The demands addressed in this research were work overload and students' misbehavior. The work overload is associated with excessive demands related to the tasks to be performed, usually motivated by the employee's lack of time or knowledge (Korpershoek et al., 2016). Those demands may be related to the pace of work, the number of hours scheduled for the execution of the tasks, or the use of new technologies the individual is not qualified.

When the number of tasks to be performed is huge, or the time available for execution is insufficient, the overload is quantitative. On the other hand, when the employee has no knowledge or ability to perform the work, the overload is considered qualitative (Dewe et al., 2010). Both types of overload usually lead to boredom, demotivation, exhaustion, drop in performance, and lack of work engagement (Molino et al., 2015).

The student's misbehavior, in turn, refers to behaviors that interfere with teaching in the classroom and cause the teachers to experience emotional suffering (Lopes et al., 2017). In that sense, those behaviors impede the effective occurrence of the teaching-learning process, as they negatively affect communication in the classroom (Çoban, 2015). According to the JD-R Theory, when they become constant, the job demands tend to undermine the individuals' energy and make them experience moments of tension, which ends up interfering in their motivation and, consequently, in their engagement (Bakker & Demerouti, 2017). Thus, both the work overload and students' misbehavior, being characterized as job demands, are expected to lead to physical and psychological exhaustion, which would interfere in the teachers' work engagement (Lopes et al., 2017).

In this sense, in a study by Fiabane et al. (2013), work overload was considered as one of the most important predictors of work engagement in Italian health professionals. It was concluded that the absence of overload, be it quantitative or qualitative, prevented employees' energies from being exhausted. This is, it protected them against emotional exhaustion. Similar findings were also found in samples of Swiss nurses and physicians (Setti & Argentero, 2011), Italian employees (Molino et al., 2015), and Finnish teachers (Hakanen et al., 2006).

Similarly, studies of student misbehavior have shown that this variable is positively associated with burnout and teacher stress (Haydon & Alter, 2017), threatening the teachers' well-being (Boyle et al., 1995). Based on these considerations, the following hypotheses were formulated:

H1: Student misbehavior is negatively related to teachers' work engagement.

H2: Work overload is negatively related to teachers' work engagement.

The personal resource investigated in this study was self-efficacy. This construct refers to the individual's perception of his ability to perform specific and necessary actions within reach of a given goal (Ritter & Lorig, 2014). According to the JD-R Theory, personal resources contribute to employee motivation, helping employees deal with emotional demands and becoming more engaged (Bakker & Demerouti, 2017). Therefore, they constitute coping strategies that favor occupational activities and improve employee well-being (Kassing et al., 2012).

In line with those assertions, studies carried out with German (Klusmann et al., 2008), Spanish (Salanova et al., 2005), Dutch (Bakker & Xanthopoulou, 2013), Iranian (Zangenehvandi et al., 2014), and Norwegian teachers (Skaalvik & Skaalvik, 2014) found that self-efficacy correlated positively with work engagement. This construct is a significant predictor of attitudes associated with demonstrating more energy, inspiration, and work involvement (Bakker & Demerouti, 2017). The hypothesis raised taking these considerations as a reference is:

H3: Teachers' self-efficacy is positively related to work engagement.

The job outcomes investigated in this study were turnover intention and job crafting behavior. Turnover intention is characterized by the thoughts and desires to request dismissal to the immediate superior (Clemens et al., 2009). Engaged employees become more motivated and start to carry out their tasks with more energy, enthusiasm, and work engagement, which is why they do not usually have turnover intentions (Oliveira & Rocha, 2017). In this sense, the study conducted by Høigaard et al. (2012) in Norwegian teachers showed that work engagement was negatively correlated with turnover intention, in a demonstration that, when teachers carry out their activities with more energy and motivation, the chances of them quitting their jobs are reduced.

The job crafting behavior, in turn, refers to the set of proactive behaviors that employees use to minimize the job demands and maximize the job resources (Bakker, 2017). Job crafting attempts seem particularly effective when employees are highly engaged in their work (Demerouti et al., 2019). In this way, research on this construct has shown a positive association between work engagement and job crafting behavior (Bakker et al., 2016). Also, the study conducted by Leana et al. (2009) showed that early American childhood education teachers who were more engaged modified their behaviors to combat the job demands and to obtain the necessary resources to achieve their tasks, thus improving the quality of service to the students. The hypothesis raised taking these considerations as a reference is:

H4: Work engagement is negatively related to turnover intention.

