Longitudinal Study of Teachers' Occupational Well-being (PDF)

Journal of Occupational Health Psychology 2018, Vol. 23, No. 2, 262–277 © 2017 American Psychological Association 1076-8998/18/$12.00 http://dx.doi.org/10.1037/ocp0000070 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended...

Journal of Occupational Health Psychology 2018, Vol. 23, No. 2, 262–277 © 2017 American Psychological Association 1076-8998/18/$12.00 http://dx.doi.org/10.1037/ocp0000070 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. A Longitudinal Study of Teachers’ Occupational Well-Being: Applying the Job Demands-Resources Model Theresa Dicke Ferdinand Stebner Australian Catholic University, Strathfield, New South Wales Ruhr-University Bochum Christina Linninger and Mareike Kunter Detlev Leutner Goethe University Duisburg-Essen University The job demands-resources model (JD-R model; Bakker & Demerouti, 2014) is well established in occupational research, and the proposed processes it posits have been replicated numerous times. Thus, the JD-R model provides an excellent framework for explaining the occupational well-being of beginning teachers—an occupation associated with particularly high levels of strain and consequently, high attrition rates. However, the model’s assumptions have to date mostly been tested piecewise, and seldom on the basis of longitudinal models. With a series of longitudinal autoregressive SEM models (N ⫽ 1,700) we tested all assumptions of the JD-R model simultaneously in one model with an applied focus on beginning teachers. We assessed self-reports of beginning teachers at three time waves: at the beginning and end (one and a half to two years later) of their preservice period, and again, one year later. Results revealed significant direct effects of resources (self-efficacy) on engagement, of demands (classroom disturbances) on strain (emotional exhaustion), and a significant reverse path of engagement on selfefficacy. Additionally, the results showed two moderation effects: Self-efficacy buffered the demandsstrain relationship, while self-efficacy also predicted engagement, especially when disturbances were high. Thus, self-efficacy in classroom management plays an important role in the teachers’ stress development process, as it will, in case of high classroom disturbances, not only buffer the strain-enhancing effects, but also boost engagement. Commitment was predicted directly by emotional exhaustion and engagement, but indirectly only by self-efficacy (via engagement). Thus, we provide strong empirical support for the JD-R model. Keywords: job demands-resources model, occupational well-being, longitudinal, beginning teachers Supplemental materials: http://dx.doi.org/10.1037/ocp0000070.supp When one reviews articles on occupational stress research there is almost no way not to come across the job demands-resources (JD-R) model (Bakker & Demerouti, 2007). The JD-R model predicts organizational outcomes, such as job performance, occupational commitment, or leaving intentions, across several occupations (Schaufeli & Taris, 2014). The core predictions of the model are represented by two dual processes: (a) the so-called health impairment process, where job demands predict job strain, which in turn predicts organizational outcomes; and (b) the motivational process, where resources pre- dict engagement, which in turn also predicts organizational outcomes (Bakker & Demerouti, 2007). These processes are intertwined through direct, indirect, and interaction effects, which have been investigated thoroughly (Schaufeli & Taris, 2014). The JD-R model is well established, and its proposed processes have been empirically replicated in a large number of studies (Hu, Schaufeli, & Taris, 2013). Thus, there is strong support for the model being able to reflect the relationship of the working environment to job- related outcomes. Methodologically, however, although several recent studies have tested the model using diary This article was published Online First February 2, 2017. Theresa Dicke, Institute for Positive Psychology and Education, Australian Catholic University, Strathfield, New South Wales; Ferdinand Stebner, Department of Research on Learning and Instruction, Ruhr-University Bochum; Christina Linninger and Mareike Kunter, Institute of Psychology, Goethe University; Detlev Leutner, Department of Instructional Psychology, University of Duisburg-Essen. The present study was conducted using data from the BilWiss Study (01JH0911) and the BilWiss–Beruf Study (01PK11007B), both funded by the Federal Ministry of Education and Research (BMBF). The authors claim responsibility for the contents of this report. The opinions expressed are not necessarily the opinions of the BMBF. We thank the members of the BilWiss team, especially Olga Kunina-Habenicht, Franziska Schulze-Stocker, Tina Seidel, Ewald Terhart, Simone Emmenlauer, Franziska Czeka, Doris Holzberger, and Hendrik Lohse-Bossenz, for their support. The study has been presented at SELF Conference 2015, Kiel, Germany and the German PAEPS Conference 2015, Kassel, Germany. Correspondence concerning this article should be addressed to Theresa Dicke, Institute for Positive Psychology and Education, Australian Catholic University, Locked Bag, 2002, Strathfield, NSW 2135, Australia. E-mail: [email protected] 262 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. TEACHERS’ OCCUPATIONAL WELL-BEING studies (e.g., Bakker, Sanz-Vergel, & Kuntze, 2015) and applying longitudinal data (e.g., Barbier, Dardenne, & Hansez, 2013; Boyd et al., 2011; Hakanen, Perhoniemi, & Toppinen-Tanner, 2008; Schaufeli, Bakker, & van Rhenen, 2009), the examination of interaction effects has often relied on cross-sectional data (e.g., Bakker, Demerouti, & Euwema, 2005; Bakker, Hakanen, Demerouti, & Xanthopoulou, 2007; Xanthopoulou, Bakker, Dollard et al., 2007). Further, the specific causal pathways of the model have so far mostly been tested individually, and rarely simultaneously in one model (for models without interactions see: Brauchli, Jenny, Füllemann, & Bauer, 2015; Frins, van Ruysseveldt, van Dam, & van den Bossche, 2016). Most do not also include longitudinal data (but see Brough et al., 2013), and are not additionally based on latent variables (but see Barbier et al., 2013). Validating all proposed effects in the same empirical model, utilizing state of the art latent variable and interaction modeling, is one aim of the present study, and will contribute to empirically strengthening the overall JD-R model, as well as JD-R theory (Bakker & Demerouti, 2014). Thus, we utilize longitudinal data to model latent autoregressive, cross-lagged effects, and interaction effects, to be able to provide strong evidence for the effects proposed by the JD-R model (Bakker & Demerouti, 2014). To validate the JD-R model, in accordance with our aim, we applied it to a large and representative sample of beginning teachers. This enabled us to further investigate and identify variables that are important for the teacher stress development process, which was the applied focus of the present study. Indeed, investigating the reasons for and consequences of teacher strain, but also teacher occupational well-being, has become increasingly important (Brenninkmeijer, van Yperen, & Buunk, 2001). Further, teachers report high work-related strain (e.g., Hakanen, Bakker, & Schaufeli, 2006), resulting in high levels of teacher attrition—an important organizational outcome—this enables a direct translation of this process to the JD-R model. The present study focuses on beginning teachers1 in and after their induction phase. In contrast to many other countries, where teachers enter their career directly after university graduation (e.g., in the U.S.), German teacher education includes an additional induction phase, which begins after university education and must be completed before teachers are officially qualified. In the induction phase, the so called referendariat, beginning teachers are allocated to schools, where they gradually begin to teach (after about 2– 6 months) after having observed more-experienced teachers giving lessons. On a weekly basis, they attend seminars outside their schools and are mentored by experienced teachers. After successfully completing their induction period of 1 and a half to 2 years, beginning teachers start working as fully qualified teachers. Indeed, this induction phase is a particularly important phase with regard to work socialization and identity formation (Solinger, van Olffen, Roe, & Hofmans, 2013; for an overview on literature on teacher identity formation and teacher work socialization see Beijaard, Meijer, & Verloop, 2004; and Huberman, 1989; Pogodzinski, 2012, respectively). Irrespective of the teacher education system, however, these difficult first years of teaching have been given a particular label in being referred to as the time of “reality shock” (Friedman, 2000). As a result of this high strain, teacher attrition is greater in the early years of teaching (Schaufeli & Enzmann, 1998). Indeed, up to 50% of beginning U.S. teachers leave the profession within 263 the first 5 years (see Hong, 2010; Ingersoll, 2012). The high attrition rates of beginning teachers are associated with high financial costs for further recruiting and managing teachers, as well as with disruptions of program continuity and planning, which impacts school effectiveness (Hong, 2010). Engaged, healthy, and committed teachers, on the other hand, show fewer turn-over intentions (Hakanen et al., 2006) and better teaching performance (Klusmann, Kunter, Trautwein, & Baumert, 2006). In the following passages we will introduce the JD-R process, based on variables that are of major importance for (beginning) teachers, including a specific demand—namely, classroom disturbances—and a specific personal resource: namely, self-efficacy in classroom management. Moreover, we also introduce how the occupational outcomes of the JD-R model play an important role for beginning teachers. Further, we review the extent to which all proposed constructs of the JD-R model are relevant and related to each other in the development process of teacher-specific outcomes. Using these context (teaching) specific variables to model the relationship of demands, resources, and outcomes proposed by the JD-R is in line with the triple-match principle, which suggests that the strongest relationships between demands and resources and an outcome are observed when all three are based on qualitatively identical dimensions (de Jonge and Dormann, 2006; for an overview see de Jonge, Dormann, & van den Tooren, 2008). The JD-R Model There are several different approaches to modeling the development of stress and strain in stress research: these include the transactional model of stress and coping (Lazarus & Folkman, 1984), the job characteristics model (JCM; Hackman & Oldham, 1980), the job demands control model (DCM; Karasek, 1979), and the effort-reward imbalance model (ERI; Siegrist, 1996), to name just a few. The job demands resources (JD-R) model (Bakker & Demerouti, 2007; Bakker, Demerouti, De Boer, & Schaufeli, 2003) in particular, however, has increased in popularity among researchers investigating occupational stress in the past decade (Hu et al., 2013; Schaufeli & Taris, 2014). Indeed, on the basis of the large numbers of studies validating the model and the applications derived from their results, its developers have been able to extend the model to a comprehensive theory, referred to as the job demands-resources theory (Bakker & Demerouti, 2014). One possible reason for this increased engagement in developing the theory is that the JD-R model is based on and combines the strengths of other well-established stress models, such as the DCM or ERI models, but extends and advances these models through its broad scope and flexibility (e.g., Bakker & Demerouti, 2014). Apart from the fact that one of the model’s unique features is that it simultaneously incorporates