Teacher burnout explained : teacher-, student-, and organisation-level variables

This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details. Author(s): Title: Year: Version: Copyright: Rights: Rights url: Please cite the original version: CC BY-NC-ND 4.0 https://creativecommons.org/licenses/by-nc-nd/4.0/ Teacher burnout explained : teacher-, student-, and organisation-level variables © 2020 The Author(s). Published by Elsevier Ltd. Published version Saloviita, Timo; Pakarinen, Eija Saloviita, T., & Pakarinen, E. (2021). Teacher burnout explained : teacher-, student-, and organisation-level variables. Teaching and Teacher Education, 97, Article 103221. https://doi.org/10.1016/j.tate.2020.103221 2021 Research paper Teacher burnout explained: Teacher-, student-, and organisation-level variables Timo Saloviita * , Eija Pakarinen University of Jyv€ askyl€ a, Finland highlights The burnout profile differed between teacher categories. Burnout was highest among subject teachers, lowest among special education teachers. Class size had a minimal association with teachers’burnout. Students with special needs had a small association with teacher burnout. Availability of additional help was associated with lower level of teacher burnout. article info Article history: Received 15 February 2019 Received in revised form 21 September 2020 Accepted 7 October 2020 Available online xxx Keywords: Teacher burnout Teacher category Inclusive education Self-efficacy Teaching assistants abstract Understanding the factors related to teacher burnout helps in creating schools which foster teachers’job satisfaction and the delivery of high-quality education. We studied teacher burnout and its three subdomains across several teacher-, student-, and organisation-level variables, including teacher category, class size, number of students with support needs, attitudes towards inclusive education, and availability of support. The participants were 4567 Finnish primary school teachers consisting of 2080 classroom teachers, 1744 subject teachers, 438 special-class and 305 resource room teachers. Several associations between teacher burnout and the background variables were observed and recommendations made based on these results. ©2020 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 1. Introduction Burnout has been identified as a psychological risk, especially in fields which involve working with customers (Loonstra, Browers, & Tomic, 2009;Maslach &Leiter, 2016). It has been defined as “a psychological syndrome emerging as a prolonged response to chronic interpersonal stressors on the job”(Maslach &Leiter, 2016, p. 103). Burnout develops gradually when work becomes unpleasant, unfulfilling, and unrewarding. The concept is usually divided into three dimensions: exhaustion, lack of accomplishment, and feelings of cynicism towards and detachment from the job (Maslach, Schaufeli, &Leiter, 2001). Of these three dimensions, emotional exhaustion, or the feeling of being emotionally drained and worn-out, is identified as a central aspect of the whole construct (Maslach &Leiter, 2016;2017). Understanding burnout among teachers, as well as the factors related to it, has awakened considerable attention (Kyriacou, 2001). Teacher burnout has been shown to have significant negative implications not only for teachers’well-being in terms of their selfrated health (Hakanen, Bakker, &Schaufeli, 2006), mental health (Schonfeld &Bianchi, 2016), and job satisfaction (Klassen, Usher, & Bong, 2010;Robinson, Bridges, Rollins, &Schumacker, 2019; Skaalvik &Skaalvik, 2009;2011) but also for student achievement (Herman, Hickmon-Rosa, &Reinke, 2018;Klusmann, Richter, & Lüdtke, 2016) and adjustment (Jennings &Greenberg, 2009; Oberle &Schonert-Reichl, 2016). Among teachers, burnout has also been related to high absenteeism, retirement, and turnover rates (Ingersoll &May 2012;Schonfeld, 2001) and a lower quality of job performance (Klusmann, Kunter, Trautwein, Lüdtke, &Baumert, 2008). As teacher burnout has negative consequences at the individual teacher-, student-, organizational, and societal levels, factors *Corresponding author. E-mail address: timo.saloviita@jyu.fi(T. Saloviita). Contents lists available at ScienceDirect Teaching and Teacher Education journal homepage: www.elsevier.com/locate/tate https://doi.org/10.1016/j.tate.2020.103221 0742-051X/©2020 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/ ). Teaching and Teacher Education 97 (2021) 103221 related to teacher burnout require more attention. Teaching has been ranked as one of the most stressful professions in various cultural and educational contexts. In the United Kingdom, teaching was one of the most stressful among 26 occupations (Johnson et al., 2005). In Finland, teachers experienced stress and burnout more frequently (12%) than other professions (8%) (Kauppinen et al., 2010). Teachers also suffered the highest level of burnout compared to other workers in the human services and white-collar jobs (Kalimo &Hakanen, 2000). The present study draws on the job demands-resources (JD-R) model (Demerouti, Bakker, Nachreiner, &Schaufeli, 2001), which proposes that teacher stress and burnout are predicted by their perceptions of two factors, namely, job demands and resources (Hakanen et al., 2006). The central assumption of this model is that job stress arises when a person’s resources have been exceeded (Bakker &Demerouti, 2007). School is a complex environment where job demands include several individualand school-level aspects, such as work overload, role conflict, school climate, conflicts with colleagues, and students’behavioural problems (Hakanen et al., 2006;Pyh€ alt€ o, Pietarinen, &Salmela-Aro, 2011; Skaalvik &Skaalvik, 2010). Job resources consist of, for example, teacher efficacy, support from one’s colleagues and principal, participation in decision-making, public recognition, and professional development (Rudow, 1999). The literature has provided consistent support for the JD-R model and the primary role of job demands and resources in burnout (see Fernet, Austin, Tr epanier, & Dussault, 2013, for a review; Hakanen et al., 2006). Job demands and resources can arise from the individual teacher-, student-, and organisational levels. Thus, the present study adopts a more integrative approach by investigating both the individual and environmental factors related to burnout. In empirical research, teachers’stress and burnout have been linked to several factors. The first category is teacher and student characteristics (Engelbrecht, Oswald, Swart, &Eloff, 2003; Klusmann et al., 2008; Kokkinos, 2007). Organisational and workrelated variables also play a significant role in teacher stress. These variables include time pressure (Hakanen et al., 2006; Kyriacou, 2001; Schaufeli &Bakker, 2004), relationships with colleagues (Hakanen et al., 2006; Leung &Lee, 2006; Schaufeli & Bakker, 2004), and the availability of various forms of support (Skaalvik &Skaalvik, 2011). Teacher burnout has been found to be related in particular to the quality of the social interactions occurring within the school community (Fernet, Guay, Sen ecal, &Austin, 2012). 1.1. Teacher-level variables Among the teacher-level factors, teacher burnout has been studied in relation to demographic variables and attitudinal dimensions. Regarding gender, female teachers have often been more exhausted than their male counterparts (Antoniou, Polychroni, & Vlachakis, 2006;Fernet et al., 2012;Pyh€ alt€ o et al., 2011; Santavirta et al., 2000). They have also reported lower levels of personal accomplishment (Lau, Yuen, &Chan, 2005), while male teachers have expressed higher levels of depersonalisation (Lau et al., 2005; Schwab et al., 1986). With regard to burnout, the association between age and work experience is less clear. Older teachers are often more exhausted than younger ones (Klusmann et al., 2008;Santavirta et al., 2000), but younger teachers have sometimes reported a higher level of emotional exhaustion than their experienced counterparts (e.g., Antoniou et al., 2006; Brunsting, Sreckovic, &Lane, 2014;Lau et al., 2005). Teacher efficacy has been defined as teachers’confidence in their individual and collective capabilities to influence students’ performance and especially their learning (Brouwers &Tomic, 2000;Klassen, Tze, Betts, &Gordon, 2011). There is emerging evidence of the positive relationship between teachers’self-efficacy and student learning (Collie, Shapka, &Perry, 2012;Klassen & Chiu, 2010). A review of 23 studies by Zee and Koomen (2016) indicated that students’motivation was predicted more consistently by teacher self-efficacy than by students’academic achievement. Lower teacher self-efficacy is associated with higher burnout and stress. Teachers with greater reliance on their capabilities, especially in regard to classroom management, have lower levels of burnout (Betoret, 2006;Dicke et al., 2014;Evers, Brouwers, &Tomic, 2002;Friedman, 2003;Friedman &Farber, 1992; Malinen &Savolainen, 2016; Skaalvik &Skaalvik, 2007,2010;Zee & Koomen, 2016). The connection between these concepts may be partly intrinsic, because professional failure and sense of inefficacy are principal factors in burnout (Friedman, 2003). It has been suggested that teachers’relatedness with students is also important for teachers’own well-being (Milatz, Lüftenegger, & Schober, 2015;Spilt, Koomen, &Thijs, 2011). In line with selfdetermination theory (Deci &Ryan, 2000), teachers have a basic need for relatedness. Relationships with students can be an important source of positive energy, enjoyment, and reward and can guide teachers’daily emotions and cognitions in the classroom (Milatz et al., 2015). Positive teacherestudent relationships are generally characterised by relatedness, respect, warmth, support, trust, and low levels of interpersonal conflict (e.g., Pianta, 1999; Roorda, Koomen, Spilt, &Oort, 2011). Positive relationships are facilitated by structures, which give teachers the possibility of developing standing relationships with each student, such as in primary school where teachers have less students than in upper levels. Warm and supportive relationships between teacher and students foster positive classroom climates (Klassen, Perry, &Frenzel, 2012) and increase positive learning outcomes (Cornelius-White, 2007;Kiuru et al., 2012;Lerkkanen et al., 2016). Teachers who devoted their time and energy to forming warm and supportive relationships with their students had higher work-related wellbeing (Spilt et al., 2011) and experienced lower levels of emotional stress and burnout than those who had more distant relationships with their students (Hoglund, Klingle, &Hosan, 2015;Jennings & Greenberg, 2009;Milatz et al., 2015). Teachers’relatedness with their students was associated with lower emotional exhaustion (Klassen, Perry, &Frentzel, 2012; Virtanen et al., 2018), higher work enthusiasm (Aldrup et al., 2018), and increased job satisfaction (Virtanen et al., 2018). Teacher attitudes towards students can also be related to their experiences of burnout. So far, studies on teacher burnout and inclusive education have concentrated on the possible factors that explain teachers’stress (Brackenreed, 2011;Forlin, 2001), while studies on the association between burnout with teachers’attitudes towards inclusion seem to be lacking. As an exception, Talmor, Reiter, and Feigin (2005) showed that teachers’attitudes towards inclusion were significantly associated with burnout: The more positive the teacher’s attitude, the greater the experience of burnout. Furthermore, higher levels of burnout have been linked to the lower level of support obtained from principals, colleagues, and students’parents (Brunsting et al., 2014;Skaalvik &Skaalvik, 2011). 