Swiss adolescents’ well-being in school

Student well-being is an issue with regard to educational effectiveness. However, little is known about Swiss students’ well-being in school. This study was conducted in the context of the project “Überprüfung des Erreichens der Grundkompetenzen ÜGK 2016” and aimed at contributing to closing this gap by investigating adolescents’ (N = 22,423) well-being in school. An analysis of six well-being in school dimensions revealed the following results: Swiss secondary students report positive attitudes, a good academic self-concept, low physical complaints and low social problems, but also a lack of enjoyment and a prevalence for worries in school. Significant differences across gender, region, migration background, and attended school type as well as associations between well-being in school and school reluctance and truancy were found.

and that positive relationships with peers and teachers support well-being in school (e.g. Hascher & Baillod, 2004). Additionally, within the context of the Programme for International Student Assessment (PISA) 2015 study, it was found that a sense of belonging at school and positive relations with teachers and classmates are generally relevant for student achievement and for Swiss adolescent students' achievement, respectively (Organisation for Economic Cooperation and Development [OECD], 2017).
Despite this gradually growing knowledge regarding children's and adolescents' well-being in school in Switzerland, there is still a paucity of representative data and little understanding of the various elements contributing to the phenomenon. There is also poor comprehension of the relation of wellbeing in school to problematic school behaviour, which may occur when a student does not feel good in school. Our study aimed to fill this gap by capitalising on the well-being in school data from the project "Überprüfung des Erreichens der Grundkompetenzen" (ÜGK, 2016). Based on the idea that well-being in school is an important precondition for students' positive development and achievement in school (e.g. Hascher, 2011;OECD, 2017), we sought to gain a deeper insight into Swiss adolescents' well-being in school and its association with school reluctance as a silent form of school refusal and school absenteeism as a visible form of school refusal.

The theoretical concept of well-being in school
Currently, a heterogeneity of conceptualisations of children's and adolescents' well-being in school exists. Essentially, there is little agreement on how to define or measure it, and different approaches coexist. For example, one approach is to regard well-being in school as an indicator of an individual's successful functioning in school, that is, as a positive response to the school environment, as may be construed and, to a degree, measured using the degree of feeling good in that milieu (De Fraine et al., 2005). Another approach is to define wellbeing in school as the result of an interplay between an individual's expectations and needs, on the one hand, and conditions of the school environment, on the other (as represented by Engels et al., 2004). Hascher (e.g. 2004Hascher (e.g. , 2011 suggests conceptualising well-being in school in close relation to psychological wellbeing theory as introduced by Diener (2000), with an integration of emotional, cognitive, and physical domains, such as positive attitudes towards school, enjoyment in school, and lack of physical complaints in school. Alternatively, well-being in school has been considered a crucial indicator for student health in school, relevant to children's and adolescents' general health (e.g. Gysin, 2018).
Regardless of the approach, most researchers agree that well-being in school is a multidimensional concept that consists of various dimensions relating T h e m a to attitudes towards school and the student's emotions; however, there is disagreement as to which dimensions contribute to well-being in school. It can be stated that, although there is an increase in empirical studies, there is a lack of theoretically based concepts. It seems that well-being in school is a general term, without a sound theory, that is used for a variety of variables, such as feeling attached to school, feeling good at school, achieving well academically, reporting a good school climate, or having positive relationships with teachers and peers. To give an example, with respect to the PISA 2015 study, student well-being was described as a combination of several psychological, cognitive, social, and physical factors (OECD, 2017, p. 62). Each dimension consisted of a bundle of variables, such as motivation, resilience, self-efficacy beliefs, and stress for the psychological dimension. For the operationalisation of this dimension, however, achievement motivation and school anxiety were selected, and information about well-being in school is missing. There is thus a severe inconsistency between the conceptual framework and the methodological approach that might be used to evaluate the phenomenon. This challenges the content validity of many results thus far obtained in the research on well-being in school.
As it turns out, the theoretical construct of well-being in school seems to be a fuzzy set that hardly helps in understanding the nature and manifestations of well-being in school. Moreover, the definitions of well-being in school are often conflated with its predictors and causes. In consequence, well-being in school fails to provide a roadmap for counteracting absence from school. For this reason, we argue that psychological research explicitly dedicated to the understanding of people's general well-being can provide clarification of the concept. Across decades, researchers in this area (e.g. Diener, 2000;Diener & Biswas-Diener, 2000;Diener & Lucas, 2000;Ryff & Keyes, 1995) have developed a comprehensive definition of the phenomenon, a deep understanding of its mental aspects and behavioural expression, and also a valid set of empirical instruments to assess it. There is general agreement that: (1) well-being consists of several dimensions that address positive and negative aspects of such feelings (i.e. positive emotions, on the one hand, and worries, on the other); (2) enjoyment is a core element of well-being; and (3) well-being consists of cognitive as well as emotional dimensions and psychological as well as physical dimensions. Based on these findings, Hascher (2003) describes well-being in school as a quality of experience characterised by the dominance of positive feelings and cognitions towards school, persons in school and the school context in comparison to negative feelings and cognitions towards school life. Wellbeing in school represents subjective, emotional and cognitive evaluations of school reality and can be seen as an imbalance of positive and negative aspects in favour of positive aspects. (p. 129) This definition encompasses the abovementioned key ideas of psychological wellbeing research and can be applied to well-being in school as a state (a temporal experience) or as a trait (a disposition). More specifically, well-being in school has T h e m a been defined as consisting of the following six dimensions: (1) positive attitudes towards school, (2) enjoyment in school, (3) positive academic self-concept, (4) worries in school, (5) physical complaints in school, and (6) social problems in school (see also Hascher, 2011Hascher, , 2012. These six dimensions represent the main indicators that have been found in psychological well-being research and have already been applied in several empirical studies in Switzerland, Germany, Austria, the Netherlands, the Czech Republic, and Luxemburg (e.g. Hascher, 2004;Hascher & Hagenauer, 2011;Morinaj & Hascher, 2018). Accordingly, these six dimensions formed the core of our study on well-being in school.

