Gender and Leadership in Public Organizational Contexts: A conjoint analysis of employee preferences for public managers in male- and female-dominated public sectors in the Netherlands

(1)

Master Thesis Public Administration

Track: Economics & Governance

Gender and Leadership in Public

Organizational Contexts

A conjoint analysis of employee preferences for public

managers in male- and female-dominated public sectors

in the Netherlands

Name: Sophie Offringa

Student ID: S2300826

Student number: S2300826

Date: 11-06-2019

Advisor: Prof. dr. S.M. Groeneveld

Second reader: Prof. dr. O.P. van Vliet

(2)

2

Foreword

I would like to thank my thesis supervisor Prof. dr. Sandra Groeneveld of Leiden University. From the start, you sparked my enthusiasm and interest in the subject of gender and leadership, and we had many interesting and analytical conversations about the subject. I would like to thank you for giving me the opportunity to participate in the research project, as well as for the guidance you provided along the way. Since our first thesis meeting in December 2018, I have learned a lot on both the subject of gender and leadership and on conjoint analysis, which was an entirely new statistical technique to me at that time. I would also like to thank my second reader, Prof. dr. Olaf van Vliet.

Finally, I would like to thank Daan Hiemstra for showing me the way in the world of programming and for your helpful comments on various parts of the thesis.

(3)

3

Abstract

In this thesis, I study the relationship between gender, gender-stereotypical traits, and transformational and transactional leadership by looking at employee preferences for public managers. Role incongruity theory of prejudice towards female leaders predicts that employees prefer male, agentic, and transformational managers. Besides, the effects are expected to be stronger in male-dominated contexts compared to female-dominated contexts. Using a conjoint survey experiment, I examined the effects of various manager attributes on employee preferences for public managers in male- and female-dominated public sectors in the Netherlands. Findings from 2,757 respondents working in the education, police, and defense sectors show that employees are indifferent between a male or female manager and that they have a clear preference for communal managers, both in male- and female-dominated contexts. Interaction effects between a manager’s gender and gender-stereotypical traits or leadership attributes are often small. However, in the male-dominated context, male, communal managers are significantly preferred over female, communal managers. On the other hand, female managers who use verbal rewards get a significantly larger boost than male managers who use verbal rewards. All in all, this study does not find strong support for role congruity theory.

(4)

4

Introduction

First mentioned by Schein (1973), “think manager – think male” refers to the association people tend to have when it comes to leadership. Forty-six years later, the phrase still appears to be accurate. A recent meta-analysis on gender and leadership shows that, despite the decrease of the gender gap in leadership emergence between 1955 and 2017, men still emergence in leadership positions more often than women. The same study reports that in the US, 39% of managers in 2014 were women, compared to 32% in 1983 (Badura et al., 2018). In the EU, only 27% of senior business roles is held by women, whereas currently, more women than men graduate from universities across the EU (Catalyst, 2019). Moreover, the share of female managers in the Netherlands (27%) one of the lowest in the EU (SCP&CBS, 2018, p.104). While today, women make up 47% of the workforce in the Netherlands as opposed to 30% in 1973 (CBS, 2019; CBS, 2010, p. 44). The question on the origin of the skewed distribution of men and women in leadership positions is frequently asked, and the figures show that it remains relevant. Since the 1970s, there has been a vast amount of research on gender and leadership, with an increasing academic and public interest in recent years (Eagly & Heilman, 2016).1

Nevertheless, the focus of this thesis is not to provide an answer to the question of the relatively low amount of women in leadership positions. Rather, I focus on the psychological and sociological mechanisms behind gender and leadership in public organizational contexts.

One of the most influential theories regarding gender and leadership is role congruity theory (Eagly & Karau, 2002; Eagly, 1987). In short, the theory states that the female gender role is incongruent with the leader role, which creates prejudice towards female leaders.2_This

prejudice originates from incongruity between the predominantly communal traits that people tend to relate to women, such as emotionally expressive, caring, and understanding, and the predominantly agentic traits that are generally associated with men, and successful leaders, such as assertive, dominant and self-confident.3_{(Eagly & Karau, 2002; Eagly, Gartzia, & Carli,}

2014; Eagly & Carli, 2003; Burgess and Borgida, 1999; Heilman, 2001). Others add that people relate men with a higher status and competence in social hierarchy and leadership than women,

1_{For recent overviews of the literature on gender and leadership, see for example Eagly et al., 2014; Koenig et al.,}

2011; Eagly & Heilman, 2016; Badura, 2018; Derue et al., 2011.

2_{I use the terms leader and manager interchangeably. Even though there are differences between the two, in this}

research the leader and managerial role are not separated (see also Koenig et al., 2011, p.616). For the research question and the hypotheses, I specifically use manager, because the survey focused on potential direct managers.

3_{I refer to agentic and communal traits combined as gender-stereotypical traits, because they are not fixed traits}

(6)

6

which also creates barriers for women in leadership positions (Ridgeway, 2001). In short, the incongruence between the gender-stereotypical traits people associate with women and what constitutes as a successful leader results in a reduced evaluation of women as an actual or potential holder of leadership positions (Eagly & Karau, 2002). However, since the increasing focus on transformational leadership – because of its positive effect on organizational effectiveness (e.g. Ng, 2017; Fletcher, 2004) – scholars have discussed a possible female leadership advantage. Because the qualities needed for transformational leadership are often related to communal traits, the incongruence between the female gender role and the leader role may diminish. Nevertheless, this female leadership advantage should not be taken for granted as long as the stereotypical image of a successful leader is culturally defined in masculine terms (Koenig, Eagly, Mitchell, & Ristikari, 2011; Eagly et al., 2014). In addition, the degree of role incongruity depends on varies other factors, such as the status of the leadership position, the organizational context, or personal characteristics of the perceiver of leadership – such as sex, age, nationality, or experience with current or previous managers (Koenig, Eagly, Mitchell, & Ristikari, 2011; Eagly et al., 2003; Duehr & Bono, 2006). Specifically, in some organizational contexts the incongruence between the female gender role and the leader role may be lower than in others, because of variation in beliefs on gender roles and leadership (Eagly & Karau, 2002; Koenig et al., 2011).