both positive and negative processes affecting occupational well-being, the model can address various kinds of job demands and job resources (including personal resources) and can be applied to a wide variety of occupational work settings (for an overview see Schaufeli & Taris, 2014; also Bakker & Demerouti, 1 In the present study, beginning teachers were followed up from the beginning of their mandatory teacher induction time (with a duration of up to 2 years) where they finalized their university education, without being as yet fully qualified teachers, until 1 year after commencing work as a fully qualified teacher. This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 264 DICKE, STEBNER, LINNINGER, KUNTER, AND LEUTNER 2007). While the creators of the JD-R model propose that, substantively, these resources and demands depend on the occupational group under investigation (Bakker & Demerouti, 2007), recent research by Clausen, Burr, and Borg (2014) has shown some evidence that the similarities among occupational groups might be greater than the differences in respect of the configuration of demands and resources. The JD-R model is based on the assumption of there being specific physical, psychological, social, or organizational aspects of the job. These can have either negative or positive effects on occupational outcomes. Thus, the model distinguishes between job demands, which refer to all aspects of the job associated with specific physiological and/or psychological costs, as they require sustained physical and/or psychological effort (Bakker & Demerouti, 2014), and job resources. This latter comprise those that: “(a) [are] functional in achieving work goals; (b) reduce job demands and the associated physiological and psychological costs; or (c) stimulate personal growth, learning, and development” (Bakker & Demerouti, 2014, p. 9). As can be seen in Figure 1, the JD-R model suggests grouping several demands and resources into general higher order factors of demands or resources. Recent research by Luchman and GonzálezMorales (2013), however, has shown that a model including several job demands as individual factors fitted the data better and produced fewer counterintuitive results. One possible reason is the distinction of demands, depending on whether they are perceived as hindrances or challenges: The former have a negative, the latter a positive effect (Crawford, LePine, & Rich, 2010). Consequently, grouping them into one overarching construct will then produce inaccurate results. The present study focuses on latent factors for resources and demands, however, consisting of only one domain per construct. In the following, the basic assumptions of the model and the interrelations between the single elements are described (see Figure 1). Additionally, there is a stepwise introduction of the general assumptions of the JD-R model, where each assumption is addressed in respect of the theoretical assumptions and empirical findings from the teacher occupational well-being applied focus. Dual Processes: Important Demands, Resources, and Occupational Outcomes for Beginning Teachers The model suggests that job demands lead to a decrease in employees’ health (e.g., high workload leads to psychological strain): This can be referred to as the health impairment process. Additionally, the model proposes that employees’ resources in- Figure 1. crease their engagement: This can be referred to as the motivational process (Bakker & Demerouti, 2014). The posited dual processes and direct effects of the model are independent of each other, and have been supported by a large number of empirical studies (for an overview see, e.g., Bakker & Demerouti, 2007; Demerouti, Bakker, Nachreiner, & Schaufeli, 2001). Indeed, a recent meta-analysis confirmed the assumed processes for a wide variety of resources (e.g., autonomy and leadership) and demands (e.g., complexity and hazards), and for a broad set of occupations, such as construction, health care, and manufacturing (Nahrgang, Morgeson, & Hofmann, 2011). The JD-R model originally focused on those resources that can be found within the work environment—that is, job resources— rather than on those that are a product of personal characteristics. Recent studies, however, have examined the influence of personal resources (positive self-evaluations related to resiliency, such as self-beliefs or optimism), thus integrating insights from the Conservation of Resources Theory (COR; Hobfoll, 1989, 2001). Researchers have found evidence for a direct effect of personal resources on engagement (Xanthopoulou, Bakker, Demerouti, & Schaufeli, 2009). Nevertheless, empirical findings for the interaction effect, and even on the overall role of personal resources, seem to be inconsistent (Bakker & Demerouti, 2014; Schaufeli & Taris, 2014). Xanthopoulou, Bakker, Demerouti, and Schaufeli (2007) found no evidence for a buffering effect of personal resources such as self-efficacy, self-esteem, and optimism on the relationship of job demands to exhaustion. In contrast, while not explicitly referring to the JD-R model, Jex and Bliese (1999; see also Dicke et al., 2014) found a buffering effect of self-efficacy on the relationship between stress and strain. In addition, in a sample of nurses, Bakker and Sanz-Vergel (2013) found evidence that a boosting effect is valid for personal resources, and that the effect of personal resources on weekly engagement was particularly high in the case of high emotional job demands. The present study focuses on these personal resources, rather than on job resources, as research has shown that these play an important role in the teacher stress process (see below). Thus, we investigate these dual processes by testing longitudinally whether teacherspecific demands (classroom disturbances) and resources (selfefficacy in classroom management) lead to an increase in strain and engagement, respectively. The JD-R Model of Bakker and Demerouti (2007). TEACHERS’ OCCUPATIONAL WELL-BEING This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. The Health Impairment Process: The Relationship of Classroom Disturbances (Job Demand) and Emotional Exhaustion (Strain) Teachers often report high levels of self-perceived work-related strain (Hakanen et al., 2006; Klassen & Chiu, 2011; Smith, Brice, Collins, Matthews, & McNamara, 2000), such as emotional exhaustion, which has been identified as the central dimension of burnout for teachers (see Cropanzano, Rupp, & Byrne, 2003; Schaufeli & Enzmann, 1998). In addition, research has found strong empirical evidence for an increase of emotional exhaustion especially within the first years of teaching (Dicke, Parker, Holzberger, Kunter, & Leutner, 2015; Klusmann Kunter, Voss, & Baumert, 2012). Concludingly, researchers investigating the major reasons for teachers’ strain—that is, their job demands—found that issues regarding the social-psychological aspects of teaching, such as managing student behavior and teacher–student relationships (Spilt, Koomen, & Thijs, 2011), rather than instructional problems (e.g., low academic student achievement, difficulty in teaching new material) predict teacher strain for beginning teachers (see, e.g., Dicke et al., 2014; Friedman, 2006). In particular, the level of disruptive student behavior seems to account for the highest amount of variance in emotional exhaustion (Kokkinos, 2007) and has been found by many researchers (e.g., Bakker et al., 2007; Skaalvik & Skaalvik, 2014) to be the major predictor of teachers’ strain. Indeed, particularly teachers at the beginning of their teaching career, perceive managing student behavior as an overly challenging job demand (Jones, 2006; Rieg, Paquette, & Chen, 2007; Veenman, 1984). The Motivational Process: The Relationship of SelfEfficacy in Classroom Management (Resource) and Engagement Teacher self-efficacy, a personal resource, can be viewed as one of the most important resources of teachers (Dicke et al., 2014; Schwarzer & Hallum, 2008), particularly in their beginning years (Klassen & Chiu, 2011; Woolfolk Hoy & BurkeSpero, 2005). As we identify classroom disturbances to be a major demand (see above), we focus on self-efficacy in classroom management, which refers to “teachers’ beliefs in their capabilities to organize and execute the courses of action required to maintain classroom order” (Brouwers & Tomic, 2000, p. 242). This is in line with the triple-match principle’s double-match of a common kind, which refers to this correspondence between a demand and a resource (de Jonge et al., 2008). Further, this domain of teacher self-efficacy has become increasingly important in research on teacher self-efficacy throughout the last decades (O’Neill & Stephenson, 2011), and has been linked to all constructs focused upon in the present study. Self-efficacy (in classroom management) is directly and positively related to engagement (Klassen et al., 2009; Skaalvik & Skaalvik, 2014). Several researchers propose that, theoretically, emotional exhaustion is the opposite pole of engagement on a shared continuum of workplace well-being (Demerouti, Mostert, & Bakker, 2010; Schaufeli & Bakker, 2003). While emotional exhaustion is characterized by feelings of fatigue and being emotionally drained (Maslach, Jackson, & Leiter, 1996; 265 see also Maslach, 1999), engagement refers to energy, enthusiasm, and dedication toward work (Schaufeli & Bakker, 2003). Empirically, this negative relationship has found support in teacher research. For example, Hakanen et al. (2006) found a high negative relationship of burnout and engagement in a large sample of teachers. Hypothesis 1: Concerning the dual processes assumed by the JD-R model, we first test stable and direct effects. We expect: a. Job demands (i.e., classroom disturbances) to positively predict emotional exhaustion and negatively predict engagement, over and above emotional exhaustion and engagement predicting themselves; b. Personal resources (i.e., self-efficacy in classroom management) to positively predict engagement and negatively predict emotional exhaustion, over and above emotional exhaustion and engagement predicting themselves. Reciprocal Relationships: How Teacher Variables Affect Each Other Recently, the authors of the JD-R model have begun to investigate how increased job demands and resources not only lead to, but can also be a result of, high strain or engagement respectively (Schaufeli & Taris, 2014). Hence, job demands are also predicted by exhaustion (Demerouti, Le Blanc, Bakker, Schaufeli, & Hox, 2009; Zapf, Dormann, & Frese, 1996), and job resources, as well as personal resources, are also predicted by engagement (Llorens, Schaufeli, Bakker, & Salanova, 2007; Schaufeli et al., 2009; Xanthopoulou et al., 2009). Salanova, Bakker, and Llorens (2006) found that music teachers’ job resources, such as social support, as well as their self-efficacy, predicted work flow, while at the same time, work flow predicted both kinds of resources. Similarly, Bakker and Bal (2010) also found evidence for reciprocal relations of job resources and engagement in beginning teachers. Applying these model assumptions to the variables of the present study, on the basis of self-efficacy theory (Bandura, 1997), classroom disturbances should be negatively related to self-efficacy in classroom management. In other words, higher levels of self-efficacy in classroom management should predict fewer classroom disturbances (Dicke et al., 2014). However, other researchers (e.g., Klassen & Chiu, 2010) have found evidence for the reverse relationship. These apparently contradictory findings suggest a reciprocal relationship of resources and demands, as stated in the JD-R model. Theoretically, while a major source of self-efficacy is mastery experience—the perception of one’s own behavior as successful (in this case, reflected by a low level of classroom disturbances as a prerequisite for successful teaching), selfefficacy also predicts performance (Bandura, 1997). So far, several researchers have found support for this reciprocal relationship (e.g., Maslach, 1999; Skaalvik & Skaalvik, 2010). Furthermore, on the basis of longitudinal data on secondary teachers, Salanova et al. (2006) found general self-efficacy to predict work-related flow and vice versa; this was also found for teacher efficacy and engagement (Simbula, Guglielmi, & Schaufeli, 2011). 