Observing Leadership as and in Networks: A Multilevel Investigation of Shared Leadership, Discrete Emotions, and Knowledge Behaviors

Introduction

Today, organizations must operate with continually emerging new technologies that are complex, dynamic, and unpredictable in commercial environments.^1,2 As a way to remain agile and flexible, organizations have adopted a more lateral and informal style of leadership, which can be termed shared leadership.^3,4 Shared team leadership facilitates the coordination and integration of knowledge among team members, enabling them to achieve higher levels of performance and creativity.^5–7 Despite the growing interest and numerous studies into shared leadership, there are still several important gaps in understanding shared leadership and its consequences.

First, there is a limited understanding of the multilevel nature of shared leadership. Given that shared leadership is a leadership network made up of multiple emergent leaders, the existing literature would benefit from the inclusion of individual-level analysis.⁸ Thus, drawing on dual network lenses,⁹ the first contribution of the current study is to introduce “shared leadership as networks” (SLAN) and “shared leadership in networks” (SLIN) to concepts to describe the multilevel attribute of shared leadership. Specifically, SLAN is defined as an emergent network of leadership whereby nodes and links represent team members and interactions related to leadership roles and influence, respectively.^10,11 SLIN is defined as emergent leaders who receive an overwhelming majority of leadership nominations (grants) by their peers.¹²

Second, recent studies have noted the potential benefits (eg, information communication and cooperation) and challenges (eg, power struggles and competition) for leaders in shared environments,^13,14 implying a potential double-edged sword effect of SLIN. Emerging leaders may either share knowledge to promote teamwork or hide knowledge to maintain their position of leadership and advantage over others. However, there no empirical research has yet explored the mechanisms at play between SLIN and knowledge behaviors. Therefore, to bridge this gap, we considered both productive (knowledge sharing) and counterproductive (knowledge hiding) knowledge behaviors, seeking to uncover the complex mechanisms between SLIN and knowledge behaviors.

One important mediation mechanism that might explain this double-edged sword effect is emotions. Indeed, the ongoing “affective revolution” has underlined the importance of emotions in the workplace,¹⁵ particularly specific discrete emotions in the workplace.¹⁶ Existing evidence suggests that pride helps to increase employees’ intention to share knowledge,¹⁷ but that fear of losing power triggers one’s pessimistic expectations about the future, thus triggering knowledge hiding behaviors.¹⁸ Pride (as a positive emotion) and fear of losing power (as a negative emotion) may have unique implications for explaining SLIN’s knowledge behaviors. However, few studies have employed a comprehensive theoretical framework to examine the above emotional mechanisms. Emotion appraisal theory (EAT) is an appropriate framework to illustrate the generation of one’s emotions as well as the subsequent coping mechanisms.^19,20 Knowledge sharing (an approach-oriented response) and knowledge hiding (an avoidance-oriented response) are the two specific coping strategies we focused on in this study.¹⁸ Hence, drawing on EAT, we aimed to investigate the mediating role of pride and fear of losing power on the relationship between SLIN and knowledge behaviors.

Meanwhile, SLAN sets boundaries for the indirect relationship between SLIN and one’s knowledge behaviors. According to EAT, team context provides the most direct setting for understanding individual emotions and coping strategies.²⁰ SLAN indicates a high density level in a leadership network, which can facilitate knowledge sharing processes within the team.²¹ Emergent leaders who are embedded in a denser leadership network are more likely to develop a more critical awareness of the benefits of taking on leadership responsibilities, resulting in enhanced pride and increased knowledge sharing behaviors. Therefore, this study explored the moderating role of SLAN in the SLIN–discrete emotions relationship, thereby illustrating when SLIN engages with more productive knowledge behaviors. The theoretical model is depicted in Figure 1. The following research questions guided this study:

Figure 1 The overall theoretical model. Notes: H1 and H2 represent proposed mediation paths from shared leadership in networks to the outcomes through pride and fear of losing power, respectively. H3 and H4 in brackets represent moderated mediation paths from shared leadership in networks to the outcomes through pride and fear of losing power, respectively.