However, there is little experimental evidence on role congruity in the actual organizational context (Koenig et al., 2011; Eagly & Karau, 2002). Besides, the links between various leader characteristics, such as gender, agentic or communal traits, or leadership style, remain unclear (see for example Derue et al., 2011). How do employees in different public organizational contexts perceive leadership? What kind of managers do these employees prefer, and which manager characteristics matter most? By examining the influence of certain manager characteristics on employee preference, I contribute to the understanding of the mechanisms behind stereotyping and prejudice towards female leaders in a public sector context. Because of role incongruence, employees may have different preferences for gender-stereotypical traits and leadership styles depending on a manager’s gender. Therefore, in this thesis I answer the following research question: “How do gender and gender-stereotypical traits of public managers affect employee preferences for managers in male- and female-dominated public sectors?”. To answer this question, I replicated a conjoint survey experiment – which is part of a Danish research project4_{– to be able to disentangle the various manager}

4_{The results of the Danish research project are due later this year. I used the conjoint design for my own research}

(7)

7

characteristics/gender, gender-stereotypical traits, and leadership behavior. The final sample consists of 2,757 Dutch public sector employees in education, police and defense, which makes it possible to compare perceived role incongruence between male- and female-dominated contexts. I analyze the conjoint experiment using the approach developed by Hainmueller, Hopkins, & Yamamoto (2014) to examine how much each of the manager characteristics, or an interaction between them, explains in terms of manager preference of public sector employees.

With this research, I contribute to the literature in the area of Public Administration in the following ways. First, by conducting the conjoint survey experiment among public sector employees, manager preferences are measured in their actual organizational context. Survey experiments are currently widely used in public management and administration research, but they mostly involve randomly selected citizens participating in online panels. Jilke and Van Ryzin (2017) therefore recommend focusing more on respondents in the actual context, such as public managers and employees, when doing survey experiments. Moreover, because the survey is held among employees in male- and female-dominated contexts, I can examine the possible variation in the contextual definition of leadership. Second, there is little research on preferences for certain types of managers, especially not which disentangles a manager’s gender, gender-stereotypical traits, and leadership styles (Stoker, Van der Velde, & Lammers, 2012). Third, I adopt a relatively unknown research method in the field of public administration. Conjoint analysis has been introduced in the field, but it has not been used very often yet (see Jilke & Van Ryzin, 2017). Specifically, I introduce the statistical approach of Hainmueller et al. (2014), which is starting to gain interest in political science, but has to my knowledge not yet been used in research on gender and leadership.

Furthermore, the research is socially relevant because by providing insight into employee preferences for managers, I examine prejudice towards female managers in the Dutch public sector. Besides, this research gains knowledge on variation in prejudice and stereotyping across different organizational settings. A good understanding of contextual and individual influences on gender prejudice may stimulate interventions to prevent gender discrimination when selecting people as managers or evaluating them (Eagly & Karau, 2002, p. 589). When people become aware of stereotypes and prejudices, it is possible to alleviate the perceived incongruity between women and leadership and foster the rise of more women in leadership positions.

The outline of my thesis is as follows. In the next chapter, I provide a theoretical framework in which I elaborate more extensively on role congruity theory, leadership and the influence of a

(8)

8

male- and female-dominated context. In the third chapter, I describe the data collection strategy and research design used to measure and analyze the conjoint survey experiment. Subsequently, in the fourth chapter, I present the results of the conjoint analysis, as well as additional analyses and robustness checks. In the last chapter, I end with a conclusion and discussion of the results.

(9)

9

Theoretical Framework

In this chapter, I review the relevant academic literature on gender and leadership that may help explain employee preferences for managers and the potential factors influencing it. By formulating hypotheses, I give an overview of the theoretical explanations that together provide an answer to the research question “How do gender and gender-stereotypical traits of public managers affect employee preferences for managers in male- and female-dominated public sectors?”. The chapter starts with an explanation of role congruity theory, the central theory used for this research. After that, I turn to leadership theory by elaborating on transformational and transactional leadership. Subsequently, I describe the relation between these leadership styles and gender-stereotypical traits. In the final section, I elaborate on the influence of the organizational context on employee preferences for managers.

Role congruity theory

An important part of the main research question is how gender and gender-stereotypical traits affect employee preferences for leaders. People evaluate leaders and potential leaders according to the prejudiced and instinctive ideas they have about them (Hoyt & Murphy, 2016). This is reflected by gender-stereotypical traits and behaviors, as well as the greater status associated with men (Eagly & Karau, 2002; Ridgeway, 2001). The ideas people have about leaders can lead to prejudices and biases against women, who do not fit the common perception of a leader (Hoyt & Murphy, 2016). An influential explanation of this phenomenon can be based on the role congruity theory of prejudice towards female leaders of Eagly & Karau (2002).

Role congruity theory developed from social role theory of sex differences and similarities, which states that differences in women’s and men’s behavior originates in the distribution of men and women into different social roles (Eagly, Wood, & Diekman, 2000). Women and men generally hold roles with typical characteristics, which follow from the traditional social role division of men as breadwinner, associated with higher status roles, and women as homemaker, associated with lower status roles (Eagly et al., 2000; Eagly & Wood, 2012). Here, it is important to make a distinction between the terms sex and gender. The term sex refers to the division of individuals into the categories of female and male, based on their reproductive functions (Wood & Eagly, 2015). Where sex is about biological aspects of men and women, gender includes the psychological aspects ascribed to these sexes based on cultural beliefs. The term gender therefore refers to the interpretation of the female and male categories

(10)

10

by societies and individuals (Eagly, 1987; Wood & Eagly, 2010; Bem, 1993; Abele, 2003; Eagly et al., 2014).5_{A key proposition of social role theory is the allocation of stereotypic}

gender beliefs into communal and agentic traits (Eagly, 1987; Eagly & Karau, 2002). A large part of the literature focuses on the agency-communion paradigm, and research indeed indicates that most of the stereotypic gender beliefs fall in these two categories (Eagly & Wood, 2012; Eagly et al., 2000; Broverman, Vogel, Broverman, Clarkson, & Rosenkrantz, 1972). Agentic and communal traits, which originate from Bakan (1966), are related to gender, such that agentic traits are more strongly ascribed to men, whereas communal traits are more strongly ascribed to women. Communal traits primarily refer to a tendency to be concerned about the welfare of others. Typical examples of communal traits are caring, understanding, friendly, helpful, sympathetic, emotionally expressive, affectionate, and interpersonally sensitive. Agentic traits refer to a tendency to be controlling, assertive and confident. Typical examples of agentic traits are self-confident, ambitious, dominant, assertive, competitive, independent, forceful, and inclined to act as a leader (Eagly & Karau, 2002; Eagly & Wood, 2012; Eagly & Johannesen-Schmidt, 2001; Eagly et al., 2000; Eagly, 1987).

Role congruity theory makes use of the construct of gender role to depict the common beliefs people hold on the traits of women and men (Eagly & Karau, 2002). This includes both descriptive norms or stereotypes, which refer to how women and men are, and prescriptive norms or stereotypes, referring to how women and men should be. Thus, gender roles prescribe the appropriateness or inappropriateness of traits and behavior of women and men. The core of role congruity theory is that the female gender role is incongruent with the leader role, which causes two forms of prejudice against female leaders. The first type of prejudice – related to descriptive gender norms – arises because people perceive women as less agentic and more communal than men, which results in being evaluated as less qualified for a leadership position. The second type of prejudice arises because agentic behavior of a woman conflicts with people’s perception of how women ought to behave, related to prescriptive gender norms (Eagly & Karau, 2002). This means that people expect women to behave communal, whereas leaders are expected to behave agentic.

