Pathways to educational aspirations: resilience as a mediator of proximal resources and risks

ABSTRACT

Educational aspirations play an important role in educational achievement, understanding more about the factors that shape aspirations should enable educators and others who support young people to better articulate educational ambitions and then to work towards their achievement. Given this, the current paper uses a form of integrative data analysis to examine two issues. First, it identifies that a social ecological measure of resilience mediates the impact which proximal risks and resources have upon young people’s educational aspirations. Second, it establishes that the linkages between these proximal risks and resources, resilience and educational aspirations are similar for young people facing high levels of adversity and young people on more or less normative developmental pathways. The paper draws on a data collected in the first wave of the Youth Transitions Study; a longitudinal mixed-methods investigation of the factors that influence transitions to adulthood for a group of vulnerable teenagers.

KEYWORDS:

• Resilience
• risk
• positive development
• educational aspirations

Introduction

It has long been recognised that academic ability does not explain all of the variance in educational achievement (Snyder et al. 2002; Strathdee and Engler 2012). Indeed, the risks of high school non-completion are not evenly distributed in the population (Strathdee and Engler 2012; Becker and Tuppat 2013). Internationally, children from ethnic minority backgrounds, including children with indigenous heritage, those from lower socio-economic status families and who face a range of other risks and challenges have lower rates of educational achievement than youth without these characteristics (Gray and Beresford 2002; Strand 2010; Easton 2013) and this differential pattern of achievement is not convincingly explained by differences in academic ability (Snyder et al. 2002; Strathdee and Engler 2012). New Zealand is no different in this regard. While it performs well in international comparisons of overall educational achievement, a significant minority of children still fail to achieve school qualifications commensurate with their abilities and face increased social and economic risks across the life course as a result (Advisory Group on Conduct Problems 2013). For instance, recent figures suggest that around 15% of all school leavers in New Zealand have no formal qualifications (Ministry of Education 2017). As is the case elsewhere, the children who are most likely to exit school prematurely in New Zealand face heightened risks in their families and neighbourhoods and who have reduced access to resources that support them in their schooling. Included in this group are ethnic minority youth, particularly Māori youth (the indigenous population), those from low socio-economic status households and who confront heightened risks in their families and neighbourhoods (Jacobsen et al. 2002). There are arguments that the school curriculum does not adequately cater for the particular needs of indigenous youth and that this explains lower rates of educational achievement by Māori (Alton-Lee 2003).

While academic ability may not be a consistent predictor of educational achievement, levels of educational achievement are strong predictors of successful transitions to adulthood as measured by indicators such as employment and income levels (Frønes 2010). Because of this powerful role that educational achievement plays in adult outcomes and because educational achievement is not explained solely by academic ability, there is considerable interest in understanding the factors that might increase the chances that young people will accumulate educational credentials commensurate with their abilities.

Aspirations as predictors of achievement

A promising area of work in this regard is research into the role of cognitive and behavioural orientations of students. This work explores the way that students think about themselves as learners and the behaviours they adopt in relation to learning tasks (Solomon and Rogers 2001). A wide range of cognitive and behavioural factors have been found to play a role. These include self efficacy (Bandura 1982), goals, feelings of helplessness and mastery (Covington 2000), optimism and hopefulness (Snyder et al. 2002) and an achievement orientation (Määttä et al. 2007). Each of these factors explain signficant amounts of variance in educational outcomes. These factors also typically show strong correlations with each other. For example, hope is positively connected to other motivational constructs, as well as to internal factors such as self-esteem, problem-solving ability, positive affective states, sense of control and educational performance (Snyder et al. 2002). Ashby and Schoon (2010) found that adolescent educational aspirations predicted later income, educational and occupational status in adulthood. Others have found a strong connection between hope and educational achievement when other factors such as intelligence, personality and prior educational achievement are controlled for (Day et al. 2010).

While factors such as hope, aspiration and achievement orientations have been linked to educational achievement and better outcomes in adulthood, less is known about the factors that influence these hopeful or positive achievement orientations. Further, much of the research on the connection between educational achievement orientations and later success has been conducted on adults rather than adolsecents, leaving unanswered questions about the extent to which patterns observed in adulthood may have been amenable to influence during childhood (Määttä et al. 2007). The current study seeks to address this gap by focusing on the connection between a range of proximal resources and risks known to be connected to educational achievement (such as resilience, prosociality, mental health status, peer issues, delinquency and minority/indigenous ethnic status) and the educational aspirations of young people.

