Introduction

The emotions that leaders display affect the amount of power they are assumed to have in the eyes of others (Clark 1990; Ridgeway and Johnson 1990). In particular, displays of anger and sadness play a crucial role in power perceptions; leaders displaying anger in negative situations have been shown to be ascribed higher levels of power than those reacting with sadness (Tiedens 2001; Tiedens et al. 2000). Although one might assume, based on these findings, that anger displays are more beneficial for leaders than displays of sadness, recent research has indicated that displaying sadness instead of anger might ultimately lead to better outcomes, such as being judged as more effective (Madera and Smith 2009) and having better leader–follower relationships (Schaubroeck and Shao 2012).

We argue that this seeming discrepancy can be explained by examining the types of power instead of the overall level of power inferred from leaders’ anger compared to sadness displays. Theoretical accounts have widely acknowledged that leaders can make use of different bases for exerting power (e.g., Blau 1964; Pfeffer 1992). According to the most prominent (Sturm and Antonakis 2015) and empirically well-supported (Hinkin and Schriesheim 1989) theoretical model by French and Raven (1959), power can be achieved through five different power bases that result from leaders’ formal position within an organization (i.e., position power) and leaders’ individual characteristics (i.e., personal power; Bass 1960; Yukl and Falbe 1991). Position power includes legitimate power (the perception that one has a legitimate right to prescribe behavior for others), reward power (the perception that one can administer rewards) and coercive power (the perception that one can hand out punishments), whereas personal power includes referent power (the perception that one can make others identify and sympathize with oneself) and expert power (the perception that one possesses valuable knowledge).

It is crucial to distinguish among these five power bases resulting from position and personal power (Yukl and Falbe 1991), because, although all power bases ultimately contribute to the same overall quantity of power, their psychological quality differs substantially (French and Raven 1959). Thus, even though followers’ extent of compliance with requests might seem identical, followers’ reasons for compliance vary considerably with leaders’ perceived power bases. For example, followers may comply with a leader considered as high in coercive power to avoid punishments, whereas they may conform to a leader high in referent power because of identification and attraction. These different psychological processes might also result in opposing follower attitudes and behaviors. Referent power, for instance, positively affects favorable leadership outcomes such as perceived leader effectiveness (Podsakoff and Schriesheim 1985) and follower loyalty (Hinkin and Schriesheim 1989), whereas coercive power exerts a negative influence on the same (Carson et al. 1993; Yukl and Falbe 1991).

Considering these different qualities of power, we aim to examine the specific power bases that followers ascribe to leaders displaying anger versus those displaying sadness in negative work situations (Study 1). We expect these power bases to explain why angry leaders are considered as more powerful than sad ones yet achieve worse leadership outcomes; thus, we further strive to explore their relationship with followers’ perceptions of leader effectiveness and followers’ leader-directed loyalty and deviance (Studies 2 and 3). We derive our hypotheses building on the emotions as social information (EASI) model (Van Kleef 2009, 2014), according to which followers make inferences about leaders’ personality, intentions and characteristics based on their emotional expressions, which, in turn, are related to followers’ leader-directed attitudes and behavior.

By demonstrating which power bases are triggered by leaders’ anger versus sadness displays, the present paper expands the often-cited but relatively sparse literature on negative emotion displays and power (Tiedens 2001; Tiedens et al. 2000). Integrating the theoretical perspective of power bases (French and Raven 1959), this research provides a deeper understanding of the different qualities of leaders’ power and elucidates the foundation on which angry compared to sad leaders’ capacity to influence followers is built.

Furthermore, this paper contributes to the general research on leader emotion displays by introducing perceived power bases as a potential answer to the question of why the association between anger displays and various leadership outcomes might be more negative than the association between displays of sadness and leadership outcomes, despite anger leading to a higher overall level of power. Thereby, we shed light on possible explanatory mechanisms behind these relationships, which remain vastly understudied (Madera and Smith 2009) despite the fact that comprehending them is essential for predicting positive and negative follower behaviors in reaction to leaders’ negative emotion displays (Van Kleef et al. 2009).

Influence of Leaders’ Anger versus Sadness Displays on Perceived Power Bases

Recent theoretical accounts increasingly stress the interpersonal functions that emotions fulfill (Keltner and Haidt 1999). According to the EASI model (Van Kleef 2009) and related frameworks (e.g., Miron-Spektor and Rafaeli 2009), the emotions that individuals display provide valuable social information to observers. Thus, observers use others’ emotions to make inferences about the emotion-displaying person and the current situation, which in turn influence their attitudes and behaviors. With power bases reflecting both the personality of leaders and the situation in which they find themselves (Bass 1960; French and Raven 1959), we expect that followers will also make inferences about leaders’ power bases based on their leaders’ emotion displays.

Following previous research that based inferences about perceptions of others’ general level of power on appraisals (i.e., interpretations of the current situation) and action tendencies (i.e., preferred behaviors) connected with specific emotions (Martorana et al. 2005; Tiedens et al. 2000), we likewise derive followers’ inferences about leaders’ power bases from these components of emotions. Anger and sadness are the distinct negative emotions that most frequently arise in the workplace (Basch and Fisher 2000; Grandey et al. 2002), are experienced and expressed in similar situations (Madera and Smith 2009) and are similarly unpleasant (Russell 1980). However, according to circumplex models of affect (Russell 1980; Russell and Barrett 1999), these emotions result from opposite underlying appraisals (Smith and Lazarus 1993) and cause opposite subsequent action tendencies (Carver and Harmon-Jones 2009; Frijda et al. 1989). With respect to underlying appraisals, anger results from a high perceived coping potential (i.e., the feeling of being able to deal with adverse circumstances), whereas individuals who perceive their coping potential to be low typically experience sadness (Smith and Ellsworth 1985). Concerning action tendencies, experiencing anger results in approaching negative circumstances (Carver and Harmon-Jones 2009), whereas experiencing sadness results in avoidant behaviors (Frijda et al. 1989). In addition to these differing approach/avoidance tendencies, feelings of anger and sadness are also related to action tendencies that are more interpersonal in nature; while angry people regularly aggress against others (Fischer and Roseman 2007), sad people often engage in affiliative, prosocial behaviors (Hess et al. 2000).

Position Power

Due to the differences between anger and sadness displays regarding coping potential, approach motivation and aggressiveness, we expected to find uniformly higher ascriptions of those power bases relating to position power for leaders displaying anger instead of sadness.

Perceived Legitimate Power

High coping potential and high approach motivation are both related to legitimate power positions (Galinsky et al. 2003; Magee and Galinsky 2008). Based on this argument, research has demonstrated that individuals displaying anger are thought to be in legitimate power positions (i.e., to be leaders), whereas those displaying sadness are assumed to occupy lower-ranking positions (Tiedens et al. 2000). In line with this finding, observers might also ascribe differing levels of legitimacy (i.e., the right to influence others’ behavior) to individuals already in leadership positions. We therefore assume that leaders displaying anger will be ascribed higher levels of legitimate power than leaders displaying sadness.

