7. Results

7.1. Descriptive statistics, distribution, reliability, and factor analysis

Table 1 presents scale means, standard deviations, skewness, kurtosis, and reliability. Skewness and kurtosis reflect
approximately normal distributions (Curran, West, & Finch, 1996). Reliability for all factors ranges between .71 and .93 (mean
Cronbach’s alpha = .82), suggesting internal consistency (Anastasi & Urbina, 1997; Sattler, 2001). The CFA comprised
covariates and service use (single-item indicators estimated without error), resilience, and three educational connectedness
factors. A first CFA operationalized service use via the ‘service user’ (No/Yes) variable and yielded a good fit to the data,
x2 = 153.35, df = 60, CFI = .942, RMSEA = .079; a second CFA operationalized service use via the ‘service use count’ variable and
yielded a good fit to the data, x2 = 146.47, df = 60, CFI = .940, RMSEA = .076; a third CFA operationalized service use via two
‘service use type’ variables (education service use; other service use) and also yielded a good fit to the data, x2 = 157.22,
df = 66, CFI = .940, RMSEA = .075. Factor loadings for resilience are acceptable, ranging from .70 to .90 (mean loading of .79), as
are loadings for academic engagement, ranging from .61 to .90 (mean loading of .83), and achievement, ranging from .90 to
.96 (mean loading of .94).
Table 2 presents correlations for all relationships between central factors. Service users are significantly lower in
resilience (r =
.32, p < .001), engagement (r =
.21, p < .001), achievement (r =
.29, p < .001), and significantly higher in
academic difficulties (r = .24, p < .001). The number of services used is significantly associated with lower resilience (r =
.40,
p < .001), engagement (r =
.28, p < .001), achievement (r =
.34, p < .001), and significantly greater academic difficulties
(r = .40, p < .001). Resilience is positively associated with engagement (r = .65, p < .001), achievement (r = .32, p < .001), and
negatively associated with academic difficulties (r =
.38, p < .001). Service use and the number of services used are
associated with lower resilience (r =
.32, p < .001 and r =
.40, p < .001, respectively). Interestingly, students using
educational services are not significantly lower in resilience or educational connectedness. Taken together, these CFA results
and correlations (in conjunction with distribution and reliability findings) provide a sound measurement and correlational
basis upon which to examine the hypothesized models that bring together all factors in multivariate models.

7.2. Exploring the roles of service use and resilience in educational connectedness

7.2.1. Service users

For service users, Model 1 explored whether participants using any type of service are lower or higher than non-users in
educational connectedness, controlling for socio-demographic covariates. Table 3 presents results. These findings show that
service users are significantly lower in academic engagement (B =
.48, p < .01) and academic achievement (B =
.85,
p < .001), and higher in academic difficulties (B = .83, p < .001). Model 2 included resilience in the analyses. These findings
show that resilience is positively associated with academic engagement (B = 1.72, p < .001) and academic achievement
(B = .84, p < .05), and negatively associated with academic difficulties (B =
1.40, p < .001). For all three dependent measures,
the inclusion of resilience reduces the effects of service use, with service use declining to non-significance for academic
engagement and declining by a significance level (.001–.01) for academic achievement and academic difficulties, signaling
some moderation as a function of resilience. Model 3 included the interaction of service use and resilience. These findings

Table 2
Correlations from CFAs: central factors.

Service user Service use count Educational service use Other service use Resilience Engage Achieve Difficulty
[154_TD$IF](N/Y) [15_TD$IF](N/Y) [15_TD$IF](N/Y)
Resilience
.32***
.40***
.11
.38*** 1.00
Engagement
.21***
.28*** .01
.33*** .65*** 1.00
Achievement
.29***
.34***
.06
.38*** .32*** .51*** 1.00
Difficulty .24*** .40*** .07 .39***
.38***
.41***
.45*** 1.00

Notes: *p < .05, **p < .01, ***p < .001; correlations reported are those in the substantive model comprising service use variables, resilience, and educational
connectedness variables; correlations reported here are zero-order bivariate associations, with partial correlations represented in the SEMs in Table 3.
 A.J. Martin et al. / International Journal of Educational Research 74 (2015) 1–12 7

Table 3
Service use, resilience, and educational connectedness: unstandardized coefficients.

