Academic well-being: School engagement and burnout

The definitions of academic well-being vary, but many agree that well-being is a multidimensional phenomenon consisting of both cognitive (what students think of school) and affective (how students feel about school) dimensions (Diener et al., 2018; Putwain et al., 2020). It has often been described as representing subjective evaluations of school, commonly including both positive and negative indicators (Diener et al., 2018; Putwain et al., 2020). Hascher (2008), for example, operationalized academic well-being as the presence of positive attitudes and enjoyment of school, positive academic self-concept and the absence of worry, physical complaints, and social problems related to school. Others have, similarly, conceptualized academic well-being as academic self-concept, perceived learning difficulties and school burnout (Korhonen et al., 2014), school burnout, engagement, school value and satisfaction with educational choice (Tuominen-Soini et al., 2012), and school burnout and engagement (Fiorilli et al., 2017).

Consequently, school engagement and burnout were chosen as indicators of academic well-being in this study, comprising both positive and negative indicators and cognitive and affective components of students' subjective well-being in school (Diener et al., 2018). Engagement and burnout are commonly combined indicators and key concepts in the job demands–resources theory that has initially been used to explain health and performance in the occupational context (Demerouti et al., 2001). Based on the rationale that school is a place where students work – they attend classes and complete structured activities with specific performance goals – the combined concepts of engagement and burnout may reasonably be extended to educational contexts (Walburg, 2014). The demands–resources theory has increasingly been used in the educational context to explain students' well-being and academic functioning in school (Study demands–resources theory: Lesener et al., 2020; Salmela-Aro & Upadyaya, 2014).

School engagement has been described as a positive, study-related state of mind comprised of energy, dedication and absorption (Salmela-Aro & Upadyaya, 2012; Schaufeli et al., 2002, 2006). Energy involves high levels of vigour and energy while studying, dedication reflects a positive cognitive attitude towards school, whereas absorption is characterized by being fully concentrated and involved in one's studies. These dimensions are highly correlated (Salmela-Aro & Upadyaya, 2012) and are commonly described as representing one overall engagement construct (e.g., Salmela-Aro et al., 2016).

School burnout has typically been approached as a multidimensional construct consisting of three highly correlated, but conceptually distinct factors: exhaustion, cynicism and inadequacy (Salmela-Aro et al., 2009; Schaufeli et al., 2002). Exhaustion reflects school-related feelings of strain due to high school demands. Cynicism is described as detached attitudes towards school and loss of interest in one's schoolwork, whereas inadequacy refers to diminished feelings of one's competence and achievements (Salmela-Aro et al., 2009). These dimensions seem to be differently associated with various school-related outcomes (e.g., achievement: Salmela-Aro & Upadyaya, 2014; motivation: Tuominen-Soini et al., 2012).

Although school engagement and burnout seem to be negatively correlated (Schaufeli et al., 2002), longitudinal studies investigating causality and their reciprocal relations are scarce. However, in line with the assumptions proposed by the demands–resources theory (Bakker & Demerouti, 2017), a meta-analysis, of which most studies were drawn from occupational contexts, found a reciprocal, negative relationship between engagement and burnout (Maricuțoiu et al., 2017). In the educational context, Salmela-Aro and Upadyaya (2014) found burnout to negatively predict engagement during one school year in upper secondary education, but not vice versa. Engels et al. (2019), in turn, did not find any relations between behavioural and cognitive components of engagement and school burnout (considered as one factor) during grades 6–9.

Academic well-being and performance

School engagement seems to have a positive, moderate association with academic performance, suggesting that highly engaged students generally perform better in school (Lei et al., 2018; Skinner et al., 2008). School burnout, in turn, seems to have negative associations with performance, indicating that students who are exhausted, have cynical attitudes towards school and feel inadequate in school, generally, have lower academic performance (Madigan & Curran, 2021). Such positive relations between engagement and performance, and on the contrary, negative associations between burnout and performance, could be explained, for example, by ‘gain- and loss spirals’ as proposed by the job demands–resources theory (Bakker & Demerouti, 2017; Hobfoll et al., 2018). Burnout may develop if the school environment imposes high levels of demands and offers few study resources, whereas the availability of contextual and personal resources fosters school engagement that leads to more positive outcomes (Salmela-Aro & Upadyaya, 2014). Individuals who are highly engaged in their work or studies likely have more resources to focus and dedicate their energy to their work and may also be more motivated to stay engaged and create their own resources over time, thus creating a ‘gain spiral’ (Bakker & Demerouti, 2017; Hobfoll et al., 2018). However, individuals who experience high strain or burnout are more likely to participate in self-undermining behaviour, creating more demands over time, thus risking creating a ‘loss spiral’ (e.g., Fiorilli et al., 2017).

Only a few studies have studied the associations between school engagement, burnout and performance longitudinally. Wang et al. (2015) and Paloș et al. (2019) found that prior academic performance (grade point average, GPA) positively predicted school engagement, whereas it negatively predicted school burnout. However, changes in GPA (i.e., the slope factor) were not related to changes in engagement or burnout during Grades 9–11 (Wang et al., 2015). Furthermore, both Paloș et al. (2019) and Salanova et al. (2010) investigated the reversed effect (i.e., of engagement and burnout on performance) among university students. While Salanova et al. (2010) only found engagement to predict subsequent performance, no significant effects were found by Paloș et al. (2019) in that direction.