Hypothesis 1a

Leaders are perceived as possessing more legitimate power when displaying anger than when displaying sadness.

Perceived Reward Power

Due to the differential tendencies to aggress (Fischer and Roseman 2007) versus affiliate (Hess et al. 2000), in the moment of their emotion expression angry leaders might be considered as less likely to reward others than sad leaders. However, angry leaders might nevertheless be assumed to possess a higher capability to reward employees (i.e., to have more reward power) than sad leaders. Because of the high coping potential (Smith and Ellsworth 1985) and the approach motivation (Carver and Harmon-Jones 2009) that are connected with anger but not with sadness, leaders showing anger might be perceived as having better access to resources that are necessary to motivate others into changing their behavior, such as incentives and rewards, than leaders displaying sadness. Hence, angry leaders should be ascribed more reward power than sad ones.

Hypothesis 1b

Leaders are perceived as possessing more reward power when displaying anger than when displaying sadness.

Perceived Coercive Power

Likewise, due to the high coping potential (Smith and Ellsworth 1985) and approach motivation (Carver and Harmon-Jones 2009) associated with anger, angry leaders may be seen as more likely than sad leaders to have the capability to coerce. In addition, due to angry people’s tendency toward aggressive behavior (Fischer and Roseman 2007), in contrast to sad people’s tendency to affiliate (Hess et al. 2000), angry leaders may also be seen as more likely to make use of this capability. We thus assume that angry leaders will be ascribed higher levels of coercive power (i.e., a higher capacity to punish employees) than sad ones (see also Gibson and Schroeder 2002; Ragins and Winkel 2011).

Hypothesis 1c

Leaders are perceived as possessing more coercive power when displaying anger than when displaying sadness.

Personal Power

We derived differing hypotheses for referent and expert power as the two elements of personal power. We assume that angry leaders will be attributed less referent power than sad leaders due to differences between anger and sadness with regard to aggression versus affiliation. We did not make specific predictions for expert power because the higher coping potential, higher approach motivation and more aggressive behavior associated with anger instead of sadness led to contradicting expectations.

Perceived Referent Power

In line with anger being connected to aggressive action tendencies (Fischer and Roseman 2007) and sadness being related to affiliative action tendencies (Hess et al. 2000), angry people are considered as low in warmth (Knutson 1996) and likeability (Koning and Van Kleef 2015), whereas sad people are judged as warm, nice and likeable (Hareli and Hess 2010; Madera and Smith 2009). Thus, angry leaders might be considered as less able to make followers feel valued and personally accepted (i.e., as possessing less referent power) than sad leaders.

Hypothesis 1d

Leaders are perceived as possessing less referent power when displaying anger than when displaying sadness.

Perceived Expert Power

Suggesting that anger displays enhance expert power compared with displays of sadness, the high coping potential (Smith and Ellsworth 1985) connected with anger but not with sadness might signal that angry leaders have the necessary capacities—such as experience and knowledge—to address a difficult situation. Moreover, tendencies to approach and avoid were shown to be unrelated to one’s level of competence (Johnson et al. 2013), suggesting that neither displaying anger nor displaying sadness would affect perceptions of expert power. However, the aggressiveness connected with anger displays (Fischer and Roseman 2007) might be interpreted as insecurity about one’s professional abilities (Fast and Chen 2009; Hareli et al. 2013) and therefore as violating the leaders’ expert role (Rafaeli and Sutton 1987), thus indicating that anger displays reduce expert power compared with displays of sadness. We did not formulate a hypothesis for expert power because of these diverging arguments.

We conducted three experimental studies to test our hypotheses and to derive causal claims about the influence of leaders’ anger versus sadness displays on perceived power bases. Furthermore, to increase generalizability, we replicated our results with different samples (students and working adults) and different induction methods, combining the external validity of videos with the internal validity of pictures accompanied by written scenario descriptions.

Study 1

Method

Participants

Study 1 was conducted in the experimental laboratory of a large German university. Participants were recruited via the existing participant pool of the laboratory as well as by distributing flyers and directly approaching students on campus. A total of 116 students (74.10 % male, M age = 21.60 years, SD age = 2.34) participated in the study in exchange for financial compensation.

Design

In line with previous research (Lewis 2000), we used videos as stimulus materials. As we wanted to control for the potential effects of leader gender on perceptions of emotion displays (Brescoll and Uhlmann 2008), we followed previous research in employing both male and female actors as our stimulus leaders (Horberg et al. 2013). Leaders were portrayed by four professional actors (two men and two women), who were dressed in business attire to increase credibility and who were rated with similar perceived age, attractiveness and leader prototypicality in two separate within-subject pretests (N total = 16 students). Participants were randomly assigned to watch one of the resulting eight videos showing a male or female leader expressing either anger or sadness.

Procedure

Upon arrival, participants were seated in separate booths of the laboratory, provided with soundproof headphones and instructed about the general procedure to follow. In line with previous research (Damen et al. 2008), participants then watched two videos on a computer monitor, which were presented as parts of a leader’s end-of-year speech. In this speech, the leader talked about the company’s previous poor financial year (Lewis 2000). After watching the video, participants answered questions about the leader using a standard online survey tool.

Emotion Display Manipulation

After a short general introduction, leaders’ emotion displays were manipulated both verbally and nonverbally. For verbal displays, leaders explicitly stated the emotions they felt with respect to their company’s dissatisfying performance. An example statement is “The operating results have to be judged as being negative,” followed by “…and I am very angry about this” (anger condition) or “…and I am very sad about this” (sadness condition). In addition to this verbal emotion display manipulation, leaders’ nonverbal expressions (tone of voice, facial and body movements) were varied between the angry and sad conditions (see Dael et al. 2012; Ekman and Friesen 1978; Lewis 2000; Marsh et al. 2005; Scherer 1986; Smith and Ellsworth 1985; Wallbott and Scherer 1986). Angry leaders frowned, narrowed their eyes, clenched their fists and used an angry tone of voice, whereas sad leaders stood with their arms hanging loosely, pressed their lower lips forward, raised their inner brows and spoke with a depressed, gloomy and rather slow voice. To enhance external validity, from all recorded material we chose video sequences in which emotion displays were clearly recognizable but sufficiently moderate for a real leader. On average, the videos lasted 1.24 min in the anger condition (SD = 0.06) and 1.30 min in the sadness condition (SD = 0.09). Videos were pretested by 73 students (79.20 % male, M age = 21.40 years, SD age = 2.03, 80.80 % having work experience) to ensure that they indeed conveyed anger and sadness.

