Leadership During Crisis: An Examination of Supervisory Leadership Behavior and Gender During COVID-19.

Due to major work disruptions caused by the coronavirus (COVID-19) pandemic, supervisors in organizations are facing leadership challenges as they attempt to manage "work from home" arrangements, the health and safety of essential workers, and workforce reductions. Accordingly, the present research seeks to understand what types of leadership employees think is most important for supervisors to exhibit when managing these crisis-related contexts and, in light of assertions that women may be better leaders during times of crisis, examines gender differences in how male and female supervisors act and how subordinates perceive and evaluate them in real (Study 1) and hypothetical (Study 2) settings. Results indicate that communal leader behaviors were more important to employees in all three crisis contexts. In Study 1, communality was a stronger predictor than agency of supervisor likability and competence. In Study 2, communality was also more positively related to likability, but agency and communality were equally predictive of competence ratings. Ratings of real supervisors suggest that women were not more communal than men when managing these crises, nor did perceptions of leader behavior differ by supervisor gender in a controlled experiment. However, evaluations of women's competence were more directly related to their display of communal behaviors than were evaluations of male supervisors. This research is helpful practically in understanding effective supervisory leadership during the COVID-19 crisis and contributes to the literature on gender and leadership in crisis contexts by attempting to disentangle gender differences in leader behaviors, perceptions, and evaluations.

Since the onset of the novel coronavirus disease 2019 (COVID-19) pandemic, the work lives of many have been significantly altered. Millions of workers were furloughed or laid off due to souring economic conditions (U.S., 2020), and many of those who remain employed as “nonessential workers” scrambled to transition to virtual work arrangements as mandated by local and state governments. Meanwhile, “essential workers” have continued reporting to work and risk exposure to the virus daily. On top of work-related challenges, individuals are also struggling to cope with personal life issues brought on by the pandemic such as loss, anxiety, and social isolation (UCSF, 2020). This unprecedented situation undoubtedly makes it more difficult for employees to continue carrying out work-related duties as they did before the pandemic.

Supervisors, whom employees view as an important source of support at work (Eisenberger et al., 2002; Kottke & Sharafinski, 1988), face a massive challenge in both supporting their subordinates during these difficult times while also continuing to encourage productivity and goal achievement. However, little is known about what constitutes effective supervisory leadership when managing crisis-related activities. Several voices in the popular press (e.g., North, 2020; Taub, 2020) have asserted that women may be more apt at providing effective crisis leadership by pointing to the pattern that nearly all countries being lauded for their response to the pandemic have women as heads of state. Further, research on the “glass cliff” suggests that in times of crisis, women are chosen as leaders because of stereotypically communal traits (Ryan & Haslam, 2007; Ryan et al., 2011), although a recent meta-analysis found the effect to be small and dependent upon many factors (Morgenroth et al., 2020).

The present paper has two primary goals: (1) to investigate what employees in three different COVID-19 induced work situations expect of their supervisors and how they perceive and evaluate leadership behaviors in the crisis and (2) to test for gender differences in how supervisors are perceived and evaluated by their subordinates, from both employee descriptions and via an experimental study controlling for behaviors in COVID-19 work contexts. To investigate leader behavior, we utilize the classic agency/communality paradigm. Agentic behaviors are considered to be achievement-oriented (e.g., independence and assertiveness) and communal behaviors are relationship-oriented (e.g., warmth and supportiveness; Fiske et al., 2002).

Supervisory Leadership During a Crisis

Working individuals impacted by the COVID-19 pandemic generally fall into one of three groups: those suddenly forced to work remotely (i.e., working from home) but did not previously, those continuing to work outside their homes as essential workers, and those laid off or furloughed. To date, most research on leadership during crises has focused on organization-specific crises (e.g., accidents and scandals; Marcus & Goodman, 1991; Wooten & James, 2008) or times of war or economic downturn (e.g., Cohen, 2012). While these contexts are similar to the present crisis in that they induce anxiety and often lead to workforce reductions, they posed different struggles for employees than the COVID-19 crisis. Furthermore, responding to organization-specific crises largely falls on executive-level leaders. While top leaders chart the course for organizations to navigate crises, supervisory leadership is also critically important as supervisors regularly interact with employees and are tasked with implementing crisis-management strategies directed from above (Wooten & James, 2008). Thus, an investigation is required to better understand what effective supervisory leadership looks like and how different leader behaviors are perceived by subordinates within three crisis contexts: managing work from home arrangements, managing layoffs/furloughs, and managing essential workers.

Evidence suggests the single best predictor of the success of a telework arrangement is supervisory leadership (Offstein et al., 2010). The most successful supervisors managing telework arrangements display both agentic behaviors (e.g., clear communication and project management) and communal behaviors (e.g., providing social support and displaying sensitivity; Bolletino et al., 1997; Offstein et al., 2010; Snyder, 2012; Taylor & Kavanaugh, 2005). However, the literature on supervising teleworkers describes situations in which subordinates and supervisors both willingly participate in the arrangement. In contrast, many who were suddenly forced to work from home due to COVID-19 might lack an adequate home environment for accomplishing work due to lack of technology, lapses in child care, and other issues. This challenge is further exacerbated by issues specific to the pandemic itself (e.g., anxiety and isolation) that are not present in typical telecommuting contexts. Therefore, while effective supervisors will display agentic, task-oriented behaviors, subordinates might view communal behaviors such as providing flexibility, social support, and sensitivity to work/family balance as more important for supervisors to display than agentic behaviors in this context of disruption.

The severe acute respiratory syndrome epidemic of 2002–2004 illustrated how public health efforts can be hampered when health care workers are infected at a high rate (Campbell, 2005). Research exploring what motivates essential workers to continue working during emergencies yields predictors related directly to supervisory practices, such as providing personal protective equipment (PPE) and enacting policies and procedures that enhance worker safety (Possamai, 2007). Essential workers are more willing to work during public health crises when they receive honest and transparent communications, flexibility, and feel there is a sufficient work safety culture (Possamai, 2007; Qureshi et al., 2013). While both agentic and communal leader behaviors are important to safe workplaces, according to Possamai (2007), a key element of developing a strong workplace safety culture is “listening to workers’ concerns” (p. 25), suggesting that communal behaviors may be seen as particularly important during a time of less safety at work.

