Fight or flight: How gender influences follower responses to unethical leader behaviour

ABSTRACT

Does unethical leader behaviour produce gendered responses? We study follower response to unethical leader behaviours of bribery and workplace harassment. Our findings, based on an experiment, suggest that leader gender, gender congruence, and implicit gender biases influence follower responses to unethical leader behaviour. These findings raise the troubling implication that unethical male leaders will remain entrenched in positions of power.

KEYWORDS:

• Gender
• unethical leadership
• follower behaviour

Introduction

A rich empirical literature suggests that followers differentially respond to a leader’s behaviour depending on whether the leader is male or female. Followers tend to perceive female leaders who discipline employees as less effective and fair than their disciplining male counterparts (Atwater et al. 2001). Followers are also less receptive to female leaders imposing discipline unless that discipline is a two-way discussion (Brett, Atwater, and Waldman 2005). Female leaders expressing anger are perceived by followers as less effective than their male counterparts, and male leaders expressing sadness are perceived by followers as less effective (Lewis 2000). At a more general level, followers evaluate female leaders more negatively than their male counterparts, especially when the leader’s style is autocratic and the setting is male-dominated (Eagly, Makhijani, and Klonsky 1992).

Although a significant body of research suggests that followers have deeply gendered responses to leaders, three limitations open the door to the present study. First, these studies describe gendered responses to normal leadership behaviour, leaving open the question of how gender influences follower responses to unethical leader behaviour. Second, these studies focus entirely on private sector settings, paving the way for research conducted in the public sector. Third, the literature on gendered responses to unethical behaviour is in its early stages of development (Bowles and Gelfand 2010; Kennedy, McDonnell, and Stephens 2016; Montgomery and Cowen 2020) and, as such, is ripe for fresh contribution.

Seeing these limitations as research opportunities, we seek to understand how gender influences follower responses to unethical leader behaviour in a public sector setting. Two types of unethical leader behaviours are of particular interest in the public sector: bribery, which harms the organization, and workplace harassment, which harms individual organizational members. And while the private sector literature has identified a range of follower responses to unethical leader behaviour, we investigated the broader binary choice of whether the followers will ‘fight’ by speaking up about the unethical behaviour, or ‘flee’, by exiting the organization (Mayes and Ganster 1988). Drawing on theories related to gender and leadership, we hypothesize that unethical female leaders are more likely to elicit fight responses, while male leaders will more likely elicit flight. Furthermore, we expect that the congruence of follower and leader gender and gender-authority biases will predictably moderate these responses.

To test these predictions, we conducted an experiment in which 245 research participants responded to vignettes about leaders’ unethical behaviour and took tests of explicit and implicit gender authority bias. The hypothetical leader in our experiment was a city manager whose unethical behaviour took the form of either bribery or workplace harassment. The leader’s gender was randomly assigned to participants. Experimental participants were asked in closed and open-ended questions how they would respond to the unethical leader in terms of exit (flight) and voice (fight) responses. The findings suggest that the nature of the unethical behaviour, leaders’ gender, followers’ gender, and followers’ gender authority bias all influence followers’ responses to leaders’ unethical behaviour in nuanced ways.

This study makes several contributions to the public administration literature. While, unfortunately, unethical leader behaviour is not uncommon, it is an understudied topic by public administration scholars (Hassan 2019). As such, we are unaware of any studies conducted in a public sector setting that examine gendered responses to leader behaviour, ethical or unethical. The role of ethics in follower responses to unethical leader is a timely topic, especially as it relates to effective public sector functioning.

In the first section, the theoretical background of the study is discussed, leading to our hypotheses about follower responses to leaders’ unethical behaviour. The second section describes the methods, detailing the experimental design, measurement, and results. The third section discusses the implications of the results, acknowledges study limitations, and outlines fertile ground for future research. The final section draws conclusions from the research results.

Theoretical background and study hypotheses

Recent public administration research points towards an array of gendered responses to normal leadership behaviour by women and men. In the context of the present study, female city managers and police officers report experiencing greater challenges to their authority by followers (Portillo 2012). In a multi-method study of gender imbalance in public service leadership, city and county managers who are women recounted questions about their competence and availability to lead given childcare demands (DeHart-Davis et al. 2020). Qualitative interview research confirms this finding, in which a sample of women city managers located in the Southeast and Northeast recalled feeling compelled to demonstrate their competence and commitment to the job to followers (Bishu and Headley 2020). These experiences lead women who are local government managers to legitimize their authority by appealing to rules (Portillo 2012) and being overly prepared with evidence to support positions (Bishu and Heckler 2020; DeHart-Davis et al. 2020).

In contrast with studies of gendered responses to normal leadership behaviour, we examine the gendered nature of follower responses to unethical leadership behaviour. Unethical leader behaviour involves ‘behaviours conducted and decisions made by organisational leaders that are illegal and/or violate moral standards, and those that impose processes and structures that promote unethical conduct by followers’ (Brown and Mitchell 2010, 588). While scholars have explored a variety of follower responses to unethical leader behaviour, including employee effort (Gan et al. 2019), resistance to change (Moutousi and May 2018), and follower unethical behaviour (Ashforth and Anand 2003), we conceptualize followers’ response to leaders’ unethical behaviour using Mayes and Ganster’s ideas of ‘fight’ or ‘flight’ responses to stress (Mayes and Ganster 1988).