In essence, the two types of prejudice originate from the incongruity between the predominantly communal traits that people tend to relate to women, and the predominantly agentic traits that are generally associated with successful leaders (Eagly & Karau, 2002). This results in a better fit between the male gender role and the leader role, because people may think

5_{These terms do not insinuate any further causes that may determine effects of sex and gender, as it is beyond the}

(11)

11

that men have or should have more agentic traits.6_{Thus, female leaders with evident communal}

characteristics are criticized because this role is incongruent with the leader role, whereas female leaders who behave highly agentic encounter a backlash effect because this leader role is incongruent with the female gender role (Hoyt & Murphy, 2015; Eagly et al., 2014; Heilman, 2001).

The masculinity of the leader role has been strongly established in previous research, as outlined in the meta-analysis of Koenig et al. (2011). They examined studies from three research paradigms on gender and leadership: the think manager – think male paradigm (Schein, 1973), the agency-communion paradigm (Powel and Butterfield, 1979), and the masculinity-femininity paradigm (Shinar, 1975). They find robust effects for the perceived masculinity of leadership in all three paradigms, over time and across countries. There is evidence that there is a better fit for men and the cultural beliefs of leadership than women, and that men therefore face fewer challenges to gain leadership positions and become successful in them (Koenig et al., 2011). They did find a slight decrease in the cultural masculinity of leadership stereotypes and conclude that currently, leadership seems to include “more feminine relational qualities, such as sensitivity, warmth, and understanding” (Koenig et al., 2011, p. 634).

Next to role incongruence or a lack of fit, status may also play a role in prejudice against female leaders. Ridgeway (2001) adds status to role incongruence by stating that the female gender role in general in society has a lower status compared to the male gender role, as argued in expectation states theory. The explanation of how female leaders behave and are evaluated resembles the role congruity theory of Eagly & Karau (2002). However, status beliefs in this context refer to the widely shared cultural beliefs about the position of status in society, and they associate “greater social significance and general competence, as well as positive and negative skills” (Ridgeway, 2001, p. 638) with one group in society over another. These status beliefs are included in gender stereotypes, which causes people to relate men with a higher status dignity and competence in social hierarchy and leadership than women. This in turn creates double standards for women, because women have to perform better to prove they are just as good, i.e. to be perceived as competent as the higher status group men (Ridgeway, 2001; Foschi, 2000; Lyness & Heilman, 2006). Because of the perceived masculinity of leadership, descriptive gender norms give rise to this double standard that might hamper women’s rise into leadership positions. Moreover, prescriptive gender norms cause a double bind that makes it more difficult for women to adequately execute a leadership position (Eagly et al., 2014).

6_{Heilman (2001) started in 1983 to refer to this as a perceived “lack of fit” for female leaders. The idea is the same}

(12)

12

In summary, the perceived incongruity between the female gender role and the social role of leaders often leads to prejudice towards female leaders. In addition, people evaluate their potential leaders in reference to implicit leadership theories, which are “intuitive and preconceived notions of what it means to be a leader” (Hoyt & Murphy, 2016, p. 388). They reflect personality traits, behaviors, and social identities, which may be incongruent with the image of a female leader (idem, 2016). This, in turn, can subsequently influence employee preferences for managers, as they will prefer a manager that fits their image of a manager. Thus, despite the increasing importance of feminine relational qualities in leadership, the masculinity of leadership is still highly present (Koenig et al., 2011). In addition, the combination of differences in status of men and women in society and the higher status associated with leadership (Ridgeway, 2001; Eagly et al., 2000; Eagly & Wood, 2012), may still lead to role incongruence for female leaders. To examine the effect of gender and gender-stereotypical traits on employee preferences, as well as the interaction between them, I therefore arrive the following hypotheses:

Hypothesis 1a: Employees prefer male managers over female managers Hypothesis 1b: Employees prefer agentic managers over communal managers

Hypothesis 2: Employees prefer male, agentic managers over female or communal managers

Transformational and transactional leadership

In combination with role congruity theory of prejudice towards female leaders, researchers also mentioned the potential of a female leadership advantage. Not because of changes in prejudice about women’s competence as leaders, but because of advantages in certain leadership styles (Eagly & Carli, 2003). Leadership is widely studied in social science, however a universal agreement on its definition remains absent. Therefore, Antonakis & Day (2018) broadly defined leadership as “a formal or informal contextually rooted and goal-influencing process that occurs between a leader and a follower, groups of followers, or institutions. The science of leadership is the systematic study of this process and its outcomes, as well as how this process depends on the leader’s traits and behaviors, observer inferences about the leader’s characteristics, and observer attributions made regarding the outcomes of the entity led.” (p. 5). Besides, leadership style is defined as “relatively stable patterns of behavior displayed by leaders” (Eagly et al., 2003, p. 569). Specifically, transformational and transactional leadership styles are used to examine employee preferences for managers in the public sector. These two goal-oriented

(13)

13

leadership styles are studied a lot in both the generic leadership literature and public administration research (Jensen et al., 2019; Judge & Piccolo, 2004; Vogel & Masal, 2015) and they are often regarded as the most relevant for the public sector (Bellé & Cantarelli, 2018, p.196; Jensen et al., 2019; Van Wart, 2013; Oberfield, 2012). Transformational leadership is often linked to increasing organizational productivity (Ng, 2017), but studies have also suggested that there is a positive relationship between goal-oriented styles in general and employee performance (Avolio, Reichard, Hannah, Walumbwa, & Chan, 2009; Wright, Moynihan, & Pandey, 2012).

According to Bass (1985), who further elaborated on transactional and transformational leadership as firstly coined by Burns (1978), both styles are used to get employees to work towards achieving the organizational goals and objectives. They are the active components of the full-range leadership theory, as opposed to the passive laissez-faire style – an overall failure to display active leadership behavior (Eagly et al., 2003; Jensen et al., 2019; Avolio, 1999). Originally, transformational leadership has been conceptualized into four components that shape employee behavior, i.e. idealized influence, inspirational motivation, individualized consideration, and intellectual stimulation. Transactional leadership covers three components, i.e. contingent reward, active exception, and passive management-by-exception (Avolio, Bass, & Jung, 1999; Bass, Avolio, Jung, & Berson, 2003). However, Van Knippenberg and Sitkin (2013) recently criticized the conventional multi-dimensional concept of transformational leadership and recommend focusing on its core, which is visionary leadership. In line with this criticism, Jensen et al. (2019) argued that existing research in public administration on leadership suffers from indistinct theoretical foundations, the confounding of leadership and its effects, and the unsuitability for public organizations. Therefore, I adopt the reconceptualization of Jensen et al. (2019) of the two leadership styles, which they established to foster a consistent and proper usage of the concepts in public administration research. They argue that the focus on the behavior of leadership ensures that the styles are not confounded with their effects.

