Having a good relationship with your boss : will it prevent a burnout or can it cause a burnout? : the curvilinear relationship between LMX and burnout and the mediating effect of role overload

(1)

Having a good relationship with your boss; will it

prevent a burnout or can it cause a burnout?

The curvilinear relationship between LMX and burnout

and the mediating effect of role overload

“

A person who feels appreciated will always do more than what

is expected

”

– Amy Rees Anderson, 2015 –

Executive Program in Management Studies Leadership & Management Track Amsterdam Business School January 2017 Cora van Triest (10901876) Supervisor: Dr. A.H.B. de Hoogh Second reader: Prof. Dr. D.N. den Hartog

(2)

Statement of Originality

This document is written by Student Cora van Triest who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

This study aimed to contribute to the literature on Leader-Member Exchange (LMX) by exploring its non-linear effects on the multidimensional construct burnout. We furthermore investigated whether role overload could explain the nature of this relationship. Data was collected in two Dutch organizations and a personal network, resulting in 346 respondents. Our findings showed that increases in LMX quality had very divergent effects on the three burnout dimensions. Unfortunately, this could not be explained by a mediating effect of role overload. Having a better relationship with one’s supervisor is in most cases beneficial for the prevention or reduction of burnout. For emotional exhaustion, the higher the LMX quality the better it was. The beneficial effect of LMX on depersonalization attenuated when the relationship’s quality became very high. The differences in having a good or very good relationship with one’s boss however, were trivial. For reduced personal accomplishment we found that with increasing LMX levels, the effects at first seem harmful, but above average LMX levels these effects appear to turn advantageous. In reducing or preventing burnout, it therefore seems better for employees to have an above average to high quality relationship with their supervisor with regards to all burnout dimensions.

Keywords: Leader-Member Exchange (LMX), Role overload, Burnout, Emotional

(4)

List of Figures

Figure 1: Conceptual model ... 27

Figure 2: The curvilinear relationship LMX - depersonalization... 44

Figure 3: The curvilinear relationship LMX - reduced personal accomplishment ... 45

Figure 4: Explored paths in mediation model ... 49

Figure 5: The curvilinear relationship role overload - emotional exhaustion ... 55

Figure 6: Linear relationship LMX – role overload ... 61

Figure 7: Curvilinear and linear relationships role overload – burnout dimensions ... 61

Figure 8: Linear and curvilinear relationships LMX – burnout dimensions ... 61

List of Tables

Table 1: Prior research on LMX, role stressors and burnout dimensions ... 11

Table 2: Coefficients and communalities for Principal Components Analysis of the role overload scale ... 34

Table 3: Means, standard deviations, correlations, and reliabilities ... 36

Table 4: Hierarchical regression for H1; curvilinear relation LMX - burnout dimensions ... 42

Table 5: Hierarchical regression for H2; role overload as predictor for the burnout dimensions .. 46

Table 6: Hierarchical regression for H3; curvilinear effect of LMX on role overload ... 48

Table 7: Mediated regression analysis for H4; role overload as mediator in the curvilinear LMX - burnout relationship ... 50

Table 8: Means and standard deviations for controls and variables in the different sample parts . 51 Table 9: Means, standard deviations, correlations, and reliabilities for sample 1 ... 52

(6)

Table 10: Means, standard deviations, correlations, and reliabilities for sample 2 ... 53 Table 11: Means, standard deviations, correlations, and reliabilities for sample 3 ... 53 Table 12: Additional hierarchical regression; curvilinear relation role overload - burnout

dimensions ... 54 Table 13: Hypotheses tested ... 56

Acknowledgements

This Master thesis is an important milestone for me in the completion of my Master in Management Studies at the University of Amsterdam. It has been a challenging yet very interesting two and a half years and I really enjoyed it. Both professionally and personally, I experienced very valuable developments. This all would not have been possible without the support from a number of people, who deserve a big thank you. First of all my supervisor dr. Annebel de Hoogh, who provided me with the perfect balance between advice, feedback, and challenges; she really boosted my self-confidence. Second, the management team of the Saxion Hospitality Business School, who gave me this opportunity. Then, of course, my family and friends who were very understanding and supportive in stressful times. Especially my partner Martijn showed a lot of patience and supported me all the way. I also wish to thank Jolanda Steunenberg-Wolters for contributing to my data collection and for the brainstorm sessions and feedback we exchanged. Furthermore, my colleagues Cindy Kuiper-Nijland, Rienk van Marle, and Geertje Tonnaer advised me in various stages of this research project and thereby facilitated important decisions and revisions. Last but not least, thank you to all the respondents who filled in my questionnaire!

(7)

7

1 Introduction

Organizations benefit highly from good relationships between employees and supervisors (Gerstner & Day, 1997). The quality of this relationship is called Leader-Member Exchange (LMX; Graen, Novak & Sommerkamp, 1982; Graen & Uhl-Bien, 1995; Liden, Wayne, & Stilwell 1993). Several factors influence this relationship, such as liking and expectations (Wayne, Shore & Liden, 1997). The quality of communication is important for the employee to know what is expected in the job and whether those expectations are being fulfilled. Furthermore, support from supervisors is firstly required during the socialization phase of new employees, and secondly throughout further employment in order to receive necessary resources for and feedback on the job-related activities and for developmental purposes. Having faith, trusting each other and feeling that you can rely on each other also are beneficial for an employee in performing the job. Thus, many positive effects are expected for employees who have a good relationship with their boss.

Gerstner and Day (1997) state that several individual, group, and organizational level outcomes can be predicted by the quality of the developed relationship between a leader and a follower. In their meta-analytic review of the LMX theory, they found significant relationships with for example satisfaction and commitment, performance and competence, turnover intentions, and role conflict and role clarity (Gerstner & Day, 1997). Interestingly, whether these

outcomes are always and only positive, is not unequivocal. Research seems to suggest mostly

positive outcomes of higher quality LMX, which is intuitively logical; the better the relationship with one’s supervisor, the better an employee seems to be informed and supported to fulfill the job requirements. Previous research has shown many cases of such positive outcomes (Burton,

(8)

8

Sablynski, & Sekiguchi, 2008; Harris, Wheeler, & Kacmar, 2009; Thomas & Lankau, 2009). The findings suggested that the higher the quality of LMX, the better it is for the employee and subsequently also the organization.

