i
Master Thesis
Narcissistic Leadership
:
The influence of gender and leader-member exchange on evaluations
of effective leadership
Student
Stella van Hoenselaar
Student number6228658
Supervisor
Prof. Dr. D. N. Den Hartog
June 2016
ii
Statement of Originality
This document is written by Student Stella van Hoenselaar 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.
iii
Abstract
Based on social role theory, this study tests the proposition that female narcissistic leaders are rated as less effective than their male counterparts, because of the violation of their traditional communal gender role. Furthermore, leader-member exchange (LMX) is studied as a possible mediator in the relationship between narcissism and perceived leader effectiveness. Research was done with the recently developed NARQ scale as a measure of narcissism, which allows for the distinction between two different dimensions; narcissistic admiration and narcissistic rivalry. Data of 123 leader-follower dyads shows that narcissism in leaders has a negative effect on perceived leader effectiveness and that specifically the rivalry dimension accounts for this effect. This negative effect, which is found for narcissism and narcissistic rivalry, is partly mediated by LMX. A moderating effect of gender is found for the narcissistic rivalry dimension, with women being rated as less effective than men, but only when LMX is controlled for. Further exploratory analysis finds that this moderating effect of gender specifically takes place at low levels of LMX, meaning that women that score high on narcissistic rivalry are rated as less effective than men, but only when quality of LMX is low. Implications and directions for future research are discussed.
.
Key words: Leadership; Narcissistic Leadership; Gender; NARQ; narcissistic rivalry;
iv
Table of Contents
Page
1. Introduction……… 1
2. Theoretical Background………... 4
2.1 The development of Narcissism as a trait……… 4
2.2 Narcissism & Leadership………... 5
2.3 Narcissism & leader- member exchange………. 9
2.4 Narcissism & Gender …….………. 11
2.5 Narcissism, Gender & LMX……… 14
3. Research Method………... 15
3.1 Procedure………... 15
3.2 Sample………..16
3.3 Measures……….. 16
4. Data analysis & Results………... 17
4.1 Data Analysis………... 17 4.2 Descriptive Statistics……… 18 4.3 Regression……… 20 4.4 Mediation………. 22 4.5 Moderation………... 24 4.6 Moderated mediation………... 26 4.7 Exploratory analysis……….30 5. Discussion……….. 32 5.1 Implications………..34
5.2 Limitation & directions for future research………. 36
6. Conclusion………. 37
Literature……… 39
1
1. Introduction
Successful companies always strive to improve business, enhance performance and maximize profit. As a result, a vast amount of research is done towards discovering ways that can help them achieve these goals and identify factors that can contribute to their success. Leadership is one of those factors. Studies suggest that effective leadership influences several aspects of firm performance (Thomas, 1988). Ever since, research is interested in what attributes make up a successful leader. A fairly new topic of interest in this field is leader narcissism, which in the context of leadership, is seen as an attribute or trait that includes grandiosity,
entitlement, dominance, and superiority (Raskin & Hall, 1979). Narcissism is found to be prevalent in leaders, but is also associated with leader emergence (Campbell, Hoffman, Campbell & Marchisio, 2011). Consequently, a discussion is held among management theorists about the effects of narcissism on effective leadership. No consensus is reached so far, however most research indicates more negative than positive effects of narcissism on leader effectiveness. (Rosenthal & Pittinsky, 2006).
The underlying factors that are responsible for the effect have remained largely unstudied. However, one factor that is associated with both leader narcissism and perceived leader effectiveness is leader-member exchange (LMX). In business theory, this term is often used to denote the quality of the relationship between leader and follower. LMX is low when relationships are characterized by downward influence and role-defined relationships. When LMX is high, relationships are characterized by mutual trust, respect, liking and reciprocal influence. High quality of LMX is associated with both high performance ratings and high leader effectiveness ratings (Kacmar, Witt, Zivnuska & Gully, 2003; Deluga & Perry, 1991). Shuns (2015, p.131) also theorizes that leaders that score high on narcissism are unlikely to form positive LMX relationships, as the leader will not accept the followers’ relevance and therefore will not seek out a positive relationship. At the same time, the followers are also
2 unlikely to seek out a positive relationship as this would imply almost complete submission to the leader. Therefore, narcissism in leaders possibly has a negative effect on perceived leader effectiveness, through a low quality of LMX.
As with several other (leadership) attributes, an age independent gender difference exists in the prevalence of narcissism, with men generally being more narcissistic than women (Grijalva, Newman, Tay, Donnellan, Harms, Robings and Yan, 2015). The research about gender differences in leadership roles does not limit itself to the topic of narcissism. Men still largely outnumber women in top management positions despite increased
awareness and efforts to change this fact (Eagly & Carli, 2007). These inequalities cannot be addressed to actual differences in the leadership behavior of men and women, which raises the question why they actually exist (Eagly, Johannesen-Schmidt, & Van Engen, 2003).
The most recognized explanation is based upon the social role theory, which explains how the same behavior is interpreted differently depending on whether it is enacted by a male or female. According to this theory, women are expected to be communal, where men are expected to be agentic (Eagly, 1987). When a person does not behave consistent with these expectations, they are evaluated more negatively, especially when showing undesirable qualities associated with the other gender. This means that because male gender roles are more related to typical leadership roles than those of women, it is easier for men to be perceived more positively than women when they engage in leadership behavior (Eagly & Karau, 2002). The behavioral expressions associated with narcissism also mainly resemble the masculine sex role, including physical expression of anger, strong need for power and an authoritative leadership style (Corry, Merritt, Mrug, & Pamp, 2008, p. 593 Grijalva, 2015). It is therefore expected that women engaging in narcissistic behavior, will be perceived more negatively than men engaging in the same behavior. This is consistent with results found in practice (De Hoogh, Den Hartog, & Nevicka, 2015).
3
This effect possibly takes place in two forms. It can influence evaluations of leader effectiveness directly, but it could also have in indirect effect by influencing the quality of LMX which in its turn influences perceived effectiveness. Evaluations of leader effectiveness are thought to depend mainly on perceived credibility and trustworthiness (Campbell, 1991; Harris & Hogan, 1992; Lombardo, Ruderman & McCauley, 1988) whereas quality of LMX includes different factors of evaluation, under which member affect about the relationship (Liden, 1985; Day & Cram, 1992). Social role theory however states that violations of the traditional gender role leads to negative evaluations, but whether these are evaluations of likeability, capability or even something else is not yet clear. Therefore, both possibilities are tested.
The aim of this study is twofold. Firstly, LMX as a possible explanation for the relationship between leader narcissism and perceived leader effectiveness is studied. Secondly, the influence of gender on this relationship is examined.