Academic engagement Academic achievement Academic difficulties

M1 M2 M3 M1 M2 M3 M1 M2 M3
Service users
Gender (FM/M)
.03
.02
.03 .10 .13 .13 .07 .03 .04
Age
.07 .13 .12
.40**
.30
.26 .10
.08
.12
NESB (N/Y) .04
.17
.18 .48 .37 .35 .09 .28 .30
SES
.15
.18
.19 .20 .17 .19 .31 .37* .34*
Service User (N/Y)
.48**
.10
.09
.85***
.69**
.75** .83*** .56** .63**
Resilience 1.72*** 1.86*** .84* .08
1.40***
.52
Service  Resil
.21
1.32**^ 1.14*^

Sum of services used

Gender (FM/M)
.02
.01
.04 .11 .13 .15 .12 .08 .05
Age
.07 .12 .11
.40**
.32*
.32* .06
.08
.08
NESB (N/Y)
.01
.18
.17 .40 .33 .31 .19 .31 .34
SES
.13
.18
.18 .23 .20 .20 .26 .32 .32
Sum Services Used
.34***
.06
.10
.59***
.50***
.45*** .76*** .60*** .53***
Resilience 1.69*** 1.94*** .65 .32
1.10***
.61
Sum Serv  Resil
.23 .31
.49

Service types used

Gender (FM/M) .05 .02 .01 .19 .21 .25 .02 .01 .01
Age .05 .19 .15
.24
.17
.14
.08
.21
.17
NESB (N/Y)
.02
.19
.16 .39 .32 .27 .18 .32 .30
SES
.12
.17
.16 .24 .21 .20 .27 .32 .30
Educational (N/Y) .08 .19 .21
.08
.03
.05 .25 .16 .14
Other (N/Y)
.88***
.36*
.41*
1.38***
1.22***
1.17*** 1.52*** 1.22*** 1.22***
Resilience 1.67*** 1.71*** .36 .03
1.17***
.54
Educational  Resil .46 .34
1.19*^
Other  Resil
.46 1.07
.30

Notes: *p < .05, **p < .01, ***p < .001; ^ see Table 4 for follow-up tests of significant interaction effects; M1: ‘covariates’ + ‘service’, M2: ‘covariates’ + ‘
service’ + ‘resilience’, M3: ‘covariates’ + ‘service’ + ‘resilience’ + ‘service  resilience’.

show a significant interaction effect for academic achievement (B =
1.32, p < .01) and academic difficulties (B = 1.14, p < .05).
Follow-up tests are presented in Table 4. These show that the effect of resilience is not significant for non-service users but
significant for service users: for service users resilience was positively associated with academic achievement (B = 1.32,
p < .001) and negatively associated with academic difficulties (B =
1.81, p < .001). This is further suggestive of a moderating
or buffering function for resilience such that resilience seems especially salient and adaptive for service users.

7.2.2. Sum of services used

For the sum of services used, Model 1 examined whether a count of a participant’s use of any of the service types is
associated with educational connectedness, controlling for socio-demographic covariates. Table 3 presents results. These
findings show that the sum of services used is associated with lower academic engagement (B =
.34, p < .001), lower
academic achievement (B =
.59, p < .001), and higher academic difficulties (B = .76, p < .001). Model 2 then included
resilience in the analyses. These findings show that resilience is positively associated with academic engagement (B = 1.69,
p < .001) but not academic achievement, and negatively associated with academic difficulties (B =
1.10, p < .001). For all
three dependent measures, the inclusion of resilience reduces the effects of service use count, with service use count
declining to non-significance for academic engagement, signaling some moderation as a function of resilience. Model 3

Table 4
Follow-up on significant interactions: association between resilience (independent variable) and educational connectedness (dependent variable) for
service use groups: unstandardized coefficients.

Resilience Resilience Resilience

# # #
Academic engagement Academic achievement Academic difficulties
Not Service Users N/A .03
.23
Service Users N/A 1.32***
1.81***
Not Educational Service Users N/A N/A
1.06*
Educational Service Users N/A N/A
2.13***

Notes: *p < .05, **p < .01, ***p < .001; parameters in table represent unstandardized B coefficients for the association between resilience and each of the
outcome measures; N/A = no significant interaction between service use and resilience.
8 A.J. Martin et al. / International Journal of Educational Research 74 (2015) 1–12

included the interaction between sum of service use and resilience. Interestingly, these findings show no significant
interaction effect for any of the dependent measures.

