Eindhoven University of Technology MASTER Keeping Employees Engaged The Effects of Job Demands, Job Autonomy and the Role of Sleep Quality Bains, S.S.

(1)

Eindhoven University of Technology

MASTER

Keeping Employees Engaged

The Effects of Job Demands, Job Autonomy and the Role of Sleep Quality

Bains, S.S.

Award date:

2020

Link to publication

Disclaimer

This document contains a student thesis (bachelor's or master's), as authored by a student at Eindhoven University of Technology. Student theses are made available in the TU/e repository upon obtaining the required degree. The grade received is not published on the document as presented in the repository. The required complexity or quality of research of student theses may vary by program, and the required minimum study period may vary in duration.

General rights

Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners

(2)

Eindhoven, December 2020

Keeping Employees Engaged: The Effects of Job Demands, Job Autonomy and the Role of Sleep

Quality

Sunjit Bains

Eindhoven University of Technology Student Identity Number: 0823945

In partial fulfillment of the requirements for the degree of

Master of Science in Operations Management and Logistics

Supervisors:

Dr. L. (Leander) van der Meij Eindhoven University of Technology, HPM Dr. S. (Sonja) Rispens Eindhoven University of Technology, HPM Dr. P. (Philippe) van de Calseyde Eindhoven University of Technology, HPM

TU/e, Department of Industrial Engineering and Innovation Sciences

(3)

I would like to thank several people from the Eindhoven University of Technology who were involved in this master thesis project. Firstly, I would like to thank Leander van der Meij for his continuous support and encouragement during the course of writing my master thesis. Leander was always available for support, but also gave me the freedom to find my own way. Secondly, I would like to thank Sonja Rispens for her supervision and feedback. Thirdly, I would like to thank Philippe van de Calseyde for assessing my thesis. Moreover, I would like to thank Sophia Frick for introducing and laying the foundation of this project. Lastly, I would like to thank Piet van Gool, who has been very valuable in the guidance of the statistical analyses of my thesis.

Furthermore, I would like to thank the 51 people who participated in the study. Without their efforts and willingness to participate, this research would not have been possible.

During my studies at the Eindhoven University of Technology, I have been able to learn a lot, both through my studies and through other activities. Industrial Engineering has taught me to attack problems in a systematic way and to always feel motivated to work towards organizational excellence. Next to my studies, I have had the opportunity to develop myself, both as a team player and as a leader. The Congress committee has enabled me to develop my organizational and leadership skills, which I am sure will be very valuable in my professional career.

Finally, I would like to thank my parents, friends and girlfriend for their support, guidance and faith in me during my studies.

Sunjit Bains December 2020

Preface

(4)

Scientific abstract

This thesis focused on researching the fluctuations of momentary work engagement in a sample of employees, within-person, moment-to-moment within the workday as a function of momentary job demands, momentary job autonomy and daily sleep quality. It was hypothesized that momentary job demands would have a direct negative effect on the experience of momentary work engagement, while momentary job autonomy was expected to directly and positively affect work engagement. Momentary job autonomy and daily sleep quality were expected to have a negative effect on the relation between momentary job demands and work engagement. Moreover, it was hypothesized that sleep quality would have a positive effect on the relation between momentary job autonomy and momentary work

engagement. A field study was carried out over the course of one week, with seven measurement occasions per day (N = 51 persons). As predicted, the direct effect of momentary job autonomy on momentary work engagement was found to be positive. The relation between momentary job demands and momentary work engagement was found to be positive as well, which was against expectations. This finding may be explained by the notion that momentary job demands may be perceived as a challenge instead of a hindrance. Furthermore, it was found that more momentary job autonomy leads to an increased positive effect of momentary job demands on momentary work engagement. Self-perceived sleep quality and self-perceived recovery were found to have the same positive effect on the relation between momentary job demands and momentary work engagement. No support was found for the moderation of daily sleep quality on the relation between momentary job autonomy and momentary work engagement. The results suggest that short-term job demands may be perceived as a challenge instead of a hindrance. Due to the perception of momentary job demands as challenge, momentary work engagement is increased. Momentary job autonomy, self-perceived sleep quality and recovery were found to have a positive effect on the relation between momentary job demands and momentary work engagement.

(5)

List of Figures and Tables

(8)

Burnout is a psychological response to stressors on the job and has become an accepted way to describe the personal agony of job stress. Burnout is considered a problem for the employee as well as the employer, as it can bring about many costs for both sides. For the employer, a burned-out employee performs less and delivers a lower quality of work. For the employee, a burnout could lead to poorer health, and problems in family relationships (Karatepe & Olugbade, 2009). Work engagement, which is considered as the opposite of burnout and refers to “a positive, fulfilling, work-related state of mind that is characterized by vigor dedication, and absorption” (González-Romá, Schaufeli, Bakker, & Lloret, 2006;

Langelaan, Bakker, van Doornen, & Schaufeli, 2006; Llorens, Schaufeli, Bakker, & Salanova, 2007;

Maslach & Leiter, 2006; Schaufeli, Salanova, González-romá, & Bakker, 2002). According to research, engaged employees are more likely to perform well at their job than disengaged employees (Schaufeli &

Bakker, 2013). Work engagement is characterized by enthusiasm, commitment and dedication, experiences which are replaced during burnout by a sense of exhaustion, cynicism and ineffectiveness (Karatepe & Olugbade, 2009; Maslach & Leiter, 2006). Thus, keeping employees engaged is an important task, which is why it might be considered relevant to investigate which factors play a role in achieving work engagement among employees.

Studies have shown that sufficient job resources could lead to more engaged employees (Bakker &

Demerouti, 2007; Demerouti, Nachreiner, Bakker, & Schaufeli, 2001). Job autonomy is a job resource that describes the degree to which a job gives freedom to an employee, the independence and discretion to the individual in scheduling work, as well as the freedom to determine the procedures for carrying out job assignments (Hackman & Oldham, 1975). Thus, job autonomy is a resource that is expected to contribute to the experience of work engagement among employees (Sonnentag, 2003). However, the importance of job autonomy may be influenced by the presence of job demands (Schaufeli & Taris, 2013). When these demands become too much for an employee to cope with and the demands require too much effort, they may turn into job stressors, which could in turn lead to symptoms of anxiety, depression, or even burnout (Schaufeli & Bakker, 2004). An increase in the job resource job autonomy may enable the employee to better cope with job demands (Meijman & Mulder, 1998). Therefore, the role of job autonomy is relevant to research, as well as its moderating role between job demands and work engagement.