In line with Jensen et al. (2019), transactional leadership refers to the use of transactions, or incentive structures, in order to align the self-interest of employees with the organizational goals. This leadership style is based on the use of contingent rewards and sanctions, of pecuniary and nonpecuniary nature (Antonakis, Avolio, & Sivasubramaniam, 2003; Jensen et al., 2019). Transactional leaders reward employees when they meet the objectives set and correct employees if they fail to meet them (Eagly et al., 2003). The contingency of the rewards and sanctions is important here, which means that the transactions

(14)

14

between a leader and employee should be directly related to individual achievements of the employee (Jensen et al., 2019). It is important to make a distinction between pecuniary rewards – such as a bonus on top of one’s salary, perks and job training – and non-pecuniary, verbal rewards – such as giving compliments for work well done. Besides, sanctions could be given in the form of penalties for errors or bad performance (Jensen et al., 2019). In short, transactional leaders use contingent rewards and sanctions to make the employees work towards the organizational goals.

Transformational leadership refers to the creation of a shared vision among employees in order to achieve the organizational goals. This is done by attempting to transform and motivate employees by managing and stimulating individual efforts, with the aim to activate their higher order intrinsic needs (Jensen et al., 2019; Judge & Piccolo, 2004). Transformational leaders encourage employees to fully develop their capabilities, so that they will contribute to the organization more effectively (Eagly et al., 2003) and “transcend their own self-interest for the sake of the organization” (Jensen et al., 2019, p. 8). According to Jensen et al. (2019), three types of behavior are relevant here. First of all, a transformational leader attempts to clarify a vision for the organization that appeals to and is desirable by the employees. Second, it is important for a transformational leader to share the vision among the employees who have to execute it, by clearly communicating the link between the actions and goals of the vision. The last element that is relevant, is to maintain this established shared vision in the short and in the long run. Thus, transformational leaders seek to create, share, and maintain a vision for the organization so that employees are encouraged to share and pursue the organizational goals (Jensen et al., 2019). This focus of transformational leadership is also called visionary leadership.

To sum up, transformational and transactional leadership both aim at achieving organizational goals. The key difference between these conceptualizations is that transformational leadership aims at encouraging employees to go beyond their self-interest by creating a shared vision among employees, whereas transactional leadership aims at establishing self-interest of employees towards achieving the organizational goals (Jensen et al., 2019).

Gender-stereotypical traits and leadership styles

The spillover and internalization of gender-specific norms, which stems from the dynamics of role incongruity and the influence of gender roles on behavior, causes women and men to differ

(15)

15

in their leadership styles (Eagly et al., 2003, p. 573). These differences can be substantial, because they may, among other things, influence people’s views about female leadership (Eagly & Johannesen-Schmidt, 2001). Agentic and communal traits are not only pertinent aspects of gender roles that can help understand leadership; they also correspond to the personal traits of transformational and transactional leadership (Eagly et al., 2003). In previous research, transformational leadership has been associated with female, communal leaders, whereas transactional leadership has been associated with male and agentic leaders (e.g Eagly & Johannesen-Schmidt, 2001; Eagly et al., 2003; Koenig et al., 2011). Increasingly, transformational leadership is argued to become more feminine, because of the relation between communal aspects and the individual consideration for employees to create a shared vision (e.g. Fletcher, 2004; Kark, Waismel-Manor, & Shamir, 2012; Eagly et al., 2014). Besides, studies have shown higher congruence between female leaders’ communal attributes and their transformational style, than their transactional styles (Eagly & Johannesen-Schmidt, 2001). Moreover, in a meta-analysis of Eagly et al. (2003), replicated by Antonakis et al. (2003), women used more transformational styles than men, together with the reward component of transactional leadership. In a research by Vinkenburg et al. (2011), participants correctly related the leadership style of women with transformational and that of men with transactional. Women may therefore prefer to use a transformational leadership style, because in this way they can conform to both their gender role and their leader role (Eagly et al., 2003). Besides, the use of transformational leadership could help to reduce the overly masculine image when women use agentic leader behavior that leads to role incongruence (Yoder, 2001). In combination with the effectiveness of transformational leadership, women may have a small advantage in leadership style that offers opportunities to increase organizational effectiveness (Eagly et al., 2014). However, women are still held to higher standards to obtain these leader roles at all. Besides, higher status beliefs associated with men presumably lead to stronger preferences for male leaders in general. Before hypothesizing a female leadership advantage, “organizations must overcome the female disadvantage inherent in cultural stereotyping of women and leadership” (Eagly et al., 2014, p. 14). Therefore, in combination with the popularity of transformational leadership and the perceived masculinity of leadership in general (Koenig et al., 2011), I come to the following hypothesis:

Hypothesis 3: Employees prefer male, transformational managers over female, transformational managers or transactional managers.

(16)

16

Leadership in male- and female-dominated organizational contexts

Not only a person’s gender can influence employee preferences for managers and their leadership style. Eagly & Karau (2002) elaborate on various moderators influencing the perceived incongruence between women and leadership, which results in variation in the effect of sex across different social settings. One moderator for the perception of leadership of women and men is the extent to which the organizational context is male- or female-dominated. The incongruence between the leader role and the female gender role is often higher in relatively male-dominated contexts. Related to descriptive stereotypes, there is a perceived misfit of women’s leadership abilities, because there is a high incongruence between their communal traits and the highly agentic traits needed for leadership positions defined in masculine terms. And if women do behave very agentically, they may encounter a backlash effect because they do not behave “as women should behave”. On the other hand, role incongruence is expected to be lower in female-dominated contexts when leadership can be realized in less masculine terms (Eagly & Karau, 2002).

Kanter (1977) was one of the first to stipulate the importance of the organizational context when studying women in the workplace. She argued that women are perceived as tokens in male-dominated contexts – as exemplary of their gender instead of as individuals. The relative number of men and women in a team or organization affect perceived gender differences the more women have a so-called token status. Consequently, the token status of women reduces when the share of women in a team or organization goes up. Others have also elaborated on the effect of a low representation of women on perceived gender stereotypes. People may think that female-dominated professions require stereotypically feminine traits, and that male-dominated professions require stereotypically masculine traits. This leads to a perceived lack of fit that is especially negative for women when the job is male gender-typed – which may be determined by e.g. occupation and function or level within an organization – and when men are a majority in the organization (Cejka & Eagly, 1999, in Heilman, 2012). Ridgeway (2001) adds that gender status beliefs differ across contexts, also depending on the representation of men and women in a situation or the cultural beliefs about the task or goal of the job. Expectation states theory therefore predicts that gender status beliefs are most salient in “mixed-sex and gender-relevant contexts” (p. 648), i.e. the prejudice against female leaders will be most apparent in contexts that are culturally identified as masculine. Besides, research has shown that men hold more stereotypical beliefs towards female leaders than women (Schein, 2001; Koenig et al., 2011), or that the effect is mostly present in male-dominated

(17)

17

contexts (Funk, 2019). The greater number of men in an organization may therefore also affect manager preference.