Few scholars have wondered however, if there only are positive outcomes of LMX or whether a good relationship with one’s supervisor can also have harmful effects. Besides, most relationships proven by previous research were linear and only a few studies examined possible non-linear effects (Harris & Kacmar, 2006; Hochwarter, 2005; Jian, 2012). The idea behind investigating non-linear effects may have been triggered by the fact that positive effects could also have been exaggerated. An employee whose relationship with the supervisor reaches high or very high quality, may feel extra pressure and can get used to working harder and taking on extra activities or responsibilities that were not required in the job description (Harris & Kacmar, 2006). As a result, positive outcomes like job satisfaction or employee well-being may decrease or fade away and eventually turn into negative outcomes like stress or burnout. This is in line with findings by Lawrence and Kacmar (2012), who state that with higher LMX job involvement increases and subsequently also stress increases. Hence, although higher LMX quality predominantly is seen as beneficial, it may not always have only bright sides.

Factors that may reduce stress levels or burnout risks have received a lot of attention in research and the impact of LMX herein has been investigated similarly (Harris & Kacmar, 2006; Hesselgraeves & Scholarios, 2014; Hochwarter, 2005; Jian, 2012; Lawrence & Kacmar, 2012; Sparr & Sonnentag, 2008; Thomas & Lankau, 2009). The present study proposes that LMX will be curvilineairly related to burnout and with the present study, we aim to investigate how this specific relationship can be explained through role overload, a specific role stressor. Thomas and Lankau (2009) found role stress to mediate the relationship between LMX (as one of the predictors of organizational socialization) and burnout, and indicated that role stress was one of

(9)

9

the most important burnout predictors. Furthermore, this study (Thomas & Lankau, 2009) measured burnout uni-dimensionally as emotional exhaustion, just like the studies by Harris and Kacmar (2006) and Lawrence and Kacmar (2012). Thomas and Lankau (2009) suggested future research that takes the multidimensionality of burnout (Maslach &Jackson, 1981) into account. They indicated the necessity to examine the effects of different LMX levels not only on the first burnout dimension emotional exhaustion, but on depersonalization and reduced personal accomplishment as well. Moreover, trying to disclose the underlying mechanism in the LMX – stress relationship will help explain why and how different levels of LMX quality can have an influence on preventing or developing stress. Harris and Kacmar (2006) state that LMX can also be “too much of a good thing” (p.65), and therefore they suggested future research on this phenomenon. They indicated the need for finding a mediator to reveal how exactly higher LMX quality may intensify stress levels.

This study proposes role overload to be a factor that can help explain the mechanism how LMX affects burnout in non-linear ways. Employees who have a better relationship with their boss feel more supported, which leads to less stress. Nonetheless, they also feel more challenged and are therefore tempted to exceed expectations by taking on extra tasks and responsibilities, which in turn can increase stress. This is in line with the two mediators in the LMX – stress relationship found by Lawrence and Kacmar (2012); role conflict and job involvement. These authors suggested to further investigate other role stressors like role overload as a mediator as well. Role overload as possible mediator in the LMX – burnout relation is logical because when employees are experiencing overloading roles, this could intensify stress levels. The perception of role overload is a natural reaction to higher job involvement as a result of trying to amount to the loyalty, affect, and contribution received from the supervisor when having a good relationship (Lawrence & Kacmar, 2012). Role overload, as one of the role stress components, has been

(10)

10

linked to both LMX and to burnout and stress (Bacharach, Bamberger, & Conley, 1991; Jian, 2012; Lawrence & Kacmar, 2012; Peiró, González-Romá, Tordera, & Mañas, 2001; Thomas & Lankau, 2009; Tordera, González-Romá, & Peiró, 2008). Considering the curvilineairity of the LMX – stress relationship found by Harris and Kacmar (2006) and the contribution of role stressors in explaining the underlying mechanism, we propose that role overload mediates the curvilinear LMX – burnout relationship. The curvilinear pattern is to be found in the first part of the mediation model; LMX is expected to be curvilineairly related to role overload, whereas role overload is positively related to the burnout dimensions emotional exhaustion, depersonalization, and reduced personal accomplishment. In sum, several relationships between LMX, role stress components, and burnout dimensions have been investigated; an overview of these previous studies is provided in Table 1.

(11)

11 Table 1: Prior research on LMX, role stressors and burnout dimensions Reference Independent variable(s) Mediator(s) Dependent variable(s) Linear? Remarks Bacharach, Bamberger & Conley (1991)

Role stress (conflict & overload)

Work-home conflict

Job satisfaction & burnout (dimension EE)

Linear Partial mediation

Burke & Greenglass (1995)

Work stressors (e.g. workload) & social support

- Burnout (dimensions EE, DP, RPA)

Linear Longitudinal

Burton, Sablynski & Sekiguchi (2008) Organizational justice (procedural, interactional, distributive) LMX Organizational Citizenship Behaviors (OCB’s)

Linear Full mediation for interactional;

LMX moderates distributive and procedural relationships with OCB’s.

Dulebohn, Bommer, Liden, Brouer & Ferris (2012)

LMX - Role states (role

ambiguity and conflict)

Linear Meta-analysis of antecedents and consequences of LMX

Harris & Kacmar (2006) LMX - Stress (job tension) Curvilinear Harris, Wheeler &

Kacmar (2009)

LMX - Job satisfaction,

turnover intentions & OCB’s

Linear Moderated by empowerment

Hesselgreaves & Scholarios (2014)

LMX Job demands Strain (job stressors) Curvilinear Linear for junior roles and curvilinear (U-shape) for senior roles;

Mediation for junior roles, not senior Hochwarter (2005) LMX - Job tension Curvilinear Moderated by negative and positive affect;

curvilinear only for low PA and high NA

Jian (2012) LMX - Role stress (conflict,

overload & ambiguity)

Curvilinear Curvilinear with conflict and overload, negatively linear with ambiguity

(12)

12 Reference Independent variable(s) Mediator(s) Dependent variable(s) Linear? Remarks

Lawrence & Kacmar (2012)

LMX Role conflict and job involvement

Stress Linear Partial mediation in one sample, full mediation in another sample

Lee & Ashfort (1993) Role stress Strain (Burnout dimension EE)

Burnout dimensions DP and RPA

Linear Longitudinal;

Role stress affects BO dimension EE, which in turn is positively related to BO dimensions DP and RPA.

Peiró, González-Romá, Tordera & Mañas (2001)

Role stress (conflict, ambiguity & overload) - Burnout (dimensions EE, DP, PA) Linear Longitudinal

Sparr & Sonnentag (2008) Fairness perceptions of feedback LMX Well-being (4 components) and perceived control at work Linear Longitudinal;

Partial mediation for job depression and turnover intentions;

Full mediation for job satisfaction and perceived control at work.