1.2. Student-level variables The student-level variables related to teacher burnout include, first and foremost, the ages of the students and their support needs. Students’age has typically been operationalized as grade level. The grade level taught has been shown to have a significant association with teacher burnout: Secondary school teachers tend to T. Saloviita and E. Pakarinen Teaching and Teacher Education 97 (2021) 103221 2 experience higher levels of depersonalisation and reduced personal accomplishment than elementary school teachers (Schwab & Iwanicki, 1982). Teachers in the upper grades have been shown to experience more burnout than those in lower grades (Arvidsson, Håkansson, Karlson, Bj€ ork, &Persson, 2016). As an exception, Antoniou, Ploumpi, and Ntalla (2013) indicated that Greek primary school teachers experienced higher emotional exhaustion than teachers of secondary education. The principle of inclusive education contains the idea of placing students with disabilities in the mainstream classrooms (United Nations Educational, Scientific and Cultural Organization [UNESCO], 1994). Increased student heterogeneity in the classroom may, in turn, have negative consequences for the teachers. In particular, students’behavioural problems have been associated with higher levels of teacher dissatisfaction and stress (Aloe, Shisler, Norris, Nickerson, &Rinker, 2014;Organization for Economic Co-operation and Development [OECD], 2014). Few studies have investigated the association between teacher burnout and the presence of students with special education needs (SEN) in the classroom. A study from Israel found that teachers’ level of depersonalisation was higher when the number of students with SEN exceeded 20% of the classroom population (Talmor, Reiter, &Fejgin, 2005). In another study, fatigue and depersonalisation among physical education teachers were higher when they had students with behaviour problems; however, their learning problems were not associated with teacher burnout (Fejgin, Talmor, & Erlich, 2005). The inclusion of students with moderate to severe intellectual disabilities was found to be somewhat stressful for regular education teachers, with the most stressful issues being the division of time between students and concerns regarding their educational outcomes (Forlin, 2001). In a replication of this study, the most stressful items were associated with teachers’perceptions of their self-competency and of the behaviour problems of the child (Brackenreed, 2011). The most frequently mentioned problems were the necessity of compromising with regard to the teaching of other students because of the presence of a child with SEN, mentioned by 83% of the teachers, the modifications needed to accommodate students with SEN (85%), and the behavioural problems of these students (80%) (Brackenreed, 2011). Some intervening variables have mediated the effects of student characteristics on teacher burnout. Teachers have suffered more stress or burnout in inclusive classrooms when they estimated that their own skills were inadequate (Forlin, 2001) or that outside support was lacking (Talmor et al., 2005). Experience as a teacher in inclusive classrooms and participation in formal training were associated with lower burnout (Forlin, 2001). Teacher burnout was also lower when certain organisational features were present, such as independence at work, a clear division of labour, rapid student assessment, effective instructional planning, and recognition of the positive consequences of inclusion. These social and organisational variables explained 24% of the total variance in the teachers’ burnout (Talmor et al., 2005). In another study, teachers, who used the versatile methods recommended in the implementation of inclusive education, suffered less stress than other teachers (Weiss, Muckenthaler, Heimlich, Kuechler, &Kiel, 2019). 1.3. Organisation-level variables Teachers’job-specific tasks as well class size, school size, and the availability of support are among the significant organisation-level variables related to burnout. Subject teachers’overall burnout has been found to be significantly higher than burnout among class and special education teachers (Pietarinen, Pyh€ alt€ o, Soini, &SalmelaAro, 2013). Pietarinen et al. (2013) also demonstrated that subject teachers experienced the highest level of inadequacy, whereas special education teachers reported the lowest levels of inadequacy. Burnout among special education teachers was examined by Brunsting et al. (2014), who reviewed 23 studies conducted on the subject between 1979 and 2013. They found that the variables associated with burnout in this population included the lack of resources, workload problems, and teachers’role ambiguity. It has also been indicated that larger class size is related to higher levels of burnout and exhaustion (French, 1993;Travers & Cooper, 1996). Furthermore, Skaalvik and Skaalvik (2010) indicated that larger school size plays a role in teacher burnout, as manifested in decreased job satisfaction, lower accomplishment, and a sense of depersonalisation. In a Finnish study by Pietarinen et al. (2013), teachers working in small schools (less than 100 pupils) experienced less burnout symptoms than those operating in medium-sized (301e450 pupils) or large schools (over 600 pupils). Burnout has also been linked to the level of support obtained (Brunsting et al., 2014;Skaalvik &Skaalvik, 2011). Thus, it could be assumed that the availability of a teaching assistant in a classroom is related to less burnout. 