Hereby, the research determines different factors influencing leader effectiveness, which can have important consequences for organizational performance. At the same time, it provides more insight in the factors that entail this relationship. Lastly it provides insight in a possible source of gender inequality in leadership whilst providing a probable means to compensate for this inequality. See figure 1. For the complete research model
4
2. Theoretical Background
2.1 The Development of Narcissism as a Trait.
The term narcissism, derived from the Greek myth of Narcissus, was first used by Ellis (1898) to describe a clinical condition of ‘perverse self-love’. This concept was further developed over time. Freud described a specific narcissistic personality type associated with confidence, arrogance and what he called ‘a limited susceptibility to influence’ (Cratsley, 2016). Later Horney (1939) suggested that the personality traits exhibited such as self-inflation and self-admiration, are not qualities a narcissist actually possesses. The idea that narcissism is a personality disorder, was further advanced by Kernberg (1989), who describes narcissistic people as having an unusual degree of self-reference in their interactions, a contradiction between inflated self-concept and inordinate need to tribute from others, shallow emotional lives, envy, and extremes of idealization and devaluation of others. He also describes narcissistic people as having a charming and engaging presence that masks an underlying coldness and lack of empathy. Currently, Narcissism is recognized as an official personality disorder. In the Diagnostic and Statistical Manual of Mental Disorders - 5 (APA, 2013), narcissistic personality disorder is defined as “a pervasive pattern of grandiosity (in fantasy and behavior), need for admiration and lack of empathy, beginning by early
adulthood and present in a variety of contexts” (Grijalva, 2015, p. 645). It continues to be a topic of interest in social and psychological research.
Often a distinction between narcissism as a psychological disorder and narcissism as a trait in the general population is made. In management research, the latter definition on narcissism is most often used. From this perspective, narcissism is described as an attribute that exists on a continuum. In this line of reasoning, Malkin (2015, From Breton, 2015 p. 92) states that narcissism can sometimes also be beneficial and necessary to develop a healthy self. He describes a spectrum where the center is a healthy amount of narcissism; ‘people
5 who safely depend on others and who can develop caring relationships through true
intimacy’. Both far ends of the spectrum are pathological, with 0 being someone that refuses all attention and care and 10 being a pathological narcissist.
What behaviors exactly make up narcissism differs slightly among theorists, but it is thought to include arrogance, entitlement, dominance, superiority, self-absorption, fragile self-esteem, hostility, lack of empathy, hypersensitivity and anger (Raskin & Hall, 1979, Rosenthal & Puttinsky 2006). In this thesis, it is the latter interpretation that is taken as a starting point for this study; the idea that narcissism is a trait that everyone possesses to a greater or lesser extent.
2.2 Narcissism and Leadership
The occurrence of narcissism in people with leadership functions is often studied, partly because the prevalence of highly narcissistic people is bigger in this environment than it is in others. There is still an ongoing debate about the relationship between narcissism in leaders and organizational outcomes, and both negative and positive effects have been found.
Narcissism is linked to several negative outcomes such as ratings of interpersonal performance and integrity (Blair, Hoffman & Helland, 2008) and white collar crime (Blicke, Schlegel, Fassbender & Klein, 2006). At the same time, narcissism is associated with several positive outcomes as well. It shows a relationship with certain aspects of psychological health, which is mediated by self-esteem (Sedikides, Rudich, Gregg, Kumashiro & Rusbult, 2004). This duality is illustrated by one of the most distinct attributes of narcissism, which is arrogance. Some research proposes that arrogance is the driving force behind the grand vision that a leader needs (Maccoby, 2004). However, it is also associated with difficulties in
interpersonal relationships (Ronningstam, 2005), which can be a hinder for successful leadership (Spector, 2003).
6 Management research found that narcissism in CEOs is related to bold actions that attract attention (strategic dynamism, number and size of acquisitions) leading to big wins or big losses (Chatterjee & Hambrick, 2007). Rosenthal and Pittinsky (2006) give an extensive review about the different effects of narcissistic leaders. They note that such leaders are generally more motivated by their need for power and admiration (Kets de Vries & Miller, 1997) than their empathetic concern for their followers and the institution they lead (Conger, 1998). They demand unquestioning devotion and loyalty from followers (Harwood, 2003). This is a key motivation for narcissists to seek leadership positions, as this will give them the power they need to structure the ‘external world’ in such a way that supports their grandiose needs and visions (Glad, 2002). Also, the lack of empathy which is associated with
narcissism (Brown & Bosson, 2011), can be inimical to good leadership (white, 1994). This is because empathy is considered a key aspect of emotional intelligence, which is seen as an important trait for good leadership (Kellet, Humprey & Sleeth, 2006).
Similarly, it is argued that narcissists tend to make management decisions that are not in the best interest of organizations and its employees/stakeholders, but rather in their own best interest (Holain, 2006; Morf & Rhodewalt, 2001). Furthermore, they reported to be poor listeners and sensitive to criticism (Maccoby, 2000). Other traits thought to decrease their leadership qualities are irrationality, inflexibility and paranoia (Glad, 2002). In addition, Judge, LePine & Ridge found that narcissism in managers is negatively related to other people’s rating of leadership.
At the same time, narcissists also possess the charisma and grand vision that are important for effective leadership (Rosenthal & Pittinsky, 2006). As visionaries, they are able to inspire great numbers of followers (Maccoby, 2000). Also the confidence and dominance that narcissists show are in some cases exactly the reasons that a group of followers selects them as a leader (Gladwell, 2002; Hogan et al., 1994, Post, 1986).
7 In conclusion, both positive and negative effects of narcissism in leaders exist.
However, there is still no clear conclusion on whether narcissism has a negative or a positive effect on leader effectiveness. This question will therefore be studied further in this thesis. This current study will focus on one specific part of leader effectiveness, namely leader effectiveness as perceived by stakeholders, and more specifically by followers. This kind of leader effectiveness (as opposed to objective, hard measures like organizational performance) is chosen for several reasons. Firstly, obtaining organizational performance measures is not compatible with the nature and restrictions of current study. Secondly, leadership traits are more likely to influence subjective leader effectiveness than objective measures. This is because leader traits directly influence the way leaders are perceived by followers, but do not directly affect objective measures of leader performance, like sales numbers. Furthermore, it is argued that the traits that predispose someone to emerge as a leader, also enable them to better translate their position into effectiveness, at least in the eyes of a stakeholder (Judge, Piccolo & Kosalka 2009).
Related to this Judge et al. (2009) also argue that according to implicit leadership theories, the more closely a leader’s traits matches stakeholder prototypes for judging effectiveness, the more effective that leader will be perceived. It is arguable that most behavioral expressions related to narcissism do not match most people’s implicit ideas about effective leadership because they include grandiosity, aggression and lack of empathy. This leads to the following hypothesis:
Hypothesis 1: Leader narcissism has a negative effect on perceived leader effectiveness.
8 The different effects found for narcissism might also be related to the way the concept of narcissism is defined and which aspects are exactly measured in different studies. The most used narcissism questionnaire is the NPI, which allows for dividing the construct of
narcissism in 3 different facets. The first of these is the exploitativeness/entitlement facet which is often seen as the most maladaptive facet of the NPI and is associated with toxic behaviors like aggression. The second facet of leadership/authority reflects an individuals’ motivation to lead and desire for authority and power. This is generally considered the most adaptive facet. The last facet of grandiose/exhibitionism illustrates the features of self-love and theatrical self-presentation (Grijalva, 2015, p. 265, 267).