Resilience as a mediator of aspirations and achievement

In addition to their own personal characteristics, children’s capacities to cope with adversity (resilience) depends upon the resources available to them in their families, neighbourhoods and the wider political, economic and social systems around them (Saewyc and Edinburgh 2010). The education system constitutes a key resilience resource that can facilitate positive development (Theron 2014). Conceptualised in this way, resilience is a set of social ecological resources that can offset the negative developmental effects of risks and which can influence the ways in which children think about themselves and their capacity to aspire to achieve educationally (Nicoll 2014). Rather than a static, individual trait, resilience is a set of resources that comprise the social environment around children (Ungar 2012). Conceived of in this way, resilience becomes relevant to efforts to increase the capacity of marginalised or vulnerable youth to participate and succeed in education because the resources around a young person can be harnessed along with their own capacities in the pursuit of enhanced educational outcomes.

Risk factors and educational aspirations and outcomes

There is a large body of research which examines the influence which individual risk factors have upon educational outcomes. For instance, the presence of mental health issues have been connected to lower levels of educational attainment (Kessler et al. 1995). Määttä, et al. (2007) found that higher levels of depression, along with lower levels of engagement in education predicted reduced educational achievement-related beliefs of students. Conduct and attention disorders (ADHD) also explain reduced rates of high school graduation (Breslau et al. 2011). Externalising risks such as delinquency have been linked to the reduced likelihood of positive outcomes, including educational achievement (Hammen et al. 2008; Gibb et al. 2010). Delinquent behaviour combined with a deviant peer group increase the likelihood of educational disengagement and other problematic outcomes that follow youth into adulthood (Washburn and Capaldi 2013). Internalising and externalising behavioural risks play a key role in disengagement from school (Martin 2011; Ungar and Liebenberg 2013). The connection between delinquent behaviour and poor educational outcomes is moderated by levels of social competence (Stepp et al. 2011). While peer problems are connected with reduced educational performance youth who report positive, supportive peer relationships are more likely to engage in positive, prosocial behaviours and are also more likely to continue with their education (van Dommelen-Gonzalez et al. 2015). As noted above, ethnic minority status has also been connected to variations in educational achievements. In the current study individual-level factors such as depression, behavioural issues and risk behaviours, peer problems and minority ethnic status are considered, not in terms of their direct effect on educational outcomes, but rather for their explanatory power as predictors of educational aspirations through the influence they have on a social ecological measure of resilience.

Understanding the experience of youth who have prematurely exited school

A limitation of much of the research conducted to date on achievement aspirations and educational outcomes is that it has been located within schools or universities (Määttä et al. 2007). This means that understanding of the factors that influence young people’s educational aspirations have been informed predominantly by the experiences of students who are participating in mainstream education yet those facing the most risks typically do not consistently attend school. This inherent sampling bias limits the capacity of these studies to explain the experiences of youth who face high levels of adversity and the ways that these adversities might be connected to their ability to adopt and sustain a hopeful orientation to their education (Ungar and Liebenberg 2013). A critical step in understanding the factors that lead to positive outcomes then is to compare samples of at-risk young people (referred to here as the Study Group; SG) with samples of young people with more normative development (referred to here as the Comparison Group; CG). Similarities in the impact of risk factors and positive resources on indicators, such as educational aspirations, between high and low risk youth, indicate the relevance of general developmental models to both groups of youth. The current study speaks to this knowledge gap.

The current paper

The purpose of the current paper is to use a form of integrative data analysis (the combination of separate samples in a single analysis) to compare the way in which resilience potentiates educational aspirations across two populations of youth (Curran and Hussong 2009). Issues that arise in the study of selected samples, such as the SG young people in the current study, include restricted range of measurement (Allen 1997; Sackett and Yang 2000), which is often indicated by statistically significant differences in mean scores in comparison with more normative groups. One consequence of this is reduced variability across measures, which leads to reductions in the magnitude of regression coefficients, and increased standard error, thus reducing the precision of the fitted regression models (and thereby affecting the potential validity of these models). Integrative data analysis provides a solution to this issue (Curran and Hussong 2009). It allows two or more samples to be combined and analysed together, employing increased power to obtain reliable parameter estimates, and allowing tests of model equivalence across samples, where samples are heterogeneous. In this particular case, Integrative data analysis allowed the examination of the extent to which a general developmental model of the predictors of educational aspirations and the mediating role of resilience applied to both the SG and the CG.