Dependent Variables

Participants rated how much the leader possessed each of the five power bases on a scale from 1 (strongly disagree) to 5 (strongly agree). The original items (four items per power base; Hinkin and Schriesheim 1989) were converted from first- to third-person format to fit our design. Sample items include “This leader can make others feel that they have commitments to meet” (legitimate power; α = .83); “This leader can provide others with special benefits” (reward power; α = .69); “This leader can make life difficult for others” (coercive power; α = .84); “This leader can make others feel valued” (referent power; α = .83); and “This leader can provide others with sound job-related advice” (expert power; α = .80).

Manipulation Check

To test whether the emotion display manipulation was successful, we assessed how much participants thought the leader was displaying anger and sadness throughout the video on a six-point scale (1 = not at all, 6 = very much) at the end of the questionnaire.

Control Variable

As outlined above, we initially controlled for leader gender (i.e., the gender of the actor portraying the randomly assigned leader) in our statistical analyses. Because including this covariate did not change our results in either size or direction, results are reported without this control variable.

Results

Analytical Strategy

We conducted separate analyses of variance (ANOVAs) for each power base.

Preliminary Data Analysis

To provide evidence for the validity of our constructs, we performed a confirmatory factor analysis (CFA). To determine model fit, we analyzed relative Chi-square (χ 2/df), for which values should be below 2.00 (Tabachnick and Fidell 2007); the root mean square error of approximation (RMSEA), for which values equal to or smaller than .08 indicate reasonable fit (Browne and Cudeck 1992); and the Comparative Fit Index (CFI), which should be above .90 to be considered as adequate (Bentler and Bonett 1980). Our hypothesized five-factor model satisfactorily fit the data: χ 2 [136] = 243.20, p < .001; χ 2/df = 1.79; RMSEA = .08; CFI = .90. Moreover, our hypothesized five-factor model fit the data better than an alternative two-factor model differentiating only between position and personal power without reflecting the five specific power bases (χ 2 [147] = 370.76, p < .001; χ 2/df = 2.52; RMSEA = .12; CFI = .79; ∆χ 2 [11] = 127.56, p < .001), as well as an alternative one-factor model in which all power bases loaded on a global power factor (χ 2 [141] = 313.89, p < .001; χ 2/df = 2.23; RMSEA = .10; CFI = .84; ∆χ 2 [5] = 70.69, p < .001).

Manipulation Check

The emotion display manipulation significantly influenced perceived leader anger, F(1, 114) = 111.13, p < .001, d = 1.95, with more anger ascribed to angry leaders (M = 5.27, SD = 0.95) than to sad ones (M = 3.15, SD = 1.21). Leader emotion displays also significantly affected perceived leader sadness, F(1, 114) = 95.50, p < .001, d = 1.84, with more sadness ascribed to sad leaders (M = 5.10, SD = 1.19) than to angry ones (M = 2.75, SD = 1.36). Thus, the emotion display manipulation was successful.

Ratings of Leaders’ Power Bases

Table 1 depicts the means, standard deviations and correlations of all variables included in the study.

Table 1 Descriptive statistics, correlations and reliabilities of all variables (Study 1)
Full size table

Position power Hypothesis 1a stated that displays of anger enhance perceived legitimate power compared with displays of sadness. Indeed, angry leaders (M = 3.95, SD = 0.75) were ascribed more legitimate power than sad leaders (M = 3.27, SD = 0.95), F(1, 114) = 18.31, p < .001, d = 0.80. However, contradicting Hypothesis 1b, which predicted that angry leaders would be ascribed more reward power than sad ones, leaders’ emotion displays did not influence their perceived reward power, F(1, 114) < 1, ns, d = 0.15 (M ang = 3.23, SD ang = 0.77; M sad = 3.12, SD sad = 0.71). In line with Hypothesis 1c, angry leaders (M = 3.95, SD = 0.67) were perceived to possess more coercive power than sad leaders (M = 2.89, SD = 0.88), F(1, 114) = 54.16, p < .001, d = 1.36.

Personal power In accordance with Hypothesis 1d, angry leaders (M = 2.68, SD = 0.82) were attributed significantly less referent power than sad ones (M = 3.39, SD = 0.84), F(1, 114) = 21.68, p < .001, d = 0.86. As assumed, emotion displays did not affect leaders’ perceived expert power, F(1, 114) < 1, ns, d = 0.16 (M ang = 2.66, SD ang = 0.79; M sad = 2.79, SD sad = 0.83).

Discussion

In line with our general predictions, Study 1 provides first evidence for the differential effects of anger versus sadness displays on the power bases attributed to leaders. Angry leaders gained two position power bases (i.e., legitimate and coercive power) but lost a personal power base (i.e., referent power). Thus, although our hypothesis on reward power was not supported, in general, leaders expressing anger seem to stress the power that comes with their position while eroding their personal power.

In sum, our results show that examining the different types of power instead of the overall level of power associated with anger versus sadness displays is a valuable approach, as these emotions actually exert differential effects on leaders’ perceived power bases. Nonetheless, the theoretical assumption that these perceived power bases might explain negative leadership outcomes associated with leaders’ anger versus sadness displays still remains to be clarified. Study 2 therefore addresses the indirect effects of leaders’ anger versus sadness displays on leadership outcomes via leaders’ perceived power bases. Specifically, we focused on three leadership outcomes that are likely to be negatively related to leaders’ anger displays: first, leaders’ perceived effectiveness, which was shown to be judged as lower for angry than for sad leaders (Madera and Smith 2009); second, followers’ loyalty toward their leaders, as the relationship quality between leaders and followers was shown to be negatively related to leader anger, but positively to leader sadness displays (Schaubroeck and Shao 2012); third, followers’ deviance toward leaders, as previous research has shown that followers’ social behavior toward leaders is negatively affected by leaders’ anger displays (Johnson and Connelly 2014).

Study 2

Theoretical Extension: Leaders’ Anger versus Sadness Displays, Perceived Power Bases and Leadership Outcomes

In line with the EASI model, we assumed that followers’ inferences about leader power bases would subsequently influence their attitudes and reactions toward the emotion-displaying leader (Van Kleef 2009), thus explaining more negative leadership outcomes (i.e., lower perceived effectiveness, lower follower loyalty and higher leader-directed deviance) for leaders displaying anger instead of sadness. On the one hand, we expected positive indirect effects of leaders’ anger versus sadness displays to emerge on leadership outcomes via those two position power bases resulting from the high coping potential and approach motivation connected with anger but not with sadness (i.e., enhanced legitimate and reward power). On the other hand, we expected these positive indirect effects to be offset by those position and personal power bases resulting from the opposing action tendencies of anger and sadness to aggress versus to affiliate (i.e., enhanced coercive and reduced referent power).