This study however uses the recently developed NARQ questionnaire. This
questionnaire allows for a different distinction; the one between narcissistic ‘admiration’ and ‘rivalry’. In their article, the developers present a process model that is based on the notion that narcissists’ main goal is to maintain a grandiose self. According to them, perseverance of the grandiose self can be achieved by two different strategies. The first strategy of narcissistic admiration maintains a grandiose self by approaching social admiration by means of self-promotion (assertive self-enhancement). In the second strategy of narcissistic rivalry, a grandiose self is maintained by preventing social failure by means of self-defense (antagonistic self-promotion). The admiration facet shows a lot of similarity with the leadership/authority facet of the NPI and also shows some overlap with the
grandiose/exhibitionism facet. The rivalry facet is mainly related to the exploitativeness/entitlement facet of the NPI.
The NARQ as a measurement tool shows good internal consistencies and stabilities and shows both convergent- and discriminant- as well as predictive validity higher than that of the NPI (Back, Küfner, Dufner, Gerlach, Rauthmann, & Denissen 2013). In this study the effects of both dimensions of narcissism are studied separately. It is argued that both will
9 have a negative effect on perceived leader effectiveness. The negative effect of narcissistic rivalry is expected to be stronger than the effect of narcissistic admiration, as the behavior associated with self-defense (like aggression and deterioration of others) will have a clear negative effect on employee’s evaluations of their leader. The behavior associated with self-enhancement (arrogance and grandiosity) will not have that strong of an effect, as the leaders goal is to have his employees admire him. However, arrogance and grandiosity in
combination with a lack of empathy, will ultimately lead to negative evaluations of the leader as well.
Hypothesis 2: Narcissistic admiration in leaders has a negative effect on perceived leader effectiveness
Hypothesis 3: Narcissistic rivalry in leaders has a negative effect on perceived leader effectives and this effect is stronger than the effect of narcissistic admiration.
2.3 Narcissism and Leader-Member Exchange.
The relationship between leader and member is operationalized as leader-member exchange (LMX). LMX theory suggests that leaders differentiate between their subordinates and develop a different style of exchange with each of them. The relationships range from being characterized by downward influence and role-defined relations (Low LMX) to those characterized by mutual trust, respect, liking and reciprocal influence (High LMX). This theory proposes that the type of exchange relationship influences many organizational outcomes (Liden, Wayne, & Stilwell, 1993, Sweetland, 1978). Earlier research already showed a direct positive relationship between LMX and satisfaction with supervision (Gerstner, & Day, 1997), evaluations of managerial performance (Murphy & Cleveland, 1991) and ratings of managerial effectiveness (Deluga & Perry, 1991).
10 As discussed earlier, Schyns (2015) theorizes that it is unlikely for narcissists to form positive LMX relationships with their subordinates as they will not acknowledge the
importance of this. Additionally, narcissists characteristically lack empathy (Brown & Bosson, 2001) and the tactics they use to gain admiration and affirmation (aggressing at others, self-aggrandizement) undermine interpersonal relationships in the long term (Morf & Rhodewalt, 2001). A study by Du (2012) confirms this idea by showing that narcissistic personality is negatively related to manager-ratings of interpersonal competencies. Therefore, it is argued that the negative effect of narcissism on leader effectiveness ratings is at least partly due to low quality relationships, leading to the following hypothesis:
Hypothesis 4: The negative effect of narcissism on perceived leader effectiveness is mediated by LMX.
Again a distinction is made between the two different dimensions of narcissism. Both the self-enhancement and self-defense behaviors associated with respectively the admiration- and rivalry dimension are linked to negative interpersonal relationships. The behaviors associated with self-defense, like aggression and deterioration of others clearly decrease interpersonal relationships (Myers & Johnson, 2003) and therefore have a negative effect on LMX. For the self-enhancement dimension, the extent to which the associated behaviors are enacted seems to be determinative. Slight to moderate self-enhancement appears to entail more social benefits than social costs (Alicke & Sedikides, 2009). In the long term, above average self-enhancement shows to have a negative effect on evaluation by others. As narcissism is associated with excessive self-enhancement it is therefore likely that this also has a negative effect on LMX. As this relationship is less clear than for the rivalry aspect of narcissism, it is expected that this effect is less strong.
11 Hypothesis 5: the negative effect of narcissistic admiration on perceived leader
effectiveness is mediated by LMX
Hypothesis 6: the negative effect of narcissistic rivalry on perceived leader effectiveness is mediated by LMX, and this effect is stronger than the effect of narcissistic admiration.
2.4 Narcissism and Gender
Other factors are also thought to influence the effect that leader narcissism has on perceived leader effectiveness. As discussed before, social role theory gives insight in the effect that gender has on how different behaviors are interpreted. Studies show this effect in several different areas. Female leaders are reported to receive lower effectiveness rating when expressing anger, whereas this effect does not occur for men (lewis, 2000). At the same time, male leaders are perceived as less effective when committing errors in a masculinized
domain, whereas this effect does not occur for women (Thoroughgood, Sawyer & Hunter, 2013). Social role theory proposes that stereotypical beliefs about males being more agentic and females being more communal stem from observations of the two sexes in specific societal roles that are thought to require different personality traits. This means that as men more often occupy the breadwinner and authority roles this cause people to perceive men as more agentic. As women are more often in communally demanding employment rolls, this causes people to infer that women are especially communal (Cejka & Eagly, 1999). Behind these beliefs is an important principle known as the correspondence bias: the inference that people’s behavior reflects their psychological disposition (Gawronski, 2003). This principle also implies that gender-stereotypical beliefs disappear when men and women occupy the same role. This has been confirmed several times (Bosak, Sczensy & Eagly, 2011). These
12 findings have important implications, namely that when social roles are taken on by men and women in around the same ratio, stereotypes might disappear.
So far however, management roles are still more often take on by men than women. In the past years this ratio has slowly changed in favor of women. Duehr and Bono (2006) test whether this increase in diversity has also led to a subsequent change in perceptions of women as a leader. They did find that male managers view of women as leaders did indeed change. This was evidenced by a greater congruence between their perceptions of women and successful managers and stronger endorsement of agentic and task-oriented leadership
characteristics for women. The same authors also note that over the years, management work has also been described in qualities that are traditionally defined as feminine, such as helping and developing others. This is an indication that a shift might be taking place towards less gender stereotypes when it comes to managerial roles in general. Partly because gender stereotypes are changing, but also because the idea of effective leadership is changing. When it comes to narcissistic leadership specifically, this is still a leadership style which mainly shows attributes that are traditionally male roles. Based on this it seems likely that female leaders that show narcissistic behavior will be evaluated as less effective than men showing this behavior.
This idea has been confirmed in a study by De Hoogh, Den Hartog and Nevicka (2015) who showed that female narcissistic leaders were perceived as less effective than male narcissistic leaders. Moreover, they found that this gender difference was only found when subordinates were male. When female subordinates served as raters, no gender bias was found in effectiveness evaluations. This study aims to replicate these results, but this time with the NARQ scale as a measure of narcissism instead of the NPI (Back, Küfner, Dufner, Gerlach, Rauthmann and Denissen, 2013). The gender of the follower will not be taken into
13 account in this study as this is not the topic of interest in this thesis. Instead, other variables are additionally researched.
Hypothesis 7: The negative relationship between leader narcissism and perceived leader effectiveness is moderated by leader gender, so that this relationship is stronger for female leaders than for male leaders.
