Leader narcissism and follower job performance: a
matter of different dimensions or curvilinearity?
A two-fold study analyzing the mediating role of leader-member-exchange in the relationship between two distinctive dimensions of leader narcissism and follower job performance and analyzing the
curvilinear effect of leader narcissism on follower job performance
By Cécile Jansen University of Amsterdam
Faculty of Economics and business Master: Business Administration Track: Leadership & Management Student number: 10146555
Student email: cecile.jansen@student.uva.nl Supervisor: dr. Annebel de Hoogh
Second reader: prof. dr. Deanne den Hartog Date: 22-06-2016
Statement of originality
This document is written by Cécile Jansen 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.
Abstract
Despite a growing body of research on leader narcissism and organizational outcomes, no consensus has been reached on the effects of leader narcissism in the organization. Especially research on its effects on follower job performance, including in-role behavior and extra-role behaviors, such as affiliative OCB and challenging OCB, is scarce in the literature. Also, leader narcissism has clear relational implications (Owens, Wallace, Waldman & David, 2015), however no study to date included relationship-based constructs in this relationship. The inconsistent findings may be due to the existence of multiple dimensions of narcissism or the hidden curvilinear effect of leader narcissism on organizational outcomes. The present study aims to tackle these inconsistent findings in two ways: 1) to test whether leader-member-exchange (LMX) mediates the relationship between two distinctive dimensions of narcissism, Admiration and Rivalry, and follower job performance, and 2) to analyze the curvilinear effect of leader narcissism on follower job performance. Data was collected from followers and leaders working in different Dutch and non-Dutch organizations, which resulted in 123 dyads. Results showed that Rivalry was negatively related to LMX, while Admiration was unrelated. Also, LMX was positively related to all three follower job performance variables. Subsequently, Rivalry had a negative relationship with follower job performance, which is mediated by LMX. Curvilinear analysis of leader narcissism on follower job performance revealed no significant effect. Overall this means that Rivalry has far-reaching negative interpersonal consequences for follower job performance.
Recommendations for future research regarding leader narcissism and follower job performance are given.
Table of Content
1. Introduction 6
2. Theoretical background 9
2.1 Multidimensionality of leader narcissism 9
2.2 Admiration, Rivalry, and its interpersonal consequences 12 2.3 Admiration, Rivalry, and its interpersonal consequences for
follower job performance 15
2.4 Curvilinear research on leader narcissism 17
3. Method 20
3.1 Participants and procedure 20
3.2 Measures 22
3.3 Control variables 23
4. Results 24
4.1 Preliminary tests and descriptive statistics 24
4.2 Results study 1 26
4.2.1 Hypothesis 1a + 1b 27
4.2.2 Hypothesis 2a, 2b, and 2c 29
4.2.3 Hypothesis 3a, 3b, and 3c 30
4.2.4 Hypothesis 3d, 3e, and 3f 31
4.3 Results study 2 34
4.3.1 Hypothesis 4a, 4b, and 4c 34
5. Discussion and implications 37
5.1 Practical implications 41
5.2 Strengths and limitations 42
5.3 Future research 44
6. Conclusion 45
7. Acknowledgements 46
8. Reference list 47
List of Tables
4. Results 24
Table 1. Means, standard deviations and correlations of variables 25 Table 2. Regression analysis of the effects of Admiration on LMX 28 Table 3. Regression analysis of the effects of Rivalry on LMX 28 Table 4. Regression analysis of the effects of LMX on in-role behavior, affiliative
OCB, and challenging OCB 30
Table 5. Bootstrapping results of LMX as a mediator in the relationship between
Rivalry and in-role behavior 32
Table 6. Bootstrapping results of LMX as a mediator in the relationship between
Rivalry and affiliative OCB 33
Table 7. Bootstrapping results of LMX as a mediator in the relationship between
Rivalry and challenging OCB 33
Table 8. Regression analysis of the curvilinear effect of leader narcissism on
in-role behavior 35
Table 9. Regression analysis of the curvilinear effect of leader narcissism on
affiliative OCB 36
Table 10. Regression analysis of the curvilinear effect of leader narcissism on
challenging OCB 36
1. Introduction
Nearly every working adult has reported that he or she worked for an intolerable boss during one point in his or her life (Blair, Hoffman & Helland, 2008). Given this prevalence of “bad” leadership behaviors, the “dark” side of leadership has become a more popular subject in leadership research for the past few decades. In the literature there is now a well-established link between narcissism and leadership. Narcissism is the “one personality constellation to which leaders tend to gravitate” (De Vries & Miller, 1985, p. 586) and narcissists are frequently found to emerge as leaders (Brunell, Gentry, Campbell, Hoffman, Kuhnert & DeMarree, 2008; Judge, LePine & Rich, 2006; Nevicka, de Hoogh, Van Vianen, Beersma & McIlwain, 2011). Since narcissists tend to frequently emerge as leaders in organizations, it is surprising to find the lack of studies investigating its effect on the job performance of
employees, since the latter is shown to be very important for the survival and profitability of firms (Koys, 2001; Fisher, McPhail & Menghetti, 2010). Follower job performance
encompasses in-role behaviors, or task behaviors that directly contribute to the organizational effectiveness and extra-role behaviors, or organizational citizenship behaviors that are not formally part of the job description, but do contribute to the organizational effectiveness, such as bringing in new ideas and voluntarily doing extra work (Borman & Motowidlo, 1997).
Nonetheless, several studies investigated the effect of leader narcissism on other organizational outcomes. Narcissistic leaders have a negative effect on the effectiveness of the organization through their exploitative behaviors (Campbell, Hoffman, Campbell & Marchisio, 2011), fail to create a climate necessary to achieve sustainable performance (Higgs, 2009), adversely influence subordinates physical and psychological well-being (Padilla, Hogan & Kaiser, 2007), downplay subordinates’ needs (Zeigler-Hill, Meyers & Clark, 2010), and inhibit information exchange between group members and thereby
negatively affect group performance (Nevicka et al., 2011). On the other hand, narcissistic leaders possess positive traits or behaviors that are beneficial for organizational outcomes. For example, the organizational performance of firms with narcissistic CEO’s is no better or worse than firms with non-narcissistic CEO’s and narcissistic CEO’s tend to engage in bolder and visible actions, acting like a trigger for organizational change (Chatterjee & Hambrick, 2007). Also, narcissistic leaders are often charismatic, which is vital to effective leadership (Rosenthal & Pittinsky, 2006), and are often considered as risk-takers and inspirational leaders by their followers (Galvin, Waldman & Balthazard, 2010).
In order to solve these inconsistent findings in research on leader narcissism and organizational outcomes, this study aims to address the shortcomings of previous studies. First, there is a possibility of multiple “flavors” or forms of narcissism (Campbell et al., 2011, p. 270). Several authors already tried to capture this multidimensionality of narcissism
(Ackerman, Witt, Donnellan, Trzesniewski, Robins & Kashy, 2010; Corry, Merrit, Mrug & Pamp, 2008; Emmons, 1984; Kubarych, Deary & Austin, 2004; Raskin & Terry, 1988), differentiating in “adaptive” and “maladaptive” narcissism. However, no consensus has been reached and the distinctiveness and internal consistencies of the measures are questionable. The Narcissistic Admiration and Rivalry Concept (NARC) developed by Back, Küfner, Dufner, Gerlach, Rautmann and Denissen (2013) seems promising, since it has a good factorial structure, internal consistencies, and stabilities. It also demonstrates convergent, discriminant and predictive validity above and beyond previous measures of narcissism (Back et al., p. 1032). Admiration encompasses self-confident and expressive behaviors, or adaptive narcissistic aspects, while Rivalry encompasses a devaluation of others and other hostile behaviors, or maladaptive aspects.