Perceived Leader Effectiveness

The two position power bases of legitimate and reward power have been associated with higher follower performance in previous research (Carson et al. 1993); thus, we assumed that leader anger versus sadness displays would exert positive indirect effects on perceived leader effectiveness via these power bases. However, in line with the finding that aggressiveness does not lead to actual conviction in followers (Schwarzwald et al. 2004), we assumed that these effects would be offset by a negative indirect effect via coercive power, which was shown to be negatively related to leaders’ perceived effectiveness (Hinkin and Schriesheim 1989). Likewise, we expected a negative indirect effect via referent power, as decreased referent power is linked with lower perceived leader performance (Mulder et al. 1986; Ward 1998).

Hypothesis 2a

Compared to displays of sadness, leaders’ displays of anger have a positive indirect effect on perceived leader effectiveness via legitimate and reward power and a negative indirect effect on perceived leader effectiveness via coercive and referent power.

Follower Loyalty

In line with the finding that the position power base of legitimate power is related to enhanced follower support of the leader (Dunne et al. 1978), we expected a positive indirect effect of leader anger versus sadness displays on loyalty to emerge via legitimate power. We did not expect such a positive effect via the position power base of reward power, because reward power might also be viewed as being manipulative and bribing (Carson et al. 1993), potentially explaining why this power base is unrelated to followers’ support of their leader (Podsakoff and Schriesheim 1985). However, we did expect a negative indirect effect on follower loyalty to emerge via the position power base of coercive power, as enhanced coercive power has been shown to be related to lower follower support (Carson et al. 1993) and commitment (Yukl and Falbe 1991). In line with the above, we also expected a negative indirect effect via the personal power base of referent power, as decreased referent power has been found to be associated with less follower commitment (Hinkin and Schriesheim 1989).

Hypothesis 2b

Compared to displays of sadness, leaders’ displays of anger have a positive indirect effect on loyalty toward leaders via legitimate power and a negative indirect effect on loyalty toward leaders via coercive and referent power.

Leader-Directed Deviance

For leader-directed deviance, we did not assume positive indirect effects of anger versus sadness displays via the position power bases of legitimate power, because the formal status of those who express anger does not deter retaliation (Wang et al. 2012), or reward power, because this power base is unrelated to followers’ compliance with the leader (Rahim 1989). However, we expected positive indirect effects on deviance to emerge via the position power base of coercive power and the personal power base of referent power. Enhanced coercive and reduced referent power seem to be related to perceptions of injustice (Mossholder et al. 1998)—a central inverse predictor of deviant behaviors (Skarlicki and Folger 1997).

Hypothesis 2c

Compared to displays of sadness, leaders’ displays of anger have a positive indirect effect on leader-directed deviance via coercive and referent power.

Figure 1 visualizes the theoretical model underlying this paper.

Fig. 1
figure 1

Theoretical model of the effects of leader anger versus sadness displays on perceived leader power bases and leadership outcomes

Full size image

Method

Participants

For Study 2, we recruited working adults who, having daily experience with leaders expressing emotions, are able to anticipate their own reactions toward angry versus sad leaders. The study was conducted in a German shopping center, where potential participants were approached and asked to participate in the study if they indicated that they were employed and had a leader. In total, 129 adults took part in the study in exchange for candy. Two participants had to be excluded from the analysis because, in contrast to their initial statement, they later reported not having a leader or not being employed.Footnote 1 Thus, the final sample consisted of 127 participants (52.40 % male, M age = 34.82 years, SD age = 11.36). Forty-four percent of the participants had a university degree, and 77.20 % worked full-time (compared with part-time) for an average of 12.77 years (SD work = 11.46), primarily in service industries rather than in production (76.40 vs. 23.60 %).

Design

As in Study 1, participants were randomly assigned to one of two emotion display conditions (anger vs. sadness) and were shown a video of either a male or a female leader. We used the same videos as in Study 1.

Procedure

Having agreed to participate in the study, participants watched the videos in separate booths in the shopping center while wearing soundproof headphones. They subsequently answered questions about the leader on a paper-and-pencil questionnaire.

Potential Mediators

Items for legitimate (α = .90), reward (α = .88), coercive (α = .85), referent (α = .91) and expert power (α = .82) were the same as in Study 1.

Dependent Variables

Perceived leader effectiveness was measured with four items (e.g., “This person is successful in his/her company”; α = .91) adapted from Johnson et al. (2008) and Wayne et al. (1997). When rating loyalty and deviance, participants were instructed to imagine that the leader was their own supervisor (for similar procedures, see Lewis 2000; Van Kleef et al. 2010). Loyalty was measured using three items (e.g., “I would feel a strong loyalty toward my leader”; α = .80) adapted from Podsakoff et al. (1990), and deviance was measured using three items (e.g., “I would act rudely toward my leader”; α = .78) adapted from Mitchell and Ambrose (2007). All dependent variables were rated on seven-point scales (1 = strongly disagree, 7 = strongly agree).

Manipulation Check

The manipulation checks were identical to those in Study 1.

Results

Analytical Strategy

We initially conducted univariate analyses of variance. Indirect effects were then analyzed with the PROCESS macro (Hayes 2013), employing a parallel multiple mediator model with bias-corrected bootstrapped confidence intervals (N = 10,000) and indirect effects resulting from the product of the a- and b-paths (Hayes 2013). Leader emotion display was dummy-coded (1 = anger, 0 = sadness). To account for possible indirect effects via power bases not contained in the hypotheses, we included all power bases in our model. Analyses were conducted both without and with leader gender as a control. Including the control variable did not change the results either in size or in direction; thus, results are reported without the control variable.

Preliminary Data Analysis

Our hypothesized model, incorporating both power bases and leadership outcomes, adequately fit the data: χ 2 [354] = 496.90, p < .001; χ 2 / df = 1.40; RMSEA = .06; CFI = .94. Moreover, our hypothesized model fit the data better than did alternative models in which the power base items loaded only on two factors representing position versus personal power (χ 2 [368] = 628.49, p < .001; χ2/df = 1.71; RMSEA = .08; CFI = .89; ∆χ 2 [14] = 131.59, p < .001) or on one overall power factor (χ 2 [364] = 628.87, p < .001; χ2/df = 1.73; RMSEA = .08; CFI = .89; ∆χ 2 [10] = 131.97, p < .001).

Manipulation Check

Leader emotion displays significantly affected perceived leader anger, F(1, 124) = 89.80, p < .001, d = 1.69, with more anger ascribed to angry leaders (M = 3.98, SD = 1.25) than to sad ones (M = 1.59, SD = 1.56). Leader emotion displays also had a significant effect on perceived leader sadness, F(1, 124) = 76.21, p < .001, d = 1.56, with more sadness ascribed to sad leaders (M = 3.28, SD = 1.42) than to angry ones (M = 1.23, SD = 1.20). Thus, the emotion display manipulation was successful.

Direct Effects on Power Bases

Table 2 depicts descriptive statistics and correlations.

