The Effects of Leader Power and Status on Subordinate Voice Behavior: The Role of Psychological Safety

Rong Yan¹, Sui Yang², Jiang Jing³
¹College of Business, Shanghai University of Finance and Economics, Shanghai 200433, China
²School of Economics and Management, University of Science and Technology Beijing, Beijing 100083, China
³School of Economics and Management, Beijing University of Posts and Telecommunications, Beijing 100876, China

Abstract

Through two experimental studies (Studies 1 and 2) and a multi-source, multi-time field survey (Study 3), this research investigates how leader power and status influence subordinate voice behavior. The findings reveal that leader power negatively affects voice behavior, whereas leader status positively promotes it. Furthermore, leader status moderates the relationship between leader power and subordinate voice: when leader status is low, the negative effect of leader power on voice is amplified; conversely, when status is high, this negative effect becomes non-significant. This moderating effect of status on power is transmitted through subordinate psychological safety. These results contribute to our understanding of how hierarchical differences in organizations (e.g., power, status) shape subordinate voice behavior.

Keywords: power, status, voice behavior, psychological safety
Classification Code: B849: C93

Introduction

Hierarchy represents one of the most common organizational structures, yet it remains a subject of ongoing controversy (Greer et al., 2018). A particularly salient debate concerns whether hierarchical differences impede employee voice (Morrison, 2011). Because voice constitutes a bottom-up information flow embedded within organizational hierarchies, clarifying how hierarchical differences affect voice is crucial. Existing research presents inconsistent findings: some studies suggest that hierarchical differences hinder voice, as subordinates worry about being ignored or fear negative repercussions from leaders (Oc et al., 2019; Weiss et al., 2017), while others indicate that hierarchy facilitates voice by channeling information upward to leaders who are the primary recipients of such input (King et al., 2009; Reitzig & Maciejovsky, 2015).

We propose that these inconsistent conclusions stem from researchers' failure to distinguish between the bases of hierarchical differences—whether they derive from power (control over resources) or status (degree of respect and admiration). Although interrelated, power and status can produce divergent effects (Blader et al., 2016). Conflating their effects may generate contradictory findings. In practice, power and status can become decoupled. For instance, in dual career ladders, technical experts may lack formal authority over rewards and punishments yet command deep respect for their expertise, while leaders who rely solely on positional power to control others may fail to earn genuine esteem. The question of whether power and status enhance or offset each other's effects on voice remains untested. Thus, discussing hierarchical differences without distinguishing power from status obscures their distinct influences and overlooks their interwoven relationship.

Research in justice, conflict, and leadership has demonstrated that power and status not only exert independent effects but also interact (Anicich et al., 2016; Fast et al., 2012), with some findings suggesting that power's effects depend on status levels (Zheng & van Dijke, 2020). We argue that in superior-subordinate relationships, employees evaluate both leader power and status when assessing voice risks, making integrated judgments that inform their voice decisions. Therefore, the voice literature must disentangle and examine the interactive effects of leader power and status.

Drawing on evolutionary theory of social structures, power and status reflect distinct pathways to hierarchical differentiation through dominance and prestige, respectively. Dominance-based power may cause subordinates to fear resource deprivation and experience anxiety (Kish-Gephart et al., 2009; Milliken et al., 2003), reducing psychological safety when considering voice. In contrast, status reflects prestige derived from subordinate admiration (Yu et al., 2019), prompting leaders to behave more benevolently and inclusively (Blader et al., 2016), which enhances subordinate psychological safety and willingness to voice suggestions. Moreover, prestige-based status can constrain leader behavior, reducing concerns about power abuse and increasing power legitimacy (Magee & Galinsky, 2008; Vial et al., 2016), thereby attenuating power's negative effects on voice.

