critical appraisal

Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=cesw20

Download by: [University of Glasgow] Date: 12 September 2016, At: 08:05

European Journal of Social Work

ISSN: 1369-1457 (Print) 1468-2664 (Online) Journal homepage: http://www.tandfonline.com/loi/cesw20

Placement, protective and risk factors in the educational success of young people in care: cross- sectional and longitudinal analyses

Robert J. Flynn , Nicholas G. Tessier & Daniel Coulombe

To cite this article: Robert J. Flynn , Nicholas G. Tessier & Daniel Coulombe (2013) Placement, protective and risk factors in the educational success of young people in care: cross- sectional and longitudinal analyses, European Journal of Social Work, 16:1, 70-87, DOI: 10.1080/13691457.2012.722985

To link to this article: http://dx.doi.org/10.1080/13691457.2012.722985

Published online: 10 Sep 2012.

Submit your article to this journal

Article views: 521

View related articles

Citing articles: 6 View citing articles

Placement, protective and risk factors in the educational success of young people in care: cross-sectional and longitudinal analyses

Des facteurs de placement, de protection, et de risque dans le succès scolaire des jeunes placés: Analyses transversales et longitudinales Robert J. Flynn, Nicholas G. Tessier & Daniel Coulombe

In the present study, we formulated and tested a basic model of the educational success of

young people in out-of-home care. We used data from 2007 to 2008 and 2008 to 2009 on

a sample of 1106 young people in care in Ontario, Canada. The youths were 12�17 years of age; 56.24% were male and 43.76% female. The indicators of educational success in

both years were the youth’s average marks and the youth’s school performance in reading,

math, science and overall, as rated by his or her caregiver. Based on resilience theory and

on a model of the influence of maltreatment on educational achievement, our model

included four categories of predictors: control variables (youth gender and age and, in the

longitudinal analyses, the year 7 value of the year 8 dependent variable), three placement

types (foster, kinship care or group homes), three risk factors (previous repetition of a

grade in school, a health-related cognitive impairment index and a measure of

behavioural difficulties) and three protective factors (caregiver involvement in the

youth’s school, caregiver educational aspirations for the young person and the youth’s

total number of internal developmental assets). Cross-sectional and longitudinal

Correspondence to: Robert J. Flynn, School of Psychology and Centre for Research on Educational and

Community Services, University of Ottawa, 34 Stewart Street, Ottawa, ON K1N 6N5, Canada. Email:

[email protected]

# 2013 Taylor & Francis

European Journal of Social Work, 2013

Vol. 16, No. 1, 70�87, http://dx.doi.org/10.1080/13691457.2012.722985

hierarchical regression analyses provided mixed support for the proposed model. The

youth’s gender, level of behavioural difficulties and number of developmental assets, and

the caregiver’s educational aspirations for the young person, emerged as the most

consistent predictors of educational success. The implications and limitations of the

findings were discussed.

Keywords: Educational Success; Youth In Care; Risk And Protective Factors; Ontario

Looking After Children project

Dans cette étude, nous avons formulé et évalué un modèle de base du succès scolaire des

jeunes qui ont été placés en dehors de leur famille d’origine. Nous avons utilisé des

données dont la saisie s’est faite en 2007�2008 et 2008�2009 auprès d’un échantillon composé de 1106 jeunes placés dans la province d’Ontario au Canada. L’âge des jeunes

étaient entre 12 et 17 ans; 56.24% étaient de sexe masculin et 43.76% de sexe féminin.

Les indicateurs du succès éducatif chaque année étaient la moyenne des notes scolaires du

jeune ainsi que sa performance en lecture, mathématiques, science, et dans l’ensemble des

matières. Notre modèle, basé sur la théorie de la résilience ainsi que sur un modèle de

l’influence de la maltraitance sur le rendement scolaire, incluait quatre catégories de

prédicteurs : des variables de contrôle statistique (le sexe et l’âge et, dans les analyses

longitudinales, la valeur dans l’an 7 de la variable formant la variable dépendante dans

l’an 8), trois types de placement (des familles d’accueil, des familles de parenté, et des

foyers de groupe), trois facteurs de risque (le fait d’avoir redoublé à l’école, un indice

composé de plusieurs difficultés cognitives reliées à la santé, et un indice des difficultés du

jeune sur le plan du comportement), et trois facteurs protecteurs (l’implication du parent

d’accueil dans la vie de l’école du jeune, les aspirations du parent d’accueil envers le

jeune, et le nombre total des acquis internes de développement du jeune). Des analyses

de régression transversales et longitudinales ont fourni un soutien partiel au modèle

proposé. Le sexe, le niveau de difficultés de comportement, le nombre d’acquis de

développement, et les aspirations du parent d’accueil envers le jeune se sont révélés les

meilleurs prédicteurs du succès éducatif. Les implications ainsi que les limites des

résultats ont été explorés.

Mots clés: succès scolaire; jeunes placés; facteurs de risque et de protection; le projet

S’Occuper des enfants en Ontario

Introduction

Many young people in out-of-home care (hereafter, ‘in care’) in Europe, North

America and other regions experience educational difficulties, including cognitive

deficits, poor problem-solving and reasoning skills, inconsistent school attendance,

below-average academic performance and low scores on standardised tests of

academic achievement in reading, writing and mathematics (Flynn et al., 2004;

European Journal of Social Work 71

Jackson, 2007; Slade & Wissow, 2007; Trout et al., 2008). In the USA, Trout et al.