Table 2 Descriptive statistics, correlations and reliabilities of all variables (Study 2)
Full size table

Position power For legitimate power, there was a significant effect of the emotion condition, F(1, 125) = 11.73, p < .001, d = 0.62. Angry leaders (M = 3.73, SD = 0.91) were thought to possess more legitimate power than sad leaders (M = 3.14, SD = 1.00), supporting Hypothesis 1a. Supporting Hypothesis 1b, angry leaders were also ascribed more reward power (M = 3.58, SD = 0.82) than sad leaders (M = 3.01, SD = 0.92), F(1, 125) = 13.36, p < .001, d = 0.65. In line with Hypothesis 1c, angry leaders (M = 4.08, SD = 0.71) were thought to have significantly more coercive power than sad leaders (M = 3.39, SD = 1.02), F(1, 125) = 19.53, p < .001, d = 0.79.

Personal power Supporting Hypothesis 1d, angry leaders (M = 2.12, SD = 0.91) were attributed significantly less referent power than sad leaders (M = 2.72, SD = 1.00), F(1, 125) = 12.67, p < .001, d = 0.63. Leader emotion displays did not affect perceptions of expert power, F(1, 125) < 1, ns, d = 0.12 (M ang = 2.37, SD ang = 0.86; M sad = 2.27, SD sad = 0.80).

Indirect Effects on Leadership Outcomes

Table 3 depicts the indirect effects of leaders’ anger versus sadness displays on leadership outcomes.

Table 3 Indirect effects (Study 2)
Full size table

Perceived leader effectiveness Regarding position power, Hypothesis 2a predicted both positive indirect effects via legitimate and reward power and a negative indirect effect via coercive power. In addition, for personal power, it predicted a negative indirect effect via referent power. The results showed a significantly positive indirect effect via legitimate power (a × b = 0.13, 95 % CI [0.01, 0.37]), a nonsignificantly positive indirect effect via reward power (a × b = 0.07, 95 % CI [−0.06, 0.22]) and a nonsignificant indirect effect via coercive power (a × b = 0.06, 95 % CI [−0.12, 0.30]). Moreover, there was indeed a significantly negative indirect effect via referent power (a × b = −0.28, 95 % CI [−0.59, −0.10]). As the positive indirect effect of position (i.e., legitimate) power was offset by the negative indirect effect via personal (i.e., referent) power, Hypothesis 2a was partly supported.

Follower loyalty Regarding position power, Hypothesis 2b predicted both a positive indirect effect via legitimate power and a negative indirect effect via coercive power. Regarding personal power, it additionally predicted a negative indirect effect via referent power. The results indeed revealed a significantly positive indirect effect via legitimate power (a × b = 0.19, 95 % CI [0.07, 0.39]), as well as significantly negative indirect effects via coercive power (a × b = −0.25, 95 % CI [−0.49, −0.09]) and referent power (a × b = −0.20, 95 % CI [−0.43, −0.06]). As the positive effect via the position power base of legitimate power was offset by the negative effects via the position power base of coercive power and the personal power base of referent power, Hypothesis 2b was fully supported.

Leader - directed deviance Hypothesis 2c predicted positive indirect effects via position (i.e., coercive) as well as personal (i.e., referent) power. Results showed that the indirect effects via coercive power (a × b = 0.28, 95 % CI [0.05, 0.63]) and referent power (a × b = 0.19, 95 % CI [0.01, 0.50]) were indeed significantly positive. Thus, Hypothesis 2c was fully supported.

Discussion

The findings from Study 2 provide further evidence for our assumption that leaders’ displays of anger and sadness differentially affect perceived leader power bases. Again, angry leaders were ascribed higher levels of position power (i.e., legitimate, reward, and coercive power), but lower levels of personal power (i.e., referent power), than sad leaders. Furthermore, there was a positive indirect effect on perceived leader effectiveness and followers’ loyalty via the position power base of legitimate power. However, the negative indirect effect via the position power base of coercive power offset this favorable effect for followers’ loyalty, and the negative indirect effect via the personal power base of referent power offset this positive effect via legitimate power for both perceived leader effectiveness and followers’ loyalty. In addition, the enhanced position power base of coercive power and the reduced personal power base of referent power both conveyed positive indirect effects on followers’ leader-directed deviance.

Having demonstrated that perceptions of power bases might explain why angry leaders achieve worse outcomes than sad leaders, in our next study we wanted to assess the robustness and generalizability of these results. We therefore tested our hypotheses using a different kind of manipulation—employing pictures combined with written scenarios—because this approach has been characterized as a purer manipulation of emotion displays than videos (Shariff et al. 2012). Moreover, in Study 3, we also added a neutral control condition in which the leader did not display any emotions. This approach allowed us not only to assess whether anger is a predictor of certain power bases compared with sadness but also to examine whether anger and sadness, compared with no emotion displays, are related to power bases and leadership outcomes.

Study 3

Method

Participants

Study 3 was conducted online. Participants were recruited via forum posts on a large German online business network, as well as via personal contacts. Every potential participant who was currently employed and had a leader at work was deemed eligible for participation in the study. Participants were incentivized with the opportunity to take part in a lottery. Overall, 233 individuals took part in our study, but 58 had to be excluded prior to the analyses because they indicated that they did not have a leader or were not employed, or because they did not provide any employment-related information.Footnote 2 The final sample thus contained 175 participants (55.20 % male, M age = 37.89 years, SD age = 10.87) who had been working in different industries for, on average, 14.79 years (SD work = 11.16) and had mostly (70.20 %) completed university studies.

Design

Participants were randomly assigned to one of three leader emotion conditions (anger vs. sadness vs. no emotion display). Leader gender was again randomly assigned.

Procedure

Participants took part in the experiment via a standard online survey tool, by clicking on the link provided to them. They first read the leaders’ end-of-year speech and then rated the leader on the measures indicated below.

Emotion Display Manipulation

The manipulation of leaders’ emotion displays occurred both verbally, by showing participants the text of a leader’s end-of-year speech, and nonverbally, by adding a picture of the fictitious leader, presumably taken during the speech. The pictures were comparable screenshots of the leaders taken from the videos used in Studies 1 and 2. In the anger and sadness conditions, the pictures were chosen to closely resemble the respective emotional expressions, whereas in the neutral condition, leaders showed neutral facial expressions and no gestures (Lewis 2000). The leaders’ text was slightly abbreviated compared to the version used in Studies 1 and 2. In the neutral condition, the text of the speech was the same as in the other two conditions, except that it conveyed only factual information and made no reference to the leaders’ emotions. All stimulus materials were pretested with 76 participants (52.10 % male, M age = 35.00 years, SD age = 10.59, 94.60 % currently employed) to ensure successful emotion manipulation.

Potential Mediators

As in our previous studies, we measured perceptions of leaders’ legitimate (α = .92), reward (α = .89), coercive (α = .88), referent (α = .93) and expert power (α = .85).