Traditionally, public organizations have been identified as masculine environments (Ferguson, 1984; Stivers, 2002 in Funk, 2019). However, there are differences in the masculinity or femininity of public organizations. The two indicators of the male- or female-dominated organizational context – cultural beliefs about the type of work and the representation of men and women in the organization – are mostly correlated (Cejka & Eagly, 1999). There is a tendency for organizations in policy areas that are traditionally seen as more masculine, such as defense, finance and economics, to have more male employees. Consequently, organizations in healthcare, education and social services, which are policy areas traditionally associated with femininity, tend to have more female employees (Funk, 2019). Therefore, despite leadership positions being treated as typically masculine and male typed, some organizations are perceived to be feminine and have more female employees than men (Paustian-Underdahl, Walker, and Woehr, 2014).

Empirical research provides evidence that perceived gender differences in the effectiveness of leadership of women and men is affected by the organizational context. A recent meta-analysis of Paustian-Underdahl et al. (2014) mostly replicated earlier findings of the meta-analysis of Eagly, Karau, & Makhijani (1995). Consistent with role congruity theory, gender differences in effectiveness were significantly moderated by whether the organization under examination was male- or female-dominated. Both meta-analyses found that male leaders were perceived to be more effective in male-dominated organizational contexts, such as in military organizations, and female leaders were perceived as slightly more effective in less male-dominated or female-dominated organizational contexts, such as in educational organizations. Because role congruity theory concerns the stereotypical beliefs people hold about what constitutes a successful leader, manager preferences are also likely to differ across male- and female-dominated contexts.

In summary, prejudice against female leaders is expected to be more apparent in male-dominated organizational contexts, because of the stronger incongruence between the leader role and the female gender role, which results in more prejudice against women. This in turn may lead to different employee preferences for leaders in respectively male- and female-dominated organizational contexts. Besides, beliefs about the higher status and competence associated with men in general, which become especially apparent in mixed-sex and gender-relevant contexts (Ridgeway, 2001), may cause stronger preferences in male-dominated

(18)

18

contexts. In line with the literature, I therefore expect employees to have a stronger prejudice towards female leaders in male-dominated contexts. This leads to the following hypotheses: Hypothesis 4a: Employee preferences for male managers are stronger in male-dominated contexts than in female-dominated contexts

Hypothesis 4b: Employee preferences for agentic managers are stronger in male-dominated contexts than in female-dominated contexts

Hypothesis 5: Employee preferences for male, agentic managers are stronger in male-dominated contexts than in female-male-dominated contexts

Hypothesis 6: Employee preferences for male, transformational managers are stronger in male-dominated contexts than in female-male-dominated contexts

All in all, testing the hypotheses provides an answer to the research question regarding the relationship between gender, gender-stereotypical traits, and leadership styles. I should point out that the hypotheses specifically focus on employee preference for gender or gender-stereotypical traits, and the interaction between gender and gender-gender-stereotypical traits and gender and leadership styles. It would also be interesting to, for instance, compare the relative preference for the different components of transformational and transactional leadership. The research design and analysis make this possible. However, role congruity theory considers above all the interaction of these styles with a manager’s gender. For reasons of clarity of the argument, I will therefore focus on testing the hypotheses and not go into other potential interesting findings. In the following chapter, I will discuss the research design and conjoint analysis in more depth.

(19)

19

Method

I conduct a statistical analysis using data collected through a conjoint survey experiment, carried out among employees in the Dutch education, police and defense sectors. It is a replication of a Danish study held under Danish high school teachers in January 2019, of which the results are due later this year. For this research I examine two extra hypotheses. The first concerns the interaction between a manager’s gender and other manager characteristics, to test role congruity theory. Because the second concerns the influence of the organizational context on manager preferences, employees in other educational subsectors and in police and defense were also asked to fill in the survey.7_{In this chapter, I elaborate on the research design and}

method used in this thesis. The chapter will be outlined as follows. I start with a description of the case selection, after which I discuss the data collection and the final sample. Next, I describe the measurement and operationalization of the variables. At the end of the chapter, I explain the estimation strategy.

Data

Case selection

In order to measure employee preferences for managers and compare them across different organizational contexts, I use the cases of the Dutch education sector and the police and defense sectors. As explained in the theoretical framework, there are two indicators to assess the masculinity or femininity of an organizational context. The first is the ratio of men and women in the organization, and the second are the stereotypical beliefs about the organizational culture. The education sector and police and defense sectors are chosen as respectively a female-dominated context and a male-female-dominated context.

Regarding the first indicator, men comprise a majority in the Dutch police and defense sectors, whereas in education there are more women than men. Within the police, 64.2% of the employees is male and within defense, 85.8% of the employees are men.8_{The average}

percentage of female employees in education is 66.5%, but there are some differences within

7_{It should be noted that in the first place the survey was held among Dutch public sector employees as part of a}

replication of a Danish research project. In order to correctly compare the results of the Danish and Dutch respondents, the design of the conjoint experiment was already fixed. My part in it was to conduct my own research using this survey design and the resulting data from it. In addition, I translated the survey to Dutch and added several extra questions on the organizational context to take into account the influence of the male- or female-dominated context.

(20)

20

subsectors. In primary education, 84.5% of employees is female, whereas in higher education 48.9% of the employees are women (Ministerie van BZK, 2018).

With respect to the second indicator, police and defense have traditionally been masculine organizations because they are commonly regarded to maintain a high masculine culture. On the other hand, the education sector is often regarded as female-dominated, or at least less male-dominated (Eagly et al., 1995; Kark & Eagly, 2010; Paustian-Underdahl et al., 2014; Rice & Barth, 2016). The type of the job may indeed be more feminine, because education is associated with communal characteristics such as nurturant and caring. There are again differences between subsectors of education. Institutions of higher education are more associated with a masculine culture (Priola, 2007; Hearn, 2001), whereas elementary education is more associated with communal traits (Croft et al., 2015). Although I mainly consider the education sector to be more feminine, the research design allows to additionally compare the results for the different educational subsectors.