7.2.3. Service use type

For service use type, Model 1 explored whether participants’ use of educational and other (community, health,
correctional) services is associated with educational connectedness, controlling for socio-demographic covariates. Table 3
presents findings. These show that educational service users are not significantly different from non-users in academic
engagement, academic achievement, and academic difficulties. Findings also show that the use of other services is associated
with lower academic engagement (B =
.88, p < .001), lower academic achievement (B =
1.38, p < .001), and higher
academic difficulties (B = 1.52, p < .001). Model 2 included resilience in the analyses. These findings show that resilience is
positively associated with academic engagement (B = 1.67, p < .001) but not academic achievement, and is negatively
associated with academic difficulties (B =
1.17, p < .001). For all three dependent measures, the inclusion of resilience
reduces the effects of other service use, with service use declining in significance level (.001–.05) for academic engagement,
signaling some moderation as a function of resilience. Model 3 included the interaction of service use and resilience. These
findings show a significant interaction effect of educational service use  resilience on academic difficulties (B =
1.19,
p < .05). Follow-up tests are presented in Table 4. These show that the effect of resilience is modest for educational non-
service users (B =
1.06, p < .05) and notably stronger for educational service users (B =
2.13, p < .001). This is further
suggestive of a moderating function for resilience such that resilience seems especially salient and adaptive for educational
service users.

8. Discussion

With particular focus on service users, the present study explored young people’s resilience from an ecological
perspective (encompassing individual, family, school, and community factors) and its role in service users’ educational
connectedness (encompassing academic engagement, achievement, and difficulties). The study found that on most
measures of service use, service users were lower than non-users in educational connectedness and that resilience was
positively associated with educational connectedness. Alongside the positive role of resilience in terms of its direct effects on
educational connectedness, there was evidence that resilience played a moderating role by way of (a) attenuating the
negative effects of service use once entered into modeling and (b) significant interactions such that resilience yielded a more
adaptive effect in the educational connectedness of service users than non-service users.

8.1. Findings of particular note

Four particularly salient findings emerged in relation to the proposed central factors: the lower levels of educational
connectedness for service users, the pattern of findings for different forms of service use, the role of resilience in educational
connectedness and moderating (or buffering) the negative outcomes of service users, and the ecological nature of resilience
as relevant to educational intervention.

8.1.1. Service use

The first noteworthy finding is that service users were, in the main, significantly lower in educational connectedness than
non-users. On the one hand, this may not be surprising: they are utilizing such services because they are at-risk and this risk
renders them vulnerable or compromised from developmental perspective (Ungar, 2005). On the other hand, it may have
been hoped that service use was a means by which educational connectedness was enhanced. The finding thus serves as a
reminder that adversity and its impacts may be so marked that service provision alone is insufficient.
To date, there has been an assumption that if the service or program exists, it will be sufficient for the individual to learn
the skills necessary to buffer stress and function adaptively (Ungar, Liebenberg, Dudding, Armstrong, & van de Vijver, 2013).
However, some research suggests that an individual’s personal agency and capacity to cope explain only part of the variance
in outcomes (Cicchetti, 2010; Sroufe, Egeland, Carlson, & Collins, 2005)—leading to conclusions that other factors are
important to consider. Some have suggested that a comprehensive system of social welfare substantially diminishes the risks
that threaten youth development, underscoring the importance of service provision alongside an individual’s personal
agency and capacity (Betancourt et al., 2010; Browne et al., 2001; Obrist, Pfeiffer, & Henley, 2010).
A second noteworthy finding relates to the pattern of results relevant to the different forms of service use tested. When
the sum of services used was included in modeling as a main effect, the role of resilience was somewhat diminished and
there were no significant interactions between sum of services and resilience—suggesting that the sum of services used
explained relatively more variance in outcomes than other operationalizations of service use. According to Coleman and
Hagell, “the more risks the individual is exposed to, or the greater the risk gradient, the greater the likelihood of a poor
outcome” (2007, p. 6). Similarly, it has been suggested that the presence of two risk factors may be sufficient to predict
academic failure (Lucio, Hunt, & Bornovalova, 2012). Although we did not assess the number of risks per se, the number of
services used may be considered a proxy. This study identified a more negative impact on educational connectedness for the
number of services used and thus it seems that in terms of educational outcomes, educational intervention may be
 A.J. Martin et al. / International Journal of Educational Research 74 (2015) 1–12 9