Second, leader narcissism has clear relational implications (Owens et al., 2015), but no study to date used this insight by adding relationship-based mediators in the link between
leader narcissism and organizational outcomes. A useful concept to include is that of leader-member-exchange (LMX). This concept focuses on the explicit one-on-one relationships that develop between leaders and followers. Including such relationship-based constructs might help explain what is happening inside the "black box" between observed leadership aspects and measured follower outcomes (Howell & Hall-Merenda, 1999).
Finally, all previous studies focused mainly on linear relationships between leader narcissism and organizational outcomes. While there can be a lack of evidence of a linear relationship, there might be a possibility of a curvilinear relationship. Most often,
hypothesized relations take on its simplest form: linear. The prevailing theoretical perspective has largely been “the more the better”. So far, there are very limited published studies
examining a curvilinear relationship between leader narcissism and organizational outcomes. Pierce and Aguinis (2013) state that doing a curvilinear analysis, or what they call the “too much of a good thing-effect”, “provides an enhancement and makes a value-added
contribution to theory and practice” (p. 316). Doing this type of analysis can solve the
inconsistencies that still exist concerning the research on leader narcissism and organizational outcomes.
Therefore, the aim of this study is two-fold: firstly, to test whether the distinctive dimensions of leader narcissism, namely Admiration and Rivalry, show a linear relationship with follower job performance (in-role and extra-role behaviors) that is mediated by LMX, and secondly, to test whether there is a curvilinear relationship between leader narcissism and follower job performance. Using different dimensions of the NARC in this study can add to the literature by identifying which aspects of leader narcissism are either beneficial or harmful for follower job performance and if this is effect is (partially) explained by the relationship between the leader and follower. Furthermore, doing a curvilinear analysis can
result in new theoretical insights and can solve current theoretical inconsistencies concerning leader narcissism and workplace outcomes.
2. Theoretical background
2.1 Multidimensionality of leader narcissism
In the literature, there has been debate about narcissism consisting of different forms relating to either positive or negative outcomes. De Vries and Miller (1985) called narcissistic leaders either “constructive” or “reactive” for the organization, respectively being good or bad for the organization. Maccoby (2000) introduces the concept of “productive narcissists”, leaders who are visionaries, can inspire people, and can convert the masses with their rhetoric. They have great visions because of their need for power, glory, and legacy. They gain followers who in turn fulfil the narcissistic leader’s need for admiration. But these productive narcissists also possess some weaknesses, for example they are sensitive to criticism, are poor listeners, and lack empathy. So, often the literature speaks of adaptive and maladaptive forms of
narcissism, respectively relating to either positive personality traits, like extraversion, or negative personality traits, like manipulation (Wink, 1991; Watson, Taylor & Morris, 1987).
This debate is in line with readings on the multi-dimensionality of narcissism as a construct. Most research on narcissism used the Narcissistic Personality Inventory (NPI; Raskin & Hall, 1979). Several, authors argued that the factor structure of the NPI is multi-dimensional and that there might be different aspects of narcissism embedded in it (Emmons, 1984; Raskin & Terry, 1988). Emmons (1984) examined the internal structure and the
multidimensionality of the NPI using principal components analysis (PCA) and exploratory factor analysis (EFA). He found that the NPI consisted of four dimensions:
Self-absorption/Self-admiration. The NPI is not a unitary construct, but rather consist of four dimensions that moderately correlate with each other. Raskin and Terry (1988) tried to modify the work of Emmons (1984). Conducting a principal-components analysis resulted in seven components of the NPI: Authority, Exhibitionism, Superiority, Vanity,
Exploitativeness, Entitlement, and Self-Sufficiency. Corry et al. (2008) also studied the multidimensional nature of the NPI. First, exploratory factor analysis showed a two-factor model of the NPI: Leadership/Authority and Exploitativeness/Entitlement. The first is best associated with adaptive narcissism and the latter with maladaptive narcissism. Second, using confirmatory factor analysis, they tested the four-factor model of Emmons (1984), the seven-factor model of Raskin and Terry (1988), the two-and three-seven-factor model of Kubarych et al. (2004), and their own two-factor model. They found that the two-factor model obtained in their own study and the one developed by Kubarych et al. (2004) showed the best fit and internal consistency and was therefore recommended for further usage.
More recently, Ackerman et al. (2010) expressed some doubts about Corry et al.’s (2008) approach. They argue that placing too much emphasis on identifying dimensions that yield scales with high levels of internal consistency, “may lead to an underextraction of key factors”. It might be, they argue, that a dimension of the NPI with a relatively small number of items can be theoretically important. Analyzing the factor structure of the NPI resulted in a three-factor solution: Leadership/Authority, Grandiose Exhibitionism, and
Entitlement/Exploitativeness. Leadership/Authority was generally linked to adaptive
outcomes, where the two latter dimensions were linked to maladaptive outcomes. However, the adaptive facets capture twenty items in total, while the maladaptive facets capture only four items.
There are, however, some concerns relating to the distinctiveness of these dimensions, how strongly they correlate to each other, and their reliabilities (Back et al., 2013).
Furthermore, the NPI predominantly assesses adaptive narcissistic behaviors, leaving the maladaptive behaviors underemphasized (Pincus & Lukowitsky, 2010). A more recent study by Back et al. (2013) developed a new measure of narcissism: NARC. They argue that the reason all previous attempts to generate dimensions of the NPI resulted in different kinds of factors, is because the NPI was originally constructed as a unidimensional measure. Also, the NPI mostly captures the adaptive aspects of narcissism, letting the maladaptive aspects underrepresented. This means that the majority of the items in the NPI tap only the assertive, dominant, and grandiose aspects of narcissism (like Leadership/Authority and Grandiose Exhibitionism), and a limited number of items the antagonistic, aggressive, and exploitative aspects (like Entitlement/Exploitativeness). Furthermore, the Leadership/Authority and Exploitativeness/Entitlement facets are quite restricted and miss many other crucial elements of narcissism. The NARC is designed to overcome these inconsistencies and is able to capture two distinct dimensions, namely Admiration and Rivalry, into one measurement. Both dimensions are positively correlated, yet they have different nomological network and distinct interpersonal consequences (Back et al., 2013, p. 1013).
The NARC is based on the idea that the goal of a narcissist is to maintain a grandiose self that is achieved by two underlying motivational dynamics: assertive self-enhancement (self-promotion) and antagonistic self-protection (self-defense). Assertive self-enhancement activates a set of behaviors that Back et al. (2013) term Admiration. This consists of striving for uniqueness, grandiose fantasies, and charmingness. These exaggerated thoughts about one’s own grandiosity and uniqueness result in self-assured, dominant, and expressive behaviors (charmingness). This in turn results in positive social outcomes, such as status, success, praise, being chosen as a leader, and evoking social interest. Antagonistic self-protection activates a set of behaviors is called Rivalry. It consists of striving for supremacy, devaluation of others, and aggressiveness. This self-protection leads to defending one’s
status, including devaluing others. This in turn leads to aggressiveness and results in negative social outcomes such as rejection, unpopularity, criticism, and a lack of trust from others (Beck et al., 2013, p. 1016). Both positive and negative social interaction outcomes feed back into the underlying motivational dynamics. For example, praise or success are accompanied by a certain ego-boost, which in turn reinforces the self-enhancement strategy. On the other hand, negative social interaction outcomes, such as criticism, is perceived as an ego-threat, which in turn feeds back into self-protection strategy.