Thomas & Lankau (2009)

Social support (LMX & mentoring)

Role stress Burnout (dimension EE)

Linear Full mediation

Tordera, González-Romá, Peiró (2008)

LMX - Role overload Linear Moderated by Psychological climate; 3 out of 4 PC dimensions strengthened the relationship Wayne, Shore & Liden

(1997)

LMX Performance, OCB’s &

doing favors

Linear

(13)

13

What seems to be unanswered is whether a curvilinear effect of LMX on burnout will hold for all burnout dimensions. Furthermore, whether role overload as a specific role stressor can help explain the particular curvilinear pattern in the relationship also is not clear yet. This inspired the following research question, which we believe can add interesting new insights to the literature:

What are the effects of different levels of LMX quality on the three burnout dimensions (emotional exhaustion, depersonalization and reduced personal accomplishment), and to what extent can role overload help explain this possibly curvilinear relationship?

The remaining chapters of this study are structured as follows. An outline of the current literature and previous findings on LMX, specifically in relation to burnout and role overload, is provided in chapter two. Chapter two also includes the hypotheses and research model. Chapter three provides an overview of the data collection, the sample, and the measurement of variables used to perform the empirical part of this study. The results found during the analysis phase are described in chapter four, after which the implications and limitations are discussed in chapter five. This last chapter also provides suggestions for further research and the conclusion of this study.

(14)

14

2 Literature review

This study will examine the proposed curvilineair relationship between LMX and the three burnout dimensions (emotional exhaustion, depersonalization, and reduced personal accomplishment), and to what extent role overload mediates this relationship. The constructs LMX, burnout, and role overload will be discussed in this chapter and hypotheses will be proposed based on theory and previous findings. A research model depicting the variables and hypotheses will finalize this chapter.

2.1 LMX

Leader-Member-Exchange (LMX; Graen, Novak & Sommerkamp, 1982; Graen & Uhl-Bien, 1995; Liden, Wayne, & Stilwell 1993) refers to the quality of the relationship between an employee and his or her supervisor. A precursor of LMX theory is Vertical Dyad Linkage (VDL; Dansereau, Graen, & Haga, 1975), where supervisors are referred to as leaders and employees are referred to as members. In this theory, leaders were said to have different kind of relationships with the members in their units. This leads to an approach of in-group versus out-group with differing perceptions of relationship quality and fairness. The members need their leaders to help determine what roles they must fulfill in the organizations. The processes through which this is established differs per member because of the different kind of relationships they have with their leaders (Dienesch & Liden, 1986). A high quality of the relationship between employee and supervisor helps the employee interpret his or her role, as indicated by Dienesch and Liden (1986), who focus on the interpersonal exchange relationship. Liden and Graen (1980) also focused on the vertical exchange between leaders and members by looking at negotiating latitude. This exchange can differ among members with different quality of relationships with their leader,

(15)

15

according to VDL theory. Subsequently, any positive or negative consequences of this relationship quality will differ among employees as well.

Social Exchange Theory (SET) was an important foundation for the development of the LMX concept. When a social exchange relationship of high quality can be established and cultivated between an employee and the supervisor, leadership processes become effective (Graen & Uhl-Bien, 1995). According to Graen and Uhl-Bien (1995), there are three domains in leadership; the leader, the follower, and their relationship. Using the third domain, the relationship, as approach to leadership, the focal point is the dyadic nature of the relationship. The quality of this relationship can affect how an employee functions and feels at work. The difference in relationship quality can be explained as “downward influence and role-defined relations” for low LMX and “mutual trust, respect, liking, and reciprocal influence” for high LMX (Danserau et al., 1975, as cited by Liden, Wayne, & Stilwell 1993, pp. 662). In the same SET perspective, it can be said that the relationship between the employee and supervisor is built on the reciprocity between the involved parties. Cropanzano and Mitchell (2005) state that guidelines for this relationship are the “rules of exchange” (p. 875). The actual exchange deals with resources provided to- and efforts shown by the employee. However, the quality of the relationship refers more to the employee’s perception of the exchange of trust and commitment in both directions. This, in turn, helps to develop the quality of the relationship.

SET theory also was the foundation for Liden and Maslyn (1998) to emphasize the multidimensionality of the LMX construct. They state that in the exchange relationship, the non-tangible items that are exchanged comprise the following four dimensions. Contribution deals with resources and efforts that are contributed by both employee and supervisor. When these exchanges become greater or higher, the quality of the relationship increases. Loyalty refers to how loyal the employee and supervisor are towards each other and to what extent they support

(16)

16

the other. The third dimension affect relates to feelings of friendship and trust between employee and supervisor, and can influence the frequency of contact. When the parties in the relationship also like each other on a personal basis besides professionally, this increases the quality of the relationship as well. Professional respect refers to how impressed an employee is with the reputation of the supervisor, based on their professional knowledge and skills. Based on these theoretical foundations, scholars have examined numerous correlates for LMX.

Previous research has shown many cases of the positive outcomes of LMX. A study by Liden and Graen (1980) showed that employees, who experience relationships of better quality with their leaders, show more positive outcomes like higher performance, job responsibility, and contribution to team work. LMX can help explain the positive relationship between interactional justice and organizational citizenship behaviors (OCBs) and performance (Burton et al., 2008). Thomas and Lankau (2009) found evidence that supervisors high on LMX were sources of greater socialization and subsequently could help reduce role stress and diminish emotional exhaustion. The quality of LMX has a positive influence on outcomes like job satisfaction and performance, turnover intentions and OCBs, especially for employees low on empowerment (Harris, Wheeler & Kacmar, 2009). Much less attention has been paid to the question whether higher LMX quality always has the effect of diminishing negative employees’ outcomes or enhancing positive outcomes.

The question is, whether indications can be found that there is a point when the optimal level of LMX quality is reached and after this, the effect will turn in the opposite direction? Most relationships proven by previous research were linear and therefore suggested that the higher the quality of LMX, the better it was. Similarly, most studies linking LMX with outcomes that might trigger burnout risks such as stress have examined this relationship in linear ways (e.g. Thomas & Lankau, 2009; Lawrence & Kacmar, 2012). Nonetheless, Hochwarter (2005) found a curvilinear

(17)

17

relation between LMX and job tension, and Jian (2012) has shown the curvilinear relationship between LMX and role stressors. Interestingly, Hesselgreaves and Scholarios (2014) found that although LMX was linearly related to strain for junior employees, there was a curvilinear relationship for senior employees. After strain at first reduced with increasing LMX quality, it increased for even higher levels of LMX quality. A very specific example is the curvilinear LMX – stress relationship found by Harris and Kacmar (2006). In their study, they examined the linear or non-linear relationship between LMX and undesired outcomes at the individual level, like stress. They found that initially the increase of LMX quality would reduce stress levels, but after the quality of LMX increased from moderately high to high, more stress was perceived indicating a U-shaped form and thus indicating a curvilinear relationship. This was based on their expectation that there will be a moment after which high LMX benefits are no longer in balance with demands or felt obligations. This is in line with research by Wayne, Shore, and Liden (1997), who found that LMX is positively related to outcomes such as Organizational Citizenship Behavior and doing favors that benefits your boss. Both these outcomes suggest that employees with higher quality LMX are likely to do extra tasks, take on extra responsibilities, or work harder than expected in their job profile, resulting in perceptions of increased workload which can lead to stress or burnout.

