Self-efficacy and health in Swedish teachers:
Validating the Norwegian Teacher Self-Efficacy Scale in a Swedish context
Josefin Brickman and Amanda Olsson
Department of Law, Psychology & Social Work (JPS), Örebro University PS3111: Examensuppsats HT20
Supervisor: Mats Liljegren. January 5, 2021
This study aimed to translate the Norwegian teacher self-efficacy scale (NTSES; Skaalvik & Skaalvik, 2007) and explore its validity and factor structure in a sample of 256 Swedish teachers. The ties between teacher self-efficacy and teacher burnout and self-efficacy and teacher well-being were also investigated. The results showed that the Swedish version of the NTSES had good internal consistency and adequate concurrent and convergent validity. However, results from a confirmatory factor analysis and two exploratory factor analyses did not support a factor structure equivalent to the original NTSES. The Swedish version of the NTSES might need some adjustments in translation and even consideration regarding removal of some items before it can truly be of use in a Swedish context.
Den aktuella studien syftade till att översätta Norwegian teacher self-efficacy scale (NTSES; Skaalvik & Skaalvik, 2007) och utforska dess validitet och faktorstruktur i ett urval bestående av 256 svenska pedagoger. Relationerna mellan teacher-self-efficacy (pedagogers upplevda självförmåga) och utbrändhet samt teacher self-efficacy och välmående undersöktes också. Resultaten visade att NTSES hade god intern konsistens samt adekvat samstämmig och konvergent validitet. Resultat från en konfirmatorisk och två explorativa faktoranalyser stöttade dock inte en faktorstruktur likvärdig den i orginalskalan. Den svenska versionen av NTSES kan behöva några justeringar i översättning samt eventuell övervägning att ta bort några frågor, innan den verkligen kan komma till användning i en svensk kontext.
Table of contents
Teacher self-efficacy in swedish teachers 5
Self-efficacy - theory and measurement 6
Teacher burnout 10
Aim and hypothesis 12
Norwegian Teacher Self- Efficacy Scale 15
Bandura's Teacher Self-Efficacy Scale 16
Copenhagen Burnout Inventory 18
Statistical analyses 18
Ethical considerations 21
Convergent and Concurrent validity 23
Results from CFA and EFA 24
Methodological considerations 30
Implications and future research 33
Teacher self-efficacy in Swedish teachers
The teaching profession is recognized as one of the most important professions in the world (UNESCO, 2009). It can be an enjoyable and satisfying occupation, as well as very demanding and stressful. During 2019 and 2020 there were approximately 210 200 teachers working in Sweden (Swedish national agency for education [SNAE], 2020b) and
approximately 45% of these teachers experienced work-related stress (SNAE, 2020a), 40% reported symptoms of depression and over 60% experienced sleep related problems (Schad & Johnsson, 2019). According to Swedish social insurance agency (SSIA, 2018) teaching is one of the professions with the highest risk of sick leave, especially for women. About 40 000 educated teachers are not working with teaching (Statistics Sweden, 2016) and, as reported by Lindqvist and Nordanger (2016), numerous new teachers leave the profession within the first few years, or sidestep working as a teacher altogether. This is a significant problem considering that Sweden has a shortage of teachers (SNAE, 2019).
An important concept related to positive teacher outcomes, regarding both health and performance, is teacher self-efficacy. Self-efficacy is derived from Social Cognitive Theory (Bandura, 1977, as cited in Zee & Koomen, 2016) and is defined as: “beliefs in one’s capabilities to organize and execute the courses of action required to produce given attainments” (Bandura, 1997, p. 3). According to Bandura, one’s sense of self-efficacy fluctuates across different aspects of life, and teacher self-efficacy is, consequently, a measure of its own. Building on social cognitive theory, Skaalvik and Skaalvik (2010) defined teacher self-efficacy as individual teachers’ belief in their own capability to organize and perform different tasks needed within the profession.
The ties between the teaching profession and self-efficacy has been widely investigated, and findings show that teacher self-efficacy is related to higher teacher motivation, better organizational skills and greater ability to handle difficulties in the
classroom (Tschannen-Moran & Woolfolk Hoy, 2001). Moreover, higher levels of teacher self-efficacy have been found to increase self-efficacy in students (Aloe et al., 2014),
correlate positively with student motivation (Tschannen-Moran & Woolfolk Hoy, 2001) and predict higher levels of student achievement (Caprara et al., 2006). A meta-analysis by Zee and Koomen (2016) found that teacher self-efficacy also increases classroom quality. Additionally, a lower sense of self-efficacy has been found in teachers who quit or leave the teaching profession, compared to teachers who stay (Hong, 2012), which is harmonious with findings that self-efficacy is related to increased levels of job satisfaction (Caprara et al., 2006; Zee & Koomen, 2016).
Self-efficacy - theory and measurement
The concept of teacher self-efficacy has, historically, been approached from two different viewpoints; Rotter’s (1966) Locus of control theory and Bandura’s (1977) Social cognitive theory (as cited in Zee & Koomen, 2016). Rotter (1966) explained that locus of control is influenced by an individual’s interaction with the environment. Through this interaction, people form general beliefs regarding to what extent they can control aspects of their life (internal control) and to what extent they cannot (external control) (Rotter, 1966). Rotter claimed that a high sense of internal control caused a high sense of teacher self-efficacy, because these teachers believed student behaviour and outcome could be shaped by education (Skaalvik & Skaalvik, 2007). Bandura, on the other hand, argued that believing education could change a student did not equal a belief in one’s own capabilities to educate well, hence pointing out the difference between self-efficacy and outcome expectancies (Zee & Koomen, 2016). Today Bandura’s self-efficacy theory is the most recognized theory in teacher self-efficacy research.
When explaining perceived self-efficacy Bandura (1997), emphasizes the importance of human agency - the desire and ability to influence our actions - and the capacity to reflect
upon and contribute to what happens to us. Self-efficacy, unlike for example self-concept, is not about a general perception of abilities or skill level, but rather a belief about what one can accomplish with the skills and abilities one has in any given situation (Bong & Skaalvik, 2003). According to Bandura (1997) this belief in one's capability fuels people's motivations, feelings and behaviours to a greater extent than any objective truth. He maintained that self-efficacy is reciprocally influenced by contextual, cognitive as well as social conditions, and can impact an individual's choice of task, including effort and persistence put into the task. For example, if people have doubts about their power to produce a desirable result, they will be less likely to act. With their sense of self-efficacy people set goals, intentions and estimate likely outcomes of their behaviours. Bandura (2012) further explains that self-efficacy is not a general trait, but a belief system that varies across different situations and areas in life. Even within a certain life area, an individual can experience diverse levels of self-efficacy
depending on, for example, setting or type of task.
Because of its multidimensional nature, the concept of self-efficacy cannot be captured by a universal, all-purpose measure (Bandura, 2012). Since self-efficacy can vary across domains and within domains, to measure self-efficacy adequately, a thorough conceptualization and a sound understanding of the particular domain is needed.