H5: Work engagement is positively related to job crafting behavior.

Studies on the mediation of work engagement are still incipient since most investigations have focused on analyzing the antecedents or consequences of this construct and not on its role as a mediator of the relationships between job demands and resources and job outcomes. According to the JD-R Theory, however, the employees who use their resources become more capable of combating the demands and presenting greater work engagement. This also makes them demonstrate more positive results at work, such as better performance and lower turnover (Bakker & Demerouti, 2017)and a greater sense of meaning and efficacy in accomplishing tasks (Xanthopoulou et al., 2009).

In line with this theory, in the research by Bakker and Xanthopoulou (2013) in Dutch elementary school teachers, self-efficacy increased work engagement, which led teachers to become more attentive to their own ability to perform activities. Similarly, Salanova et al. (2005) found that college teachers, when they felt effective, increased their engagement and, consequently, began to demonstrate more involvement in accomplishing their tasks.

Self-efficacy is, therefore, a personal resource that probably leads the employee to work with greater energy, enthusiasm, emotional well-being, self-regulation, persistence, perseverance, and involvement in carrying out their tasks, that is, to show more work engagement (Skaalvik & Skaalvik, 2014). Such a state of engagement allows the individual to actively participate in tasks, aim to reach goals and exert greater control over events (Bakker & Demerouti, 2018). It would thus be expected that work engagement would mediate the relationships of self-efficacy with work outcomes, such as the turnover intention and job crafting behaviors. In other words, the expectation is that work engagement acts as a mediator of the negative relationship between self-efficacy and the turnover intention and the positive relationship between self-efficacy and job crafting behaviors.

On the other hand, work overload and student misbehavior are characterized as demands that impose barriers fulfilling tasks and make the occupational environment unfavorable, consequently leading to negative work activities (Çoban, 2015). In this sense, those demands drain workers' energy, leading them to experience states of tension, physical and emotional exhaustion, and less involvement with work, making it difficult to change behaviors in favor of occupational activities (Bakker & Demerouti, 2018). Then, it can be assumed that work engagement acts as a mediator of the positive relations between those demands and turnover intention and the negative relationships between these demands and the job crafting behaviors. Based on these considerations, the following hypotheses were raised:

H6: Work engagement acts as a mediator of the negative relationship of self-efficacy with turnover intention.

H7: Work engagement acts as a mediator of the positive relationship of self-efficacy with job crafting behaviors.

H8: Work engagement acts as a mediator of the positive relationship of students' misbehavior with the turnover intention.

H9: Work engagement acts as a mediator of the negative relationship of students' misbehavior to job crafting behaviors.

H10: Work engagement acts as a mediator of the positive relationship between work overload and turnover intention.

H11: Work engagement acts as a mediator of the negative relationship between work overload and job crafting behaviors.

The direct effect of the independent variables (student misbehavior, work overload, and self-efficacy) on the dependent variables (turnover intention and job crafting behavior) was initially observed to check the hypotheses. The results evidenced that these effects were significant for work overload on turnover intention (WOL - INT: β = 0.19, p < 0.01) and on the job crafting behavior (WOL - JCB: β = 0.12; p < 0.05), and for self-efficacy on the turnover intention (SEF - INT: β = -0.11, p < 0.05). The inclusion of work engagement as a mediating variable in the model indicated that: the students' misbehavior was negatively and significantly associated with the engagement (MIS - ENG: β = -0.08; p < 0.05), which confirmed hypothesis 1; that the overload was also negatively and significantly correlated with the engagement (WOL - ENG: β = -0.08; p < 0.05), which permitted the confirmation of hypothesis 2; that self-efficacy was positively and significantly related to engagement (SEF- ENG: β = 0.70; p < 0.01), thus confirming hypothesis 3. Regarding the relationship between the mediator variable and the dependent variables, the results showed that there was a negative and significant association between engagement and turnover intention (ENG – INT: β = -0.29; p < 0.01), which permitted confirming hypothesis 4, as well as a positive and significant relationship between the engagement and job crafting behaviors (ENG - JCB: β = 0.24; p < 0.01), which confirmed hypothesis 5.