266 DICKE, STEBNER, LINNINGER, KUNTER, AND LEUTNER Hypothesis 2: Concerning reciprocal effects we expect: a. Job demands and personal resources, as well as emotional exhaustion and engagement, to be negatively related cross-sectionally; b. Personal resources and engagement to positively predict each other longitudinally. This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Interactions: Boosting and Buffering Effects on Teachers’ Occupational Well-Being Furthermore, the JD-R model assumes interaction effects of job demands and resources on strain and engagement. The first of these effects, the so-called “buffering effect” (Bakker & Demerouti, 2007), proposes that high resources will decrease the impact of job demands on strain. Thus, researchers were able to show how job resources could buffer the demands-strain relationship (e.g., Bakker et al., 2005; van Woerkom, Bakker, & Nishii, 2016; Xanthopoulou et al., 2007). Garrick et al. (2014) provide evidence for several interaction effects in a sample of school teachers, where psychological safety climate moderated, on the one hand, the relationships of job demands with fatigue and engagement and, on the other hand, also the relationship of recovery with fatigue and engagement. The second interaction effect proposes that the positive effect of resources on engagement will be particularly pronounced in the case of high demands, and could thus be best described as a “boosting effect” (Bakker et al., 2007). Hence, Bakker et al. (2007) showed that several job resources (supervisor support, innovativeness, appreciation, and organizational climate) predicted teachers’ work engagement, particularly when pupil misbehavior was regarded as an important work demand. Similar findings were provided by other studies (e.g., Hakanen, Bakker, & Demerouti, 2005; Bakker, van Veldhoven, & Xanthopoulou, 2010). Several studies have provided empirical evidence for these moderation effects (although mostly these have been tested in isolation from each other) and all additional assumed relationships within the model. Furthermore, according to Hu, Schaufeli, and Taris (2013), the evidence for these interactions is still weak and inconsistent. Focusing on those variables in the beginning teacher context that are important for the present study, research showing a negative relationship of teacher efficacy and burnout is plentiful (e.g., Dicke et al., 2015; Federici & Skaalvik, 2012; Friedman, 2006; Skaalvik & Skaalvik, 2010). Utilizing a meta-analytic approach to sum up recent and state-of-the-art research on the relationship of self-efficacy to classroom management and burnout, Aloe, Amo, and Shanahan (2014) showed that high levels of self-efficacy in classroom management are associated with lower levels of emotional exhaustion, at an average correlation of r ⫽ ⫺.28. More importantly however, teacher self-efficacy—that is, self-efficacy in classroom management—is able to buffer the effect of demands on strain and, in particular, emotional exhaustion (Dicke et al., 2014; Schwarzer & Hallum, 2008): This is in line with the assumed moderating effects of personal resources on the demandstrain relationship in the JD-R model. Research testing the relationship between classroom disturbances and engagement has yielded inconsistent results. Klusmann, Kunter, Trautwein, Lüdtke, and Baumert (2008) could not find any significant effect of student misbehavior on engagement. Bakker et al. (2007), however, found a negative effect of class- room disturbances on engagement. Moreover, they found that high job resources (such as job control, supervisor support, information, good organizational climate, innovativeness, and appreciation) were able to buffer this negative effect. Furthermore, they provided evidence for a boosting effect, in which these job demands positively predicted teachers’ work engagement, particularly when pupil misbehavior was high. The resources (self-efficacy in classroom management), demands (classroom disturbances), and outcomes (emotional exhaustion and engagement) of interest for the present study suggest a triple match, as the above stated studies could show the strong interrelation and shared domain of these variables. This match, according to triple-match principle (de Jonge et al., 2008), should increase chances to observe the expected moderation effects. Hypothesis 3: Regarding possible interaction effects of the JD-R model, we hypothesize an interaction of self-efficacy in classroom management and classroom disturbances, in predicting engagement (booster effect) and emotional exhaustion (buffer effect). Occupational Outcomes: Teachers’ Occupational Commitment The model also assumes that strain and engagement, in turn, both lead to different levels of occupational outcomes, such as job performance (Bakker, Demerouti, & Verbeke, 2004), commitment, ill-health (Hakanen et al., 2006), or absences (Bakker et al., 2003). For beginning teachers, Bakker and Bal (2010) found reports of weekly work engagement predicting job performance. Especially interesting for the present study are the findings of Hakanen et al. (2006), who found that burnout mediated the effect of job demands on ill-health, while work engagement mediated the effect of resources on organizational commitment. Wang, Lu, and Siu (2015), on the basis of longitudinal data with a sample of Chinese employees, confirmed a mediating effect of engagement where job characteristics predict occupational outcomes. As noted above, up to 50% of beginning teachers leave the profession within the first 5 years (see Hong, 2010; Ingersoll, 2012) and yet, lack of occupational commitment, the intention to leave, and actually leaving the profession, are empirically distinct concepts, as the first refers to an attitude, the second to an intention and the third to an action. They are nevertheless related, as in many cases low occupational commitment will lead to an intention to quit, which then can eventually lead to teacher attrition (Klassen & Chiu, 2011). Thus for the present study, as the last, and slightly more distal occupational outcome of the JD-R model, we propose occupational commitment. Based on Meyer, Becker, and Vandenberghe’s (2004) definition of organizational commitment, which targets a specific organization rather than a union or profession (Meyer, Allen, & Smith, 1993), occupational commitment can be defined as a force that binds an individual to an occupation, and thereby reduces the likelihood of turnover. However, research on occupational commitment of teachers is scarce (Klassen & Chiu, 2011), although it has generated interesting findings regarding its relationship to the aforementioned other constructs. Hakanen et al. (2006), as mentioned above, showed that high levels of burnout can lead to a decrease of occupational commitment, while engage- This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. TEACHERS’ OCCUPATIONAL WELL-BEING ment showed the reverse relationship: that is, positively predicting occupational commitment of teachers. Other researchers have found teachers’ job stressors, such as classroom disturbances, to be negatively predicted by occupational commitment (Jepson & Forrest, 2006), but also to predict occupational commitment (Klassen & Chiu, 2011). However, both studies were conducted with cross-sectional data, which does not allow for causal interpretations. In the present study we focus on the indirect effect of classroom disturbances on occupational commitment via burnout, as proposed by the JD-R model and empirical findings by Hakanen et al. (2006). Importantly, we base our analysis on longitudinal data, to provide further evidence of prior stress and resources predicting later occupational commitment (Klassen & Chiu, 2011). Hypothesis 4: In addition to the hypothesized direct effects, when including occupational commitment as a more distant occupational outcome in the model, we expect classroom disturbances to indirectly predict occupational commitment via emotional exhaustion, while self-efficacy in classroom management indirectly predicts occupational commitment via engagement. In these processes, emotional exhaustion negatively predicts occupational commitment, while engagement positively predicts occupational commitment. The Present Study The present study aims to test all hypothesized main assumptions of the JD-R model simultaneously in a sample of beginning teachers. In the past, researchers have often focused on testing isolated parts of the model, depending on their area of interest. Exceptions based on longitudinal data, for example, are Brough et al. (2013) who, testing the assumptions of the model and its proposed interactions in a wide array of occupational settings (finance, health, education, manufacturing [eyeglasses], and nongovernment organizations) found evidence for the motivational process, but little evidence for the health impairment and interaction processes. Further, using three waves of longitudinal data in a sample of Belgian public administration workers, Barbier et al. (2013) expanded the traditional JD-R model by including perceived stigma and group identification as predictors of occupational well-being at work. In contrast to the aforementioned study by Brough et al. (2013), they found stronger evidence for the health impairment process, and little evidence for the motivation process. In respect of interaction effects, the results were inconsistent. Thus, the present study contributes to substantive issues and research gaps regarding the relationship of job demands (classroom disturbances), personal resources (self-efficacy in classroom management), strain (emotional exhaustion), engagement, and the occupational commitment of teachers, thereby empirically strengthening the JD-R model and, thus, JD-R theory. Moreover, we apply state of the art statistical methods by modeling all relationships and possible interactions (a) simultaneously, thus controlling for possible influences of one variable on the others; (b) on the basis of latent models, thus controlling for measurement error; and (c) on the basis of longitudinal data, thus providing strong evidence for the causal assumptions of the model. 