2.2 Burnout

Freudenberger (1975), who worked as a psychiatrist in a clinic with volunteers, first identified burnout as a phenomenon in the 1970’s. He saw the volunteers getting burned out from their work and he therefore started writing about this. Burnout was further conceptualized by social psychologist Maslach and her colleagues (Maslach, 1976; Maslach & Jackson, 1981; Maslach, Jackson, & Leiter, 1997), who were intrigued by the question how emotional arousal on

(18)

18

the job affects people who work in ‘human services’. Burnout originated in the environment where employees work with human beings and has gained a lot of attention among researchers since then. The burnout syndrome is described by Schaufeli, Leiter, and Maslach (2009) as draining of energy and losing the ability to contribute to and accomplish the requirements of the job. The burnout syndrome has three dimensions; emotional exhaustion, depersonalization, and reduced personal accomplishment (Maslach & Jackson, 1981). Maslach and Jackson (1981) consider emotional exhaustion to be the key component of burnout. They say this is somewhat connected to depersonalization, which refers to one’s relation to other people with whom one interacts, also referred to as ‘clients’. The third component, reduced personal accomplishment, is related to oneself and has low correlations with the other two. According to Jackson, Schwab, and Schuler (1986), the three dimensions comprise the exhaustion that results from continuous psychological and emotional demands (emotional exhaustion), becoming detached from the people one works with and no longer seeing them as individuals but rather as objects (depersonalization), and no longer having belief in one’s efficacy to achieve results or contribute to goals (reduced personal accomplishment). Altogether, burnout is a complex syndrome that causes many unfavorable outcomes (Burke & Greenglass, 1995; Leiter & Maslach, 1988; Schaufeli & Buunk, 1996).

Many researchers see the three burnout components as mere dimensions that are distinctive and to a certain degree correlated, but it is also argued that the components function as stages in the process according to which burnout develops. Maslach already proposed that emotional exhaustion, depersonalization and reduced personal accomplishment follow each other in that specific order (Leiter & Maslach, 1988). One specific view on the development process of the burnout syndrome is that of Cordes and Dougherty (1993). They state that during the first stage, emotional exhaustion, perceived high demands from both the organization and an

(19)

19

employee’s personal beliefs can cause a feeling of exhaustion. Furthermore, the second stage in the burnout development process, depersonalization, is a response to emotional exhaustion. An employee tries to cope with this exhaustion by creating a distance towards recipients, becoming more cynical and adopting a more negative attitude. After reaching higher levels of depersonalization, employees can become so detached from their work that they no longer feel competent to fulfill the requirements of the job. This is the third stage of burnout; reduced personal accomplishment. Some studies suggest that the strain that results from role stress is emotional exhaustion and can be labeled as the first stage of burnout (Lee & Ashfort, 1993). As a result, the two other dimensions can occur at the same time. Depersonalization is then seen as a coping strategy to deal with the exhaustion, with reduced personal accomplishment as a result from both the other stages. Hence, it can be said that burnout development is a complicated process based on mutually influencing dimensions, rather than a mere stress syndrome.

Most research has been done on emotional exhaustion because it is said to be the first stage in the development process of burnout and therefore most variance and strongest correlates are found with this dimension (Lee & Ashfort, 1993; Peiró et al., 2001). This dimension is investigated most often and related to many different antecedents and outcomes (Bacharach, Bamberger & Conley, 1991; Thomas & Lankau, 2009). However, the two following stages in the process, depersonalization and reduced personal accomplishment are distinctive dimensions as well and worth investigating how they correlate with antecedents as well. Because the three dimensions are believed to be subsequent stages in the development process as a whole, it is very interesting to see whether correlates with one stage also correlate with the other stages, and how. These three different stages are all substantial components of burnout and therefore theory distinguishes them as such. Nonetheless, not much research has focused on these different stages as the whole burnout syndrome. We therefore specifically propose that LMX is curvilineairly

(20)

20

related to all three burnout dimensions. Emotional exhaustion is expected to decrease when the employee perceives more support from the supervisor, for example in the form of provision of resources. However, when more extreme levels of contribution or loyalty are perceived, the employee may try to reciprocate this and subsequently experience higher levels of emotional exhaustion. Depersonalization as a coping strategy for emotional exhaustion can at first reduce when more support is experienced, resulting in being more compassionate towards one’s clients. However, when experiencing greater support, taking on more challenging tasks can also result in taking more distance from the people one works with. Investing time and energy in new and more challenging skills and competences may occur at the expense of contact with clients, either voluntary or involuntary. Finally, perceptions of personal accomplishment initially grow when a supervisor repays an employee’s professional respect but can also decline with better relationship quality when the employees takes on tasks or responsibilities that are too demanding or transcend the employee’s capabilities. The burnout syndrome as a whole is therefore expected to be influenced by different LMX levels in non-linear ways.