Because much of the downsizing due to COVID-19 could be temporary, the manner in which organizations handled layoffs and furloughs is likely consequential in shaping employees’ attitudes upon returning to work. When laid off employees report perceptions of procedural fairness and support during the process, they remain more committed and loyal to the organization (Bies et al., 1993; Brockner et al., 1994; Naumann et al., 1998). While executive leaders ultimately make downsizing-related decisions, supervisors play an important role in delivering the news and managing the fallout. Research on reducing the impact severity of layoffs suggests combining clear, direct, and empathetic communication when announcing layoffs with immediate social support results in individuals maintaining dignity and belief that the organization cares about their well-being (Brockner et al., 1994; Feldman, 1994). One study concluded that high levels of supervisor support can also reduce attrition among layoff survivors (Erickson & Roloff, 2007), and others similarly highlighted the importance of supervisors exhibiting communal behavior during layoffs (e.g., Mansour-Cole & Scott, 1998; Richter et al., 2018).

While it is clear that agency and communion are both important for supervisors to exhibit in crisis situations, the evidence in these contexts taken together suggests relative importance of communal behavior.

H1: Employees in all three crisis-related work situations (i.e., work from home, laid off/furloughed, and essential workers) will view communal behaviors as being more important for supervisors to exhibit than agentic behaviors.

Leadership and Gender

It is also important to examine who is excelling at providing effective leadership during the COVID-19 crisis. One narrative in the popular press is that women might be more adept at leading during the pandemic than men (e.g., North, 2020; Wittenberg-Cox, 2020; Zalis, 2020). Some evidence for this assertion has come from the fact that many countries being lauded for their handling of the pandemic have women heads of state (e.g., New Zealand, Germany, and Taiwan), while men are most often leading countries that have been criticized for their pandemic response (e.g., the United States, Brazil, and Russia; Fioramonti et al., 2020). An analysis of the U.S. governors likewise concluded that states with female governors have experienced fewer deaths than states with male governors (Sergent & Stajkovic, 2020). The popular press narrative has attributed the success of women leaders during the COVID-19 crisis to gender differences in leadership; specifically, that women leaders’ display of communal qualities such as honesty, empathy, compassion, collaboration, and humility is advantageous during this time of crisis. This argument is based on two assumptions. The first is that communal leadership is more effective than agentic leadership when leading during a crisis, an assumption consistent with the evidence summarized earlier that suggests the relative importance of communal behaviors in crisis contexts. While this literature focuses on top-level leadership, it suggests that displays of communal behaviors by supervisors should be more strongly linked to favorable evaluations than agentic behaviors. Indeed, Cole et al. (2006) illustrated the importance of supervisor support to employee reactions during an organizational crisis. Two types of follower evaluations of leaders have been most commonly studied in research on gender and leadership: perceptions of effectiveness or competence and perceptions of likability (Forsyth et al., 1997). The present research considers both outcomes as some evidence indicates evaluations of liking are more closely linked to communal behavior and evaluations of competence to agentic behavior (Wojciszke et al., 2009). However, in the current pandemic-related crisis contexts, we hypothesize the following:

H2: When managing employees in the three pandemic-related crisis contexts, supervisors’ communal behaviors will be more positively related to evaluations of leader competence and likability than agentic behaviors.

The second assumption is that women are exhibiting greater levels of communal leadership than men during the pandemic. Social role theory (Eagly, 1987) suggests that individuals will act in ways that are consistent with their culturally defined gender roles. Men's traditional gender role encourages them to display agentic traits such as assertiveness and self-confidence, while women are expected to engage in communal behaviors such as being kind and showing concern for others (Eagly, 1987). Research supports that prescriptive gendered stereotypes regarding agency and communion have endured into the 21st century (Duehr & Bono, 2006; Eagly et al., 2019; Zehnter et al., 2018).

Women are believed to excel in displaying a more communal and men in an agentic leadership style due to behavioral constraints placed upon them by prescriptive role expectations (Eagly & Karau, 2002). Evaluative penalties, or “backlash” effects, have been found when leaders of both genders engage in counter-role behavior (e.g., Heilman & Wallen, 2010; Rudman, 1998), perhaps conditioning them to remain within these gender role constraints. Evidence for gender differences in leader behavior is decidedly mixed, although research focused specifically at the supervisory level is lacking. One meta-analysis comparing styles of men and women across leadership levels found no differences in task-oriented behaviors and only a small difference in interpersonal-oriented behaviors favoring women (Eagly & Johnson, 1990). Another meta-analysis found that women display more transformational leadership behaviors (e.g., individualized consideration) than men, who engage in more transactional behaviors (e.g., management by exception; Eagly et al., 2003). Studies of top leaders also support the belief that women are more communal leaders (e.g., more caring and relationship-oriented) and men are more agentic (e.g., more forceful and task-oriented; Lyness & Heilman, 2006; Rosette & Tost, 2010).

More specific to crisis leadership, the “glass cliff” phenomenon suggests women are more often chosen to lead organizations during times of crisis (Morgenroth et al., 2020; Ryan & Haslam, 2007; Ryan et al., 2011). However, glass cliff research focuses on leader selection based on gender and often assumes that men and women display gender-stereotypic behaviors (i.e., agentic and communal, respectively) once in the role. That is, this literature has not adequately disentangled gender from stereotypic behaviors in understanding the phenomenon of why women are appointed to leadership positions during precarious times. Views of ideal leaders during crises are associated with the female stereotype but it is unclear whether this is due to this stereotype being synonymous with unsuccessful leadership or the perceived value in communal characteristics associated with women (Ryan et al., 2011). Is it truly “think crisis–think female” or rather, “think crisis–think communality?”