particularly critical as service use increases. Following from this, it would be interesting to explore the extent to which
multiple interventions within or across services have significant impact on educational connectedness.
Our tests for service types used provided further insight into who is and is not at educational risk and what services may
be relatively more effective than others in assisting educational connectedness. For example, it was encouraging to note that
educational service users were not markedly different from non-users in educational connectedness. Thus, educational
service use may be doing its part in terms of bringing about parity in educational outcomes. It is also probable that
educational services are used by a wider range of students, including those not at-risk on other fronts and this would also be a
factor explaining parity in educational connectedness. Indeed, in contrast, other service use (community, health,
correctional), once partialing out shared variance with educational service use (i.e., entered into the model alongside
educational service use), was consistently associated with lower educational connectedness.
The role of service providers in young people’s resilience also brings into consideration recent work examining
‘community resilience’ (Ungar, 2012). Communities are potentially supportive ecologies for youth who are at risk.
Community resilience refers to the extent to which various forms of capital are available to residents. Young people will fare
better in families that reside in communities with sufficient access to human capital (e.g., opportunities to work and gain
knowledge/education), social capital (e.g., support networks, sense of belonging and identity), financial capital (Bush &
Peterson, 2012; Masten & Coatsworth, 1998; Obrist et al., 2010; Seccombe, 2002), and physical capital (e.g., adequate/safe
housing, transportation, energy). Service providers play a potentially vital role in connecting young people and their families
to each of these forms of capital, thereby aiding the development of young people’s outcomes.

8.1.2. Resilience
A third noteworthy finding relates to the salient role played by resilience in its association with educational
connectedness (particularly academic engagement in all models) or moderating some negative outcomes for service users,
beyond the effects of service use on educational connectedness (and beyond the effects of socio-demographic covariates).
Where resilience did not remain a significant factor in the final models (Models 2 and 3), it was due to the presence of its
interaction with service use or due to the more salient role of the sum of services used (discussed above). This suggests the
potential importance of resilience as a focus for intervention. Further, because some evidence suggested that service users
may have especially benefited from resilience (more than non-users), there may be a place for service providers to directly
consider how they are promoting resilience in their stakeholders. To the extent that they do this, they are likely to contribute
to young people’s educational connectedness.
How service providers promote young service users’ resilience leads to the fourth noteworthy finding: the ecological
multidimensionality of resilience and its implications for educational connectedness. Consistent with Ungar et al. (2013), we
conceptualized resilience from an ecological perspective, comprising individual, family, school, and community dimensions.
When approached from this multidimensional perspective, there can be improved targeting of the elements within this
resilience constellation. There can also be better identification of specific elements that may represent a relative strength to
sustain or relative weakness to address. For example, at school, the young person may have good literacy, numeracy, and
relationships with peers and teachers; these would be sustained. On the other hand, there may be particular risks in the
family context that need close and ongoing attention.

8.1.3. Addressing educational connectedness directly

Alongside attention to service providers and resilience as two factors in the process underpinning educational
connectedness, there is also a place for direct educational intervention to enhance educational connectedness. Again, this is
where a multidimensional approach to educational connectedness is helpful. For example, understanding that at-risk
service users may have lower levels of prior achievement (one of the indicators of educational connectedness), allows
educators to consider whether additional instruction and skill development in literacy and numeracy (just two important
achievement foundations) is important to nurture academic functioning (Liem & Martin, 2013). Or, self-regulatory strategies
(e.g., Zimmerman, 2002) to address academic engagement (another indicator of educational connectedness) such as
planning, task management and persistence may be a useful focus for intervention for at-risk youth.
In researching and addressing the role of resilience in educational connectedness, consideration might also be given to
cognate factors that have come into recent attention, namely, the roles of academic buoyancy and academic resilience.
Academic buoyancy is defined as students’ ability to deal with everyday academic setback (i.e., minor adversity; Martin &
Marsh, 2009 see also Malmberg, Hall, & Martin, 2013; Putwain, Conners, Symes & Douglas-Osborn, 2012; Putwain & Daly,
2013). Academic resilience is defined as an ability to deal with acute and/or chronic academic setback and challenge (i.e.,
major adversity; Martin, 2013; Martin & Marsh, 2009). At-risk youth experience both types of adversity (minor and major) to
a greater extent than other youth (Martin, 2014a, 2014b) and this negatively impacts their educational outcomes (Martin &
Marsh, 2008). Thus, alongside addressing resilience from an ecological perspective through family, school and community
approaches, attention might also be directed to enhancing academic buoyancy and resilience.