Data collection

Employee preferences for managers with varying gender, gender-stereotypical traits and leadership styles are measured by conducting a survey experiment among Dutch public sectors employees. Participants in this research are members from Flitspanel, which is an online panel of employees in the public sector initiated by the Dutch Ministry of the Interior and Kingdom Relations.9_{The survey was sent to panel members working in education, police and defense.}

For the education sector, the target groups were employees in primary education, secondary education, intermediate vocational education (MBO in Dutch), higher professional education (HBO in Dutch), and scientific education. Besides, employees working in police and defense, both military staff and civilian staff, were targeted. At the time of data collection in April 2019, all 9,094 members of the panel working in the sectors covered in this research have been approached by an e-mail with an invitation to participate in the online survey.10_{The survey}

took approximately ten to fifteen minutes to complete and I included the original Dutch version of the survey in Appendix A. After three weeks and two reminder e-mails, 2,783 members had participated in the survey experiment, which results in a response rate of 30.6%. Of these respondents, 26 were removed because of data quality and missing values, so the remaining sample consists of 2,757 respondents – a response rate of 30.3%.

9_See_{https://www.flitspanel.nl/}_{for more information on the panel}

10_{In the e-mail, the survey was announced as a research on preferences for various types of managers. Nothing}

(21)

21

Final sample

Descriptive statistics of respondent’s background characteristics and other survey questions used in this research can be found in the Appendix (Table B1). The average percentage of female respondents and the mean age in the final sample do not match the population averages, both for the entire sample as for the sectors separately. Of the final sample, 44.5% of the respondents is female and the mean age is 54.8 years, whereas of all employees working in the relevant sectors 57.7% is female and the mean age is 44.1 years. This can be explained by the fact that members in the Flitspanel are not representative of the population, as shown in Table 1. In addition, a comparison between Flitspanel and the final sample shows that female respondents were less likely to respond across all sectors, which can be explained by the perceived masculinity of the leadership topic. Because in the final sample female respondents are underrepresented and the respondent’s mean age is higher, I do various subgroup analyses in the next chapter to see whether there are notable differences in results and to be able to reflect on their validity in the discussion.

Additionally, 613 of the 2,757 respondents indicated that they currently have a leadership position. Of respondents in police and defense, even more than one third indicated to have a leadership position. I decided to include these respondents, because employees who have a leadership position will have managers themselves, so they will be able to state their manager preference. Besides, if I was to exclude these respondents from the final sample, the sample size of the male-dominated context would become too small to make a comparison. To be sure, I compared the results of the main analysis for all respondents with the results when I exclude the respondents with a leadership position. This resulted in highly similar outcomes and no statistically significant differences (see Tables C1a & C1b in the Appendix). I therefore decided to use the data of all respondents, including those with a leadership position, to have enough observations for the comparison and increase the statistical power of the analysis.

Table 1. Population, Flitspanel, and sample characteristics

Public sector employees Flitspanel Final sample

Age % Women Age % Women Age % Women

Education 44.6 66.5 55.2 52.4 54.9 49.9

Police 45.2 35.8 54.5 25.7 54.9 20.8

Defense 39.1 14.2 52.3 13.6 52.9 11.6

Total 44.1 57.7 54.9 47.0 54.8 44.5

(22)

22

Method of Analysis

The main part in this research is the conjoint analysis of employee preferences for public managers with various characteristics. In the survey experiment distributed among public sector employees, a conjoint design is used to measure manager preferences. One major advantage of conjoint experiments over traditional survey experiments is the ability to simultaneously manipulate a wide range of variables, instead of only one manipulation mostly possible in traditional survey experiments (Hainmueller et al., 2014; Rao, 2014). Conjoint analysis originates from marketing research, where it is still widely used, to determine consumer preferences for specific attributes of a product. Currently, it is increasingly being adopted in other research fields, such as business and political science.11_{Conjoint experiments can be}

performed in various ways and analyzed with multiple techniques, but essentially it comes down to respondents having to choose between or rate hypothetical profiles of persons or products, which combine various attributes in a randomized order across respondents.

Figure 1 shows how the conjoint experiment looked like for the respondents in this research. Five times in a row, respondents were shown a set of two potential direct managers with seven varying attributes. For each set, so five times in total, they had to answer three questions about the potential direct managers on a seven-point scale. Each attribute had two different categories (see Table 2), which were randomly assigned per task and respondent for each manager profile. In total, seven attributes with each two possible levels result in 128 possible combinations of manager profiles. The randomization of the manager profiles makes it possible to assess the relevance of each of the attributes for a respondent’s manager preference, and which category of the attribute is more positively evaluated (Hainmueller et al., 2014).

11_{see for example Rao, 2014; Raghavarao, Wiley, and Chitturi, 2011; Hainmueller et al. (2015); Ono & Yamada}

(23)

23

Figure 1. Experimental Design

Measurement

Here, I discuss the measurement of the concepts used in the main research question and the hypotheses. The main outcome variable of interest in this research is taken from the first question below the conjoint design, which directly measures a respondent’s preference for a certain type of manager on a seven-point scale. I refer to this as the dependent variable prefer. The other two questions can also be interpreted as employee preference but in a slightly different way. These dependent variables, respectively referred to as assignment and enjoy are therefore included as a robustness check. To analyze manager preference, I coded the responses on the seven-point scale as a discrete variable, giving the attribute levels shown for the first manager a weight of 1 if the respondent chose one of the first three options and a weight of 0 if

(24)

24

the respondent chose one of the last three options on the scale. An equal weight of 0.5 was given to all attribute categories shown for both managers when the respondent selected the middle option “Indifferent”.12_{Originally, the experiment was set up to use the seven-point scale}

as an interval scale giving equal weight to each of the options. However, I decided to analyze the data use a three-point scale, which comes closer to a choice-based conjoint design. I am in fact interested in preferences for certain types of managers, and not in the extent to which respondents prefer a certain type. Thus, I argue that when a respondent chooses ‘Almost certainly’ or ‘Probably’, he or she prefers the attributes of that manager profile over the attributes of the other manager profile.13

The independent variables are the manager characteristics – called attributes in the conjoint design – which are measured for two different categories, called levels, per attribute. Table 2 presents an overview of the theoretical concepts and their measurement in the conjoint experiment. They are all literary translated from the Danish design and follow logically from the literature as outlined in the theoretical chapter. Except for the attribute levels of ‘Goal’, because the Danish research only focuses on high school teachers and the levels were defined in sector-specific terms. I redefined the levels to make them relevant for all the sectors in this study.

12_{With the first three options I mean “Certainly A – Almost certainly A - Probably A” for the first set of managers.}

With the last three options I mean “Probably B – Almost certainly B – Certainly B”. In total each respondent had to rate five sets of managers, so the others were called manager C, D, E, F, G, H, I, J. See Table B2 in the Appendix for an overview of the coding of the scales.