Method

Sample

The data presented in this paper was collected as part of the New Zealand Youth Transitions Research Programme, a longitudinal, mixed-methods study of the transition to young adulthood for a group of vulnerable teenagers (Sanders and Munford 2017). The study commenced in 2009 and six waves of data collection occurred between 2009 and 2016. The current paper focuses on data collected in the first wave (n = 1371). The research was approved by the University Ethics Review Board prior to any data collection commencing (Sanders and Munford 2017).

The analysis compares 778 youth who were progressing in a more or less normative fashion through high school (CG) with 593 individuals who were facing significant risks and challenges (Study group; SG). To be included in the SG youth needed to have one or more of the following characteristics at the time of their first interview: they had prematurely stopped attending mainstream schools (prior to age 16; the mandated school leaving age), they were involved in one or more of the major service systems (juvenile justice, child welfare, mental health, or attending an alternative education programme), or they were living independently or were homeless while attending high school. The CG were recruited from the same communities as the SG (Sanders and Munford 2014). The mean age of the SG was 15.3 years (SD = 1.13) and for the CG was 15.0 years (SD = 1.34), with ages ranging from 12 to 17 years at the time of the first interview.

It should be noted that because the present study focused on vulnerable young people with atypically high levels of risk, the final sample was non-representative. For example, the ethnic and gender mix of the sample was included an overrepresentation of Māori (the indigenous people of New Zealand) (47% for the SG; 31% for the CG, compared with 20% for this age bracket at the 2013 census) and Pacific youth (20% SG, 23% CG) compared with 8.4% at the 2013 census, see Statistics New Zealand, 2013), and more males (59%) than females.

In terms of exposure to risks, a majority of SG youth (76%, compared with 11% in the CG) reported histories of familial abuse and neglect. Furthermore, self-reports of lifetime rates of mental health service use showed that, to the time of the first interview, 72% of the participants in the SG had been clients of mental health services (21% in the CG); 97% had been recipients of educational interventions additional to mainstream classroom programing (51% in the CG); 87% had received services from family and child or youth services (including statutory child protection services) (25% in the CG), and 62% reported some involvement with juvenile justice services (4.5% in the CG).

Data gathering

Participants were identified by schools, community organisations, recreational clubs and a range of services that provided therapeutic support to youth. A community saturation approach was taken whereby client records were examined for all relevant youth organisations in five geographic locations across New Zealand to identify potential youth; once an organisation’s list had been exhausted, researchers moved to the next organisation and reviewed their client lists to identify additional youth (Sanders and Munford 2014). The study involved the administration of a survey instrument during a one-to-one interview with participants annually for three years (Sanders and Munford 2014). The refusal rates for the first phase of the study were 2.5% for the SG, and 12% for the CG. Youth were interviewed by trained interviewers who assisted youth in completing the questionnaire (Sanders and Munford 2014). All measures were obtained via self-report.

Measures

Educational aspiration

Young people were asked to identify the level of formal education they hoped to achieve. This was measured by a Guttman scale representing level of achievement aspired to. These were ‘0 = do not expect to complete formal educational qualifications’; ‘1–3 = expect to complete NCEA Level 1, 2 or 3)’; 4 = ‘expect to complete tertiary trade or similar qualification’; 5 = ‘expect to attend polytech/university’; to 6 = ‘expect to comple a university degree’ (Sanders et al. 2015).

Predictors

For the purposes of the present study, predictors were chosen on the basis of being significantly (p < .05) associated with either the measure of resilience or the measure of educational aspiration (see Table 1). The following predictors were employed in the analyses.

Table 1. Spearman rank correlations between mental health and behavioural predictor variables, and CYRM-28 Resilience and educational aspiration.

Predictor	CYRM-28 resilience	Educational achievement aspiration
SDQ prosocial behaviour	.45***	.13***
SDQ peer problems	−.22***	−.09**
SDQ conduct problems	−.35***	−.20***
Delinquency	−.31***	−.25***
Depression symptoms	−.34***	−.12***
Māori ethnicity	−.04	−.11***

**p < .001.

***p < .0001.

Peer problems

The peer problems subscale (five items) of the SDQ (Goodman et al. 1998) questionnaire was used. This assesses the level of involvement in age-appropriate peer relationships, tendency to social isolation and preference for interaction with adults over peers. Items are measured on a 3-point scale from 0 = not true to 2 = certainly true with some items being reverse scored. Lower scores indicate normative peer relationships. The reliability coefficient in the current study was .52.