Mayes and Ganster map their fight or flight responses onto Hirschman’s (1990) exit-voice-loyalty responses to dissatisfaction with an organization. Voice involves remaining with the organization and using sanctioned and non-sanctioned mechanisms to make issues known and thus, in a sense, is a type of ‘fight’ behaviour (Rusbult et al. 1988). Exit entails a type of retreat and a recognition that the workplace situation is beyond one’s influence, and thus is a type of ‘flight’ behaviour (Travis and Mor Barak 2010). As the following section illustrates, we expect that fight or flight responses will be determined not only by leader gender, but also the interaction of leader and follower gender, and gender-authority biases. We begin our arguments with theory related to leader gender and follower responses to unethical leader behaviour.

Leader gender and follower responses to unethical leader behaviour

We draw on theories of role congruity and status characteristics theories to argue that followers will respond to unethical leader behaviour based on the gender of the leader. Within the role congruity perspective, women and men are expected to behave differently because they occupy distinct social roles. Eagly and Karau (2002) defined social roles as the expectations for people who occupy a certain position or social category (p. 574); these expectations pertain to how group members actually behave (descriptive norms) as well as how they should behave (injunctive norms).

Present-day gender norms emerged from the historical division of labour, in which women worked as homemakers and men as breadwinners (Eagly and Kite 1987; Stivers 2002). This traditional division produced the belief that women behave in a communal manner and men behave in an agentic manner. Communal behaviour and attributes cluster around relationships with others, and include being helpful, nurturing, gentle (Johnson et al. 2008), friendly, unselfish, concerned with others, and emotionally expressive (Eagly 1993). Agentic behaviour, by contrast, is self-oriented and incorporates independence, mastery, assertiveness, competence (Eagly, Makhijani, and Klonsky 1992), dominance (Williams and Tiedens 2015), strength, masculinity, tyranny (Johnson et al. 2008), control, aggression, ambition, forcefulness, and self-sufficiency (Brems and Johnson 1990).

Role congruity theory argues that because leadership is associated with agentic attributes, women who occupy leadership roles risk violating gender expectations (Eagly and Carli 2003; Eagly and Karau 2002). Empirical evidence largely supports this expectation. Dominant behaviour in the form of making direct demands appears to reduce women’s perceived likeability and ability to be hired (Williams and Tiedens 2015). Self-promoting women are perceived as more competent but less likeable and hireable (Rudman 1998). In addition, women who attempt to negotiate a higher salary are more likely to be penalized (by not being hired or having uncooperative colleagues) than men who engage in such negotiations (Bowles, Babcock, and Lai 2007). Further, agentic female job candidates are less likely than androgynous applicants to be hired for feminized job assignments (Rudman and Glick 2001).

Women incur gender penalties by merely assuming a leadership role, but male leaders tend to be evaluated more favourably than female leaders. This advantage is intensified in male-dominated environments (Eagly, Makhijani, and Klonsky 1992). The negative effects of breaking with stereotype constitute a sort of ‘backlash’ against female leaders that affects hiring, salary negotiations, promotion, and peer sabotage (Rudman and Phelan 2008).

Because unethical behaviour defies the prescriptive female stereotype and contradicts communal values, this behaviour can be considered a type of agentic action for which women are punished more harshly than their male counterparts. This has prompted researchers to explore gender differences in responses to unethical leader behaviour (Heilman 2012; Kennedy, McDonnell, and Stephens 2016). For example, an experimental study of workplace deviance (conducted as part of a large multi-method study) found that white male research subjects punished other white male deviants more leniently, on average, than female or black deviants (Bowles and Gelfand 2010). In another study on gendered reactions to unethical behaviour that included both experimental and field research, Kennedy and colleagues found that women were expected to act more ethically than their otherwise identical male counterparts and were punished more harshly for ethical violations (Kennedy, McDonnell, and Stephens 2016). Most recently, experimental evidence suggests that research participants penalize firms that commit ethical errors more if they are led by women than by men (Montgomery and Cowen 2020). Based on this evidence, we expect that followers will ‘fight back’ against unethical female leaders as a way of pushing back against their agentic behaviour.

In contrast with the expected pushback against unethical female leaders, we expect followers of unethical male leaders to feel less empowered in challenging a high-status male. Thus, the followers more likely to ‘flee’, that is, look for another job. We return to status theory to make this argument. Society accords men higher cultural dominance, respect, and status, which translates into greater organizational power for male leaders (Ridgeway 2011). This phenomenon is seen in research showing that agentic men in organizations receive positive or neutral responses, while women receive negative responses. To illustrate, men engaging in task-oriented behaviours in team meetings were more likely to emerge as leaders than their female counterparts (Schlamp, Gerpott, and Voelpel 2021); innovative men have been observed receiving higher performance evaluations than innovative women (Luksyte, Unsworth, and Avery 2018); and intimidating male supervisors did not suffer lower likeability ratings, and their performance appraisals were more positive than intimidating female supervisors (Bolino and Turnley 2003). We apply this evidence to the expectation that followers of unethical male leaders feel less empowered to ‘fight back’, and thus choose to ‘flee’, searching for another job instead.