(2008) conducted a comprehensive review of the American educational research

conducted on young people in care and published in journals during the period of

1940�2006. The 29 studies reviewed described the academic status of a total of 13,401 young people in care; they had a mean age of 12.9 years, and 52% were male. On the

whole, the young people in care had a much higher level of risks in school

functioning than the general youth population, with frequent changes in schools and

high levels of grade retention, suspension and school dropout. The young people in

care were nearly three times as likely to be involved in special education as their age

peers in the general population, tended to score in the low to low-average range on

measures of academic achievement, and were often rated by their teachers as

academically at risk. Trout et al. (2008) suggested that many youths in care presented

academic deficits similar to those identified in reviews of the academic status of other

at-risk populations, including children with emotional and behavioural disorders and

maltreated children who had been reported to child welfare agencies.

In the UK, Jackson (2007) reviewed the progress made over the last 20 years in

improving the educational outcomes of young people in care. She found that

considerably better data on their academic status now existed and that coordination

had improved between local educational and child welfare services. In addition, the

issue of education for children in care had risen to the top of the policy agenda, with

local authorities now required to promote their educational attainment. Berridge

(2007) noted that there was recent evidence of a slight improvement in outcomes in

the UK. The proportion of young people in care obtaining one General Certificate of

Secondary Education (GCSE) or equivalent had increased from 53% to 60% between

2002 and 2005, but this improved level of achievement was still much lower than the

level of 96% on the same criterion in the general child population. Moreover, the

proportion of young people in care who had obtained five or more GCSEs had

increased only from 8% to 11% during 2002�2005, whereas in the general population the level had increased from 52% to 56%.

Jackson (2007) agreed with Berridge (2007) that there had been little attempt in

the UK to understand the basic reasons for the achievement gap, while disagreeing

with his view that the answer lay in the characteristics of the families of origin of

children in care rather than in weaknesses of the care system itself. Jackson (2007)

suggested that discussions of foster parent recruitment and selection had largely

ignored the large body of research that identified a strong link between children’s

academic performance and the educational level and expectations of their parents or

caregivers. She added that the problem in the UK of ensuring an adequate education

for young people in the care system also characterised the child welfare systems of

other English-speaking countries. This appears to be true for Australia (Cashmore

et al., 2007) and the USA (Trout et al., 2008), and also, as we shall see, of Canada. The

educational attainment gap also appears to characterise other countries, however,

such as Sweden (Vinnerljung et al., 2005) and Norway (Iversen et al., 2010). Weyts

(2004), cited in Iversen et al. (2010), for example, found that the reading skills of

72 R. J. Flynn et al.

Norwegian children in care were no better than those of children in care in England,

the Netherlands and Spain.

In Canada, the relatively few studies conducted to date on the educational

achievement of young people in care provide a sketch similar to the portrait in the

countries previously mentioned. In the province of Ontario, Flynn and Biro (1998)

found that children in care had higher rates of grade retention and school suspension

than their age peers in the general population. Flynn et al. (2004) compared the

ratings made by caregivers in Ontario of the educational performance of children and

adolescents in their care with the ratings made by parents in the general Canadian

population of their own children’s educational progress. Eighty per cent of the young

people in care aged 10�15 years and 78% of the children in care aged 5�9 years were rated by their foster parents as performing in the same range as the lowest third of the

children in the general population, who had been rated by their parents on the same

composite measure of reading, spelling, math and overall educational performance.

In more recent research in Ontario, Miller et al. (2008) pointed to some possible

reasons for the achievement gap. Sixty-eight per cent of the young people in care aged

10�15 years in their sample had changed schools three or more times for reasons unrelated to normal progression through the school system, and the percentage

repeating a grade was 16% among 5�9 year olds in care, 27% among 10�15 year olds and 32% among 16�20 year olds.

In Ontario, as elsewhere, academic difficulties seem especially prevalent among

boys in care. Miller et al. (2009) found that girls in care were less likely than boys to

undergo assessments for learning-related problems (58% vs. 79%) or to receive

special academic help at school (49% vs. 69%). Caregivers also rated the girls’ school

performance more highly: 24% of the girls were rated as performing ‘Very Well’ or

‘Well’ in written work, compared with 13% of the boys; 41% of the girls (vs. 28% of

the boys) were seen as doing ‘Very Well’ or ‘Well’ in reading; and 29% of the girls

(vs. 20%) were rated as doing ‘Very Well’ or ‘Well’ overall. Only in math were equal

proportions of boys (23%) and girls (24%) rated as performing ‘Very Well’ or ‘Well’.

The girls also tended to be more positive about education-related matters than the

boys: 37% (vs. 28%) said, for example, that they read ‘for fun’ every day, and 40%

(vs. 26%) aspired to attain one or more university degrees.

Recently, the Ontario Association of Children’s Aid Societies (OACAS, 2010), in

collaboration with 43 of its local member agencies, carried out a review of the files of

4694 Crown Wards or former Crown Wards (i.e., young people in relatively long-

term out-of-home care). The youths were 16�20 years of age and had been in school during 2008�2009. OACAS compared the results from 2008 to 2009 with those from a similar study conducted in 2006�2007. The results showed that the youths in care had results that fell well short of those of their age peers in the general population,

although some progress had been made in the two-year interval since the initial

study. The percentage of 16 and 17 year olds not attending any educational

programme (Ontario requires school attendance up to age 18) had declined from

14% to 7%. Graduation from secondary school had increased by 2%, from 42% to

European Journal of Social Work 73

44%, compared, however, with a larger increase in the general youth population,

from 75% to 79%. The number of former Crown Wards aged 18�20 who were enrolled in post-secondary education (PSE) had increased from 21% to 23%,

compared to 39% in the general population. Of those in PSE, 81% were now in

community colleges, including apprenticeship programmes (vs. 84% two years

earlier), compared with 19% in university (vs. 16% two years earlier).