This research examines how different hierarchical bases (power, status) influence voice behavior, why these effects occur, and how status moderates power's influence. Our contributions are threefold. First, we resolve debates about hierarchy's effects on voice by proposing that dominance-based power inhibits voice while prestige-based status promotes it, and we test whether status moderates power's effects. This clarifies how distinct hierarchical bases and their interplay shape voice. Second, we advance power research by showing that power's effects on voice depend on status. Prior work often conflated power with status or examined power in isolation, overlooking how status constrains power use and provides legitimacy (Henrich & Gil-White, 2001). Identifying status as a boundary condition clarifies power's true effects. Third, we contribute to leadership research by examining how soft influence (status/prestige) operates alongside hard power. While scholars have emphasized control and coercion, less attention has been paid to how status shapes power's effects in organizational settings (H. J. Li et al., 2016). Our work illuminates how these two fundamental sources of leader influence affect subordinate perceptions (psychological safety) and proactive behaviors (voice).

1.1 The Distinction and Relationship Between Power and Status

While some research overlooks differences between power and status, others highlight their distinct characteristics (Galinsky et al., 2015; Magee & Galinsky, 2008; To et al., 2020; Yu et al., 2019). First, their foundations differ: evolutionary theory posits that power is dominance-based whereas status is prestige-based (Cheng et al., 2013; Henrich & Gil-White, 2001). Power refers to the ability to control valuable resources and influence others through rewards and punishments (Magee & Galinsky, 2008). Status reflects the degree to which an individual is respected and admired by others in the organization (Blader & Yu, 2017). Dominance and prestige represent two fundamental pathways to hierarchical differentiation, with power and status as their respective manifestations (Case et al., 2018; Cheng et al., 2013).

Second, their sources differ. Power typically derives from formal positions, while status represents a social perception of prestige granted by others (Anderson & Brion, 2014), reflecting informal hierarchical differences. Third, they affect interpersonal dependence differently. Powerholders control resources and enjoy independence, whereas high-status individuals, dependent on others' evaluations, must attend to subordinates' perspectives to maintain their standing (Blader et al., 2016). Fourth, status depends more heavily on legitimacy than power. Power can exist without others' consent and thus may lack legitimacy, whereas status's fairness and reasonableness are critical to its maintenance (Hays, 2013). Fifth, power hierarchies are more stable than status hierarchies, as wresting resource control from powerholders is difficult, whereas status, being subjectively perceived, fluctuates more readily with perceived contributions and competence (Hays & Bendersky, 2015).

Although power and status sometimes correlate positively (To et al., 2020), as both command attention and constitute important influence sources (Cheng et al., 2013; Fragale et al., 2011), they are not always aligned. First, their acquisition paths differ. Power can be obtained through ability and contribution but also through connections and political maneuvering (To et al., 2020), whereas status derives primarily from competence and team contribution, making it difficult to acquire through politics alone (Blader & Yu, 2017; Fiske et al., 2007; Swencionis & Fiske, 2016). Second, high power can produce negative consequences, such as increased self-interest, unethical behavior, and abusive supervision (Gruenfeld et al., 2008; Y. Liu et al., 2020; Rus et al., 2012; Wisse & Sleebos, 2016), which undermine respect and admiration (Willer et al., 2012). Thus, high power does not guarantee high status, and the two can diverge (Cheng et al., 2013; Fragale et al., 2011).

1.2 Leader Power and Subordinate Voice Behavior

We propose that leader power inhibits subordinate voice. According to evolutionary theory, high-power leaders control both positive (rewards, promotion opportunities) and negative (undesirable tasks) resources (Henrich & Gil-White, 2001; McClanahan, 2020). Voice may directly impact subordinates' daily work, career development, and incentives, creating concerns about negative consequences and discouraging candid expression of opinions, dissatisfaction, or concerns. Voice challenges the status quo and may express disapproval of current workflows, potentially creating conflict, embarrassing leaders (Duan & Ling, 2011), and triggering resistance or dissatisfaction (Morrison, 2014). Voice risks may include negative performance evaluations (Burris, 2012), undesirable assignments, or even termination (Grant, 2013; Morrison, 2014). When leader power is high, these risks become more salient, suppressing voice.