It is also arguable that the relationships between the two dimension of narcissism as divided by the NARQ and leader effectiveness are influenced differently by the gender of the leader. The behavior shown in narcissistic rivalry does more resemble the traditional male role than the narcissistic admiration dimension, due to its assertive nature, elements of hostility and proneness to social conflict (Lange, Hagemeyer, & Crusius, 2014). However, the self-enhancement dimension also contains some behaviors that are associated with the more agentic male role, like grandiosity and entitlement. Therefore, it is theorized that for women the expression of both dimension of narcissism as proposed by the NARQ will lead to lower perceived leader effectiveness than for men, but that this effect is stronger for the rivalry aspect of narcissism than the admiration aspect. This leads to the following hypothesis:
Hypothesis 8: The negative relationship between leader narcissistic admiration and perceived leader effectiveness is moderated by leader gender, so that this relationship is stronger for female leaders than for male leaders.
Hypothesis 9: The negative relationship between leader narcissistic rivalry and perceived leadereffectiveness is moderated by leader gender, so that this relationship is
14 stronger for female leaders than for male leaders. This moderating effect is expected to be stronger than for the narcissistic admiration dimension.
2.5 Narcissism, Gender and LMX
As previously discussed, gender is expected to have an influence on overall ratings of perceived leader effectiveness. However, it is also possible that gender has an influence on the quality of leader-member exchange. As previously argued, women are expected to be evaluated as less effective than men, when showing narcissistic behavior. In this evaluation managers are rated on factors like credibility and trustworthiness as perceived by the subordinate (Campbell,1991; Harris & Hogan, 1992; Lombardo, Ruderman and
McCauley,1988). Along the same line of reasoning, it can be argued that when women show narcissistic behavior they are perceived as less likable than men, which is a factor known to be associated more with LMX than with perceived effectiveness (Liden, 1985; Day & Cram, 1992). This lower quality of LMX can in its turn lead to lower perceived leader effectiveness. Leading to the following hypothesis:
Hypothesis 10: Leader narcissism is related to perceived leader effectiveness via conditional indirect effects, such that the interaction between gender and leader narcissism is related to LMX, which in its turn is related to perceived leader effectiveness.
Both dimensions of narcissism are again analyzed separately as well and a stronger effect is again expected for narcissistic rivalry. Along the same line of reasoning both the behaviors associated with the rivalry dimension as well as the admiration dimension are associated more with the traditional agentic male role. Not only will this lead to women being rated as less effective than men, it will also lead to a lower quality of the relationship between leader
15 and member (LMX). There are however more behaviors in the rivalry dimension which are more clearly associated with the male role. Therefore, a stronger effect is expected for the rivalry dimension than for the admiration dimension. Leading to the following hypothesis:
Hypothesis 11: Leader narcissistic admiration is related to perceived leader effectiveness via conditional indirect effects, such that the interaction between gender and leader narcissism is related to LMX which in its turn is related to perceived leader
effectiveness.
Hypothesis 12: Leader narcissistic rivalry is related to perceived leader effectiveness via conditional indirect effects, such that the interaction between gender and leader narcissism is related to LMX which in its turn is related to perceived leader effectiveness.
3. Research Method
The next section explains how the research has been set up, how the data was collected and which measures were applied.
3.1 Procedure
For this study leader-follower dyads are researched, meaning that managers from different industrial sectors and one of their subordinates both participate in the study. They both fill out different surveys, which are administered online through a program called Qualtrics. For both questionnaires there was a Dutch as well as an English version available. As this study was carried out by a research group of five students, the survey was made up out of all
questionnaires that the students needed in order to answer their individual research questions. All research questions were related to narcissism. Data collection was done as a group, meaning that each student collected data of approximately 25 leader-follower dyads. The
16 leader- and follower surveys were connected together through a unique code which was provided when surveys weere distributed and were filled in by the respondents at the start of the survey.
3.2 Sample
For this study, 450 questionnaires were sent out. A total of 258 people responded, leading to a response rate of 57,3 %. In some cases, only the leader or follower questionnaire was filled in. These cases were excluded, leaving the data of 123 dyads for the analysis. Out of the leaders, 62 men and 57 were women. Under the followers were 48 men and 71 women.
3.3 Measures
All questionnaires that were used could be answered on a 7 point Likert-scale that ranges from 1 (not agree at all) to 7 (completely agree). See Appendix A for a complete overview of all survey questions.
Leader narcissism is measured with a scale adapted from Back, Kufner, Dufner, Gerlach, Rauthmann and Denissen (2013) called the NARQ. This scale consists of 18 items that includes measures of both different dimension of narcissism as discussed before, namely Narcissistic Admiration (9 items) and Narcissistic Rivalry (9 Items). Example questions for both dimensions are respectively ‘I deserve to be seen as a great personality’ and ‘I enjoy it when another person is inferior to me’. Reported Cronbach’s Alpha is .83 for the whole scale, .87 for admiration and .83 for rivalry.
Leader-member exchange is measured through a scale adapted from Liden and
Maslyn, (1998) and exists of 11 items, with questions such as ‘My supervisor is a lot of fun to work with’. The reported Cronbach’s alpha for this scale is .89.
17 Perceived leader effectiveness is measured by a 3-item scale adopted from De Hoogh, Den Hartog and Koopman (2005). This scale contains items like ‘The overall performance of my leader is satisfactory’. Reported Cronbach’s Alpha is .86Leader gender was measured through one self-report item, which was coded 1 for men and 2 for women.
Several control variables were included under which age, as narcissism is thought to decline with age (Foster, Campbell, & Twenge, 2003). Organizational tenure (amount of time worked at current organization) is also included as it is reported that the attributes narcissists exhibit leads them to sometimes make positive first impressions, but these same attributes are maladaptive in the long run. (Back, Schmukle, & Egloff, 2010).
4. Data analysis & Results
The next paragraph explains how data analysis was done in order to facilitate replication.
4.1 Data Analysis
Data analysis was done with IBM SPSS Statistics 24. First, reliability measures for all questionnaires were obtained. The NARQ scale for narcissism has high reliability, with a Cronbach’s Alpha of .892. The corrected item-total correlations indicate that all the items have a good correlation with the total score of the scale (all above .30). Also, none of the items would substantially affect reliability if they were deleted. Admiration as a separate scale has a Cronbach’s alpha of .875. The corrected item-total correlations indicate that the question: ‘mostly, I am very adept at dealing with other people’ does not show a good correlation with the total score of the scale (.277). Removing this question significantly increased reliability (.11), which led to the elimination of this question. The final Cronbach’s alpha was .886. The rivalry dimension of the NARQ shows a Cronbach’s alpha of .866 and all corrected item-total correlations are above .30.
18 The Leader effectiveness scale and the LMX scale both show high reliability with Cronbach’s Alphas of respectively .875 and .909. Scale means were computed by averaging the items belonging to each scale. Three respondents had one missing value on the NARQ scale, and one respondent had one missing value on the LMX scale. The data of these
respondents has still been used in the analysis as they are not thought to significantly alter the personal average. Data was tested on normality through measures of skewness and kurtosis and the Kolmogorov-Smirnov test. On this basis it was concluded that only the scale of narcissistic admiration has data that is normally distributed.