1.4. Aims of the study Better understanding of the factors related to teacher stress and burnout helps to create school environments that foster teachers’ job commitment, prevent dropout from the profession, and raise the quality of education. While numerous aspects of teacher burnout have been thoroughly investigated, others deserve further study, and many of them were selected to this study. We used a large and representative sample of Finnish primary school teachers from different categories to survey the possible factors explaining teacher burnout. First, we investigated the structural validity of the Friedman Burnout Scale among a large Finnish sample of teachers. It was expected that the suggested three-factor structure would fit the data well (H1). The further aim of the present study was to examine the extent to which various teacher-, student-, and organisation-level variables were related to teacher burnout. The individual teacher-related variables selected for this study were the teachers’age and gender, attitudes towards inclusive education, sense of self-efficacy, relatedness with students, and perceptions of the sufficiency and availability of support. At the teacher level, it was expected that higher levels of burnout would be associated with older age (H2), the male gender (H3), negative attitudes towards inclusive education (H4), a lower sense of efficacy (H5), lower relatedness with students (H6), and lower perceptions of sufficiency and availability of support (H7). The student-related variables that were selected were grade level taught by teachers, the number of students with SEN and intensive support needs (ISN) in the class. It was hypothesised that burnout would increase for teachers in the upper grades (H8) and when the number of students with support needs increased (H9). The organisational variables included teacher category, principal’s duties, class size, size of school and municipality, and availability of teaching assistants (TAs). It was expected that subject teachers (H10) and teachers with larger classes (H11) and no teaching assistant (H12) would experience higher levels of burnout. 2. Methods 2.1. Participants The participants of this study were 4567 Finnish primary school teachers. The criterion for inclusion in the study was that the participant was currently working as a primary school teacher. The participants’distribution across current teacher position, formal T. Saloviita and E. Pakarinen Teaching and Teacher Education 97 (2021) 103221 3 qualification, and gender is presented in Table 1. Of the participants, 262 also worked as school principals or vice-principals. The mean age of the participants was 45.7 years (SD ¼9.5). In terms of age, the sample was representative of the Finnish teacher population (Finnish National Agency for Education, 2016). The number of formally qualified teachers in this study (96%) corresponded with the number obtained in a large Finnish study conducted in 2016 (95%) (Finnish National Agency for Education, 2016). The percentage of female teachers in each category was also close to the numbers reported in a previous large study (Finnish National Agency for Education, 2016). The subject teachers were classified into four subgroups based on their majors: languages (N ¼667), science and mathematics (N ¼481), humanities (N ¼204) and arts, crafts and physical education (N ¼472). 2.2. The Finnish primary school and its teachers To qualify as a primary school teacher of any type, the completion of five years of university studies, leading to a master’s degree, is almost always a necessity (Government of Finland, 1998). The classroom teacher instructs Grades one to sixdthat is, students between the ages of 7 and 12. The subject teacher is mainly responsible for the upper schooldthat is Grades seven to nine (ages 13 to 15). Special education teachers are either special-class teachers or resource room teachers. The former instruct selfcontained classrooms of students with SEN and have at most 10 students in their class. Resource room teachers are special education teachers who encounter a changing body of students in their offices. They give elementary speech therapy, instruct in difficult school subjects, or merely provide temporary placement for students who exhibit disturbing behaviour. In 2017, about 22% of primary school students participated in this activity, which is known as “part-time special education”(Statistics Finland, 2018). Primary school students in Finland are categorised according to their needs for general, intensified, or special support (Government of Finland, 2010). “Students with intensified support needs”(ISN) is a Finnish specialty which came into force in 2010 through a change in the school legislation. The change created a continuum of supports where ISN falls between the supports for typical students and those designated as SEN. The aim of the new three-tier system was to limit the ever-increasing number of students being categorised as having SEN and transferred to special education classrooms. The official ISN category was created to make it easier for the teacher to obtain additional help. ISN was also intended to be the necessary first step before a student can be identified as having a need for special support, which is equivalent to the expression “special educational needs”(SEN) in many other countries. Although, in principle, students with ISN remain in their mainstream classrooms, students with SEN are usually transferred to special education classrooms. Such transfers are probably more common in Finland than in any other European country (EADSNE, 2018; Statistics Finland, 2018). 