Additionally, Bandura recommends that items need to be phrased with words like “can do” or “able to” in order to measure capability beliefs, and should include an obstacle to challenge the sense of self-efficacy (Bandura, 2006). If the items describe scenarios that are easily performed, everyone will perceive themselves capable of doing it and, hence, everyone will score highly on self-efficacy (Bandura, 1997). This would result in a scale with low
Various researchers have tried to capture the concept of teacher self-efficacy in various ways (Skaalvik & Skaalvik 2007). Tschannen-Moran and Woolfolk Hoy (2001)
reported that many teacher self-efficacy scales show instability in factor structure and that commonly, items were either too specific for practical usage or not specific enough. In turn Tschannen-Moran and Woolfolk Hoy’s own scale (2001), Teacher’s Sense of Efficacy Scale (TSES), was criticized by Skaalvik and Skaalvik (2007) for lacking sufficient number of obstacles to challenge sense of self-efficacy. A literature review by Klassen et al. (2011) shows that the research field in general is associated with diverging conceptualizations of self-efficacy and survey questions not falling within Bandura's recommendations for item construction.
To be a teacher is also to be tied to a cultural context, and following Bandura's (2006) advice to create items specifically tailored to the area of self-efficacy intended to be
measured, it would be wise to use a scale adapted to the cultural conditions in which the teachers operate. There is no validated teacher self-efficacy scale available for measurement in the Swedish context. The Norwegian Teacher Self-Efficacy Scale (NTSES; Skaalvik & Skaalvik, 2007) is a teacher self-efficacy scale developed in Norway, a country with a similar cultural context and similar demands on teachers. The scale is multidimensional, containing six subscales, and the items are constructed using Bandura’s recommendations mentioned above. The subscales were designed by analysing the Norwegian national curriculum and through interviews with teachers working in Norway.
The interest around well-being, in general, is growing (VanderWeele et al., 2020), with voices insisting that it should be a fundamental aspect of the healthcare system (Plough, 2015). Well-being in teachers has shown to decrease levels of burnout and depression in teachers (Capone & Petrillo, 2020) as well as reduce psychological distress in students (Harding et al., 2019). Moreover, a study on special education teachers in China showed that
teachers with higher levels of subjective well-being had a lower intention to leave the profession (Fu et al., 2020).
Since well-being can be experienced in various ways and represent different things to different cultures, groups and individuals, there is no official and universal definition of well-being (World Health Organisation, n.d.). However, Ed Diener has been researching the concept of Subjective well-being (SWB) for several decades (see Diener, 1984; 2000; Diener et. al., 2006; Tov & Diener, 2013) and explains it as: “people's cognitive and affective evaluations of their lives” (Diener, 2000, p. 34). Some scientists carefully suggest that a synonym for SWB might be happiness (Tov & Diener, 2013), however, although positive emotions are an important factor in life satisfaction (Diener et. al., 2006), most researchers agree that a complete absence of negative emotions or life experiences does not necessarily equal a higher SWB (Holt et al., 2012).
According to Diener (1984) SWB consists of three factors; life satisfaction (LS), positive affect (PA) and negative affect (NA), and tend to be judged by individuals as a general feeling, based on their life situation as a whole. It is still unclear exactly how these factors interact with each other. Busseri and Sarava (2010) discuss several ways LS, PA and NA may be related, one example being that PA and NA are positively and negatively correlated to LS, respectively. This would correspond with the fact that WHO considers higher SBW equal to mental health (Jahoda, 1958, as cited in Topp et al., 2015).
When measuring well-being it is recommended to use a multiple item scale in order to capture different aspects of what can be considered feeling well (Ryff et al., 2020).
Research shows that in collective cultures, such as Japan, well-being is more dependent on collective aspects, while more independently oriented cultures, such as Sweden, well-being is, to a higher degree, personal (Yoo et al., 2016). Therefore, measures of well-being in independent cultures should include questions about the individual (Ryff et al., 2020).
Previous research shows a moderate correlation between teacher self-efficacy and teacher well-being (Harding et al., 2019; Huang et al., 2019; Zee & Koomen, 2016).
Another important concept related to teacher self-efficacy is burnout. In the teaching profession, burnout is more prevalent than in other professions (García-Carmona et al., 2019). Arvidsson et al. (2016) investigated burnout in Swedish teachers and found that 15% had high levels of burnout. Teachers' emotional exhaustion is significantly associated with decrease of job satisfaction and an increased motivation to leave the occupation (Skaalvik & Skaalvik, 2011). Studies done in both Norway and Italy have shown a significant negative association between teacher burnout and teacher self-efficacy (Avanzi et al., 2013; Skaalvik & Skaalvik, 2010; 2014). In Sweden, Arvidsson et al. (2016) found that increased burnout was linked to a lower sense of self-efficacy. According to a meta-analysis by Shoji et al. (2016), a moderate negative correlation was found for self-efficacy and burnout. García-Ros and colleagues (2015), using regression analysis, found that self-efficacy was predictive of burnout. However, a recent study conducted by Kim and Burić (2020) using a cross-lagged panel design found that burnout better predicts future teacher self-efficacy than vice versa. Maslach (2001) defines burnout as a syndrome with three major components; exhaustion, cynicism and inefficacy, all which are caused by occupational stress. Her definition has gained huge recognition and is, for example, used in International
Classification of Diseases (ICD) 11th revision (World Health Organisation, 2018). Maslach (2001) argues that there are five common elements for the phenomenon; (1) the predominant symptoms exist of emotional exhaustion and dysmorphic symptoms, (2) the symptoms are mental and not physical exhaustion, (3) the symptoms are associated with work, (4) the symptoms occur in people who do not experience other mental health issues and (5) because
of negative attitudes and behaviours there is a decrease in the level of efficacy at work. This conceptualization is the basis for the Maslach Burnout Inventory (MBI).
However, Maslach's definition and measure have several flaws according to Kristenssen and colleagues (2005). First, Maslach argues that burnout is mainly found in client related work, and the original MBI only measures burnout in such occupations. To both strengthen or falsify this hypothesis a measure including other professions should be used. Recent research shows that burnout is not only caused by paid work and client related work but can be caused by studiying, role-conflict, and volunteering (Bianchi et al., 2018; Kremer, 2016; Kristensen et al., 2005). Kristenssen et al. (2005), secondly argue that MBI only measures one symptom (exhaustion), one coping strategy for the symptom (cynicism) and a consequence of the symptom (inefficacy at work). Factor analysis have shown that these three components are in fact three distinct and different aspects, and not subfactors included in one syndrome (Kristensen et al., 2005). Other distinguished researchers in the field have defined burnout to not only be psychological in character, but argue that it also incorporates physical and emotional exhaustion (Pines & Aronson, 1988; Shirom, 1989), caused by being exposed to an emotionally demanding environment for a long duration of time (Pines & Aronson, 1988). Copenhagen Burnout Inventory (CBI) builds upon these definitions rather than Maslach's definition. The CBI measures fatigue and exhaustion within three different dimensions of a person's life; personal burnout, work related burnout and client related burnout (Kristensen et al., 2005), and in all three dimensions, physical exhaustion is included as a symptom. The CBI captures the variation of how individuals experience burnout, their belief about the cause of the burnout and it can be used in all professions. The structure of the CBI also enables people who are not currently working to be included in studies of burnout.
Aim and hypothesis
Despite the amount of research and literature on teacher self-efficacy, the concept has scarcely been investigated in Sweden and no teacher-self efficacy scales have been translated into Swedish. Hence, the main purpose of this study was to validate a Swedish version of the Norwegian Teacher Self-Efficacy Scale (NTSES) in a Swedish context. The study also sought to investigate if demographic variables were related to self-efficacy scores. Further, the hypotheses were that the swedish NTSES would have a positive correlation with subjective well-being, a negative correlation with burnout and a positive correlation with another teacher self-efficacy scale.