Dependent Variables

Items for measuring perceived leader effectiveness (α = .89), loyalty (α = .84) and deviance toward the leader (α = .75) were the same as in Study 2.

Manipulation Check

Manipulation checks were identical to those in Studies 1 and 2.

Results

Analytical Strategy

We first conducted univariate ANOVAs for each dependent variable to identify potential differences among our three conditions. In the case of significant differences, ANOVAs were followed by Tukey’s post hoc tests to identify the conditions between which these significant differences had emerged. Second, we conducted parallel mediation analysis, both with and without leader gender as a control variable. During mediation analysis, the three emotion conditions were represented by two dummy variables for leader anger (1 = anger, 0 = sadness and neutral) and sadness (1= sadness, 0 = anger and neutral). When making comparisons between two emotion conditions, only participants assigned to those two conditions were included in the analyses. Because the results remained comparable in size and direction when including leader gender as a covariate in the analyses, the results will be reported without this control variable.

Preliminary Data Analysis

Our hypothesized model, incorporating both power bases and leadership outcomes, adequately fit the data: χ 2 [356] = 575.72, p < .001; χ 2/df = 1.62; RMSEA = .06; CFI = .94. Moreover, our hypothesized model fit the data better than did alternative models in which the power base items loaded on two factors, representing position and personal power (χ 2 [372] = 680.71, p < .001; χ 2/df = 1.83; RMSEA = .07; CFI = .92; ∆χ 2 [16] = 104.99, p < .001), or on one overall power factor (χ 2 [369] = 951.32, p < .001; χ2/df = 2.58; RMSEA = .10; CFI = .84; ∆χ 2 [13] = 375.60, p < .001).

Manipulation Check

Leaders’ emotion displays significantly affected perceived leader anger, F(2, 172) = 53.48, p < .001, with more anger being ascribed to angry leaders (M = 5.31, SD = 0.99) than to sad leaders (M = 2.95, SD = 1.35, p < .001, d = 1.99) or to leaders not displaying emotions (M = 3.62, SD = 1.42, p < .001, d = 1.38). Leaders’ emotion displays also significantly affected perceived leader sadness, F(2, 172) = 62.09, p < .001, with more sadness being ascribed to sad leaders (M = 4.63, SD = 1.15) than to angry ones (M = 2.32, SD = 1.12, p < .001, d = 2.04) or to those not displaying emotions (M = 3.02, SD = 1.13, p < .01, d = 1.41). Thus, the emotion display manipulation was successful.

Direct Effects on Power Bases

Table 4 depicts descriptive statistics and correlations.

Table 4 Descriptive statistics, correlations and reliabilities of all variables (Study 3)
Full size table

Position power Leader emotion displays significantly affected perceived legitimate power, F(2, 172) = 10.77, p < .001. Supporting Hypothesis 1a, angry leaders (M = 3.83, SD = 0.82) were viewed as having more legitimate power than sad leaders (M = 3.09, SD = 0.93, p < .001, d = 0.84). Legitimate power ratings did not differ between angry and emotionally neutral leaders (M = 3.52, SD = 0.84, ns, d = 0.37), whereas sad leaders were ascribed significantly less legitimate power (p < .05, d = 0.49) than leaders who did not display emotions. Leader emotion displays also significantly affected perceived reward power, F(2, 172) = 5.45, p < .01. Providing evidence for Hypothesis 1b, angry leaders (M = 3.37, SD = 0.85) were thought to possess more reward power than sad leaders (M = 2.84, SD = 0.86, p < .01, d = 0.62). Compared with leaders who did not display emotions (M = 3.04, SD = 0.89), angry leaders were ascribed marginally more reward power (p < .10, d = 0.38). Reward power ratings did not differ between the neutral and the sadness condition (ns, d = 0.23). Leader emotion displays also significantly affected perceived coercive power, F(2, 171) = 23.06, p < .001. Supporting Hypothesis 1c, angry leaders (M = 4.15, SD = 0.61) were ascribed more coercive power than sad leaders (M = 3.08, SD = 0.95, p < .001, d = 1.34). Compared with leaders who did not display emotions (M = 3.52, SD = 0.94) angry leaders were viewed as having significantly more coercive power (p < .001, d = 0.80), whereas sad leaders were viewed as having significantly less (p < .05, d = 0.47).

Personal power Leader emotion displays significantly influenced perceived referent power, F(2, 171) = 10.29, p < .001. In line with Hypothesis 1d, angry leaders (M = 2.03, SD = 0.94) were ascribed significantly less referent power than sad leaders (M = 2.85, SD = 1.00, p < .001, d = 0.85). Compared with leaders who did not display emotions (M = 2.52, SD = 0.98), angry leaders were ascribed significantly less referent power (p < .05, d = 0.51), whereas referent power ratings for sad and neutral leaders did not differ (ns, d = 0.33). As in Studies 1 and 2, leader emotion displays did not affect perceptions of expert power, F(2, 171) = 1.19, ns, d ang versus sad = 0.26, d ang versus neut = 0.22, d sad versus neut = 0.03 (M ang = 2.35, SD ang = 0.92; M sad = 2.56, SD sad = 0.69; M neut = 2.54, SD neut = 0.81).

Indirect Effects on Leadership Outcomes

Table 5 depicts the indirect effects for all three emotion display condition comparisons. In addition, focal comparisons are highlighted below.

Table 5 Indirect effects (Study 3)
Full size table

Perceived leader effectiveness For position power, Hypothesis 2a predicted both positive indirect effects via legitimate and reward power and a negative indirect effect via coercive power. For personal power, it predicted a negative indirect effect via referent power. Results showed a significantly positive indirect effect via legitimate power (a × b = 0.24, 95 % CI [0.07, 0.51]), a nonsignificantly positive indirect effect via reward power (a × b = 0.11, 95 % CI [−0.01, 0.29]), a significantly negative indirect effect via coercive power (a × b = −0.40, 95 % CI [−0.79, −0.10]) and a nonsignificantly negative indirect effect via referent power (a × b = −0.13, 95 % CI [−0.47, 0.12]). The positive indirect effect via the position power base of legitimate power was offset by the negative indirect effect via the position power base of coercive power; thus, Hypothesis 2a was partly supported. The positive indirect effect via legitimate power was also revealed when comparing anger with neutral displays (a × b = 0.09, 95 % CI [0.00, 0.28]), whereas this indirect effect was significantly negative when comparing sadness with neutral displays (a × b = −0.13, 95 % CI [−0.31, −0.03]). The negative indirect effect via coercive power also emerged when comparing anger with neutral displays (a × b = −0.24, 95 % CI [−0.52, −0.05]), whereas this indirect effect was significantly positive when comparing sadness with neutral displays (a × b = 0.11, 95 % CI [0.01, 0.32]).