Importance and prevalence of well-being in school Why well-being in school matters
The role of well-being in school can be theoretically framed and empirically analysed from three perspectives: (a) Well-being in school can be seen as a prerequisite of adaptive student behaviour and student achievement, such as engagement, as was shown by Gutman and Vorhaus (2012) and as argued by PISA 2015 (OECD, 2017). From this theoretical perspective, well-being in school is one of the preconditions that fosters successful learning in school. Thus, well-being in school can be empirically investigated as an independent variable, with a particular emphasis on its positive impact on student learning, behaviour, and academic achievement. With this approach, the importance of well-being in school is legitimised through its positive functions in fostering adaptive student behaviours and serving as a supporting resource.
(b) With respect to deviant behaviour, well-being in school can serve a preventive function. Like theories on school climate (for an overview, see Thapa et al., 2013), well-being in school is thought to serve as a protective resource for students to cope with challenges and problems in schools. It strengthens students' capacity to deal with adversive situations and to manage troubles in school. Empirically speaking, as an independent variable, well-being in school serves a positive role by hampering maladaptive behaviour and by preventing the development of negative approaches to learning and development in school.
(c) Furthermore, well-being in school can be seen as an educational goal in itself that merits promotion. This perspective, among others, is represented in health education research (e.g. Klasen et al., 2017) and school quality or school effectiveness research (e.g. Chodkiewicz & Boyle, 2017). Well-being in school is thus an issue that every school should address as a means of responding to students' basic psychological needs (Ryan & Deci, 2001). In line with this approach, it can be investigated empirically as a dependent variable leading to the question of which factors contribute to well-being in school.
T h e m a Prevalence of well-being in Swiss secondary schools Thus far, it has been found that adolescent students in Switzerland report moderate to high well-being in school scores (Hascher, 2004). They show weaker scores on enjoyment in school, however, and have reported worrying about schools' academic demands. Given the fact that school is a very important environment in adolescent students' lives, it is worth noting that PISA 2015 data revealed that Swiss adolesents, compared to their international peers, reported an above-average level of satisfaction with life (OECD average = 7.3, Swiss adolescents average = 7.7), with approximately 80% of the adolescents reporting being very satisfied (around 40%) or satisfied (around 40%) with life (OECD, 2017). Both studies, however, also found a substantial number of students who evidenced problematic patterns. In 1999, for example, 10% of the students exhibited negative patterns in well-being in school domains (such as frequent worries concerning school; Hascher, 2004). In PISA 2015, approximately 48% of the Swiss adolescents agreed or strongly agreed with the statement, "I often worry that it will be difficult for me to take a test", while 56% agreed or strongly agreed with the statement, "I worry that I will get poor grades at school" (OECD, 2017, p. 85). Furthermore, Hascher (2004) found that Swiss adolescents' well-being in school gradually decreases during secondary I education. In PISA 2015, it was shown that, in general, students' sense of belonging at school had decreased significantly over the past 12 years, with a difference of 11% between 2003 and 2015 and a difference of 12% for Swiss adolescent students between 2012 and 2015.
With respect to gender differences, it was found that girls report more positive emotions and attitudes towards school but also more worries (Hascher, 2004). In terms of satisfaction with life, PISA 2015 also found gender differences in favour of Swiss boys.
There is, however, a lack of reliable data on Swiss students' well-being in school; moreover, little is known about the different dimensions of well-being in school. There is also a lack of reliable data on subgroup-specific differences, although there are indicators that girls and boys, students with and without a migration background, students from different regions, and students attending different school types differ with respect to their levels of well-being in school (Hascher, 2004;OECD, 2017).