267 Method Participants Participants were teacher candidates working in German schools in the state of North-Rhine Westphalia. For the present study it is important to note that the referendariat (induction period) in North-Rhine-Westphalia, the state where this study was conducted, underwent a reform in the year 2011. One major change of this reform was the duration of the referendariat, as it was shortened from 2 years to 1 and a half years. Further, the reform aimed to implement a more person-oriented approach, with fewer exams and more individual coaching (Kunter, Linninger, Schulze– Stocker, Kunina-Habenicht, & Lohse-Bossenz, 2013). However, initial results evaluating the reformed referendariat show little to no differences between development characteristics of both cohorts (those that underwent the old referendariat and those that underwent the later version; Dicke et al., 2016; Kunter et al., 2013). The sample of this study (N ⫽ 1,763) consisted of 27.9% male and 72.0% female teacher candidates with a mean age of M ⫽ 27.49 years, SD ⫽ 4.15 (beginning of the study; Time 1). Within this sample, 7.7% taught at primary schools, 26.9% in lower secondary school tracks, 38.8% in academic school tracks, and 26.6% in vocational high school tracks. Further, 70.6% of the sample took part in the prereform induction time (duration approximately 2 years), while 29.4% underwent the reformed induction (duration approximately 1 and a half years). Study Design Data on these teacher candidates in their induction phase and 1 year beyond, were collected as part of a large research project on teacher professional development. In the present study we focus on data assessed at three points in time: First, at the beginning of the induction phase (Time 1); second, at the end of the induction phase, irrespective of the duration of the induction phase (between 1.5–2 years; Time 2), and finally, one year after having commenced working as a fullyfledged teacher (between 2.5–3 years; Time 3). At Time 1 (the beginning of the induction phase), the teacher candidates (N ⫽ 1,763) were approached during their seminars, where they voluntarily completed the questionnaire (and a test battery that is not focus of the present study). The participation rate was very high (⬎ 90%), indicating a representative sample of teacher candidates in North-Rhine-Westphalia. The study applied a multimatrix design, with 12 booklets (with approximately N ⫽ 300 per booklet). The advantage of the multi matrix technique lies in being able to include more variables in the data set without adding extra time for each participant to fill in the questionnaire, because not every scale is completed by each participant (see Graham, Taylor, Olchowski, & Cumsille, 2006). For the scales in the present study this design reduced the length of each individual questionnaire/test, but we ensured that each scale was met with a sufficient sample size (Enders, 2010; see also Graham, 2012). This led to the effect of random “missings by design” at Time 1. To obtain longitudinal data at Time 2 (end of the induction phase) and Time 3 (1 year after starting in a regular teaching job), a reduced subsample of all participants of Time 1 (N ⫽ 671) was administered a reduced number of scales, without matrix sampling. 268 DICKE, STEBNER, LINNINGER, KUNTER, AND LEUTNER This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Participants for this longitudinal sample were selected, ensuring a representative sample of school types and regions. In effect, we treated data as missing completely at random (MCAR) at Time 1, and missing at random at Times 2 and 3 (Parker, Marsh, Morin, Seaton, & Van Zanden, 2015). Missing data was handled using the full integrated maximum likelihood (FIML) approach (Enders, 2010). This allowed for the inclusion of all participants who had filled in at least one scale at one time wave. Hence, we used all available data for the total sample. The average covariance coverage rate was 18%; the actual figure of the covariance coverage depended on the number of scales the person had filled in. Measures Classroom disturbances, self-efficacy in classroom management, engagement, and emotional exhaustion were measured at Time 1 and Time 2. Occupational commitment was measured at Time 3, when participants were fully employed and could relate to the construct. All variables were modeled as latent variables with multiple indicators. McDonald’s (1999) omega, which reflects the proportion of variance in the scale scores accounted for by a general latent factor, is reported as a measure of internal consistency (see also, e.g., Zinbarg, Yovel, Revelle, & McDonald, 2006). Emotional exhaustion was measured with a short version of the emotional exhaustion subscale of the German version (Enzmann & Kleiber, 1989) of the Maslach Burnout Inventory (MBI: Maslach, Jackson, & Leiter, 1996). Items (e.g., “I feel emotionally drained from my work as a teacher”) were measured on a 4-point Likert scale (omega 0.73/0.77). Self-efficacy in classroom management was measured with an adapted, and shortened German version of the self-efficacy in classroom management subscale of the Teachers’ Sense of Efficacy Scale (TSES), developed by Tschannen-Moran and Woolfolk Hoy (2001). Items (e.g., “How well can you respond to defiant students?”) were measured on a 6-point Likert scale (omega 0.84/0.87). Classroom disturbances were measured with a scale developed by Baumert et al. (2008). Items (e.g., “At the beginning of the lesson it takes a long time until the students calm down and start to work”) were measured on a 6-point Likert scale (omega 0.74/0.78). Engagement was measured with a German translation of the Utrecht Engagement Scales (Schaufeli & Bakker, 2003). Items (e.g., “When I wake up in the morning, I look forward to going to work”) were measured on a 4-point Likert scale (omega 0.90/0.81). Occupational commitment was measured by a German translation of the Occupational Commitment Scale by Meyer, Allen, and Smith (1993). Items (e.g., “I am proud to be a teacher”) were measured on a 4-point Likert scale (omega.72). Analysis CFI, TLI, and RMSEA are reported as fit indices, in addition to the ␹2 value, as the ␹2 value depends on sample size, where even small amounts of misfit can lead to significant ␹2 values when sample sizes are moderate to large (Chen, 2007). For the RMSEA, values ⱕ.05 are taken to reflect a good fit, values between.05 and.08 an adequate fit (Browne & Cudeck, 1993). For TLI and CFI, values of.90 or higher are considered a satisfactory fit, whereas values above.95 are considered excellent fit (McDonald & Marsh, 1990; Hu & Bentler, 1999). All interactions were modeled with latent variables (for details please see supplemental materials). Longitudinal invariance was achieved for all constructs (see supplemental materials). To test the proposed longitudinal effects of our hypothesized model, we utilized autoregressive (or cross-lagged panel) models. These models allow for testing the antecedent effects of one variable on change in another variable, while controlling for stability of the variable over time (Duncan, Duncan, & Strycker, 2006; Parker, Lüdtke, Trautwein, & Roberts, 2012). For details please see supplemental materials. Results Descriptives A number of items and the latent correlations of the latent factors representing all variables are presented in Table 1. All variables correlate significantly with each other at different time points, with the exception of self-efficacy in classroom management. The prediction (i.e., regression) of earlier self-efficacy in classroom management on itself over time however, is significant (see stability [lagged] model).2 For an overview of model fit for all subsequent models, please see Table A in the supplemental materials. Hypotheses 1 and 2: Replicating the Dual Processes and Possible Reciprocal Effects of the JD-R Model: Autoregressive Models In a series of autoregressive models we tested for stable, direct, reverse, and reciprocal effects. Stability (lagged) model. First, we set up a model implying stable effects of the variables over time. Thus, the model included autoregressive paths of the variables at Time 1 on themselves at Time 2. In addition, all variables were allowed to correlate with each other within Time 1 and within Time 2. Model fit was good, with ␹2 ⫽ 769, df ⫽ 305, p ⬍.001, CFI ⫽.94, TLI ⫽.94, RMSEA ⫽.03. All variables showed moderate levels of stability, significantly predicting themselves positively with emotional exhaustion ␤ ⫽.64, p ⬍.001, classroom disturbances ␤ ⫽.60, p ⬍.001, engagement ␤ ⫽.54, p ⬍.001, and self-efficacy in classroom management ␤ ⫽.53, p ⬍.001. In addition, the hypothesized (Hypothesis 2a) cross-sectional relationships of self-efficacy in classroom management and classroom disturbances (Time 1: ␤ ⫽ ⫺.27, p ⬍.001, Time 2: ␤ ⫽ ⫺.35, p ⬍.001) as well as emotional exhaustion and engagement (Time 1: ␤ ⫽ ⫺.57, p ⬍.001, Time 2: ␤ ⫽ ⫺.521, p ⬍.001) were confirmed. This structure of cross-sectional results remained stable for all subsequent models. Direct model. In addition to the paths of the stability model, the direct model included cross-lagged paths for testing whether 2 This is most likely due to a suppression effect, where the addition of other predictors to the former bivariate correlation will reduce error variance and thus, can result in significant coefficients. TEACHERS’ OCCUPATIONAL WELL-BEING 269 Table 1 Number of Items and Latent Correlations of All Scales Scales This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 1. 2. 3. 4. 5. 6. 7. 8. 9. Emotional exhaustion T1 Emotional exhaustion T2 Self-efficacy in classroom management T1 Self-efficacy in classroom management T2 Classroom disturbances T1 Classroom disturbances T2 Engagement T1 Engagement T2 Occupational commitment T3 No. of items 1 2 3 4 5 6 7 8 3 3 4 4 3 3 3 3 6 —.63ⴱⴱ ⫺.29ⴱⴱ ⫺.17ⴱ.28ⴱⴱ.18ⴱ ⫺.56ⴱⴱ ⫺.35ⴱⴱ ⫺.40ⴱ — ⫺.54ⴱⴱ ⫺.33ⴱ.54ⴱⴱ.39ⴱⴱ ⫺.47ⴱⴱ ⫺.58ⴱⴱ ⫺.48ⴱⴱ —.054 ⫺.26ⴱⴱ ⫺.33ⴱ.30ⴱⴱ.53ⴱⴱ.39 — ⫺.15 ⫺.34ⴱⴱ.32ⴱⴱ.35ⴱⴱ.24ⴱ —.77ⴱⴱ ⫺.22ⴱⴱ ⫺.05 ⫺.46ⴱ — ⫺.19ⴱ ⫺.24ⴱⴱ ⫺.32ⴱ —.58ⴱⴱ.36ⴱ —.58ⴱⴱ Note. T1 ⫽ Time 1; T2 ⫽ Time 2; T3 ⫽ Time 3. ⴱ Correlation is significant at p ⬍.05. ⴱⴱ Correlation is significant at p ⬍.01. prior self-efficacy in classroom management is related to later engagement, and whether earlier classroom disturbances are related to later emotional exhaustion. Model fit remained good and did not drop below our suggested cut-off values for model comparison with ␹2 ⫽ 745, df ⫽ 301, p ⬍.001, CFI ⫽.94, TLI ⫽.94, RMSEA ⫽.03. All autoregressive paths remained positive and significant. Cross-lagged paths revealed that classroom disturbances significantly predicted emotional exhaustion, ␤ ⫽.42, p ⬍.05, but not engagement, ␤ ⫽.08, p ⫽.524 (Hypothesis 1a). Further, self-efficacy in classroom management significantly predicted emotional exhaustion, with ␤ ⫽ ⫺.29, p ⬍.05 and engagement with ␤ ⫽.42, p ⬍.001 (Hypothesis 1b). Reverse model. To identify whether the relationships between the variables were stronger in a reverse matter (i.e., in contrast to the assumed direction of the dual processes), we also investigated whether emotional exhaustion and engagement at Time 1 predicted classroom disturbances and self-efficacy in classroom management at Time 2, respectively. Model fit was again good, with ␹2 ⫽ 755, df ⫽ 301, p ⬍.001, CFI ⫽.94, TLI ⫽.94, RMSEA ⫽.03, and was still acceptable with regard to cut-off values for model comparison. All autoregressive paths remained positive and significant, except for self-efficacy in classroom management, with ␤ ⫽ ⫺.20, p ⫽.514. In this model, however, only the cross-lagged path where engagement predicts self-efficacy in classroom management was significant, with ␤ ⫽.35, p ⬍.001. All other cross-lagged paths failed to reach significance, with engagement predicting classroom disturbances (␤ ⫽ ⫺.04, p ⫽.676), and emotional exhaustion predicting self-efficacy in classroom management (␤ ⫽ ⫺.03, p ⫽.727), and classroom disturbances (␤ ⫽.02, p ⫽.875), indicating that only prior engagement seems to influence later self-efficacy in classroom management. Reciprocal model. Having found evidence for the direct model, and for one reverse path, we aimed to examine possible reciprocal effects. Hence, in this model we allowed the crosslagged paths of the direct model, as well as the cross-lagged paths of the reverse model: that is, as opposed to our hypotheses. Thus, prior classroom disturbances and prior self-efficacy in classroom management both predicted later emotional exhaustion and engagement, while earlier emotional exhaustion and engagement were likewise related both to later self-efficacy in classroom management and to classroom disturbances (see Figure A(d) in the supplemental materials). Model fit remained similar to the other models, with ␹2 ⫽ 728, df ⫽ 297, p ⬍.001, CFI ⫽.94, TLI ⫽.94, RMSEA ⫽.03, and did not drop below our suggested cut-off values for model comparison. Again, all autoregressive paths, except for self-efficacy in classroom management, with ␤ ⫽ ⫺.11, p ⫽.651, remained positive and significant. Results revealed that the same cross-lagged paths were significant as in the direct and reverse models. Therefore, classroom disturbances significantly predicted emotional exhaustion, ␤ ⫽.40, p ⬍.05, but not engagement, ␤ ⫽.12, p ⫽.325. Further, self-efficacy in classroom management significantly predicted emotional exhaustion, with ␤ ⫽ ⫺.34, p ⬍.05, and