Depression

Depression symptoms were measured using the 12-item version of the Center for Epidemiological Studies Depression Scale (CES-D-12-NLSCY: Poulin et al. 2005). Participants rated each item on a 4-point scale from 0 = rarely or none of the time to 3 = all of the time with some items being reverse scored. Higher scores indicate higher levels of depression symptoms. The reliability coefficient in the current study was 0.80.

Delinquency

The Delinquency subscale (five items) of the 4-H study of positive development was used (Theokas and Lerner 2006). Items were rated on a 5-point scale from 1 = Never to 5 = 5 or more times. The reliability coefficient in the current study was 0.88.

Prosocial behaviour

The five item Prosocial subscale of the Strengths and Difficulties Questionnaire (SDQ) (Goodman et al. 1998) questionnaire was used. This assesses the extent to which youth behave in a kind, sharing manner. Items are measured on a 3-point scale from 0 = not true to 2 = certainly true with some items being reverse scored. Higher scores indicate higher levels of prosocial behaviour. The reliability coefficient in the current study was 0.63.

Resilience

Resilience was assessed using a social ecological measure (Masten et al. 1999; Rutter 2000), the Child and Youth Resilience Measure – 28 (CYRM-28: Liebenberg et al. 2012). The 28 items were rated on a 5-point scale from 1 = Does not describe me at all to 5 = Describes me a lot. The alpha coefficient in the present study was 0.88.

Māori ethnicity

Participants were classified as New Zealand Māori if they indicated that their primary ethnicity was Māori (38% of the sample).

Statistical analyses

Bivariate associations between predictors, CYRM-28 resilience and educational aspiration

In order to establish the existence of associations between predictors, the measure of resilience, and educational aspirations, the first step of the analysis involved estimating the bivariate associations between each predictor, CYRM-28 resilience and the educational aspirations outcome using Spearman rank correlations. For these estimates, data were data pooled across both the SG and CG. Statistical significance was set at 0.05 for all analyses.

Differences between study group and comparison group

The second step of the analytic procedure aimed to establish differences between the SG and the CG across all measures in the study. In order to do this, differences between the SG and the CG across the predictors, CYRM-28 resilience and educational aspiration were tested using independent samples t-tests (for continuous measures) and chi-square tests (for dichotomous measures), using SAS 9.4 (SAS Institute, Inc. 1999). These analyses were used to demonstrate the heterogeneity of the samples across the measures included in the study.

Path analytic regression modelling of the possible mediating effect of CYRM-28 resilience on the association between predictors and educational aspirations

In the final step of the analyses, the set of predictors, the CYRM-28 resilience measure and the educational aspiration outcome measure were entered into a path analytic regression model fitted to the data using MPlus v.7 (Muthén and Muthén 1998–2012). This allowed tests of the assumption of integrative data analysis that the model fit equally well for both the SG and the CG.

The model fitting proceeded in two stages:

1. First, all predictors were entered simultaneously into a model using pooled data from both the SG and the CG. In addition, in order to control for possible variability in model estimating due to geographical region, each model contained a dummy variable (not shown) representing the location in New Zealand at which the data were collected (Auckland; Manawatu; Kapiti; Wellington; Otago).

2. Next, the simultaneous models were tested for equivalence across the SG and CG by specifying that the parameters (coefficients) for the model be constrained to equal across both groups. The model was then adjusted to improve fit by removing the equivalence constraints for selected parameters, as suggested by modification indices. The test of significance for the fit of the equivalent models was given by a chi-square test, with a non-statistically significant (p > .05) result suggesting adequate model fit. Additional fit indices were provided by the Root Mean Square Error of Approximation (RMSEA) and the Comparative Fit Index (CFI). In well-fitting models the RMSEA should be less than .05 and the CFI close to 1.