The replenishment of job resources is also expected to play a substantial role in achieving a state of work engagement. An important factor in the replenishment of job resources is recovery. Good recovery is helped by good sleep quality, and thus sleep may play a vital role in work engagement (Ebert et al., 2015;

Meijman & Mulder, 1998; Schaufeli et al., 2009). Therefore, ensuring high sleep quality among

employees may be vital for maintaining an engaged workforce (Hahn, Binnewies, Sonnentag, & Mojza, 2011). Job resources are expected to moderate the relation between job demands and work

engagement. Therefore, sleep quality is expected to also play a moderating role in this relationship (Bakker, Demerouti, & Sanz-Vergel, 2014; Breevaart & Bakker, 2018).

This thesis focused on researching the fluctuations of work engagement in a sample of employees, within-person, moment-to-moment within the workday as a function of momentary job demands and momentary job autonomy. Sleep quality was included as a moderating variable to investigate its role in achieving a state of work engagement. This research contributes to the literature by investigating the role of job autonomy, job demands and sleep quality and their potential influence on work engagement at a momentary level. Previous research has mostly focused on general or day-level effects of these variables, which is why it is interesting to investigate whether the same relations exist at the within- person, moment-to-moment level during the workday (i.e., the momentary level).

1 Introduction

(9)

Research suggests that engaged employees are more likely to perform well at their job than disengaged employees (Schaufeli & Bakker, 2013). There are several reasons for the success of engaged employees.

These reasons include that engaged employees are more proactive and take more personal initiative, that they set high goals because they feel competent, that they are intrinsically motivated and therefore prepared to go the extra mile and that they are cooperative and friendly, which makes them more attractive to work with. Furthermore, engaged employees experience plenty of positive emotions, which makes them more capable to process information. Finally, engaged employees are healthier, which makes them less susceptible to become sick and go on sick leave (Schaufeli & Bakker, 2013). These findings suggest that having engaged employees may be beneficial for job performance, and thus an advantage for employers. However, it may be questioned how work engagement among employees can be achieved.

Work engagement

High performance is important for company success, which is why employers strongly desire having engaged employees as members of their staff (Ufuophu-Biri & Iwu, 2014). Moreover, more engaged employees tend to increase a company’s profits (Macey, Schneider, Barbera, & Young, 2009). Since work engagement is thus related to performance, employers might wish to achieve this state for their

employees (Ufuophu-Biri & Iwu, 2014). Work engagement can be defined as “a positive, affective- motivational state of fulfillment that is characterized by vigor, dedication, and absorption” (Schaufeli et al., 2002). The three components that characterize work engagement all describe different aspects of an employees’ experience of being engaged. The first component, vigor, can be defined by high levels of energy and mental flexibility while at work, the enthusiasm to devote oneself to one’s work and perseverance to continue even when this seems very hard (Schaufeli et al., 2002). Hence, an employee who experiences vigor during his duties has a higher chance of being more ambitious in his work and is also more likely to be able to endure hardships during his work (Mauno, Kinnunen, & Ruokolainen, 2007). The second component of work engagement is dedication. Dedication describes a feeling of significance, enthusiasm, inspiration, pride and challenge. Dedication shares some similarities with the more common known concept of job involvement, which refers to the degree to which an employee is committed to its job (Brown, 1996; Cooper-Hakim & Viswesvaran, 2005). The third and final component of work engagement is absorption. Absorption can be defined by a feeling of being fully concentrated and intensely absorbed in one’s work so that one does not have a sense of time. Detaching oneself from work can then even sometimes lead to hardship and suffering for an employee. Absorbed people tend to be in a certain state of mind where they are so heavily entangled in an activity that they think of nothing else (Schaufeli et al., 2002).

Although work engagement is characterized by the three aforementioned components, it is believed that the core components of work engagement encompass the dimensions of vigor and dedication, which are also the exact opposites of the dimensions of burnout namely, emotional exhaustion and cynicism, respectively (González-Romá et al., 2006; Langelaan et al., 2006; Llorens et al., 2007). Burnout can be defined by three dimensions namely, professional inefficacy, cynicism and exhaustion (Maslach & Leiter, 2016). Furthermore, work engaged employees are proactive and are characterized by taking initiative when projects tend to get stuck, whereas burned-out employees are more likely to feel displaced and take on a more passive stance towards their projects, which further illustrates why having work engaged employees is desirable (Schaufeli & Bakker, 2013). In addition, engaged employees are better able to cope with the demands of their job, and are thus less likely to find themselves in a state of burnout (Barkhuizen, Rothmann, & Van De Vijver, 2014).

2 Theoretical background

(10)

Work engagement has been conceptualized as a relatively stable individual difference variable, but research has shown that there are within-person variations in the experience of work-engagement (Macey & Schneider, 2008; Sonnentag, 2003; Xanthopoulou & Bakker, 2012). Moreover, experience sampling studies and diary studies have shown extensive within-person variations in work-related affective experiences, as well as fluctuations from one day to another (Sonnentag, 2003). This implies that work engagement not only differs between employees, but also within-person and from one day to another (Bakker & Bal, 2010). Therefore, scientific research has defined daily work engagement as a short-term, fulfilling, positive and work-related state of mind that is characterized by dedication, vigor and absorption and fluctuates within individuals over a short period of time (Breevaart, Bakker, &

Demerouti, 2014). Thus, the definition of daily work engagement is rather similar to the general definition of work engagement. However, there are two differences. Firstly, daily work engagement refers to a short-term state of mind, and secondly, this state of mind happens over a short period of time, namely one day (Breevaart et al., 2014; Schaufeli et al., 2002).

So far, we have discussed what work engaged employees are and why they are more likely to perform better at their job. Although work engaged employees experience feelings of vigor, dedication and absorption at their work, this may be confused with being too attached to one’s work. When an

employee is too attached to his work he could be a workaholic (Bakker, Schaufeli, Leiter, & Taris, 2008).

However, in contrast to workaholism, positive consequences of highly engaged employees include a better work performance, more organizational commitment and better health (Bakker, Hakanen, Demerouti, & Xanthopoulou, 2007; Schaufeli & Taris, 2013). In order to maintain or increase employee work engagement, several factors seem to be important, for example, having sufficient job resources to cope with job demands (Bakker & Demerouti, 2007; Demerouti, Nachreiner, Bakker, & Schaufeli, 2001;

Schaufeli & Bakker, 2004).