Burnout involves employees’ wellbeing and stress levels, and these have been investigated in relation to LMX in previous research (overview in Table 1). Sparr and Sonnentag (2008) for example, found LMX to be a mediator in fairness perceptions of feedback

relationships with well-being at work and perceived control at work. Lawrence and Kacmar

(2012) showed that the influence of LMX on stress is mediated by role conflict and job involvement, and the effect of LMX on stress found by Harris and Kacmar (2006) indicated a U-shape relation where at first higher LMX decreased stress, but later increased it. In line with these findings, we expect to find LMX to influence the burnout syndrome as a whole in similar ways. The earlier indicated effects of LMX on emotional exhaustion are expected to hold for the other burnout dimensions as well, and are anticipated to be curvilineair relationships. In the same way

(21)

21

that emotional exhaustion first decreases and later increases with higher levels of LMX quality, depersonalization and reduced personal accomplishment can also be influenced in this curvilineair way. Accordingly, the first set of hypotheses to be tested in this study refers to the curvilineairity of the direct effects of LMX on the burnout dimensions. LMX is expected to be curvilineairly related to the three burnout dimensions emotional exhaustion (H1a), depersonalization (H1b) and reduced personal accomplishment (H1c), such that both low and high levels of LMX are expected to be associated with higher levels of the three burnout dimensions, whereas moderate levels of LMX are expected to be associated with lower levels of the tree burnout dimensions:

H1a: LMX is curvilineairly related to emotional exhaustion, such that across LMX levels

the negative relationship turns into a positive relationship (best represented as U-shape)

H1b: LMX is curvilineairly related to depersonalization, such that across LMX levels the

negative relationship turns into a positive relationship (best represented as U-shape)

H1c: LMX is curvilineairly related to reduced personal accomplishment, such that across

LMX levels the negative relationship turns into a positive relationship (best represented as U-shape)

2.3 Role overload as mediator

When an employee experiences stress because of the perception of having a lack of resources to fulfill the demands of the job, this can be considered role overload (Kahn, Wolfe, Quinn, Snoek, & Rosenthal, 1964). Sales (1969) examined role overload in relation to health

(22)

22

risks (coronary disease) and described it likewise as not having enough available time to fulfill one’s obligations. In a more recent study by Tordera, González-Romá and Peiró (2008), role overload is characterized as one of the fastest growing and most important type of role stressors. Role overload is seen as a component of role stress. Role stress originally consisted of role conflict and role ambiguity (Jackson & Schuler, 1985), and Peterson et al. (1995) added role overload as distinctive third component. Newton and Keenan (1987) define it as quantitative role overload, referring to the amount of work that one should apply at the job. Role overload can be considered a possible antecedent for burnout (Bacharach, Bamberger & Conley, 1991; Burke & Greenglass, 1995; Peiró et al., 2001).

Experts differ in opinion about which of the three burnout stages result from overburdening roles (Lee & Ashforth, 1993; Peiró et al., 2001). Roles that are perceived to be too heavy can lead to feelings of depletion and fatigue as well as taking more distance from work and its recipients and becoming more cynical. Moreover, losing confidence and the belief in one’s own competence is not improbable as concomitant as well. When experiencing higher demands in the job than the recognized resources available, employees are likely to perceive more exhaustion, may lose confidence in their own competence, or try to cope with it by taking more distance (Burke & Greenglass, 1995). Role overload has been proven by Peiró et.al (2001) to be positively related to two of the burnout dimensions; emotional exhaustion and depersonalization. For the third dimension reduced personal accomplishment, they found that it could only be predicted by role ambiguity and not by role overload. In this study, we propose that reduced personal accomplishment can also be influenced by role overload. Employees who experience role overload will not only become emotionally exhausted and try to cope with this by taking more distance. They will lose confidence in their capabilities as well, as a result of the lack of

(23)

23

control over their tasks and responsibilities. We therefore consider the burnout syndrome as a whole to be affected by role overload.

To conclude, the second set of hypotheses to be tested in this study concerns the effect of

role overload on the burnout dimensions.Specifically, role overload is expected to be positively

related to the three burnout dimensions emotional exhaustion (H2a), depersonalization (H2b) and

reduced personal accomplishment (H2c):

H2a: Role overload is positively related to emotional exhaustion

H2b: Role overload is positively related to depersonalization

H2c: Role overload is positively related to reduced personal accomplishment

Additionally, it is also logical to link LMX with role overload because, as rooted in social exchange theory, higher quality LMX relationships can be expected to include better communication and more perceived support and mutual trust. As a result, employees will be less likely to experience overloading roles when LMX quality increases. Dulebohn, Bommer, Liden, Brouer, and Ferris (2012) showed a significant negative relationship between LMX and role ambiguity and role conflict, but did not take role overload into account. Nonetheless, with higher LMX quality, employees will presumably also receive more challenging tasks from the supervisor in exchange for the efforts of the employee. This in turn may result in perceptions of heavier workload when the employees aim to fulfill everything they feel is expected of them. Therefore, role overload might eventually increase after it has initially decreased because of better quality of the relationship. For example, employees who are very loyal to their supervisor might feel more pressure and are more tempted to want to live up to the expectations and fully embrace the offered challenges which can result in working harder and trying to constantly

(24)

24

exceed those expectations. This can, in turn, lead to more perceived role overload for those with the highest levels of LMX quality. Previous studies showed reasons to believe there are non-linear effects of better LMX quality (Harris & Kacmar, 2006; Hesselgreaves & Scholarios, 2014; Hochwarter, 2005; Jian, 2012). Scandura (1999) found that employees with high LMX quality show higher performance, whereas low-LMX employees only perform at a level that is appropriate based on the rewards they receive. This may subsequently result in higher levels of role overload for high-LMX employees. Hence, the bright side of a good relationship with one’s boss turns detrimental.

Based on the assumption that better LMX quality can trigger employees to work harder and as a result perceive more role overload, we assume that relationships between LMX and role overload may not be linear. Wayne, Shore, and Liden (1997) found support for their hypotheses that LMX would be positively related to OCB’s and doing favors for the supervisor by the employee. This suggests that employees with higher LMX quality are likely to be satisfied and feel they can handle extra tasks, but are also likely to engage in activities or responsibilities beyond their job description which in turn might cause feelings of overload. This also supports our expectation that LMX will be related with role overload in a curvilinear way. Not much has been investigated in regards to possible non-linear effects of LMX on role overload. As indicated in Table 1, only a few studies found curvilinear effects of LMX on role overload (Jian, 2012) and on job tension (Hochwarter, 2005). Whereas higher quality of the relationship may ultimately lead to higher levels of role overload, incorporating this concept may also resolve the question in what ways LMX influences the burnout process.

It is commonly known that overloading roles are related to higher employee stress levels. If people experience a workload that is too heavy for them, or they feel they are not provided with enough resources to carry out their tasks, this may cause stress and harm their feelings of

(25)

25

well-being and induce health risks. Burnout is said to be caused by, among others, high demands in the job and a lack of resources to deal with these demands as shown by several studies in this field (Bacharach, Bamberger & Conley, 1991; Burke & Greenglass, 1995; Peiró et.al, 2001). Some studies only examined the effects of burnout antecedents such as role overload on one dimension, emotional exhaustion, since this is the first stage of burnout. However, the other burnout stages can be influenced in similar ways as well. This was proven by Peiró et al. (2001), who found that role overload, as part of role stress, could predict both emotional exhaustion and depersonalization but not reduced personal accomplishment. The latter burnout dimension was only predicted by role ambiguity. In a review of literature on burnout, Cordes and Dougherty (1993) propose a framework that indicates that components of role stress are important antecedents in the development process of burnout. They state that role overload is a demand stressor that influences emotional exhaustion and reduced personal accomplishment. Furthermore, organizational and personal social support as coping resources are presented as moderators in the stages of the burnout development process, which can be related to perceived levels of relationship quality with supervisors.