The indirect effect of self-efficacy on the turnover intention was negative and significant (SEF - ENG - INT: β = -0.20; p < 0.01). Similarly, the indirect effect of the relationship between self-efficacy and job crafting behavior (SEF - ENG - JCB: β = 0.17; p < 0.01) was also statistically significant. It was also observed that, in the presence of the mediator variable, the direct effects of self-efficacy on turnover intention and of self-efficacy on job crafting behavior were not significant. In summary, engagement mediated the relationship between self-efficacy and the turnover intention, as well as between self-efficacy and job crafting behavior, which confirmed hypotheses 6 and 7.

The indirect effect of students' misbehavior on the turnover intention (SEF - ENG - INT: β = 0.02; p > 0.05) was not statistically significant. In addition, the indirect effect of the students' misbehavior on the job crafting behavior (MIS - ENG - JCB: β = -0.02; p > 0.05) was not statistically significant either. It was also verified that, in the presence of the mediator variable, the direct effects of the misbehavior on the turnover intention and of the misbehavior on the job crafting behavior were not significant. In this sense, engagement did not mediate the relationship between those variables, which impeded the confirmation of hypotheses 8 and 9.

Similarly, the indirect effect of the work overload on the turnover intention (WOL - ENG - INT: β = 0.02; p > 0.05), as well as the indirect effect of the work overload on the job crafting behavior (WOL - ENG - JCB: β = -0.02, p > 0.05), were not statistically significant, thus impeding the confirmation of hypotheses 10 and 11. The final model is displayed in Figure 1. Their fit indices were equal to: c² (501) =6826.849; CFI = 0.92; TLI = 0.91; RMSEA = 0.08.

Figure 1 Figure 1 Graphical representation of the model tested with the standardized parameters

This study aimed to test a model of relationships between job demands (work overload and misbehavior of students) and personal resources (self-efficacy) and job outcomes (turnover intention and job crafting behaviors), as well as the mediating role of work engagement in those relationships. The results showed that students' misbehavior was negatively related to work engagement, which confirmed hypothesis 1. This result converges with the previous findings, based on which it was also concluded that a negative relationship exists between the students' misbehavior and the work engagement (Lopes et al., 2017; Hakanen et al., 2006). Students' misbehavior implies an increase in the psychological costs associated with the teaching activity, which would cause the teachers to become unmotivated to carry out their school activities. In other words, disobedience, disrespect for the teacher, among others, can end the energy, inspiration, and engagement of the teacher in the work tasks that need to be developed (Korpershoek et al., 2016; Hakanen et al., 2006).

The work overload was negatively and significantly related to the work engagement, which confirmed hypothesis 2. It was verified that high workloads, whether quantitative or qualitative, interfere with the teachers' work engagement. This result corroborates the study by Hakanen et al. (2006), who also observed negative relationships between work overload and engagement. It is possible that the teachers perceive the fact that they have many tasks to accomplish in their daily work (such as correcting tests and assignments and registering grades), within short deadlines, besides having to participate in other activities, such as board meetings and meetings with parents of students, as an obstacle that will reduce their energy, inspiration, and engagement in teaching tasks (Molino et al., 2015; Timms et al., 2007).

The results about the negative relationships of work overload and misbehavior of students with work engagement can also be viewed as a confirmation of the proposition of the Job Demands and Resources Theory that job demands lead to losses to the physical and psychological health of the workers. In this sense, they contribute to diminishing their work engagement (Bakker, 2017).

It was also observed that self-efficacy was positively related to work engagement, which fully confirmed hypothesis 3. In this sense, the findings permit affirming that teachers who more strongly believe that they can perform their work activities manifest greater engagement, also evidenced in Xanthopoulou et al. (2007). This may be because when teachers realize they can accomplish their tasks and set goals on how to achieve them, they become more motivated to make efforts and persist in carrying out these activities (Bakker & Demerouti, 2017; Simbula et al., 2011). In other words, personal resources such as self-efficacy help employees to actively approach their job demands and effectively deal with them, which makes them feel more engaged in their work (Bakker & Demerouti, 2018).

It was also verified that the work engagement was negatively related to the turnover intention, which confirmed hypothesis 4 and converging with other studies, such as Gupta and Shareen (2017) and Schaufeli and Bakker (2004). It has been demonstrated, then, that when teachers experience high levels of energy, being absorbed in the performance of their tasks, and dedicated to them, the thought of leaving their jobs ends up being distanced. This may be since the teacher, when motivated, inspired, and dedicated to his work, ends up developing strategies to perform his tasks with greater disposition and to face the demands inherent to the school context, which is why he does not develop the desire to quit his job (Saks, 2006).