engagement, with ␤ ⫽.47, p ⬍.001. In addition, engagement predicted self-efficacy in classroom management significantly, with ␤ ⫽.35, p ⬍.001 (Hypothesis 2b). All other cross-lagged paths failed to reach significance, with engagement predicting classroom disturbances with ␤ ⫽ ⫺.03, p ⫽.728, and emotional exhaustion predicting self-efficacy in classroom management with ␤ ⫽ ⫺.01, p ⫽.894, and classroom disturbances with ␤ ⫽ ⫺.04, p ⫽.699. Overall, our results indicate that a high level of earlier classroom disturbances is related to high levels of later emotional exhaustion. In addition, high levels of earlier self-efficacy in classroom management are related to high levels of later engagement, and vice versa. On the basis of these results, we based all further analysis on the reciprocal model, but, in the interest of parsimony, omitted the near zero nonsignificant reverse paths. Thus, our model included all direct paths, but only the prediction of self-efficacy at Time 2 through engagement at Time 1, as a reverse path.3 Hypothesis 3: Setting up the Core of the JD-R Model: Adding Latent Interactions In the third step, we then added the proposed latent interaction of self-efficacy in classroom management and classroom disturbances as a predictor for engagement and emotional exhaustion (see Figure 2). After adding the proposed latent interaction variable (Selfefficacy in Classroom Management ⫻ Classroom Disturbances) to the reciprocal model, the model fit was good with ␹2 ⫽ 884, df ⫽ 3 Adding these paths to the model did not alter the structure of results. This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 270 DICKE, STEBNER, LINNINGER, KUNTER, AND LEUTNER Figure 2. Final model including latent interactions and latent indirect effects with standardized regression coefficients. ⴱ p ⬍.05. ⴱⴱ p ⬍.001. ns ⫽ p ⬎.05. T1 ⫽ Time 1; T2 ⫽ Time 2; T3 ⫽ Time 3; Self-efficacy ⫽ Self-efficacy in classroom management; Disturbances ⫽ Classroom disruptions; Exhaustion ⫽ Emotional exhaustion; Disturbances ⫻ Self-efficacy ⫽ interaction of self-efficacy in classroom management and classroom disturbances; Commitment ⫽ occupational commitment. The bold arrows and bold regression coefficients indicate indirect effects, while a bold variable indicates an interaction effect. 372, p ⬍.001, CFI ⫽.94, TLI ⫽.93, RMSEA ⫽.03. All autoregressive paths were positive and significant. Further, results revealed the same structure as above with regard to cross-lagged paths. Moreover, both assumed that latent interaction effects were statistically significant. The interaction negatively predicted emotional exhaustion, with ␤ ⫽ ⫺.27, p ⬍.05, and engagement, positively and slightly more strongly, with ␤ ⫽.46, p ⬍.05.4 This indicates that higher self-efficacy in classroom management diminishes the effect of classroom disturbances on emotional exhaustion, in the sense of the proposed buffer effect in the JD-R model. The effect of self-efficacy in classroom management on engagement is stronger in the case of high classroom disturbances; this indicates a booster effect, as proposed by the JD-R model (see Figure 3). tional exhaustion, with ␤ ⫽ ⫺.27, p ⬍.05 and engagement positively, and slightly higher, with ␤ ⫽.46, p ⬍.05. Further, occupational commitment was directly predicted by engagement, with ␤ ⫽.41, p ⬍.05 as well as emotional exhaustion, with ␤ ⫽ ⫺.27, p ⬍.05. However, regarding the assumed indirect effects, only self-efficacy in classroom management significantly predicted occupational commitment via engagement, with ␤ ⫽.28, p ⬍.05. The prediction of occupational commitment by classroom disturbances through emotional exhaustion failed to reach significance with ␤ ⫽ ⫺.09, p ⫽.11 (for the final model see Figure 2). Adding covariates resulted in decreased fit, but did not change the structure of results (see supplemental materials). Hypothesis 4: Investigating the Prediction of Distant Occupational Outcomes: Adding Indirect Effects The present study aimed to test all assumptions of the JD-R model, including the proposed interaction and indirect effects simultaneously, in a sample of beginning teachers, on the basis of longitudinal data. Overall, our results provide strong evidence for (a) the causal effects proposed by the JD-R model (Bakker & Demerouti, 2014); and (b) the importance of the applied variables in the teacher stress process. Finally, we additionally included the latent variable occupational commitment to the model and, thus, the proposed indirect effects (see Figure 2) of resources (self-efficacy in classroom management) on occupational commitment via engagement, and of demands (classroom disturbances) on occupational commitment through strain (emotional exhaustion). Adding the indirect effect resulted in a decreased but acceptable model fit, with ␹2 ⫽ 1,438, df ⫽ 553, p ⬍.001, CFI ⫽.90, TLI ⫽.89, RMSEA ⫽.03, whereby the general structure of results compared with the model including only latent interactions, remained stable. Again, both assumed latent interaction effects to be statistically significant. The interaction negatively predicted emo- Discussion 4 Results based on the LMS approach (Klein & Moosbrugger, 2000) revealed the same structure of results. However, the LMS approach with TYPE ⫽ RANDOM and ALGORITHM ⫽ INTEGRATION does not allow for calculating commonly used fit statistics such as RMSEA, CFI, or TLI. This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. TEACHERS’ OCCUPATIONAL WELL-BEING 271 Figure 3. Interaction effects of self-efficacy in classroom management and classroom disturbances at Time 1 on (a) emotional exhaustion and (b) engagement at Time 2. The Dual Processes of the JD-R Model (Hypotheses 1 and 2) In a first step we replicated the dual processes of the JD-R model (Bakker & Demerouti, 2007, 2014) with our proposed variables. As hypothesized for the health impairment process (Hypothesis 1a), prior high job demands—in our case, classroom disturbances—increased emotional exhaustion up to 2 years later. These findings are in line with the findings of several researchers (e.g., Aloe et al., 2014; Skaalvik & Skaalvik, 2014) and reflect important evidence for the heavily discussed sequential ordering of job stress and strain with teachers (Brouwers & Tomic, 2000; see Dicke et al., 2014, for an overview). Contrary to our expectations and to the assumptions of triple-match principle (de Jonge et al., 2008), however, classroom disturbances did not reduce engagement. A possible explanation is that classroom disturbances and engagement are not based on matching dimensions with regard to triple-match principle (de Jonge et al., 2008). Further, Crawford et al. (2010) found that the effect of demands on engagement depended on whether the demands were perceived as a hindrance (negative effect) or a challenge (positive effect). Although it seems intuitive to assume classroom disturbances to be a hindrance, it could be interesting further investigating in future research whether they might also be perceived as a potential challenge. There might also be individual differences according to whether classroom disturbances are viewed as a hindrance or a challenge, that would result in an overall nonsignificant effect. Also, as expected in Hypothesis 1b (the motivational process), we found prior personal resources (self-efficacy in classroom management) to positively predict later engagement. Additionally, as assumed in Hypothesis 2, we not only found negative cross-sectional reciprocal effects (correlations) between self-efficacy in classroom management and classroom disturbances, as well as between engagement and emotional exhaustion, but also a longitudinal reciprocal effect, where earlier engagement also predicted later resources. This latter effect means that a high level of self-efficacy in classroom management predicts high job engagement, while high initial job engagement will also predict higher self-efficacy in classroom management. Many researchers (Llorens et al., 2007; Schaufeli et al., 2009; Xanthopoulou et al., 2009) have proposed such reciprocal effects to be part of the JD-R model for many occupations. Such a relationship can be characterized as a gain spiral (Halbesleben & Wheeler, 2015; Hobfoll, 2001; see Llorens et al., 2007, for an overview), where resource gain leads to enhanced motivation, which in turn leads to further resource gain. Within our design, we were not able to test this gain spiral explicitly, as it would be necessary to test for actual resource gains over time and preferably, to focus on shorter time intervals than in the present study (Salanova, Schaufeli, Xanthopoulou, & Bakker, 2010). We thus see our results as an indication that the occurrence of a gain spiral is likely, but cannot conclude that it actually occurred. It would be interesting in future research to apply this analysis to a similar context as the present study. The results also shed light on the dual meaning of self-efficacy as both resource and outcome (Holzberger, Phillip, & Kunter, 2013; Spurk & Abele, 2014). From an applied perspective, selfefficacy is most often seen as a resource that is an important predictor for several aspects of occupational outcomes, such as job satisfaction (Caprara, Barbaranelli, Borgogni, & Steca, 2003) and job performance (Cherian & Jacob, 2013) and engagement (Skaalvik & Skaalvik, 2014). In our study, we found that self-efficacy could, in addition, potentially be altered through enhancing (in this case) engagement. This is an important finding, as interventions to enhance occupational well-being could target self-efficacy in classroom management or also engagement, taking advantage of a most likely indirect effect of self-efficacy in classroom management and a direct effect of engagement on occupational well-being (Wang, Lu, & Siu, 2015). Overall, our results were largely in line with recent metaanalytic findings on several occupations, noting where resources were positively related to engagement and negatively to strain, while job demands are positively related to burnout (Crawford et al., 2010; Nahrgang et al., 2011). Adding the Interaction of Job Demands and Resources: The Core of the JD-R Model (Hypothesis 3) In our next step we successfully added the proposed interaction effect of job demands and resources (Bakker & Demerouti, 2007; Bakker et al., 2007) to our model; the results were as hypothesized and additionally, were in line with triple-match principle (de Jonge This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 272 DICKE, STEBNER, LINNINGER, KUNTER, AND LEUTNER et al., 2008). First, having a high level of self-efficacy (personal resource) buffered the health impairment process—that is, weakened the link between classroom disturbances (job demand) and emotional exhaustion. This finding adds further support to the moderating role of self-efficacy in the relationship between demands and strain (Dicke et al., 2014; Jex & Bliese, 1999) and is in line with findings on buffer effects in other occupations, such as mental health care (van Woerkom et al., 2016). Jex and Bliese (1999) found self-efficacy and collective efficacy to moderate the stressor-strain relationship in a sample of U.S. Army soldiers. The variables of this study had largely been assessed with instruments targeting the army-specific environment. For our sample of beginning teachers there has been extensive research on the positive effects of employing classroom management strategies on student behaviors and student discipline (for an overview, see Emmer & Stough, 2001). Further, research on the benefits of having professional knowledge (including knowledge of classroom management), which can be imparted to beginning teachers during their studies (Terhart et al., 2012), shows to be effective in buffering an increase in emotional exhaustion of preservice teachers (Dicke et al., 2015; Klusmann et al., 2012). However, our results reveal that emotional exhaustion is predicted by classroom disruptions, depending on the level of self-efficacy