2.3. Data collection In total, 95 preservice teacher participating in a scientific methodology course collected data in the spring term 2017. Each student or a group of students obtained a sample of Finnish municipalities. Of the 317 Finnish municipalities, only 302 were needed to fulfil the data collection requirements given to the participants. After collecting teachers’email addresses from the schools’official websites, the students sent each potential participant an email containing a link to the survey. The cover letter of the email stated that participation was fully anonymous and voluntary. Individual participants, schools, and municipalities could not be identified from the data. The ethical standards of the Finnish Advisory Board on Research Integrity (2012) were followed in the data collection. If the teachers’addresses were not available on the schools’websites, the schools were omitted from the study. Between 14,000 and 15,000 letters were distributed, and approximately 30e33% of the participants who were approached completed and returned the survey. 2.4. Survey instrument The survey included individual items on TA support (“I would prefer to have help in my class from a special education teacher than from a teaching assistant”), evaluation of adequacy of support (“In my classroom, the support is sufficient”and “I believe that in my classroom, I get enough assistance and support if needed”) and demographic variables. These variables are identified in the relevant passages of the Results section. The survey also contained scales measuring teachers’burnout, attitudes towards inclusion, sense of self-efficacy, relatedness with students, and available sources of support. Teachers were asked to answer on a five-point Likert scale ranging from “strongly disagree”(score one) to “strongly agree”(score five) with a neutral midpoint. 2.4.1. The Friedman Teacher Burnout scale (FTBS) The teachers’burnout was measured using Friedman’s (1999) adaptation of the second edition of the Maslach Burnout Scale (Maslach &Jackson, 1986). The original scale was divided into three subdomainsdexhaustion, lack of accomplishment, and depersonalisation. Friedman retained the three-factor structure but reduced the scale items from 22 to 14 and adapted their wording to make it applicable to teachers. Five items were used to measure emotional exhaustion or overextension: (a) “I feel exhausted from teaching”, (b) “I feel burned out from teaching”, (c) “I feel worn out from teaching”, (d) “I feel wiped out by the end of a day of teaching”, and (e) “I feel physically worn out by teaching”. In the section titled “lack of accomplishment”, the participants’feelings of incompetence and lack of achievement were measured using five items: (a) “I feel I could have better used my professional and personal capabilities in a profession other than teaching”, (b) “I do not feel that I fulfil myself in teaching”, (c) “I feel that I am not doing so well as a teacher”, (d) “I think that if I had to choose again, I would not choose teaching”, and (e) “I feel my expectations of teaching have not been met”. The section on depersonalisation measured negative and impersonal sentiments towards students using four items: (a) “I feel that my students do not really try enough”, (b) “I feel that my students do not really care about being good students”, (c) “I feel that my students do not really want to learn”, (d) “I think that I would rather have better students than those I have now”. The psychometric properties of the FTBS are presented in the results section. Table 1 Study participants. Teacher category N Qualified % Gender Male % Female % Total % Classroom teacher 2080 97 19 81 100 Subject teacher 1744 98 23 77 100 Special-class teacher 438 86 16 84 100 Resource room teacher 305 93 7 93 100 Total 4567 96 81 19 100 T. Saloviita and E. Pakarinen Teaching and Teacher Education 97 (2021) 103221 4 2.4.2. Teacher Attitudes towards Inclusive Education Scale short form Teachers’attitudes towards inclusive education were measured using three items: (a) “The best result is achieved if each child with SEN is placed in a special education classroom that best suits to him or her”, (b) “The education of students with SEN should be arranged as far as possible in mainstream classrooms”, and (c) “Children with special educational needs learn best in their own special education classes, in which they have specially trained teachers”. These items were selected from the 10-item Teacher Attitudes towards Inclusive Education Scale (TAIS) (Saloviita, 2015) based on their high itemetotal correlations. In a Finnish sample, the TAIS-SF explained 86% of the variance of the original TAIS scale (Saloviita &Tolvanen, 2017). The reliability of the TAIS-SF in the present sample was a ¼0.82, which was a good level. 