The purpose of the present study was to formulate and test a basic model of

educational success among young people in care that included the standard control

variables of gender and age and selected placement, protective and risk factors. In

formulating the model, we were guided by two theoretical frameworks. First, we drew

upon Masten’s (2006) conceptualisation of resilience theory: ‘Resilience refers to

positive patterns of functioning or development during or following exposure to

adversity, or, more simply, to good adaptation in a context of risk’ (p. 4). Masten

(2006) noted that ‘Direct predictors of better outcomes often are described as assets

or resources’ (p. 6). As predictors of academic achievement, good examples of assets

would be high-quality parenting or higher IQ scores. In the risk-related child-welfare

context of providing care for formerly abused or neglected young people, assets may

be called protective factors because they appear to play an especially important role in

positive adaptation. Risk factors, on the other hand, are predictors of undesired

outcomes. In the context of educational performance, abusive or neglectful parenting

or severe poverty would be good examples. Masten (2006) suggested that a typical

‘short list’ of factors associated with resilience in children and youth includes

relationships and parenting (e.g., strong links with one or more effective parental

figures; high-quality parenting that provides affection, monitoring and expectations);

individual differences (e.g., learning and problem-solving skills; self-control of

attention, emotional arousal and impulses); and community context (e.g., effective

schools; positive organisations).

Second, we used the framework proposed by Slade and Wissow (2007), in which

maltreatment is hypothesised as influencing educational outcomes through two main

pathways. The first pathway consists of mental health problems stemming from abuse

or neglect, including disruptive classroom behaviours, suspensions or difficulties of

concentration and motivation. The second pathway comprises inadequate cognitive

stimulation at home, lower-quality informal and formal education, and poorly

developed academic skills in word knowledge, literacy and numerical reasoning. Slade

and Wissow (2007) further hypothesise that the maltreated youth’s mental health

difficulties and low academic skills raise his or her risk of not adhering to behavioural

norms at school, obtaining less support from teachers and classmates, doing poorly

on homework assignments and tests and ultimately performing inadequately in

school.

In the present study, we defined educational success in terms of two outcomes: the

youth’s average marks during the last year in school, and his or her school performance

as rated by the caregiver on a composite measure of reading, math, science and

overall performance. Based on the literature reviewed, we included four categories of

74 R. J. Flynn et al.

predictors in our basic model of educational success. The first category consisted of

the standard control variables of gender and age, although we also saw female gender

as a protective factor because of girls’ greater educational success than boys (Miller

et al., 2009). Regarding age, we had no expectation that older youths would perform

any better or worse than younger youths. The second category of predictors,

corresponding to Masten’s (2006) ‘community context’ factor, comprised the type of

placement in which the young person had been living, whether a foster home, kinship

care home or group home. In line with McClung and Gayle’s (2010) findings

regarding the role of placement type, we anticipated that youths living in smaller

settings (i.e., foster or kinship care homes) would succeed better in school than those

residing in larger settings (i.e., group homes). The third category of predictors

consisted of three risk factors that we believed would be negatively associated with

educational success: having previously repeated a grade in school (Flynn & Biro,

1998), a lower level of cognitive functioning (Masten, 2006) and a higher level of

behavioural difficulties (Slade and Wissow, 2007). The fourth category of predictors

comprised three protective factors suggested by Masten’s emphasis on the key role of

assets or resources in promoting better adaptation. Two parenting-related assets were,

respectively, a greater degree of involvement by the parental figure (caregiver) with

the young person’s school, and a higher level of aspirations on the part of the

caregiver regarding the young person’s eventual level of educational attainment. The

third resource was the young person’s level of internal developmental assets, chosen

because of prior evidence that a greater number of developmental assets is associated

with greater educational success both in the general population (Scales et al., 2006)

and in young people in care (Flynn & Tessier, 2011).

Method

Participants and Service Context

The sample consisted of 1106 young people in care, aged 12�17; 56.24% were male and 43.76% female. The young people had participated in both year 7 (2007�2008) and year 8 (2008�2009) of the Ontario Looking after Children (OnLAC) project (Flynn et al., 2006; Flynn et al., 2009), which annually monitors the service needs

and developmental outcomes of children and youth who have been in care for a

year or more in the province. The OnLAC project is mandated by the provincial

government in local Children’s Aid Societies (CASs) across Ontario to encourage

more data-based decision-making about children’s needs, improve the quality of the

substitute parenting they receive and enhance their short-term and long-term

outcomes.

At the time the data were gathered, child welfare services in Ontario were

provided or supervised by a network of 53 government-funded CASs, the number of

which was beginning to be reduced through amalgamations in a search for greater

efficiency and sustainability. There were approximately 18,500 children and youths in

out-of-home care, over half of whom were teenagers (Commission to Promote

European Journal of Social Work 75

Sustainable Child Welfare, 2010). Excluding older youths in supported transitional or

independent living, 80% of the days of care provided in Ontario in 2009�2010 were spent in family-based care (i.e., family foster care or kinship care), 15% in group care

and 5% in other settings (e.g., hospitals, youth justice settings or children’s mental

health settings). Approximately 40% of expenditures in child welfare in Ontario were

allocated to in-care services (Commission to Promote Sustainable Child Welfare,

2010).

Instrument

The child welfare worker responsible for a given young person in care administered

the data collection instrument from which all the measures in the present study were

taken, namely, the second Canadian adaptation of the Assessment and Action Record

from Looking after Children (AAR-C2-2006; Flynn et al., 2009). The AAR-C2-2006

consists of eight age-appropriate formats, each of which comprises a family of

instruments. Administration of the tool is done annually, in the form of a structured

conversational interview in which the young person in care (if aged 10 or over), his or

her caregiver and his or her child welfare worker take part. The information gathered

in the AAR-C2-2006 interview is used to carry out a major revision, each year, of the

young person’s plan of care for the ensuing 12 months.