in classroom management. From a practical perspective, this implies that the effectiveness of trying to reduce emotional exhaustion through decreasing classroom disruptions depends on the level of self-efficacy in classroom management. This would imply that preservice teacher education or training programs need to be more closely matched to beginning teachers’ needs to ensure a well-balanced return on investment (Dewe, 2004). Second, the positive effect of self-efficacy (personal resource) on engagement was especially strong in case of high perceived classroom disturbances (job demand), providing evidence for the assumed boosting effect (Bakker et al., 2010) that was also reported in a sample of dentists (Hakanen et al., 2005). Bakker et al. (2010) predicted task enjoyment and organizational commitment through the interactions of a high number of job resources (e.g., skill utilization, learning opportunities, and autonomy) and job demands (workload and emotional demands). They showed that in a sample of mixed employees, 15 of the 16 hypothesized interactions were significant for task enjoyment and 13 of the 16 interactions were significant for organizational commitment. Hakanen et al. (2005) then, similarly to the current study, conclude that resources in dentistry gain salience and relevance under stressful conditions. These studies and the present study are thus in line with COR theory (Hobfoll, 2001) as they indicate that as long as there are enough resources available, high demands can be perceived as being positive, that is, a challenge, enabling the employee to thrive (Bakker et al., 2010). This again highlights the importance for beginning teachers to acquire resources such as self-efficacy in classroom management early on. Thus, being able to draw on these resources when necessary will not only protect them from developing strain, but will also increase their engagement. Both of these findings, the boosting and the buffering effect, highlight the important role of self-efficacy in classroom management in respect of stress development. On the one hand, selfefficacy will prevent job demands from leading to high job strain. On the other hand, self-efficacy in classroom management becomes increasingly important as a predictor of engagement in the case of high job demands. It is of course important to keep in mind that empirically, although relying on longitudinal data, the analysis of the present study is based on path modeling, and predictions cannot be interpreted as causal relationships. Although the inclusion of baseline measures and various covariates is a step toward approximating a quasi-experimental situation, we cannot rule out the possibility of other omitted variables. However, other research based on (quasi-) experimental designs shows further evidence of the assumed processes stated in the JD-R model (for an overview see Bakker & Demerouti, 2014; Dicke, Elling, Schmeck, & Leutner, 2015; Peterson, Luthans, Avolio, Walumbwa, & Zhang, 2011). A Distant Occupational Outcome: Adding the Indirect Effects (Hypothesis 4) Last, we added occupational commitment as a more distal occupation outcome to our model, enabling us to test for indirect effects of job demands via emotional exhaustion, and job resources via engagement. Despite occupational commitment being significantly directly predicted by both engagement (positively) and emotional exhaustion (negatively), as proposed by Bakker, Demerouti, and Verbeke (2004; Bakker & Bal, 2010), our hypothesis regarding indirect effects was only partly confirmed. Hence, as opposed to findings by Hakanen et al. (2006), who found significant indirect effects of resources and job demands on occupational outcomes, we only found a significant positive indirect effect of self-efficacy in classroom management on occupational commitment through engagement, in line with findings by Wang et al. (2015) with a sample of Chinese employees. A possible reason is that our data were collected over a period of 2 and a half to 3 years, while Hakanen et al. (2006) relied on cross-sectional data, and Bakker et al. (2003) investigated effects over 1 year only. Substantively, as well as practically, it can be considered promising that the motivational process seems to have a stronger influence on occupational commitment than does the health impairment process. There has been an ongoing trend in psychology of focusing on fostering occupational well-being, rather than on pathology: This has been termed positive psychology (Seligman & Csikszentmihalyi, 2014). It is within the scope of the positive psychology approach to foster the motivational aspects of work, instead of focusing on interventions dealing with and thereby putting focus on, the negative aspects. Overall, our results provide strong evidence for the JD-R model. First, we successfully replicated the health impairment process, as well as the motivation process, where high demands lead to higher stress and high levels of resources lead to more engagement, respectively. Second, we found a significant interaction of demands and resources on strain (buffering effect) and engagement (boosting effect). Third, our data revealed cross-sectional reciprocal effects between demands and resources and between strain and engagement, as well as longitudinal reciprocal effects of resources and engagement. Finally, we also provide sound evidence for the indirect effects of resources and demands on occupational outcomes, such as commitment, via engagement and strain, respectively. TEACHERS’ OCCUPATIONAL WELL-BEING This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Strengths, Limitations, and Directions for Further Research In our model we were able to control for all proposed relationships of the JD-R model, while testing all proposed effects. Thus, we took possible interfering influences into account, thereby providing particularly strong evidence for both theoretical and empirical assumptions of the model. The application of advanced methods such as latent interactions was made feasible by the availability of a large, representative sample of teacher candidates at Time 1 and a low dropout at Time 2. Despite these strengths, the present study has some limitations. Due to a large initial test battery we utilized economical 4-point response scales. This can lead to unreliability due to the small number of response options, particularly when using scaled scores. In the present study, however, we used latent variable modeling and thus, accounted for measurement error (unreliability) in all our models. Further, although it has been argued that self-reports are appropriate in assessing inner psychological processes such as emotional exhaustion and self-efficacy (Howard, 1994), it can be an advantage to include other, more objective measures (Dewe & Trenberth, 2004; Lazarus, 2000; Schmitt, 1994). Hence, it would be interesting to include other, objective criteria reflecting our variables of interest. For example, multiple measurement methods, including self-reported classroom disturbances but also other sources, such as students’ perception of disturbances, could be utilized to validate collected data. Nevertheless, for the Disturbance Scale applied in the current study, Kunter and Baumert (2006) compared teachers’ self-report with ratings from the students taught by their respective teachers, and showed that both perspectives converge. For exhaustion, physiological information such as cortisol levels, and for engagement, student, colleague and supervisor ratings, could provide additional insights. Additionally, other variables reflecting not only job demands (e.g., time pressure; Skaalvik & Skaalvik, 2010, 2011), and resources, for example, social support (Van der Doef & Maes, 1999), but also occupational outcomes, could be introduced in the model. Furthermore, by measuring occupational commitment solely at Time 3, we were not able to control for prior commitment or to model a fully cross-sectional model of all variables. As described in the introduction, the beginning teachers in the present study are not yet full employees or fully fledged members of the occupation, with full responsibility, authorities, and status. The teacher induction period is rather a phase of being in-between: partly a teacher in training and partly still a student (who still is subject to being graded and who needs to undertake several exams). Thus, assessing occupational commitment during the teacher induction phase might not reflect the same occupational commitment as that experienced after actually entering the occupation as a full member. Future research should examine teacher commitment at different career stages, to investigate how occupational commitment changes not only in intensity or patterns (for an analysis of such patterns see Solinger et al., 2013), but also in meaning across these stages. Although we were able to rely on a large teacher sample, we only included German teachers, although there are great differences in teacher induction programs internationally. Teachers in Germany might report higher levels of well- 273 being compared with teachers of other nationalities, due to the long induction phase, which can enable a smoother transition to the occupation. However, we consider our results potentially informative for research on occupational outcomes in general, as the results found in the current sample reflect findings from other occupations in many countries (e.g., Cherian & Jacob, 2013; Crawford et al., 2010; Hakanen et al., 2005; Jex & Bliese, 1999; Nahrgang et al., 2011; van Woerkom et al., 2016; Wang et al., 2015; Xanthopoulou et al., 2009). Nevertheless, future research should investigate whether the proposed processes are generalizable to different cultural backgrounds. The method of testing the model as a whole should, additionally, be applied to other occupations for further validation of core variables linked to these professions as well as of the overall generalizability of the model. Conclusion While the assumptions of the JD-R model have been successfully tested for various settings and occupational groups, so far this has been done in a piecewise manner. Building on that research, the present study tested all assumptions simultaneously: that is, in one model in a sample of beginning teachers. Thereby, strong evidence for the complex interplay of variables that have been found to be important within the (beginning) teachers’ stress development process, namely, classroom disturbances, selfefficacy in classroom management, emotional exhaustion (strain), engagement, and occupational commitment (organizational outcome) are presented. Further, the results stress the important role of self-efficacy in classroom management as an important resource that, in case of high classroom disturbances, will not only buffer the strain-enhancing effects, but also boost engagement. Thus, by applying the model to those variables most relevant to the sample, the study provides strong evidence for the validity of the JD-R model. References Aloe, A. M., Amo, L. C., & Shanahan, M. E. (2014). Classroom management self-efficacy and burnout: A multivariate meta-analysis. Educational Psychology Review, 26, 101–126. http://dx.doi.org/10.1007/ s10648-013-9244-0 Bakker, A. B., & Bal, M. P. (2010). Weekly work engagement and performance: A study among starting teachers. Journal of Occupational and Organizational Psychology, 83, 189 –206. http://dx.doi.org/10.1348/ 096317909X402596 Bakker, A. B., & Demerouti, E. (2007). The job demands-resources model: State of the art. Journal of Managerial Psychology, 22, 309 –328. http://dx.doi.org/10.1108/02683940710733115 Bakker, A. B., & Demerouti, E. (2014). Job demands–resources theory. Well-Being, 3, 1–28. Bakker, A. B., Demerouti, E., De Boer, E., & Schaufeli, W. B. (2003). Job demands and job resources as predictors of absence duration and frequency. Journal of Vocational Behavior, 62, 341–356. http://dx.doi.org/ 10.1016/S0001-8791(02)00030-1 