Well-being in school and students' school refusal
Given the crucial role that school plays in children's and adolescents' lives, the possible outcomes of low well-being in school are evident. With respect to Bronfenbrenner's ecological systems theory (1979), school belongs to the microsystem that is the individual's most intimate environment for development. Negative interactions within this environment negatively affect an individual's development. Accordingly, it has been discussed that disliking school is one crucial reason for school dropout (e.g. Finn, 1989). Empirically, it has been found that well-being in school is negatively related to health-risk behaviour T h e m a (McNeely & Falci, 2004) and to disruptive behaviour (Närhi et al., 2014). Recently, Morinaj and Hascher (2018) were able to demonstrate that dimensions of well-being in school are negatively related to domains of alienation from school, which is thought to engender deviant behaviour, school absenteeism, and, eventually, school dropout. Well-being in school can thus be regarded as a protective resource against school dropout; however, data on the relationship between the different dimensions of well-being in school and facets of absenteeism, such as school reluctance and truancy, in Swiss schools is missing. Thus, in the context of understanding that school absenteeism results from an interplay of individual and school factors (Sälzer, 2010), it is necessary to investigate how (low) well-being in school is associated with absenteeism.

The present study
Considering the lack of representative data on a theoretically well-sounded concept of well-being in school, this study aimed to investigate Swiss students' well-being in school at the end of the obligatory school years in the context of the ÜGK 2016 study. The following research questions (RQs) were addressed: RQ1: How good do Swiss adolescent students feel in school? RQ2: What gender differences can be found in Swiss adolecents' well-being in school? RQ3: What differences with respect to students' origins can be found in Swiss adolescents' well-being in school? RQ4: What regional differences can be found in Swiss adolescents' well-being in school? RQ5: What differences with respect to the attended school type can be found in Swiss adolescents' well-being in school? RQ6: How is Swiss adolescents' well-being in school related to school reluctance and school absenteeism? As previous research has showed rather high scores on well-being in school among Swiss adolescents (Hascher, 2004), we expected Swiss adolescents to report an overall positive evaluation of their well-being in school (RQ1). According to earlier studies (cf. Hascher, 2004), which have uncovered gender differences, we assumed that girls would report higher scores not only on the positive dimensions but also on the negative dimensions of well-being in school (RQ2). The analyses pertaining to RQ 3, 4 and 5 are exploratory, as prevalence rates addressing these specific subgroups are not available so far; thus, no specific hypotheses were proposed. The testing of differences, however, is relevant, as an understanding of subgroup-specific differences could help to identify weaknesses in the school system and the vulnerability of certain groups and could subsequently help to improve well-being in school. Finally, we assumed that well-being in school is negatively related to school reluctance and school absenteeism (RQ 6) (Sälzer, 2010;Stamm, 2012).
T h e m a Method Sample Altogether, 22,423 students in the last year of compulsory education (grade 9 / grade HarmoS 11; mean age = 15.88 years 1 ) participated in the ÜGK 2016 study (Nidegger, 2019). Of these students, 51.2% (n = 11,479) were male, and 48.8% (n = 10,944) were female. With respect to the region studied, 73.2% (n = 16,409) of the sample came from the German-speaking part of the country, 23.4% (n = 5,257) were from the French-speaking part, and 3.4% (n = 757) were from the Italian-speaking portion of Switzerland.
Concerning the students' origins, 71.6% (n = 15,875) of the students were native to Switzerland; 18.9% (n = 4,191) had a migration background as second-generation youth, and 9.5% (n = 2,098) were first-generation students. Coding was applied in the style of the PISA study: Students were classified as "native" if they were born in Switzerland or if at least one parent was born in the country; students with a migration background were differentiated as either second-generation youth (students were born in Switzerland, but their parents were born abroad) or first-generation youth (students and their parents were born abroad).
With respect to the school type, 33.8% of the students (n = 7,400) attended a school type with basic requirements, and 66.2% (n = 14,525) attended a school type with high/advanced requirements. 2