13_{The results of the original interval scale do not alter the conclusions, only the significance and the size of the}

effects goes down a bit. Besides, the relative importance of the attributes remains the same. The analysis is provided as a robustness check in the next chapter.

Table 2. Manager profiles’ attributes and their levels in the conjoint experiment

Attributes Levels Concepts

Goal Focuses on the efficient use of the available financial resources Transformational leadership Focuses on realizing the social task of the organization Transactional leadership Self-confident Is not particularly self-confident and ambitious

Is very self-confident and ambitious Agentic traits Caring Is not particularly caring and understanding

Is very caring and understanding Communal traits

Vision Does not clarify a vision for the organization

Clarifies a vision for the organization Transformational leadership Compliment Does not compliment work well done

Compliments work well done Transactional: verbal rewards

Gender Male Male

Female Female

Rewards Does not reward employees

(25)

25

First, to examine how a manager’s gender affects employee preference for managers, the levels of the attribute gender are employed as either male or female. Second, gender-stereotypical traits are divided into two separate attributes, one for agentic traits and one for communal traits. Agentic traits are operationalized into the attribute self-confident, in which the levels denote whether the potential direct manager is very self-confident and ambitious or not. Communal traits are operationalized into the attribute caring, in which the levels refer to a very caring and understanding manager or a not very caring and understanding manager. Third, the leadership styles are measured with four different attributes and are operationalized in line with the suggestions of Jensen et al. (2019). Transformational leadership is therefore measured by referring to a manager who focuses on pursuing the social task of the organization and a manager who clarifies a vision for the organization. This is in line with the concepts of transformational leadership, particularly the visionary leadership elements of this style. Transactional leadership is operationalized into a manager who focuses on the financial goal of the organization. Besides, the contingent rewards component is divided into two parts. The pecuniary rewards component refers to a manager who rewards employees when they perform well on the job with for instance wage supplements or courses. Non-pecuniary rewards are operationalized into a manager who compliments employees for work well done. The use of sanctions is not used in this research, because for the original Danish experiment it was argued that sanctions are less relevant to study in the context of education (Fjendbo, 2019, paper under review).

Next to the conjoint experiment, various other questions were included in the survey to allow for extra analyses.14_{Before the experiment, respondents were asked to indicate the sector}

they currently work in, which allows me to compare the results for respondents working in the male-dominated context, i.e. in police and defense, with respondents working in the female-dominated context, i.e. education. Directly after the experiment, respondents were asked to rank the seven attributes in order of importance for manager preference. As a robustness check, I compare the order of the attributes from this question with the relative importance of the attributes from the conjoint experiment. In addition, respondents were asked various questions about their background, such as age, gender, whether they have a leadership position, whether they have experience with both male and female managers, and about the gender of their current direct manager. All these characteristics may have an influence on manager preference, so I compare the results for various subgroups of respondents depending on these characteristics as

14_{I did not use all the survey questions in this study, so I only describe the relevant extra survey questions for my}

(26)

26

additional analyses. Finally, I included extra questions at the end of the survey to gain more information about the masculinity or femininity of the organizational context. This was measured by the employee-perceived ratio of men and women in their organization, and the importance of various agentic and communal traits within the organization, measured on a five-point Likert-scale ranging from “unimportant” to “very important”.15_{Together, this gives an}

indication of the masculinity or femininity of organizational context as perceived by respondents.

Statistical approach

In order to test the hypotheses, I am not only interested in the effect of the attributes on manager preference, but also in the differences in effects conditional on the attribute gender. I therefore used the statistical approach as developed by Hainmueller et al. (2014). First, because there is no need to make a lot of statistical assumptions when using this approach, which leads to a higher internal validity than the more model-dependent methods used in conjoint analysis. Second, and more importantly, the approach has the possibility to estimate not only the effect of each of the attributes, but also the interaction effects between the attributes. Besides, it allows to easily compare the results for various subgroups of respondents (idem, 2014).

In short, the approach estimates the Average Marginal Component Effect (AMCE) of each attribute level on the probability that the manager will be preferred, averaged over any combination of the other attributes. Because of the random assignment of the attributes, which causes the distributions of all other attributes to be on average the same, the AMCE can be nonparametrically identified (Hainmueller & Hopkins, 2015; but see Hainmueller et al. (2014) for a comprehensive explanation of the approach). The effect of a certain attribute level is measured against a baseline, which is always the first level of the attribute in Table 2. For example, the effect of gender is measured with ‘male’ as the baseline, so that a positive value means that there is a higher preference for a female manager and a negative value that there is a higher preference for a male manager. The same principle applies to all other attributes, and the effects should be interpreted as follows. First, the AMCE says something about the difference of a level with its baseline within the attribute. The AMCE refers to the average change in probability that a certain type of manager will be preferred when it has the specified

15_{The importance of five communal and five agentic traits in the organization had to be rated on a five-point}

Likert-scale ranging from “unimportant” to “very important”. For the analysis, I recoded the answers from 0 to 1 and present the average for the communal traits and the average for the agentic traits. A t-test is used to compare the statistical significance of the differences in means.

(27)

27

attribute level (the second level in Table 2) instead of the baseline level (Hainmueller et al., 2014). Thus, the bigger and more significant the effect, the more the level (e.g. “very confident and ambitious”) is likely to be preferred over its baseline level (e.g. “not very self-confident and ambitious”). Second, the AMCEs can also be compared to each other, which allows to compare the relative importance of preference of the attributes. The higher the AMCE, the higher the probability that a respondent will choose that level over other attribute-levels. The AMCEs can therefore be interpreted as the mean importance of that attribute-level for a potential direct manager, taken over all respondents (idem, 2014).

In addition, the approach allows to estimate the Average Component Interaction Effect (ACIE), which measures the effect of certain attributes conditional on a chosen other attribute, in my case gender. The effects of gender-stereotypical traits and leadership styles are expected to vary depending on the hypothetical manager’s gender, and this is measured with the ACIE. In other words, the ACIEs say something about how a manager’s gender may condition employee preferences for the other attributes (Hainmueller et al., 2014).

Lastly, the effects may also vary depending on respondent characteristics, such as gender, age and or the sector they work in. The approach allows to estimate the AMCEs and ACIEs conditional on respondent characteristics, divided into various subgroups. I subsequently use an F-test to test the statistical significance of the difference between the results of the subgroups of respondents, according to the same approach in Hainmueller and Hopkins (2015) and Horiuchi et al. (2018).