Parviainen et al. (2020) investigated the longitudinal effects of arithmetic and reading skills on school burnout among slightly younger students and found that lower arithmetic skills in 6th grade indirectly predicted higher levels of exhaustion and cynicism in upper secondary education, mediated by exhaustion and changes in cynicism in lower secondary education. However, surprisingly, they also found that better reading skills in 6th grade predicted higher levels of cynicism in upper secondary education. Engels et al. (2019), on the other hand, did not find any associations between performance, burnout and engagement among students in grades 6–7.

Overall, these findings suggest that the associations between performance, engagement and burnout are not always straightforward. A few studies have also shown that some students seem to be both engaged and burned out by school, while still performing well (Tuominen-Soini & Salmela-Aro, 2014; Widlund et al., 2018). Cross-lagged studies investigating the associations between school burnout and other educational outcomes have found similar results; Salmela-Aro and Upadyaya (2017) found positive relations between exhaustion, educational aspirations and educational attainment over time.

Furthermore, a relevant factor to consider when investigating the dynamics of students' well-being and performance seems to be gender. Although there are no notable gender differences in Finnish boys' and girls' mathematics performance (OECD, 2019), girls consistently report higher levels of exhaustion and feelings of inadequacy in school (Finnish Institute for Health and Welfare, 2022; Tuominen-Soini & Salmela-Aro, 2014; Widlund et al., 2018). Some have also found girls to be slightly more engaged in their studies (Tuominen-Soini & Salmela-Aro, 2014). However, when it comes to developmental trends, Wang et al. (2015) did not find any gender differences in the development of either engagement or burnout during the adolescent years.

Participants and procedure

Data came from a longitudinal project (FRAM: Adolescents' well-being and learning in future society) at Åbo Akademi University, utilizing an accelerated design, following students over 4 years. APA ethical standards were carefully followed in the conduct of the whole project. Participation in the study was voluntary, informed consent forms were collected from the students' parents, and the participants were assured of the confidentiality of their responses. Five public lower secondary schools from different regions of Swedish-speaking areas of Finland participated in the data collection (Swedish is the second official language in Finland). Overall, Finland is a relatively homogenous country regarding both socioeconomic status and ethnicity (Official Statistics of Finland, 2022).

The participants were recruited in lower secondary school in the fall (T1) and spring (T2) of the school year 2016–2017, when they were in Grade 7 (Cohort 1) and Grade 9 (Cohort 2). The same participants were followed up 2 years later, in the fall (T3) and spring (T4) of the school year 2018–2019, when they were in Grade 9 (Cohort 1) and Grade 11, that is, studying in upper secondary education for the second year (Cohort 2), respectively. However, due to relatively large attrition rates (ca 60%–70%) in Cohort 2 after students transitioned from lower to upper secondary education, partly due to difficulties in locating students, and many deciding to withdraw participation, only data collected in Grades 7 and 9 were included in this study. Consequently, we combined data from Cohort 1 (fall and spring in 7th and 9th grade) and Cohort 2 (fall and spring in 9th grade). Time points for Cohort 2, for which we did not have data, were entered as incomplete data. The missing data patterns were examined with Little's MCAR test, which indicated that the survey data (i.e., engagement and burnout) was missing completely at random in both the first (MCAR) (χ²[983] = 988.322, p = .446) and second (MCAR) (χ²[133] = 154.753, p = .095) cohort. The missing data in mathematics performance was missing completely at random in Cohort 2 (MCAR) (χ²[4] = 9.287, p = .54), but not in Cohort 1 (MCAR) (χ²[28] = 59.944, p = .000). The full information maximum likelihood approach implemented in Mplus, taking all available information into account when estimating model parameters, was used to deal with missing data (Graham, 2008). All students who participated at least once across the measurement waves were included in this study, resulting in 1131 students (see Table 1).

Measures

Data analysis strategy

All analyses were carried out in Mplus (Muthén & Muthén, 1998–2017). Analyses began with examining the structural validity and stability of the measures through confirmatory factor analysis (CFA; see Table S1) as well as establishing measurement invariance across time through longitudinal confirmatory factor analysis (LCFA; see Table S2) and across cohorts through multiple group confirmatory factor analysis (MGCFA; see Table S3). As students' mathematics test score was based on an item response theory model, it was not possible to conduct CFAs on that measure.

Next, a random intercept cross-lagged panel model (RI-CLPM) was fitted to the data, following the procedure as described by Hamaker et al. (2015). Autoregressive parameters reflect the extent to which variability in one measure at one time point (e.g., engagement at T4) can be explained by variability in the same measure from the immediately preceding time point (e.g., engagement at T3). The cross-lagged parameters, in turn, estimate the degree to which variability in one measure at a time point (e.g., engagement at T4) can be explained by variability in a different measure at a preceding time point (e.g., burnout at T3). The standard cross-lagged panel model (CLPM) approach has received criticism, stating that stable differences between individuals over time cannot be estimated separately from changes in measures within an individual (Hamaker et al., 2015). Therefore, it is challenging to determine the degree to which bidirectional patterns of the longitudinal association between measures can be attributed to changes occurring within students or relative changes between students.