Measures
Students' well-being in school was part of the core questionnaire: All students answered these questions. Well-being in school was assessed using the conceptualisation developed and introduced by Hascher (2004). Six dimensions of well-being in school were distinguished, which were to be rated on a 6-point Likert-type scale (1 = never; 6 = very often). 1) Positive attitude towards school (PAS) (three items, e.g. "I like to go to school"; Cronbach's alpha = .78 3 ) 2) Enjoyment in school (EIS) (three items, e.g. "Have you experienced joy because of teachers' friendliness in the past few weeks?"; Cronbach´s alpha = .80) 3) Positive academic self-concept (PASC) (three items, e.g. "I don´t have problems mastering school tasks"; Cronbach´s alpha = .84.). 4) Worries in school (WIS) (three items, e.g. "Have you been worried about your school grades in the past few weeks?"; Cronbach´s alpha = .75) 5) Social problems in school (SPC) (three items, e.g. "Have you had problems with your classmates in the past few weeks?"; Cronbach´s alpha = .82) 6) Physical complaints in school (PCS) (four items; e.g. "Have you had a severe headache in school in the past few weeks?"; Cronbach´s alpha = .77) In addition, school reluctance was measured based on three items (e.g. "How frequently has it happened during the past few weeks that you wished that school T h e m a was over?"; Cronbach´s alpha = .66; 1 = never; 6 = very often; Hagenauer & Hascher, 2012). Finally, school absenteeism (truancy) was assessed using three items based on the PISA 2012 test (OECD, 2014): Students were asked how often they had arrived late for school in the two weeks prior to the ÜGK test; in addition, they were asked if they had skipped some lessons across a day of school in those two weeks (Cronbach´s alpha = .65; 4-point Likert scale; for a detailed overview of the psychometrical values of the used scales, see Sacchi & Oesch, 2017).
For the present analyses, mean scores were calculated (instead of using the available factor scores in the UEGK database) to build the scales, as previous research on students' well-being in school using the described scale has relied on this procedure, which makes the results across studies comparable.

Data analysis
Descriptive statistics, intercorrelations, mean differences based on independent t-tests, and multiple regression analyses (method: Enter) were conducted by means of the International Database Analyser (IDB), made available by the International Association for the Evaluation of Educational Achievement (IEA), that was programmed for the analyses of the Core Skills Assessment Switzerland https://www.iea.nl/data-tools/tools). The IDB analyser considers the complex sampling design and calculates correct standard errors (using the balanced repeated replication [BRR] method based on 120 replicate weights).
In addition, the software Mplus 7.3 (Muthen & Muthen, 1998 was applied to calculate confirmatory factor analyses (CFAs) and to test for measurement invariance (as these procedures are not implemented in the IDB analyser). The student weight was considered a weight variable. To account for clustering of the data and the stratification of the sample, the cluster (school ID; 830 clusters) and stratification (59 strata) option was applied using the "Type=Complex" command. The complex sampling design and the multilevel-structured data were thereby accounted for. Without level-2 predictors we did not run multi-level analysis. For the present analysis, however, we used the Mplus command "Type=Complex", which considers the non-independence of observations when calculating standard errors and chi-squared tests (Muthen & Muthen, 1998. The interclass correlation coefficients (ICCs) could only be calculated at the school level, as the sample size at the class level was too small. At the school level (school id; mean cluster size: 26.73 students), the ICCs were rather low for all scales: Positive attitudes towards school: ICC = .062; Enjoyment in school: ICC = .050; Positive academic self-concept: ICC = .019; Worries in school: ICC = .067; Social problems in school: ICC = .022; Physical complaints in school: ICC = .031; School reluctance: ICC = .052, and Truancy: ICC = .081.
MLR estimation was applied to account for the non-normality in the data and to calculate robust standard errors.
Model fit was assessed by several commonly used fit indices: CFI, TLI, RMSEA and SRMR. The model fit is good when RMSEA and SRMR are less T h e m a than .06 and when CFI and TLI values are above .95. The fit of a model is considered acceptable when RMSEA and SRMR are less than .08 and CFI and TLI fall between .90 and .95 (Hu & Bentler, 1999).
Missing values were very rare. They were below 1% for all variables (n = 128 missing values (0.6 %) for "like going to school" and n = 201 missing values (0.9 %) for "feel sick with agitation"), except for migration background, for which missing values were 1.2 %. Thus, missing values were not replaced 4 .