Bakker, A. B., Demerouti, E., & Euwema, M. C. (2005). Job resources buffer the impact of job demands on burnout. Journal of Occupational Health Psychology, 10, 170 –180. http://dx.doi.org/10.1037/1076-8998.10.2.170 Bakker, A. B., Demerouti, E., & Verbeke, W. (2004). Using the job demands-resources model to predict burnout and performance. Human Resource Management, 43, 83–104. http://dx.doi.org/10.1002/hrm.20004 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 274 DICKE, STEBNER, LINNINGER, KUNTER, AND LEUTNER Bakker, A. B., Hakanen, J. J., Demerouti, E., & Xanthopoulou, D. (2007). Job resources boost work engagement, particularly when job demands are high. Journal of Educational Psychology, 99, 274 –284. http://dx.doi.org/10.1037/0022-0663.99.2.274 Bakker, A. B., & Sanz-Vergel, A. I. (2013). Weekly work engagement and flourishing: The role of hindrance and challenge job demands. Journal of Vocational Behavior, 83, 397– 409. http://dx.doi.org/10.1016/j.jvb.2013.06.008 Bakker, A. B., Sanz-Vergel, A. I., & Kuntze, J. (2015). Student engagement and performance: A weekly diary study on the role of openness. Motivation and Emotion, 39, 49 – 62. http://dx.doi.org/10.1007/s11031014-9422-5 Bakker, A. B., van Veldhoven, M., & Xanthopoulou, D. (2010). Beyond the demand-control model. Journal of Personnel Psychology, 9, 3–16. http://dx.doi.org/10.1027/1866-5888/a000006 Bandura, A. (1997). Self-efficacy: The exercise of control. New York, NY: Freeman. Barbier, M., Dardenne, B., & Hansez, I. (2013). A longitudinal test of the Job Demands–Resources model using perceived stigma and social identity. European Journal of Work and Organizational Psychology, 22, 532–546. http://dx.doi.org/10.1080/1359432X.2012.698056 Baumert, J., Blum, W., Brunner, M., Dubberke, T., Jordan, A., Klusmann, U., & Yi-Miau, T. (2008). Professionswissen von Lehrkräften, kognitiv aktivierender Mathematikunterricht und die Entwicklung von mathematischer Kompetenz (COACTIV): Dokumentation der Erhebungsinstrumente [Professional knowledge of teachers, cognitive activating math lessons and the development of mathematical competence (COACTIV): Documentation of the scales]. Berlin, Germany: Max-Planck-Institut für Bildungsforschung. Beijaard, D., Meijer, P. C., & Verloop, N. (2004). Reconsidering research on teachers’ professional identity. Teaching and Teacher Education, 20, 107–128. http://dx.doi.org/10.1016/j.tate.2003.07.001 Boyd, C. M., Bakker, A. B., Pignata, S., Winefield, A. H., Gillespie, N., & Stough, C. (2011). A longitudinal test of the job demands-resources model among Australian university academics. Applied Psychology: An International Review, 60, 112–140. Brauchli, R., Jenny, G. J., Füllemann, D., & Bauer, G. F. (2015). Towards a job demands-resources health model: Empirical testing with generalizable indicators of job demands, job resources, and comprehensive health outcomes. BioMed Research International, 2015. Article ID 959621. http://dx.doi.org/10.1155/2015/959621 Brenninkmeijer, V., van Yperen, N. W., & Buunk, B. P. (2001). I am not a better teacher, but others are doing worse: Burnout and perceptions of superiority among teachers. Social Psychology of Education, 4, 259 – 274. http://dx.doi.org/10.1023/A:1011376503306 Brough, P., Timms, C., Siu, O. L., Kalliath, T., O’Driscoll, M., Sit, C.,... Lu, C. Q. (2013). Validation of the job demands-resources model in cross-national samples: Cross-sectional and longitudinal predictions of psychological strain and work engagement. Human Relations, 66, 1311– 1335. http://dx.doi.org/10.1177/0018726712472915 Brouwers, A., & Tomic, W. (2000). A longitudinal study of teacher burnout and perceived self-efficacy in classroom management. Teaching and Teacher Education, 16, 239 –253. http://dx.doi.org/10.1016/S0742051X(99)00057-8 Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A. Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 136 –162). Beverly Hills, CA: SAGE. Caprara, G. V., Barbaranelli, C., Borgogni, L., & Steca, P. (2003). Efficacy beliefs as determinants of teachers’ job satisfaction. Journal of Educational Psychology, 95, 821– 832. http://dx.doi.org/10.1037/0022-0663.95.4.821 Chen, F. F. (2007). Sensitivity of goodness of fit indexes to lack of measurement invariance. Structural Equation Modeling, 14, 464 –504. http://dx.doi.org/10.1080/10705510701301834 Cherian, J., & Jacob, J. (2013). Impact of self-efficacy on motivation and performance of employees. International Journal of Business and Management, 8, 80 – 88. http://dx.doi.org/10.5539/ijbm.v8n14p80 Clausen, T., Burr, H., & Borg, V. (2014). Do psychosocial job demands and job resources predict long-term sickness absence? An analysis of register-based outcomes using pooled data on 39,408 individuals in four occupational groups. International Archives of Occupational and Environmental Health, 87, 909 –917. http://dx.doi.org/10.1007/s00420-0140936-7 Crawford, E. R., Lepine, J. A., & Rich, B. L. (2010). Linking job demands and resources to employee engagement and burnout: A theoretical extension and meta-analytic test. Journal of Applied Psychology, 95, 834 – 848. http://dx.doi.org/10.1037/a0019364 Cropanzano, R., Rupp, D. E., & Byrne, Z. S. (2003). The relationship of emotional exhaustion to work attitudes, job performance, and organizational citizenship behaviors. Journal of Applied Psychology, 88, 160 – 169. http://dx.doi.org/10.1037/0021-9010.88.1.160 de Jonge, J., & Dormann, C. (2006). Stressors, resources, and strain at work: A longitudinal test of the triple-match principle. Journal of Applied Psychology, 91, 1359 –1374. http://dx.doi.org/10.1037/00219010.91.5.1359 de Jonge, J., Dormann, C., & van den Tooren, M. (2008). The demandinduced strain compensation model: Renewed theoretical considerations and empirical evidence. In K. Näswall, J. Hellgren, & M. Sverke (Eds.), The individual in the changing working life (pp. 67– 87). Oxford, UK: Cambridge University Press. http://dx.doi.org/10.1017/CBO9780 511490064.004 Demerouti, E., Bakker, A. B., Nachreiner, F., & Schaufeli, W. B. (2001). The job demands-resources model of burnout. Journal of Applied Psychology, 86, 499 –512. http://dx.doi.org/10.1037/0021-9010.86.3.499 Demerouti, E., Le Blanc, P. M., Bakker, A. B., Schaufeli, W. B., & Hox, J. (2009). Present but sick: A three-wave study on job demands, presenteeism and burnout. Career Development International, 14, 50 – 68. http://dx.doi.org/10.1108/13620430910933574 Demerouti, E., Mostert, K., & Bakker, A. B. (2010). Burnout and work engagement: A thorough investigation of the independency of both constructs. Journal of Occupational Health Psychology, 15, 209 –222. http://dx.doi.org/10.1037/a0019408 Dewe, P. (2004). Work stress and coping: Theory, research and practice. British Journal of Guidance & Counselling, 32, 139 –142. http://dx.doi.org/10.1080/03069880410001693138 Dewe, P., & Trenberth, L. (2004). Work stress and coping: Drawing together research and practice. British Journal of Guidance & Counselling, 32, 143–156. http://dx.doi.org/10.1080/03069880410001693138 Dicke, T., Elling, J., Schmeck, A., & Leutner, D. (2015). Reducing reality shock: The effects of classroom management skills training on beginning teachers. Teaching and Teacher Education, 48, 1–12. http://dx.doi.org/10.1016/j.tate.2015.01.013 Dicke, T., Holzberger, D., Kunina–Habenicht, O., Linninger, C., SchulzeStocker, F., Seidel, T.,... Kunter, M. (2016). Doppelter Praxisschock auf dem Weg ins Lehramt? Verlauf und potentielle Einflussfaktoren emotionaler Erschöpfung während des Vorbereitungsdienstes und nach dem Berufseintritt [Double reality shock when becoming a teacher? The development and potential influencing factors of emotional exhaustion during teacher induction time, and after having worked as fully licensed teacher]. Psychologie in Erziehung und Unterricht, 63, 244 –257. Dicke, T., Parker, P. D., Holzberger, D., Kunter, M., & Leutner, D. (2015). Beginning teachers’ efficacy and emotional exhaustion: Latent changes, reciprocity, and the influence of professional knowledge. Contemporary Educational Psychology, 41, 62–72. http://dx.doi.org/10.1016/j.cedpsych.2014.11.003 Dicke, T., Parker, P. D., Marsh, H. W., Kunter, M., Schmeck, A., & Leutner, D. (2014). Self-efficacy in classroom management, classroom disturbances, and emotional exhaustion: A moderated mediation analy- This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. TEACHERS’ OCCUPATIONAL WELL-BEING sis of teacher candidates. Journal of Educational Psychology, 106, 569 –583. http://dx.doi.org/10.1037/a0035504 Duncan, S. C., Duncan, T. E., & Strycker, L. A. (2006). Alcohol use from ages 9 to 16: A cohort-sequential latent growth model. Drug and Alcohol Dependence, 81, 71– 81. http://dx.doi.org/10.1016/j.drugalcdep.2005.06.001 Emmer, E. T., & Stough, L. M. (2001). Classroom management: A critical part of educational psychology, with implications for teacher education. Educational Psychologist, 36, 103–112. http://dx.doi.org/10.1207/ S15326985EP3602_5 Enders, C. K. (2010). Applied missing data analysis. New York, NY: Guilford Press. Enzmann, D., & Kleiber, D. (1989). Helfer-Leiden: Stress und Burnout in psychosozialen Berufen [Helpers suffering: Stress and burnout in psychosocial work areas]. Heidelberg, Germany: Roland Asanger Verlag. Federici, R. A., & Skaalvik, E. M. (2012). Principal self-efficacy: Relations with burnout, job satisfaction and motivation to quit. Social Psychology of Education, 15, 295–320. http://dx.doi.org/10.1007/s11218-0129183-5 Friedman, I. A. (2000). Burnout in teachers: Shattered dreams of impeccable professional performance. Journal of Clinical Psychology, 56, 595– 606. http://dx.doi.org/10.1002/(SICI)1097-4679(200005)56: 5⬍595::AID-JCLP2⬎3.0.CO;2-Q Friedman, I. (2006). Classroom management and teacher stress and burnout. In C. M. Evertson & C. S. Weinstein (Eds.), Handbook of classroom management: Research, practice, and contemporary issues (pp. 925– 944). Mahwah, NJ: Erlbaum. Frins, W., van Ruysseveldt, J., van Dam, K., & van den Bossche, S. N. (2016). Older employees’ desired retirement age: A JD-R perspective. Journal of Managerial Psychology, 31, 34 – 49. http://dx.doi.org/10.1108/JMP-05-2013-0133 Garrick, A., Mak, A. S., Cathcart, S., Winwood, P. C., Bakker, A. B., & Lushington, K. (2014). Psychosocial safety climate moderating the effects of daily job demands and recovery on fatigue and work engagement. Journal of Occupational and Organizational Psychology, 87, 694 –714. http://dx.doi.org/10.1111/joop.12069 Graham, J. W. (2012). Missing data: Analysis and design. New York, NY: Springer Science & Business Media. http://dx.doi.org/10.1007/978-14614-4018-5 Graham, J. W., Taylor, B. J., Olchowski, A. E., & Cumsille, P. E. (2006). Planned missing data designs in psychological research. Psychological Methods, 11, 323–343. http://dx.doi.org/10.1037/1082-989X.11.4.323 Hackman, J. R., & Oldham, G. R. (1980). Work redesign. Reading, MA: Addison Wesley. Hakanen, J. J., Bakker, A. B., & Demerouti, E. (2005). How dentists cope with their job demands and stay engaged: The moderating role of job resources. European Journal of Oral Sciences, 113, 479 – 487. http://dx.doi.org/10.1111/j.1600-0722.2005.00250.x Hakanen, J. J., Bakker, A. B., & Schaufeli, W. B. (2006). Burnout and work engagement among teachers. Journal of School Psychology, 43, 495–513. http://dx.doi.org/10.1016/j.jsp.2005.11.001 Hakanen, J. J., Perhoniemi, R., & Toppinen-Tanner, S. (2008). Positive gain spirals at work: From job resources to work engagement, personal initiative and work-unit innovativeness. Journal of Vocational Behavior, 73, 78 –91. http://dx.doi.org/10.1016/j.jvb.2008.01.003 Halbesleben, J. R., & Wheeler, A. R. (2015). To invest or not? The role of coworker support and trust in daily reciprocal gain spirals of helping behavior. Journal of Management, 41, 1628 –1650. http://dx.doi.org/10.1177/0149206312455246 Hobfoll, S. E. (1989). Conservation of resources. A new attempt at conceptualizing stress. American Psychologist, 44, 513–524. http://dx.doi.org/10.1037/0003-066X.44.3.513 Hobfoll, S. E. (2001). The influence of culture, community, and the nested self in the stress process: Advancing conservation of resources theory. 