Because followers generally reward men and punish women for agentic behaviour, we expect these behavioural patterns to translate to follower responses to unethical leader behaviour. Specifically, we expect that the cultural dominance, respect, and greater status accorded men (Ridgeway 2011) will lead followers to be more likely to flee unethical leader behaviour by exiting the organization. By contrast, we expect the tendency to punish agentic women (Bowles and Gelfand 2010; Kennedy, McDonnell, and Stephens 2016) engaging in unethical behaviour will stimulate a fight response in followers, leading them to be more likely to push back against the leaders by raising their concerns in the workplace. Placing these expectations in the context of specific unethical leader behaviours, we developed the following hypotheses:

Hypothesis 1: Followers are more likely to choose a fight (voice) response when female leaders engage in unethical behaviour than when male leaders engage in unethical behaviour.

Hypothesis 2: Followers are more likely to choose a flight (exit) response when male leaders engage in unethical behaviour than when female leaders engage in unethical behaviour.

Leader-follower gender congruence and responses to unethical leader behaviour

The congruence of demographic characteristics between leaders and followers has been shown to affect follower behaviour (Tsui and O’reilly 1989), a finding that has led to subsequent research on the gender congruence between leaders and followers (Piatak, Mohr, and McDonald 2020). The underlying assumption of gender congruence research is that same-gender dyads of leaders and followers lead to favourable behaviours as followers look to same-gender leaders as role models or seek to reciprocate any benefits that accrue as a result of demographic representation (Marvel 2015). This assumption has been supported in national surveys of K-12 teachers and principals in the US, in which male teachers with male principals were more satisfied and stated lower turnover intentions than male teachers with female principals (Grissom, Nicholson-Crotty, and Keiser 2012), and female teachers with female principals tend to work more hours than female teachers with male principals (Marvel 2015).

In contradictory evidence, an experimental study found that gender congruence between hypothetical managers (whose gender was randomized) and research participants failed to produce different levels of willingness to break rules, either for a good employee or for a grant applicant submitting a late application (Piatak, Mohr, and McDonald 2020). Scholarship on gender congruence has also produced contradictory findings. Another survey study of K-12 teachers in Denmark found that male teachers with male principals were more apt to break rules and less supportive of school goals than male teachers with female principals (Pedersen and Nielsen 2016). Two studies conducted within a state agency found that gender incongruent managers and employees led to higher employee performance ratings (Hassan and Hatmaker 2015) and greater managerial support (Hatmaker and Hassan 2021).

Although some evidence suggests that gender congruence between leaders and followers has led to favourable follower responses in some settings, we argue that follower responses to unethical leader behaviour engenders differing behavioural dynamics. Specifically, we argue that gender congruence between followers and unethical leaders can lead to a sense of social betrayal when same-gender leaders violate personally held behavioural expectations (Elangovan and Shapiro 1998). Employee organizational citizenship behaviours and performance decrease with the perception that their organizations have breached their obligations to them (Restubog et al. 2010). Given evidence that followers set higher standards for leaders of their respective genders than of opposite gender (Cooper 1997), we pose the following hypotheses:

Hypothesis 3: Female followers with female leaders and male followers with male leaders are more likely to engage in fight (voice) responses to unethical leader behaviour.

Hypothesis 4: Female followers with male leaders and male followers with female leaders are more likely to engage in flight (exit) responses to unethical leader behaviour.

Gender authority bias and follower responses to unethical leader behaviour

Researchers on gender and leadership have detected a pronounced bias against women in positions of authority. Experimental evidence has shown this bias among research participants who associated men with high authority and women with low authority (Rudman and Kilianski 2000). More recently, gender bias against women in positions of authority has been observed in an array of public management settings, ranging from local government (Bishu and Heckler 2020; Bishu and Headley 2020; DeHart-Davis et al. 2020) to policing (Bishu and Headley 2020; Portillo 2012) to the military (Doan and Portillo 2017, 2019). This bias fluctuates based on the setting in which women hold authority, a concept known as translocational positionality (Anthias 2012; Doan and Portillo 2017). Given that public service organizations are a setting where women are underrepresented as leaders (DeHart-Davis et al. 2020), we argue that gender authority bias is another important factor for understanding followers’ attitudes towards leaders’ unethical behaviour.

We based our expectation about the importance of gender authority bias on social status theory, which explains the persistence and resilience of gender-based inequality in spite of major social and economic transformations (Ridgeway 2001, 2011). Ridgeway (2011) posited that the basis for persistent inequality lies in gender status beliefs reinforced by ‘positional inequalities between men and women that provide men with more resources and power, on average, than women have’ (27). Thus, widely held cultural beliefs attribute competence and social significance to certain categories of people rather than to the disproportionate power or resources they hold (Ridgeway 2001). While status beliefs characterize a culture or society, individuals can vary in how closely their own beliefs align with cultural assumptions (Ridgeway 2001, 643). Thus, it is important to understand whether followers’ gender authority bias interacts with leaders’ gender to influence followers’ responses.

Social status theory postulates that unethical female behaviour violates behavioural expectations and thus trigger greater resistance – backlash and dislike – than unethical male behaviour (Correll and Ridgeway 2006). In addition, social status theory anticipates that lower status group members experience three consequences relevant to follower response: greater scrutiny of performance, a lower level of legitimacy, and less influence (Correll and Ridgeway 2006). In the context of leaders’ unethical behaviour, followers who accord male leaders a higher status than female leaders will interpret unethical behaviour from female leaders more harshly, be less cooperative in interactions with female leaders, and more readily challenge the legitimacy of women as leaders compared with their male counterparts. This reasoning leads to the following hypotheses:

Hypothesis 5: Followers with gender authority bias (favouring men over women) will be more likely to fight (voice) unethical female than male leaders.