Confirmatory factor analysis
First, we tested the factor structure of the well-being in school scales. Hascher (2004Hascher ( , 2007, as noted above, proposes six dimensions of well-being in school (comprising three positive and three negative components). We compared three concurring models: (1) six dimensions of well-being in school; (2) two dimensions (positive vs. negative components) and (3) one dimension (a so-called g-factor model). According to the theory on well-being in school (Hascher, 2004), we expected the best fit to be achieved by the six-dimensional model with six latent factors that were correlated. Measurement errors were not correlated. A second-order factor was not hypothesised, therefore a second-order model was not specified.
The results of the confirmatory factor analysis (CFA) revealed a satisfactory fit for the six-factor model of well-being in school (six distinct but correlated dimensions of student well-being in school), while the other two models yielded unsatisfactory fit statistics. Thus, the six-factor structure of well-being in school was supported by the data (see Table 1). We also tested for measurement invariance. The results suggest scalar invariance for all variables. Details on these results can be obtained from the authors.  With respect to the scale mean, the results indicated that the positive indicators of well-being in school (PAS, EIS, PASC) were rated higher than the negative scales (WIS, SPS, PCS) by the students. The comparatively low mean level pertaining to enjoyment in school (EIS; M = 3.52) and the relatively high mean value pertaining to worries in school (WIS; M = 3.22) merit attention. The correlations confirmed the expectations. The negative dimensions of wellbeing in school correlated negatively but weakly with the positive scales, while the correlations within the negative dimensions were positive on a moderate level. The same was found for the positive scales. School reluctance correlated on a moderate level with indicators of well-being in school, while the correlations with truancy were somewhat weaker (see Table 2). Note. The manifest correlations were calculated with the IDB analyser, accounting for the complex sampling design (www.iea.nl) (below the diagonal), the latent correlations were calculated with Mplus (above the diagonal). In both analyses, missing values were deleted listwise. Abbreviations: PAS = Positive attitudes towards school; EIS = Enjoyment in school; PASC = Positive academic self-concept; WIS = Worries in school; SPS = Social problems in school; PCS = Physical complaints in school; SRT = School reluctance; TRU = Truancy. *** p < .001; ** p < .01; * p < .05

Group differences in well-being in school
According to our first research question, we were interested in whether particular student subgroups exhibited differences in well-being in school. To test for group differences, the analysis module of the IDB analyser (percentages and means) was used (see https://www.iea.nl/data-tools/tools). The results with respect to gender, region, origins, and school type are displayed in Figures 1 to 4. As the p value is not informative for large sample sizes, as even minor differences become significant, we also report the effect size (Cohen´s d), which reveals the practical significance of the group differences. 5 T h e m a Gender: Female students reported more positive attitudes towards school (t = 11.93; d = 0.22; p < .001) and more enjoyment in school (t = 5.04; d = 0.10; p < .001); however, their positive self-concept was lower (t = -5.98; d = 0.12; p < .001) than that of boys. Girls also exhibited more worries in school (t = 20.19; d = 0.37; p < .001), more social problems (t = 8.31; d = 0.14; p < .001), and more physical complaints in school (t = 26.59; d = 0.44; p < .001). The difference was particularly high for physical complaints and worries in school, while the other effect sizes were rather small (see Figure 1).

Figure 1: Differences in well-being in school, comparing girls and boys
Region: In terms of the region in which the students resided, some pronounced differences were observed. Students from the Italian-speaking part of Switzerland reported higher enjoyment in school than their German-speaking (t = 10.11; d = 0.42; p < .001) and French-speaking (t = 13.38; d = 0.56; p < .001) counterparts; however, they also exhibited more worries in school than students in the Germanspeaking (t = 19.53; d = 0.75; p < .001) and French-speaking (t = 11.51; d = 0.46; p < .001) regions. Moreover, students from the Italian-speaking part experienced more social problems in school than their German-speaking (t = 7.63; d = 0.37; p < .001) and French-speaking (t = 6.13; d = 0.28 p < .001) counterparts. In addition, their physical complaints were higher than those of the Germanspeaking (t = 5.77; d = 0.27; p < .001) and French-speaking students (t = 2.24; d = 0.12; p < .05). The positive attitudes towards school of the Italian-speaking students were also slightly higher than those of the French-speaking students (t = 2.15, d = 0.10, p < .05), while they were not significantly different from those of the German-speaking students (p > .05). Furthermore, the Italian-speaking students' academic self-concept was not significantly different from that of the French-and German-speaking students (p > .05).

Figure 2: Differences in well-being in school, comparing students from the German-speaking part of Switzerland with students from the French-speaking part and the Italian-speaking part of Switzerland
Students' origins: Taking a look at the positive dimensions of well-being in school, students with a migration background reported more positive attitudes towards school (t = 7.83; d = 0.18 for second-generation youth and t = 3.35; d = 0.11 for first-generation youth; p < .001) and more enjoyment in school (t = 7.63; d = 0.19 for second-generation youth and t = 7.13; d = 0.22 for firstgeneration youth; p < .001) than their Swiss native counterparts. Academic selfconcept did not differ substantially (t = 1.59; d = 0.04 for second-generation youth; p > .05; and t = 2.20; d = 0.06 for first-generation youth; p <. 05).
The comparison also revealed that students with a migration background exhibited more worries (t = 16.83; d = 0.36 for second-generation youth and t = 12.05; d = 0.39 for first-generation youth; p < .001) and more physical complaints in school (t = 12.47; d = 0.29 for second-generation and t = 9.54; T h e m a d = 0.29 for first-generation students; p < .001). Interestingly, first-generation students also reported more social problems in school than their native counterparts (t = 5.83; d = 0.17; p < .001), while this difference was negligible for the second-generation students (t = 1.52; d = 0.03; p > .05).
First-and second-generation students did not differ with respect to wellbeing in school (p > .05) (see Figure 3).