Follower loyalty For position power, Hypothesis 2b predicted both a positive indirect effect via legitimate power and a negative indirect effect via coercive power. In addition, it predicted a negative indirect effect via personal (i.e., referent) power. Results showed a significantly positive indirect effect via legitimate power (a × b = 0.23, 95 % CI [0.08, 0.49]), a nonsignificantly negative indirect effect via coercive power (a × b = −0.26, 95 % CI [−0.68, 0.13]) and a significantly negative indirect effect via referent power (a × b = −0.43, 95 % CI [−0.80, −0.18]). Hence, in line with Hypothesis 2b, the positive indirect effect via position (i.e., legitimate) power was offset via the personal power base of referent power. A positive indirect effect via legitimate power (a × b = 0.07, 95 % CI [0.00, 0.22]) and a negative indirect effect via referent power (a × b = −0.18, 95 % CI [−0.39, −0.05]) also emerged when comparing anger and neutral displays. When comparing sadness and neutral displays, the indirect effect via legitimate power was significantly negative (a × b = −0.15, 95 % CI [−0.40, −0.02]), whereas the indirect effect via referent power was insignificantly positive (a × b = 0.17, 95 % CI [−0.02, 0.43]).

Leader - directed deviance Hypothesis 2c posited negative indirect effects via position (i.e., coercive) as well as personal (i.e., referent) power. Results showed a significantly positive indirect effect of anger versus sadness displays via coercive power (a × b = 0.57, 95 % CI [0.24, 1.04]) and a nonsignificantly positive indirect effect via referent power (a × b = 0.07, 95 % CI [−0.20, 0.41]). Hence, in line with Hypothesis 2c, there was a negative indirect effect via the position power base of coercive power. In comparison with the neutral condition, the indirect effects of anger (a × b = 0.22, 95 % CI [−0.03, 0.54]) and sadness displays (a × b = −0.11, 95 % CI [−0.37, 0.00]) via coercive power were not significant.

Discussion

Study 3 provides further evidence for the differential effects of leaders’ anger versus sadness displays on perceptions of their power bases. Again, angry leaders were ascribed more position (i.e., legitimate, reward and coercive) power but less personal (i.e., referent) power than sad leaders. In addition, in line with Study 2, the positive indirect effects on perceived leader effectiveness and followers’ loyalty via the position power base of legitimate power were offset by negative indirect effects via the position power base of coercive power for perceived leader effectiveness and the personal power base of referent power for followers’ loyalty. Furthermore, perceptions of enhanced coercive power were related to enhanced follower deviance.

By incorporating a neutral control condition, the findings of Study 3 also allow us to make conclusions about the absolute and distinct influences of anger and sadness displays on perceptions of leaders’ power bases. Compared with no emotion displays, leaders’ displays of anger led to higher levels of the position power bases of reward and coercive power but lower levels of the personal power base of referent power. Also compared to no emotion displays, leaders’ sadness displays resulted in lower levels of the two position power bases of legitimate and coercive power. These findings indicate that the effects we expected when comparing anger and sadness displays resulted from the unique effects of both emotional expressions relative to a no emotion baseline condition, showing that anger and sadness in and of themselves can be considered as predictors of leader power bases.

General Discussion

We set up this line of research to find out whether perceptions of leaders’ power bases might explain why angry leaders are considered to be more powerful than sad leaders (Tiedens et al. 2000; Tiedens 2001) but still achieve worse leadership outcomes (Madera and Smith 2009; Schaubroeck and Shao 2012). As expected, angry leaders were viewed as possessing higher levels of position (i.e., legitimate, reward and coercive) power but lower levels of personal (i.e., referent) power than sad leaders. Hence, followers seem to perceive leaders displaying anger, in comparison with leaders showing sadness, as more strongly stressing their legitimate position within the organizational hierarchy and the control over punishments and rewards that is available to them. In contrast, followers seem to consider leaders displaying sadness as more strongly appealing to them on a personal level than leaders displaying anger.

Furthermore, while the enhanced position power base of legitimate power seems to be positively related to favorable leadership outcomes, the enhanced position power base of coercive power and the reduced personal power base of referent power are both negatively associated with favorable leadership outcomes. Leaders’ anger versus sadness displays have a positive indirect effect on leaders’ perceived effectiveness via legitimate power, but this effect is offset by likewise emerging negative indirect effects via coercive (Study 3) and referent power (Study 2). Followers’ loyalty shows the same pattern, with a positive indirect effect via legitimate power being offset by negative indirect effects via coercive (Study 2) and referent power (Studies 2 and 3). Finally, anger displays indirectly increase the possibility that followers will show deviant behavior against their leaders via perceptions of enhanced coercive (Studies 2 and 3) and reduced referent power (Study 2). Overall, these findings indicate that perceptions of leaders’ power bases might indeed explain why angry leaders are seen as more powerful than sad leaders, but achieve worse leadership outcomes.

Theoretical Implications

Our findings contribute to the literature on leaders’ emotion displays and power (Tiedens 2001; Tiedens et al. 2000) by demonstrating that followers will infer not only leaders’ general level of power but also the specific power bases they possess based on whether they express anger or sadness in response to negative situations at work. Although leaders gain position power—namely legitimate, reward and coercive power—when displaying anger instead of sadness, displaying sadness leads to higher levels of personal power—namely referent power—than displaying anger. With these findings, we demonstrate that the often-proclaimed association “anger = power, sadness ≠ power” might be too simplistic, providing an important caveat to the assumption that angry leaders will be considered as more powerful than sad ones on all dimensions. In addition, we offer an answer to the current question of how individuals acquire different power bases (Sturm and Antonakis 2015) by outlining the critical roles of anger and sadness displays during negative organizational situations. How leaders acquire different power bases in the eyes of their followers in turn is of essential importance because, although all power bases might eventually lead to the same overall level of power, they nevertheless differ crucially with respect to their psychological quality and therefore the solidness and sustainability of leaders’ influence (French and Raven 1959; Yukl and Falbe 1991).

These quality differences are also reflected in our results on the indirect effects of leader anger versus sadness displays on leadership outcomes. Whereas leaders’ anger versus sadness expressions enhanced position power but lowered personal power, both enhanced coercive power (i.e., position power) and reduced referent power (i.e., personal power) were adversely related to leadership outcomes; via these power bases, leaders’ anger versus sadness displays were negatively associated with perceived leader effectiveness and followers’ loyalty but were positively associated with followers’ deviant intentions toward the leader. Thus, by considering perceptions of these power bases as potential explanatory mechanisms, our studies might help to clarify why previous studies have found more negative effects of leaders’ anger than sadness displays on leadership outcomes (e.g., Madera and Smith 2009; Schaubroeck and Shao 2012), despite angry leaders’ higher levels of overall power (Tiedens et al. 2000; Tiedens 2001).