The AAR-C2-2006 consists of questions that cover nine areas: a background

section, completed mainly by the child welfare worker, that provides basic descriptive

information on the young person, caregiver and child welfare worker; seven sections,

rated mainly by the young person in care and his or her caregiver, that assess the

youth’s service needs and developmental outcomes in each of the Looking After

Children domains, namely, health, education, identity, social and family relation-

ships, social presentation, emotional and behavioural development and self-care

skills; and a final section, adapted from the work of the Search Institute (Scales et al.,

2000), in which the child welfare worker rates the young person’s acquisition of 40

different developmental assets (Flynn et al., 2009).

Criterion (Outcome) Measures

Two criterion measures of educational success were selected from the year 7 AAR-C2-

2006 data for the eventual cross-sectional analyses, and the same two measures were

taken from the year 8 data for the longitudinal analyses. The first measure was the

average marks that the young person in care had attained during the previous year in

school or during the last year he or she had been enrolled in school. The possible

values were 4 (B50%), 5 (51�60%), 6 (61�70%), 7 (71�80%), 8 (81�90%) or 9 (90�100%). The second measure was the young person’s school performance, as assessed by the caregiver on a four-item composite scale consisting of ratings of how

well the youth had done in school in years 7 and 8 on language and reading,

mathematics, science and overall. Each item was rated on a 3-point scale: 0 �Very

76 R. J. Flynn et al.

Poor or Poor; 1 �Average; and 2 �Very Well or Well. The total score on school performance could range from zero to eight.

Predictor Measures Control variables

Female gender was assigned the value of one and male gender the value of zero. The

young person’s age was his or her age in years as of the date that the AAR-C2-2006

interview had begun in OnLAC year 7.

Placement type

Three dichotomous variables were used to represent the type of placement setting in

which the young person in care had resided in OnLAC year 7: Foster Home (1 �Yes, 0 �Other), Kinship Care Home (1 �Yes, 0 �Other) or Group Home (1 �Yes, 0 �Other). In the regression analyses, the group home dichotomy served as the reference category and was thus omitted.

Risk-factor measures

The measures of the three risk factors were taken from the AAR-C2-2006. The first

was a dichotomy that indicated whether the young person in care had ever repeated a

grade in school (1 �Yes, 0 �No). The child welfare worker provided this information, with assistance, as needed, from the caregiver and young person. The

second risk-factor measure was a health-related Cognitive Impairments Index that we

created. The index consisted of the youth’s total number of cognitively related long-

term health conditions (out of a maximum of four), as rated by the youth’s child

welfare worker. These health conditions had lasted or been expected to last for 6

months or more, had been diagnosed by a health professional, and, by their very

nature, were likely to pose a challenge to the youth’s cognitive functioning. The child

welfare worker indicated which of the following health conditions the youth had:

Learning Disability (1 �Yes, 0 �No), Developmental Disability (1 �Yes, 0 �No), Attention-Deficit Disorder (1 �Yes, 0 �No) and Fetal Alcohol Syndrome (1 �Yes, 0 �No). The score on the index could range from zero to four.

The third risk-factor measure was the youth’s score on the Total Difficulties Scale of

the Strengths and Difficulties Questionnaire (SDQ; Goodman, 1997), which is

embedded in the AAR-C2-2006. The SDQ Total Difficulties Scale, composed of 20

behavioural items rated by the caregiver (0 �Not True, 1 �Somewhat True and 2 �True), covers the domains of emotional symptoms, conduct problems, hyperactivity/inattention and peer problems. The total score could range between 0

and 40.

Protective-factor measures

The three protective-factor measures were also taken from the AAR-C2-2006. The

first two were educationally relevant aspects of high-quality parenting on the part of

the caregiver. The Caregiver School Involvement Index consisted of the number of

European Journal of Social Work 77

school activities (out of a maximum of eight) in which the caregiver reported having

been involved during the current or last school year, such as volunteering in the

young person’s class or attending a school event in which the young person had

participated. The second protective-factor measure, Caregiver Aspirations, consisted

of the caregiver’s expressed hope that the young person in care would achieve a

certain level of education (1 �Primary or elementary school; 2 �Secondary or high school; 3 �Trade, technical, vocational school or business college; 4 �Community college or nursing school; 5 �University).

The third protective-factor measure consisted of the number of Internal

Developmental Assets (out of a maximum of 20) possessed by the young person

in care. The 20 asset items were rated by the youth’s child welfare worker

(1 �Yes, 0 �Uncertain or No). The internal assets covered four areas: the youth’s commitment to learning (e.g., ‘The young person is motivated to do well

in school’); the young person’s positive values (e.g., ‘The young person accepts and

takes personal responsibility’); the youth’s social competencies (e.g., ‘The young

person knows how to plan ahead and make choices’); and the young person’s positive

identity (e.g., ‘The young person feels that he/she has control over ‘‘things that

happen to me’’’).

Data Analysis Preliminary analyses

We began by assessing whether we would need to conduct multi-level analyses of our

data, in which the young people in care would be nested within their respective local

CASs. We found that this would not be necessary, as the amount of overall variance

accounted for by CASs in our two measures of educational success was very small

and statistically non-significant. We also evaluated whether we would need to

control for the particular geographic region (out of a total of six) in Ontario within

which the youth’s CASs was located. This, too, turned out to be unnecessary, as

geographic region was not significantly related to either measure of educational

success.