The participants in this study were 258 teachers between the ages 20-74 years (M = 43.84, SD = 11.56). 75.8% were women, 22.3% were men and 2.0% did not wish to report gender. 92.6% of the participants had a degree in teaching and 88.0% were authorized teachers according to the Swedish National Agency for Education. Teaching experience ranged from 0-45 years (M = 15.13, SD = 10.92). Almost all participants (96.0%) were working at the time of the survey, 0.4% were on parental leave, 1.6% were on sick leave and 2.0% were absent for other reasons. The participants worked in pre-school (9.8%), lower school (28.9%), middle school (30.1%), upper school (29.3%) and high school (25.4%), with 19.0% working in two or more age groups. The size of the schools that participants worked at varied between 100 students or less (5.9%), 100-300 students (33.9%), 300-500 students (29.1%), 500-700 students (16.9%), 700-1000 students (8.3%) and above 1000 students (5.9%). 78.9% of the sample worked in public schools, 19.9% in private schools, 0.4% in special education and 0.8% in an unspecified type of school. Most participants worked full
time (83.6%), with 94.0% having permanent employment. 5.5% were working for more than one employer.
A power analysis with an anticipated effect size of 0.3, a desired statistical power of 0.8, one latent variable, six observed variables and a probability level of 0.05 yielded a minimum sample size of 200 participants.
A small number of missing values was detected in the sample. One participant had a great extent of unanswered questions throughout the survey, and yet another participant did not report any demographics at all. These two participants were left out of the data analysis by means of listwise deletion (Berntson et al., 2016). Through descriptive analysis, additional missing values were found on single items for 31 of the participants. Further investigation revealed the missing values to be non-systematic and mean imputation (MI) was used to replace the missing values. The final number of participants was 256, which exceeded the minimum amount according to the power analysis.
Analysis of recruitment characteristics
To control for systematic bias in the sample, a one-way ANOVA:s were performed to compare teacher self-efficacy level on the different ways of recruitment. The results showed that there was a significant overall difference between the four different groups of
recruitment of teachers on the average level of teacher self-efficacy, F(3, 252) = 4.52, p < .01. Specifically, teachers recruited through their principals (M=131.34, SD=16.24) had a higher mean average than did teachers recruited through a friend (M=118.43, SD=16.31).
To recruit teachers, emails were sent to headmasters of schools in 25 different municipalities within Sweden, containing information about the study and a request to forward the information to the teachers at their school/s. Participants were also recruited by posting information about the study along with a link to the survey on Facebook, and through
acquaintances spreading the word. Individual emails were sent to 750 principals and 69 of them answered that they would spread the survey to the teachers at their school. 47
headmasters replied that they would not spread the survey due to a high workload or similar reasons, the other 634 did not respond. 85.2 % of the participants were recruited via their principal.
The survey was available online for four weeks during October and November 2020. The participants answered four different inventories, two that were translated from English to Swedish for the sake of this study and two measures which already had Swedish versions. At the beginning of the survey participants completed a demographic form, thereafter the two teacher self-efficacy scales, the well-being scale and the burnout scale. No personal
information was recorded, and all responders were anonymous. Participants were allowed to go back and forth in the questionnaire, if they felt the need to change responses. There was no time limit and none of the items were compulsory.
Teachers working with children in preschool, grades 1-9 and high school were included in our sample. Special education teachers, teachers working with children with intellectual disabilities, and teachers working in specialized schools (e.g. school for deaf children) were included. Teachers working with adults were not included. The survey clearly stated that only teachers currently working with teaching were of interest for this study, thereby excluding retired teachers and educated teachers working in other professions. Teachers momentarily absent from the profession, such as teachers on sick leave, parental leave or otherwise unable to work at the present moment were included. Both authorized and non-authorized teachers were invited to participate.
Norwegian Teacher Self- Efficacy Scale
The Norwegian teacher self-efficacy scale (NTSES; Skaalvik & Skaalvik, 2007) is a scale used for measuring sense of self-efficacy across 6 domains within the teaching
profession. The domains measured are (1) Instruction, (2) Adapt instruction to individual needs, (3) Motivate students, (4) Maintain discipline, (5) Cooperate with colleagues and parents and (6) Cope with change. The scale has a total of 24 items, with an overall
instruction “How certain are you that you can:” followed by statements. Each item was rated on a 7-point Likert scale: (1) not certain at all, (3) quite uncertain, (5) quite certain and (7) absolutely certain. Sample items consist of: “Explain subject matters so that most students understand the basic principles” (Instruction); “Provide realistic challenge for all students even in mixed ability classes?” (Adapt instruction to individual needs); “Motivate students who show low interest in schoolwork” (Motivate students); “Get all students to behave politely and respect the teachers” (Maintaining discipline); “Cooperate well with most parents” (Cooperate with colleagues and parents) and “Manage instruction even when the curriculum is changed” (Cope with change). The internal reliability (Cronbach’s Alpha) for the subscales were: .83, .90, .83, .91, .77 and .88, respectively (Skaalvik & Skaalvik, 2010).
Translation of NTSES was done using the back-translation model proposed by Brislin (1970). The scale was translated into Swedish by both authors, independently. Next, the two versions were compared and merged into a single version. The translation was then translated back into English by another independent psychology student, making it possible to compare the original English version of the scale and the back-translated version of the scale in search of ambiguities. The Swedish translation was modified and revised into its final version. Some items were culturally adapted to better fit the Swedish context. For example, “discipline” is
not a word often used within Swedish schools and neither is “aggressive”. Hence, these phrases were translated to “keep order” and “acting out” respectively.
To ensure face and content validity, four teachers were asked to look through the Swedish version of NTSES before the survey was distributed to the participants. Their expert knowledge about the teaching profession gave us additional feedback regarding translation and cultural adaptation, hence, some items were rephrased. The full translation of the scale can be found in appendix A.
Bandura's Teacher Self-Efficacy Scale
The Teacher Self-efficacy scale (Bandura, 2006) is an inventory that assesses teachers' perception of their ability to succeed in specific tasks as teachers. The different domains measured are (1) efficacy to influence decision making, (2) instructional
self-efficacy, (3) disciplinary self-self-efficacy, (4) efficacy to enlist parental involvement, (5) efficacy to enlist community involvement and (6) efficacy to create a positive school climate.
Bandura’s teacher self-efficacy scale (n.d.; 2006) was produced for a North American context. In the present study, Bandura’s teacher self-efficacy scale was used for convergent validation.
The original 30-item version (Bandura, n.d.) has been validated by Hoy and Spero (2005). The items of the original scale were rated on a 9-point Likert scale with 1
representing “nothing” and 9 representing “a great deal”. Cronbach’s alpha was 0.92 (Hoy & Spero, 2005). Bandura presented an updated version of the scale (Pajares & Urdan, 2006), where the items “How much can you do to get churches involved in working with the school” and “How much can you do to influence the class sizes in your school” had been removed. The latter version also had a different response format, with an 11-point Likert scale where the value 0 represented “cannot do at all” and the value 100 represented “highly certain can do”. The latter 28 item version was used in the present study, the responses were summarized
and then divided by the number of items in order to standardize the results on a scale from 0 to 100.
Using Brislin’s method (1970), both authors independently translated the scale into Swedish, and the two versions were then reconciled into one. Back-translation was done from Swedish to English by an independent translator according to the ITC- guidelines (Gregoire, 2018). The original English version and the back-translated version were compared, showing no major changes in meaning, though minor grammatical adjustments were done to the final Swedish version. For example, the word “parents” was replaced with “caregivers” since this was judged to be more culturally appropriate in the Swedish context.