A RI-CLPM bypasses these challenges by separating within-person variability from between-person variability. Consequently, each construct was decomposed into a stable between-student part and a varying within-student part. The cross-lagged effects are thus interpreted as the quantities that express the extent to which the variables influence each other within individuals. To capture stable trait-like differences between students in performance, school engagement and burnout constructs, five overarching random intercept factors were included, one for each measure. These reflect the trait aspects of performance, school engagement and burnout over time. The four observed scores for each time point were the indicators of each random intercept, with all factor loadings constrained to 1.

The within-student varying part was captured by regressing each observed score on its own latent factor. The resulting 20 latent factors (i.e., one for each construct per measurement wave) were subsequently used to specify within-time associations, autoregressive stability paths and cross-lagged paths. The error variances of the observed scores were constrained to zero, ensuring that all variation in the observed measures was entirely captured by the within-person and between-person latent factor structures. Lastly, gender was added to the model as a time-invariant between-level predictor.

Due to the complexity of the model, composite scores were used for school engagement and burnout. In all analyses, chi-square (χ²), the comparative fit index (CFI: cut-off value close to >.95), the Tucker–Lewis Index (TLI: cut-off value close to >.95), and the root mean square error of approximation (RMSEA; cut-off value close to <.05) were used as model-fit indices (Marsh et al., 2004). Descriptive statistics (means and standard deviations) and internal consistencies for each time point and cohort are presented in Table 2 and correlations between all variables in Table 3.

Cross lagged pathways

The RI-CLPM fitted the data very well [χ²(75) = 102.781, p = .0183, RMSEA = .018, CFI = .997, TLI = .991] and explained 27% of the variance in the latent mathematics performance variable, 33% in engagement, 23% in exhaustion, 18% in cynicism and 20% of the variance in inadequacy at T4 (i.e., end of 9th grade). The effect of gender was significant on the random intercepts of exhaustion and inadequacy, suggesting that girls felt more exhausted and inadequate in school in general, compared to boys. The between-student correlations indicated that students with higher mathematics performance were also, generally, more engaged (r = .23, p < .05), less exhausted (r = −.22, p < .05), less cynical (r = −.29, p < .05) and felt less inadequate in school (r = −.27, p < .05) across measurement waves. The between-student correlations between engagement and burnout constructs were also significant and negative (exhaustion: r = −.32, p < .05; cynicism: r = −.51, p < .05; inadequacy: r = −.42, p < .05).

Standardized results are depicted in Figure 1.

Limitations and future directions

Due to the complexity of the model, sum scores were used instead of latent factors (engagement and burnout constructs) in the analyses. Consequently, not all measurement errors are parcelled out in the measurement models, which might underestimate relations between variables. We also found that the Cronbach's alphas for inadequacy were rather low in 7th grade, indicating that the items might not be as adequately representative of inadequacy among younger students, or that more items may be needed to fully capture students' feelings of inadequacy. Furthermore, although combining cohort data comes with several benefits (e.g., increased sample size and time efficiency), it also creates more uncertainty in the data due to increased missing values. Although most data were found to be missing completely at random, math performance in the first cohort was not, which should also be acknowledged when interpreting the results.

Our findings revealed that the relationships between academic well-being and performance seemed to become more prominent over time during adolescence, and therefore, it would be important for future studies to invest such relations among older students as well, particularly over the educational transition to upper secondary education as this has been found to spark changes in students' well-being and motivation. It should also be acknowledged that only mathematics performance was used to represent academic performance in this study. Thus, it would be important for future research to investigate the relationship between academic well-being with performance in other domains as well (e.g., language arts or science), and, to use domain-specific aspects of students' well-being in school (e.g., domain-specific emotional cost).

Lastly, research indicates that the role of demands and resources for an individual's engagement and perceived burnout might be affected by societal cultural beliefs and values (Rattrie et al., 2020). Therefore, the country-specific context should be acknowledged when making conclusions about our findings, and the relation between students' performance, engagement, and burnout should be tested in different cultures in future work.

Practical implications

The findings demonstrate the importance of students' emotional engagement with schoolwork, but also, the importance of acknowledging possible strains students' negative feelings towards school might have on their performance. The results indicate that experiences of school burnout, particularly feelings of inadequacy and cynical attitudes, may affect students' performance negatively over time. Therefore, both academic performance and well-being should be supported in school. It is also important to emphasize that even though exhaustion may go hand in hand with higher performance for some, prolonged feelings of strain may result in more negative outcomes later on. Considering that there were negative effects of school burnout, not only long-term but also within school years, it would be important to implement resources to support students' academic well-being early on. Schools should, for example, make student welfare services (i.e., school psychologists and health care) easily available and focus on enhancing adequate coping strategies for students, to help them prevent and handle possible feelings of burnout (van Loon et al., 2020).