Taken together, this literature on prescriptive stereotypes, leadership style, and leadership in crisis suggests followers will see gender differences in leader behavior, particularly in a crisis context.

H3a: When managing employees in the three pandemic-related crisis contexts, female supervisors will be perceived as engaging in more communal behaviors than men.

H3b: When managing employees in the three pandemic-related crisis contexts, male supervisors will be perceived as engaging in more agentic behaviors than women.

It is important for research on gender and leadership to examine leader behaviors and evaluations in conjunction as there is evidence leader behaviors are sometimes rewarded and penalized differently for men and women leaders. Heilman’s (1983, 2001) lack of fit model and role congruity theory (Eagly & Karau, 2002) offer explanations for this bias by suggesting that the gender role to which women are ascribed is dissimilar from the perceived role of a leader, traditionally viewed as masculine in nature. This incongruence, known as descriptive bias, leads to women being viewed as less competent leaders than men (Eagly & Karau, 2002). Additionally, women who adopt a more masculine leadership style (i.e., counter-role) can face prescriptive bias, which results in being penalized in evaluations despite exhibiting similar behaviors as men (Eagly & Karau, 2002; Grenny & Maxfield, 2015; Wang et al., 2013). While there is also evidence that men can be penalized for engaging in counter-role leader behaviors (Heilman & Wallen, 2010), the support for backlash against women for incongruent behavior is more consistently documented (Rudman, 1998; Rudman & Glick, 2001). Considering this question is especially critical in crisis situations that might require leaders to utilize different leader behaviors than they typically exhibit. Thus, we adopt an expectancy violation approach and hypothesize the following:

H4a: The relationship between perceptions of communality and evaluations of leader competence and likability will be moderated by supervisor gender such that communality will be more positively related to competence and likability for women than men.

H4b: The relationship between perceptions of agency and evaluations of leader competence and likability will be moderated by supervisor gender such that agency will be more positively related to competence and likability for men than for women.

Study 1 Method

Sample

Participants were recruited and compensated via Qualtrics Panels. We purposively sampled 312 adults in the United States during May 2020 in three subgroups impacted by the COVID-19 crisis: full-time workers (i.e., at least 35 hours per week) who normally work outside of their homes but have been forced to work from home during April and/or May 2020 because of COVID-19 restrictions (n = 106), full-time workers who have continued to work outside of their homes due to the essential nature of their work (n = 103), and individuals who had been working full time but were laid off or furloughed due to workforce reductions that occurred between March and May 2020 (n = 103). We sampled for equivalent numbers of men and women (53.2% female; age M = 40.3 years, SD = 12.4 years), and removed any individuals that did not provide full, sensible answers to qualitative questions. One individual was also excluded because they reported their supervisor's gender as “Other.” Responses represented 40 U.S. states, with NY (9.3%), FL (8.0%), and CA (7.1%) most commonly represented. Most (89.1%) had not experienced symptoms or tested positive for COVID-19 themselves, but 17.9% reported that at least one close friend or family member had tested positive and another 10% suspect a close friend or family member had the virus but was not tested. Just over half of the respondents’ supervisors were men (53.5%) and the mean estimated age of supervisors was 47.2 years old (SD = 10.9 years). Subordinate–supervisor relationships had a mean tenure of 5 years (range 1–41 years).

Perceived Agency and Communality

Leader behaviors were measured with two, six-item scales adapted from Scott and Brown (2006; agentic α = 0.79; communal α = 0.96; see the Appendix). Each leader quality was translated to its behavioral form (e.g., dedicated to dedication). The response scale was modified to ask respondents to rate on a 5-point Likert rating scale (1 = never, 5 = very often) how often their direct immediate supervisor has exhibited each leader behavior when managing one of the three crisis-related situations since March 1 (i.e., “work from home” for those who are working from home; “health and safety” for those who are continuing to work outside the home as essential workers; or “workforce reductions” for those who were laid off or furloughed).

Leader Behavior Expectations

Prior to rating their supervisor's behavior, individuals were instructed to rank the top three behaviors they thought were most important for their supervisor to demonstrate when managing the present crisis (i.e., work from home management, health/safety management, or layoff/furlough management) from the same list of agentic and communal behaviors (adapted from Scott & Brown, 2006) .

Qualitative Measures

Respondents were asked to describe something their supervisor has done well and something their supervisor could have done better in managing with respect to their situation (i.e., work from home arrangements, employee health and safety, or layoffs/furloughs). Gender pronouns describing supervisors were masked prior to coding responses. Two coders were trained to code (1) communal and agentic adjectives according to an agentic/communal words dictionary adapted from Pietraszkiewicz et al. (2019; see the Appendix), (2) task-oriented and relations-oriented leadership behaviors as taxonomized by Yukl et al. (2002), and (3) responses from essential workers for mention of provision (or lack thereof) of PPE. Across all variables, inter-rater agreement was 90.82%. The third author served as the tiebreaker to resolve all discrepancies.

Leader Likability

Leader likability was measured using a four-item scale previously used by Turban et al. (1990) and Engle and Lord (1997; α = 0.94). A sample item is “I get along well with my supervisor.” Responses were rated on a 7-point Likert agreement scale (1 = strongly disagree, 7 = strongly agree).

Leader Competence

Leader competence was assessed using a four-item scale of perceived leader competence (α = 0.94). Previous studies have used these items to assess subordinates’ perceptions of supervisor competence (e.g., Fiske et al., 2002; Heflick & Goldenberg, 2009; Rudman & Glick, 1999; Rudman et al., 2012). Each item begins with the stem “I think that my supervisor is…” and concludes with the items “competent,” “capable,” “intelligent,” and “confident,” and responses were rated on a 7-point Likert scale.1

Demographics

Respondent gender, age, state, job title, supervisor gender, supervisor age, supervisor tenure, and questions regarding COVID health status were asked.