Job resources, job demands and work engagement

Employees have a bigger chance of becoming more engaged in their work when they have sufficient job resources (Bakker & Demerouti, 2007). Moreover, daily work engagement may be enhanced by a motivational process that is stimulated through the presence of job resources (Breevaart et al., 2014).

Job resources refer to those physical, organizational and social aspects of the job that could reduce the job demands, help in the achievement of work goals and stimulate the employees personal growth, learning and development (Bakker & Demerouti, 2007; Schaufeli & Bakker, 2004). This means that when an employee has sufficient resources, the employee should experience support, for example, through performance feedback, autonomy, social support, skill variety and learning opportunities (Bakker &

Demerouti, 2007).

The importance of resources could be further explained by making use of the conservation of resources (COR) theory (Schaufeli & Bakker, 2013). The COR theory can be seen as a motivation theory where the main principle states that employees wish to gain resources, hold on to their existing resources or regain lost resources (Hobfoll, 2001; Schaufeli & Bakker, 2013). When an employee gains resources, it becomes easier to perform the work duties, which would make the employee more successful and encourage more positive feelings towards the job (Reis, Hoppe, & Schröder, 2015). This increase in resources could spark a repeating effect (i.e. it is easier to gain more resources, when an employee already has

resources) when the resources are accumulated, turning the resource gain into a cycle; which is conveniently called a gain cycle (Schaufeli & Bakker, 2013). Therefore, when an employee is more engaged, it is more likely that the accumulation of resources already in his possession will grow

(Schaufeli & Bakker, 2013). Finally, in line with ideas from the COR theory, the increase in job resources will lead to more work engagement (Schaufeli, Bakker, & van Rhenen, 2009).

(11)

However, when job resources are limited or job demands are too high, a situation could arise where job stress (created by this situation) may become a problem (Bakker et al., 2005). Job demands apply to those physical, organizational or social aspects of work that need sustained physical or mental effort and are therefore related to physiological and psychological costs (Bakker, Demerouti, & Euwema, 2005;

Demerouti & Bakker, 2011). When these demands become too much for an employee to cope with and the demands require too much effort they may turn into job stressors, which could in turn lead to symptoms of anxiety, depression, or even burnout (Schaufeli & Bakker, 2004). Job resources play an important role when trying to improve work engagement when there are a lot of job demands present (Schaufeli & Taris, 2013). This could also be interpreted as that job resources moderate the effects of job demands (Schaufeli & Taris, 2013). So for employees to decrease the risk of developing symptoms of burnout and to be able to create a situation where the employee feels work engaged, it is important to have job resources available (Bakker et al., 2005; Schaufeli & Taris, 2013).

Despite the possible negative consequences of job demands on employees, one may ask whether there are also positive outcomes of job demands. A meta-analysis study implicates that the relationship between job demands and work engagement relies upon the perception of the demand by the employee (Crawford, LePine, & Rich, 2010). If the perception of the employee is that he or she interprets the job demand as a challenge, the job demand is positively related to work engagement since the job demand may act as an intrinsic motivator (Sawang, 2012). An example of a challenging job demand could be intellectual demand, where an employee is given a difficult project but has the chance to develop professional skills at the same time. Another example of a job demand could be when an employee is given great responsibility, which was sought after by the employee. Receiving such great responsibility should then have a motivating effect on the employee, which may then result in work engagement.

However, when the employee perceives the job demand as an obstruction, the job demand is negatively related to work engagement (Sawang, 2012). Furthermore, it is more likely that an employee will be less stimulated and engaged with their work, when job demands are low (e.g. assembly line work, routine jobs) (Sawang, 2012). When job demands increase up to a point where the demands become more challenging but not too high, it is more likely that the employee will become more engaged in his work.

Nevertheless, when the job demands become too much (overload), this engagement could decrease, and turn into exhaustion (Sawang, 2012). Thus, it may be stated that some job demands can have positive consequences, but that an overload of demands may lead to exhaustion.

The relationship between job resources, job demands, and work engagement may be further explained using the Job Demands and Resources (JD-R) model. The JD-R model combines the effects of both job demands and resources, and can be seen as an overall foundation for promoting work engaged employees (Demerouti et al., 2001). The JD-R model suggests that burnout might be provoked by multiple aspects related to the work environment (Bakker et al., 2003). According to the JD-R model, becoming vulnerable to long-term job demands and lack of job resources are prognostic for burnout (Hakanen & Schaufeli, 2012). Job demands seem to be positively related to exhaustion, whereas job resources seem to be negatively related to cynicism and lack of professional efficacy, which are all factors of burnout (Bianchi, Schonfeld, & Laurent, 2015; Llorens, Bakker, Schaufeli, & Salanova, 2006).

The effects of job resources and job demands can induce two psychological processes (Llorens et al., 2006). Firstly, the job demands process could be described as the health impairment process, where the long-term job demands could lead to a deterioration of health (Hakanen, Bakker, & Schaufeli, 2006;

Schaufeli & Bakker, 2004) and sick leave (Bakker, Demerouti, de Boer, & Schaufeli, 2003). Secondly, the job resources process could be described as the motivational process (Llorens et al., 2006). The

motivational process stimulates employee’s motivation, because of the availability of more job resources (Hackman & Oldham, 1980), which in turn stimulates an increase in work engagement and more

progressive work outcomes, such as organizational commitment and employee performance (Salanova, Agut, & Peiró, 2005). According to the JD-R model, job demands can moderate the effect of job

resources on work engagement, whereas job resources can moderate the effect of job demands on burnout. Furthermore, the JD-R model illustrates how job resources could reduce the negative effects

(12)

that job demands might have on the development of burnout. Hence, the risk of burnout increases, and work engagement decreases when job demands are high and job resources are low. In the JD-R model, not only the negative psychological state is described (i.e. burnout), but also a positive corresponding state (i.e. work engagement) (Bauer, Hämmig, Schaufeli, & Taris, 2014). Thus, the model illustrates that when an employee experiences sufficient job resources to meet the job demands, this is related to work engagement through the motivational process (Van Den Broeck, Vansteenkiste, De Witte, & Lens, 2008).

Job autonomy and work engagement

As described in the JD-R model, having sufficient job resources is essential for reaching a state of work engagement. Job autonomy is a job resource that describes the degree to which a job gives freedom to an employee, the independence and discretion to the individual in scheduling work, as well as the freedom to determine the procedures for carrying out job assignments (Hackman & Oldham, 1975).