Results

Correlations between predictor measures, CYRM-28 resilience, and educational aspirations

Table 1 shows the Spearman correlation coefficients for the associations between the measures of mental health, behaviour and ethnicity, and: (a) the CYRM-28 measure of resilience; and (b) the measure of educational aspirations. The Table shows that the mental health and behavioural measures were significantly correlated with both CYRM-28 resilience and educational aspirations. Māori ethnicity, on the other hand, was significantly correlated only with the educational aspirations outcome measure.

Tests of differences between study group and comparison group on predictors, CYRM-28 resilience, and educational aspirations

Table 2 shows the mean scores and percentages (for Māori ethnicity) across the predictors, and for the CYRM-28 resilience measure and the educational aspirations measure. Table 2 also reports the results of independent t-tests comparing the SG and CG across the continuous measures. The table shows that SG and CG members differed significantly (p < .05) across all measures. SG members had: lower scores on the SDQ prosocial measure; higher scores on SDQ peer problems and conduct problems; higher scores on the measures of delinquency and depression symptoms; and were more likely to be of Māori ethnicity. Furthermore, SG members had lower scores on the measure of resilience, and lower educational aspirations.

Table 2. Comparisons of SG and CG on measures of predictors, CYRM-28 resilience, and educational aspiration.

Measure	SG	CG	X^2/t	p
	n = 593	n = 773
% Māori ethnicity	47.4	31.3	6.14	<.0001
Mean (SEM) SDQ prosocial behaviour	7.72 (.07)	7.91 (.06)	−2.04	<.05
Mean (SEM) SDQ peer problems	2.49 (.07)	1.53 (.05)	11.26	<.0001
Mean (SEM) SDQ conduct problems	4.53 (.10)	2.09 (.07)	20.37	<.0001
Mean (SEM) delinquency score	10.28 (.30)	2.53 (.14)	25.37	<.0001
Mean (SEM) depression score	9.93 (.25)	6.85 (.18)	10.25	<.0001
Mean (SEM) CYRM resilience score	103.44 (.66)	110.70 (.53)	−8.70	<.0001
Mean (SEM) educational achievement aspirations	3.55 (.09)	4.44 (.07)	−7.82	<.0001

Note: Nonparametric median tests showed the same pattern of significant (p < .05) differences.

Path analytic regression modelling of the possible mediating effect of CYRM-28 resilience on the association between predictors and educational aspirations

In order to obtain a test of the mediating effect of CYRM-28 resilience on the associations between the predictors and educational aspirations, the next step of the analyses employed a path analytic regression model. This model was able to: (a) account for the mediating effect of CYRM-28 resilience in the association between the predictors and the educational aspiration outcome; and (b) the fit of the model across both the SG and CG (see Methods).

The tests of model goodness-of-fit showed that a largely constrained model (the equivalent model for both the SG and CG) was well-fitting (X² (24) = 26.7, p > .30; CFI = 0.995; RMSEA = 0.013). Modification indices suggested that 2/32 model parameters (6.3%) needed to be unconstrained across the SG and CG in order to improve model fit (these were the pathways from delinquency to CYRM-28 resilience and from depression symptoms to CYRM-28 resilience). Overall, the results of the tests of model fit suggested that it was not necessary to fit separate models for each group.

The final fitted path model is shown in Figure 1, with arrows going from predictors to outcome measures. The Figure shows the standardised regression coefficients for the pathways from predictors to CYRM-28 resilience, and to educational aspirations, as well as the pathway from CYRM-28 resilience to educational aspirations. For the two pathways that were unconstrained (pathways from delinquency to CYRM-28 resilience and from depression symptoms to CYRM-28 resilience) the Figure shows separate estimates of standardised regression coefficients for the SG and the CG. The Figure shows:

1. The pathways from SDQ prosocial, SDQ conduct problems, and depression symptoms to CYRM-28 were all statistically significant (all p-values < .01), while the direct pathways from these variables to educational aspirations were statistically non-significant (all p values >.05), suggesting that the linkages between prosocial behaviour, conduct problems and symptoms of depression and educational aspirations were mediated via CYRM-28 resilience. The magnitude of the pathway from SDQ prosocial to CYRM-28 resilience was the largest in the model, with the next largest being depression.

2. For the measure of delinquency, the pathway to CYRM resilience for the SG was statistically non-significant (p > .05) whereas for the CG the pathway it was statistically significant (p < .05). The pathway from the measure of delinquency to educational aspirations was statistically non-significant, suggesting that the linkage between educational aspirations and delinquency was mediated by CYRM-28 resilience, but only for the CG. The adjusted association between delinquency and resilience was relatively weak (−0.08) for the CG.