Hypothesis 6: Followers with gender authority bias (favouring men over women) will be more likely to flee (exit) unethical male than female leaders.

Methods

Study design and participants

The study experiment had a 2 (Unethical Behaviour Type) × 2 (Leader’s Gender) between-subjects design and was administered online using Qualtrics in 2016. Figure 1 presents the experimental design. We recruited participants through listservs of the School of Government at the University of North Carolina-Chapel Hill (UNC), GovLoop, the Emerging Local Government Leaders network, and the Alliance for Innovation Knowledge Network.² These listservs cater to the needs of local government professionals working in middle and upper management. In return for participating, participants were given access to a free webinar on emotional intelligence offered through the UNC School of Government. Participation was voluntary and anonymous, with no identifiers collected.¹

Figure 1. Survey experimental design.

Each participant was randomly assigned to one of the four treatment conditions via the randomization option in the Qualtrics program. In total, 245 participants read and responded to the unethical behaviour vignette. We also collected data on gender authority biases and demographic characteristics; 11 participants did not complete the gender authority bias questions/tasks, and 18 respondents did not provide demographic information. Among the 227 participants who provided demographic information, 75% were female, 85% were younger than 50 years old, 35% had an associate degree, 40% had a bachelor’s degree, 85% worked for the local government, and about 50% had held a management position at some point.

Experimental manipulations

Again, per Brown and Mitchell (2010), unethical leader behaviour is defined as ‘behaviours conducted and decisions made by organisational leaders that are illegal and/or violate moral standards, and those that impose processes and structures that promote unethical conduct by followers’. Borrowing from this definition, we focus on two unethical behaviours that are both illegal and immoral: bribery and workplace harassment. Bribery was chosen because it is a criminal act that harms the public organization, particularly at the local government level (Nelson and Afonso 2019). Workplace harassment was chosen it can be illegal and harmful to individual organizational members (Neall and Tuckey 2014). To design realistic vignettes featuring these behaviours, we reviewed media articles that reported on the prosecution/firing of local government officials for taking bribes and for mistreatment of employees. In the pilot testing phase, we reviewed open-ended responses in which participants explained the rationale for their decisions. All respondents stated that they would take action. The underlying rationale for this choice was that taking bribes was unacceptable and unethical, and mistreating employees violated human resource norms and created an unhealthy and toxic workplace.

Study participants were asked to imagine themselves in the role of a department head. The following vignette described their role and controlled for extraneous variables (job satisfaction, job tenure, investment in work, relationships with colleagues, professional reputation) that could influence follower response to a work situation:

You work for a medium-sized U.S. city as a department head. Your job involves work that you find enjoyable, and you feel satisfied with your job. You have held the position for a reasonably long period of time; you have worked there for two years. You have invested time and energy in your job, and you are well-respected by your colleagues and peers. At a recent professional conference, some of your peers inquired if you would consider working for them instead.

In the text that followed this introductory material, we introduced the leader in the form of the city manager in charge of the department head (i.e. the participant). We manipulated the leader’s gender by using different names and we controlled for potential confounding variables (leader tenure, performance, impact). Thus, the only difference between the vignettes for the female leader and the male leader were their names: ‘Rebecca Sue Carlson’ and ‘Robert Steven Carlson’. Our goal was to use names that clearly and unambiguously signal gender. Given Tabak et al.’s (2005) success in in signaling gender, we decided to use the same names they used in their study. The text continued as follows (female leader version):

Your City Manager, Rebecca Sue Carlson, was sworn into office in December 2012. As city manager, Rebecca is responsible for overseeing the city’s $525 million budget and 2,600 full-time employees and managing city operations. During her tenure, the city has attracted residents and businesses from throughout the region. At the annual awards program of your state’s city and county management association in 2015, Rebecca was recognized with ‘City Manager of the Year’ award, honouring individuals who make significant contributions to local government in your state.

We manipulated the unethical behaviour via the following text. The first vignette focuses on bribery:

However, disconcerting information has come to your attention, through your professional contacts. You have credible knowledge that Rebecca is accepting under-the-table payments from Sigma Health Associates, a health care company, which is seeking permission to build a clinic in town.

The second vignette focuses on workplace harassment:

Rebecca, however, is also known for making racist and sexist comments to employees and resorting to aggressive physical actions. This past week you attended a meeting at which Rebecca proclaimed zero tolerance for slackers and threatened public humiliation and layoffs for poorly performing city employees. You are concerned about the City Manager’s pattern of harassing and intimidating behaviours.

After the participants read the full vignette, we collected data on their exit and voice responses, their explicit and implicit biases against women in authority (gender authority bias), and their demographic characteristics.

Measures

We used Rusbult et al. (1988) items for measuring an exit response on a scale from 1 (definitely would not react this way) to 9 (definitely would react this way). The three exit response items were: 1) I would think about quitting my job, 2) I would contact my peers who are interested in hiring me, and 3) I would explore other job openings. Following Rusbult et al. (1988), we averaged the three item scores (α = 0.82).