Additionally, high-power leaders enjoy greater behavioral freedom and face fewer social constraints, potentially displaying aggressive or unfriendly behaviors that inhibit voice. Compared to low-power leaders, high-power leaders are less dependent on others and less constrained by social norms (Galinsky et al., 2008). They show less concern for subordinates' feelings and face, often dominating conversations (Tost et al., 2013), lacking patience, using direct rather than tactful communication, and interrupting frequently, whereas low-power individuals behave more politely (Hall et al., 2005). Such behaviors can make subordinates feel unsafe and reluctant to speak up. Therefore, we hypothesize:

Hypothesis 1: Leader power negatively affects subordinate voice behavior.

1.3 Leader Status and Subordinate Voice Behavior

We propose that leader status promotes subordinate voice. Evolutionary theory suggests that subordinates communicate frequently with high-status leaders for two reasons. First, status correlates highly with expertise, competence, and team contribution (Blader & Yu, 2017; Fiske et al., 2007; Swencionis & Fiske, 2016). Since voice involves expressing work-related information, ideas, and opinions (Morrison, 2014), subordinates believe high-status leaders possess the competence to evaluate suggestions appropriately and understand their perspectives, making voice more likely. Indeed, leader expertise positively correlates with voice behavior (Kumar & Mishra, 2017).

Second, high-status leaders receive greater affection, indicating positive leader-subordinate relationships. This admiration fosters identification with the leader, promoting voice (Kumar & Mishra, 2017; W. Liu et al., 2010). Moreover, to maintain high status, leaders must nurture relationships, behaving more benevolently and openly (Yu et al., 2019) and displaying prosocial behaviors (McClanahan, 2020), such as considering subordinates' perspectives (Blader et al., 2016) and providing opportunities for expression (Blader & Chen, 2012). This reduces voice-related concerns, as subordinates feel their input will be properly evaluated and good ideas recognized (Xiang et al., 2019; Zhu & Akhtar, 2019). We hypothesize:

Hypothesis 2: Leader status positively affects subordinate voice behavior.

1.4 The Mediating Role of Psychological Safety

We propose that leader power and status influence voice through psychological safety, defined as the belief that expressing opinions, suggestions, or concerns will not result in punishment, criticism, or unfair treatment (Liang et al., 2012). As noted, high leader power raises concerns about negative voice consequences, manifesting as reduced psychological safety. Power differences may lead subordinates to worry that their input will be poorly evaluated or that speaking up will anger the leader (Nembhard & Edmondson, 2006). High-power leaders' more aggressive, dominant, and impolite behaviors further diminish psychological safety (Morrison & Rothman, 2009), making others feel cold (Fragale et al., 2011), anxious (Bunderson & Reagans, 2011), or fearful (Cheng et al., 2013).

Conversely, high leader status, reflecting trust and admiration, enhances psychological safety. Interacting with liked and trusted leaders makes subordinates feel safer (Podsakoff et al., 1990). Additionally, status-driven leaders' friendliness, openness, and fairness help alleviate voice-related concerns (Detert & Burris, 2007; Milliken et al., 2003). When psychological safety is high, subordinates perceive lower voice costs and risks, becoming more willing to speak candidly; when absent, they remain silent for fear of negative evaluation (Edmondson & Lei, 2014; Milliken et al., 2003). We hypothesize:

Hypothesis 3: Subordinate psychological safety mediates the relationship between leader power (status) and subordinate voice behavior.

1.5 The Moderating Role of Leader Status

We further propose that leader status moderates the negative effect of leader power on voice. Evolutionary theory suggests that status, depending on others' evaluations, constrains leader behavior: actions deviating from subordinate expectations erode respect and affection (Case et al., 2018; Henrich & Gil-White, 2001). Status should therefore constrain power use, inhibiting abuse and self-serving behavior. When leader status is high, subordinates' anxiety and fear about leader power decrease, increasing confidence that power will be exercised fairly and that voice will not trigger retaliation. When status is low, leaders face no such constraints and need not worry that misuse of power will damage their reputation. Subordinates then become more concerned that voice will challenge authority and provoke negative consequences.