3. The pathway from Māori ethnicity to CYRM-28 was statistically non-significant, whereas the direct pathway to educational aspirations was statistically significant, indicating that CYRM-28 resilience did not mediate the pathway from ethnicity to educational aspirations. However, the effect size of the negative association between Māori ethnicity and educational aspirations was quite small (−0.06).

4. The pathways from SDQ peer problems to both CYRM-28 resilience and educational aspirations were statistically non-significant (both p-values > .05).

Figure 1. Fitted path analytic regression model of the association between predictors and educational aspiration, mediated by CYRM-28 Resilience (standardised regression coefficients shown). *p < .05. **p < .01. ***p < .001.

In addition, it should be noted that the R² values for CYRM-28 resilience and educational aspirations were .32 and .07, respectively (both p-values < .001). The results of these analyses suggest that the effects of four predictors on educational aspirations were mediated by CYRM-28: prosocial behaviour; conduct problems, depression symptoms, and delinquency (for the CG only). Relatively large effect sizes were found for prosocial behaviour (positive) and depression (negative).

Discussion

The current paper had two purposes. First, it was concerned with understanding whether or not resilience mediated the impact which proximal risks and resources have upon young people’s educational aspirations. The second aim was to establish whether the linkages between these proximal risks and resources, resilience and educational aspirations were similar for young people facing high levels of adversity (Study Group; SG) and young people on more or less normative developmental pathways (CG). Given that educational aspirations are known to play an important role in educational achievement, understanding more about the factors that shape aspirations should enable educators and others who support young people to better articulate educational ambitions and then to work towards their achievement. As Määttä et al. (2007) have argued, in working to effectively support the development of educational aspirations among youth, it is important to identify those factors that are amenable to influence and the factors included in the model are all able to be modified by the actions of educators and others who work with youth.

While there were significant differences between the SG and CG in terms of the impact that two of the measures (delinquency and depression) had upon resilience, the analysis suggested that a common set of factors combined together to explain the educational aspirations of this diverse group of young people. The final model showed that resilience was a key mechanism mediating the impact which the risks and resources around youth have upon their educational aspirations. This pattern held true regardless of their background circumstances (i.e. irrespective of whether they were in the SG or the CG). Comprising as it does individual, relational and contextual components, the model clearly articulates the significant role that resilience resources play in potentiating educational aspirations. The wider literature has established that educational aspirations are significant drivers of educational success (Snyder et al. 2002; Day et al. 2010). The challenge, then, is for educators and other professionals to focus upon enhancing the resilience resources around young people because this will facilitate educational aspirations. Indeed, the CYRM-28 taps a range of contextual dimensions relating to education, community, culture and spiritual connection as well as individual and family components. All of these aspects of resilience can be supported and bolstered in educational settings. In particular, the model suggests that paying attention to the presence of risk factors, most notably depression and, to a lesser extent, delinquency, as well as to the positive aspects such as prosociality will enhance youth resilience. Indeed, the fact that the measure of prosociality had the largest magnitude in the final fitted model implies its importance in resilience (and in turn, influencing educational aspirations). Ensuring that youth receive appropriate mental health support, for instance, is likely to bolster young people’s resilience and in the process facilitate their capacity to formulate educational aspirations. Alternatively, school personnel could directly target specifically educational components of resilience by ensuring that the school environment is inclusive and creates a strong sense of belonging for all students.

Neither ethnicity nor peer group factors impacted upon resilience levels among youth in this model. Indeed, peer relationships had no impact on educational aspirations either. On the other hand, Māori ethnicity did have a small but significant direct impact on educational aspirations. This suggests that Māori youth may face particular challenges not confronted by other youth in sustaining educational aspirations. These may not be able to be totally mitigated by bolstering resilience. The pattern observed here is likely to have been influenced at least partially by the significantly higher proportion of Māori youth in the SG, and the related fact that SG youth were exposed to greater overall risks, fewer resources and lower resilience (see Table 2). However, it is recognised that indigenous youth may face particular challenges in sustaining educational aspirations and schools clearly have a role in adopting practices that foster a sense of belonging for indigenous youth and with ensuring programmes are relevant and meaningful for all youth (Gray and Beresford 2002).