To adequate the scale for the Swedish teacher context, four teachers were recruited as experts. The experts provided feedback on language and cultural adaptation to ensure the face validity of the scale.
Subjective well-being was measured using WHO-5, a 5-item scale developed by World Health Organisation (Topp et al., 2015). The items are phrased positively in order to measure well-being as opposed to distress (Bech et al., 2003); (1) “I have felt cheerful and in good spirits”; (2) I have felt calm and relaxed”; (3) “I have felt active and vigorous”; (4) ”I woke up feeling fresh and rested” and (5) “My daily life has been filled with things that interest me”. Responses were measured on a 6-point Likert scale ranging from 0 (none of the time) to 5 (all of the time). The responses were multiplied by four in order to standardize the results on a scale from 0 to 100 (Topp et al., 2015).
WHO-5 has adequate validity, has been used across many different fields of study and additionally serves as a valid screening tool for depression (Topp et al., 2015). The measure has a good reliability with a Cronbach’s alpha of 0.93 (Sisask et al., 2008).
Copenhagen Burnout Inventory
The Copenhagen Burnout Inventory (CBI; Kristensen et al., 2005) is a 19-item inventory that measures 3 different aspects of burnout; personal burnout (PB), work related burnout (WRB) and client related burnout (CRB). Sample items included: “How often do you feel worn out” (PB), “Do you feel that every working hour is tiring for you?” (WRB) and “Does it drain your energy to work with clients” (CRB).
Items were answered using a 5-point Likert scale, including two different versions; “always” to “never/almost never” (12 items) and “a very high degree” to “a very low degree” (7 items). Each answer generates a value between 0-100 (0, 25, 50, 75 or 100), which are then summarized and divided by the number of items, to generate a standardised score, as
recommended by Kristensen and colleagues (2005). The value 0 implies the lowest degree of burnout and the value 100 the highest degree of burnout. The cut-off for burnout is a score of 50 (Kristensen et al., 2005). In the present study the Swedish version of the CBI was used. Validation for the Swedish context has been done in three different studies (Arneson, 2006; Arneson & Liljegren, 2005; Södergren, 2005). To better fit the work context of teachers, the word “client” was changed to “student” in the CRB domain, as has been done in previous research (Campos et al., 2013). Therefore, the subfactor was changed from client related burnout (CRB) to student related burnout (SRB). The order of the items was scrambled to prevent stereotyped response patterns, as suggested by the authors. Cronbach's alpha for the Swedish version of the CBI was 0.92 (Arneson, 2006). In this study, only the scale as a whole was used for validation purposes.
Firstly, descriptive analyses were run (means, standard deviations, range, skewness and kurtosis) on all variables and missing values were handled. The Little’s missing
data set were missing non-systematically. Looking at only missing answers the missing values amounted to 0.2% of the complete data set, with the method listwise deletion 12% of the participants would have been removed. Since such a small amount of the data was lost, the missing values were replaced via mean imputation (MI) (Berntson, 2016).
Kolmogorov-Smirnov tests were conducted on NTSES, Bandura's Teacher Self-efficacy Scale, CBI and WHO-5, to control for normality. If the test was significant with a p <.05, and therefore not normally distributed, both Spearman and Pearson’s r were used during correlation analyses.
Thereafter, a series of one-way analysis of variance (ANOVA) were conducted to control for systematic differences in self-efficacy (NTSES-results) between demographic variables in our sample. Correlation analyses were also run to investigate the relationship between teacher self-efficacy and the continuous demographic variables.
To investigate convergent validity, a correlation analysis including the NTSES and the Bandura's Teacher Self-efficacy scale was conducted. Convergent validity can be defined as “the extent to which two measures capture a common construct” (Carlsson & Herdman, 2012 s. 18) and is used to ensure that a scale is measuring what it is intended to measure. According to Carlsson and Herdman (2012) a value of r > .70 is recommended to imply good convergent validity and values below r < . 50 are not recommended.
Since Skaalvik and Skaalvik (2007) used concurrent validity to validate the original NTSES, a similar procedure will be used to validate the Swedish version. Concurrent validity refers to “The correlation of a measure with performance on another measure or criterion at the same point in time” (Kazdin, 2010 s. 359). In this case, correlation analyses including teacher self-efficacy and two other constructs, shown to be related in previous research, was run. Hence, it was expected that the Swedish NTSES would have a positive moderate
2019; Huang et al., 2019; Zee & Koomen, 2016), but have a negative moderate correlation with teacher burnout, at about -.33, as seen in the meta-analysis by Shoji (2016).
Cronbach's alpha was calculated for both the Bandura Teacher Self-efficacy Scale and the NTSES as a whole. Cronbach's alpha was also calculated for all the subscales of the NTSES, to explore the internal validity of the measure. Originally, the NTSES includes six subscales (Skaalvik & Skaalvik, 2007), and the aim of this study was to test if this model structure was consistent in the Swedish version.
To investigate the factorial structure of the scale, a confirmatory factor analysis (CFA) was run. This was decided a-priori since CFA is the proper test to use when
investigating a scale with a beforehand hypothesized factor structure (Byrne, 2001). Three different models were analysed using CFA in Amos 26 Graphics, and in all analyses error terms were set free. Model 1 consisted of all items of the NTSES and six primary factors, based on the original factor structure. Model 2 specified six primary factors and one second order factor, also consistent with Skaalvik and Skaalvik (2007). Model 3 specified only one primary factor, to control for unidimensionality of the scale. To ensure the model fit, the chi-square test statistic was used, along with the comparative fit index (CFI), the incremental fit index (IFI), the Tucker–Lewis index (TLI), and the root-mean-square error of approximation (RMSEA) (Hu & Bentler, 1999; Byrne, 2001; Schweizer, 2010). The chi-square test
compares the expected and observed variance matrices and reveals how big the difference is, a low chi-square value indicates a small difference, hence, a good fit of the model.
Acceptable fit of the model is also confirmed if values on the CFI, IFI and TLI are > 0.9 and considered good fit if values are > 0.95 (Hu & Bentler, 1999; Byrne, 2001; Schweizer, 2010; Berntson et al., 2016). The RMSEA controls for the residuals in the model, a value < 0.05 on the RMSEA indicates a good fit of the model (Berntson et al., 2016).
The results of the CFA revealed the need to run additional tests, and two exploratory factor analysis (EFA) were conducted to further investigate the factor structure of the Swedish NTSES. The EFA:s were done using Principal Axis Factoring in SPSS. The first EFA was conducted with the instruction to choose factors with eigenvalues above 1, the second EFA was conducted using six fixed factors. For both analyses promax rotation was chosen since theory suggested the factors would correlate (Field, 2013).
Self-efficacy is adequately measured when the items contain an obstacle or difficulty (Bandura, 2006). This sort of wording may have induced a sense of discomfort or generated feelings of inadequacy in the participants. Since some of the participants were recruited through their principal, the questionnaire may also have incited fear of critique from superiors and worries regarding the usage of submitted answers. However, participants were provided full information on the study and had to give their consent to participate before being able to enter the survey. Additionally, participants were informed that the survey was anonymous and that only the authors had access to the submitted answers. The benefits of researching teacher self-efficacy in a Swedish context were judged to be of higher importance than any discomfort individual participants may have experienced on the behalf of the survey.
Boxplots produced in SPSS were used to detect outliers in the Teacher self-efficacy scores (NTSES scores), no outliers were found. To test for normal distribution the
Kolmogorov-Smirnov test in SPSS was run for the four different inventories. The results showed that the CBI was not normally distributed, D(256) = 0.08, p < .01, neither was the WHO-5, D(256) = 0.14, p < .01. The NTSES and the Bandura teacher self-efficacy scale had normal distribution.