Hierarchical regression analyses

We related our two criteria of educational success, average marks and school

performance, to the four categories of predictors (controls, placement type, risk

factors and protective factors), in a series of hierarchical regression analyses. Two

cross-sectional analyses were conducted, in which the year 7 outcomes served as the

dependent variables. Similarly, two longitudinal analyses were carried out, in which

the year 8 outcomes were the dependent variables and in which the year 7 values of

the criterion variables were entered as control variables in step 1 (along with gender

and age). These longitudinal analyses tested the ability of our predictive models to

account for change from year 7 to year 8 in the two educational outcomes.

78 R. J. Flynn et al.

Results

Descriptive Results

Paired t-tests (not shown) revealed that, on the outcome of young people’s average

marks, there was no significant mean change from year 7 (M�6.54, SD �1.01) to year 8 (M�6.56, SD�1.04; t(784) �0.29, p�0.77). There was also no change on the other outcome, school performance, between year 7 (M�4.21, SD�2.37) and year 8 (M�4.32, SD�2.26; t(857) �1.42, p�0.16).

Table 1 presents basic descriptive information on the study variables. On some, the

effective sample size was B1106 because, for example, some young people were in

ungraded classrooms and thus had no data on the outcome of average marks. Other

youths were in placement settings that were neither foster, kinship, nor group homes,

such as mental health or juvenile justice residential settings and were eliminated from

the analyses. On other variables, the caregiver or child welfare worker were uncertain

about the young person’s previous scholastic history (e.g., regarding whether the

young person had previously repeated a grade).

Over half of the young people (52.1%) had no health-related cognitive

impairments, whereas 29.1% had only one, 13.7% had two, 4.3% had three and

0.8% had the maximum of four. The internal consistency coefficients (Cronbach’s

alphas) for four of the multi-item measures were excellent, in the 0.80s. On our two

constructed indexes, internal consistency was lower. It was acceptable (0.62) in the

case of the eight-item Caregiver Involvement in School Index but marginal on the

Table 1 Means (or percentages), standard deviations and Cronbach’s alphas for study

variables

Variable N Mean (or %) SD Cronbach’s Alpha

Outcomes Average marks*Year 7 894 6.51 1.03 � Average marks*Year 8 878 6.54 1.03 � School performance in Year 7 975 4.11 2.39 0.89 School performance in Year 8 942 4.28 2.27 0.88

Control variables Gender (1 �Female, 0 � Male) 1106 43.76% � � Age (in years) 1106 13.99 1.34 �

Placement type Foster home (1 �Yes, 0 � Other) 1058 72.40% � � Kinship care home (1 � Yes, 0 �Other) 1058 7.09% � � Group home (1 �Yes, 0 � Other) 1058 20.51% � �

Risk factors Previously repeated a grade (1 � Yes; 0 �No) 825 20.73% � � Cognitive impairment index 1106 0.73 0.91 0.45 SDQ total difficulties 1045 12.64 7.44 0.87

Protective factors Caregiver involvement in school 902 3.10 1.67 0.62 Caregiver aspirations 855 3.79 1.01 � Internal developmental assets 1106 12.75 5.10 0.88

European Journal of Social Work 79

brief Cognitive Impairments Index (Cronbach’s alpha �0.45). Despite this, the latter correlated significantly and in the expected direction with virtually all of the other

variables (see Table 2).

Predictors of Educational Success Inter-correlations

Table 2 showed, as anticipated, that the four measures of educational success were

positively and significantly inter-correlated. Other findings were also as expected: the

girls had better outcomes than the boys in both years; all 12 of the correlations

between the risk factors and outcomes were negative and statistically significant, and

10 of the 12 correlations between the protective factors and outcomes were positive

and significant. On the other hand, the correlations between the three types of homes

with the educational outcomes were weak, although in the expected direction, and

only half were statistically significant.

Hierarchical Regressions Average marks

Table 3 displays the results for the cross-sectional (left-hand panel) and longitudinal

(right-hand panel) regression models for the outcome of average marks. In the cross-

sectional model, the control (step 1), risk (step 3) and protective factors (step 4) all

accounted for statistically significant increments in the amount of variance accounted

for in year 7 average marks, and there was a trend in the same direction in the case of

the placement types (step 2).

The cross-sectional model as a whole accounted for 20.5% of the variance in year 7

average marks, with the risk and protective factors together explaining 17.2%. As

predicted, youths in the foster and kinship care homes had higher average marks than

those in group homes (step 2), although the relationship was modest. Once the three

risk factors had been entered into the model, however (at step 3), the beta coefficients

for the placement types were reduced to nearly zero, indicating that their association

with educational success was probably mediated by the risk factors. In the final model

(at step 4), the youth’s total number of internal developmental assets was the

strongest predictor of average marks. Caregiver aspirations and youth behavioural

difficulties were also statistically significant predictors, and there was a trend in this

direction in the instance of female gender and previous repetition of a grade in

school.

As previously noted, there was no significant mean change in the youths’ average

marks between year 7 and year 8. Thus, it was not surprising that the longitudinal

model explained little additional variance in year 8 average marks, once the role of

year 7 average marks and female gender (which was associated with improved marks

at all four steps) had been taken into account. Only two additional predictors*the caregiver’s level of involvement in school activities and the youth’s level of internal

developmental assets*were significantly associated with year 8 average marks.

80 R. J. Flynn et al.

Table 2 Inter-correlation matrix

Variables 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15.