While the findings indicate that, broadly speaking, the same factors influence educational aspirations irrespective of overall levels of adversity, there are nonetheless some significant differences in how these processes work themselves out for SG and CG youth. First, the SG were clearly exposed to far greater risks, and had significantly fewer resources with which to manage these risks than CG youth. SG youth confronted an accumulation of disadvantage and it could be expected that they would confront greater challenges in successfully participating in mainstream educational institutions (Sanders and Munford 2015). Given this, it was somewhat surprising that prosociality and conduct problems operated in largely similar ways to boost (prosociality) or undermine (conduct problems) resilience for both groups. It was also surprising to observe that both delinquency and depression appeared to be of significantly greater importance in terms of their impact upon resilience for CG youth than was the case for SG youth. While the CG had significantly less depression, this appeared to have a proportionately greater impact upon their resilience resources than was the case for SG youth. This impact which depression has upon young people’s resilience is particularly important because of the connection between depression and achievement orientations that has been demonstrated elsewhere (Määttä et al. 2007). Ungar (2004) has observed that different factors boost resilience at different levels of risk. It may well be that the pattern in the current study confirms this overall principle of the differential impact of risks upon resilience depending on the overall level of adversity that youth confront.

In a similar way, while CG youth had profoundly lower mean delinquency scores (see Table 2); these lower levels of delinquency had a small but significant impact upon their resilience. Although SG youth reported much greater levels of delinquency, these levels did not, in and of themselves, impact upon their resilience suggesting some support for the argument of hidden resilience among young people facing high levels of adversity (Ungar 2004). Norm breaking behaviour could be argued to comprise a form of bounded agency (Aaltonen 2013) wherein young people facing significant challenges use deviant acts to demonstrate agency and these powerful acts then contribute to their overall resilience (Lumby 2012). Elsewhere boys have been found to use norm breaking behaviour to manage the sense of powerlessness in their lives (Määttä et al. 2007). When faced with high levels of adversity young people will draw coping strategies from the resources that are to hand. These will not always necessarily be prosocial: ‘without the opportunity to take advantage of a healthy environment, a child will use maladaptive forms of coping to maintain wellbeing’ (Ungar 2013, p. 330). Rather than undermining resilience, these acts become powerful forms of coping in dangerous environments and as such comprise resilience resources.

It should be noted that the present study has several limitations, the most important of which is that the sample, drawn from at-risk youth, is non-representative of the New Zealand population. However, as the focus of this research is on young people exposed to higher than normal levels of risk, the generalisability of the research to the population as a whole was not a specific goal. Also, all data were obtained by self-report, which may have led to errors and biases arising from this method. In addition, it should be noted that many of the measures in the study (such as resilience and mental health disorders) may not be fully conceptually distinct from one another. However, because these measures were only moderately correlated in all cases, the extent to which there may have been conceptual overlap between measures is unlikely to pose significant problems for study validity.

Conclusion

Educational aspirations appear to emerge out of a complex web of individual-level risks and resources that are then critically shaped by wider contextual resilience resources. These resilience resources implicate not only young people, but their families, friends, neighbourhoods and wider social systems. Attempting to shape aspirations then becomes a more ecologically driven endeavour that addresses the context within which the young person is embedded as well as the characteristics and challenges particular to that young person. The model clearly shows that resilience is able to mitigate risks and potentiate resources around youth enabling young people to imagine achieving educational credentials.

It could be argued that fitting the same model to the SG and CG was inappropriate, because of more limited variability within the SG due to the selected nature of the sample (Sackett and Yang 2000). However, the regression analysis found a common model could be fitted across both groups that linked predictors to educational aspirations; only 6.3% of model parameters needed to vary across groups. These results suggest that any threat to validity posed by range restriction is not a critical concern for analyses of SG youth in the Youth Transitions Study. More generally, the results suggest that factors that shape educational aspirations are similar for both vulnerable youth and those who confront substantially less risk. This then in turn implies that our understanding of the places to focus attention when seeking to have an impact upon young people’s educational aspirations is largely similar regardless of levels of risks and resilience. Broad developmental frameworks can therefore be used to explore the mechanisms by which youth facing significant adversity may be supported to function within educational settings. However, the magnitude of the differences between the CG and SG seen in this analysis, underscores the need to consider the specific ways in which risks and resources interact more closely.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by Ministry of Business, Innovation and Employment [grant number MAUX0901].