In combination, the previous findings indicate that role overload can explain the underlying mechanism in the relationship between LMX and the burnout dimensions and above all, indicate where the expected curvilinear pattern appears. This effect is expected because positive effects of LMX quality can result in employees’ behavior evolving into exaggerating their efforts and therefore higher stress levels. At first, when LMX quality gets higher, employees will feel more supported and get more or better feedback which results in less role stress. However, these positive outcomes can turn around when the employee also feels more challenged and keeps on trying to meet even higher expectations and continues to work harder and take on supplementary tasks. This may result in role overload and subsequently in higher burnout levels.

(26)

26

This is in line with previous research on the nonlinear relationship between the quality of LMX with role overload by Jian (2012). Moreover, several studies also suggested that role stress can increase levels of stress or burnout (Thomas & Lankau, 2009; Peiró et al., 2001; Bacharach et al., 1991). Thomas and Lankau (2009) only used the two role stress components conflict and ambiguity as mediators in the relationship between LMX as one form of social support and emotional exhaustion as key component for burnout. Building on their design, including the role stressor overload as well as all three burnout dimensions, will add interesting new insights to the literature and previous findings.

To summarize, role overload is expected to help explain how LMX influences the three burnout dimensions in curvilineair ways, in line with the findings by Harris and Kacmar (2006). This is stated in the third hypothesis and fourth set of hypotheses to be tested in this study. Hypothesis H3 states that the curvilineair pattern (U-shaped across different levels of LMX) that is expected to exist between LMX and role overload appears in the first part of the mediation model:

H3: LMX is curvilineairly related to role overload, such that across LMX levels the

negative relationship turns into a positive relationship (U-shape)

Subsequently, for the second part of the mediation model and as stated in Hypothesis H2, role overload is expected to have a positive relationship with the three burnout dimensions. The last set of hypotheses refers to role overload explaining the process in which LMX affects the burnout dimensions in a curvilinear way:

(27)

27

H4a: Role overload mediates the curvilineair relationship between LMX and emotional

exhaustion

H4b: Role overload mediates the curvilineair relationship between LMX and

depersonalization

H4c: Role overload mediates the curvilineair relationship between LMX and reduced

personal accomplishment

2.4 Research model

The conceptual model for this study is shown in Figure 1. It depicts the independent variable LMX, the three burnout dimensions as dependent variables, the mediator role overload, and the hypotheses to be tested.

(28)

28

3 Data and method

The following chapter describes the empirical part of this study. The data collection procedure and the sample collected are discussed, after which the measurement of variables and

statistical procedure are explained.

3.1 Sample and procedure

Data was collected using an online questionnaire through Qualtrics. The sampling technique was mainly non-probability, because no sampling frame for the whole population was available. The questionnaire was distributed among three different target groups. First, in the researcher’s own organization (sector education & academic research), sampling frames were available for three departments, and all these employees were invited to participate through a combination of personal email and staff newsletters (response rates 48%, 42%, and 26%). For the other departments, convenience and snowball sampling were used through personal emails. A total of 178 respondents filled in the questionnaire in this first part of the sample between 22 September and 7 November (243 employees started the questionnaire; response rate = 73%). The second part of the sample was a company in the sector business services & consultancy. Via the HR manager, all 237 employees of 3 business units were invited to participate through a personal email after the survey was announced in the staff newsletter. From the 142 employees who started the questionnaire, 119 respondents finished it (response rate 50% from sampling frame or 84% from the 142 employees who started the questionnaire) between 6 October and 7 November. The third part of the sample consists of individuals in the researcher’s own personal network that were targeted through email and social media on the basis of convenience, self-selection, and snowball sampling between 22 September and 23 October. From the 65 people who started the

(29)

29

questionnaire, 50 have finished it (77%). For collecting data from individuals in organizational research, an average response rate of 52.7% was found among 1607 studies in 17 refereed academic journals (Baruch & Holtom, 2008). Compared to this, the response rate in this study is reasonably good.

The English and Dutch version of the general questionnaire can be found in Appendices I and II. For the two specific organizations, an extra question was added to each of the questionnaires to ask for the respondent’s department within that organization. In total, the sample consisted of 347 employees but with one questionnaire terminated at an early stage, the actual data set contains 346 cases.

3.2 Measures

The three constructs from the conceptual model; LMX, role overload, and burnout are translated into five variables because burnout is measured as three separate dimensions according the research question and hypotheses. All constructs are measured as self-report by the employees, which means that the respondents filled in their own perception of LMX, role overload, and burnout levels. The following section will describe the variables and measurements used in this study.

LMX

LMX is operationalized in this study as the quality of the relationship between an employee and his or her supervisor. The commonly used scale by Liden and Maslyn (1998) is used and consists of the four dimensions contribution, professional respect, affect, and loyalty. Contribution deals with resources and efforts that are contributed by both employee and supervisor. When these exchanges become greater or higher, the quality of the relationship

(30)

30

increases. Loyalty refers to how loyal the employee and supervisor are towards each other and to what extent they support the other. Although affect may play a smaller role in the exchange relationship, this third dimension is important because it can influence the frequency of contact and feelings of friendship and trust between employee and supervisor. When the parties in the relationship not only professionally, but also personally, like each other; this increases the quality of the relationship as well. Professional respect relates to how impressed an employee is with the reputation of the supervisor, based on professional knowledge and skills. With three items each, this results in a twelve-item scale. These items are statements, measured on a 7-point Likert scale ranging from 1 – strongly disagree to 7 – strongly agree. Higher scores express an employee’s perception of better quality of the relationship. An example item is “I do not mind working my hardest for my supervisor”. With a Cronbach’s α of .889, this scale has a good reliability.