1. Average marks*Year 7 � 2. Average marks*Year 8 43*** � 3. School performance*Year 7 64*** 35*** � 4. School performance * Year 8 37*** 60*** 48*** � 5. Gender (1 � F, 0 � M) 13*** 18*** 13*** 13*** � 6. Age (in years) �01 �03 00 �07* 02 � 7. Foster home (1 � Y, 0 � N) 01 04 07* 04 10*** �09** � 8. Kinship care home

(1 � Y; 0 � N) 04 08* 07* 06 02 �02 �45*** �

9. Group home (1 � Y, 0 � N) �04 �11** �13*** �09** �13*** 11*** �82*** �14*** � 10. Previously repeated a grade

(1 � Y; 0 � N) �10** �10** �13*** �13*** �04 �00 04 �06 �00 �

11. Cognitive impairment index �11*** �12*** �23*** �19*** �20*** �10*** �02 �12*** 10** 13*** � 12. SDQ total difficulties �27*** �19*** �37*** �24*** �10*** �01 �18*** �12*** 27*** 08* 31*** � 13. Caregiver involvement 10** 16*** 03 05 �01 �22*** �03 01 03 08* 11*** 04 � 14. Caregiver aspirations 26*** 16*** 35*** 28*** 15*** �06 07* 06 �13*** �13*** �34*** �29*** 03 � 15. Internal developmental assets 34*** 28*** 43*** 33*** 18*** �06* 21*** 11*** �30*** �06 �24*** �54*** 07* 29*** �

Note: Decimals omitted in correlations. Correlations are pair-wise; the number of cases on which the correlations were based varied between 652 (for the correlation between

Previously Repeated a Grade and Caregiver Aspirations) and 1106.

*p B 0.05 (2-tailed); **p B 0.01 (2-tailed); ***p B 0.001 (2-tailed).

E u

ro p

ea n

Jo u

rn a

l o f

S o cia

l W

o rk

8 1

Table 3 Beta coefficients in hierarchical regressions of average marks on control, placement, risk and protective variables

Outcome variable: Average marks in Year 7 Outcome variable: Average marks in Year 8 Cross-sectional regression (b) (N�531) Longitudinal regression (b) (N�488)

Predictors Step 1 Step 2 Step 3 Step 4 Predictors Step 1 Step 2 Step 3 Step 4

Average marks (Year 7) 0.42*** 0.42*** 0.40*** 0.37*** Female gender 0.15*** 0.14* 0.11** 0.07

$ Female gender 0.14*** 0.14*** 0.12** 0.11**

Age �0.03 �0.02 �0.05 �0.02 Age �0.02 �0.02 �0.03 �0.00 Foster home 0.10

$ 0.01 �0.04 Foster home �0.00 �0.02 �0.03

Kinship care home 0.11* 0.03 0.02 Kinship care home 0.05 0.03 0.03 Previously repeated a grade �0.11* �0.08$ Previously repeated a grade �0.02 �0.03 Cognitive impairments index �0.04 0.00 Cognitive impairments index �0.07 �0.07 SDQ total difficulties �0.28*** �0.13* SDQ total difficulties �0.02 0.02 Caregiver involvement in

school 0.03 Caregiver involvement in

school 0.10*

Caregiver aspirations 0.18*** Caregiver aspirations 0.02 Internal developmental assets 0.25*** Internal developmental assets 0.10*

DR 2

0.024* 0.009 $

0.095** 0.077** DR 2

0.214*** 0.003 0.006 0.018**

Note: *p B 0.05 (2-tailed); **p B 0.01 (2-tailed); ***p B 0.001 (2-tailed); $ p B 0.10 (2-tailed).

8 2

R .

J. F

ly n

n et

a l.

School performance

Table 4 shows the results for the two models for school performance. In the cross-

sectional model, the results were similar to those for average marks. The control, risk

and protective factors all accounted for statistically significant increments in the

variance, and the findings for step 2 were at the level of a trend. The cross-sectional

model as a whole explained 28.2% of the variance in year 7 school performance, with

the risk factor of behavioural difficulties especially important as a predictor. The

foster and kinship care homes were again modestly associated with better school

performance, but their beta coefficients were reduced to near zero once the three risk

factors (all of which were significantly and negatively associated with school

performance) had entered the model. In the final model, at step 4, the risk factor

of behavioural difficulties and the protective factors of caregiver aspirations and

youth internal developmental assets were all significantly related to school

performance.

In the longitudinal analyses, the lack of significant change in year-to-year school

performance meant that the model had relatively little power to predict change in

year 8 school performance, once year 7 school performance had been taken into

account. Interestingly, the girls had relatively consistent improved school perfor-

mance, at all four steps, whereas the older youths had fairly consistent worse

performance, at all four steps. In the final model (at step 4), caregiver aspirations for

the youth’s educational attainment was the only protective factor that was

significantly associated with better school performance.

Discussion

The findings provide some, albeit mixed, support for our basic model of educational

success and have implications for improving the educational success of young people

in care. First, in the cross-sectional regression model for average marks (i.e., in year

7), all four steps in the regression model were associated with increments in the

amount of variance accounted for, either at statistically significant levels (steps 1, 3

and 4) or at the level of a trend (step 2). The risk (9.5%) and protective factors

(7.7%) explained important proportions of the variance in average marks. Similarly,

with respect to school performance, statistically significant increments in the variance

accounted for were also found at steps 1, 2 and 4, and a trend in this direction was

seen at step 2, with the risk (17.4%) and protective factors (8.3%) accounting for

important increments in the amount of variance explained in school performance.

The fact that in the longitudinal analyses steps 2�4 accounted for much less additional variance in the two outcomes was no doubt due to the fact that, on

average, there was little or no change to explain.