Besides Back et al.’s (2013) different underlying approach to measuring the
narcissism construct, the NARC also shows some similarities with previous studies analyzing the factorial structure of the NPI mentioned above. Compared to the study of Ackerman et al. (2010) on the factorial structure of the NPI, Admiration is closely correlated to
Leadership/Authority (r = .47) and Rivalry to Exploitativeness/Entitlement (r = .47). Furthermore, compared to the facets developed by Emmons (1984), Admiration is closely correlated to the Self-Sufficiency (r = .51), Authority (r = .44), and Superiority (r = .45) dimensions, while Rivalry is closely correlated to Exploitativeness-Entitlement (r = .55) dimension. Overall this means that, taking the strong correlations of the NARC’s dimensions with earlier developed dimensions of narcissism and the positive and negative social
outcomes of the NARC into consideration, Admiration can be seen as adaptive narcissism, while Rivalry can be seen as maladaptive narcissism
2.2 Admiration, Rivalry, and its interpersonal consequences
Previous research has shown that narcissistic leaders are not very skilled in building positive interpersonal relationships in the workplace. For example, a narcissistic leader inflicts damage on others by bullying and coercion and damages the well-being of the employees
(Aasland, Skogstad & Einarsen, 2008). They also damage the trust of the employee (Benson & Hogan, 2008), create a toxic work atmosphere by showing a lack of empathy toward employees (Goldman, 2006), and have poor management rankings on interpersonal performance and integrity (Blair et al., 2008).
A theory that is based on the interpersonal relationship between leaders and follower is LMX theory. It focuses on the dyadic relationship leaders have with their followers. It is based on the principle that leaders have different types of exchange relationships with their followers and that the quality of these relationships impact leader and follower attitudes and behaviors (Ilies, Nahrgang & Morgeson, 2007). The relationship is based on social exchange theory, which suggests that the follower and the leader each must offer something that the other party sees as valuable and that this “exchange” is seen as fair and equitable (Wayne, Shore & Liden, 1997). A high-quality exchange relationship is based on high levels of mutual trust, interaction, support, respect, liking and reciprocal influence (Ilies et al., 2007; Liden & Maslyn, 1998). Low quality exchange relationships are characterized by formal, role-defined interactions and exchanges that result in hierarchical downward influence between the leader and follower (Janssen & van Yperen, 2004).
Back et al. (2013) analyzed the interpersonal consequences of both Rivalry and Admiration. The behaviors expressed by people high on Admiration are dominant, self-confident, and expressive. All behaviors are motivated by a feeling of self-enhancement, in which the narcissist aims to feel good about one-self and reinstates one’s grandiose self by feeling admired a special. Admiration positively correlates to certain aspects that are necessary for maintaining close relationships with others, such as gratitude and direct
problem-focused reactions in conflicts. Furthermore, people high in Admiration are perceived by others as sociable, assertive, attractive, and competent. Finally, they also view other people as attractive. Overall, people high in Admiration can have positive consequences for
social interaction due to their assertive self-enhancement behaviors. This will positively influence aspects such as social interaction, respect, and liking, all of which are antecedents for maintaining high-quality exchange relationships. This leads to the following hypothesis:
Hypothesis 1a: There is a positive relationship between Admiration and LMX
The behavioral aspects that characterize the Rivalry dimension of the NARC are the devaluation of others, striving for supremacy, and aggressive and hostile reactions towards others. All behaviors are motivated by a feeling of self-protection in order to feel good about one-self and “protect” one’s grandiose self. Back et al. (2013) found that Rivalry negatively correlates to aspects necessary for maintaining close interpersonal relationships, such as empathy, interpersonal trust, forgiveness, gratitude, and direct problem-focused reactions in conflicts. Thus, the previously found negative effects of leader narcissism on maintaining positive interpersonal relationships could be attributed to the maladaptive aspect of
narcissism. Furthermore, people high in Rivalry are seen as less trustworthy and likable by others. Additionally, they perceive others as more negative (aggressive, less trustworthy, and narcissistic). Overall, they can have negative consequences for social interaction due to their antagonistic self-protective behaviors. This will negatively influence aspects such as social interaction, respect, and liking, all of which are antecedents for maintaining high-quality exchange relationships. This leads to the following hypothesis:
2.3 Admiration, Rivalry, and its interpersonal consequences for follower job performance In this study, follower job performance is comprised of two aspects: in-role behavior and extra-role behavior. In-role behavior (IRB), or so-called task performance, is defined as “behaviors recognized by formal reward systems and are part of the requirements described in job descriptions” (Organ, Podsakoff, & MacKenzie, 2005). Extra-role behavior, or so-called organizational citizenship behavior (OCB), is defined as “individual behavior that is discretionary, not directly or explicitly recognized by the formal reward system, and which, in aggregate, promotes the efficient and effective functioning of the organization” (Organ et al., 2005). In the literature, there are different forms of OCB’s identified. Van Dyne and LePine (1998) distinguished these behaviors between those that are affiliative or those that are challenging. Affiliative behaviors include promoting and supporting the existing work processes and relationships, thus maintaining the status quo. Affiliative OCB (OCBA) is interpersonal and cooperative (McAllister, Kamdar, Morrison & Turban, 2007), such as helping other colleagues and showing courtesy towards other people. Challenging behaviors include questioning and improving current work processes and relationships, thus challenging the status quo (Van Dyne & LePine, 1988). Challenging OCB (OCBC) is change-oriented behavior, such as inflicting change on how things are currently done at work or expressing new ideas (McAllister et al., 2007).
The reason why people at work engage in OCB is because when they feel they are treated well by their superior, they feel a need to reciprocate a positive contribution to the organization which is beyond their formal task description (Organ, 1988). So, an important antecedent in how well employees believe they are treated positively is that of the quality of the relationship with their supervisor, or their perceived LMX (Van Dyne, Kamdar &
Joireman, 2008). Because of a social exchange between leader and follower, it is argued that people who are treated positively by others feel obligated to respond positively or return a
favor instead (Walumbwa, Mayer, Wang, Wang, Workman & Christensen, 2011). So, when there is a high-quality exchange relationship, employees exert extra effort on behalf of the leader (Wayne & Green, 1993).