However, since non-parametric tests were not available for all variables, parametric tests were used throughout the research. Descriptives and correlations were obtained and lastly hypotheses were tested through regression, mediation- and moderation analysis.
4.2 Descriptive Statistics
Table 1 shows means, standard deviations and correlations of all variables used in the analysis. Both Narcissistic rivalry and admiration show a positive correlation with the construct of narcissism, which is an indicator that these constructs are indeed mutually related. It is noteworthy that leader narcissism has a negative correlation with leader effectiveness, which is an indicator that a negative relationship between the two variables does exist. However, looking at admiration and rivalry separately, data shows that only rivalry is significantly negatively correlated with leader effectiveness. This same pattern shows for narcissism and LMX. Furthermore, LMX shows a positive correlation with leader effectiveness, which is an indicator that a high quality of LMX is possibly conducive for higher effectiveness ratings. Lastly the table shows that leader effectiveness and LMX are positively correlated, which is an indicator that the constructs measured through these questionnaires might also show a certain amount of overlap.
19 Table 1: Means, Standard Deviations, Correlations
Variables M SD 1 2 3 4 5 6 7 8 9 1. Gender 1.48 0.50 - 2. Age 3.75 1.16 -.07 - 3. Tenure 8.47 6.26 -.00 .50** - 4. Relation 4.40 4.79 -.11 .28** 0,54** - 5. Narcissism 3.46 0.83 -.04 -.13 -.20 -.04 (.89) 6. Admiration 4.29 1.05 -.06 -.15 -.19 -.05 .86** (.89) 7. Rivalry 2.49 1.00 -.13 -.08 -.16 -.01 .85** .45** (.87) 8. Leader effectiveness 5.73 0.99 -.08 .18 .18 .03 -.20* -.08 -.31** (.88) 9. LMX 5.55 0.87 -.03 .15 .14 .09 -.20* -.10 -.29** .72** (.91)
**. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed).
20
4.3 Regression
Hierarchical multiple regression was performed in order to examine whether leader
narcissism predicts perceived leader effectiveness, after controlling for age and tenure. Table 2 shows relevant results of the analysis.
Table 2. Hierarchical regression model of perceived leader effectiveness (1)
R R2 R2 Change B SE β T Step 1 .205 .042 .042 Age .103 .106 .115 .976 Tenure .021 .020 .122 1.041 Step 2 .292 .085 .043* Age .085 .104 .094 .809 Tenure .015 .020 .090 .775 Narcissism -.265 .128 -.213* -2.076
*Measure is significant at the .05 significance level (2-tailed).
In the first step, two predictors were entered: age and tenure. This model was statistically not significant ( F(2,92) = 2.020 P = 0.138 ) and explained 4.2% of the variance. After narcissism is entered in the second step, the complete model explains 8.5% of the total variance, which makes it significant F(3,91) = 2.833, P<.05. The introduction of leader narcissism explains an additional 5.5%, after controlling for gender and age (R2 Change = .043; F (1,91) = 4.312, p <0.05). In the final model, only one predictor is statistically significant, which is leader narcissism (B = -.265, p <.05). Based on these results it is concluded that in line with
hypothesis one, leader narcissism has a negative influence on perceived leader effectiveness. Through a second hierarchical regression, the individual effects of the two different dimensions of narcissism are analyzed. Table 3 shows relevant results of the analysis
21 Table 3. Hierarchical regression model of perceived leader effectiveness (2)
R R2 R2 Change B SE β T Step 1 .205 .042 .042 Age .103 .106 .115 .976 Tenure .021 .020 .122 1.041 Step 2 .207 .043 .001 Age .101 .107 .112 .947 Tenure .020 .020 .118 .993 Admiration -.028 .104 -.029 -.273 Step 3 .412 .170 .127** Age .089 .100 .098 .885 Tenure .016 .019 .095 .851 Admiration .138 .107 .139 1.289 Rivalry -.403 .109 -.398** 3.707
**measure is significant at the 0.001 level (2-tailed)
In the first step, the same two predictors were entered: age and tenure. This model was statistically not significant ( F(2,92) = 2.020 P = 0.138 ) and explained 4.2% of the variance. In the second model narcissistic admiration was entered, and the model is also not significant F(3,91)=1.358, P = .261. This model explains 4.3% of the variance. The introduction of narcissistic admiration explains an additional 1.1% after controlling for age and tenure F(1,91)=.074, P =.786. In the third step, narcissistic rivalry was entered, and this model is statistically significant F(4,90)= 4.596, P<0.05. This model explains 17% of the variance. The introduction or narcissistic rivalry explains an additional 13.3%, F(1,90) = 13,74, P<0.01. In the final model, only one predictor was significant, which was narcissistic rivalry (B=-.398, P<.001). From this it is concluded that the negative effect of leader narcissism on perceived leader effectiveness is specifically caused by one of the two dimension of leader narcissism, namely the narcissistic rivalry dimension.
22
4.4 Mediation
The extension of SPSS called Process (Hayes, 2013) was used to examine whether the effect of narcissism on perceived leader effectiveness is mediated by LMX.
Firstly, LMX is studied as a mediator in the relationship between narcissism and perceived leader effectiveness and later the two dimension are studied separately. See table 4 for the details of the analysis.
Table 4. Analysis of the mediating effect of LMX - Narcissism
Consequent M (LMX) Y (Effectiveness)
Antecedent Coeff. SE p Coeff. SE p
X (NARQ) a -.229 .093 <.05 -.079 .078 .314 M (LMX) --- --- --- .800 .075 <.001 Constant I1 6.344 .330 <.001 1.56 .544 <.01 R2 = .048 R2 = .512 F(1,121) = 6.075, p < 0.05 F (2,120) = 63.652, p<.001
The indirect effect of -.183 means that two people that differ one standard unit in their
narcissism rating, differ -0.183 units in their rating of LMX. A higher score on narcissism has a negative effect on working relationship, which in turn translates to a lower rating of
perceived effectiveness. This indirect effect is statistically significant from zero, as revealed by a 95% BC bootstrap confidence interval that is entirely under zero (-.340 to -.047). From this it is concluded that the negative effect of narcissism on perceived leader effectiveness is partly explained through a low quality of leader-member exchange.
Secondly, the analysis was run again with narcissistic admiration as the independent variable. Table 5 shows the relevant detail of the analysis.
23 Table 5. Analysis of the Mediating effect of LMX - Admiration
Consequent M (LMX) Y (Effectiveness)
Antecedent Coeff. SE p Coeff. SE p
X (ADM) a -.080 .075 .291 -.010 .061 .876
M (LMX) --- --- --- .816 .073 <.001
Constant I1 5.895 .332 <.001 1.246 .508 <.05
R2 =.009 R2 = .512
F(1,121) = 1.125, p =.291 F (2,120) = 62.632, p<.001
For narcissistic admiration an indirect effect of -.065 is found, which is not significant, shown by a bootstrap interval that includes zero (-.197 to .080). This means narcissistic admiration does not have an indirect effect on perceived effectiveness through LMX. Lastly, the analysis was run with narcissistic rivalry as the independent variable. See table 6 for details of the analysis.