Our findings also validate current propositions that anger displays sometimes lead to positive outcomes (Lindebaum and Jordan 2012). Indeed, the higher perceived position power base of legitimate power triggered by leaders’ anger (vs. sadness) displays was positively associated with perceived leader effectiveness and follower loyalty. Thus, anger expressions are connected to positive outcomes, because angry leaders are recognized as formal authorities. Stressing one’s position power by displaying anger can hence actually lead to positive effects, which, however, are likely to be offset by the negative effects emerging via enhanced coercive and reduced referent power.

Practical Implications

Our results have valuable implications for practice by showing that subordinates form impressions of leaders when they view their emotion displays in negative work situations. Leaders often believe that they should show anger to make subordinates more compliant and to be seen as effective (Lindebaum and Fielden 2011). Instead, although leaders might benefit from stressing the position power base of legitimate power, anger displays might backfire by causing subordinates to infer that the leader has a lot of coercive power but little referent power. The same is true for sadness, which also comes with specific costs (lower legitimate power) and benefits (decreased coercive and increased referent power). In summary, these findings indicate that it is very important for leaders to consciously reflect on the emotions they display.

In line with this argument, we believe that our findings might also inform leader selection and development; that is, applicants for leadership positions might be selected based on their ability to handle difficulties in an emotionally adequate way. To do so, companies might screen for emotional intelligence, which allows the suitable expression of emotions and the ability to handle emotionally charged situations (George 2000). Leader development should focus on familiarizing leaders with the upsides and downsides of anger and sadness displays in order to enhance their awareness of the potential consequences of negative emotion displays at work.

Limitations and Opportunities for Future Research

Although our study has valuable implications for both theory and practice, we acknowledge some limitations to the present work. Due to the research question that we wanted to answer, we focused on the emotion expressions of anger and sadness. However, it might also be interesting to compare these emotions’ effects to the effects of other negative emotions, such as disappointment or contempt. Anger expressions aim at changing others’ behaviors in order to be able to continue a relationship and might hence be perceived as more prosocial in nature than displays of contempt, which aim at ending a relationship (Fischer and Roseman 2007). In comparison, expressing disappointment communicates that expectations have not been met (Frijda 1986) and might be less harmful than expressing anger with regard to coercive and referent power, while possibly also keeping the other position power bases (legitimate and reward power) at higher levels than displays of sadness. Analyzing the effects of anger in comparison with these other emotions would be valuable, because doing so allows for providing more specific recommendations to leaders, possibly also with respect to emotions that are currently not in their repertoire.

A potential limitation for the generalizability of our results on displays of anger versus sadness is that the specific type of situation in which leaders’ emotion displays occur might affect the inferences that followers make. In line with previous studies on the effects of distinct negative emotions (Lewis 2000; Tiedens 2001), we formulated our scenarios generally to examine basic inferences resulting from leaders’ anger versus sadness displays. However, it is possible that the target of leaders’ emotion displays affects followers’ inferences (Lelieveld et al. 2011); for example, if followers are direct targets of leaders’ anger, this might even more strongly enhance leaders’ coercive power and reduce their referent power. Furthermore, whereas our scenario clearly specified that the leader displayed emotions because the company had performed poorly, when leaders display anger that is unspecified or unjustified this might likewise more strongly increase coercive and reduce referent power. In addition, the context in which emotion displays occur might be crucial, with anger displays possibly leading to more favorable inferences in male-dominated, aggressive contexts such as the construction industry (Lindebaum and Fielden 2011), and sadness displays being more functional in female-dominated, communal contexts such as health care and education (Judge and Livingston 2008). Considering these potential moderators might constitute a valuable avenue for future research.

Although our experimental manipulation allows us to make causal claims about the influence of anger versus sadness displays on power bases, the concurrent measurement of power bases and leadership outcomes does not allow us to provide definitive evidence for our assumed causal sequence (power bases preceding leadership outcomes; Maxwell and Cole 2007). We derived this proposed sequence by closely following the EASI model (Van Kleef 2009), which theorizes that the inferences observers make due to a target’s emotion display are followed by reactions in observers’ attitudes and behaviors. Additionally, power bases have been theoretically assumed (French and Raven 1959) and empirically demonstrated (e.g., Carson et al. 1993; Hinkin and Schriesheim 1989; Yukl and Falbe 1991) to be antecedents of leadership outcomes. Nevertheless, making use of a longitudinal design would allow also for empirically establishing our proposed sequence.

Additionally, a question concerning issues of causality and sequence that has yet to be answered is whether observers ground their direct inferences about leaders’ power bases in their knowledge of the emotions those in power positions typically express. In line with previous research (Martorana et al. 2005; Tiedens et al. 2000), we assumed that followers’ inferences result from the appraisals and action tendencies related to anger and sadness. Nevertheless, researchers on emotion expressions also highlight reverse processes, according to which observers use their knowledge of the emotions individuals typically express in certain positions to make inferences (Hareli and Hess 2010; Van Kleef 2009). Furthermore, Tiedens et al. (2000) explicitly demonstrated both causal directions, with individuals in high-power positions seen as more likely to express anger and angry individuals seen as more likely to be in high-power positions. In accordance with these findings, future research might also demonstrate the theoretical connection between specific power bases and emotion displays, for example by experimentally inducing position power in contrast to personal power to observe resulting emotion expressions. Moreover, directly measuring appraisals and action tendencies as mediators of the relationship between emotion displays and power bases might help to explain the extent to which power bases are directly derived from followers’ knowledge of the emotion expressions of those in power positions.

Finally, because participants did not rate their own but unknown leaders, and because leadership outcomes involved intentions instead of actual behavior, our leadership ratings were hypothetical in nature. Aiming to elicit lifelike inferences in participants, we designed our materials to be as realistic as possible and used professional actors to impersonate leaders in our videos. Moreover, indicating that hypothetical ratings are indeed likely to transfer to the field, previous research has shown that experimental and field studies produce highly similar follower inferences in response to leaders’ emotion expressions (Schaubroeck and Shao 2012) and that behavioral intentions are the best predictors of subsequent actual behavior (Ajzen 1991; Sheeran 2002). Nevertheless, replicating results in the field would also allow for the inclusion of moderators such as the quality of followers’ relationship with their leaders (Dienesch and Liden 1986) and objective leadership outcomes such as productivity (Kaiser et al. 2008). Because the latter have been shown to sometimes deviate from the subjective measures of leaders’ effectiveness used within this study (Lord and Maher 1991; Visser et al. 2013), these examinations would provide valuable indications about our results’ robustness.

Conclusion

Our findings help to solve the apparent discrepancy between the outcomes of anger displays and sadness displays found by previous research. They indicate that the power bases ascribed to angry versus sad leaders can account for the observation that angry leaders are seen as more powerful but achieve worse outcomes than sad leaders. In consequence, although angry leaders might be considered as more powerful in general, their resulting power seems to rest upon a weak foundation.