8.2. Limitations and future directions

There are some limitations to consider when considering the present findings. First, the data are self-reported and thus
future research should collect ‘objective’ data from those within the young person’s resilience ecology: family, school, and
10 A.J. Martin et al. / International Journal of Educational Research 74 (2015) 1–12

community. Second, the data are cross-sectional; longitudinal data are needed to further support claims about resilience
predicting educational connectedness and exploring the extent to which lower resilience leads to greater service use.
Further, resilience operates as part of a process (Morales, 2000; Rutter, 1987) and so it is likely that reduced educational
connectedness depletes resilience and longitudinal data are necessary to test this ordering of factors. Third, we were unable
to determine which services were mandated and which were non-mandated. Further research should code for this and test
whether this has a bearing on young people’s outcomes. Fourth, alongside quantitative research, qualitative methods would
shed additional light on the processes relevant to service use, resilience, and educational connectedness.
Fifth, whereas our study adopted a variable-centered approach, there is merit in person-centered approaches to resilience
such that we investigate resilient (or non-resilient) young service users and the factors that comprise their profile (Masten,
2001). Sixth, in future efforts to shed light on service users, we suggest larger and broader samples be recruited. Our sample
was not large (however, power calculations on our final SEMs comprising 8 and 9 independent variables at p < .05 with
N = 249 yielded power values of .97 and .99 for explained variance of .10 and .25, respectively) and so we would recommend
expanding on these numbers and capturing a wider range of service users. Seventh, we recognize that school and community
center samples are quite different groups, with the possibility that some sites might not be representative on some key
factors (e.g., service use type) or that differences in socioeconomic status between sites might disproportionately impact
results. We did our best to guard against this by also modeling “low-level” services such as educational services that students
in all school types and sites are likely to utilize. We also included SES and language background as covariates in all models,
thereby partialling out variance in outcome measures that is attributable to SES-related constructs. Notwithstanding these
efforts, sample bias is important to consider when interpreting findings. Finally, taking into account these factors and the
present findings, there is a place for systemic intervention research within the service sector that seeks to assess different
aspects of the process (e.g., nature and length of service use, individual youth factors, family factors, academic skill and
issues) in order to determine the most viable channels for assisting at-risk young people in their academic journey.

9. Conclusion

Many youth are deemed at-risk and require support, assistance, and advocacy from one or more support services. Primary
aims of these services are to enhance young people’s life effectiveness and to maintain positive connections to “normative”
institutions such as school. The present study investigated the roles of service use and resilience in young people’s
educational connectedness (academic engagement, achievement, and difficulty). We found that service users tended to be
lower than non-users in educational connectedness. It was also the case that resilience was positively associated with
educational connectedness. In addition, resilience played a moderating role in that it seemed to yield a more positive effect in
the educational connectedness of service users than non-service users. These findings hold implications for service providers
and educators as they seek to assist at-risk youth in efforts to enhance their academic pathways.
The findings also open up new questions and avenues for further research. As noted earlier, this includes qualitative
research to better understand contexts and mechanisms in the resilience and educational connectedness process, person-
centered research to identify thriving and struggling profiles of youth on connectedness and resilience dimensions, and
larger and broader samples that can shed further light on service use and service users. There is also scope for intervention
research that builds on our findings. This work would seek to promote resilience in service users’ lives and then examine how
this impacts educational outcomes. From a more systemic perspective, it is important to recognize that young people often
rely on more than one service and service provider at any given time. Thus, research might also investigate the effect of
coordinated services (where, for example, mental health, community, and educational services are integrated or where there
is a case manager coordinating all three for the young service user) and non-coordinated services (where, for example,
mental health, community, and educational services operate independently and with no personnel to integrate or coordinate
them) on at-risk service users’ educational outcomes. Taken together, then, the findings presented here not only contribute
to current understanding of at-risk young service users, they also identify exciting areas for further research to assist these
young people as they navigate their complex and challenging worlds.

Acknowledgements

Thanks are extended to the Spencer Foundation (funding the research), participating schools and community centers, and
Julian Wood for data collection.