Additionally, status confers legitimacy on power. High status means the powerful leader also enjoys respect and admiration, rendering power more legitimate (Halevy et al., 2011; Ji et al., 2019) and reducing subordinate anxiety, thereby mitigating power's negative effects. When status is low, leaders lack support and affection, prompting subordinates to question power legitimacy and fear arbitrary punishment or self-serving authority maintenance. Legitimate power allocation increases trust (van der Toorn et al., 2011) and reduces power's negative voice effects, whereas illegitimate allocation exacerbates them. For example, when subordinates perceive power distribution as unfair, voice decreases (J. Li et al., 2014). Research on extra-role behaviors similarly shows that female leaders struggle to gain high status despite high power, reducing subordinates' citizenship behaviors (Vial et al., 2016). As voice represents a key extra-role behavior, we expect power's negative effects to intensify when status is low. We hypothesize:

Hypothesis 4: Leader status moderates the negative effect of leader power on subordinate voice, such that the negative effect weakens when status is high and strengthens when status is low.

We further propose a mediated moderation model. When leader status is high, power-induced psychological safety concerns diminish, reducing power's negative voice effects. Conversely, when status is low, power-induced insecurity intensifies, amplifying the suppression of voice. We hypothesize:

Hypothesis 5: The moderating effect of leader status on leader power is transmitted to subordinate voice behavior through the mediating role of psychological safety, representing a mediated moderation pattern.

We test these hypotheses using multiple designs (experiments, field survey) and samples (Western, Chinese) to establish internal and external validity and ensure replicability.

Study 1

2.1.1 Sample

Prior to data collection, we used G*Power (Faul et al., 2007) to calculate required sample size. Selecting F-test with medium effect size (f = 0.25; Cohen, 1988), desired power of 0.80, and α = 0.05, we determined a minimum of 128 participants. We ultimately recruited 163 full-time workers from Prolific, compensating each £1.25 (approximately ¥8.77). The sample was 57.67% male, with mean age 31.48 years (SD = 8.90, range 18–59) and average work experience 10.00 years (SD = 8.69); 93.25% were from European and American countries.

2.1.2 Design and Procedure

Study 1 employed a 2 (leader power: high vs. low) × 2 (leader status: high vs. low) between-subjects design. In Part 1, participants assumed the role of "marketing manager assistant" and read descriptions of their immediate supervisor's power and status characteristics. Across conditions, participants read: "The marketing manager has complete (almost no) control over your daily work and monthly performance evaluations. He also can (cannot) determine your pay level and promotion opportunities. Moreover (However), compared to managers at the same level, he is very (not) respected and admired by colleagues, and holds a very high (extremely low) position and prestige in the company."

In Part 2, participants completed an adapted hidden profile task. They read an email from the marketing manager about selecting an athlete for a competition in a U.S. national park. The competition involved photographing sunrise from a mountain peak, analyzing air pollution, photographing 20 rare plant species, and delivering oxygen supplies to hiking camps. The manager noted two finalists: Candidate A, a marathon runner who was strong, energetic, had excellent endurance, and held the state marathon record, with whom the manager had a pleasant conversation; and Candidate B, a fitness instructor who was also physically strong and confident about winning, but whom the manager deemed unsuitable, preferring Candidate A. Participants then read additional background information: "Further investigation reveals that although Candidate A is strong, he knows nothing about plants and animals and has poor time management; Candidate B is a former environmental scientist who won a National Geographic photography award. Based on this information, you realize Candidate B is more suitable." Finally, participants wrote a reply email to the marketing manager expressing their opinions about the candidates and provided demographic information. To simulate realistic work conditions and avoid demand effects, we imposed no length requirements.

2.2.1 Manipulation Checks

We measured perceived leader power using a 5-item scale (sample item: "To what extent does the marketing manager have control over you?"; Cronbach's α = 0.94). High-power (n = 82) and low-power (n = 81) groups differed significantly (M_high = 6.92, SD = 1.76 vs. M_low = 3.22, SD = 1.57), t(161) = 14.18, p < 0.001, Cohen's d = 2.22. Perceived leader status was measured with a 4-item scale (sample item: "To what extent is the marketing manager admired?"; Cronbach's α = 0.96). High-status (n = 83) and low-status (n = 80) groups differed significantly (M_high = 7.07, SD = 2.00 vs. M_low = 3.48, SD = 1.88), t(161) = 11.80, p < 0.001, Cohen's d = 1.85.