RQ1. How does a review on shared leadership help us to define and distinguish SLAN and SLIN?

RQ2. What mechanisms guide someone in a SLIN context to engage in productive and counterproductive knowledge behaviors, such as pride or fear of losing power?

RQ3. How does SLAN as a social context affect the knowledge behaviors of SLIN?

By addressing these research questions, this study offers three significant theoretical contributions. First, by defining SLAN and SLIN, our work bridges the disjointed domain of shared leadership and lays the foundation for further development of the concept of multilevel shared leadership. Second, we shift attention from the whole network to central nodes and investigate both the positive and negative effects of SLIN on actors (leaders), which answers recent calls to focus on individuals in networks²² and on individual-level outcomes.¹¹ Further, this study provides an integrated theoretical framework for clarifying how SLIN links knowledge behaviors between leaders. Third, this study identifies SLAN as a contingent factor that may enable SLIN to engage in more knowledge sharing behaviors, thus expanding the knowledge on the utility boundaries of shared leadership. This study also has practical significance as it provides ways in which organizations can influence knowledge practices through emotional mechanisms. More importantly, it shows the top-down impact of SLAN on one’s behaviors and how it helps practitioners to reinforce the positive side of SLIN.

Theory and Hypotheses

SLAN and SLIN

In this section, we address RQ1. Shared leadership can be regarded as an “emergent process” whereby multiple individuals are identified as leaders through continuous “claiming-granting” interactions,^23,24 resulting in a broad distribution of leadership roles and influence among team members.¹¹ Thus, the multilevel nature of shared leadership can be observed. Specifically, emergent leaders are observed at the individual level, while emergent team phenomenon is observed at the team level. However, the current understanding and development of this multilevel structure are hindered by two main issues. First, there is a very little discussion of emergent leaders at the individual level, typically mentioned only in descriptions or measurements of shared leadership. Second, as of yet there is no integrated perspective to help us understand the connections and distinctions between individual-level and team-level shared leadership. Thus, for the current study, we reviewed related terms, definitions, and measurements in the shared leadership literature to provide support for this multilevel structure (see Table 1).

Table 1 Summary of Sample Terms, Definitions and Measurements of Shared Leadership at the Team Level and Individual Level

We specify and differentiate between “shared leadership as networks” (SLAN) and “shared leadership in networks” (SLIN), referring to the relevant terms in leadership networks as proposed by Crater et al.⁹ As illustrated in Table 1, although definitions of shared leadership at the team level vary, one commonality is that shared leadership is conceptualized as “networks.”^10,12,25 In a network, multiple individuals emerge as both informal leaders and followers. Thus, we have defined SLAN as an emergent network of leadership, with nodes representing team members and links representing interactions related to leadership roles and influence.^10,11 In terms of measurement, shared leadership can be assessed by diverse network indexes, such as density, centralization, network efficiency, and strength. In our study, density was employed to represent the whole network structure as it is the most frequently used index used to measure network structure^8,26,27 and has demonstrated adequate validity.²³ For individual-level shared leadership, the discriminant criteria and discussions are relatively inconsistent across the literature. Some researchers conceptualize it as “each team member individually perceives and accepts the leadership influence from their peers”,⁸ while others conceptualize it as “some team members who are identified as the nodes receiving the greatest number of ties.”^10,25,28 However, the issue with the former is that all team members are implicitly identified as emergent leaders and assigned individual-level scores when measured. As such, we define SLIN as emergent leaders those who receive the overwhelming majority of leadership nominations (grants) by their peers.¹² Given that leadership is about direct influence on others, we used in-degree centrality to measure and identify emergent leaders in leadership sociograms.^29,30 In the next two sections, we will explore the consequences of SLIN and how these consequences are shaped by the team context (SLAN).