Research has found evidence for the positive and significant relationship between job autonomy and work engagement (Bakker & Demerouti, 2007; Demerouti et al., 2001; Hakanen et al., 2006; Schaufeli &

Bakker, 2004; Vera, Martínez, Lorente, & Chambel, 2016). Furthermore, there are also models and theories which support the positive effects and significance of job autonomy. For instance, the demand- control model (DCM) states that when an employee has job autonomy, job stress is reduced, even under the pressure of work overload or excessive job demands (Karasek, 1979, 1998). Moreover, according to the effort and recovery theory (Meijman & Mulder, 1998), job resources (i.e. job autonomy) and employees’ effort to perform a job well, are more likely to improve the working conditions (Schaufeli &

Taris, 2013). Therefore, job autonomy could enlarge the employees’ intrinsic motivation and as a result could boost work engagement, learning at work and organizational responsibility (Taipale, Selander, Anttila, & Nätti, 2011). From the known job resources, it has been expressed that job autonomy is an example of a job resource that bolsters work engagement (Schaufeli & Salanova, 2007). Moreover, a sense of control and autonomy during work are more likely to decrease emotional exhaustion, reduced personal accomplishment and cynicism, which are the dimensions of burnout (Alarcon, 2011). All in all, the aforementioned scientific research suggests that the availability of job autonomy may reduce the effects of the dimensions that predict burnout. Furthermore, job autonomy may stimulate the motivational process of the employee mentioned in the JD-R model, and thus also stimulate work engagement (Alarcon, 2011; Van Den Broeck et al., 2008). Thus, an increase in the job resource job autonomy may enable the employee to better cope with job demands. Therefore, job autonomy may also moderate the relationship between job demands and work engagement (Meijman & Mulder, 1998).

Sleep quality and work engagement

As mentioned previously, having sufficient resources available is important for achieving a state of work engagement. Therefore, being able to replenish resources may play a substantial role. As mentioned in the effort and recovery theory, recovery is important for the replenishment of job resources (Meijman &

Mulder, 1998). Thus, recovery, which is in part achieved through sleep, may play a vital role in stimulating work engagement (Ebert et al., 2015; Meijman & Mulder, 1998; Schaufeli et al., 2009). A large number of scientific research has focused on the importance of sleep.

As a result of this research, sleep has often been referred to as the most basic and important factor in the recovery process (Ebert et al., 2015). Moreover, for employee recovery, higher levels of sleep quality may play an important role, since sleep quality is a relevant factor for employee well-being (Hahn et al., 2011). The term ‘’sleep quality’’ is an important and complex construct, that has not yet been defined in the literature with an established definition (Crivello, Barsocchi, Girolami, & Palumbo, 2019). An example of a measure that is often used to capture sleep quality, is the Pittsburgh Sleep Quality Index (PSQI). The PSQI is used to portray a global view of sleep quality through the use of subjective, self-reported diaries

(13)

or questionnaires (Crivello et al., 2019; Krystal & Edinger, 2008). For the purpose of this research, it was assumed that sleep quality includes not only the more easily quantifiable components of sleep, such as sleep latency, sleep efficiency and total hours of sleep, but that sleep quality also includes subjective indices such as self-perceived quality of sleep and self-perceived recovery. Thus, sleep quality might be defined as a subjective measure of sleep, that tries to capture not only the quantifiable components of sleep, but also a person’s subjective experience of sleep (Buysse, Reynolds, Monk, Berman & Kupfer, 1989; Crivello et al., 2019; Krystal & Edinger, 2008; Pilcher, Ginter, & Sadowsky, 1997). In this study, several aspects of sleep have been used to measure sleep quality. The quantifiable components of sleep included in this study are total sleep, sleep latency and sleep efficiency. Sleep latency is defined as the time between bed and the onset of sleep (Van Der Schuur, Baumgartner, & Sumter, 2019). Moreover, total sleep describes the total hours slept during the night. Lastly, sleep efficiency can be defined as the ratio between the total sleep time and the time spent attempting to initially fall asleep and sleep discontinuity (Reed & Sacco, 2016). These objective measures of sleep quality could be accompanied by two subjective measures of sleep, for instance, self-perceived sleep quality and self-perceived recovery.

The inclusion of both subjective and objective measures to assess sleep quality could give more insight in the employees’ recovery, which might affect an employees’ work engagement during the next day (Sonnentag, 2003).

Although few studies have researched the relationship between sleep quality and work engagement, the effort-recovery model could offer a hopeful theoretical explanation for the relationship between sleep quality and work engagement (Barber, Grawitch, & Munz, 2013). According to this model, employees need to recover from the job demands from one day in order to be able to perform well on the next day (Meijman & Mulder, 1998). When employees go to sleep, their resource levels have the opportunity to reenergize so that they may be used by the employee during the next day (Baumeister, Muraven, & Tice, 2000). Following the doctrine of the COR theory, restoration of resources enables the employee to gain more resources (Schaufeli et al., 2009). Subsequently, as also illustrated by the JD-R model, an employee with more resources is then able to experience higher levels of work engagement (Salanova et al., 2005;

Schaufeli et al., 2009). Thus, the revival of an employees’ resources while sleeping may enable the employee to experience more work engagement on the following day (Kühnel, Zacher, de Bloom, &

Bledow, 2017; Salanova et al., 2005; Schaufeli et al., 2009; Sonnentag, 2003). Could it therefore be argued that higher sleep quality leads to a decreased effect of job demands on work engagement, meaning that employees are then better able to cope with job demands?

As mentioned previously, job resources moderate the effects of job demands (Schaufeli & Taris, 2013).

Furthermore, the restoration of resources during sleep provides the employee with the necessary fuel to handle the job demands at work (Bakker et al., 2014; Breevaart & Bakker, 2018). Because energy is needed to restore the resources that are used to cope with job demands, it suggests that for employees to become more engaged, they need a good recovery (Sheng, Wang, Hong, Zhu, & Zhang, 2019). Hence, since an employee needs resources to buffer the effects of job demands, this means that higher sleep quality could positively affect the level of an employee’s daily work engagement and thus buffer the negative effects of job demands (Kühnel et al., 2017; Salanova et al., 2005; Schaufeli et al., 2009;

Schaufeli & Taris, 2013; Sonnentag, 2003). Therefore, the relationship between job demands and daily work engagement may be moderated by sleep quality (Kühnel et al., 2017; Salanova et al., 2005;

Schaufeli et al., 2009; Schaufeli & Taris, 2013; Sonnentag, 2003).