Study 1 Results

Hypotheses Tests

Means, standard deviations, and intercorrelations are presented in Table 1. H1 was evaluated using chi-square (χ2) goodness-of-fit tests to determine whether communal and agentic behaviors were equally frequently selected as the first, second, and third most important behaviors when managing the crisis. Frequencies are presented in Table 2. H1 was supported; across groups, employees selected communal behaviors more often than agentic behaviors as their first ranked, χ2(1, N = 308) = 43.69, p < .001, second ranked, χ2(1, N = 306) = 53.54, p < .001, and third ranked, χ2(1, N = 304) = 27.84, p < .001, most important behaviors. Further analysis within each group yielded similar results (Full results are available in the Supplemental material) suggesting that individuals prioritize communality in their expectations of supervisory leadership in all three work contexts.

Table 1.

Study 1: Descriptive Statistics and Bivariate Correlations.

Variable	M (SD)	1	2	3	4	5	6	7	8	9	10	11	12
1. Perceived agency	3.44 (0.83)	(.79)
2. Perceived communality	3.75 (1.09)	.67	.96)
3. Competence	5.60 (1.46)	.59	.67	(.94)
4. Likability	5.16 (1.57)	.53	.73	.80	(.94)
5. Supervisor gender	1.46 (0.50)	.06	.06	−.02	.00	—
6. Virtual management	0.34 (0.47)	.08	.11	.13	.12	−.02	—
7. Health management	0.33 (0.47)	.07	.06	.06	−.00	.05	−.50	—
8. Reduction management	0.33 (0.47)	−.14	−.17	−.19	−.12	−.02	−.50	−.50	—
9. Supervisor tenure	5.02 (5.97)	.11	.09	.06	.09	−.11	−.09	.09	.00	—
10. Supervisor age	47.17 (10.94)	−.04	−.05	−.13	−.11	.07	.00	.07	−.07	.29	—
11. Respondent age	40.31 (12.43)	−.12	−.05	.01	−.06	.02	−.15	.13	.02	.25	.29	—
12. Respondent gender	1.53 (0.50)	−.10	−.05	−.10	−.11	.52	−.00	.00	.00	−.10	.04	−.02	—

Note. N = 311. Bolded values are significant at p < .05. Cronbach's alpha reported on the diagonal. Agency and communality were rated on a 5-point scale, while competence and likability were rated on a 7-point scale. All scale measures are oriented such that a higher mean indicates greater levels. Supervisor gender and respondent gender were dummy coded, 1 = male, 2 = female. Each management condition was dummy coded 1 or 0 as appropriate. Supervisor tenure, supervisor age, and respondent age were indicated in years.

Table 2.

Study 1: Most Important Leader Behaviors—Top 3 Selection Frequency by Group.

Behavior	Percent by group
	Total sample	Work from home	Essential workers	Laid off/furloughed
Honestyb	42.95	33.96	41.75	53.40
Understandingb	42.63	48.11	44.66	34.95
Careb	42.31	44.34	43.69	38.84
Compassionb	36.22	30.19	40.78	37.86
Dedicationa	32.05	37.74	33.98	24.27
Sensitivityb	27.56	27.36	21.36	33.98
Intelligencea	25.32	31.13	24.27	20.39
Determinationa	18.59	20.76	18.45	16.51
Sympathyb	9.30	10.38	6.80	10.68
Aggressivenessa	6.09	2.83	4.85	10.68
Charismaa	5.77	7.55	4.85	4.85
Competitivenessa	5.45	4.72	6.80	4.85

Note. Subscripts denote agentic behaviors (a) and communal behaviors (b). Table reflects the frequency by which each behavior was selected as one of the top 3 ranked behaviors for each respondent; thus, each column sums to 300%.

To evaluate H2 and H4, we used two hierarchical multiple regression models to consider the roles of perceived communality and agency in predicting evaluations of leader competence and likability, respectively. In each model, supervisor age was included in the first step as a control due to its association with the outcome variables (Table 1). Results indicated that communality and agency were both significant predictors of competence (see Table 3), but a comparison of their relative weights showed that communality (RW = 0.29) was a significantly stronger predictor than agency (RW = 0.19; 95% CI [−0.18, −0.02]). In the model predicting likability (Table 4), communality (RW = 0.39) was a significant predictor but agency (RW = 0.14) was not; the difference in relative weights was significant (95% CI [−0.35, −0.17]). Thus, H2 was supported. To test the moderating role of gender (H4a and H4b), two additional steps were added to each regression model. Supervisor gender (Step 3) did not predict competence over and above communality and agency, ΔR2 = 0.003, p = .15. However, the communality–supervisor gender interaction was significant. Simple slopes analysis indicated that perceived communality is more positively related to evaluations of competence for women than men supervisors, supporting H4a (see Figure 1). The agency–gender interaction was not significant. In the model predicting likability (Table 4), supervisor gender did not predict competence over and above communality and agency (ΔR2 = 0.001, p = .33), nor was either interaction term significant. Thus, H4a was partially supported but H4b was not supported.

Table 3.

Study 1: Regression Models Predicting Competence as a Function of Supervisor Gender, (a) Perceived Communality, (b) Perceived Agency, and Their Interactions.

Variable	Competence
	Step 1						Step 2
	b	β	t	p	RW	RS–RW	b	β	t	p	RW	RS–RW
R2 = 0.51; F(6, 304) = 52.02, p < .001
Intercept	5.60						5.60
Supervisor age	−0.02	−0.13	−2.21	.03			−.01	−0.09	−2.17	.03	0.01	2.30
Communality							0.66	0.49	8.91	<.01	0.29	58.94
Agency							0.45	0.25	4.61	<.01	0.19	38.76
Supervisor gender
Communality × supervisor gender
Agency × supervisor gender

Note. Age, communality, and agency centered. Gender dummy coded, 1 = male, 0 = female.

Table 4.

Study 1: Regression Models Predicting Likability as a Function of Supervisor Gender: (a) Perceived Communality and (b) Perceived Agency, and Their Interactions.