Several studies have found positive relationships between LMX and follower job performance, including OCB’s (Bauer, Erdogan, Liden & Wayne, 2006; Walumbwa,
Cropanzano & Hartnell, 2009; Walumbwa et al., 2011; Wang, Law, Hackett, Wang & Chen, 2005). Furthermore, a meta-analysis conducted by Ilies et al. (2007) revealed that high-quality exchange relationships between leaders and followers predict both IRB and OCB’s, suggesting that both aspects of follower job performance should be considered in doing leadership research. LMX theory deals directly with leadership as a driver of forms of motivation. Subsequently, motivation is presumed to lead to more tangible outcomes, such as follower job performance and OCB’s (Hiller, DeChurch, Murase & Doty, 2011, p. 31). Therefore, it is expected that when employees have high-quality exchange relationships with their leader, such as a high level of liking, loyalty, and respect, this will create a safe
psychological climate in which employees are willing to fulfil their prescribed tasks, are willing to help other people in the organization, and feel safe to express their ideas. This leads to the following hypotheses:
Hypothesis 2a: There is a positive relationship between LMX and IRB Hypotheses 2b: There is a positive relationship between LMX and OCBA Hypothesis 2c: There is a positive relationship between LMX and OCBC
Overall, since leaders high in Admiration can have positive consequences for social interaction due to their assertive self-enhancement behaviors, it is expected that this will lead to high-quality exchange relationships with their employees and therefore to increased follower job performance. On the contrary, leaders high in Rivalry can have negative
consequences for social interaction due to their antagonistic self-protective behaviors, which will lead to low-quality exchange relationships with their employees and therefore to
decreased follower job performance. This leads to the following hypotheses:
Hypothesis 3a-c: LMX mediates the positive relationship between Admiration and follower job performance (IRB, OCBA, and OCBC)
Hypothesis 3d-f: LMX mediates the negative relationship between Rivalry and follower job performance (IRB, OCBA, and OCBC)
2.4 Curvilinear research on leader narcissism
Analyzing the relationship between leader narcissism and follower job performance by a disentanglement of narcissism’s dimensions and incorporating its interpersonal implications seems promising. However, another interesting way of analyzing the link between leader narcissism and follower job performance is by doing a curvilinear research. In the literature, there are several studies examining the curvilinear relationship between leader narcissism and organizational outcomes, such as leadership performance, effectiveness, and group creativity. First, Benson and Campbell (2007) found evidence for nonlinear relationships between narcissistic behaviors (ego-centered, intimidating, manipulation, micro-managing, and passive-aggressive) and leadership performance. They argue that it is important for leadership performance to have mid-range levels of narcissistic behaviors, such as
entitlement, visionary-thinking, self-confidence, and risk-taking behavior. A leader who does not feel a sense of entitlement or self-confidence, is unlikely to aspire for a leadership
position. Especially, risk-taking behavior and creative-thinking can distinguish mediocre leaders from more effective ones. But, when the levels of these behaviors become too high, it shows an inability to learn from mistakes, to share credits, and to avoid unnecessary risks. They become so over-confident in their thinking that others see them as out of touch with the situation.
Second, Goncalo, Flynn and Kim (2010) found that a narcissistic personality has a curvilinear relationship with other criteria at the team-level. First, they found that narcissists evaluate themselves as creative talents, while this was not supported by objective measures of their performance. Second, they found that having more narcissists on a team facilitated the creativity of both the group process (systematic thinking) and the group’s product up to a point at which increasing narcissism in the group became detrimental. This implies that having more narcissists on a team can contribute to group creativity even if they are not that creative on their own. However, having too many narcissists on a team eventually provides diminishing returns.
Finally, Grijalva, Harms, Newman, Gaddis, and Frayley (2015) conducted a meta-analysis on narcissism and leadership research. They found that “the nil linear relationship between narcissism and leadership effectiveness masks an underlying curvilinear trend”. This means that there could be an optimal or midrange level of leader narcissism in relation to leadership effectiveness. The theoretical reasoning behind this finding is based on a
behavioral threshold theory that explains curvilinear effects. Grijalva et al. (2015) argue that a curvilinear effect, which takes an inverted U-shape, between an independent variable X and a dependent variable Y is explained by the threshold of the items. Narcissism has behavioral manifestations and different thresholds for individuals to engage in it. If an item has a high
threshold, this means individuals engage in the behavior when they possess a high level of the trait. Other behaviors are only engaged in when they possess a low level of the trait (low threshold item). This means that if X has a low threshold, or is ‘easy’, it will be more
positively related to Y, whereas if X has a high threshold, or is ‘difficult’, it will be more negatively related to Y. So, low threshold items tend to be more adaptive, while high threshold items tend to be more maladaptive.
Ackerman, Donnellan and Robins (2012) examined the varying threshold of narcissism items of the NPI using Item Response Theory. They found that items that are more difficult, or have higher thresholds, are often related to only higher levels of narcissism. Easy items, or items with lower threshold, are often related to both lower and higher levels of narcissism. Higher levels of narcissism are characterized by maladaptive aspects, while low to moderate levels of narcissism are characterized by more adaptive aspects. To put simply, easy questions on a test are often answered correct by all people, no matter how high or low their intelligence level, while difficult questions are only answered correctly by people who possess higher levels of intelligence. This implies that when respondents possess low to average levels of narcissism, only adaptive aspects of narcissism are found. However, when the level of narcissism in respondents rises, this will lead to more maladaptive narcissistic aspects.
Still, no study to date conducted a curvilinear analysis of leader narcissism on follower job performance, while this has shown to be of great importance for the
effectiveness of the organization (Koys, 2001; Fisher et al., 2010). Back et al. (2013) argued that people high in Admiration show more adaptive aspects of narcissism due to positive social interaction outcomes, while people high in Rivalry show maladaptive aspects of narcissism, due to negative social interaction outcomes. The NARC captures both adaptive (Admiration) and maladaptive (Rivalry) aspects of leader narcissism into one construct. So,
when measuring leader narcissism as a whole construct, low to average levels of leader narcissism reflects Admiration, leading to adaptive aspects, but when levels of leader narcissism rise, Rivalry becomes apparent, thus leading to more maladaptive aspects. Eventually, the maladaptive aspects overshadow the adaptive aspects when level of leader narcissism rise, turning the positive relationship to Y into a negative relationship, creating an inverted U-shape pattern. So, the above reasoning for a curvilinear trend by Grijalva et al. (2015) also apply to the hypothesis of this current study: increasing levels of leader narcissism capturing low threshold items (Admiration) result in more adaptive aspects of leader narcissism, thus leading to better follower job performance. However, as levels of leader narcissism rise, and high threshold items are endorsed by respondents (Rivalry), more maladaptive aspects of leader narcissism become apparent and the relationship between leader narcissism and follower job performance becomes negative. This leads to the following hypothesis:
Hypothesis 4a-c: Leader narcissism will have an inverted U-shaped relationship to follower job performance (IRB, OCBA, OCBC). The relationship is initially positive, but becomes negative as levels of leader narcissism increase. Therefore, follower job performance (IRB, OCBA, OCBC) will benefit most from moderate levels of leader narcissism.
3. Method
3.1 Participants and procedure
This study is carried out by using a cross-sectional survey design. Two separate
self-administered surveys were developed, including an invitation letter, and were distributed to managers and employees of Dutch and non-Dutch companies over a period of six weeks. The surveys were available in Dutch and English, depending on the nationality of the respondents.
Each survey consisted of a short introduction in which the study was shortly outlined, how long the survey would approximately take, and ethical procedures were presented. In order to say something about the interplay between a leader’s narcissistic aspects and follower
outcomes, dyads are included in this study. Employees completed a survey regarding aspects relating to their managers, and vice versa. In order to make the followers and leaders a dyad, a unique survey code was distributed, which each dyad had to fill in on the first question of the survey. The survey was provided to the leader, who subsequently chose the subordinate to fill in the other survey. In some cases, the survey was provided to the follower, who then distributed it to their leader. As a result, the sample consists of Dutch and non-Dutch leader/follower-dyads working in different organizations.
Nonprobability sampling methods were used, such as convenience sampling and purposive sampling. Convenience sampling means the members of the population are
selected based on their ease of access. Since it is difficult to collect approximately 130 dyads, this type of sampling method suits well. However, we also used purposive sampling, meaning that we chose respondents who we thought were appropriate/suitable for the study. This implies that we were mainly interested in office workers. Respondents who responded
positively to participating in this study received an invitation consisting of a brief explanation of the study, a unique survey code, and a weblink to the online survey. Both online and printed versions of the survey were used in this study. The survey started with a short informative introduction including the voluntary character of the survey. Participation was voluntary and the answers of the respondents were held confidential. Because the data is collected with a group of five students, and different sampling techniques are used, the sample could be more diverse than using a single sampling method with only one researcher.