Table 6. Analysis of the Mediating effect of LMX - Rivalry
Consequent M (LMX ) Y (Effectiveness )
Antecedent Coeff. SE p Coeff. SE p
X (RIV) a -.249 .076 <.05 -.109 .065 .097
M (LMX) --- --- --- .781 .075 <.001
Constant I1 6.171 .202 <.001 1.672 .494 <.01
R2 =.083 R2 = .522
F(1,121) = 10.874, p < 0.05 F (2,120) = 65.474, p<.001
The indirect effect of -.194 means that two people that differ one standard unit in their narcissistic rivalry rating, differ -0.183 units in their rating of LMX. A higher score on narcissism has a negative effect on working relationship, which in turn translates to a lower rating of perceived effectiveness. This indirect effect is statistically significant from zero, as revealed by a 95% BC bootstrap confidence interval that is entirely under zero (.381 to -.068). From this it can be concluded, that just like narcissism, narcissistic rivalry has a
24 negative effect on perceived leader effectiveness which is partly explained through a low quality of LMX.
4.5 moderation
With the same extension of SPSS it is examined whether the effect of narcissism on
perceived leader effectiveness is influenced by gender, in such a way that female narcissistic leaders are perceived as less effective than male narcissistic leaders. Similar to previous analysis, the analysis is done for narcissism and for its two dimensions separately. Table 7 shows the results of the moderating effect of gender with narcissism as the independent variable.
Table 7. Analysis of the moderating effect of gender - Narcissism
Model Summary
R R-sq F df1 df2 P
.2.38 .0566 2.301 3 115 0.081
R-square increase due to interaction
R2-chng F df1 df2 P
INT_1 .0102 1.248 1 115 .266
Conditional effects of X on Y at values of the moderator(s):
Gender Effect SE T P
Male -.0966 .168 -.574 .567
Female -.342 .141 -2.421 .017
By adding gender as a moderating variable, the model is able to explain an additional 0.12% of perceived leader effectiveness, which is not a significant amount F(1,115) = 1.248, P = 0.266. From this, it can be concluded that gender does not have an influence on the
relationship between leader narcissism and perceived leader effectiveness. In other words, there is no difference in effectiveness ratings between narcissistic men and women. The same analysis is run again for narcissistic admiration. See table 8 for relevant details of this
25 Table 8. Analysis of the moderating effect of gender - Admiration
Model Summary
R R-sq F df1 df2 P
.115 .013 .512 3 115 0.675
R-square increase due to interaction
R2-chng F df1 df2 P
INT_1 .002 .248 1 115 .619
Conditional effects of X on Y at values of the moderator(s):
Gender Effect SE T P
Male -.019 .132 -.147 .883
Female -.107 .117 -.916 .362
Adding narcissistic admiration as a moderating variable, the model is able to explain an extra 0.2% of the variance of perceived leader effectiveness, which is statistically not significant, F (1,115) = .248, P = 0.619. This means that also for people that score high on narcissistic admiration, there is no gender difference in perceived effectiveness.
Lastly, the analysis is run for narcissistic rivalry. Table 9 shows the relevant details of this analysis.
Table 9. Analysis of the moderating effect of gender - Rivalry
Model Summary
R R-sq F df1 df2 P
.323 .104 4.452 3 115 0.0054
R-square increase due to interaction
R2-chng F df1 df2 P
INT_1 .0175 2.244 1 115 .137
Conditional effects of X on Y at values of the moderator(s):
Gender Effect SE T P
Male -.019 .132 -.147 .883
Female -.107 .117 -.916 .362
Adding narcissistic rivalry as a moderating variable, the model is able to explain an extra 0.175% of the variance of perceived leader effectiveness, which is not significant F(1,115) = 2.244, p = .137. From these analyses it is concluded that gender does not have an influence
26 on the relationship between narcissism, or its separate dimension and perceived leader
effectiveness.
4.6 Moderated Mediation
As the final part of the analysis, gender is not only researched as a moderator on the direct effect of narcissism on perceived leader effectiveness but also on the indirect effect between these variables as mediated by LMX. Table 10 shows the relevant details of this analysis, with narcissism as the independent variable.
Table 10. Analysis of moderated mediation - Narcissism
Consequent M (LMX) Y (Effectiveness)
Antecedent Coeff. SE p Coeff. SE p
NARQ a1 -.475 .321 .142 C’1 .536 .578 .040 LMX - - - - B1 .815 .074 <.001 Gender a2 -.534 .691 .441 C’2 1.424 .0550 <.05 Gender X NARQ a3 .167 .1933 .862 C’4 -.381 .154 <.05 Constant I1 7.121 1.149 <.001 I2 -.794 1.055 .454 R2=.048 R2=.542 F(3,115) = 1,921, p =.13 F (4,114) = 33.758, p<.001
Gender Unstandardized Boot effects SE LLCI ULCI
Male -.251 .121 -.498 -.015
Female -.115 .081 -.281 -.041
Results indicate the effect of NARQ on perceived leader effectiveness to be contingent on the employee’s gender, as evidenced by a statistically significant interaction between gender and NARQ in the model of Effectiveness (c’4= -.381, p<.05). A closer inspection of the
27 effect between NARQ and perceived leader effectiveness, appears as significantly only within the group of female leaders (effect=-.227, SE = .099, CL -.424 to -.030) compared to male leaders (effect = -.155, SE=.120, CL -.083 to .392). The indirect effect of NARQ on leader effectiveness via LMX was not found to be contingent on gender.
The same analysis is run with admiration as an independent variable, of which the details are shown in table 11.
Table 11. Analysis of moderated mediation – Admiration
Consequent M (LMX) Y (Effectiveness)
Antecedent Coeff. SE p Coeff. SE p
ADM a1 -.401 .252 .113 C’1 .404 .202 .048 LMX - - - - B1 .834 .074 <.001 Gender a2 -.849 .681 .215 C’2 1.246 .543 <.05 ADM X Gender a3 .213 .154 .168 C’4 -.266 .123 <.05 Constant I1 7.158 1.102 <.001 I2 -.760 .1.022 .458 R2=.025 R2=.534 F(3,115) = .961, p =.414 F (4,114) = 32.588, p<.001
Gender Unstandardized Boot effects SE LLCI ULCI
Male -.157 .101 -.362 .034
Female -.0208 .104 -.132 .288
Results indicate the effect of narcissistic admiration on perceived leader effectiveness to be contingent on the leaders’ gender, as evidenced by a statistically significant interaction between Gender and NARQ in the model of Effectiveness (c’4= -.266, p<.05). However, a closer look indicates that both the individual direct or indirect in itself are not statistically significant, as illustrated by bootstrap levels that include the zero point, for both the direct and indirect effect.
28 Lastly, the same analysis is done with narcissistic rivalry as the independent variable. Relevant details are shown in table 12.