References

Alva, S. A. (1991). Academic invulnerability among Mexican-American students: the importance of protective and resources and appraisals. Hispanic Journal
of Behavioral Sciences, 13, 18–34. http://dx.doi.org/10.1177/07399863910131002.
Anastasi, A., & Urbina, S. (1997). Psychological testing. Upper Saddle River, NJ: Prentice Hall.
Armstrong, M. I., Stroul, B. A., & Boothroyd, R. A. (2005). Intercepts of resilience and systems of care. In M. Ungar (Ed.), Handbook for working with children and
youth: Pathways to resilience across cultures and contexts (pp. 387–404). Thousand Oaks, CA: Sage.
Betancourt, T. S., Borisovea, I. I., Williams, T. P., Brennan, R. T., Whitfield, T. H., de la Soudiere, M., et al. (2010). Sierra Leone’s former child soldiers: a follow-up
study of psychosocial adjustment and community reintegration. Child Development, 81, 1076–1094.
Browne, G., Roberts, J., Byrne, C., Gafni, A., Weir, R., & Majumdar, B. (2001). The costs and effects of addressing the needs of vulnerable populations: results of
10 years of research. Canadian Journal of Nursing Research, 33(397), 65–76.
 A.J. Martin et al. / International Journal of Educational Research 74 (2015) 1–12 11