2.4.3. Teacher sense of efficacy in Inclusive Education Scale Teachers’sense of self-efficacy was measured using a three-item scale constructed on the basis of the Teachers’Sense of Efficacy Scale (TSES) (Tschannen-Moran &Woolfolk Hoy, 2001). The original scale contained 24 items divided into three factors. In the present study, three items were chosen so that each represented a factorial dimension discovered in TSES. The dimension of efficacy for instructional strategies was covered by the item (a) “I can teach many kinds of students, including students with special educational needs”; the dimension of efficacy in classroom management was measured by the item (b) “I can keep good order in my classroom”; and the dimension of efficacy for student engagement was assessed by the item (c) “I get students to rely on their own abilities”. The reliability of the TSEIES, as measured by Cronbach’s alpha, was a ¼0.61, which was just above the poor level. The reliability was somewhat low because the first item focusing on children with special education needs differed from the other two. 2.4.4. Relatedness with students scale Four items from the Teachers’Interactional Style Scale (Aunola, Lerkkanen, Poikkeus, &Nurmi, 2005;Kiuru et al., 2012)were selected to measure teachers’positive and warm behaviour towards their students. The original scale measured the extent to which the teachers perceived that their relationships with their students were characterised by affection, sensitivity, and responsiveness to the needs and interests of the children (Kiuru et al., 2012). The chosen items were (a) “I like to talk with my students”, (b) “In the company of my students, I am usually uncomplicated and relaxed”,(c) “I like to ask how my students are doing”, and (d) “I often show my students that I care about them”.The reliability of the RSS, as measured by Cronbach’s alpha, was a ¼0.84, thereby indicating a good level. 2.5. Analytic strategy The data were analysed using the IBM SPSS Statistics 24 and Mplus version 8.0 (Muth en &Muth en, 1998e2015). First, confirmatory factor analysis (CFA) was used to examine the structural validity of the FTBS measure as a three-factor model. The goodnessof-fit of the estimated model was evaluated using the Chi-Square test, the Comparative Fit Index (CFI), the Tucker-Lewin Index (TLI), the Root Mean Square Error of Approximation (RMSEA), and the Standardised Root Mean Square Error of Approximation (SRMR). A non-significant Chi-Square value, CFI and TLI values above 0.95, a RMSEA value below 0.06, and an SRMR value below 0.08 indicated a good fit with the data (Muth en &Muth en, 1998e2015). Second, burnout was analysed in relation to teacherlevel variables, which consisted of the teachers’gender, age, and attitudes. The teachers’evaluations of the adequacy of support were considered as teacher-level variables, while the actual presence of teaching assistants (TA) was considered as an organisationlevel variable. Third, student-level variables, that is, student age and support needs, were analysed in relation to teacher burnout. Next, teacher category (classroom, subject, special class, or resource room teacher), class size, size of the school and municipality, and TA availability were used as organisational variables. The relation of teacher burnout to different factors was investigated using the ttest, hierarchical regression analysis, the one-way analysis of variance (ANOVA), and the two-way analysis of variance. Bonferroni post hoc tests were used for all multiple comparisons. Effect sizes are also provided because large sample sizes tend to derive meaning from statistical significance. 3. Results 3.1. Psychometric properties of the FTBS The distribution of the FTBS deviated from normalcy, having a thick right tail in the direction of higher burnout. The value of the skewness was 0.298, and the standard error of the skewness, SD ¼0.038, indicating that the score was more than twice its standard error. The reliability of the FTBS was a ¼0.90 or excellent. The reliabilities of the subscales were of good leveld a ¼0.87 for exhaustion, a ¼0.80 for lack of accomplishment, and a ¼0.84 for depersonalisation. CFA was performed for the FTBS to study the structural validity of the measure among a sample of Finnish teachers. The results indicated that the three-factor structure described the data satisfactorily: c 2 (74) ¼1825.926, p<0.001, SRMR ¼0.049, RMSEA ¼0.072, CFI ¼0.931, and TLI ¼0.915. The modification indices (MI ¼358.095) suggested that the model fit would increase after allowing the residual correlation between items “I feel I could have better used my professional and personal capabilities in a profession other than teaching”and “If I had to choose again, I would not choose teaching”and between the items “I feel burned out from teaching”and “I feel wiped out by the end of a day of teaching”(MI ¼222.942). The correlated residuals seemed justifiable, as the suggested model seemed to have a shared method variance due to the different wording, compared to other indicators, and specific item content (see Brown, 2015). The correlated residuals among the factor indicators in CFA might represent a minor factor (see Cole, Ciesla, &Steiger, 2007). After these modifications, the suggested three-factor structure fit the data well: c 2(72) ¼1288.662, p <0.001, SRMR ¼0.044, RMSEA ¼0.061, CFI ¼0.952, and TLI ¼0.939. In addition, to test whether the suggested three-factor structure fit across all teacher groups, we carried out multiple-group analyses. This was done by creating a nonrestricted model and comparing its fit to another model that was fully restricted across teacher categories (i.e., subject teachers, classroom teachers, and special education teachers). The chi-square likelihood ratio test between the non-restricted and the restricted models suggested that there were no differences in the models between the teacher categories. 