Second, the girls in our sample, as predicted, experienced greater educational

success than the boys, on both outcomes. The sequential reductions in the size of the

beta coefficient for female gender at each step of the mode, especially pronounced in

the cross-sectional models, suggested that the girls’ educational advantage was partly

European Journal of Social Work 83

Table 4 Beta coefficients in hierarchical regressions of school performance on control, placement, risk and protective variables

Outcome variable: school performance in Year 7 Outcome variable: school performance in Year 8 Cross-sectional regression (b) (N�565) Longitudinal regression (b) (N�518)

Predictors Step 1 Step 2 Step 3 Step 4 Predictors Step 1 Step 2 Step 3 Step 4

School Performance (Year 7) 0.42*** 0.42*** 0.37*** 0.33*** Female gender 0.12** 0.12** 0.06 0.02 Female gender 0.10* 0.10* 0.09* 0.08

$

Age 0.02 �0.02 �0.00 0.03 Age �0.08* �0.08* �0.09* �0.07$

Foster home 0.09 $ �0.02 �0.06 Foster home �0.02 �0.05 �0.06

Kinship care home 0.11* �0.00 �0.02 Kinship care home 0.04 0.01 0.01 Previously repeated grade �0.11** �0.08* Previously repeated a grade �0.04 �0.03 Cognitive impairments index �0.11* �0.04 Cognitive impairments index �0.05 �0.03 SDQ total difficulties �0.37*** �0.22*** SDQ total difficulties �0.09$ �0.05 Caregiver involvement in

school 0.03 Caregiver involvement in

school 0.04

Caregiver aspirations 0.23*** Caregiver aspirations 0.10* Internal developmental assets 0.21*** Internal developmental assets 0.06

DR 2

0.016* 0.009 $

0.174*** 0.083*** DR 2

0.200*** 0.003 0.012 $

0.012*

Note: *pB0.05 (2-tailed); **pB0.01 (2-tailed); ***pB0.001; (2-tailed); $ pB0.10 (2-tailed).

8 4

R .

J. F

ly n

n et

a l.

mediated by lower levels of the risk factors and higher levels of the protective factors.

Despite this, the girls’ educational advantage tended to persist even after the risk and

protective factors had been introduced into the model.

Third, age had no relationship to educational success, in three of the four

regression models. In the fourth, however, age was consistently and negatively

associated with improved school performance, indicating perhaps that older youths

tend to fall further behind as the academic curriculum becomes more demanding.

Fourth, we found, as expected, that young people in foster and kinship care homes

had better educational outcomes than those in group homes, at least in the cross-

sectional models. This advantage was modest, however, and disappeared when the

risk factors had been taken into account. This was probably due to mediation and

may reflect a selection rather than a programme effect, with more turbulent youths

being selected out of foster or kinship care and into group care.

Fifth, it is clear that young people in care*of both genders*would benefit from lower levels of behavioural difficulties. The cross-sectional results for this latter

variable were congruent with Slade and Wissow’s (2007) hypothesis that poor

behavioural skills have a serious negative impact on the educational success of

maltreated youth. Previous repetition of a grade in school was also predictive of lower

educational success in the cross-sectional models, even when the youth’s level of

health-related cognitive functioning had been taken into account. This indicated that

effective action to prevent young people from repeating a grade is likely to pay

dividends.

Sixth, in all four models, the protective factors were associated with a statistically

significant increment in the amount of variance explained in the two indicators of

educational success. The young person’s level of internal developmental assets was

particularly important for educational success (except in the case of improved school

performance), which is congruent with the findings of Scales et al. (2006).

Seventh, it was noteworthy that caregiver attitudes and behaviour were related to

both indicators of educational success. Higher educational aspirations on the part of

caregivers were associated with better outcomes in three of the four regression

models, and caregiver involvement in a greater number of school activities predicted

significant improvement in the youth’s average marks. These results are consistent

with Jackson and Ajayi’s (2007) position that caregivers are an important resource for

improving educational outcomes and should be recognised as such. The recruitment

and training of carers should thus take explicit account of their influential role in the

educational achievement of young people in care (Dill et al., 2012; Ferguson &

Wolkow, 2012; Jackson & Cameron, 2012).

The present study had a number of limitations. The data were correlational in

nature, and the longitudinal analyses covered only a 12 month period. Also, the

effective sample size was considerably reduced due to incomplete data on some

variables. In addition, our index of health-related cognitive impairment was a rather

rudimentary measure of current cognitive functioning, and our measure of caregiver

aspirations for the young person in care was but a single item. Despite these

European Journal of Social Work 85

limitations, our analyses were based on relatively large samples and employed

comprehensive measures of two important predictors, the risk factor of behavioural

difficulties and the protective factor of internal developmental assets. In future, with

the accumulation of large samples and additional longitudinal data, we plan to carry

out multi-year analyses of the educational trajectories of subgroups of young people

in care, in the hope of identifying those in particular need of intervention. As Trout

et al. (2008) commented, many young people in care appear to need intensive and

effective assistance, if their educational careers are to be as successful as possible.

Acknowledgements

We gratefully acknowledge the financial support of this study by the Ontario

Association of Children’s Aid Societies and the Ontario Ministry of Children and

Youth Services. We also thank the many young people in care, caregivers and child

welfare workers in local Children’s Aid Societies who participated in the Ontario

Looking After Children project, from which the study data were drawn.

References

Berridge, D. (2007) ‘Theory and explanation in child welfare: Education and looked after children’,

Child and Family Social Work, vol. 12, no. 1, pp. 1�10. Cashmore, J., Paxman, M. & Townsend, M. (2007) ‘The educational outcomes of young people 4�5

years after leaving care: An Australian perspective’, Adoption and Fostering, vol. 31, no. 1, pp.