Bush, K. R., & Peterson, G. W. (2012). Parent-child relationships in diverse contexts. In G. W. Peterson, & K. R. Bush (Eds.), Handbook of marriage and family (pp.
275–302). New York, NY: Springer.
Chang, L., McBride-Chang, C., Stewart, S., & Au, E. (2003). Life satisfaction, self-concept, and family relations in Chinese adolescents and children.
International Journal of Behavioral Development, 27, 182–189. http://dx.doi.org/10.1080/01650250244000182.
Cicchetti, D. (2010). A developmental psychopathology perspective on bipolar disorder. In D. J. Miklowitz, & D. Cicchetti (Eds.), Understanding 400 bipolar
disorder: a developmental psychopathology perspective (pp. 1–32). New York, NY: 401 Guilford.
Coleman, J., & Hagell, A. (2007). The nature of risk and resilience in adolescence. In J. Coleman, & A. Hagell (Eds.), Adolescence, risk and resilience (pp. 1–16).
West Sussex: John Wiley & Sons Ltd..
Collie, R.J., Martin, A.J., Bottrell, D., Armstrong, D., Ungar, M., & Liebenberg, L., 2014. Social support, academic adversity, and academic buoyancy: a person-
centered analysis. Submitted for publication.
Curran, P. J., West, S. G., & Finch, J. F. (1996). The robustness of test statistics to nonnormality and specification error in confirmatory factor analysis.
Psychological Methods 1, 16–29. http://dx.doi.org/10.1037/1082-989X.1.1.16.
Dupree, D., Spencer, M. B., & Bell, S. (1997). African American children. In G. Johnson-Powell, & J. Yamamoto (Eds.), Transcultural child development:
psychological assessment and treatment (pp. 237–268). New York, NY: John Wiley & Sons.
Fergusson, D. M., & Horwood, L. J. (2003). Resilience to childhood adversity: Results of a 21-year study. In S. S. Luthar (Ed.), Resilience and vulnerability:
adaptation in the context of childhood adversities (pp. 130–155). New York, NY: Cambridge University Press. http://dx.doi.org/10.1017/
CBO9780511615788.008.
Gore, S., & Eckenrode, J. (1994). Context and process in research on risk and resilience. In R. Haggerty, L. Sherrod, N. Garmezy, & M. Rutter (Eds.), Stress, risk
and resilience in children and adolescents: processes, mechanisms and interventions New York, NY: Cambridge University Press.
Graham, J. W., & Hoffer, S. M. (2000). Multiple imputation in multivariate research. In T. D. Little, K. U. Schnable, & J. Baumert (Eds.), Modeling longitudinal and
multilevel data: Practical issues, applied approaches, and specific examples (pp. 201–218). Mahwah, NJ: Erlbaum.
Hattie, J. (2009). Visible learning. Oxford: Routledge.
Jöreskog, K. G., & Sörbom, D. (2006). LISREL 8.80. Chicago: Scientific Software International.
Liebenberg, L., Ungar, M., & Van de Vijver, F. R. R. (2012). Validation of the child and youth resilience measure-28 (CYRM-28) among Canadian youth with
complex needs. Research on Social Work Practice, 22, 219–226. http://dx.doi.org/10.1177/1049731511428619.
Liem, G. A. D., & Martin, A. J. (2013). Direct instruction and academic achievement. In J. Hattie, & E. Anderman (Eds.), International guide to student
achievement (pp. 366–368). Oxford: Routledge.
Lucio, R., Hunt, E., & Bornovalova, M. (2012). Identifying the necessary and sufficient number of risk factors for predicting academic failure. Developmental
Psychology, 48, 422–428. http://dx.doi.org/10.1037/a0025939.
MacDonald, R. (2007). Social exclusion, risk, and young adulthood. In J. Coleman, & A. Hagell (Eds.), Adolescence, risk, and resilience (pp. 143–164). London:
John Wiley.
Malmberg, L.-E., Hall, J., & Martin, A. J. (2013). Academic buoyancy in secondary school: exploring patterns of convergence in mathematics, science, English
and physical education. Learning and Individual Differences, 23, 262–266. http://dx.doi.org/10.1016/j.lindif.2012.07.014.
Marsh, H. W. (2007). Self-concept theory, measurement and research into practice. Leicester: British Psychological Society.
Martin, A. J. (2007). Examining a multidimensional model of student motivation and engagement using a construct validation approach. British Journal of
Educational Psychology 77, 413–440. http://dx.doi.org/10.1348/000709906X118036.
Martin, A. J. (2009). Motivation and engagement across the academic lifespan: a developmental construct validity study of elementary school, high school,
and university/college students. Educational and Psychological Measurement, 69, 794–824. http://dx.doi.org/10.1177/0013164409332214.
Martin, A. J. (2010). The motivation and engagement scale. Sydney: Lifelong Achievement Group. www.lifelongachievement.com.
Martin, A. J. (2013). Academic buoyancy and academic resilience: exploring ‘everyday’ and ‘classic’ resilience in the face of academic adversity. School
Psychology International, 34, 488–500. http://dx.doi.org/10.1177/0143034312472759.
Martin, A. J. (2014a). Academic buoyancy and academic outcomes: towards a further understanding of students with ADHD, students without ADHD, and
academic buoyancy itself. British Journal of Educational Psychology, 84, 86–107.
Martin, A. J. (2014b). The role of ADHD in academic adversity: disentangling ADHD effects from other personal and contextual factors. School Psychology
Quarterly, 29, 395–408. http://dx.doi.org/10.1037/spq0000069.
Martin, A. J., & Marsh, H. W. (2008). Academic buoyancy: towards an understanding of students’ everyday academic resilience. Journal of School Psychology,
46, 53–83. http://dx.doi.org/10.1016/j.jsp.2007.01.002.
Martin, A. J., & Marsh, H. W. (2009). Academic resilience and academic buoyancy: multidimensional and hierarchical conceptual framing of causes,
correlates, and cognate constructs. Oxford Review of Education, 35, 353–370. http://dx.doi.org/10.1080/03054980902934639.
Masten, A. S. (2001). Ordinary magic: resilience processes in development. American Psychologist 56, 227–238. http://dx.doi.org/10.1037/0003-
066X.56.3.227.
Masten, A., Best, K., & Garmezy, N. (1990). Resilience and development: contributions from the study of children who overcome adversity. Development and
Psychopathology, 2, 425–444. http://dx.doi.org/10.1017/s0954579400005812.
Masten, A. S., & Coatsworth, J. D. (1998). The development of competence in favorable and unfavorable environments: lessons from research on successful
children. American Psychologist 53, 205–220. http://dx.doi.org/10.1037/0003-066X.53.2.205.
McDonald, R. P., & Marsh, H. W. (1990). Choosing a multivariate model: noncentrality and goodness-of-fit. Psychological Bulletin, 107, 247–255. http://dx.doi.
org/10.1037/0033-2909.107.2.247.
Morales, E. E. (2000). A contextual understanding of the process of educational resilience: high achieving Dominican American students and the ‘resilience
cycle’. Innovative Higher Education 25, 7–22. http://dx.doi.org/10.1023/A:1007580217973.
Muthén, L. K., & Muthén, B. O. (2013). Mplus user’s guide. Los Angeles: Muthén n.
Obrist, B., Pfeiffer, C., & Henley, R. (2010). Multi-layered social resilience: a new approach in mitigation research. Progress in Developmental Studies, 10,
283–293.
Pavis, S., & Cunningham-Burley, S. (1999). Male youth street culture: understanding the context of health-related behaviors. Health Education Research 14,
583–596. http://dx.doi.org/10.1093/her/14.5.583.
Putwain, D. W., Connors, L., Symes, W., & Douglas-Osborn, E. (2012). Is academic buoyancy anything more than adaptive coping? Anxiety. Stress and Coping
25, 349–358. http://dx.doi.org/10.1080/10615806.2011.582459.
Putwain, D. W., & Daly, T. (2013). Do clusters of test anxiety and academic buoyancy differentially predict academic performance? Learning and Individual
Differences, 27, 157–162. http://dx.doi.org/10.1016/j.lindif.2013.07.010.
Reschly, A. L., & Christenson, S. L. (2012). Jingle, jangle, and conceptual haziness: Evolution and future directions of the engagement construct. In S. L.
Christenson, A. L. Reschly, & C. Wylie (Eds.), Handbook of research on student engagement (pp. 3–19). New York, NY: Springer.
Roberts, K. (1995). Youth and employment in modern Britain. Open University Press: Milton Keynes.
Rutter, M. (1985). Resilience in the face of adversity: protective factors and resistance to psychiatric disorder. British Journal of Psychiatry 147, 598–611. http://
dx.doi.org/10.1192/bjp.147.6.598.
Rutter, M. (1987). Psychosocial resilience and protective mechanisms. American Journal of Orthopsychiatry, 37, 317–331.
Rutter, M. (2006). Genes and behavior: nature-nurture interplay explained. London: Blackwell Publishing. http://dx.doi.org/10.1101/gad.1363606.
Sattler, J. M. (2001). Assessment of children: cognitive applications. San Diego: Jerome M. Sattler Publishing Inc..
Schmitt, N. (2012). Development of rationale and measures of non-cognitive college student potential. Educational Psychologist 47, 18–29. http://dx.doi.org/
10.1080/00461520.2011.610680.
12 A.J. Martin et al. / International Journal of Educational Research 74 (2015) 1–12