Variable	Likability
	Step 1						Step 2
	b	β	t	p	RW	RS–RW	b	β	t	p	RW	RS–RW
R2 = 0.55; F(6,304) = 62.74, p < .001
Intercept	5.17						5.17
Supervisor age	−0.02	−0.11	−2.01	.05			−0.01	−0.08	−1.97	.05	0.01	1.65
Communality							0.99	0.68	13.15	<.01	0.39	72.79
Agency							0.13	0.07	1.28	.20	0.14	25.56
Supervisor gender
Communality × supervisor gender
Agency × supervisor gender

Note. Age, communality, and agency centered. Gender dummy coded, 1 = male, 0 = female.

Figure 1.

Study 1: interaction between perceived communality and supervisor gender on competence.

H3a and H3b were tested using both ratings and coded data to examine whether perceptions of agency and communality differed as a function of supervisor gender. In the test for agency, respondent age and supervisor–subordinate tenure were added as controls due to their significant associations with agency (Table 1). A Multivariate analysis of variance (MANOVA) indicated no significant differences between male and female supervisors in ratings of communality, F(1,309) = 1.07, p = .30, or in ratings of agency, F(1, 307) = 1.87, p = .17 (see Table 5 for means and standard deviations by gender). As another method of testing our hypotheses, subordinates’ descriptions of their manager's behavior (Table 6) were coded for use of agentic and communal terms. T-tests indicated no differences by supervisor gender in the number of agentic or communal words used to describe what their supervisor did well or in what they could have done better (Table 7). Of note, agentic words were used more often than communal overall. Coded task and relations-oriented behaviors (Table 8) were also compared between men and women supervisors. No differences were found in the frequencies in which men and women utilized these eight behaviors in the descriptions of what they did well or what they could have done better (Table 9). Of note, the most frequently mentioned behaviors in both types of descriptions were supporting and clarifying. In all tests, H3a and H3b were not supported. Additionally, there were no differences between men and women supervisors in provision of PPE, χ2(1, N = 103) = 0.02, p = .90, or lack of providing PPE, χ2(1, N = 103) = 0.06, p = .81, to essential workers.

Table 5.

Study 1: Means and SDs for Perceived Communality, Perceived Agency, and Competence by Supervisor Gender.

		Male supervisor	Female supervisor
Communality	M	3.69	3.82
	SD	1.08	1.09
Agency	M	3.40	3.50
	SD	0.86	0.79
Competence	M	5.63	5.57
	SD	1.32	1.61
Likability	M	5.16	5.17
	SD	1.48	1.67

Note. Agency and communality were rated on a 5-point scale, while competence and likability were rated on a 7-point scale. All measures are oriented such that a higher mean indicates greater levels. Means for communality, agency, competence, and likability were not significantly different by gender at p < .05.

Table 6.

Study 1: Agentic and Communal Behavior Exemplars.

Work context	Behavior	Exemplars
Managing work from home arrangements—did well	Agentic	“He did a great job of keeping us informed on a weekly basis of the COVID-19 situation and its affect [sic] on our work place. He implemented weekly ZOOM faculty meetings for updates and questions/answers. He also helped organize the distribution of technology and hot spots to students who needed these.”
	Communal	“Been very understanding and flexible with work duties and responsibilities. She is supportive and fosters a supportive team who steps in when other team members need support.”
Managing work from home arrangements—could have done better	Agentic	“Communicate earlier on. Take a leadership role instead of waiting.”
	Communal	“He does not engage one on one or recognize the difficulties some of us are confronting.”
Managing essential workers—did well	Agentic	“My boss enforced the mandate that all employees wear masks and gloves, that customers must also wear masks when entering the restaurant….”
	Communal	“He/She has been very flexible with childcare. If we are late, he/she is understanding and is not marking it against us which is very helpful since I had to stop our babysitter for now and my husband is still working.”
Managing essential workers—could have done better	Agentic	“He could have made sure on a daily basis that we have all of the PPE needed to do our job safely.”
	Communal	“She could have realized that we were important and essential [sic] workers that needed her support. She could have treated us like humans and not like slaves.”
Managing workforce reductions—did well	Agentic	“…She also has taken the initiative to try to help us find temporary placement and check-in on us every few weeks.”
	Communal	“My supervisor took the time to explain what was going on. That the company was closing because of COVID stay at home orders. And he personally pulled each employee aside, thanking them for their hard work and [sic] assuring [sic] us we'd all return when the company reopened [sic].”
Managing workforce reductions—could have done better	Agentic	“He should listen better, and provided [sic] better info. At a time like that confident leadership is the key. Which he was locking [sic] by a great deal.”
	Communal	“Sympathy and taking care of employees.”

Note. N = 312. Variables were coded as the number of words describing each theme. Exemplars include responses in which one or more agentic or communal words were coded from the text.

Table 7.

Study 1: Coded Agentic and Communal Words by Supervisor Gender.

Behavior	Did well			Could have done better
	Number of words coded		M Difference	Number of words coded		M difference
	Male supervisor	Female supervisor	t	Male supervisor	Female supervisor	t
Agentic	92	84	−0.37	45	26	1.62
Communal	38	36	−0.32	21	12	1.02

Note. NMen = 167, NWomen = 144. Variables were coded as the number of words describing each theme. No mean differences were found at p < .05.

Table 8.

Study 1: Task-Oriented and Relations-Oriented Behavior Exemplars.