Table 1 shows the mean, standard deviation, range and Cronbach’s alphas for the NTSES, the Swedish version of the Bandura Teacher Self-Efficacy Scale, the CBI and the WHO-5 and their subscales. Scale. Five of the NTSES subscales showed cronbach's alpha scores above the cut-off value of .70 for acceptable reliability, four of which reached good reliability at .80 (Field, 2013). The subscale Cooperate with colleagues and parents had a
Psychometric properties of the Swedish version of the NTSES, Swedish version of the Bandura Teacher Self-Efficacy Scale, CBI and WHO-5.
Scale M SD Range Cronbach's α
NTSES Total scale 129.91 16.72 83-168 .94
Instruction 23.09 2.97 16-28 .80
Adapt instruction to individual needs 21.84 3.79 4-28 .87
Motivate students 20.50 3.37 12-28 .83
Maintain discipline 20.13 4.22 4-28 .85
Cooperate with colleagues and Parents 21.26 3.06 13-28 .62
Cope with change 21.91 3.38 10-28 .75
Bandura’s teacher self-efficacy scale 69.54 14.29 24-100 .93 Efficacy to Influence Decision Making 70.18 18.30 7-100 .66
Instructional self-efficacy 69.98 15.24 24-100 .88
Disciplinary self-efficacy 76.45 17.86 10-100 .85
Efficacy to enlist parental involvement 71.63 18.90 0-100 .77 Efficacy to enlist community involvement 49.18 28.67 0-100 .89 Efficacy to create a positive school climate 73.13 16.25 13-100 .86
CBI total 35.19 17.86 4-86 .94 Personal burnout 39.91 21.35 0-96 .91 Work-related burnout 35.53 16.28 11-82 .75 Student-related burnout 30.07 21.91 0-92 .91 WHO-5 63.58 20.87 12-100 .89 Note: n = 256
Cronbach’s alpha below .70, which is considered questionable and may indicate unreliability (Field, 2013). Cronbach’s alpha for the whole scale was above .90, suggesting high reliability in the measure. The remaining scales and subscales revealed adequate reliability, with the exception of the Efficacy to influence decision making subscale from Bandura Teacher Self-Efficacy
A series of one-way ANOVA:s were performed to compare teacher self-efficacy level on the different conditions in the descriptive variables. The results revealed no significant difference in teacher self-efficacy between any of the descriptive variables.
The relationship between NTSES scores and the continuous descriptive variables (age, teaching experience and working percentage) were explored by means of Pearson's correlation analysis. The results showed that there was a significant positive correlation between teacher self-efficacy and age, r(254) = .282, p <.001, and a significant positive correlation between teacher self-efficacy and teaching experience, r(254) = .246, p <.001. No significant correlation between teacher self-efficacy and working percentage was detected.
Convergent and Concurrent validity
The relationships between NTSES and Bandura’s teacher self-efficacy scale, teacher well-being and teacher burnout was investigated through Pearson's and Spearman's
correlation analysis. In case of insignificant difference between the two tests, only Pearson’s r will be presented. Table 2 shows mean, standard deviation and correlation with NTSES results for the Bandura’s teacher self-efficacy scale, WHO-5 and CBI. The results revealed a large, positive correlation with Bandura’s teacher self-efficacy scale (convergent validity), a moderate, positive correlation with WHO-5 and a moderate, negative correlation with CBI (concurrent validity).
Descriptive statistics and correlations for study variables
Variable n M SD NTSES
Bandura’s teacher self-efficacy scale 256 69.54 14.29 .69**
WHO-5 256 63,6 20,9 .34**
CBI 256 35,2 17,9 -.43**
Note: ** p < .01 level (2-tailed).
Results from CFA and EFA
A confirmatory factor analysis (CFA) was used to test the dimensions of the Swedish version of NTSES. The analysis was based on the theory of the original NTSES structure, with 6 dimensions. One of the assumptions for running a CFA is multivariate normal distribution in the data (Byrne, 2001). Following the example of Fiorilli et al. (2015) the subscales were explored through skewness and kurtosis values. As can be seen in table 3 the criterion of multivariate normal distribution was not met, since not all the hypothesized subscales were normally distributed. Most subscales fall within the cut-off value of 1 for skewness and kurtosis (Tabachnick & Fidell, 2013 as cited in Fiorilli et al., 2015), but Adapt instruction to individual needs does not. However, since most of the data indicated sufficient normal distribution, and the aim was to replicate the same analyses as Skaalvik and Skaalvik (2007), a CFA was still judged appropriate.
Exploratory descriptive statistics of the hypothesised subscales within NTSES
Subscale M SD Skewness Kurtosis
Instruction 23.09 3.67 -.08 -.64
Adapt instruction to individual needs 21.84 3.79 -1.11 3.87
Motivation 20.50 3.37 -.04 -.33
Maintain discipline 20.13 4.22 -.59 .54
Cooperate with colleagues and parents 21.26 3.06 -.11 -.40
Cope with change 21.91 3.38 -.24 -.14
Model 1 tested a six-factor solution, in Model 2 a six-factor solution with one second order factor was tested and finally, in Model 3, a unidimensional solution, with all 24 items loading on one single construct, was tested. As is shown in table 4, all models had a poor fit to the data, and the CFA provided no support for the hypothesis of a scale with a similar factor structure to the original NTSES.
Results from a Confirmatory Factor Analysis of the Swedish version of Norwegian Self-Efficacy Scale
Model X2 df TLI CFI IFI RMSEA
Model 1 610.99 237 0.86 0.88 0.88 0.08
Model 2 695.06 246 0.84 0.86 0.86 0.09
Model 3 1077.84 252 0.72 0.74 0.74 0.11
Note. Model 1 specified six primary factors. Model 2 specified six primary factors and one second order factor. Model 3 specified one primary factor. No error terms were set free.
To further test the dimensional structure of the Swedish version of the 24-item NTSES, two exploratory factor analysis (EFA), with maximum likelihood extraction and promax rotation, were performed. The first EFA was run with eigenvalues larger than 1 and the second EFA was run with six fixed factors, in accordance with the original NTSES structure. The Kaiser-Meyer-Olkin (KMO) test revealed a value above the minimum criterion of .5 (Field, 2013), KMO = .92, indicating that the sample size was adequate for the analysis. The first EFA revealed two factors with eigenvalues larger than Kaiser’s cut-off value of 1 (Field, 2013) and two factors with eigenvalues slightly below 1 (.96 and .68). The four factors, in combination, explained 52.48% of the overall variations in the measure.
According to Field (2013), if the sample size is larger than 250, Kaiser’s criterion of extracting factors with eigenvalues above 1 is only accurate if the average communality is above .6. Since the average communality for the present EFA was .52, the factor extraction
was additionally based on the scree plot, as well as eigenvalues. The scree plot supported retaining four factors, which also fit the theoretical basis for the original NTSES better than retaining only 2 factors.