Role overload

Role overload is explained as a shortage of an employee’s resources to meet the requirements of the job (Peterson et al., 1995), and was added as a third component to the role stress construct that until then consisted of role conflict and role ambiguity (Kahn et al., 1964). Most scales for role overload only consist of three to five items and have both overlapping and distinctive items. For the present study, we combined items from different scales (Brown, Jones & Leigh, 2005; Ho, Ang & Straub, 2003; Perrewé et al., 2005) for the measurement of the role overload construct. We chose a combination of seven items that, in our opinion, represents the best suitable scale to measure role overload in relation to the other constructs in this study. The basis consisted of the original four items from Brown, Jones, and Leigh (2005) who used the conceptual definition by Kahn et al. (1964) and previous research by House (1980) and Singh (2000). Two additional items were added from Ho, Ang, and Straub’s scale (2003), and one last

(31)

31

item was added from the scale by Perrewé et.al. (2005). Appendix IV provides an overview of more detailed information on this scale adaptation and all selected items.

The Cronbach’s α is for the present study’s role overload scale is .893. The corrected item-total correlation was between .575 and .816 for the seven items, and no single item would improve the scale’s reliability when deleted. All role overload items in our adapted scale are measured on a 7-point Likert scale ranging from 1 – strongly disagree to 7 – strongly agree. An example item is “I do not have enough help and resources to get the work done”. Higher scores refer to higher perceptions of role overload for most of the items. One item is counter-indicative (“The amount of work I am asked to do is fair”) and for this item, a higher score reflects a lower role overload perception.

Burnout

Schaufeli, Leiter and Maslach (2009) describe burnout as draining of energy and losing the ability to contribute to and accomplish the requirements of the job. According to Maslach and Jackson (1981), it is a syndrome consisting of the three dimensions emotional exhaustion, depersonalization and (reduced) personal accomplishment. Measurement of burnout scores is mostly conducted by using the Maslach Burnout Inventory (MBI) developed by Maslach and colleagues (Maslach, et al., 1997; Maslach & Jackson, 1981). The MBI scale is used in the present study to measure burnout and consists of 22 items in total. All items in the burnout scale are measured as statements on a 7-point Likert scale ranging from 1 – never to 7 – always.

Emotional exhaustion is the first burnout dimension and is defined by Jackson, Schwab,

and Schuler (1986) as the exhaustion that results from continuous psychological and emotional demands. This dimension consists of nine items (with a Cronbach’s α of .887), of which one example item is “I feel emotionally drained from my work”. The second dimension,

(32)

32

depersonalization, refers to the relationship with others in the sense that one becomes detached

from the people one works with and no longer sees them as individuals but rather as objects (Jackson, Schwab & Schuler, 1986). The depersonalization dimension has a Cronbach’s α of .775, and has five items. An example item is “I don’t really care what happens to some recipients”. For these first two dimensions, a higher score means a higher burnout perception. With eight items, personal accomplishment is the third dimension (Cronbach’s α = .869). This dimension is related to oneself and refers to having belief in one’s own efficacy to achieve results or contribute to goals (Jackson, Schwab & Schuler, 1986). This dimension is reverse-coded in the MBI, and higher scores therefore relate to lower perceived levels of burnout. An example item is; “I deal very effectively with the problems of my recipients”. In the present study, the items in this dimension are recoded to represent the fifth variable reduced personal accomplishment.

Control variables

Finally, a number of control variables were included. Literature and results from previous research (Cordes & Dougherty, 1993; Marginson, 2006; Pretty, McCarthy & Catano, 1992; Schaufeli & Buunk, 1996) indicated that gender, age, and organizational tenure are known to be correlated to the main variables of interest; role stressors and burnout.

3.3 Statistical procedure

Data sets from the three samples in Qualtrics were imported in the Statistical software Package for Social Sciences (SPSS, version 22), and merged to one sample. Frequency checks were run to check for errors, missing values or outliers. One case was terminated after a few questions and therefore excluded from further analyses. In the items of all five main variables, no errors or missing data were found. In the control variables, some missing values were found for

(33)

33

age (1%) and tenure (9%). In age, there was one error where a respondent filled in 99, but no further errors or missing values occurred. The next step was to recode counter-indicative items. In the role overload scale, one of the seven items was reverse-coded. On this item, “The amount of work that is asked of me, is fair”, a higher score represents lower perception of role overload, whereas the other six items all represent higher role overload perceptions when scores are higher. For this reason, item rRO_7 was recoded into a new item; RO_7. All items in the third BO dimension (RPA) were counter-indicative compared to the other two dimensions, since it measured personal accomplishment with higher scores indicating less burnout. Therefore, all eight items were recoded into reduced personal accomplishment, in order to be able to relate the independent and mediator variables to this dependent variable in the same way as to the other two burnout dimensions.

For the LMX and three burnout dimension constructs, validated scales have been used and the reliabilities for these scales were good, given de Cronbach’s α provided in the measures section of the previous chapter. For the role overload construct however, an adapted scale was used which was a combination of items from different scales (see appendix IV). Therefore, a factor analysis (Principal Components Analysis with a fixed number of factors) was conducted in order to check whether the chosen items represented the underlying phenomenon ‘role overload’. Data was suitable for a factor analysis (Pallant, 2013), since the sample size is big enough (N = 346) and the intercorrelations among items were all above .30 (ranging from .399 to .689). The Kaiser-Meyer-Olkin test result was above .60 (.886) and the Bartlett’s Test of Sphericity was significant at the <.01 level. One component had Eigenvalue > 1, explained 61% of the variance, and also the scree plot confirmed one factor for the items in this scale. The loadings of the items on the component are quite strong, as shown in Table 2. With a Cronbach’s Alpha of .893, the internal consistency reliability of the role overload scale is considered very good. The corrected

(34)

34

item-total correlation of all items above .30 indicates good correlation with the scale’s total score. Furthermore, no single item would improve reliability of the scale when it was deleted.

Table 2: Coefficients and communalities for Principal Components Analysis of the role overload scale

For each of the variables in the conceptual model, scale means were computed by calculating the mean score of all items belonging to a scale. For an overview of the items in the different scales, please refer to appendix III for LMX, appendix IV for role overload and appendix V for burnout. Next to the linear LMX scale mean, a squared LMX term was calculated to enable us to test curvilineairity.

Component 1

Commu-nalities

RO6 My workload feels too heavy .878 .772

RO1 The amount of work I have to do interferes with how well the work gets done .808 .652 RO2 I do not have enough time to get the work done .800 .641 RO3 I do not have enough help and resources to get the work done .799 .638 RO4 I have to try to satisfy too many different people .774 .598 RO5 I have been given too much responsibility .712 .507 RO7 The amount of work that is asked of me, is NOT fair (recoded) .679 .462

(35)

35

4 Data analysis and results

This chapter first provides the descriptive statistics of the sample and comments on the distribution of the data. After that, statistical analyses that were undertaken are explained, and results of the model analysis are described.