Work context	Behavior	Exemplars
Managing work from home arrangements	Planning	“Management created ‘work from home’ kits for each employee, well in advance of Orders to do so, which consisted of laptops, extra screen, keyboard, mouse, all cords, and had all necessary software pre-loaded by the IT dept. Also established a voluntary micro-essential team to handle mail, etc. at the office.”
	Clarifying	“Made clear explanations on how to effectively work from home with the limited technologies available for us and also clearly stating each role as assigned to each individual.”
	Supporting	“She called to check in on how I was doing balancing my personal and work life. It meant a lot because she asked about my family and mental well-being. She knew I was stressed.”
	Empowering	My supervisor has been very flexible. They have allowed us to work from wherever we would like. And we have had autonomy in completing our jobs.
Managing essential workers	Planning	“The supervisor decided to set up equipment for work from home for most of the staff.”
	Clarifying	“Communication about the daily changing requirements of PPE use.”
	Supporting	“Provided face masks early before they were required. They provided hand sanitizer early too.”
	Empowering	“My supervisor allows us to refuse taking patient and employee temps before entering the hospital. We aren't clinical and shouldn't be doing them while we have furloughed nurses.”
Managing workforce reductions	Planning	“They called us individually days before it happened. Once the governor said to close down, we were already prepared on what was to come. Working at a bar, we have ensured everything is ready to go when reopening.”
	Clarifying	“Called each frontline employee directly letting them know what he/she know. Also, checked with those same employees every couple of weeks to check in with them and update them with anything he/she knew and to see if employees had any questions.”
	Supporting	“My supervisor exercised a lot of care and compassion when performing the lay off…It really gives me hope that I will be called back to work and that they actually care about me.”
	Empowering	“Given me space and freedom to work from home. He also has been allowing more personal leave.”

Note. N = 312. Planning and clarifying were the two most common task-oriented behaviors; supporting and empowering were the two most common relations-oriented behaviors. Variables were coded as 1 = theme is present, 0 = not present.

Table 9.

Study 1: Coded Task and Relations-Oriented Behaviors by Supervisor Gender.

Behavior	Did well			Could have done better
	Frequency (%)		Difference	Frequency (%)		Difference
	Men	Women	χ2	Men	Women	χ2
Planninga	10.18	6.94	1.02	11.98	16.67	1.40
Clarifyinga	51.50	40.97	3.44	22.16	27.78	1.31
Monitoringa	8.98	11.11	0.39	3.59	4.17	0.07
Supportingb	55.69	59.72	0.52	50.30	48.61	0.09
Recognizingb	0.59	2.08	1.34	1.20	2.08	0.38
Developingb	0.59	0	0.87	0.59	0.69	0.01
Consultingb	1.20	0.69	0.21	1.20	0.69	0.21
Empoweringb	8.4	6.25	0.51	0	2.08	3.51

Note. NMen = 167, NWomen = 144. Frequency % indicates the number of responses in which the theme was present out of the total number of responses for each group. No χ2 values were significant at p < .05. Subscripts denote task (a) and relations-oriented (b) behaviors.

Exploratory Analyses

(Full results for all exploratory analyses can be found in the Supplemental material) rater gender is a commonly considered moderator in studies of how female leaders are evaluated (e.g., Eagly et al., 1992), so one question is whether male and female participants differ in their perceptions and evaluations of leaders in our study. We reran our regression analyses considering participant gender, supervisor gender, and the interaction of participant and supervisor gender as predictors of competence. Neither participant gender nor the interaction was significant. We then repeated these analyses for predicting perceptions of agency, communality, and likability. These interactions were not significant. We also considered whether work conditions related to variability in how supervisor behaviors were evaluated. Work conditions are related to competence, communality, and likability, but not agency (Tables S5–S7). Simple slopes indicated that ratings for competence and communality were lower for supervisors in the layoff/furlough group than the other two groups. Ratings for likability were lower for supervisors in the layoff/furlough group compared to the work from home group. We also regressed evaluations of competence and likability on the interactions between work conditions and perceptions of agency and communality. The interactions between work conditions and communality significantly predicted competence, such that high ratings of communality led to higher competence evaluations for the work from home group than the other two groups (see Figure S1). However, the interactions between work conditions and agency did not predict competence. No interactions between work conditions and agency or communality were significant for predicting likability.

Study 2 Method

Study 2 was conducted to replicate our hypothesis tests using an experimental design to provide greater control.

Procedure

Vignettes were created to reflect a 2 (male vs. female) × 2 (high vs. low agency) × 2 (high vs. low communality) design of supervisors managing subordinates who have been forced to work remotely from home due to the COVID-19 pandemic. After providing an initial description of the situation and examples of challenges employees might be facing, each vignette provided two paragraphs of behavioral examples of how the fictional supervisor has led during the crisis. Participants were randomly presented one of eight different vignettes (see the Appendix), answered two attention checks, and rated the behaviors, competence, and likability of the supervisor.

Participants were recruited and compensated via Qualtrics Panels to ensure 50% male and female (N = 256; mean age = 46 years; SD = 17.4 years).

Measures

The same measures as in Study 1 were used to assess leader behavior expectations, agency (α = 0.87) and communality (α = 0.96), competence (α = 0.93), and likability (α = 0.97).

Study 2 Results

Hypothesis Tests

Means, standard deviations, and intercorrelations are shown in Table 10. H1 was supported as respondents selected communal behaviors more often than agentic behaviors as their first, χ2(1, N = 256) = 8.27, p < .001, second, χ2(1, N = 256) = 31.64, p < .001, and third ranked, χ2(1, N = 256) = 21.39, p < .01, most important behaviors for supervisors to display when managing the working from home crisis (see Table 11).

Table 10.

Study 2: Descriptive Statistics and Bivariate Correlations.

Variable	M (SD)	1	2	3	4	5	6	7
1. Perceived agency	3.44 (0.96)	(.87)
2. Perceived communality	3.37 (1.23)	.26	(.96)
3. Competence	5.37 (1.35)	.61	.51	(.93)
4. Likability	4.50 (1.76)	.19	.83	.59	(.97)
5. Supervisor gender	1.50 (0.50)	.10	.03	.08	−.00	—
6. Respondent gender	1.49 (0.50)	−.23	−.07	−.11	.01	−.17	—
7. Respondent age	46.27 (17.31)	.17	.00	.09	−.03	.10	−.42	—

Note. N = 249. Bolded values are significant at p < .05. Cronbach's alpha reported on the diagonal. Agentic and communal behaviors were rated on a 5-point scale and competence and likability were rated on a 7-point scale. All scale measures are oriented such that a higher mean indicates greater levels. Supervisor gender and respondent gender were dummy coded, 1 = male, 2 = female. Respondent age was indicated in years.