2.2.2 Voice Behavior Coding

Two independent coders evaluated participants' emails. First, they reviewed the voice behavior definition: "employees expressing work-related suggestions, ideas, or concerns to their immediate supervisor" (Liang et al., 2012). Second, they examined example emails and a 4-point coding scheme: 1 = "fully supports Candidate A, no additional suggestions"; 2 = "somewhat recommends Candidate B, few suggestions"; 3 = "recommends Candidate B, some suggestions"; 4 = "strongly recommends Candidate B, many suggestions." Third, coders independently rated emails while blind to experimental conditions. Fourth, they discussed discrepancies; unresolved cases were reviewed with the authors. Inter-coder correlation was 0.99.

2.2.3 Hypothesis Testing

We conducted ANOVA with leader power as the independent variable, leader status as moderator, gender and tenure as covariates, and voice behavior as the dependent variable. Results showed a significant main effect of leader power, F(1, 157) = 4.85, p = 0.03, partial η² = 0.03; a non-significant main effect of status, F(1, 157) = 0.82, p = 0.37, partial η² = 0.01; and a significant status × power interaction, F(1, 157) = 4.52, p = 0.04, partial η² = 0.03.

Simple effects analysis revealed that in the low-status condition, leader power negatively affected voice (M_low_power = 3.46, SD = 0.87 vs. M_high_power = 2.86, SD = 1.03), F(1, 157) = 9.14, p = 0.003, partial η² = 0.06. In the high-status condition, power's effect was non-significant (M_low_power = 3.31, SD = 0.91 vs. M_high_power = 3.30, SD = 0.90), F(1, 157) = 0.00, p = 0.96. Hypotheses 1 and 4 were supported.

Figure 1 The Effects of Leader Power and Status on Voice Behavior (Study 1)

Study 2

3.1.1 Sample

Using the same power analysis as Study 1, we recruited 189 full-time workers from Prolific who had supervisory experience and regular interaction with their supervisors, compensating each £0.68 (approximately ¥4.48). The sample was 68.78% male, with mean age 29.45 years (SD = 7.18, range 18–54), average tenure 7.46 years (SD = 7.05), 66.16% holding bachelor's degrees or higher, and 87.83% from European and American countries.

3.1.2 Design and Procedure

Study 2 employed a 2 (leader power: high vs. low) × 2 (leader status: high vs. low) between-subjects design. In Part 1, participants recalled a current or former supervisor with specific power and status characteristics, reading: "He/she has great (little) control over resources you value (e.g., work tasks, performance evaluation, pay, promotion). He/she can (cannot) decide whether to provide or withdraw these resources and can (cannot) punish you. Moreover (However), compared to leaders at the same level, he/she is more (less) respected and admired, and holds higher (lower) prestige and authority in the company."

In Part 2, participants recalled whether they had suggestions or concerns about work procedures while working with this supervisor. In Part 3, they imagined the supervisor asking for improvement suggestions in a routine meeting and reported how they would respond and their psychological safety in that scenario.

3.1.3 Measures

Voice behavior. We focused on "work procedures" for three reasons: (1) to immerse participants quickly, (2) to enhance experimental control by bounding content, and (3) to ensure generalizability across organizations. We adapted Liang et al.'s (2012) scale to focus on work procedures and the supervisor as target, merging redundant items to create a 4-item measure (sample item: "In this meeting, I will proactively make suggestions for improving work procedures"; Cronbach's α = 0.85).

Psychological safety. We adapted Liang et al.'s (2012) 5-item scale with the supervisor as target (sample item: "I feel safe expressing my true opinions to my supervisor in this meeting"; Cronbach's α = 0.90).

3.2.1 Manipulation Checks

Perceived leader power was measured with a 3-item scale (sample item: "He/she has high power over me"; Cronbach's α = 0.80). High-power (n = 95) and low-power (n = 94) groups differed significantly (M_high = 5.39, SD = 1.05 vs. M_low = 4.68, SD = 1.18), t(187) = 4.39, p < 0.001, Cohen's d = 0.64. Perceived leader status used the same 4-item scale as Study 1 (Cronbach's α = 0.90). High-status (n = 94) and low-status (n = 95) groups differed significantly (M_high = 5.34, SD = 1.08 vs. M_low = 4.18, SD = 1.38), t(187) = 6.46, p < 0.001, Cohen's d = 0.94.