4.1 Preliminary analyses Descriptive statistics

The 346 employees in the sample have either a fulltime (57%) or part-time (43%) job, and 96% of the respondents have Dutch nationality. The gender distribution is 39% male and 61% female, and the average age is 43 within a range from 18 to 65. The educational level of the largest part of the sample (43%) is undergraduate (HBO), followed by graduate (WO, 33%) and vocational education (MBO, 16%). The remaining respondents have secondary education, PHD, or other education. Respondents have worked with their organization between 0 and 44 years with a mean of 9.5, and under their supervisor for an average of 2.4 years within a range from 0 to 20 years. Most employees perceive the relationship with their supervisor as far better than average on a 7-point Likert scale for LMX (M = 5.15; SD = .90). Role overload is reported as almost average (M = 3.97; SD = 1.29) and burnout levels vary from rather low to rather below average (emotional exhaustion M = 2.32; SD = .84, depersonalization M = 1.71; SD = .73, reduced personal accomplishment M = 3.15; SD = .89). Because means for LMX are rather high and for the burnout dimensions rather low, the total sample was split to see whether the means and SD’s deviated between the different parts of the sample. Results of these additional descriptive statistics can be found in section 4.5.1. All means, SD’s, intercorrelations, and reliabilities of the total sample can be found in the correlation matrix in Table 3.

(36)

36

Table 3: Means, standard deviations, correlations, and reliabilities

From the control variables, only age shows a significant correlation (p < .01) with two burnout dimensions, but not with the other main variables. There is also no correlation between the other controls and main variables of interest. Age has a small negative correlation with depersonalization (r = -.246) and reduced personal accomplishment (r = -.252), indicating that older people perceive slightly lower scores on these burnout dimensions. All main variables in the conceptual model have significant correlations (p < .01) among each other. LMX has a small negative correlation with role overload (r = .162) and all three burnout dimensions (r = .219; -.278; -.271), which implies that with a better employee – supervisor relationship, perceptions of role overload and burnout become somewhat lower. Role overload has a small positive correlation with reduced personal accomplishment (r = .264), a moderately positive correlation with depersonalization (r = .370) and a strong positive correlation with the first burnout dimensions emotional exhaustion (r = .540). This demonstrates that higher role overload scores are related to higher burnout scores, especially in the first dimension emotional exhaustion. Within the burnout dimensions, there is also a strong positive correlation between emotional exhaustion and depersonalization (r = .525) , and a moderate positive correlation between both

M SD 1 2 3 4 5 6 7 8

Controls

1. Age 42.78 11.10

-2. Gender (1 = male, 2 = female) 1.61 0.49 -.138*

-3. Tenure (years) 9.55 8.80 .566** -.108 -Variables 4. LMX quality 5.14 0.91 -.023 -.070 .003 (.89) 5. Role Overload 3.97 1.23 -.075 .043 -.038 -.162** (.89) 6. Emotional Exhaustion 2.32 0.84 -.105 .102 .001 -.219** .540** (.89) 7. Depersonalization 1.71 0.73 -.246** -.077 -.107 -.278** .370** .525** (.78)

8. Reduced Personal Accomplishment 3.15 0,89 -.252** .067 -.045 -.271** .264** .348** .397** (.87)

Note: N = 346, except for age N = 342, tenure N = 326. Reliabilities are reported on the diagonal.

* Correlation is significant at the 0.05 level (2-tailed). ** Correlation is significant at the 0.01 level (2-tailed).

(37)

37

emotional exhaustion (r = .348) and depersonalization (r = .397) with reduced personal accomplishment. It seems that, in line with the literature, higher scores of one burnout dimension correspond with higher scores in the other dimensions (Cordes & Dougherty, 1993; Lee & Ashfort, 1993).

The correlations found between the main variables show that the hypotheses in this study are worth investigating. Because correlations are not too high (above r = .8 or .9), no multicollinearity issues are expected (Pallant, 2013). Because the control variables did not all show correlations with the main variables, it was first investigated whether they were worth integrating in the model. We therefore examined whether, when included in the analyses, the controls significantly influenced the results. A number of pre-tests (some of the hypotheses to be tested) have been conducted for a comparison between a) regression models with all three controls, b) only age as control (the only control with significant correlation), and c) no controls. This showed us differences in both significance levels of the models and in percentages of variance explained. In one case, it mattered for the hypothesis to be supported or not. This indicated that the controls do have an influence and therefore they will be included in further analyses.

Distribution of data

Normality of the data was checked visually (histograms with normal curve and Q-Q plots), and also calculated (skewness and kurtosis statistics). Data for role overload and reduced personal accomplishment can be considered as normally distributed. In the data for the other variables, some normality issues were detected. The results of normality checks and the issues found will be explained in the following part of this section.

(38)

38

LMX data was not normally distributed, but for most of the items negatively skewed. This means that higher LMX scores were more frequent than lower LMX scores, implying that the larger amount of employees perceive a better quality relationship with their supervisors. Two items have a normal distribution, whereas four items have a moderate negative skewness (between -.5 and -1), five items have substantial negative skewness (between -1 and -2), and one item is extremely negatively skewed (-2.107). Three items have a very small or small negative kurtosis, so distribution is a bit flatter than normal. All other items have varying positive kurtosis, indicating a more pointy distribution. The computed scale mean for LMX has a skewness of -1.088 and kurtosis of 1.650.

For role overload, data is rather normally distributed because skewness is close enough to zero (between -.5 and .5 for all seven items) and kurtosis is only slightly negative (between -.7 and 1.1) indicating a somewhat flatter than normal distribution (skewness .027 and kurtosis -.664 for the computed scale mean).

Emotional exhaustion data (9 items) is not normally distributed. On three items data is moderately positively skewed (between .5 and 1); whereas six items are substantially positively skewed (between 1 and 2). Kurtosis is close to zero for two items and the other items have fluctuating positive kurtosis ranging between .563 and 6.097. For the computed scale mean, skewness is 1.198 and kurtosis is 2.403. Depersonalization (5 items) is also not normally distributed because skewness is substantially (4 items between 1 and 2) or extremely (1 item 2.488) positive. The larger part of the sample scored in the low range on the depersonalization items. Distribution is also very pointy, since kurtosis ranges between 2.054 and 9.398 (with 1.942 skewness and 7.076 kurtosis for the computed scale mean). The positive skewness in the first two burnout dimensions means that more respondents reported lower scores on these dimensions, indicating lower burnout levels. For reduced personal accomplishment, data is rather normally

Having a good relationship with your boss : will it prevent a burnout or can it cause a burnout? : the curvilinear relationship between LMX and burnout and the mediating effect of role overload