Table 11.

Study 2: Most Important Leader Behaviors—Top 3 Selection Frequency.

Behavior	Sample N%
Understandingb	50.78
Compassionb	37.11
Careb	35.16
Dedicationa	32.42
Determinationa	31.64
Honestyb	27.73
Sensitivityb	26.95
Intelligencea	24.22
Charismaa	14.45
Sympathyb	13.28
Aggressivenessa	3.52
Competitivenessa	2.73

Note. Subscripts denote agentic behaviors (a) and communal behaviors (b). Table reflects the frequency by which each behavior was selected as one of the top 3 ranked behaviors for each respondent; thus, the column sums to 300%.

H2 and H4 were evaluated using two hierarchical multiple regression models regressing competence and likability, respectively, on perceived communality and agency. In the model predicting competence (Table 12), both communality (RW = 0.20) and agency (RW = 0.30) were significant predictors in Step 1, and relative weights analysis indicated they did not significantly differ in relative strength (95% CI [−0.03, 0.24]). In Step 1 of the model predicting likability (Table 13), communality (RW = .67) was a significant predictor but agency (RW = .02) was not; the difference in relative weights was significant (95% CI [−0.73, −0.54]). As a result, H2 was partially supported in Study 2. For H4a and H4b, neither supervisor gender (Step 2) nor the interaction terms (Step 3) predicted competence over and above perceived communality or agency (see Table 12), suggesting gender is not a moderator in the model predicting competence. In the model predicting likability (Table 13), neither supervisor gender (Step 2) nor the communality–gender interaction (Step 3) was significant. However, the agency–gender interaction was significant. Simple slopes analysis indicated that perceptions of higher agency were associated with a slight increase in likability for the male supervisor but a slight decrease for the female supervisor (Figure 2). Thus, H4 was partially supported.

Table 12.

Study 2: Hierarchical Regression Models Predicting Competence as a Function of Perceived Communality, Perceived Agency, Supervisor Gender, and Their Interactions.

Variable	Competence
	Step 1						Step 2						Step 3
	b	β	T	p	RW	RS–RW	b	β	t	p	RW	RS–RW	b	β	t	p	RW	RS–RW
R2 = 0.51; F(5,244) = 50.39, p < .001
Intercept	5.35						5.37						5.39
Communality	0.42	0.38	8.11	<.01	0.20	39.69	0.42	0.38	8.09	<.01	0.20	39.61	0.33	0.30	4.38	<.01	0.13	26.21
Agency	0.71	0.50	10.76	<.01	0.30	60.31	0.71	0.50	10.66	<.01	0.30	59.78	0.67	0.47	7.45	<.01	0.22	44.05
Supervisor gender							−0.04	−0.02	−.35	.72	<0.01	0.61	−0.04	−0.02	−0.35	.72	<0.01	0.52
Communality × supervisor gender													0.18	0.12	1.78	.08	0.07	13.49
Agency × supervisor gender													0.14	0.06	1.02	.31	0.08	15.73

Note. Communality and agency centered. Gender dummy coded, 1 = male, 0 = female.

Table 13.

Study 2: Hierarchical Regression Models Predicting Likability as a Function of Perceived Communality, Perceived Agency, Supervisor Gender and Their Interactions.

Variable	Likability
	Step 1						Step 2						Step 3
	b	β	t	p	RW	RS–RW	b	β	t	p	RW	RS–RW	b	β	t	p	RW	RS–RW
R2 = 0.70; F(5,243) = 112.95, p < .001
Intercept	4.49						4.44						4.46
Communality	1.20	0.84	22.66	<.01	0.67	97.35	1.20	0.84	22.64	<.01	0.67	97.27	1.18	0.82	15.42	<.01	0.48	69.39
Agency	−0.06	−0.03	−0.89	.38	0.02	2.65	−0.06	−0.03	−0.81	.42	0.02	2.66	−0.21	−0.11	−2.27	.02	0.01	1.95
Supervisor gender							0.09	0.03	0.70	.49	<0.01	0.07	0.09	0.03	0.71	.48	<0.01	0.07
Communality × supervisor gender													0.07	0.03	0.64	.53	0.19	27.11
Agency × supervisor gender													0.36	0.13	2.66	<.01	0.01	1.49

Note. Communality and agency centered. Gender dummy coded, 1 = male, 0 = female.

Figure 2.

Study 2: interaction between supervisor gender and agency on likability.

H3a and H3b examined whether perceptions of leader behavior differed as a function of supervisor gender. A MANOVA indicated no significant differences in how participants rated the male and female supervisor on communality, F(1,254) = 0.21, MSE = 0.32, p = .65, nor on agency, F(1,254) = 3.16, MSE = 2.88, p = .08. Group means are presented in Table 14.

Table 14.

Study 2: Means and SDs for Perceived Communality, Perceived Agency, Competence, and Likability by Supervisor Gender.

		Men	Women
Communality	M	3.33	3.41
	SD	1.23	1.23
Agency	M	3.34	3.54
	SD	0.87	1.03
Competence	M	5.26	5.48
	SD	1.33	1.37
Likability	M	4.50	4.49
	SD	1.80	1.74

Note. Agency and communality were rated on a 5-point scale, while competence and likability were rated on a 7-point scale. All measures are oriented such that a higher mean indicates greater levels. Means for communality, agency, competence, and likability were not significantly different by gender at p < .05.

As in Study 1, neither participant gender nor the interaction between participant and supervisor gender predicted competence, communality, or likability ratings. However, participant gender was a significant predictor of agency perceptions such that men perceived higher levels of agentic behaviors than did women (Table S2). Further, the participant–supervisor gender interaction indicated a cross-sex bias in perceptions in which men perceive higher levels of agentic behaviors from the female supervisor than the male and women perceive higher levels of agency from the male supervisor than the female (Figure S2).