3.2. Teacher-level variables and burnout The correlation between age as a continuous variable and burnout was not statistically significant. For a closer study, age was divided into groups of ten years and an analysis of variance was conducted. When analysed in age groups of 10 years using a polynomial contrast, a linear trend was observed, thereby indicating the lowering of the total burnout score along with age: F(4, 4382) ¼3.31, p¼0.010 ( h 2 ¼0.003). The effect was statistically T. Saloviita and E. Pakarinen Teaching and Teacher Education 97 (2021) 103221 5 significant only in exhaustion factor, F(4, 4466) ¼6.72, p<0.000 ( h 2 ¼0.006). Age did not have an effect on depersonalisation F(4, 4477) ¼1.10, p¼0.355 ( h 2 ¼0.001) or lack of accomplishment F(4, 4480) ¼2.36, p¼0.051 ( h 2 ¼0.002). Among the classroom teachers, the males had higher total burnout than the females: t(1981) ¼3.27, p¼0.001, d¼0.19. The effect was found in the lack of accomplishment, t(2036) ¼4.40, p<0.000, d¼0.25, and the depersonalisation factor, t(2029) ¼5.23, p<0.000, d¼0.27, but not in exhaustion. Among the subject teachers, no differences existed between genders in regard to total burnout; however, the female teachers exhibited higher levels of exhaustion than their male counterparts: t(1703) ¼3.02, p¼0.003, d¼0.17. Among the special-class teachers, no differences existed between the genders in regard to in the total burnout; however, the male teachers expressed higher levels of depersonalisation: t(420) ¼3.86, p<0.000, d¼0.54. Furthermore, the results of ANCOVA indicated that male teachers had higher lack of accomplishment (F(1721) ¼11.18, p<0.01, h 2 ¼0.006) and depersonalisation (F(1715) ¼11.10, p<0.01, h 2 ¼0.006) than females, when accounting for students’age and teacher category. The correlations between burnout and attitudes towards inclusive education, self-efficacy, and relatedness with students are provided in Table 2. The results show that higher burnout was associated with more negative attitudes towards inclusion, a lower sense of self-efficacy, and weaker relatedness with the students. In regard to all three variables, the correlations were on the same approximate level for all four teacher categories. The correlation between burnout and support for inclusion remained lower than in general among only the special-class teachers: (r¼0.118, p<0.05). Burnout also correlated positively with the statement “Ability-based groups should be returned to school”(Table 2). The greatest effect was found in the factor of depersonalisation; the effect size in regard to depersonalisation between the teachers who agreed or disagreed with the statement was d¼0.43. The teachers were also asked to respond to the statement “The integration of a student with SEN into the mainstream classroom causes additional work for the teacher”. The expectation of additional work caused by inclusion was associated with somewhat higher burnout among classroom teachers, r¼0.156, p<0.001; subject teachers, r¼0.150, p<0.001; and special-class teachers, r¼0.211, p<0.001, but not among resource room teachers. The teachers’perceptions regarding sufficiency of support were surveyed using the following single statements: “In my classroom, the support is sufficient”and “I believe that in my classroom, I get enough assistance and support if needed”. The correlations between these items and burnout are provided in Table 2. Teachers’ preferences were measured by the statement “I would prefer to have help in my class from a special education teacher than from a teaching assistant”. This statement received a positive response from 60% of the classroom teachers and 45% of the subject teachers, while a smaller portion (19% and 18%, respectively) preferred receiving help from TAs. The classroom teachers who preferred receiving help from special education teachers scored higher in regard to burnout than those who preferred receiving help from TAs: F(2, 1457) ¼8.88, p<0.000. The same was true among subject teachers: F(2, 1298) ¼7.28, p¼0.001. A hierarchical regression analysis with stepwise method was conducted to investigate the role of teacher-level variables on burnout. In line with our hypotheses, the results (Table 3) indicated that relatedness with students, teacher efficacy, perceptions of sufficiency of support, and attitudes towards inclusion were negatively associated with burnout and all its subdomains. Trust in getting support if needed was negatively linked to other burnout scores but not to depersonalisation. 3.3. Student-level variables in relation to burnout The student-related variables in this study were grade level, the number of SEN and ISN students in the teachers’classrooms. Their effects on teacher burnout were studied jointly with the presence of TAs. The grade levels of the classroom teachers varied between 1 and 6. A linear trend existed in the teachers’burnout, F(1) ¼9.81, p¼0.002, with the teachers of upper-grade students reporting higher levels of burnout than the others. When the elementary Grades 1e2 and 3e6 were compared, the difference was observed in the total score, p¼0.002, with an effect size of d¼0.16. The teachers in Grades 3e6 felt more unaccomplishment, p¼0.043; and depersonalisation, p<0.000; but no more exhaustion, p¼0.418, compared to the teachers at the elementary level. A linear trend was also found in the subdomain of depersonalisation, in which a further comparison between Grades 1 and 6 produced an effect size of d¼0.47. However, no differences existed between the teachers who were instructing only one grade and those who had students from several grades in their classrooms. The average number of students in these multiple-grade classrooms was 18, which undercut the other classrooms by