Figure 3: Differences in well-being in school, comparing students without a migration background with students with a migration background (second-and first-generation youth)
School type: Students who attended a school with basic requirements differed from students who attended a school with advanced requirements, particularly in terms of their academic self-concept (in favour of the latter group; t = -14.81; d = 0.59; p < .001). The other effect sizes can be classified as smaller but still significant. Students who attended schools with basic requirements displayed less positive attitudes towards school (t = -8.21; d = 0.18; p < .001); however, they nonetheless reported more enjoyment in school (t = 6.95; d = 0.13; p < .001). With respect to the negative dimensions, students attending a school with basic requirements tended to show disadvantages: They exhibited more worries (t = 8.29, d = 0.14; p < .001), more social problems (t = 9.13; d = 0.16; p < .001), and more physical complaints in school (t = 10.16; d = 0.20; p < .001; see Figure 4).
T h e m a

Figure 4: Differences in well-being in school, comparing students attending schools with basic requirements to students attending schools with high/advanced requirements
Explaining school reluctance and truancy by well-being in school Finally, we ran multiple regressions (method: Enter) to explain school reluctance and truancy by well-being in school while controlling for gender, region, students' origins, and school type. The analysis module of the IDB analyser (linear regression) was used (see https://www.iea.nl/data-tools/tools). The test for multicollinearity of the predictors yielded satisfactory results. The variance inflation factor (VIF) was between 1.27 and 1.61, while the tolerance value was between 0.62 and 0.79, which indicates some correlation between the predictors, but no severe problem with multicollineartiy occurred.
Again, it is important to consider the effect sizes to meaningfully interpret the results, as predictors become significantly easier due to the large sample size. Based on the recommendation introduced by Keith (2019), standardised regression coefficients can also be interpreted with respect to effect size: ß < .05 = no meaningful effect; ß > .05 and < .10 = small effect; ß > .10 and < .25 = moderate effect; and ß > .25 = large effect. We only report the results of Model 2, incorporating all the predictors in the text (for all the coefficients of models 1 and 2, see tables 3 and 4).
School reluctance: Well-being in school contributed significantly to the explanation of school reluctance. While the control variables (gender, region, students' origins, and school type) only explained 3% of the variance in school reluctance, the explained variance increased to 37% once the dimensions of well-being in T h e m a school were entered into the model. When students report less positive attitudes towards school (ß = -.35) and less enjoyment (ß = -.18), school reluctance increases. By contrast, school reluctance is higher when students experience worries (ß = .09), social problems (ß = .15), and physical complaints (ß = .22) in school. Interestingly, school reluctance was also higher when students reported a high academic self-concept (ß = .06). This association, however, might be due to a suppressor effect, as the bivariate correlation between these two variables was clearly negative.
Moreover, under control of the well-being in school dimensions, girls exhibited less school reluctance than boys (ß = -.08). In addition, school reluctance among students from the French-speaking part of Switzerland (ß = .09) was higher than that of students in the German-speaking region. Differences with the Italian-speaking students were minimal and not meaningful (ß = .02), again in favour of the students from the German-speaking regions. Migration background did not substantially explain school reluctance. Finally, students who attended schools with basic demands reported higher school reluctance than students who attended schools with advanced demands (ß = .04); however, this difference only became evident when the well-being in school scales were entered into the model and is very low (indicating no meaningful effect). The mean values were nearly the same (M = 3.05 for students attending schools with basic requirements, compared to M = 3.06 for students attending schools with advanced requirements). Note. Multiple linear regression based on manifest variables analysed by means of the IDB analyser (Version: Core Skills Assessment Switzerland); *** p < .001; ** p < .01; * p < .05. Gender: 1 = male, 2 = female; Migration background: reference category is "no migration background"; Region: reference category is "the German-speaking part of Switzerland"; School type: 1 = basic requirements, 2 = high/advanced requirements.
T h e m a Truancy: Truancy was less well explained by well-being in school than school reluctance: Only 11% of the variance could be explained once all predictors had been entered into the model. Truancy was higher when students reported less positive attitudes towards school (ß = -.13), when they experienced less enjoyment in school (ß = -.05), and when they exhibited a lower academic selfconcept (ß = -.04). Truancy was also higher when students had social problems in school (ß = .05) and when they experienced physical complaints in school (ß = .13). Worries in school did not explain truancy.
With respect to the demographic variables, under control of the dimensions of well-being in school, girls reported less truancy (ß = -.07). The same can be said for students who attended schools with advanced requirements (ß = -.04). By contrast, students reported higher truancy in the French-(ß = .12) and Italianspeaking (ß = .03) parts of Switzerland than in the German-speaking regions. In addition, students with a migration background reported higher truancy (ß = .08 for second-generation and ß = .09 for first-generation youth. Overall, the effects are very small (see Table 4). Note. Multiple linear regression based on manifest variables analysed by means of the IDB analyser (Version: Core Skills Assessment Switzerland); *** p < .001; ** p < .01; * p < .05. Gender: 1 = male, 2 = female; Migration background: reference category is "no migration background"; Region: reference category is "the German-speaking part of Switzerland"; School type: 1 = basic requirements, 2 = high/advanced requirements.
T h e m a