Table 12. Analysis of moderated mediation –narcissistic rivalry
Consequent M (LMX) Y (Effectiveness)
Antecedent Coeff. SE p Coeff. SE p
RIV a1 -.260 .263 .324 C’1 .343 .213 .110 LMX - - - - B1 .785 .075 <.001 Gender a2 -.039 .427 .928 C’2 .778 .346 <.05 RIV X Gender a3 .014 .159 .929 C’4 -.278 .128 <.05 Constant I1 6.192 .721 <.001 I2 .388 .745 .604 R2=.075 R2=.542 F(3,115) = 3,121, p <.05 F (4,114) = 33.713, p<.001
Gender Unstandardized Boot effects SE LLCI ULCI
Male -.1932 .093 -.380 -.0189
Female -.182 .109 -.515 -.0322
Results indicate the effect of narcissistic rivalry on perceived leader effectiveness to be contingent on the leaders’ gender, as evidenced by a statistically significant interaction between Gender and NARQ in the model of Effectiveness (c’4= -.578, p<.05). A closer inspection of the conditional effects indicates only the direct effect to be conditioned by gender. The effect between narcissistic rivalry and perceived leader effectiveness, appears as significantly only within the group of female leaders (effect=.214, SE = .085, CL .382 to -.046) compared to male leaders (effect = .065, SE=.099, CL -.132 to .261). The indirect effect of NARQ on leader effectiveness via LMX was not found to be contingent on gender.
Results from this analysis are in contrast with previously found results that showed no effect of gender on the direct relationship between narcissism and gender. This difference is due to a difference in statistical analyses used. In this analysis the moderating effect of gender
29 is studied for both the direct and indirect effect of narcissistic leadership on perceived leader effectiveness. Because of this, when the moderating effect of gender on the direct effect is tested for, LMX is used as a control variable. Therefore, this analysis shows that when LMX is taken as a control variable, gender does indeed have a moderating effect. This means that the effect of gender on effectivity only exists when the influence of the relationship between leader and member (LMX) is kept at the same level.
As previously argued, LMX and perceived effectiveness show overlap in their measurement, indicating that the factors that make up these constructs possibly also partly overlap. When a control variable is added in an analysis, the variance of the independent variable that is explained by the control variable is take out of the equation. In this case, the variance of effectiveness explained by LMX was taken out of the equation when looking at the moderating effect of gender. The part of variance in effectiveness rating that remains, seems to be sensitive for violations of the traditional gender roles, with women being perceived as less effective than men when scoring high on narcissistic rivalry. The question remains what exactly is this ‘part’ of perceived effectiveness that remains when you take out the overlap with LMX.
Effectiveness is thought to include the relationship between leader and member as well, but to be mainly dependent on evaluations of credibility and trustworthiness
(Campbell,1991; Harris & Hogan, 1992; Lombardo, Ruderman and McCauley,1988). Found results indicate that a gender difference does exist in the way narcissistic men and women are perceived and that this effect is not due to the leader member relationship, or LMX (the more affective component), but another part of effectiveness. Whether this part indeed includes evaluations of credibility and trustworthiness or even other factors, needs to be researched further. The fact that the moderating effect of gender is only found when LMX is controlled for, also is an indicator that LMX can have a possible buffering effect on this relationship.
30 This means that a good relationship can possibly diminish the found gender difference in evaluations of effectiveness.
4.7 Exploratory Analysis
In order to explore this possibility further, a three-way interaction analysis between narcissism, gender and LMX is done in Process. Again narcissism is studied as an independent variable, followed by an analysis of both its separate dimensions. Table 13 shows the details relevant details of the analysis.
Table 13. Three-way interaction - Narcissism
Effectiveness (Y) Antecedent Coeff SE P NARQ 1.391 1.608 .389 Gender 4.369 4.039 .282 NARQ X Gender -1.816 1.067 .092 LMX .854 1.102 .439 NARQ X LMX -.180 .292 .539 Gender X LMX -.584 .722 .421 NARQ X Gender X LMX .276 .193 .156 Constant -.546 6.151 .930 R2 = .598 F(7,111)=23.590, p<0.001
The three-way interaction explains an extra 0,74% of the total variance of perceived effectiveness, which is not statistically significant (P = .156). The analysis is done again with narcissistic admiration as the independent variable. Details of this analysis are shown in table 14.
31 Table 14. Three-way interaction - Admiration
Effectiveness (Y) Antecedent Coeff SE P ADM -1.857 1.233 .135 Gender -6.278 30335 .063 ADM X Gender .842 .739 .257 LMX -1.609 1.007 .113 ADM X LMX .395 .220 .076 Gender X LMX 1.327 .594 .028 ADM X Gender X LMX -.194 .133 .146 Constant 13.132 5.689 .0.023 R2 = .585 F(7,111)=22.366, p<0.001
This three-way interaction is able to explain an additional 0,8% of the variance, which is also not statistically significant (p =.146).
Lastly, the analysis is run with narcissistic rivalry as the independent variable. Details of this analysis are shown in table 15.
Table 15. Three-way interaction - Rivalry
Effectiveness (Y) Antecedent Coeff SE P RIV 3.058 1.196 <.05 Gender 5.648 2.432 <.05 RIV X Gender -2.538 .756 <.05 LMX 1.788 .667 <.01 RIV X LMX -.523 .218 <.05 Gender X LMX -.913 .430 <.05 RIV X Gender X LMX .430 .137 <.05 Constant -4.854 3.748 .198 R2 = .615 F(7,111)=25.324, p<0.001
In this analysis, the three-way interaction does explain a significant amount of variance, namely 3,41% (p<.01). A graph is plotted in order to visualize the three-way interaction, as shown in graph 1.
32 Graph 1. Three-way interaction between Rivalry, LMX and Gender
The graph shows, that only on lower levels of LMX, females are rated as significantly less effective than men when they score high on narcissistic rivalry. On high level of LMX, this effect does not occur.
5. Discussion
This study analyzed the relationship between narcissistic leadership and perceived leader effectiveness. For this study the NARQ scale was used as a measure of narcissism, which allowed for the distinction between two dimensions of narcissism, namely narcissistic
admiration and narcissistic rivalry. It was found that narcissism in leaders does indeed have a negative effective on their effectiveness as perceived by the subordinate. When looking at the two dimensions separately, the same effect is found for the rivalry dimension of narcissism
3 3.5 4 4.5 5 5.5 6 6.5 7
Low RIV High RIV
Per ce iv e d E ff e ctiv e n e ss Female, High LMX Female, Low LMX Male, High LMX Male, Low LMX
33 associated with self-defense in order to persevere a grandiose self. The same effect was nonexistent for the admiration dimension, which is associated with self-promotion.
Additionally, LMX was found to partially mediate the effect of both narcissism and narcissistic rivalry on perceived leader effectiveness. This implies that narcissistic leaders are perceived as less effective partly due to a more negative working relationship with their subordinates. Furthermore, gender was analyzed as a moderator in the relationship between narcissistic leadership and perceived leader effectiveness. Based on the social role theory and previous research (Cejka & Eagly, 1999, Bosak, Sczensy & Eagly, 2011, De Hoogh, Den Hartog & Nevicka, 2015) it was theorized that narcissistic female leaders would be rated as less effective than male narcissistic leaders due to their violation of the traditional female gender role. However, results did not confirm this.
Likewise, gender was examined as a moderator not only on the direct effect between narcissism and effectiveness but also on the indirect effect of these variables as mediated by LMX. From this analysis a moderating effect was found on the direct relationship between both narcissism and the rivalry dimension on perceived effectiveness. This is a striking result as previous analysis did not find this moderating effect. However, when both the direct and indirect effect are analyzed, LMX is used as a control variable when testing for the direct effect, which led to a significant moderation effect. This means that the moderating effect of gender on the relationship between narcissistic leadership and leader effectiveness only exists when the influence of LMX is kept at the same level.