From the 450 respondents who received an invitation to the study, a total of 258 respondents filled in the survey, resulting in a response rate of 57,3%. However, some
employees and managers did not complete the survey or one of them did not fill it in
completely while the other one did. This resulted in unusable dyads and after removing these, the data consisted of 123 dyads, leading to a final response rate of 54,6%. Of these 123 dyads, 22,8% were non-Dutch followers and leaders. Of the 123 leaders, 50,4% were male and 46,3% were female. The majority was 25-34 years (35%) (18-24: 13%, 35-44: 21,2%, 45-54: 22%, 55-64: 7,3%). The mean time working with the organization and working with the employee was rated at respectively 8,47 years (SD = 6,26) and 4,4 years (SD = 4,79). The majority of leaders had contact with their employee on a daily basis (60,2%), 37,4% on a weekly basis, and 0,8% on a monthly basis. Of the 123 employees, 39% were male and 57.7% were female. The majority was 25-34 years old (42,3%) (18-24: 39%, 35-44: 10,6%, 45-54: 4,1%, 55-64: 4,1%). The mean time working with the organization was 5,04 years (SD = 6,85)
3.2 Measures
The variables used in this study are measured using both the survey for the leader and the follower. All variables in the two surveys are operationalized by using scales selected from extant literature, and were, where necessary, translated to Dutch. An overview of all the items are presented in Appendix 1. The items were rated on a seven-point Likert scale ranging from strongly disagree (1) to strongly agree (7). At the end of each survey, some questions about demographics were incorporated like age, gender, and tenure.
Leader narcissism (NARQ) (α = .89) is assessed using the NARQ-scale adopted from Back et al. (2013). This scale consists of 18 items total and possesses two underlying and distinct dimensions, namely Admiration (ADM) (α = .88) and Rivalry (RIV) (α = .87). Sample items of Admiration include, “Most of the time I am able to draw people’s attention to myself
in conversations” and “I am great”. Sample items of Rivalry include “I react annoyed if another person steals the show from me” and “I want my rivals to fail”.
To measure In-role Behavior (IRB) (α = .85), Williams and Anderson’s (1991) 7-item measurement of in-role behavior is used. Sample items include, “Adequately completes assigned duties” and “Meets formal performance requirements of the job”.
Affiliative OCB (OCBA) (α = .77) is measured using MacKenzie, Podsakoff and Fetter’s (1993) 7-item scale. Sample items include, “Respects other people’s rights to common/shared resources (including clerical help, materials, etc.)” and “Is always ready to help or to lend a helping hand to those around him/her”.
Challenging OCB (OCBC) (α = .85) is measured using Van Dyne and Le Pine’s (1998) scale consisting of 6 items. Sample items include, “This particular co-worker develops and makes recommendations concerning issues that affect this workgroup” and “This
particular co-worker communicates his/her opinions about work issues to others in this group even if his/her opinion is different and other in the group disagree with him/her”.
To assess Leader-Member-Exchange (LMX) (α = .91), Liden and Maslyn’s (1998) 11-item scale is used. Sample 11-items include, “I like my supervisor very much as a person” and “I am willing to apply extra efforts, beyond those normally required, to further interests of my workgroup”.
3.3 Control variables
In order to avoid any spurious relationships between the independent variable, mediator, and outcome variables, this study included relationship tenure (Rel) as a control variable.
Research has shown that perceived LMX by employees is higher if the relationship between the employee and manager developed over time (Maslyn & Uhl-Bien, 2001).
4. Results
4.1 Preliminary tests and descriptive statistics
First, counterindicative items were recoded where necessary. After running reliability analyses, NARQ (α = .89) consisted of eighteen items, Rivalry (α = .87) consisted of nine items, Admiration consisted of nine items (α = .88), LMX (α = .91) consisted of eleven items, IRB (α = .85) consisted of seven items, OCBA consisted of seven items (α = .77), and OCBC (α = .85) consisted of six items. (see Table 1). All scales showed high reliabilities (α > .70) and the ‘Cronbach’s alpha if item is deleted’ were below 0.1 for all scales. Therefore, no items were removed after doing reliability analysis. After this step, scale means were computed that were suitable for further analyses.
Running tests on normality showed that all variables were not normally distributed. Analyzing the Q-Q-plots of each of the seven variables revealed that the dots did not fall right on the line (suggesting kurtosis) and, in fact created an S-like pattern (which suggests skew). The Kolmogorov-Smirnov-test was significant for all variables, indicating
non-normality. However, regression analysis is quite robust to violations of normality (Bohrnstedt & Carter, 1971) and tests can proceed in large samples even when the assumption for
normality is not justified.
The mean scores, standard deviations, and Pearson Correlations of the variables in this study are displayed in the correlation matrix (Table 1). Results show that NARQ positively correlates to Admiration (r = .84, p = 0.00) and Rivalry (.85, p = 0.00). NARQ negatively correlates to LMX (r = -.22, p = 0.02) and IRB (-.31, p = 0.00), suggesting that
higher levels of leader narcissism are associated with lower levels of LMX and IRB. However, NARQ non-significantly correlates to OCBA (p = 0.18) and OCBC (p = 0.27). Also, Admiration and Rivalry positively correlate to each other (r = .42, p = 0.00).
Furthermore, all three follower job performance variables, IRB (r = .43, p = 0.00), OCBA (r = .35, p = 0.00), and OCBC (r = .36, p = 0.00), positively correlate to LMX, suggesting that higher levels of LMX are associated with higher levels of follower job performance. All three follower job performance variables (IRB, OCBA, OCBC) positively correlate to each other. Finally, Rivalry negatively correlates to LMX (r = -.29, p = 0.00), IRB (r = -.43, p = 0.00), and OCBA (r = -.19. p = 0.03), while Admiration non-significantly correlates to all three follower job performance variables. This could imply that Rivalry has far more negative consequences for the relationship between leaders and followers and the job performance of followers, while Admiration does not make any (positive) impact at all.
Table 1. Means, standard deviations and Pearson correlations of variables
M SD 1 2 3 4 5 6 7 8 1. Rel 4.40 4.79 - 2. NARQ 3.46 .83 -.04 (.89) 3. Adm 4.44 .97 -.05 .84** (.88) 4. Riv 2.49 1.00 -.01 .85** .42** (.87) 5. LMX 5.55 .87 .09 -.22* -.08 -.29** (.91) 6. IRB 5.81 .71 -.08 -.31** -.09 -.43** .43** (.85) 7. OCBA 5.60 .71 .00 -.12 -.01 -.19* .35** .54** (.77) 8. OCBC 5.27 .87 .15 -.10 .00 -.16 .36** .47** .64** (.85) ** Correlation is significant at p ≤ 0.01 (2-tailed).
4.2 Results study 1
In this section the results of the first study are presented. Baron and Kenny (1986) argued that there are several conditions that must be met to support a mediation effect. The first step in Baron and Kenny’s procedure for mediation is that the independent variable is correlated to the dependent variable. Second, the independent variable must also correlate to the mediator variable. Third, there is a correlation between the mediator variable and the outcome variable. If all conditions are met, one can proceed to test if mediation exist using multiple regression. If the effect of the mediator on the dependent variable remains significant when controlled for the independent variable, there is mediation. Full mediation occurs when the effect of the independent variable on the dependent variable becomes non-significant when the mediator is controlled. If the effect of the independent variable on the dependent variable remains
significant, the results support partial mediation.