(14)

Research model

In this research, momentary work engagement, momentary job demands and momentary autonomy of an employee were measured for one individual at several moments during the day. So far, only research on work engagement on a daily level or person level has been done. To the best of my knowledge, no other research has been done that had previously investigated these relations at a momentary level, and this research may thus be considered a gap in the literature.

To summarize, according to the JD-R model, it was expected in this research that momentary job demands negatively affect momentary work engagement (H1). Furthermore, higher daily sleep quality may lead to the replenishment and enhancement of the effects of job resources needed by an employee for coping with momentary job demands and, as a consequence, for achieving a state of momentary work engagement. Therefore, high daily sleep quality, which is measured through self-perceived sleep quality, self-perceived recovery, observed sleep latency, observed sleep efficiency and observed total sleep time, may buffer the effect of momentary job demands on momentary work engagement, and low daily sleep quality may enhance the effects of momentary job demands on momentary work

engagement (H2). Additionally, more momentary job autonomy was expected lead to more momentary work engagement, as it should positively affect an employees’ momentary work engagement directly through the presence of the job resource momentary job autonomy (H3). Furthermore, momentary job autonomy was expected to enhance an employees’ ability to cope with momentary job demands, thus buffering the effect of momentary job demands on momentary work engagement (H4). Moreover, better daily sleep quality may improve momentary work engagement and replenish job resources (i.e.

job autonomy), so sleep quality might enhance the effect between momentary job autonomy and momentary work engagement. Therefore, high daily sleep quality may enhance the effects of

momentary job autonomy on momentary work engagement and low daily sleep quality may hamper the effects that momentary job autonomy has on momentary work engagement (H5). Figure 1 illustrates research model and the possible relations between momentary job demands, momentary job autonomy, daily sleep quality and momentary work engagement.

Figure 1: Research Model

(15)

Research design

This section discusses the research design, participants, procedure, measures and analyses used in this research. To investigate work-related experiences throughout the day, a field study using the Experience Sampling Method (ESM) was initiated (Oerlemans & Bakker, 2013). To track subjective sleep in terms of timing, duration and quality, daily diaries were used. Light exposure, sleep timing and sleep duration were monitored objectively using wearable sensors that were worn during working days and on the weekend. ESM was used to collect data on job demands, job autonomy and work engagement in seven instances during the day and when the participant was awake, and he would maintain a sleep diary once a day when waking up. Per subject, this resulted in a total of seven (1 x 7) observations pertaining to sleep, and a maximum of 49 (7 x 7) observations for each of the ESM variables. The demographic information was collected before the start of the experiment.

Participants

This thesis was part of a larger study conducted by a PhD candidate, that researched the short-term effects of light exposure on burnout-related symptoms. Convenience sampling was used to select participants by asking students attending a master’s course in Performance Enhancement at the Eindhoven University of Technology to recruit one participant each, amounting up to 33 participants.

Furthermore, a group of four students from a research project recruited eight participants and a graduating masters student recruited ten participants. Participation in this study was voluntary and confidential. Participants were required to be 30 years or older and employed for at least four days a week. The total number of participants was 51, of which 33 were men and 18 women. The average age of the participants was 49.33 years (SD = 9.17), ranging from 30 to 65 years. Most participants (N = 40) had completed higher education, while the remaining participants (N = 11) had finished lower education.

The sectors in which participants worked varied from construction to transport, education, government, health, business, finance, communication and industry. Participants had to work 36.32 hours per week (SD = 5.40) on average according to their contract, ranging from 16 to 45-hour contracts. However, when asked how much the participants actually work this was 41.98 hours per week (SD = 9.96) on average, ranging from 16 to 70 hours per week.

Procedure

The study had a length of seven days, starting on a Saturday night, in which participants answered questionnaires and maintained a sleep diary. Participants used an ESM application on their smart phone or tablet to answer the questionnaires and keep their sleep diary. Before the first day of the study, the students met with their participants, who were asked to sign an informed consent form. After signing the informed consent form, the participants were provided with instructions about the activity watch and ESM app. Furthermore, during this same meeting, participants were asked to fill out an intake

questionnaire.

Once a participant started its seven-day period, monitoring took place and the participant received seven daily notifications for the ESM questionnaires. These notifications and the immediately following ESM questionnaires appeared at semi-random intervals during the waking day on the ESM app. The

notifications would appear on the smart phone and would have to be opened by the participant. If the participant failed to respond to the questionnaire notification the first time, they would receive a reminder after ten minutes and again after twenty minutes. If the participant failed to respond after

3 Research Method

(16)

both reminders, the questionnaire would expire and not be taken into account. Participants would also maintain a sleep diary each morning, where they were asked to fill in a short sleep diary following their night’s sleep.

Measures

Variables

Gender, age, education level, working hours, actual working hours and work sector were measured using one question, respectively. Work engagement does not seem to be significantly influenced by these variables, so therefore they were not included in the subsequent analyses (Becker, 2005).

Momentary job demands. Momentary job demands were measured using the Job Content

Questionnaire (JCQ; Karasek et al., 1998). There was one item where job demands were assessed with the question: “Right now, my activity requires that I work hard”. The response format ranges from 1 (I disagree) to 7 (I agree).

Momentary job autonomy. Momentary job autonomy was measured using the Questionnaire on the Experience and Evaluation of Work (QEEW; Van Veldhoven & Broersen, 2003). One item from the QEEW was used to assess momentary levels of job autonomy with the question ‘Right now, I can decide what I do’ and scored on a seven-point Likert scale (1 = I disagree, 7 = I agree).

Daily sleep quality. Sleep quality is a conglomerate variable which is researched in sub-variables.

Self-perceived sleep quality. One item was used to measure the self-perceived sleep quality of the participant with the question: “How would you rate the quality of your sleep?”. The response format ranges from 1 (very bad) to 5 (very good).

Self-perceived recovery. One item measured the self-perceived recovery of the participant. This was assessed with the item: “How rested do you feel?”. Here the response format ranged from 1 (not rested at all) to 5 (completely rested).

Observed sleep latency, observed sleep efficiency and observed total sleep. Data on sleep latency, sleep efficiency and total sleep were collected using the Actigraphy (activity watch), that measured physical activity at the wrist and allows for quantification of sleep-wake patterns. These measurements were therefore collected objectively. In the tested hypotheses where sleep quality was included (hypothesis 2 and 5), all the sub-variables were added separately in the model to include the full spectrum of sleep quality.