Discussion and Conclusion
This study aimed to gain a deeper insight into Swiss adolescents' well-being in school. The results of the representative sample confirmed earlier findings (Hascher, 2004): Adolescent Swiss students generally experience a high level of well-being in school. This result corresponds with the findings of National Health Report (Schweizerisches Gesundheutsobservatorium, 2020). A closer look at the six dimensions of well-being in school revealed that they display positive attitudes towards school and also exhibit a positive academic self-concept with scores in the upper third of the scale range (1-6). This is combined with few social problems and physical complaints in school (both scores below 2); however, students reported less enjoyment and a prevalence of worries in school (both scores between 3-4). Thus, these two dimensions seem to be specifically exposed in the Swiss school context. The role of worries was also reported in the PISA 2015 results, where 48% of the students in Switzerland agreed or strongly agreed that they worry about difficult tests and 56% agreed that they worry about poor grades (OECD, 2017). The average across OECD countries, however, was even higher (around 60% of secondary students worry about difficult tests and 65% about poor grades in school). Given the fact that enjoyment and worries are directly related to learning motivation, learning behaviour, and educational outcomes (Pekrun, 2014), these findings point to an important aspect for improving both the academic learning process and instructional approaches. The results also revealed interesting subgroup-specific differences. These support the idea of taking a differentiated view on students' sense of well-being in school (consisting of independent yet interrelated sub-dimensions, as opposed to a common factor of general well-being).
(a) Gender differences were found in all six well-being in school dimensions, with differences in four of six well-being in school dimensions in favour of boys and to the disadvantage of girls. As found in earlier studies (e.g. Hascher, 2004), girls showed an ambivalent well-being in school pattern, as they are more prone to the positive well-being in school dimensions (positive attitudes and enjoyment) and are also more vulnerable to all three negative dimensions. Additionally, they reported lower scores in academic self-concept, which is in line with their lower domain-specific self-concept in mathematics (see Girnat, Hagenauer & Hascher, in this issue). Thus, girls seem to be more vulnerable regarding achievement pressure and social conflicts. Interestingly, gender differences are also a vivid and controversial topic in research on adult well-being (for an overview, see Batz & Tay, 2017) and have been discussed with respect to differences in need fulfilment, societal expectations, values, and biological differences. The openness of women to intense emotional experiences has also been discussed. With respect to the education system, further research is needed to test different explanations, such as socialisation effects. Accordingly, it might be worth investigating how schools could better capitalise on girls' resources regarding positive emotions and T h e m a attitudes towards school while simultaneously considering their susceptibility to developing academic worries, physical complaints, and peer problems. The student-teacher relationship (Liu et al., 2015) and the social dynamics in the classroom (Nordlander & Olofsdotter Stensöta, 2014) could play a key role.
(b) Regional differences were found in four of six well-being in school dimensions; no differences were found in the two positive dimensions positive attitudes towards school and positive academic self-concept. Overall, students from the German-speaking parts reported the most favourable scores in well-being in school. Students from the Italian-speaking parts reported the highest scores regarding worries, social problems, and physical complaints in school; however, at the same time, their enjoyment in school was higher than that of students in the German-speaking or French-speaking parts. This finding is of high interest, as students from the Italian-speaking regions reported predominantly high scores in general well-being (Castelli, 2019). Further evidence is needed, but it may be concluded that this group of students needs further attention with respect to their well-being in school because the negative dimensions clearly override their positive experiences, whereas their high scores in enjoyment in school indicate their openness to positive experiences. Future research may also analyse if cultural differences come into play when students report on the emotional dimensions of well-being in school, such as enjoyment, worries, and problems. Considering that the study was conducted in the last obligatory school year and close to trajectories into secondary II education, a dominance of negative experiences might harm students' further development.
(c) Differences with respect to students' migration background were found in four of six well-being in school dimensions (no differences in positive academic self-concept), revealing an ambivalent pattern for students with a migration background, who scored higher in two positive dimensions but also in two negative dimensions. For first-generation students, social problems were also more pronounced and might point to integration challenges. The higher scores for worries and physical complaints add to the existing evidence that, in the Swiss school system, students with a migration background are an at-risk group in terms of academic achievement (Breit, 2009). The results, however, also challenge prevailing findings concerning migrant students' unfavourable learning emotions (cf. Brandenberger et al., 2017), as students with a migration background also exhibited more positive attitudes towards school and reported more enjoyment in school. Similarly to girls' well-being in school, migrant students showed positive preconditions for learning (e.g. positive attitudes, positive emotions, and high motivation) but seemed to suffer from achievement pressure. One explanation might lie in the high expectations of migrant families regarding their childrens' education, which, in turn, can become a burden for the adolescents; for example, if parents' support for students' learning is insufficient (e.g. Plunkett et al., 2009) or the school lacks a supportive culture for migrant students (Mansel & Spaiser, 2010), the responsibility for educational success is T h e m a imposed on the children.