The above reasoning claims that mediation can only be established when there is a significant effect to be mediated (significant effect between the independent and dependent variable). However, there is growing consensus that such a prerequisite for mediation is not needed (Preacher & Hayes, 2004; MacKinnon, Lockwood, Hoffman, West & Sheets, 2002; Zhao, Lynch & Chen, 2010). Furthermore, the absence of an effect between the independent and dependent variable after controlling for a mediator should not lead to conclusions of ‘full’ mediation (MacKinnon et al., 2002; Rucker, Preacher, Tormala, & Petty, 2011; Shrout & Bolger, 2002; Zhao et al., 2010). Therefore, in this study, we focus on the strength of the indirect effect rather than the magnitude of the direct effect compared to the indirect effect. Emphasizing the effect size of the indirect effect can provide valuable insights as to what extent the mediator is the most important mediator in the relationship, questioning the possibility of other (important) mediators in the relationship (Rucker et al., 2011). This does not mean that the total and direct effect are not important, since they can be of theoretical
value, but they will not serve as a prerequisite for having a mediation effect in this study. Also, “if there are theoretical reasons to predict the presence of an indirect effect […], researchers should explore these effects regardless of the significance of the total or direct effect” (Rucker et al., 2011, p. 368). This is the reason why the hypotheses stated in study 1 are theory-driven, rather than serving as conditions for mediation analysis.
4.2.1 Hypothesis 1a + 1b
Hypothesis 1a and 1b are tested by using multiple hierarchical regression. First we
investigate the ability of Admiration to positively predict levels of LMX, after controlling for relationship tenure (H1a) (see Table 2). In the first step of hierarchical multiple regression, the control variable is entered. This model is statistically non-significant (F (1, 87) = .698; p = 0.41). After entry of Admiration in step 2, the model is statistically non-significant (F (2, 86) = .586; p = 0.56), after controlling for relationship tenure. As a result, Hypothesis 1a is not supported. This means that there is no support for a positive relationship between Admiration and LMX.
Second, we investigate the ability of Rivalry to negatively predict levels of LMX, after controlling for relationship tenure (H1b) (see Table 3). In the first step of hierarchical multiple regression, the control variable is entered (see above). After entry of Rivalry in step 2, the total variance explained by the model as a whole is 9% (F (2, 86) = 4,244; p = 0.02), after controlling for relationship tenure. Rivalry negatively relates to LMX (b = -.29; p = 0.01). Thus, Hypothesis 1b is supported, meaning that there is a negative relationship between Rivalry and LMX.
Table 2. Regression analysis of the effects of Admiration on LMX
Variable Model 1 Model 2 (H1a)
Step 1 Rel .09 .09 Step 2 ADM -.07 DR2 .01 .01 R2 .01 .01 Degrees of freedom 1, 87 2, 86 Overall F .70 .59 Note. ** p ≤ 0.01, * p ≤ 0.01
Table 3. Regression analysis of the effects of Rivalry on LMX
Variable Model 1 Model 2 (H1b)
Step 1 Rel .09 .09 Step 2 RIV -.29** DR2 .01 .08 R2 .01 .09 Degrees of freedom 1, 87 2, 86 Overall F .70 4.244 Note. ** p ≤ 0.01, * p ≤ 0.05
4.2.2 Hypothesis 2a, 2b, and 2c
In order to test whether LMX positively predicts all three follower job performance variables, again, multiple hierarchical regression is used. The results of doing these analyses are
presented in Table 4. In the first step of each of the three hierarchical regressions, the control variable is entered. For IRB (F (1, 86) = .548; p = 0.46), OCBA (F (1, 87) = .00; p = 1.00), and for OCBC (F (1, 87) = 2.035; p = 0.16), the first model is non-significant.
Then, in the second step of each of the three hierarchical regressions, LMX is entered in the model, and LMX positively relates to all three follower job performance variables, after controlling for relationship tenure. For IRB, the total variance explained by the model as a whole is 20% (F (2, 85) = 10.316; p = 0.00). LMX positively relates to IRB (b = .44; p = 0.00). Thus, Hypothesis 2a is supported. Then, for OCBA, the total variance explained by the model as a whole is 12% (F (2, 86) = 5.897; p = 0.00). LMX positively relates to OCBA (b = .35; p = 0.00). Thus, Hypothesis 2b is also supported. Finally, for OCBC, the total variance explained by the model as a whole is 14% (F (2, 86) = 7.071; p = 0.00). LMX, again, positively relates to OCBC (b = .35; p = 0.00). Thus, Hypothesis 2c is supported. Overall, this means that there is a positive relationship between LMX and IRB (H2a), OCBA (H2b), and OCBC (H2c).
4.2.3 Hypothesis 3a, 3b, and 3c
In order to test whether LMX mediates the positive relationship between Admiration and all three follower job performance variables, we used the Process Macro developed by Hayes (2013) using model 4.
Regression analysis is used to investigate whether LMX mediates the positive
relationship between Admiration and all three follower job performance variables. However, Hypothesis 1a already indicated that Admiration is non-significantly related to LMX,
suggesting that the effect of Admiration on IRB, OBCA, and OCBC could not be mediated by LMX. Using Process, this effect resulted in a bootstrap CI of [-.0224, .0541], suggesting a non-significant effect. Therefore, further analyses concerning such a mediation are irrelevant.
Table 4. Regression analysis of the effects of LMX on in-role behavior, affiliative OCB, and challenging OCB
Variables
IRB OCBA OCBC
Model 1 Model 2 (H2a) Model 1 Model 2 (H2b) Model 1 Model 2 (H2c) Step 1 Relationship tenure -.08 -.12 .00 -.03 .15 .12 Step 2 LMX .44** .35** .35** DR2 .06 .19 .00 .12 .02 .12 R2 .01 .20 .00 .12 .02 .14 Degrees of freedom 1, 86 2, 85 1, 87 2, 86 1, 87 2, 86 Overall F .548 10,316 .000 5.897 2.035 7.071 Note. ** p ≤ 0.01, * p ≤ 0.05
This means Hypothesis 3a-c are not supported, meaning that LMX does not mediate the positive relationship between Admiration and all three follower job performance variables.
4.2.4 Hypothesis 3d, 3e, and 3f
Previous regression analyses in this study already indicated a negative relationship between Rivalry and LMX (H1b) and a positive relationship between LMX and all three follower job performance variables (H2a-c). In order to test whether LMX mediates the negative
relationship between Rivalry and all three follower job performance variables, we used the Process Macro developed by Hayes (2013) using model 4.
Analyzing the direct effect of Rivalry on IRB reveals a negative effect (b = -.27; p = 0.00), however, LMX remains a significant predictor of IRB (b = .30; p = 0.00).
Approximately 37% of the variance in IRB is accounted for by Rivalry and LMX (R2 = .37). Then, the indirect effect is tested using a bootstrap estimation approach with 1000 samples. These results indicate that the indirect effect is negative and significant (effect = -.09, 95% CI: -,2072; -.0243). This effect is moderately strong (Baron & Kenny, 1986) and Hypothesis 3d is supported (see Table 5). This means that LMX mediates the negative relationship between Rivalry and IRB.
Analyzing the direct effect of Rivalry on OCBA reveals a negative effect (b = -.13; p = 0.05), however, LMX remains a significant predictor of OCBA (b = .27; p = 0.00)
Approximately 21% of the variance explained in OCBA is accounted for by Rivalry and LMX (R2 = .21). Then, the indirect effect is tested using a bootstrap estimation approach with 1000 samples. These results indicate that the indirect effect is negative and significant (effect = -.08, 95% CI: -.2487; -.0029). This effect is moderately strong and Hypothesis 3e is
supported (see Table 6). This means that LMX mediates the negative relationship between Rivalry and OCBA.