Momentary work engagement. It is assumed that the dimensions of burnout and work engagement are bipolar dimensions (Demerouti, Bakker, Vardakou, & Kantas, 2003; Demerouti, Mostert, & Bakker, 2010;

Demerouti et al., 2001; González-Romá et al., 2006). This assumption is also reflected in the OLBI where both positively worded items as well as negatively worded items are included. In order to be assess work engagement using the OLBI, the negatively frames items need to be recoded (Demerouti & Bakker, 2008;

Demerouti et al., 2003). Therefore, momentary levels of work engagement were assessed using the Oldenburg Burnout Inventory (OLBI; Demerouti, Bakker, Vardakou, & Kantas, 2003). The Utrecht Work Engagement Scale (UWES; Schaufeli et al., 2002) is a commonly used instrument for measuring work engagement. However, for this research the OLBI was used for assessing work engagement, because the OLBI essentially captures the same constructs as the UWES (Demerouti et al., 2010). Therefore, the OLBI is a reasonable alternative for measuring work engagement.

The two dimensions of the OLBI (dedication and vigor) were measured in two items each, of which one was asked in a reversed manner. This means that the dimensions of dedication and vigor had one positive and one negative question. In accordance with previous research, this thesis followed the same strategy of measuring work engagement. The two items used to measure the items of dedication were as

(17)

follows: (1) ‘At this moment, I am experiencing my activity as a real challenge’) where one was asked in a reversed manner (2) ‘I am sickened by my current activity’), and both were scored on a seven-point Likert scale (1 = I disagree, 7 = I agree). The two items used to assess vigor were as follows: (1) “I have enough energy for my current activity”, where one was asked reversed (2) “Right now, I feel emotionally drained”. The response format for these two questions were also scored on a seven-point Likert scale (1

= I disagree, 7 = I agree). Since research has shown that burnout and work engagement can be seen as each other’s opposites, the measured items were inverted so that they would represent the dimensions of dedication (opposite of disengagement) and vigor (opposite of emotional exhaustion) (González-Romá et al., 2006).

Questionnaires and measurement types

This research employed three different measurement types. The first type was the intake questionnaire, where the participant was asked on his demographics (age, gender, education level, weekly working hours on contract, actual weekly working hours, work sector and working days per week). Furthermore, the intake questionnaire measured the general work engagement levels using the Oldenburg Burnout Inventory (OLBI; Demerouti et al., 2003), the general job resources and demands using the Job Content Questionnaire (JCQ; Karasek et al., 1998), and the general sleep quality was assessed using the Pittsburgh Sleep Quality Index (PSQI; Buysse et al., 1989).

The second measurement type used was the wearable sensors. Participants wore two devices during the data collection period. The first device is a light logger that measures the amount of light a subject gets exposed to. The participant was therefore asked to wear the light logger as close to the eyes as possible.

The data collected by the light logger was not used in this research and will therefore not be mentioned further. The activity watch was the second device worn by the participants. This watch measured the physical activity at the wrist and allows for quantification of sleep wake patterns. Because it measured sleep wake patterns, it was necessary for the subject to wear the activity watch for the entirety of the experiment.

The third and final measurement type was the ESM items. Four ESM items measured momentary work engagement adapted from the OLBI (Demerouti et al., 2003) and two items adapted from the

Questionnaire on the Experience and Evaluation of Work (QEEW) were used to assess the current level of autonomy (Van Veldhoven & Broersen, 2003). In addition to ESM items, participants were asked to complete the Consensus sleep diary by Carney et al. (2012) every morning pertaining to the previous night’s sleep. An item on type of day (workday or work-free day for previous day) and use of alarm clock in the morning are added to the diary. The level of sleep quality was measured in the sleep diary with one item (‘How would you rate the quality of your sleep?’) and was scored on a five-point Likert scale (1

= very bad, 5 = very good).

Statistical Analyses

The collected data, obtained from the questionnaires and diaries, was first mapped in a dataset using Microsoft Excel software. Thereafter, the ESM variables used to test the hypotheses were divided into datasets and structured using SPSS software. SPSS was further used to do find the descriptive statistics of the variables and to transform the data, such that it could be used for the analyses. This transformation was needed so that the analyses software could read the data. In order to analyze the data, multilevel analyses were performed using MLwiN software (Rasbash, Browne, Healy, Cameron, & Charlton, 2000).

Moreover, for interpretation of the interaction effects for the moderator analyses, simple slopes analyses was conducted using Microsoft Excel software (Dawson, 2014; Dawson & Richter, 2006;

Toothaker, Aiken, & West, 1994). The collected data has a hierarchical structure with three levels where ESM variables measured at a moment were nested within days and days nested within persons.

Multilevel analysis was found to be the best way to analyze such interdependent observations using the

(18)

hierarchical linear modelling approach (Raudenbush & Bryk, 2002). A null model was developed to randomize intercepts at all levels. Independent variables were person mean-centered in order to avoid multicollinearity, which is especially valuable for measuring interactions between variables. (Bakker &

Bal, 2010). In order to test the hypotheses a model was created for each hypothesis separately. Before testing the models and analyzing the data, the data was checked for missing values and structured so that it would be appropriate to use. When creating the models in order to test the hypotheses, momentary work engagement was the construct variable and momentary job demands and job

autonomy were the explanatory variables. Sleep quality with underlying variables acted as a moderator in the models. Job autonomy also acted as a moderator in one of the models.

For hypothesis 1, the tested model (model 1) included the dependent variable work engagement, and the independent variable job demands both on a momentary level. For hypothesis 2, a model (model 2) including dependent variable work engagement and independent variable job demands both on

momentary level, but now also sleep quality was included on a day level. For the third hypothesis, job autonomy was the independent variable and direct effect on work engagement was assessed (model 3).

In hypothesis 4, the effect of the independent variable job demands on the dependent variable work engagement was measured with job autonomy as a moderator (model 4). Lastly, hypothesis 5 was tested by calculating the effect of the independent variable job autonomy on the dependent variable work engagement, with sleep quality as a moderator (model 5).

Data

Construct variable

Because the construct variable momentary work engagement is built up out of four items, it needed to be determined whether they could be taken together so that they would form one single construct.

Because the data is multilevel it was not possible to determine correlations with the raw data alone.