Seven items loaded on factor 1 which explained 39.49% of the overall variation, six items loaded on factor 2 which explained 6.17% of the variation, seven items loaded on factor 3 which explained 4% of the variation and four items loaded on factor 4 which
Results from an Exploratory Factor Analysis of the Swedish version of Norwegian Teacher Self-Efficacy Scale (NTSES)
Item Factor loading
1 2 3 4
Instruction 1 .417 .104 .278 -.134
Instruction 2 .792 -.093 .053 .022
Adapt instruction to individual needs 1 .628 .118 -.071 .046 Adapt instruction to individual needs 2 .826 .097 -.116 -.002 Adapt instruction to individual needs 3 .857 -.032 -.105 .058 Adapt instruction to individual needs 4 .873 -.069 -.130 .136
Cope with change 2 .589 -.148 .293 .029
Motivate students 1 .121 .673 -.077 .033
Motivate students 2 .099 .770 -.130 -.022
Motivate students 3 .350 .481 .028 -.019
Motivate students 4 .027 .725 -.089 .078
Cooperate with colleagues and parents 1 -.193 .475 .105 .208 Cooperate with colleagues and parents 3 -.276 .516 .117 .379
Instruction 3 .221 .189 .487 -.109
Instruction 4 .303 .240 .382 -.144
Cope with change 1 .007 .133 .447 .129
Cope with change 3 .186 .081 .590 -.067
Cope with change 4 .123 -.254 .581 .196
Cooperate with colleagues and parents 2 -.033 -.196 .616 .142 Cooperate with colleagues and parents 4 -.236 .077 .643 .026
Maintain discipline 1 .232 .077 .014 .587
Maintain discipline 2 -.035 .133 .089 .650
Maintain discipline 3 .097 .034 .080 .733
Maintain discipline 4 .196 .141 .024 .449
Note. N = 256. Extraction Method: Principal Axis Factoring. Rotation Method: Promax with Kaiser Normalization. Factor loadings above .30 are in bold.
explained 2.81% of the variation. The loadings for factor 1 items ranged between .42 and .87, the loadings for factor 2 items varied between .46 and .77, the loadings for factor 3 items ranged between .38 and .64, and respectively the loadings for factor 4 items ranged between .45 and .73. The factor loadings, after rotation, can be seen in table 5.
Results from an Exploratory Factor analysis with 6 fixed factors of the Swedish version of Norwegian Teacher Self-Efficacy Scale (NTSES)
Item Factor loading
1 2 3 4 5 6 Instruction 1 .456 .100 .172 -.048 -.032 .283 Instruction 2 .834 -.104 .008 .089 -.102 .207 Instruction 4 .429 .201 .055 -.148 .251 .187 Adapt instruction 1 .695 -.082 .012 -.001 .070 -.046 Adapt instruction 2 .911 -.201 .013 -.019 .062 .045 Adapt instruction 3 .866 -.056 .041 .049 -.180 -.060 Adapt instruction 4 .934 -.077 -.109 .069 -.063 -.136
Cope with change 2 .666 .215 -.173 .011 .004 .063
Cope with change 1 .043 .532 .022 .064 .068 -.144
Cope with change 3 .307 .422 -.082 -.070 .216 .157
Cope with change 4 .095 .514 -.067 .274 -.191 .135
Cooperate 2 -.041 .594 -.087 .178 -.126 .051 Cooperate 4 -.280 .929 .009 -.092 -.027 -.225 Instruction 3 .269 .325 .184 -.040 .051 .238 Motivate st. 1 .182 .024 .380 -.067 .271 -.210 Motivate st. 2 -.043 -.078 .974 .032 -.120 .042 Motivate st. 3 .363 .165 .395 -.076 -.001 -.177 Motivate st. 4 -.078 -.053 .806 .135 -.021 .027 Maintain discipline 1 .261 -.001 -.053 .530 .173 -.105 Maintain discipline 2 -.082 -.031 .118 .742 .123 .060 Maintain discipline 3 .061 .127 .024 .698 .036 -.148 Maintain discipline 4 .166 .230 .082 .383 -.047 -.340 Cooperate 1 -.027 -.094 -.115 .054 .875 -.026 Cooperate 3 -.180 -.056 .074 .315 .671 .007
Extraction Method: Principal Axis Factoring. Rotation Method: Promax with Kaiser Normalization. Factor loadings above .30 are in bold.
Overall, the exploratory factor analysis did not support the original NTSES factor structure, with six factors. The items for Motivation, Maintain discipline and Adapt instruction to individual needs seem to form subscales relatively similar to the original NTSES, while subscales Instruction, Cope with change and Cooperate with colleagues and parents do not.
The second EFA revealed six factors, explaining 68.1% of the overall variations in the measure. Eight items loaded on factor 1 which explained 41.4% of the overall variation, six items loaded on factor 2 which explained 8% of the variation, four items loaded on factor 3 which explained 6.3% of the variation, four items loaded on factor 4 which explained 4.6% of
the variation and 2 items loaded on factor 5 which explained 4.1% of the variation. The sixth factor explained 3.7% of the variation but had no items loading above the cut-off value of 0.3, as recommended by Field (2013). The loadings for factor 1 items ranged between .43 and .93, the loadings for factor 2 items varied between .33 and .93, the loadings for factor 3 items ranged between .40 and .97, the loadings for factor 4 items ranged between .38 and .74, the loadings for factor 5 items varied between .67 and .88 and finally, the loadings for factor 6 items varied between .01 and .28. The factor loadings, post rotation, for the 6 factors can be seen in table 6.
To summarize, the second EFA did not support the original 6-factor NTSES structure. As with the first EFA, the subscales, Instruction, Cope with change and Cooperate with colleagues and parents have been split up, while the items for Motivation, Maintain discipline and Adapt instruction to individual needs stay together. The two items asking specifically about cooperation with parents have formed their own dimension (factor 5).
Teacher self-efficacy is a widely researched concept related to the teaching
profession. It has shown ties to lower levels of burnout in teachers (Shoji et al., 2016) and higher levels of teacher well-being (Huang, 2019). Teachers who decide to keep teaching have shown to experience higher levels of self-efficacy compared to their former colleagues, who have left the profession (Hong, 2012). Additionally, teacher self-efficacy is also related to student outcomes, such as student motivation, achievement and student self-efficacy (Aloe el al., 2014; Capara et al., 2006; Moran & Woolfolk Hoy, 2001).
The aim of the present study was to validate a Swedish version of the Norwegian teacher self-efficacy scale (NTSES; Skaalvik & Skaalvik, 2007) in a Swedish context. Bandura’s teacher self-efficacy scale was used to test the convergent validity of the scale, while the constructs teacher well-being (WHO-5) and teacher burnout (CBI) were used for
measuring concurrent validity. The dimensions of the Swedish version of NTSES were explored by means of both Confirmatory factor analysis (CFA) and Exploratory factor analysis (EFA). The results revealed a Swedish NTSES with mostly high reliability, and with a moderate negative correlation to teacher burnout and a large and moderate positive
correlation to Bandura’s teacher self-efficacy scale and teacher well-being, respectively. Neither the CFA nor the two EFA’s provided any support for a 6-factor structure, coinciding with the original NTSES.
Despite the fact that the Swedish version of NTSES does not seem to have the exact same dimensions as the original scale, the two EFA reveals factor structures that evidently have things in common with it. Especially in the EFA with 6 fixed factors, three of the original six scales are intact (Adapt instruction to individual needs, Maintain discipline and Motivation), while two others (Instruction and Cope with change) have three out of four items staying together. Additionally, the CFA reveals values only slightly below the cut-off for a satisfying model fit. This indicates that the Swedish version of NTSES is measuring constructs similar to the original NTSES.