Analyzing the direct effect of Rivalry on OCBC reveals a nonsignificant effect (b = -.13; p = 0.14), while LMX remained a significant predictor of OCBC (b = .36; p = 0.00). Approximately 21% of the variance explained in OCBC is accounted for by Rivalry and LMX (R2 = .21). However, the absence of a direct effect does not necessarily mean there is
no indirect effect. Several authors argue that there still could be an indirect effect in absence of a total or direct effect (MacKinnon et al., 2002; Rucker et al., 2011; Shrout & Bolger, 2002; Zhao et al., 2010). The results indicate that the indirect effect is negative and significant (effect = -.10, 95% CI: -.2502; -.0280). This effect is moderately strong and Hypothesis 3f is supported (see Table 7). This means that LMX mediates the negative relationship between Rivalry and OCBC.
Table. 5 Bootstrapping results of LMX as a mediator in the relationship between Rivalry and in-role behavior
Variables
Bootstrapping 95% Confidence Interval
Effect Boot SE Lower Upper
Indirect effect
LMX -.0904 .0467 -.2072 -.0243
Direct effect of RIV on IRB
-.2684 .0656 -.3989 -.1379
Note. Relationship tenure was included as control variable
Table. 7 Bootstrapping results of LMX as a mediator in the relationship between Rivalry and challenging OCB
Variables
Bootstrapping 95% Confidence Interval
Effect Boot SE Lower Upper
Indirect effect
LMX -.1042 .0550 -.2502 -.0248
Direct effect of RIV on OCBC
-.1262 .0842 -.2937 -.0413
Note. Relationship tenure was included as control variable
Table. 6 Bootstrapping results of LMX as a mediator in the relationship between Rivalry and affiliative OCB
Variables
Bootstrapping 95% Confidence Interval
Effect Boot SE Lower Upper
Indirect effect
LMX -.0800 .0566 -.2487 -.0029
Direct effect of RIV on OCBA
-.1314 .0671 -.2648 -.0019
4.3 Results study 2
4.3.1 Hypothesis 4a, 4b, and 4c
This study used hierarchical regression analysis to test the curvilinear effect of leader
narcissism on all three follower job performance variables as predicted by Hypothesis 4a, 4b, and 4c. Therefore, there will be three tables, each discussing the effect according to a
follower job performance variable. The predictor and moderator variable are centered to reduce possible multicollinearity (Aiken & West, 1991).
First, the curvilinear effect of leader narcissism on IRB is analyzed (Table 8). In the first step (Model 1), the control variable is entered. No significant effect is found (b = -.08, p = 0.46). In Step 2 (Model 2), NARQ is entered, which has a negative effect on IRB (b = -.31, p = 0.00) and accounted for an additional 9.9% of the variance. In Step 3 (Model 3), the quadratic term of NARQ (i.e., NARQ squared) is added to the equation, which is non-significant (b = .01 p = 0.96). Thus Hypothesis 4a is not supported, suggesting there is no curvilinear effect of leader narcissism on IRB.
Second, the curvilinear effect of leader narcissism on OCBA is analyzed (Table 9). In the first step (Model 1), the control variable is entered. No significant effect is found (b = .00, p = 0.99). In Step 2 (Model 2), NARQ is entered and, which has a non-significant effect on OCBA (b = -.12, p = 0.26). In Step 3 (Model 3), the quadratic term of NARQ (i.e., NARQ squared) is added to the equation, which is non-significant (b = .07, p = 0.51). Thus
Hypothesis 4b is not supported, suggesting there is no curvilinear effect of leader narcissism on OCBA.
Third, the curvilinear effect of leader narcissism on OCBC is analyzed (Table 10). In the first step (Model 1), the control variable is entered. No significant effect is found (b = .15,
p = 0.16). In Step 2 (Model 2), NARQ is entered, which has a non-significant effect on OCBC (b = -.09, p = 0.39). In Step 3 (Model 3), the quadratic term of NARQ (i.e., NARQ squared) is added to the equation, which is non-significant (b = .08, p = 0.45). Thus
Hypothesis 4c is not supported, suggesting there is no curvilinear effect of leader narcissism on OCBC.
Overall, the results indicate that there is no significant curvilinear effect of leader narcissism on all three variables of follower job performance1.
Table 8. Regression analysis of the curvilinear effect of leader narcissism on in-role behavior
Variable Model 1 Model 2 Model 3 (H4a)
Control variable Rel -.08 -.09 -.09 NARQ NARQ -.31** -.32** NARQ2 .01 DR2 .006 .01 .00 R2 .006 .11 .11 Degrees of freedom 1, 86 2, 85 3, 84 Overall F .55 4.97 3.27 Note. ** p ≤ 0.01, * p ≤ 0.05
1 Additional curvilinear analysis on the relationship between each of the different dimensions of leader
narcissism and all three follower job performance variables revealed no significant effect.
Table 10. Regression analysis of the curvilinear effect of leader narcissism on challenging OCB
Variable Model 1 Model 2 Model 3 (H4c)
Control variable Rel .15 .14 .16 NARQ NARQ -.09 -.11 NARQ2 .08 DR2 .02 .01 .01 R2 .02 .03 .04 Degrees of freedom 1, 87 2, 86 3, 85 Overall F .00 .65 .57 Note. ** p ≤ 0.01, * p ≤ 0.05
Table 9. Regression analysis of the curvilinear effect of leader narcissism on affiliative OCB
Variable Model 1 Model 2 Model 3 (H4b)
Control variable Rel .00 -.01 -.01 NARQ NARQ -.12 -.14 NARQ2 .07 DR2 .00 .02 .01 R2 .00 .02 .02 Degrees of freedom 1, 87 2, 86 3, 85 Overall F .00 .65 .57 Note. ** p ≤ 0.01, * p ≤ 0.05
5. Discussion and implications
This is the first study analyzing the effects of leader narcissism on follower job performance in a dyadic work context. The purpose of this study was to solve current inconsistent findings on the relationship between leader narcissism and organizational outcomes. A part of this problem was that there might be different “flavors” of leader narcissism hidden in the construct that each have a different effect on organizational outcomes and that the relationship is not linear, but curvilinear. Therefore, we first looked at whether the
relationship between the different dimensions of leader narcissism (developed by Back et al., 2013) and follower job performance might be explained by the perceived LMX of the
follower. Second, we analyzed whether there might be a curvilinear relationship between leader narcissism as a whole and follower job performance. In this study, the NARC and its distinctive dimensions, namely Admiration and Rivalry, was used to test their interpersonal consequences in relation to follower job performance. Additionally, the NARC as a whole was used to look for its curvilinear effect on follower job performance. Follower job performance is comprised of IRB, OCBA, and OCBC, because all three performance measures seem relevant in doing leadership research related to LMX (Ilies et al., 2007) and because LMX has shown to predict IRB and OCB’s (Bauer et al., 2006; Walumbwa et al., 2009; Walumbwa et al., 2011; Wang et al., 2005)
The first two hypotheses of this study examined how Admiration and Rivalry each related to LMX. It was expected that Admiration would show a positive relationship to LMX and Rivalry a negative relationship to LMX, which would subsequently lead to respectively a positive effect and a negative effect on follower job performance. Rivalry indeed showed a negative relationship to LMX and subsequently had a negative indirect relationship with all three follower job performance variables. However, Admiration showed a non-significant effect to LMX and had no indirect effect on all three follower job performance variables.