Moreover, since the data was multilevel and there were only four items, it was also not possible to perform a factor analysis or find reliable Cronbach Alpha’s. Therefore, in order to determine the

construct variable(s), the data needed to be altered in order to analyze the correlations of the variables, while respecting the multilevel nature of the data. The data was altered by first calculating the mean scores for each person (person mean) for DED1, DED2, VIG1 and VIG2. In order to calculate the within- person mean-centered scores, the person means were subtracted from all the raw scores resulting in N=1298 scores. Next, in order to calculate the person-centered scores, the person mean was taken once for every person (N=51). After these scores were calculated, the construct item correlations were calculated for the within-person (above the diagonal) and person centered (below the diagonal) scores.

The results are described in Table 1. These correlations show that there exists a very weak correlation between the items of dedication DED1 and DED2. Moreover, it shows a similar weak correlation between the items of vigor, VIG1 and VIG2. Other correlations were also too weak, which is why it was not

possible to combine items. Therefore, the four items together were considered to represent the construct work engagement.

(19)

Table 1: Means, standard deviations and correlations of construct items

Mean Standard

deviation

DED1 DED2 VIG1 VIG2

DED1 3.23 .86 - .17** -.07** .05

DED2 5.71 .66 -.07 - .21** .12**

VIG1 6.07 .67 -.32* .50** - .34**

VIG2 6.18 .76 -.32* .49** .64** -

** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed). Correlations below the diagonal are person-centered correlations (N= 51). Correlations above the diagonal are within-person mean-centered correlations (N=1298).

Null model

The null model is meant to provide a base model on which can be built further in the analyses, which includes momentary, day and person level random effects nested in a constant variable.

This means that only the dependent variables DED1, DED2, VIG1 and VIG2 were measured over the three levels. Mathematically, this model is described as follows:

𝐷𝐸𝐷1_!"# = 𝛽_$+ 𝑣_$#+ 𝑢_$"#+ 𝑒_$!"#

𝐷𝐸𝐷2_!"# = 𝛽_$+ 𝑣_$#+ 𝑢_$"#+ 𝑒_$!"#

𝑉𝐼𝐺1_!"# = 𝛽_$+ 𝑣_$#+ 𝑢_$"#+ 𝑒_$!"#

𝑉𝐼𝐺2_!"# = 𝛽_$+ 𝑣_$#+ 𝑢_$"#+ 𝑒_$!"#

Where:

In the model, 𝛽_! represents the intercept for the constant, 𝑣_!" represents person level variance, 𝑢_!#"

represents the day level variance and 𝑒!$#" represents the momentary level variance.

To determine whether multilevel analysis is appropriate for this data, two tests were performed. First, a test to see whether unexplained variance was present at all levels. For DED1 the null model showed 14.09% unexplained variance on the person level, 4.87% unexplained variance the day level and 81.04%

unexplained on the momentary level. For DED2 the null model showed 17.20% unexplained variance on the person level, 2.28% unexplained variance the day level and 80.52% unexplained on the momentary level. The null model for VIG1 showed unexplained variance of 28.69%, 4.14% and 67.18% on the person, day and momentary level, respectively. Finally, for VIG2 the null model showed 37.31% unexplained variance on the person level, 5.76% unexplained variance the day level and 59.93% unexplained on the momentary level. This means that the first test shows evidence for variances on multiple levels, which would suggest multilevel analysis would be appropriate.

The second test performed to determine multilevel analysis was, by determining if the three levels are also necessary in order to perform the analysis. By calculating the deviance scores (-2*log(lh)) on all three levels the chi-square difference (D-2*log(lh)) could be found. This D-2*log(lh) could then determine if there was a significant difference between the models of the different levels. The -2*log(lh) of the

𝑣$#~𝑁(0, 𝜎_%$^& ) 𝑢$"#~𝑁(0, 𝜎_'$^& ) 𝑒_$!"#~𝑁(0, 𝜎_($^& )

(20)

multiple levels were compared and subtracted in order to determine if multilevel analysis would be appropriate for this data. As can be seen in

Table 2, all the null models have significantly better models when fitted on three levels instead of one or two levels. This means that the second test also suggests multilevel modeling. Therefore, the analysis could be continued in a multilevel fashion.

Table 2: Deviance scores and Chi-square tests of null models

DED1 DED2 VIG1 VIG2

-2*log(lh) three levels 5310.38 4459.29 3760.98 3699.30 -2*log(lh) two levels 5363.64 4538.87 3894.30 3852.37 -2*log(lh) one level 5453.91 4627.47 4109.43 4175.65

D-2*log(lh) level two and three 53.26 79.57 133.32 153.08

D-2*log(lh) level one and two 90.27 88.60 215.13 323.28

P-value for all D-2*log(lh) <.001 <.001 <.001 <.001

(21)

Descriptive Statistics

The means, standard deviations and correlations of the observed variables are reported in

Table 3. In order to determine how the momentary variables correlate with each other, the within- person mean-centered correlations for all moments (N=1298) were assessed, which can be seen in Table 3 above the diagonal. Furthermore, the correlation between the four items of work engagement were insufficient. Below the diagonal, the variables were centered on the highest level (person level; N=51) where the momentary variables were person-mean scores, and the day level variables were raw scores.

However, a strong correlation exists between DED1 and job demands. There also seems to be a strong correlation between total sleep time and sleep efficiency.

Table 3: Means, standard deviations and correlations of the independent variables.

Mean Standard deviation

1 2 3 4 5 6 7 8 9 10 11

1. DED1 3.23 .86 - .18** -.07** .05 .73** -.39**

2. DED2 5.71 .66 -.07 - .21** .12** .14** .13**

3. VIG1 6.07 .67 -.32* .50** - .34** -.03 .11**

4. VIG2 6.18 .76 -.32* .49** .64** - .02 .05

5. Job

Demand 2.99 .91 .84** -.11 -.34* -.29* - -.39**

6. Job

Autonomy 5.15 .97 -.18 .04 .26 0.11 -.24 -

7. Sleep

quality 3.67 .78 .02 .23 .19 .42** .15 .08 -

8.

Recovery 3.47 .74 -.06 .04 .39** .28** -.05 .16 .51** -

9. Total

sleep time 368.95 111.65 -.18 .02 .12 -.01 -.17 .15 -.21 -.13 - 10. Sleep

latency 15.93 34.38 .03 -.16 -.13 -.13 .22 -.14 -.27 -.06 .20 -

11. Sleep efficiency

.78 .22 -.13 .01 .15 -.04 -.17 .04 -.14 .04 .83** .11 -

** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed). Correlations below the diagonal are person-mean centered for variables 1-6 and person centered for variables 7-11 (N= 51). Correlations above the diagonal are within-person mean-centered correlations (N=1298). Means and standard deviations for 9 and 10 are reported in minutes.