The Swedish NTSES shows, mainly, adequate internal consistency. These results are in agreement with the original scale (Skaalvik & Skaalvik, 2007; 2010), which showed similar reliability, and also revealed the subscale Cooperating with colleagues and parents to be the least reliable measure. The reliability of the Swedish version of the NTSES is also in line with a previous longitudinal study conducted by Avanzi et al. (2013). Avanzi and colleagues translated the NTSES into Italian, and found a comparable internal consistency in the measures, with a downtrend in the Cronbach’s alpha values for Cooperating with
colleagues and parents at both time-points.
That the subscale Cooperating with colleagues and parents seems to be the least dependable scale in the Swedish NTSES is further implied by the EFA:s. The items regarding
cooperation with parents and the items asking about cooperation with colleagues have separated in both EFA:s, although the split-up is clearer in the EFA with 6 fixed factors, where cooperation with parents have formed their own factor. This could support the speculations by Avanzi et al. (2013), that working well with parents and managing
relationships with co-workers represents two different dimensions of teacher self-efficacy. The Swedish version of NTSES shows satisfying concurrent and convergent validity. Correlations between teacher burnout and teacher well-being are consistent with previous research (Avanzi et al., 2013; Harding et al., 2019; Huang et al., 2019; Skaalvik & Skaalvik, 2014) and the results suggest a significant covariance with Bandura’s teacher self-efficacy scale. Bandura’s scale measures, in part, different dimensions of teacher self-efficacy than NTSES. This might explain why the correlation is high, but still just below .7. In sum, the Swedish version of the NTSES seems to measure teacher self-efficacy, even though the domains within it remain somewhat unclear.
The results also reveal a significant relationship between teacher self-efficacy and age, and between teacher self-efficacy and teaching experience. Since age and teaching experience also correlate quite strongly, they seem to go hand in hand - the older teachers have more teaching experience. Bandura (1997) explains that one of the ways to increase self-efficacy is through mastery experiences, referring to previous successful experiences of teaching that will enhance the level of belief in your abilities. Older teachers will probably have had a lot more chances to experience successful teaching, and therefore, their self-efficacy will be higher than their younger colleagues.
Despite following the International Test Commission (ITC) guidelines for translation of scales (Gregoire, 2018), the translation is a limitation in the present study. The Swedish version of NTSES was translated from the English version of the NTSES, even though the
original version is in Norwegian. Access to the Norwegian version was gained later in the process of this study, but could not be used as a base for translation. Through comparison of the Swedish and Norwegian version, differences in meaning for certain items became evident, possibly leading to the items loading on different factors in the exploratory factor analysis. For example, item four in the Cope with change subscale was translated using the word instruction, which is similar to the English version, yet the Norwegian version uses a word with a much broader meaning, incorporating the whole teaching situation. This could explain why the item loaded on the same factor as the Instructions and Adapt instructions to individual needs items, instead of loading with the other Cope with change items.
Another limitation may be the cultural adaption of the scale. To ensure face validity and cultural appropriateness, four teachers completed the survey beforehand and provided feedback. Several of these teachers expressed concern about items in the Cope with change subscale, saying they did not experience pressure to use methods they were not comfortable with. It might be that the changes made in the Norwegian school system are not similar to those made in Sweden, and that, therefore, the items belonging to the subscale did not come across as relevant to our sample. However, when comparing the English NTSES and the Norwegian NTSES the words used in the Cope with change subscale were different in meaning. One of the items translates (from Norwegian) to “forced to use methods you don’t believe in” which implies something more important than using a method that wouldn’t have been your choice, which is the wording in the English version. Translating the item with “don’t believe in'' might have made it equally relevant in the Swedish context.
Nevertheless, one of the most explicit recommendations by Bandura (1997; 2006) when measuring self-efficacy, is the tailoring of the scale to the domains and subdomains you wish to measure. Since the scale used in this study was translated, alterations in items for the purpose of cultural adaption was limited to change of wording, and altering whole subscales
was not possible. A Swedish teacher self-efficacy scale might need an even further investigation into the role and challenges of Swedish teachers. Additionally, following Bandura’s recommendations to a T could also indicate that a self-efficacy scale needs to be updated regularly, to keep up with the changing times. An item asking about a sense of self-efficacy regarding a change in curriculum, for example, could be relevant at one point in time, but feel totally out of place a few years later.
In order to make insurance concurrent validity of the Swedish NTSES a possibility, the second teacher self-efficacy scale was also translated, since no teacher self-efficacy scales in Swedish existed. This is not ideal, even though a scale from Bandura himself was used and thus, at least, ensuring that the concept was properly captured. The double translations could imply that the large correlation between NTSES and Bandura’s teacher self-efficacy scale is due to similar wordings or other resemblances related to translations.
The sample in this study was one of convenience, due to the time limitation of the thesis. The teachers participating had relatively high self-efficacy and high levels of well-being. As seen in the results, teachers recruited by their principals had a higher level of self-efficacy than did the teachers recruited by friends. This may be because of the headmasters of the schools acting as a filter, since the headmasters who declined to participate often
mentioned that their teachers were under a lot of pressure and stress. Literature shows that being in a stressful workplace for a prolonged amount of time is a risk factor for burnout (Pines & Aronson, 1988). This may have resulted in the present study not reaching teachers experiencing higher levels of burnout and possibly lower levels of self-efficacy, hence, giving this study a non-representative sample.
A limitation of this study could be the sample size. According to the a-priori power analysis in the present study the minimum number of participants should be at least 200, this was exceeded with a final sample of 256 participants. Nonetheless, this is a small sample
when conducting a factor analysis, since the underlying assumptions of the test are linked to large sample theory (Byrne, 2001). One of the underlying assumptions of the CFA is
multivariate normality. When analysing the data, the NTSES as a total was normally
distributed, but when checking each of the subscales, the Adapt to individual needs subscale was not normally distributed. A larger sample is more robust and would probably yield a normal distribution for each subscale. Another way to check for multivariate normality, used by Fiorilli et al. (2015), is to analyse the distribution for each item. The sample in the present study was far too small to handle an analysis of such small elements of the data, therefore this approach was ruled out.
A possible limitation, related to the burnout measure, is the use of the word student instead of client. This appropriated the measure for the school context, even so, some studies suggest conflict or relations with parents may be more draining than the relationships with students (Skaalvik & Skaalvik, 2010). This was also commented on by one of the teachers in the panel saying students were not the problem. A participant provided feedback via email, expressing that the most draining part of the work was all the meetings with colleagues, taking time from preparation for teaching. The decision not to include a colleague burnout and parent burnout subscale was done based on the length of the survey, not wanting to tire the participants. In hindsight, this exclusion might have decreased the mean level of burnout measured in our sample.
Implications and future research
Recent research has found self-efficacy to be both predictive of burnout and affected by burnout (García-Ros et al., 2015; Kim & Burić, 2020). Kim and Burićs’ (2020) suggestion is to focus on interventions lowering levels of burnout instead of those enhancing
self-efficacy. They argue that a low degree of self-efficacy is the product of burnout based on their results and hypothesize that this is caused by the feeling of achieving less at work due to
the burnout, hence a decrease in mastery experiences. García-Ros and colleagues (2015) instead found self-efficacy to be predictive of burnout and argued that activities aimed to raise teachers’ levels of self-efficacy should be prioritized as a preventive action against burnout. The present study adds to the literature, showing that there is a significant relation between the two constructs, but not how they are related. The same is true for the correlation between self-efficacy and well-being, time and time again studies find a correlation but not a causal relationship. These relations need to be investigated further through longitudinal studies, to shed light on how to improve teacher well-being and decrease levels of burnout.