(d) With respect to the educational achievement level, differences in five of six well-being in school dimensions could be found. Students assigned to groups with basic requirements reported more worries, social problems, and physical complaints than students attending schools/classes with advanced requirements. Towards the end of compulsory education, they might be fully aware of their lower position in the educational system, which directly impedes their academic self concept. Specifically with respect to their pronouncedly lower academic self-concept, students who attend schools with basic demands appear to need more academic support. This can be related to the PISA 2012 finding that Swiss students in secondary schools with advanced demands seemed to be more strongly supported by extra tutoring than students in schools with basic demands (SKBF, 2018). Nevertheless, one exception can be found, as students with basic demands reported more enjoyment in school than students with advanced demands.
The results also confirmed that well-being in school is related to school reluctance and (to a smaller extent) to truancy: Five of six dimensions of well-being in school served as predictors for school reluctance. This indicates the preventive function of positive attitudes towards school and enjoyment in school, on the one hand, and a risk factor of worries, social problems, and physical complaints, on the other, with respect to the potential for troublesome school behaviour. Interestingly, only the three positive dimensions contributed to the explanation of truancy, while the negative dimensions did not, pointing again to the preventive potential of the positive dimensions of well-being in school. These findings can inform schools on how to make a difference in coping with such issues (Stamm, 2012) and supporting especially vulnerable students attending schools with basic demands, at-risk students with migration backgrounds, and male students because these three groups exhibited higher rates of truancy and dropping out (Stamm, 2012) and also showed higher scores in school reluctance and truancy in our study.
In sum, the results of this study show the empirical relevance of well-being in school; however, as a sub-study of the national large-scale assessment, it also bears some limitations, such as the cross-sectional design, the limited opportunities to analyse factors that potentially impact well-being in school, such as teacherstudent relationships or teaching quality, and the lack of data that would enable higher Cronbach's alpha values for school reluctance and school avoidance.
With respect to further research, it can be concluded that, as of yet, little is known about the factors that contribute to well-being in school and how wellbeing in schools can be increased. There is evidence that demonstrates the importance of teachers with respect to teaching quality and learning outcomes and the social-emotional classroom climate (e.g. Fend & Sandmeier, 2004;Pianta, 1999). From an educational perspective, it might also be important to determine how the specific quality of a school environment affects the phenomenon. Schools offer T h e m a different learning opportunities and support for student learning; this leads to variations in students' perceptions of their school (Roeser et al., 2000), which, in turn, contribute to well-being in school. Schools' different cultures and climates also play a role. Thus, future research might benefit from a closer look into the characteristics of schools and their instructional programmes. In taking such a look, researchers could examine predictors at the individual level and school level (and their interactions) simultaneously by using multi-level analysis, as school effects (as context effects) with respect to the quality of instruction might be substantial as well. For the well-being of students, the school-level effects, based on the ICCs, were rather low. Finally, well-being in school according to the theoretical conceptualisation provided by Hascher (2004) has not been tested in different countries so far. For future research, it seems worthwhile to conduct more international comparisons on students' well-being in school. Section II). Schools with high and advanced requirements were coded as "schools with high requirements", and schools with basic and low requirements were coded with "schools with basic requirements". Schools that did not differ with respect to requirements were coded as missing values. 3 Cronbach´s alpha values were calculated in SPSS, considering the student weight variable (smp_w_stubw). They are also reported by Sacchi and Oesch (2017). 4 The analyses in Mplus were also computed using the full information maximum likelihood estimation (FIML) procedure to impute missing values. The results based on FIML imputation did not (substantially) differ from the analyses using listwise deletion. Variations were not found before the third place after the decimal point. Thererfore, all results are reported based on listwise deletion of missing values. 5 To calculate the effect sizes, we relied on the following formula: Cohen's d = (M2 -M1) ⁄ SD pooled ; SD pooled = √((SD 1 2 + SD 2 2 ) ⁄ 2). The following thresholds, according to Cohen (1988), were applied: d = 0.20: small effect; d = 0.50: medium effect; d = 0.80: large effect.
T h e m a Keywords: Well-being in school, adolescence, large-scale assessment, school reluctance, school absenteeism