Dual Emotions and Knowledge Behaviors of SLIN

In this section, by introducing the Positive Emotion Path and the Negative Emotion Path, we address our RQ2. Specifically, we illustrate how pride and fear of losing power are elicited and how each influences subsequent responses. Emotional appraisal theory (EAT) interprets the generation of and tactics for coping with emotions from an appraisal perspective.^19,20 Emotions are motivated by a broader kind of stimulus associated with benefits or harms.¹⁹ SLIN signifies the centrality and status of emergent leaders, and such a network position contains both resources and risks.³¹ Thus, SLIN could be regarded as a stimulus at the individual level. According to EAT, differential appraisals of stimulus determine emotion valence. Evaluations of having taken a leadership role that brings one closer to their goals or desired benefits will invoke pride, whereas evaluations that pull one further away from their goals or result in losses will invoke fear of losing power. To cope with these emotions, individuals may initiate different responses.³² In line with Fang,¹⁸ we proposed that knowledge sharing represents approach-oriented responses, whereas knowledge hiding reflects avoidance-oriented responses.

The Path of Positive Emotion: SLIN, Pride, and Knowledge Behaviors

Pride is a positive emotion that results from perceived benefits,²⁰ and at its core comprises self-worth, self-efficacy, and a sense of achievement.³³ Based on EAT, as leaders are involved in daily shared leadership activities, they can perceive benefits from exerting both self-directed and other-directed influence, and in turn experience pride. The reasons are as follows: First, leadership distribution creates a particularly motivating environment that reinforces one’s mission and intrinsic motivation.³⁴ Emergent leaders are more inspired to carry out self-directed work, which enhances the meaning of their work and self-worth.³⁵ Second, compared with non-emergent leaders, SLIN is linked with a large number of members in the network. Interacting with different members leads to the development of excellent verbal persuasion abilities in SLIN, which helps the leaders to be effective in their other-directed work, also enhancing their sense of pride.³⁶ Therefore, SLIN positively predicts pride.

According to Lazarus,²⁰ positive emotions have action tendencies, which are primarily reflected in one’s desire to take actions to maintain or enhance those emotional experiences. Following the core theme of pride,³³ pride indicates that individuals are confident in their ability to cope with tough situations (high coping potential) and hold a high level of self-value (high future expectations), both of which lead to approach-oriented responses, ie, knowledge sharing. Because knowledge sharing is generally considered to be an active and prosocial behavior, participating in these activities contributes to maintaining favorable emotional experiences. Research has shown that pride positively influences knowledge sharing.¹⁷ Thus, we proposed that:

H1a: SLIN is positively and indirectly associated with knowledge sharing via pride.

Knowledge hiding, on the other hand, is viewed as an avoidance-oriented response in which individuals protect themselves by concealing facts and information from others.¹⁸ Pride, as a self-conscious emotion, encourages individuals to avoid the degradation of their self-esteem or self-concept,³⁶ which leads to counterproductive work behaviors,³⁷ ie, knowledge hiding.³⁸ However, pride also stems from self-efficacy. A SLIN with high self-efficacy may believe in their own competency and be less likely to conceal knowledge for fear of making mistakes.³⁹ Hence, we propose that:

H1b: SLIN is negatively and indirectly associated with knowledge hiding via pride.

The Path of Negative Emotion: SLIN, Fear of Losing Power and Knowledge Behaviors

Given the potential dysfunctional influences of sharing leaders,^13,14 SLIN with high in-degree centrality may lead to negative cognitions and behaviors. Fear of losing power (FLP) is a negative emotion that stems from perceived losses or threats that are associated with power and personal values.^18,20,40 According to EAT, SLIN is more likely to generate fear of losing power when one experiences potential losses or threats.¹⁹ It may cost a great deal of time and energy to develop and maintain an effective leader-follower relationship, so a SLIN that involves more links may lead to role conflicts and overload.^13,41 In such a predicament, leaders are more vulnerable to anxiety or fear.⁴² Furthermore, more links in the network imply more access to information from other powerful actors,⁴³ also increasing the likelihood of information devaluation because of unfair transactions with less-central members. Thus, SLIN may experience strong fear of losing power.