Nevertheless, there are still practical implications of using self-efficacy measurements. Completing the NTSES may aid teachers in reflecting upon one's own strengths and weaknesses. For school leaders it can help assess the need for support, specific training or education to help teachers further develop their skills and get the needed resources for this cause. The present study has found a good internal consistency for the translated scale as a whole, however the same cannot be said for the subscales. Additional validation and adjustments of the Swedish version of the NTSES are needed before it can be of use in Sweden.
Recommendations for future research is a larger sample size and a recruiting method which enables an unbiased sample. For example, using the same procedure as Skaalvik and Skaalvik (2011), working closer to the participants by communicating with representatives for the teachers, instead of principals, and having time set aside for the survey during the teacher’s workday. This would lead to a more representative sample where all teachers at a school get the chance to participate, not having to use their leisure time. The Swedish version of the NTSES also needs to be compared to the original language in order to ensure an adequate translation. Adjustments in wording might lead to a factor structure more like the original NTSES than has been revealed in this study.
As mentioned above, the Cooperate with parents and colleagues subscale had a lower internal consistency than the other subscales, a result mirrored in previous research (Skaalvik & Skaalvik, 2007; Avanzi, 2013). Additionally, in the present study the factor analyses revealed that the items concerning parents and the items concerning colleagues loaded on different factors. Therefore, suggestions for future studies are to split this subscale into two separate ones, a cooperate with colleagues subscale and a separate cooperate with parents subscale. Further interviews with educators may be needed to understand their role and tasks related to both colleagues and parents, which may yield new items to add to the pre-existing ones. This also ties together with feedback given from participants about the burnout
measure. For further investigation of burnout prevalence among teachers future research may benefit by adding parent related burnout as a subscale when using the CBI. This could give a more nuanced picture and provide important information of what teachers experience as causes of burnout.
In summary, even though the Swedish version of NTSES seems to measure teacher-self efficacy as a whole, it needs adjustments in translation and, perhaps even at a subscale level, to truly be of use in a Swedish context. Even though the results indicate good
concurrent and convergent validity, the Bandura Teacher Self-Efficacy Scale was also translated, which might facilitate the correlation, and the sample mostly seems to be teachers feeling well enough not to be protected by their principal. This makes the results less
generalizable. However, the Swedish NTSES still has a factor structure relatable to the original, and with some adjustments, the scale could be a useful tool in trying to capture teacher self-efficacy.
Aloe, A. M., Amo, L. C., & Shanahan, M. E. (2014). Classroom management self-efﬁcacy and burnout: A multivariate meta-analysis. Educational Psychology Review, 26(1), 101 –126.
Arneson, H. (2006). Empowerment and health promotion in working life (PhD dissertation). Institutionen för hälsa och samhälle, Linköping. Retrieved from
Arneson, H., & Liljegren, M. (2005). Svensk utprovning av Copenhagen Burnout Inventory (CBI). [Swedish evaluation of the Copenhagen Burnout Inventory (CBI)].
Unpublished manuscript, Linköping University.
Arvidsson, I., Håkansson, C., Karlson, B., Björk, J., & Persson, R. (2016). Burnout among Swedish school teachers – A cross-sectional analysis. BMC Public Health, 16(823), 1–11. https://doi.org/10.1186/s12889-016-3498-7
Avanzi, L., Miglioretti, M., Velasco, V., Balducci, C., Vecchio, L., Fraccaroli, F., & Skaalvik, E. M. (2013). Cross-validation of the Norwegian Teacher’s Self-Efficacy Scale (NTSES). Teaching and Teacher Education, 31, 69–78.
Bandura, A. (n.d.). Bandura’s Instrument Teacher Self-Efficacy Scale. Retrieved October 27, 2020, from http://www.till.org.uk/wp-content/uploads/2019/03/Teacher-Self-Efficacy -Scale -Bandura-1998.pdf
Bandura, A. (1997). Self-efficacy: The exercise of control. New York: Freeman.
Bandura A. (2006). Guide for constructing self-efficacy scales. In Pajares, F. & Urdan, T.C. (Eds.), Self-efficacy beliefs of adolescents (pp. 307 - 337). Greenwich, CT:
Information Age Publishing.
Journal of Management, 38(1), 9–44.
Bech, P., Olsen, L. R., Kjoller, M., & Rasmussen, N. K. (2003). Measuring well-being rather than the absence of distress symptoms: A comparison of the SF-36 Mental Health subscale and the WHO-Five well-being scale. International Journal of Methods in Psychiatric Research, 12(2), 85–91. https://doi.org/10.1002/mpr.145
Berntson, E., Bernhard-Oettel, C., Hellgren, J., Näswall, K., & Sverke, M. (2016). Enkätmetodik [Survey methodology] (1 ed.). Stockholm: Natur & kultur. Bianchi, R., Mayor, E., Schonfeld, I. S., & Laurent, E. (2018). Burnout and depressive symptoms are not primarily linked to perceived organizational problems. Psychology, Health & Medicine, 23(9), 1094–1105.
Bong, M., & Skaalvik, E. M. (2003). Academic self-concept and self-efficacy: How different are they really? Educational Psychology Review, 15(1), 1–40.
Brislin, R. W. (1970). Back-translation for cross-cultural research. Journal of Cross-cultural Psychology, 1, 185e216.
Busseri, M. A., & Sadava, S. W. (2011). A review of the tripartite structure of subjective well-being: Implications for conceptualization, operationalization, analysis, and synthesis. Personality and Social Psychology Review, 15(3), 290.
Byrne, B.M. (2001). Structural equation modeling with AMOS: basic concepts, applications, and programming". Mahwah: Lawrence Erlbaum Associates, Publishers.
Campos, J. A. D. B., Carlotto, M. S., & Marôco, J. (2013). Copenhagen Burnout
Brazil. Psicologia: Reflexão e Crítica, 26(1), 87–97.
Capone, V., & Petrillo, G. (2020). Mental health in teachers: Relationships with job
satisfaction, efficacy beliefs, burnout and depression. Current Psychology: A Journal for Diverse Perspectives on Diverse Psychological Issues, 39(5), 1757–1766.
Caprara, G. V., Barbaranelli, C., Steca, P., & Malone, P. S. (2006). Teachers’ self-efficacy beliefs as determinants of job satisfaction and students’ academic achievement: A study at the school level. Journal of School Psychology, 44(6), 473–490.
Carlson, K. D., & Herdman, A. O. (2012). Understanding the impact of convergent validity on research results. Organizational Research Methods, 15(1), 17–32.
Diener, E. (1984). Subjective well-being. Psychological Bulletin, 95, 542–575.
Diener, E. (2000). Subjective well-being: The science of happiness and a proposal for a national index. American Psychologist, 55(1), 34.
Diener, E., Tamir, M., & Scollon, C. N. (2006). Happiness, Life Satisfaction, and Fulfillment: The Social Psychology of Subjective Well-Being. In P. A. M. Van Lange (Ed.), Bridging social psychology: Benefits of transdisciplinary approaches. (pp. 319–324). Lawrence Erlbaum Associates Publishers.
Field, A. (2013). Discovering statistics using IBM SPSS statistics: and sex and drugs and rock 'n' roll. (4. ed.) Los Angeles: Sage.
Fiorilli, C., De Stasio, S., Benevene, P., Iezzi, D. F., Pepe, A., & Albanese, O. (2015). Copenhagen Burnout Inventory (CBI): A validation study in an Italian teacher group. TPM-Testing, Psychometrics, Methodology in Applied Psychology, 22(4), 537–551.