Estimation strategy

For the estimation of the conjoint experiment data, I used the “cjoint” package in R (ver. 2.1.0) developed by Strezhnev et al. (2016) and parts of the R-script from the replication package of Horiuchi et al. (2018). Each of the 2,757 respondents rated five sets of potential direct managers, with two hypothetical manager profiles per set. This leads to 27,570 observations for the outcome variable ‘prefer’ when using the entire sample. Standard errors are clustered by the respondent to correct for dependencies of the observed outcomes across the rated profiles by a single respondent (see also Hainmueller et al., 2014, p.16-17), and to control for unobserved differences between respondents. For the main analysis, I used a combined model to estimate both the AMCEs and ACIEs with the coded three-point scale for the outcome variable prefer. In the next chapter, I present the findings of the conjoint analysis, extra analyses, and robustness checks.

(28)

28

Analysis

In this chapter, I present the findings of the conjoint analysis. I estimate the effect of each of the attributes on the probability that a certain type of manager will be preferred over the other, as well as the interaction effects between the gender attribute and each of the other attributes. The analysis is divided into four parts and is structured as follows. First, I conduct a conjoint analysis of employee preferences for all sectors together to test hypotheses 1a, 1b, 2 and 3. I estimate the Average Marginal Component Effect (AMCE) and the Average Component Interaction Effect (ACIE) of the attributes on manager preferences by considering all respondents. Second, I test hypotheses 4a, 4b, 5, and 6 by estimating the AMCE and ACIE respectively for the education sector and the police and defense sector and compare the results. Third, I conduct additional analyses by comparing the results for different groups of respondents based on gender, age, experience with (previous) managers, and whether the respondent has a leadership position him- or herself. Fourth, I take a closer look at the organizational context by looking at the employee-perceived masculinity or femininity of their context and providing a within-sector analysis. Then, at the end of the chapter, I conduct several robustness checks.

Effects of manager attributes on employee preferences

Figure 2 shows the results of the AMCE for all respondents, using the three-point scale for the dependent variable prefer. The dots represent the point estimates and the horizontal bars indicate 95 percent confidence intervals. The results represent the estimated AMCE of each attribute level on the probability that a respondent chose a potential direct manager with that attribute level (e.g. ‘Female), compared to a potential direct manager with the baseline level of the same attribute (e.g. ‘Male’). The figure also reveals the relative importance of the attributes for respondent’s manager preference. A detailed version of the results can be found in Table 3 column 3, where the clustered standard errors and an indication of the significance are also included. The second column in the table shows the level that is measured against the baseline, i.e. the attribute level for which the effects are presented. This means that, for example, concerning the attribute ‘Self-confident’ (column 1, row 3), the level ‘Is very self-confident and ambitious’ (column 2, row 3) is on average 4.9 percentage points (column 3, row 3) preferred over its baseline level ‘Is not very self-confident and ambitious’.

(29)

29

Figure 2. AMCE of hypothetical managers’ attributes on employee preference.

Note: Dots represent the estimated AMCE of the randomly assigned manager attributes on the probability of being preferred as a potential direct manager. Horizontal bars represent 95% confidence intervals.

Employee preferences for a manager’s gender and his or her gender-stereotypical traits First, it is expected that that employees prefer male managers over female managers (hypothesis 1a). The results of the AMCEs of the hypothetical manager’s attributes for all respondents (Figure 2 and Table 3 column 3) do not support this. Regarding the attribute gender, the effect of a female manager over the baseline male manager is close to zero and highly insignificant. It can therefore be concluded that, on average for all respondents, gender is irrelevant when it comes to manager preferences of public sector employees. This is reinforced by the fact that the AMCEs of all other attributes are bigger and significant.

Second, regarding hypothesis 1b, there is no support for the expectation that employees prefer agentic over communal managers. A manager with communal traits such as caring and understanding is on average 24 percentage points more likely to be preferred over a manager who does not have these traits. Besides, a manager who is self-confident and ambitious, related to agentic traits, is only 4.9 percentage points preferred over a manager who does not have these characteristics. Because the conjoint design enables to compare the relative importance of the attributes and their levels on the same scale, the higher AMCE for the attribute caring indicates employees have a more pronounced preference for communal traits than agentic traits in hypothetical public managers. Interestingly, caring has the highest AMCE of all attributes,

(30)

30

which means that, on average, employees in the public sector consider the presence of communal traits in a manager to be the most important concerning manager preference.

Third, it is expected that employees prefer male, agentic managers over female or communal managers (hypothesis 2). Even though employees find a manager’s gender irrelevant in terms of preference, the effects of the gender-stereotypical traits attributes – self-confident and communal – may be different conditional on a manager’s gender. Role incongruence is therefore measured by the ACIE, which measures the effect of the other attributes conditional on the attribute gender. For the hypothesis to be confirmed, this means that the effect of the attribute self-confident should be higher for a male manager than for a female-manager, and that the effect of self-confident should be higher than the effect of the attribute caring for both male and female managers. The results shown in Table 3 do not support this. Each ACIE (column 6) indicates the percentage point difference between that specific attribute effect for female managers (column 5) and for male managers (column 4). A positive effect refers to a higher preference for the attribute conditional on a female manager, whereas a negative effect refers to a higher preference for the attribute conditional on a male manager.16_{Regarding this}

interaction between gender and gender-stereotypical traits, there is a significant negative interaction effect for female, agentic managers. This means that male managers who are self-confident and ambitious get a significantly larger stimulus (2.4 percentage points) than female managers with these traits. The interaction between gender and the attribute caring is not significant, which means that, on average, the preference for communal traits is equal conditional on a male or female manager. But again, conditional on both male and female

16_{This also means that the ACIE estimates for male managers are the same as the estimates for female managers,}

but then the other way around (positive becomes negative and vice versa). Table 3. Results AMCE and ACIE for all respondents

Attribute Level AMCE SE AMCE: Male SE AMCE: Female SE ACIE SE

Gender Female -0.002 0.0056 - - - -

Self-confident Self-confident 0.049*** 0.0061 0.061*** 0.0081 0.037*** 0.0078 -0.024* 0.0102

Caring Caring 0.240*** 0.0063 0.245*** 0.0080 0.235*** 0.0078 -0.010 0.0096

Goal Social goal 0.085*** 0.0065 0.089*** 0.0082 0.082*** 0.0086 -0.007 0.0105

Vision Vision 0.196*** 0.0065 0.189*** 0.0081 0.203*** 0.0081 0.015 0.0097

Compliment Compliment 0.211*** 0.0058 0.195*** 0.0079 0.226*** 0.0077 0.031** 0.0104

Rewards Rewards 0.162*** 0.0062 0.165*** 0.0080 0.159*** 0.0078 -0.006 0.0098

Observations 2757 2757 2757 2757

Gender and Leadership in Public Organizational Contexts: A conjoint analysis of employee preferences for public managers in male- and female-dominated public sectors in the Netherlands

Master Thesis Public Administration

Track: Economics & Governance