4 Results

(22)

Hypotheses

Hypothesis 1: Momentary job demands negatively affect momentary work engagement.

The first hypothesis stated that momentary job demands negatively affect momentary work engagement. Momentary levels of job demands were found to positively affect the two items of

dedication (DED1: g = .77, se = .02, p < .001; DED2: g = .10, p < .001). Moreover, momentary job demands were unrelated to items of vigor. Adding momentary job demands to the model including the intercept only, improved the model fit (DED1: D-2*log(lh) = 951.38, p < .001; DED2: D-2*log(lh) = 24.00, p < .001).

However, adding momentary job demands to the model including the intercept only for the items of vigor, did not improve the model fit. More momentary job demands were found to lead to more dedication (DED1 and DED2), which is the opposite of what was hypothesized. Thus, momentary job demands were found to lead to more dedication. Momentary job demands were not found to affect the items vigor (VIG1 and VIG2) in any direction. In sum, hypothesis 1 was not supported for any of the items of work engagement, and the results of the multilevel analyses are shown in Table 4.

Table 4: Multilevel analyses on the effects of job demands on work engagement on a momentary level.

DED1 DED2 VIG1 VIG2

Estimate SE Z value Estimate SE Z value Estimate SE Z value Estimate SE Z value Constant 3.27 .07 61.32** 5.69 .09 61.17** 6.04 .09 64.13** 6.13 .11 55.67**

Job

demands 0.77 .02 37.78** .10 .02 4.93** -.01 .02 -.89 .02 .02 1.32

-2*log(lh) 4359.00 4435.30 3760.20 3219.73

D -2*log(lh) 951.38** 24.00** .78 .07

Df 1 1 1 1

Level 3 intercept variance

(person) .64 .14 .00% .35 .09 .05% .40 .09 .00% .56 .12 .07%

(day) .07 .04 3.12% .06 .04 .00% .06 .02 .15% .09 .03 .00%

(moment) 1.48 .06 42.88% 1.62 .07 1.99% .93 .04 .00% .85 .04 .20%

Note: Model 1 was compared to the Null Model with the intercept as the only predictor, which can be found in Appendix 7.1. ** p<.001.

(23)

Hypothesis 2: Moderation of daily sleep quality on the relation between momentary job demands and momentary work engagement.

The second hypothesis stated that high sleep quality may buffer the effects that momentary job

demands have on momentary work engagement and low sleep quality may enhance the negative effects of momentary job demands on momentary work engagement. In the analyses, the four items of work engagement were compared to the five variables (self-perceived sleep quality, self-perceived recovery, observed total sleep time, observed sleep latency and observed sleep efficiency) of sleep quality. The moderating variables are discussed separately in this section.

Self-perceived sleep quality

The results of the analysis showed that self-perceived sleep quality of the previous night directly led to an increase in DED2 (g = .22, se = .06, p = .001) and VIG2 (g = .17, se = .05, p = .001). Results of the analysis showed no direct effect between self-perceived sleep quality, DED1 (g = .06, se = .26, p =.797) and VIG1 (g = .10, se = .05, p = .051). Self-perceived sleep quality did not affect the relation between momentary job demands and the items of dedication was found (DED1: g = -.02, se = .04, p = .564; DED2:

g = -.05, se = .04, p = .184). The moderation of self-perceived sleep quality also did not affect the relation between momentary job demands and the items of vigor (VIG1: g = -.05, se = .03, p = .109; VIG2: g = -.04, se = .03, p = .155). Adding self-perceived sleep quality to the model that only included the intercept and momentary job demands improved the model fit for DED2 (D-2*log(lh) = 122.31, p < .001) and VIG2 (D- 2*log(lh) = 98.77, p < .001).

Randomization of slope was performed for momentary job demands to find out if there was a within- person day-level effect of momentary job demands on DED2. Self-perceived sleep quality was found to moderate the relation between momentary job demands and DED2 (g = -.08, se = .04, p = .050). The slope variance also revealed that within days, momentary job demands led to more DED2 (s²= .03, se = .01 p = .006). When performing randomization of slope for momentary job demands, the model showed an improved fit (D-2*log(lh) = 3.71*) over the model without the interaction, but with the

randomization. The results of this randomization can be found in Table 5.

Finally, in order to interpret the interaction effect and direction of self-perceived sleep quality, a simple slopes test was performed. The results show that under the condition of ‘low’ self-perceived sleep quality of the previous night (one standard deviation below the sample mean), momentary job demands led to an increase in DED2 (t = 4.49, p < .001). In addition, results show that for ‘high’ sleep quality (one deviation above the sample mean), job demands did not relate to DED2 (t = 1.67, p = .118). Figure 2 shows that for ‘low’ sleep quality, DED2 increases as momentary job demands also increase. The results of the analysis performed for this hypothesis can be found in Appendix 7.2. Despite the fact that there was a positive moderation for the second item of dedication, this was in the wrong direction and only for one out of four items of work engagement. Therefore, for the moderator self-perceived sleep quality, hypothesis 2 was rejected for DED1, DED2, VIG1 and VIG2.

(24)

Table 5: Randomization of slopes in analysis of moderation of sleep quality of relation between job demands and DED2.

Base model Randomized model Interaction model

Estimate SE Z value Estimate SE Z value Estimate SE Z value Constant 5.69 .09 61.01** 5.68 .10 58.87** 5.68 .10 58.52**

Job demands .10 .02 4.86** .11 .02 4.46** .11 .02 4.51**

Sleep quality .22 .06 3.55** .18 .06 2.98* .22 .06 3.42**

Job demands x sleep

quality -.08 .04 -1.97*

-2*log(lh) 4312.98 4280.64 4276.93

D -2*log(lh) 122.32** 32.34** 3.71*

Df 1 1 2

Randomized slope

job demands .03 .01 .03 .01

Level 3 intercept

variance (person) .36 .09 .38 .09 .40 .09

Level 2 intercept

variance (day) .05 .04 .08 .04 .05 .02

Level 1 intercept

variance (moment) 1.60 .07 1.48 .07 .93 .04

Note: * p < .05, **p £ .001.

Figure 2: Simple slope of cross-level interaction between job demands, self-perceived sleep quality and DED2.