only two students. Among the special-class teachers, those who instructed students in Grades 1 and 2 had a lower overall level of burnout, t(71) ¼2.22, p¼0.030, and a lower level of depersonalisation, t(75) ¼3.66, p<0.000, than the teachers of students in Grades 3 to 6. Table 2 Pearson correlations between study variables and teachers’burnout (N ¼4567). Variable Burnout Exhaustion Lack of accomplishment Depersonalisation Attitudes towards inclusive education -.266** -.202** -.220** -.239** Ability-based groups should be returned .232** .122** .197** .278** Teacher efficacy -.338** -.242** -.304** -.293** Relatedness with Students -.279** -.179** -.278** -.239** Availability of support -.315** -.303** -.248** -.225** Trust in getting support -.330** -.336** -.280** -.200** Gender .055** -.027 .076** .102** Age -.033*-.056** -.011 -.005 Class size .078** .118** .069** -.010 Students with support needs (SEN) .079** .033*.050** .131** Students with support needs (ISN) .099** .087** .046** .120** Size of the school .067** .065** .046** .060** Size of the municipality .017 .057** .013 -.040** Note. r** ¼p<0.001 *¼p<0.01. T. Saloviita and E. Pakarinen Teaching and Teacher Education 97 (2021) 103221 6 Table 3 Teacher-level Variables in Relation to Burnout (Stepwise regression, N ¼4156). Variable Burnout total Exhaustion Lack of accomplishment Depersonalisation b B(SE) D R 2 b B(SE) D R 2 b B(SE) D R 2 b B(SE) D R 2 Model 1 Intercept 33.995 (.487) 10.807 (.202) 8.868 (.164) Gender .055*** 1.381 (.387) .003*** ee .076*** .799 (.160) .006*** .105**** .894 (.130) .011*** Model 2 Intercept 35.507 (.865) 15.089 (.327) Gender .056*** 1.406 (.387) ee Age -.033*-.034 (.016) .001*-.052** -.024 (.007) .003** ee ee Model 3 Intercept 54.123 (1.159) 20.879 (.486) 18.530 (.421) 14.831(.343) Gender .047** 1.169 (.366) ee .068*** .717 (.153) .097*** .826 (.125) Age .002 .002 (.015) -.028þ-.013 (.007) ee ee Teacher efficacy -.333*** 1.745 (.077) .109*** -.236*** -.549 (.035) .055*** -.301*** -.666 (.032) .091*** -.287*** -.514 (.026) .082*** Model 4 Intercept 58.981 (1.171) 22.348(.495) 20.006(.431) 16.167(.349) Gender .026*.649 (.357) ee .052*** .542 (.151) .079*** .669(.123) Age -.010 -.010 (.015) -.038*-.017 (.007) ee ee Teacher efficacy -.297*** 1.557 (.076) -.208*** -.484 (.035) -.274*** -.604 (.302) -.256*** -.459 (.026) Attitudes towards inclusion -.223*** -.774 (.050) .048*** -.174*** -.267 (.023) .029*** -.180** -.262 (.021) .031*** -.199*** -.236 (.017) .038*** Model 5 Intercept 65.822 (1.3779) 23.930 (.589) 23.011 (.525) 17.981 (.428) Gender .012 .307 (.356) ee .037*.387 (.150) .068*** .578 (.122) Age -.015 -.015 (0.15) -.041** -.019 (.007) ee ee Teacher efficacy -.235*** 1.233 (.083) -.174*** -.404 (.038) -.208*** -.459 (.035) -.207*** -.372(.029) Attitudes towards inclusion -.216*** -.752 (.050) -.170*** -.261 (.023) -.173*** -.252 (.021) -.193*** -.229 (.017) Relatedness with Students -.144*** -.582 (.064) .017*** -.080*** -.144 (.029) .005*** -.155*** -.263 (.027) .019*** -.115*** -.159 (.022) .011*** Model 6 Intercept 66.698(1.329) 24.848(568) 23.408(.512) Gender .041** 1.019 (.345) ee .061*** .638 (.147) Age -.010 -.010 (.014) -.035*-.016 (.007) ee Teacher efficacy -.195*** 1.023 (.081) -.130*** -.302 (.037) -.173*** -.381 (.035) Attitudes towards inclusion -.166*** -.576 (.049) -.119*** -.183 (.022) -.130*** -.190 (.021) Relatedness with Students -.145*** -.583 (.061) -.084*** -.150 (.028) -.155*** -.264 (.026) Trust in getting support -.254*** 2.120 (.118) .059*** -.272*** 1.002 (.054) .069*** -.218*** -.765 (.050) .044*** Model 7 Intercept 67.026(1.325) 24.936(.567) 23.493(.513) 18.324(424) Gender .037** .935 (.344) ee .059*** .620 (.147) .075*** .643(.121) Age -.008 -.008 (.014) -.034*-.015 (.006) ee ee (continued on next page) T. Saloviita and E. Pakarinen Teaching and Teacher Education 97 (2021) 103221 7 Table 3 (continued ) Variable Burnout total Exhaustion Lack of accomplishment Depersonalisation b B(SE) D R 2 b B(SE) D R 2 b B(SE) D R 2 b B(SE) D R 2 Teacher efficacy -.189*** -.993 (.081) -.125*** -.290 (.037) -.170*** -.375 (.035) -.182*** -.326(.029) Attitudes towards inclusion -.163*** -.565 (.049) -.116*** -.178 (.022) -.129*** -.188 (.021) -.168*** -.199(.017) Relatedness with Students -.144*** -.582 (.061) -.083*** -.149 (.028) -.155*** -.263 (.026) -.115*** -.158(.022) Trust in getting support -.171*** 1.425 (.169) -.194*** -.714 (.077) -.177*** -.618 (.073) ee Availab. of support -.115*** -.953 (.166) .006*** -.108*** -.395 (.077) .005*** -.058** -.200 (.071) .002** -.154*** -.433 (.041) .022*** Model 8 Intercept 62.227(1.387) 24.117 (.610) 21.778(.548) 15.945(.446) Gender .029*.723 (.340) ee .052*** .547 (.146) .064*** .541(.118) Age -.012 -.012 (.014) -.035*-.016 (.006) ee ee Teacher efficacy -.189*** -.990 (.080) -.125*** -.290 (.037) -.169*** .547 (.146) -.181*** -.325(.028) Attitudes towards inclusion -.125*** -.435 (.050) -.102*** -.157 (.023) -.098*** -.142 (.021) -.115*** -.136(.017) Relatedness with Students -.138*** -.555 (.060) -.081*** -.144 (.028) -.150*** -.255 (.026) -.106*** -.146 (.021) Trust in getting support -.173*** 1.445 (.167) -.194*** -.717 (.077) -.178*** -.622 (.072) ee Availab. of support -.106*** -.873 (.164) -.105*** -.383 (.076) -.051*-.176 (.071) -.143*** -.402(.040) Ability-based groups should be returned .145*** 1.218 (.117) .019*** .053*** .196 (.054) .003*** .121*** .424 (.050) .013*** .205*** .586(.041) .038*** Total R 2 .261 .168 .206 .201 Note: þp<0.10; *p<0.05; **p<0.01; ***p<0.001. b ¼Standardised estimate, B ¼Unstandardised estimate. T. Saloviita and E. Pakarinen Teaching and Teacher Education 97 (2021) 103221 8