These results contradict the findings by Back et al. (2013), who found that people high in Admiration show self-assured and assertive behaviors and are perceived by others as
attractive, sociable, and assertive. They also found that such people show behaviors that that are necessary for maintaining close relationships, such as such as gratitude and direct
problem-focused reaction in conflicts (Back et al., 2013). This would subsequently lead to positive social interaction outcomes, which is a prerequisite for high-quality exchange relationships (Ilies et al., 2007).
A possible explanation for why these positive social interaction outcomes might be overestimated can be traced back to the underlying motivational mechanism of Admiration: self-enhancement. It is not only aimed at having a high self-esteem, but also repeatedly reinstating one’s ‘grandiose’ self by feeling admired and special (Back et al., 2013, p. 1016). This is reflected most clearly in the first six items of this dimension, which emphasize how good one feels about oneself. For example: “I am great” or “I will someday be famous” are questions that do not need to be answered in relation to another person. This means that Admiration is not necessarily measured against an interpersonal background. On the contrary, the underlying motivational mechanism of Rivalry, self-protection, “triggers not only passive intrapersonal reactions such as the devaluation of others but also active social reactions such as revenge” (Back et al., 2013, p. 1016). All items in this dimension need to be answered in relation to another person, for example “I react annoyed if another person steals the show from me” and “Other people are worth nothing.”. This means that Rivalry is solely measured against an interpersonal background. The lack of interpersonal background in the Admiration dimension and its “self-focused” nature might explain its non-significant effect on LMX, contrary to Rivalry, that has clear negative interpersonal implications and is more “other-focused” in nature.
Another explanation for why Admiration had no significant positive effect on LMX could be that an important driver for high-quality exchange relationships is missing in the aspects that are necessary for maintaining close relationships with others: empathy. Back et al. (2013) showed that Rivalry had a strong correlation to a lack of empathy, which is detrimental for maintaining good relationships with other people, while Admiration was not significantly related to empathy. Research has shown that when a leader possesses empathy, or the ability to understand and recognize the feelings of others, developing relationships of mutual trust with subordinates become easier (Mahsud, Yukl & Prussia, 2010). Also, self-rated leader-empathy has shown to predict higher levels of follower ratings of LMX (Gordon, 2007). In line with this, Rivalry showed a negative relationship with LMX, presumably because of a significant lack of empathy, and Admiration was not related at all to LMX, what might be due to a non-significant presence of empathy. Does this mean that a narcissist cannot maintain high-quality exchange relationships at all? A recent study developed by Hepper, Hart and Sedikides (2014) showed that narcissists can be moved to empathize with another person if the right conditions are set. If narcissists were instructed to take the perspective of another person into account, they were capable of emphatic responding. This implies that targeting training and interventions relating to perspective-taking might help improve a narcissist interpersonal relations, and thus the ability to develop higher-quality exchange relationships.
The second study investigated whether there was a curvilinear effect of leader narcissism on all three follower job performance variables. The results showed that there is no significant curvilinear effect. A possible explanation for this could be that the behavioral threshold theory of the NPI developed by Ackerman et al. (2012) does not entirely fit with the NARC. This theory argued that ‘easy’ items, or those that are low in threshold, are endorsed in when the respondent possesses low to average levels of narcissism. These items
reflect more adaptive aspects of narcissism. Items that are high in threshold, or ‘difficult’ items, are endorsed in when the respondent possess higher levels of narcissism, in which maladaptive aspects of narcissism come into play. Eventually, the maladaptive aspects overshadow the adaptive aspects when the level of leader narcissism rise, turning the positive relationship to Y into a negative relationship, creating an inverted U-shape pattern. However, the behavioral threshold of the Admiration dimension might be higher or reflect more
‘difficult’ items. For example, “I am great” or “One day I will be famous” (NARQ) are more extreme items than “I think I am a special person” or “I will be a success” (NPI-40). There is a difference between thinking you are special or actually stating it as a fact. This suggests that if respondents score high on adaptive aspects of the NARC, this could actually reflect a higher level of narcissism and is thus more maladaptive in reality. Therefore, the extent to which Admiration shows adaptive aspects of narcissism, as argued by Back et al. (2013), is questionable. This could explain why there is no curvilinear effect of leader narcissism on follower job performance, since a moderate amount of leader narcissism actually reflects more maladaptive aspects, thus harming follower job performance.
To conclude, there is a negative relationship between Rivalry and all three follower job performance variables and this relationship is mediated by LMX. This means that leaders high on Rivalry aspects are associated with lowered levels of LMX perceived by their
followers, which, subsequently, negatively relates to follower job performance. LMX is also positively related to all three follower job performance variables. The results suggest that including relationship-based constructs, such as LMX, in examining the relationship between leadership and follower outcomes is highly relevant. It can explain what is happening inside the "black box" between observed leadership aspects and measured follower outcomes (Howell & Hall-Merenda, 1999). However, there is no significant positive relationship between Admiration and follower job performance that is mediated by LMX. This might be
due to a lack of interpersonal background in the measurement of the construct or to the fact that it is unrelated to empathy, which is an important prerequisite for developing high-quality exchange relationships. Finally, there is an absence of a curvilinear effect of leader
narcissism on follower job performance, what might be caused by the underestimation of the behavioral threshold for Admiration or overestimating its adaptive aspects earlier stated by Back et al. (2013).
5.1 Practical implications
The current research aims to address the need in providing a better understanding on how leader narcissism affects the job performance of followers. This understanding has important practical implications. Firstly, because in-role behavior, but also extra-role behaviors have shown to be of great importance for the profitability and survival of firms (Koys, 2001; Fisher et al., 2010). Secondly, because narcissists are frequently found to emerge as leaders (Brunell et al., 2008; Judge et al., 2006; Nevicka et al., 2011), research on this topic is important for practical reasons.
The results of this study indicate that an important explanatory mechanism between leadership and the job performance of followers is the quality of the exchange relationship between managers and employees. Strong correlations exist between LMX and follower job performance, but also between leadership traits, such as Rivalry, and LMX. This means that the job performance of employees could be indirectly affected by how a manager’s
personality come into play in the workplace. Therefore, special attention should be paid to developing high-quality exchange relationships between managers and employees, but also caution should be paid to the “maladaptive” narcissistic traits of managers.
Finally, when selecting applicants into new leadership roles, measuring their level of narcissism on the whole construct seems less effective than measuring it on each of the two dimensions, because each dimension has different outcomes. Especially the Rivalry aspects of a leader has far more negative interpersonal consequences than Admiration has. Therefore, it is unwise to select applicants high on Rivalry, because this is negatively related to aspects that are necessary for maintaining high-quality exchange relationships with employees, which is subsequently negatively related to employee job performance. However, if people score high on Admiration, this does not necessarily mean that high-quality exchange relationship between managers and employees cannot develop. If adequate training efforts are designed to develop the ability of leaders on perspective-taking, additional research has shown that these people can engage in empathic behaviors (Hepper et al., 2014), which is beneficial for developing a high level of LMX as perceived by the employees.
5.2 Strengths and limitations
A strength of this study is its use of multi-source data. To reduce common method variance, this study collected the measures of the predictor and criterion variables from different sources: followers and leaders (Podsakoff, MacKenzie, Lee & Podsakoff, 2003). Specifically, an additional strength is that follower job performance is rated by the
supervisor. Often promotions or salary decisions are made by the supervisor, which makes supervisor-rated follower job performance practically important in this study. This shows a good indication of how well employees are actually performing. Including peer-reported job performance ratings could broaden the view on the job performance of employees. However, these type of job performance ratings are also biased, because employees might want to avoid conflicts with their peers and therefore might overestimate this rating.