We also used Rusbult et al.’s (1988) items for measuring a voice response on a scale from 1 (definitely would not react this way) to 9 (definitely would react this way). The three voice response items were: 1) I would ask my co-workers for advice on what to do, 2) I would independently research my options for taking action, and 3) I would discuss the problems with the city attorney’s office. Because the reliability score for these items was low (α = 0.25), we reviewed the literature to identify other measures of voice. Our review suggested that although alternative measures have been proposed and used, researchers have reported low internal consistency for these measures as well (e.g. LePine and Van Dyne 1998).

Given these poor metrics for previously used voice measures, we decided to measure the voice response by coding open-ended responses to the question, ‘What are you likely to do?’ (see Supplementary/Appendix A for additional details about coding open-ended responses for the voice response). We had six coders rate the strength of each response (from 1 = weak to 7 = strong) and then calculated the median score (see Supplementary/Appendix B for examples of gradations of voice response). We used the median score to aggregate the coder ratings rather than the mean or the mode because prior research has found that the median score is more robust and accurate when multiple raters’ scores are aggregated (Garcin, Faltings, and Jurca 2009). For the six coders, the single measure’s intraclass correlation coefficient was .70 and the average measure’s intraclass correlation coefficient was .93; both scores indicated good intercoder reliability.

We used both explicit and implicit measures of gender authority bias because recent evidence indicates that implicit measures are less susceptible to motivated measurement distortion (Steffens 2004). We measured explicit attitudes towards gender and authority using the Gender and Authority Measure (GAM), a 15-item self-report instrument (Rudman and Kilianski 2000). Respondents indicated their ‘preference for male versus female authorities in five areas of social influence’ (Rudman and Kilianski 2000, 1318) – legitimate, expert, reward, coercive, and referent – which are grounded in French and Raven (1959) study of social power. As in Rudman and Kilianski (2000) study, respondents in our study indicated their agreement with each item on a scale from 1 (strongly disagree) to 5 (strongly agree). The final GAM score was the average of the scores for the 15 items, with relevant items reverse-scored (α = 0.76). A high GAM score reflects a preference for males in positions of authority over females in positions of authority.

We measured implicit gender and authority biases with an implicit association test (IAT) modelled after those used by Greenwald, Nosek, and Banaji (2003) and Rudman and Kilianski (2000) (see Supplementary/Appendix C for additional details regarding the IAT design). Like the GAM, a positive score on the IAT measure reflects an implicit bias in favour of men in positions of high authority while a negative score reflects an implicit preference for women in positions of high authority. The mean score for the gender authority IAT was 0.11. This is within the range of mean IAT scores reported in Marvel (2016), which surveyed studies that have used IATs to measure implicit attitudes/biases. The correlation between the two measures of gender authority bias (implicit bias [IAT] and explicit bias [GAM]) is r(225) = .25, p < 0.001, which is similar to the positive correlation of .21 reported in Rudman and Kilianski (2000, 1324). Table 1 presents descriptive statistics and bivariate correlations for the study variables.

Table 1. Descriptive statistics for the study sample

	Variables	Mean	SD	1		2	3		4
1	Exit response	4.32	2.07
2	Voice response	5.03	1.80	0.22	*
3	GAM (explicit gender authority bias)	2.72	0.46	−0.07		−0.12
4	IAT (implicit gender authority bias)	0.11	0.34	−0.02		−0.09	0.25	*
5	Followers’ gender (1 = female; 0 = male)	0.74	0.44	−0.01		−0.06	−0.15	*	−0.19	*

n = 245

* p < 0.05

Manipulation checks

We performed an indirect manipulation check (Trevino and Youngblood 1990; Hauser, Ellsworth, and Gonzalez 2018) by conducting a content analysis of subjects’ responses to the open-ended question ‘What are you likely to do? Briefly explain why’. We scored responses as 1 for bribery if the respondent used terms like ‘under-the-table payment’, ‘bribes’, ‘conflict of interest’, ‘money’, or ‘gifts’, and 0 otherwise. Similarly, we scored responses as 1 for workplace harassment if they included terms like ‘racist’, ‘sexist’, ‘slacker’, or ‘hostile working environment’, and 0 otherwise. Only 6% of respondents stated that they would do nothing, which suggests that the vignettes portrayed significant unethical behaviours that would prompt most individuals to consider taking action, avoiding a ceiling effect. In addition, 69% of respondents in the bribery treatment group and 63% of respondents in the workplace harassment treatment group correctly identified the treatment condition in their responses even though we did not explicitly ask them to do so.

Results

Hypothesis 1 posited that followers are more likely to choose a fight (voice) response in the wake of leaders’ unethical behaviour when their leader is female than when their leader is male. Hypothesis 2 suggested the converse, arguing that followers will choose a ‘flight’ response to unethical behaviour when the leader is male compared with a female leader. To test Hypotheses 1 and 2, we conducted one-way ANOVA tests on the data. Table 2 presents the results, which are reported for one-tailed hypothesis testing.