275 Applied Psychology, 50, 337–370. http://dx.doi.org/10.1111/1464-0597.00062 Holzberger, D., Philipp, A., & Kunter, M. (2013). How teachers’ selfefficacy is related to instructional quality: A longitudinal analysis. Journal of Educational Psychology, 105, 774 –786. http://dx.doi.org/10.1037/a0032198 Hong, J. Y. (2010). Pre-service and beginning teachers’ professional identity and its relation to dropping out of the profession. Teaching and Teacher Education, 26, 1530 –1543. http://dx.doi.org/10.1016/j.tate.2010.06.003 Howard, G. (1994). Why do people say nasty things about self-reports? Journal of Organizational Behavior, 15, 399 – 404. http://dx.doi.org/10.1002/job.4030150505 Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55. http://dx.doi.org/10.1080/ 10705519909540118 Hu, Q., Schaufeli, W. B., & Taris, T. W. (2013). Does equity mediate the effects of job demands and job resources on work outcomes? An extension of the job demands-resources model. Career Development International, 18, 357–376. http://dx.doi.org/10.1108/CDI-12-20120126 Huberman, M. A. (1989). The professional life cycle of teachers. Teachers College Record, 91, 31–57. Ingersoll, R. M. (2012). Beginning teacher induction: What the data tell us. Phi Delta Kappan, 93, 47–51. http://dx.doi.org/10.1177/ 003172171209300811 Jepson, E., & Forrest, S. (2006). Individual contributory factors in teacher stress: The role of achievement striving and occupational commitment. The British Journal of Educational Psychology, 76, 183–197. http://dx.doi.org/10.1348/000709905X37299 Jex, S. M., & Bliese, P. D. (1999). Efficacy beliefs as a moderator of the impact of work-related stressors: A multilevel study. Journal of Applied Psychology, 84, 349 –361. http://dx.doi.org/10.1037/0021-9010.84.3.349 Jones, V. (2006). How do teachers learn to be effective classroom managers. In C. M. Evertson & C. S. Weinstein (Eds.), Handbook of classroom management: Research, practice, and contemporary issues (pp. 887–908). Mahwah, NJ: Erlbaum. Karasek, R. A., Jr. (1979). Job demands, job decision latitude, and mental strain: Implications for job redesign. Administrative Science Quarterly, 24, 285–308. http://dx.doi.org/10.2307/2392498 Klassen, R. M., Bong, M., Usher, E. L., Chong, W. H., Huan, V. S., Wong, I. Y. F., & Georgiou, T. (2009). Exploring the validity of the Teachers’ Self-Efficacy Scale in five countries. Contemporary Educational Psychology, 34, 67–76. http://dx.doi.org/10.1016/j.cedpsych.2008.08.001 Klassen, R. M., & Chiu, M. M. (2010). Effects on teachers’ self-efficacy and job satisfaction: Teacher gender, years of experience, and job stress. Journal of Educational Psychology, 102, 741–756. http://dx.doi.org/10.1037/a0019237 Klassen, R. M., & Chiu, M. M. (2011). The occupational commitment and intention to quit of practicing and pre-service teachers: Influence of self-efficacy, job stress, and teaching context. Contemporary Educational Psychology, 36, 114 –129. http://dx.doi.org/10.1016/j.cedpsych.2011.01.002 Klein, A., & Moosbrugger, H. (2000). Maximum likelihood estimation of latent interaction effects with the LMS method. Psychometrika, 65, 457– 474. http://dx.doi.org/10.1007/BF02296338 Klusmann, U., Kunter, M., Trautwein, U., & Baumert, J. (2006). Lehrerbelastung und Unterrichtsqualität aus der Perspektive von Lehrenden und Lernenden [Teachers’ and students’ perspective of teacher stress and lesson quality]. Zeitschrift fur Padagogische Psychologie, 20, 161– 173. http://dx.doi.org/10.1024/1010-0652.20.3.161 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 276 DICKE, STEBNER, LINNINGER, KUNTER, AND LEUTNER Klusmann, U., Kunter, M., Trautwein, U., Lüdtke, O., & Baumert, J. (2008). Engagement and emotional exhaustion in teachers: Does the school context make a difference? Applied Psychology, 57(Suppl. 1), 127–151. http://dx.doi.org/10.1111/j.1464-0597.2008.00358.x Klusmann, U., Kunter, M., Voss, T., & Baumert, J. (2012). Berufliche Beanspruchung angehender Lehrkräfte: Die Effekte von Persönlichkeit, pädagogischer Vorerfahrung und professioneller Kompetenz [Occupational stress of beginning teachers: The effects of personality, pedagogical experience and professional competence]. Zeitschrift fur Padagogische Psychologie, 26, 275–290. http://dx.doi.org/10.1024/1010-0652/ a000078 Kokkinos, C. M. (2007). Job stressors, personality and burnout in primary school teachers. The British Journal of Educational Psychology, 77, 229 –243. http://dx.doi.org/10.1348/000709905X90344 Kunter, M., & Baumert, J. (2006). Who is the expert? Construct and criteria validity of student and teacher ratings of instruction. Learning Environments Research, 9, 231–251. http://dx.doi.org/10.1007/s10984006-9015-7 Kunter, M., & Linninger, C. Schulze–Stocker, F., Kunina–Habenicht, O., & Lohse–Bossenz, H. (2013). Evaluation des reformierten Vorbereitungsdienstes in Nordrhein–Westfalen [An evaluation of the reformed teacher induction in North Rhine Westphalia]. Retrieved from https:// zfsl.nrw.de/docs/bp/Ministerium/Presse/Pressekonferenzen/2014/ 2014_05_12-Vorbereitungsdienst/03d-BilWiss-Gesamtbericht.pdf Lazarus, R. S. (2000). Toward better research on stress and coping. American Psychologist, 55, 665– 673. http://dx.doi.org/10.1037/0003066X.55.6.665 Lazarus, R. S., & Folkman, S. (1984). Stress, appraisal, and coping. New York, NY: Springer. Llorens, S., Schaufeli, W., Bakker, A., & Salanova, M. (2007). Does a positive gain spiral of resources, efficacy beliefs and engagement exist? Computers in Human Behavior, 23, 825– 841. http://dx.doi.org/10.1016/ j.chb.2004.11.012 Luchman, J. N., & González-Morales, M. G. (2013). Demands, control, and support: A meta-analytic review of work characteristics interrelationships. Journal of Occupational Health Psychology, 18, 37–52. http:// dx.doi.org/10.1037/a0030541 Maslach, C. (1999). Progress in understanding teacher burnout. In R. Vandenberghe & A. M. Huberman (Eds.), Understanding and preventing teacher stress: A sourcebook of international research and practice (pp. 211–222). Cambridge, UK: Cambridge University Press. http://dx.doi.org/10.1017/CBO9780511527784.014 Maslach, C., Jackson, S. E., & Leiter, M. P. (1996). Maslach Burnout Inventory: Manual (3rd ed.). Palo Alto, CA: Consulting Psychologists Press. McDonald, R. P. (1999). Test theory: A unified treatment. Mahwah, NJ: Erlbaum. McDonald, R. P., & Marsh, H. W. (1990). Choosing a multivariate model: Noncentrality and goodness-of-fit. Psychological Bulletin, 107, 247– 255. http://dx.doi.org/10.1037/0033-2909.107.2.247 Meyer, J. P., Allen, N. J., & Smith, C. A. (1993). Commitment to organizations and occupations: Extension and test of a three-component conceptualization. Journal of Applied Psychology, 78, 538 –551. http:// dx.doi.org/10.1037/0021-9010.78.4.538 Meyer, J. P., Becker, T. E., & Vandenberghe, C. (2004). Employee commitment and motivation: A conceptual analysis and integrative model. Journal of Applied Psychology, 89, 991–1007. http://dx.doi.org/ 10.1037/0021-9010.89.6.991 Nahrgang, J. D., Morgeson, F. P., & Hofmann, D. A. (2011). Safety at work: A meta-analytic investigation of the link between job demands, job resources, burnout, engagement, and safety outcomes. Journal of Applied Psychology, 96, 71–94. http://dx.doi.org/10.1037/a0021484 O’Neill, S. C., & Stephenson, J. (2011). The measurement of classroom management self-efficacy: A review of measurement instrument devel- opment and influences. Educational Psychology, 31, 261–299. http://dx.doi.org/10.1080/01443410.2010.545344 Parker, P. D., Lüdtke, O., Trautwein, U., & Roberts, B. W. (2012). Personality and relationship quality during the transition from high school to early adulthood. Journal of Personality, 80, 1061–1089. http:// dx.doi.org/10.1111/j.1467-6494.2012.00766.x Parker, P. D., Marsh, H. W., Morin, A. J., Seaton, M., & Van Zanden, B. (2015). If one goes up the other must come down: Examining ipsative relationships between math and English self-concept trajectories across high school. The British Journal of Educational Psychology, 85, 172– 191. http://dx.doi.org/10.1111/bjep.12050 Peterson, S. J., Luthans, F., Avolio, B. J., Walumbwa, F. O., & Zhang, Z. (2011). Psychological capital and employee performance: A latent growth modeling approach. Personnel Psychology, 64, 427– 450. http:// dx.doi.org/10.1111/j.1744-6570.2011.01215.x Pogodzinski, B. (2012). Socialization of novice teachers. Journal of School Leadership, 22, 982–1023. Rieg, S. A., Paquette, K. R., & Chen, Y. (2007). Coping with stress: An investigation of novice teachers’ stressors in the elementary classroom. Education, 128, 211–226. Salanova, M., Bakker, A. B., & Llorens, S. (2006). Flow at work: Evidence for an upward spiral of personal and organizational resources. Journal of Happiness Studies, 7, 1–22. http://dx.doi.org/10.1007/s10902-0058854-8 Salanova, M., Schaufeli, W. B., Xanthopoulou, D., & Bakker, A. B. (2010). The gain spiral of resources and work engagement: Sustaining a positive worklife. In A. B. Bakker & M. P. Leiter (Eds.), Work engagement: A handbook of essential theory and research (pp. 118 –131). New York, NY: Psychology Press. Schaufeli, W. B., & Bakker, A. B. (2003). The Utrecht Work Engagement Scale: Test manual. Utrecht, the Netherlands: Department of Social and Organizational Psychology. Schaufeli, W. B., Bakker, A. B., & Van Rhenen, W. (2009). How changes in job demands and resources predict burnout, work engagement, and sickness absenteeism. Journal of Organizational Behavior, 30, 893–917. http://dx.doi.org/10.1002/job.595 Schaufeli, W. B., & Enzmann, D. (1998). The burnout companion to study and practice. Philadelphia, PA: Taylor and Francis. Schaufeli, W. B., & Taris, T. W. (2014). A critical review of the job demands-resources model: Implications for improving work and health. Bridging Occupational, Organizational and Public Health, 2014, 43– 68. http://dx.doi.org/10.1007/978-94-007-5640-3_4 Schmitt, N. (1994). Method bias: The importance of theory and measurement. Journal of Organizational Behavior, 15, 393–398. http://dx.doi.org/10.1002/job.4030150504 Schwarzer, R., & Hallum, S. (2008). Perceived teacher self-efficacy as a predictor of job stress and burnout: Mediation Analyses. Applied Psychology, 57, 152–171. http://dx.doi.org/10.1111/j.1464-0597.2008.00359.x Seligman, M. E., & Csikszentmihalyi, M. (2014). Positive psychology: An introduction. Rotterdam, the Netherlands: Springer. Siegrist, J. (1996). Adverse health effects of high-effort/low-reward conditions. Journal of Occupational Health Psychology, 1, 27– 41. http:// dx.doi.org/10.1037/1076-8998.1.1.27 Simbula, S., Guglielmi, D., & Schaufeli, W. B. (2011). A three-wave study of job resources, self-efficacy, and work engagement among Italian schoolteachers. European Journal of Work and Organizational Psychology, 20, 285–304. http://dx.doi.org/10.1080/13594320903513916 Skaalvik, E. M., & Skaalvik, S. (2010). Teacher self-efficacy and teacher burnout. Teaching and Teacher Education, 26, 1059 –1069. http://dx.doi.org/10.1016/j.tate.2009.11.001 Skaalvik, E. M., & Skaalvik, S. (2011). Teacher job satisfaction and motivation to leave the teaching profession: Relations with school context, feeling of belonging, and emotional exhaustion. Teaching and This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. TEACHERS’ OCCUPATIONAL WELL-BEING Teacher Education, 27, 1029 –1038. http://dx.doi.org/10.1016/j.tate.2011.04.001 Skaalvik, E. M., & Skaalvik, S. (2014). Teacher self-efficacy and perceived autonomy: Relations with teacher engagement, job satisfaction, and emotional exhaustion. Psychological Reports, 114, 68 –77. http://dx.doi.org/10.2466/14.02.PR0.114k14w0 Smith, A., Brice, C., Collins, A., Matthews, V., & McNamara, R. (2000). The scale of occupational stress: A further analysis of the impact of demographic factor