50�61. Commission to Promote Sustainable Child Welfare. (2010, December) Future directions for in-care

services in a sustainable child welfare system, Working paper no. 3, Toronto, ON, Commission

to Promote Sustainable Child Welfare.

Dill, K., Flynn, R. J., Hollingshead, M. & Fernandes, A. (2012) ‘Improving the educational

achievement of young people in out-of-home care’, Children and Youth Services Review,

vol. 34, pp. 1081�1083. doi:10.1016/j.childyouth.2012.01.031 Ferguson, H. B. & Wolkow, K. (2012) ‘Educating children and youth in care: A review of barriers to

school progress and strategies for change’, Children and Youth Services Review, vol. 34,

pp. 1143�1149. doi:10.1016/j.childyouth.2012.01.034 Flynn, R. J. & Biro, C. (1998) Comparing developmental outcomes for children in care with those of

other children in Canada. Children and Society, vol. 12, pp. 228�233. Flynn, R. J. & Tessier, N. G. (2011) ‘Promotive and risk factors as predictors of educational

outcomes in supported transitional living: Extended care and maintenance in Ontario,

Canada’, Children and Youth Services Review, vol. 33, pp. 2498�2503. Flynn, R. J., Dudding, P. M. & Barber, J. G. (eds) (2006) Promoting Resilience in Child Welfare,

Ottawa, ON, University of Ottawa Press.

Flynn, R. J., Ghazal, H., Legault, L., Vandermeulen, G. & Petrick, S. (2004) ‘Use of population

measures and norms to identify resilient outcomes in young people in care: An exploratory

study’, Child and Family Social Work, vol. 9, pp. 65�79. Flynn, R. J., Vincent, C. & Legault, L. (2009) User’s Manual for the AAR-C2-2006, Ottawa, ON,

Centre for Research on Educational and Community Services, University of Ottawa.

Goodman, R. (1997) ‘The strengths and difficulties questionnaire: A research note’, Journal of Child

Psychology and Psychiatry, vol. 38, pp. 581�586.

86 R. J. Flynn et al.

Iversen, A. C., Hetland, H., Havik, T. & Stormark, K. M. (2010) ‘Learning difficulties and academic

competence among children in contact with the child welfare sytem’, Child and Family Social

Work, vol. 15, pp. 307�314. Jackson, S. (2007) ‘Progress at last?’, Adoption and Fostering, vol. 31, no. 1, pp. 3�5. Jackson, S. & Ajayi, S. (2007) ‘Foster care and higher education’, Adoption and Fostering, vol. 31, no. 1,

pp. 62�72. Jackson, S. & Cameron, C. (2012) ‘Leaving care: Looking ahead and aiming higher’, Children and

Youth Services Review, vol. 34, pp. 1107�1114. Masten, A. S. (2006) ‘Promoting resilience in development: A general framework for systems of

care’, in Promoting Resilience in Child Welfare, eds R. J. Flynn, P. M. Dudding & J. G. Barber,

Ottawa, ON, University of Ottawa Press, pp. 3�17. McClung, M. & Gayle, V. (2010) ‘Exploring the care effects of multiple factors on the educational

achievement of children looked after at home and away from home: An investigation of two

Scottish local authorities’, Child and Family Social Work, vol. 15, pp. 409�431. Miller, M., Flynn, R. & Vandermeulen, G. (2008) Looking After Children in Ontario: Good Parenting,

Good Outcomes: Ontario Provincial Report (Year Six), Reports for 0�4, 5�9, 10�15, and 16� 20 year olds, Ottawa, ON, Centre for Research on Educational and Community Services,

University of Ottawa.

Miller, M., Vincent, C. & Flynn, R. (2009) Looking After Children in Ontario: Good Parenting, Good

Outcomes. Ontario Provincial Report (Year Seven), Report for 10�15 year olds, Ottawa, ON, Centre for Research on Educational and Community Services, University of Ottawa.

Ontario Association of Children’s Aid Societies (OACAS). (2010) Gateway To Success: Cycle Two,

OACAS survey of the educational attainment of Crown Wards and former Crown Wards,

ages 16 through 20, during the 2008�2009 academic year, Toronto, ON, Ontario Association of Children’s Aid Societies.

Scales, P. C., Benson, P. L., Leffert, N. & Blyth, D. A. (2000) ‘Contribution of developmental assets

to the prediction of thriving among adolescents’, Applied Developmental Science, vol. 4, pp.

27�46. Scales, P. C., Benson, P. L., Roehlkparain, E. C., Sesma, A. & van Dulmen, M. (2006) ‘The role of

developmental assets in predicting academic achievement: A longitudinal study’, Journal of

Adolescence, vol. 29, pp. 691�708. Slade, E. P. & Wissow, L. S. (2007) ‘The influence of childhood maltreatment on adolescents’

academic performance’, Economics of Education Review, vol. 26, pp. 604�614. Trout, A. L., Hagaman, J., Casey, K., Reid, R. & Epstein, M. H. (2008) ‘The academic status of

children and youth in out-of-home care: A review of the literature’, Children and Youth

Services Review, vol. 30, pp. 979�994. Vinnerljung, B., Öman, M. & Gunnarson, T. (2005) ‘Educational attainment of former child welfare

clients: A Swedish national cohort study’, International Journal of Social Welfare, vol. 14, pp.

265�276. Weyts, A. (2004) ‘The educational achievements of looked after children: Do welfare systems make a

difference?’, Adoption and Fostering, vol. 28, no. 3, pp. 7�19.

European Journal of Social Work 87