3.2.2 Hypothesis Testing

ANOVA results with leader power as independent variable, status as moderator, gender, education, and tenure as covariates, and psychological safety and voice as dependent variables showed significant main effects of power on voice, F(1, 182) = 3.97, p = 0.048, partial η² = 0.02, and on psychological safety, F(1, 182) = 4.35, p = 0.04, partial η² = 0.02. Status also showed significant main effects on voice, F(1, 182) = 10.40, p = 0.001, partial η² = 0.05, and psychological safety, F(1, 182) = 29.92, p < 0.001, partial η² = 0.14.

Regression analyses tested mediation. As shown in Table 1 (Model 3), psychological safety positively predicted voice when controlling for power and status (B = 0.46, SE = 0.06, p < 0.001). Using Preacher and Hayes' (2004) PROCESS macro, indirect effects through psychological safety were significant for both power (-0.17, 95% CI = [-0.36, -0.01]) and status (0.45, 95% CI = [0.26, 0.70]). Hypotheses 1–3 were supported.

The status × power interaction on voice was significant, F(1, 182) = 6.27, p = 0.01, partial η² = 0.03. Simple effects showed that in the low-status condition, power negatively affected voice (M_low_power = 4.63, SD = 1.06 vs. M_high_power = 3.89, SD = 1.35), F(1, 182) = 10.11, p = 0.002, partial η² = 0.05. In the high-status condition, power's effect was non-significant (M_low_power = 4.73, SD = 1.07 vs. M_high_power = 4.83, SD = 0.88), F(1, 182) = 0.15, p = 0.70. Hierarchical regression (Table 1, Model 4) also showed a significant interaction (B = 0.82, SE = 0.33, p = 0.01). Hypothesis 4 was supported.

Table 1 Hierarchical Regression Analysis of Voice Behavior (Study 2)

We tested Hypothesis 5 (mediated moderation) following Edwards and Lambert (2007). First, the power × status interaction significantly predicted psychological safety, F(1, 182) = 5.82, p = 0.02, partial η² = 0.03. As shown in Figure 3, in the low-status condition, power negatively affected psychological safety (M_low_power = 3.78, SD = 1.41 vs. M_high_power = 2.95, SD = 1.18), F(1, 182) = 10.10, partial η² = 0.05. In the high-status condition, power's effect was non-significant (M_low_power = 4.30, SD = 1.26 vs. M_high_power = 4.38, SD = 1.03), F(1, 182) = 0.06, p = 0.80. Regression analysis (Table 1, Model 8) also showed a significant interaction (B = 0.88, SE = 0.36, p = 0.02).

Second, controlling for the interaction, psychological safety significantly predicted voice (Table 1, Model 5: B = 0.45, SE = 0.06, p < 0.001), rendering the interaction effect non-significant (B = 0.42, SE = 0.29, p = 0.15). These results support Hypothesis 5.

Figure 2 The Effects of Leader Power and Status on Voice Behavior (Study 2)
Figure 3 The Effects of Leader Power and Status on Psychological Safety (Study 2)

Study 3

4.1.1 Sample

Participants were employees and supervisors from a home building materials company in southern China producing lighting, ceiling systems, bathroom fixtures, and flooring. With HR assistance, we invited 591 employees and 130 supervisors. To reduce common method bias, we used a multi-wave, multi-source design. At Time 1 (T1), employees rated supervisor power and status. At Time 2 (T2, six months later), employees reported psychological safety while supervisors rated employee voice behavior.

To reduce supervisor burden, we limited subordinate sampling to eight per supervisor. When teams had ≤8 members, all received surveys; when >8, we randomly selected eight. After matching, we obtained 346 employee (58.54% response) and 111 supervisor (85.38% response) questionnaires. In the final sample, 80.18% of teams had ≥50% participation, with supervisors rating an average of 3.12 subordinates. Among 346 employees, 44.22% were male, mean age was 31.71 years (SD =