Table 2. Follower Responses to Unethical Leader Behaviour Based on Leader Gender

		Fight (Voice)				Flight (Exit)
		Descriptive statistics
		n	M	SD	SE	n	M	SD	SE
Bribery	Female leader	66	4.00	1.66	0.20	66	4.13	2.14	0.26
	Male leader	63	4.04	1.93	0.24	63	3.95	2.00	0.25
		Summary of ANOVA
		df	MS	F	p	df	MS	F	p
Between groups		1	0.05	0.02	.450	1	1.03	0.24	.313
Within groups		127	3.23			127	4.30
		Descriptive statistics
		n	M	SD	SE	n	M	SD	SE
Workplace Harassment	Female leader	56	3.30	1.67	0.22	56	4.35	2.14	0.29
	Male leader	60	2.80	1.43	0.18	60	4.91	1.90	0.25
		Summary of ANOVA
		df	MS	F	p	df	MS	F	p
Between groups		1	7.087	2.96	.044	1	9.11	2.24	.069
Within groups		114	2.4			114	4.08

As Table 2 indicates, followers do not differ in their likelihood of choosing a fight response to bribery by male and female leaders (F(1, 127) = 0.02, p = 0.45). However, followers are more likely to fight a female than male leader who engaged in workplace harassment (F(1, 114) = 2.96, p = 0.044), as expected. Thus, we have mixed support for Hypothesis 1. Followers also do not differ between male and female leaders in their flight response to bribery, F(1, 127) = 0.24, p = 0.313. By contrast, followers are more likely to flee male leaders in response to workplace harassment (F(1, 114) = 2.24, p = 0.069). Hypothesis 2 also receives mixed support.

Hypotheses 3 and 4 predicted an interaction between followers’ gender and leaders’ gender, while Hypotheses 5 and 6 predict an interaction between gender authority bias and leaders’ gender. We used the SPSS General Linear Model procedure to test these four hypotheses; the results are presented in Tables 3 and 4. Table 3 focuses on fight (voice) responses to both bribery and workplace harassment combined, while Table 4 focuses on flight (exit) responses to bribery and workplace harassment combined. Both tables report results for two measures of gender-authority bias: an implicit measure (i.e. IAT) and an explicit measure (i.e. GAM).

Table 3. Modelling influences on fight (voice) responses to both types of deviant behaviour

		B	t	p	F	df	p	R^2
Overall model					2.12	34,192	0.001	0.27
	Leader gender	−1.42	−0.90	.370
	Follower gender	−0.98	−2.52	.012
	Follower gender X Leader gender	1.69	3.05	.003
	IAT	1.19	1.98	.049
	GAM	0.04	0.09	.925
	IAT X Leader gender	−1.68	−2.03	.044
	GAM X Leader gender	0.25	0.47	.642
	Deviant behaviour	−0.91	−3.92	.000
	Age 20–29	0.79	1.27	.205
	Age 30–39	0.12	0.27	.785
	Age 40–49	−0.10	−0.24	.813
	Age 50–59	−0.41	−1.02	.311
	Age 60–69	0^a
	Age 70 and up	0^a
	Edu high school grad	0.14	0.13	.898
	Edu some college credit	0.86	1.01	.316
	Edu vocational training	0.18	0.12	.907
	Edu associate degree	0.19	0.23	.820
	Edu bachelor degree	0.70	0.92	.357
	Edu master degree	0.61	0.82	.416
	Edu professional degree	1.38	1.42	.157
	Edu doctorate degree	0^a
	Race black	−0.27	−0.66	.513
	Race hispanic	1.17	1.44	.151
	Race asian	1.16	0.55	.584
	Race other	1.34	2.60	.010
	Mgmt middle	−0.01	−0.04	.967
	Mgmt team leader	−0.17	−0.33	.745
	Mgmt supervisor	−0.58	−1.38	.168
	Mgmt none	−0.24	−0.75	.457
	Mgmt na	0.23	0.18	.859
	Work state gov	−0.31	−0.32	.749
	Work local gov	0.60	0.71	.479
	Work non-profit	−1.08	−0.69	.491
	Work private sector	−0.53	−0.50	.620
	Work independent consulting	0^a
	Work not working	−0.21	−0.16	.875
	Work other	−0.17	−0.16	.877

^aThis parameter is set to zero because it is redundant.

Table 4. Modelling influences on ‘flight’ exit responses to both types of deviant behaviour

		B	t	p	F	df	p	R^2
Overall model					1.41	34, 192	.079	0.20
	Leader gender	−0.20	−0.10	.919
	Follower gender	0.10	0.20	.841
	Follower gender X Leader gender	−0.58	−0.84	.400
	IAT	−0.93	−1.24	.217
	GAM	−0.18	−0.37	.709
	IAT X Leader gender	1.67	1.61	.109
	GAM X Leader gender	0.09	0.14	.887
	Deviant behaviour	0.35	1.23	.222
	Age 20–29	0.23	0.30	.765
	Age 30–39	−0.13	−0.24	.810
	Age 40–49	−0.09	−0.16	.870
	Age 50–59	0.27	0.53	.598
	Age 60–69	0^a
	Age 70 and up	0^a
	Edu high school grad	−2.71	−1.94	.054
	Edu some college credit	−2.77	−2.60	.010
	Edu vocational training	−0.22	−0.11	.911
	Edu associate degree	−1.61	−1.52	.131
	Edu bachelor degree	−1.71	−1.82	.070
	Edu master degree	−1.19	−1.28	.202
	Edu professional degree	−0.63	−0.52	.607
	Edu doctorate degree	0^a
	Race black	−0.23	−0.45	.652
	Race hispanic	−1.20	−1.19	.237
	Race asian	−5.90	−2.24	.026
	Race other	−0.27	−0.43	.671
	Mgmt middle	0.03	0.07	.945
	Mgmt team leader	1.67	2.56	.011
	Mgmt supervisor	1.28	2.44	.016
	Mgmt none	1.17	2.86	.005
	Mgmt na	−0.05	−0.03	.973
	Work state gov	−0.83	−0.68	.497
	Work local gov	−0.40	−0.38	.704
	Work non-profit	4.19	2.16	.032
	Work private sector	−0.70	−0.53	.599
	Work independent consulting	0^a
	Work not working	−0.50	−0.31	.758
	Work other	0.61	0.45	.654