This inspired an exploratory three-way interaction analysis towards the possible effect of LMX on the moderating effect of gender. From this it was found that there is indeed a moderating effect of gender on the relationship between narcissistic rivalry and perceived leader effectiveness. When women score high on narcissistic rivalry, they are perceived as less effective than men, but only when the quality of LMX is low. No gender difference is
34 found on high levels of LMX. This means that women, more than men benefit from a positive leader-member relation, specifically when they score high on narcissistic rivalry.
5.1 Implications
Current results are another step towards untangling the effects of narcissistic leadership on leader effectiveness. That previous research finds both positive and negative effects, seems to be at least partly rooted in the different conceptualizations of narcissism that are used in these studies. As noted by Kansi (2003), research into narcissism suffers from a lack of conceptual clarity. Studies done with the NARQ scale for narcissism can help clarify some
incongruences. Researchers talking about the positive effects, possibly mainly focus on the confidence and dominance associated with self-enhancement (the admiration dimension), whereas the negative effects found are associated more with the rivalry dimension; aggression as self-defense. This study is an indication that the different effects found for narcissism originate in different dimensions of narcissism. The negative effects seem to stems specifically from the rivalry dimension of narcissism. Therefore, when discussing the effects that narcissistic leaders have, one should be careful to specify on which dimension this specific person scores highest. When two people both score high on narcissism, but one scores particularly high on the admiration dimension, and the other one on the rivalry dimension, their effects on their subordinates can be completely different. Further research should therefore focus on creating a unified idea of narcissism and its dimension, which will consequently allow for a better classification of its effects.
Furthermore, an interesting interplay between narcissism, gender, LMX and their effect on leader effectiveness has been revealed. First it was found that gender did not moderate the effect between narcissism and perceived leader effectiveness. Later however, exploratory analysis showed that women are indeed rated as less effective than men when
35 they score high on narcissism, but only on the rivalry dimension and only when quality of LMX is low. This result indicates that the quality of LMX is an important asset for
narcissistic women, as it can possible be a buffer for the effects of the found gender difference. This effect can be explained by the contact hypothesis, which states that
stereotyping can be reduced when people have more interpersonal contact (Turner & Crisp, 2010). Personal contact is thought to make people focus less on the general beliefs they hold about a certain group a person belongs to (in this case gender) but focus more on the
attributes specific to that person. Consequently, a better relationship between leader and follower makes people less sensitive for role incongruity. In general, higher quality of LMX is therefore likely to reduce the influence that gender has on perceived leader effectiveness of narcissistic leaders.
Lastly, the moderated mediation analysis reveals a conceptual issue between perceived leader effectiveness and LMX. As stated before, both constructs show a certain overlap. The moderated mediation analysis shows that only when LMX is controlled for, a moderation effect of gender on the relationship between narcissistic rivalry and perceived leader effectiveness occurs. When a control variable is added in a statistical analysis, the variance of the dependent variable explained by control variable is taken out of the equation. In other words, if you take out the part of overlap between LMX and effectiveness, what remains is the part of effectiveness that gender has a moderating effect on. However, what exactly does remain? As stated before LMX is an evaluation based on a more affective component, whereas effectiveness evaluations depend more on capability (credibility and trustworthiness). This is an indicator that the when females are highly narcissistic, they are not rated as less likable than men, but rather as less capable then men. This is an important direction for future research and could help to extend the social role theory. It is known that the violation of traditional gender roles does have a negative effect on people’s evaluations of
36 this person. Future research however, should focus on establishing what factors make up this evaluation, and which of those are affected by violations of gender roles.
5.2 Limitations & Directions for Future Research
In this study, certain methodological flaws exist and it is therefore recommended that future research focusses itself towards both cancelling out these flaws associated with narcissism research and extending crucial theory.
Firstly, in this study the organizational sector the dyad works in is not controlled for. However, it is shown that different maladaptive leadership styles tend to be more prevalent is some sectors than others, for example in the financial sector (Boddy, 2010). It is possible that when this research was solely focused on the financial sector, higher averages of narcissism would have been found, leading to a different effect. Therefore, studies should be done that look at the different sectors, to see where research on narcissism is specifically relevant.
Secondly, the fact that the measures are self-reported could lead to several biases. In this case, the questionnaire for leader narcissism is answered by the leaders themselves. As narcissism can be viewed as a sensitive subject, it is possible that a social desirability bias does exist. This means that leaders possible fill in answers in a manner that will be viewed positively by others, instead of filling in an accurate evaluation of themselves. In relation to this, it is found people that score higher on narcissism, have a bigger tendency to
self-enhance on self-report measures than other people do (John & Robins, 1994). Future research could be done in an experimental form, where participants rate fictional described managers that differ in level of narcissism and gender on their effectiveness.
Lastly, the before mentioned incongruence between the several interpretations of narcissism makes it difficult to merge different study results. An agreement should be
37 different dimensions. Consequently, this will allow for a clearer structuration of the different effects of narcissism in leaders.
Current results raise the question of what factors make up an evaluation of leader effectiveness and where this evaluation differs from evaluations of LMX. Related to this, in needs to be studied on which evaluations the violation of one’s traditional gender role has a negative effect. The factors that both evaluations are based on, seem not to be sensitive for violations of the traditional gender role, whereas there are other factors that are specifically responsible for evaluations of effectiveness, on which these violations do have an effect. There is clues that the overlap of evaluations of effectiveness and LMX entails a more affective component of the relationship between leader and follower, whereas the part that differentiates effectiveness form LMX comes from evaluations of the capability of the leader (Campbell, 1991; Harris & Hogan, 1992; Lombardo, Ruderman and McCauley, 1988; Liden, 1985; Day & Cram, 1992). This would mean that the violation of gender roles mainly relates to measures of capability and not so much of likeability. More research is needed however to confirm these ideas. This could be an important extension of our knowledge of gender inequalities and social role theory.
6. Conclusion
This study aimed to add understanding towards the role of gender and LMX in the
relationship between narcissism and leader effectiveness. Results show that leader narcissism has a negative effect on perceived leader effectiveness and that specifically the narcissistic rivalry dimension is responsible for this effect, rather than the narcissistic admiration dimension. Furthermore, it is concluded that gender has a moderating effect on the relationship between narcissistic rivalry and perceived leader effectiveness, with women being rated as less effective than men, but only when the quality of LMX is low. With high
38 quality of LMX this gender difference does not occur, indicating that a good working
relationships can be an important buffer for narcissistic female leaders.
Current study also raised the question towards what factors exactly determine how followers rate the effectiveness of their leaders. The fact that a gender difference of perceived effectiveness in people that score high on narcissistic rivalry was only found when LMX was controlled for, indicates that LMX is partly responsible for effectiveness ratings. However, there are other factors that influence evaluations of effectiveness which do not overlap with evaluations of LMX. It is his part that is sensitive for violations of traditional gender roles. Further research needs to focus on examining what factors exactly make up these evaluations that are sensitive for violations of gender roles and hereby extending a crucial part of social role theory. Despite its limitations, it can be said that this study gives valuable insights in the way different aspects of narcissism influence perceived leader effectiveness.