Discussion

While the popular narrative focuses on women as better leaders in crises, our studies show that at the level of supervisory leadership employees desire communality regardless of the gender of the actor, and there were no gender differences in the display of agentic or communal behaviors, whether assessed by ratings of actual supervisors, use of adjectives in describing those supervisors, coding of supervisor behavior, or in the perceptions of supervisors when the behaviors were manipulated. Thus, this research does not suggest that male and female supervisors act differently in crisis contexts. One caveat emerged in a Study 2 exploratory analysis where evidence indicated a cross-sex rater bias in perceptions of supervisor agency (i.e., female participants perceived the male supervisor as more agentic and vice versa).

In both a hypothetical crisis situation and reality, communality was more indicative of supervisor likability, in line with prior research on communality and likability (Wojciszke et al., 2009). Communality was a stronger predictor of competence evaluations than agency in Study 1 but not in Study 2; however, both agency and communality were predictive of competence in each study. It is possible that when imagining a crisis situation, we believe competent leadership is marked by both communality and agency, but when evaluating actual supervisors in real crisis situations we discover communality is more important. It also may be that those performing a job they are familiar with are less concerned with concrete guidance (Study 1) than those considering a hypothetical context of a job they do not hold (Study 2).

In actual work contexts, evaluations of women's competence were more directly related to their display of communal behaviors than were evaluations of male supervisor competence among employees describing their supervisors; differentiation by gender in competence ratings did not occur when individuals were considering a hypothetical situation. That is, while individuals may not evaluate male and female leader behaviors differently in the abstract, they do in reality; women who do not display communality may be penalized in competence evaluations more than men who show less of these behaviors, suggesting a double standard for women. Additionally, Study 2 results indicated agentic leader behaviors differed in their relation to leader likability as a function of supervisor gender; perceived agency was positively related to likability for the male supervisor but negatively related to likability when the supervisor was female, in line with research on backlash effects for women displaying counter-role stereotypic behavior (Rudman, 1998; Rudman & Glick, 2001). While there were some differences across the actual and hypothetical contexts, both hint at potential greater penalties for female than male leaders in evaluations by followers.

Limitations and Implications

The use of two studies enabled us to consider the same questions in actual supervisory interactions and in a controlled, artificial context. Our use of multiple measures of agency and communality in Study 1 yielded consistent conclusions and lends credence to findings, albeit other operationalizations (and the use of other dictionaries) that might capture a fuller range of leadership behavior should be explored. We were limited in types of pandemic-related workplace contexts we could examine; comparisons to those in contexts less affected (e.g., no change to work situation) could be informative. We included those laid off because they represented a large segment of the U.S. workforce early in the pandemic (Morath, 2020) and these situations were presumed initially to be temporary; however, they would have less supervisory interaction than other workers from which to make their evaluations. Future research should also consider conditions of absence of communality (agency) rather than just low levels of these behaviors so as to better understand what drives leader perceptions.

While we were unable to directly compare leader behaviors before and during the pandemic, we note that the prepandemic literature shows the persistence of prescriptive stereotypes of women as more communal and men as more agentic (e.g., Duehr & Bono, 2006; Eagly et al., 2019; Zehnter et al., 2018) despite little differences in behavior (Eagly & Johnson, 1990) as well as backlash toward agentic women (Rudman, 1998; Rudman & Glick, 2001), in line with findings here. This may seem contrary to the suggestions of the glass cliff literature regarding a preference for women as leaders in crisis; however, glass cliff studies often do not disentangle agency and communality from leader sex, focus more on choosing a leader than how the leader acts once in the role, and are primarily about top-level leaders. Indeed, both agency and communality have been found in this literature to be considered important, particularly in crises where there is high uncertainty (Post et al., 2019; Ryan et al., 2011). In terms of theoretical implications, research often confounds gender differences in actual behaviors, perceptions of behaviors, and evaluations of effectiveness. By using coded descriptions of actual behaviors as well as measuring perceptions of behaviors and evaluations in Study 1, and by manipulating behavior in Study 2 we provide a more definitive look at the arguments regarding women's leadership superiority in crises.

From a practical standpoint, this research lends insight into how supervisors have managed three situations brought on by the COVID-19 pandemic and how employees in turn have evaluated their leadership efforts. Practical advice for top leadership during COVID-19 (e.g., Kerrissey and Edmondson, 2020) echoes the importance of both agency and communality noted here. Both dimensions were also emphasized in recent empirical work on leadership during the pandemic (Sergent & Stajkovic, 2020) but that study differs from our work in several key ways: a focus on top leaders as opposed to supervisors, no inclusion of the viewpoint of followers, and a focus on leader statements rather than behavior. Our findings in Study 1 that communal leadership practices are most often desired by employees suggest that supervisors need to focus greater energies on engaging in communal behaviors such as honesty, support, care, compassion, sensitivity, and sympathy during the COVID-19 and other crises. Organizations may wish to align performance evaluations of supervisors with the context, engage 360-feedback and other upward evaluation tools to provide greater real-time feedback to help supervisors align support with employee needs, and continue to emphasize attention to gender bias in evaluations.

In conclusion, crisis leadership is not just something performed by those at the top, and leaders at all levels of organizations can have a profound impact on how workers navigate challenging times. Assumptions about the importance of leader gender in a crisis may be faulty; focusing attention on the value of relational in addition to task-oriented leader behavior in challenging as well as normal times while continuing to guard against gender bias in how behaviors are viewed seems to be a more worthwhile endeavor.

Click here for additional data file.

Supplemental material, sj-docx-1-jlo-10.1177_15480518211010761 for Leadership During Crisis: An Examination of Supervisory Leadership Behavior and Gender During COVID-19 by Connor J. Eichenauer, Ann Marie Ryan and Jo M. Alanis in Journal of Leadership & Organizational Studies