^aThis parameter is set to zero because it is redundant.

The results reported in Table 3 support the hypotheses regarding interaction effects for fight (voice) effects. As shown in Figure 2, both female followers with female leaders and male followers with male leaders are more likely to fight (voice) than gender incongruent leaders and followers. Thus Hypothesis 3 is supported. Also suggested by this model, female followers are less likely than male followers to choose a fight response, and the type of unethical behaviour makes a difference such that workplace harassment is more likely to induce a fight response than bribery.

Figure 2. Interaction effects of leader and follower gender on fight (Voice) response (Hypothesis 4).

Table 3 also indicates an interaction effect between the IAT and the choice of a fight response, but in the opposite direction expected. As Figure 3 depicts, followers with a stronger preference for male leaders (revealed by higher IAT scores) are more likely to choose a voice response for male versus female leaders. Similarly, as the preference of female leaders increases (revealed by lower IAT scores), followers are more likely to choose a voice response for female versus male leaders. However, the interaction term for Gender and Authority (the explicit bias measure) and leader gender is not statistically significant, B = 0.25, t(192) = 0.47, p = 0.642. Thus, Hypothesis 5 is not supported.

Figure 3. Interaction effects of followers gender-authority IAT bias and leaders gender on voice response (Hypothesis 5).

The results reported in Table 4, however, do not support the notion of an interaction effect between follower gender and leader gender on the likelihood of followers choosing a flight (exit) response. The interaction term for leaders’ gender and followers’ gender is not statistically significant, B = −0.58, t(192) = −0.84, p = 0.40. Similarly, the results provide no empirical support for Hypothesis 6, which states that followers’ gender-authority bias and leaders’ gender have an interacting effect on followers’ exit response. Neither the interaction term for IAT (the implicit bias measure) and leaders’ gender, B = 1.67, t(192) = 1.61, p = 0.109, nor the interaction term for GAM (the explicit bias measure) and leaders’ gender, B = 0.09, t(192) = 0.14, p = 0.887, is statistically significant. Furthermore, the type of unethical leader behaviour – bribery or workplace harassment – has no significantly different effect on the extent of flight responses.

Discussion

This research sought to test the influence of gender on follower responses to unethical leader behaviour, particularly within the public sector. Drawing on a range of theories relating to gender and leadership, we hypothesized that follower responses to unethical leader behaviour would depend on the gender of the leader, the gender of the follower, the interaction of the two, and the follower’s gender authority bias. In a vignette experiment with 245 participants, we examined two types of unethical leader behaviour, workplace harassment and bribery, as well as two types of follower responses, voicing concerns, construed as fight (voice) behaviour, and searching for another job, construed as flight (exit) behaviour (Mayes and Ganster 1988). The results of the experiment suggest that follower responses to unethical leader behaviour appear gendered, but with considerable nuance to the results.

Before discussing the implications of our hypothesis testing, one notable finding is that the type of unethical leader behaviour makes a difference to follower responses. In particular, workplace harassment evoked greater fight responses than did bribery, but did not differ from bribery in the extent of flight response evoked. Furthermore, follower responses to workplace harassment evoked greater fight reactions to female than male leaders, whereas flight responses to bribery did not differ between male and female leaders. What is it about workplace harassment that triggers different reactions than bribery? It could be that workplace harassment is experienced by followers more personally, while bribery is experienced more remotely. Prior research has shown that unethical leader behaviour that occurs at a personal level poses a more direct threat to followers’ resources, including self-esteem, socioeconomic status, continued employment, and sense of mastery than unethical behaviours that transcend the individual (Halbesleben et al. 2014; Hobfoll 1989; Vogel and Mitchell 2017). Thus, unethical leader behaviour that hits closer to home may produce more fight than flight. This is a topic ripe for future research.

Now to the implications of our results. The first and second hypotheses drew on role congruity theory (Eagly and Karau 2002) to predict that the gender of the leader would determine whether followers engaged in a fight (voice) response or a flight (exit) response. In particular, we expected that unethical women leaders would elicit a voice response, while unethical male leaders would elicit an exit response. These expectations were supported, but only when workplace harassment was the unethical leader behaviour in question. In case of workplace harassment by a male leader, followers were more likely to flee than to fight back and seek recourse within the organizational apparatus. In contrast, respondents with unethical female leaders preferred to stay with the organization and utilize internal redress mechanisms. These results lend credence to the notion that because unethical behaviour is agentic in nature, it contradicts the feminine role of communal and nurturing behaviour, thus engendering greater resistance among followers. In this respect, our research results are consistent with those found by Kennedy, McDonnell, and Stephens (2016) as well as Bowles and Gelfand (2010), which found greater punishment for female leaders for ethical violations.