Work engagement was also positively related to job crafting behaviors, thus confirming hypothesis 5. Thus, it was found that high levels of energy, inspiration, and work engagement maximize the proactive behaviors towards the achievement of the outlined work goals, confirming the study by Bakker et al. (2016). It seems, hence, that the affective-motivational states are fundamental stimulators for the teacher to find resources and, consequently, to face the demands, thus carrying out his/her occupational activities with greater proactivity and greater possibilities to accomplish the outlined work goals (Beer et al., 2016).

Work engagement is, therefore, a predictor of important employee outcomes. In this way, engaged employees are goal-directed and focus all their energy on the tasks (Bakker & Albrecht, 2018). This can explain why their engagement diminishes their turnover intention and increases their job crafting behaviors (Demerouti et al., 2019; Bakker, 2017).

Another noteworthy point is that work engagement fully mediated the negative relationship of self-efficacy with the turnover intention, which permitted confirming hypothesis 6. Therefore, the findings showed that when teachers have a clearer view of their task accomplishment skills, they feel more energetic, inspired, and engaged in their work. As a result, they become less inclined to quit their job, probably because they can develop more coping strategies regarding the job demands. This result was also found in the study by Chen and Chen (2012) and Bakker et al. (2016), considering the mediating role of engagement in the relationships between job resources and turnover intention.

Furthermore, the findings showed that work engagement also mediated the positive relationship of self-efficacy with job crafting behavior, which confirmed hypothesis 7. This result may be since the teachers who believe more in their abilities to carry out specific actions to achieve a particular goal develop greater enthusiasm, inspiration, and engagement in their tasks, which triggers proactive behavior to change the way to accomplish school tasks. Such actions probably end up improving classroom management and problem-solving (Bakker et al., 2016).

Nevertheless, the work engagement mediated neither the positive relationship of the students' misbehavior with the turnover intention nor the negative relationship between the students’ misbehavior and the job crafting behavior, which impeded the confirmation of hypotheses 8 and 9. This may happen because the teachers do not perceive the students' misbehavior as a demand but as habitual conduct inherent in the school context, which needs to be managed and administered as part of their classroom activities. In other words, they are probably perceived as expected behaviors, which need to be controlled over the class period (Çoban, 2015; van Wingerden et al., 2017).

Similarly, it was verified that the work engagement mediated neither the positive relationship between the work overload and the turnover intention nor the negative relationship of the work overload with the job crafting behaviors, which prevented the confirmation of hypotheses 10 and 11. Those results may be because the teachers do not perceive the work overload as a demand, but rather as something inherent to the practice of the profession, that is, as something perceived as ordinary, even when faced with a large number of tasks (Molino et al., 2015). That assertion needs further confirmation in future research, though.

Overall, the findings are in accordance with the JD-R Theory (Bakker & Demerouti, 2018). The demands, such as work overload, require physical and psychological effort, drain the worker's energy, thus constituting dysfunctional aspects. Personal resources, in turn, facilitate the adaptation to the circumstances of life (Bakker & Demerouti, 2018; Xanthopoulou et al., 2007). Thus, JD-R Theory constitutes an overarching framework applied to various occupational settings, irrespective of the specific job demands and job resources involved (Bakker & Demerouti, 2017).

Regarding the limitations of the research, its cross-sectional design prevents the verification of causal relationships between the research variables. Another limitation concerns the fit indices of some of the scales (work overload and student misbehavior), which were not very good, which indicates the need to reevaluate these scales. In addition, the sample only included Brazilian teachers, which is why the results obtained should be generalized with caution to samples of other nationalities. In addition, the study verified only the impact of personal resources and job demands, not including, in the tested model, the job resources, which are also part of the JD-R Theory.

As for a future research agenda, it would be interesting to carry out studies with longitudinal designs to deepen the understanding of the research variables. In addition, its integration into a job resource and demand model, including other personal resources, such as self-esteem and optimism, would contribute to the test of a more robust research model. It would also be interesting to compare such a model in public and private education teachers as most of the studies so far have been using public school teachers.

In any case, teachers with firmer self-efficacy beliefs tend to experience higher levels of work engagement and, consequently, develop more proactive goal-oriented behaviors and lower turnover intention, which will probably favor their good performance in educational institutions. Regarding the practical implications, training programs could be implemented to develop feelings of self-efficacy and attitudes of greater work engagement in the teachers, which could make them more proactive and guarantee that they continue in their jobs.