Schumacker, R. E., & Lomax, R. G. (1996). A beginner’s guide to structural equation modeling. Mahwah, NJ: Erlbaum. http://dx.doi.org/10.1080/
10705519609540025.
Schunk, D. H., Pintrich, P. R., & Meece, J. L. (2002). Motivation in education: theory, research, and application. Upper Saddle River, NJ: Merrill.
Seccombe, K. (2002). “Beating the odds” versus “changing the odds”: poverty, resilience and family policy. Journal of Marriage and Family 64, 384–394. http://
dx.doi.org/10.1111/j.1741-3737.2002.00384.x.
Sroufe, L. A., Egeland, B., Carlson, E. A., & Collins, W. A. (2005). The development of the person: the Minnesota study of risk and adaptation from birth to adulthood.
New York, NY: Guilford.
Stemler, S. E. (2012). What should university admissions tests predict? Educational Psychologist 47, 5–17. http://dx.doi.org/10.1080/00461520.2011.611444.
Ungar, M. (2005). Pathways to resilience among children in child welfare, corrections, mental health and educational settings: navigation and negotiation.
Child and Youth Care Forum 34, 423–444. http://dx.doi.org/10.1007/s10566-005-7756-6.
Ungar, M. (Ed.), ()The social ecology of resilience: a handbook of theory and practice. New York, NY: Springer. http://dx.doi.org/10.1007/978-1-4614-0586-3.
Ungar, M., & Liebenberg, L. (2008). Pathways to resilience: formal service and informal support use patterns among youth in challenging social ecologies. Research
manual. Halifax, NS: Resilience Research Centre, Dalhousie University.
Ungar, M., & Liebenberg, L. (2011). Assessing resilience across cultures using mixed-methods: construction of the child and youth resilience measure-28.
Journal of Mixed-Methods Research 5, 126–149. http://dx.doi.org/10.1177/1558689811400607.
Ungar, M., Liebenberg, L., Dudding, P., Armstrong, M., & van de Vijver, F. J. (2013). Patterns of service use, individual and contextual risk factors, and resilience
among adolescents using multiple psychosocial services. Child Abuse Negl., 37, 150–159 doi: org/10.1016/j.chiabu.2012.05.007.
Werner, E., & Smith, R. (1988). Vulnerable but invincible: a longitudinal study of resilient children and youth. New York, NY: Adams, Bannister & Cox.
Zimmerman, B. J. (2002). Achieving self-regulation: the trial and triumph of adolescence. In F. Pajares, & T. Urdan (Eds.), Academic motivation of adolescents
(pp. 1–27). Connecticut: Information Age Publishing.