For those concerned with discriminatory attitudes towards women, these findings raise the troubling implication that unethical male leaders will more easily remain in positions of power. Given the prevalence of men in top organizational positions, this scenario highlights the challenges of addressing unethical behaviour at the highest organizational echelons. Another implication of this finding derives from the subjective nature of what counts as unethical leader behaviour. Such a determination has a profound influence on the career trajectories of leaders (Bowles and Gelfand 2010), particularly those who are risk takers and perhaps more vulnerable to making ethical lapses in judgement (Borins 2000). If ethical lapses are perceived and not based in fact, the negative consequences for women’s careers can be catastrophic.

Our third and fourth hypotheses expected follower and leader gender to interact in predicting follower responses to unethical leader behaviour. Specifically, we anticipated that followers of unethical leaders of the same gender would produce more fight (voice) behaviour, while followers of unethical leaders of the opposite gender would produce more flight (exit) behaviour. These expectations were supported by the notion that followers can experience a sense of social betrayal when leaders of their own gender undertake wrongdoing (Restubog et al. 2010). Interestingly, our findings contradict those of an experiment by Bowles and Gelfand (2010), which showed male research participants punishing misbehaving female employees more harshly than misbehaving male employees. These contradictory results might be due to the fact that the experiment by Bowles and Gelfand involved bad behaviour by employees rather than leaders, the latter of whom may be held to a higher standard by followers.

Our fifth and sixth hypotheses anticipated that followers’ gender authority biases would interact with leader gender, such that followers favouring male over female leaders would engage more in fight behaviour against female leaders (the fifth hypothesis) and greater flight behaviour from male leaders (the sixth hypothesis). Neither hypothesis was supported. For the fifth hypothesis, we found that followers that favoured male leaders were more likely to fight unethical male leaders and, conversely, followers that favoured female leaders were more likely to fight unethical female leaders. These relationships were found only for the Implicit Association Test (IAT). The Gender and Authority Measure (GAM), on the other hand, produced no significant interaction effects. Regarding the sixth hypothesis, neither measure of gender authority bias affected the likelihood that a follower would flee (exit) the organization. Social betrayal may again be at work here, as gender-biased followers (towards either male or female leaders) tend to fight workplace harassers of their preferred leader gender. The absence of an interaction effect between gender authority bias and bribery could result from the perception that bribery is such an extreme offence that is impervious to gender bias. Future research should consider replication to test the persistence of this finding.

This study has two significant limitations. First, despite our best efforts to use the most recent measures, some of the study measures should be updated and improved (e.g. the measures of voice response and explicit gender-authority bias). In particular, voice response can be vertical or horizontal (Dowding et al. 2000), but the measures used in the study could not discriminate between these two types of responses. Horizontal voice entails discussing problematic conditions with external peers or co-workers who do not have power over the individual exercising voice. Vertical voice describes interactions with supervisors or external individuals/authorities who may be in a position to exercise power over the individual. Including a measure that identifies different types of voice responses would provide better insight into followers’ responses.

A second limitation is the experimental design, which lacks the external validity of field studies. While we sought to overcome this limitation by enrolling practitioners as survey participants, there is no guarantee that behaviour in the lab translates to behaviour in the field. As stated in the discussion section, further field research in this area could overcome this limitation and expand our understanding of follower responses to unethical leader behaviour.

Conclusion

This study suggests that when male leaders engage in unethical behaviours, followers are reluctant to oppose them and more likely to have a flight response and exit the organization. When female leaders engage in the same behaviours, they receive internal pushback from followers (a fight response). Implicit gender-authority biases partly explain why followers fight back against deviant female leaders and flee from deviant male leaders. With men still constituting the majority of public sector organizational leaders, the results imply that gender norms and biases are obstacles to addressing the average public sector leader’s unethical behaviour.

Disclosure statement

The authors have no financial interest in the research conducted for this paper.

Supplementary material

Supplemental data for this article can be accessed at https://doi.org/10.1080/14719037.2021.2000220.

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

This research was not funded by any external parties.

Notes on contributors

Sheela Pandey

Sheela Pandey is an Assistant Professor of Management in the School of Business Administration at Pennsylvania State University, Harrisburg. Her scholarship focuses on entrepreneurship, leadership and strategic management.

Leisha DeHart-Davis

Leisha DeHart-Davis is Distinguished Term Coates Professor of Public Administration and Government at the University of North Carolina at Chapel Hill School of Government. She studies public sector organizational behavior.

Sanjay Pandey

Sanjay K. Pandey is Shapiro Professor of Public Policy and Public Administration at Trachtenberg School, The George Washington University. Professor Pandey’s scholarship focuses on public management and deals with questions central to leading and managing public organizations.

Sucheta Ahlawat

Sucheta Ahlawa is a Professor at Kean University’s School of Management and Marketing. She studies consumer behaviour, marketing strategy and social entrepreneurship.

Notes

1. This research was approved as exempt by the University of Chapel Hill at North Carolina Office of Human Research Ethics, #16-027. Consent forms were waived given that research participation was anonymous.