Negative emotions are associated primarily with withdrawal or avoidance tendencies.⁴⁴ Fear of losing power enhances uncertainty,⁴⁰ as individuals are unable to anticipate when they might suffer losses and their impacts, which then results in a sense of powerless. Teo and Yu⁴⁵ have put forth that individuals are inclined to terminate transactions in response to activities with high uncertainty. Knowledge sharing is an activity that involves uncertainty because there are no explicit rules with regards to how to reward knowledge owners for their contributions.⁴⁶ Consequently, the potentiality of adverse consequences may decrease one’s desire to share knowledge. Previous research has found that FLP negatively affects knowledge sharing.^18,40 As such, we hypothesized that:

H2a: SLIN is negatively and indirectly associated with knowledge hiding via fear of losing power.

Fear tends to awaken pessimistic judgments and encourage avoidance-oriented responses.⁴⁷ When individuals experience strong fear, the avoidance coping system is triggered to help one escape from potential threats, or to create a sense of psychological safety.⁴⁸ Fang found that FLP positively predicted knowledge hiding.¹⁸ Hence, we proposed that:

H2b: SLIN is positively and indirectly associated with knowledge hiding via fear of losing power.

The Moderating Role of SLAN

By introducing team-level stimulus (SLAN), we explored the boundary conditions of the effects of SLIN on knowledge behaviors, thus addressing our RQ3. EAT suggests that social context is crucial in shaping one’s emotions and emotional coping strategies.²⁰ In teams, leadership provide the most instantaneous social context for individuals. Previous studies revealed that leadership significantly influences employees’ perceptions and their behaviors.^49,50 Thus, SLAN (representing the density of team leadership networks) may affect the emotions of leaders and indirectly impact their behaviors.

SLAN strengthens leaders’ perceptions of benefits making them more likely to experience pride. In a denser network, more team members exert leadership responsibilities, which include team planning, problem-solving, mentoring and development, as well as support and consideration.⁵¹ Team members are influenced by others while also exerting their own influence.³⁴ As a result, a SLIN can gain direct experience by leading others, as well as indirect experience by observing other leaders,⁵² which together enhances their sense of accomplishment and pride. Furthermore, an increase in shared leadership density facilitates the development of a transactive memory system, implying that teams develop a strong collective memory structure that allows their members to confidently rely on each other’s expertise.²⁷ This contributes to increased self-efficacy in completing tasks. Recent research also shows that a high density SLAN increases individuals’ creative self-efficacy.⁵³ In other words, SLAN strengthens the positive relationship between SLIN and pride. Combining H1a and H1b, we predicted that the indirect influence of SLIN on knowledge behaviors via pride would be contingent on the level of SLAN, as reflected in the following:

H3: SLAN moderates the indirect effects of SLIN on (a) knowledge sharing and (b) knowledge hiding through pride. Specifically, these indirect effects are stronger under higher (vs lower) levels of SLAN.

In a dense leadership network, most team members lead and motivate each other, which shapes the common team culture.⁵⁴ Emergent leaders are inspired to fulfill their leadership responsibilities and believe that other team members will respond actively to their leadership efforts,²⁵ fostering a pleasant, trusting, psychologically safe workplace in which free expression of ideas is encouraged.^55,56 Thus, SLIN experiences less FLP. In addition, a high level of leadership network density also represents a high degree of mutual level and socialization.⁵⁷ In such a network, selfishness and unethical perceptions and behaviors are undesirable due to strong organizational norms and consensus,⁵⁸ therefore further weakening FLP. In short, SLAN weakens the positive relationship between SLIN and FLP. Combining H2a and H2b, we proposed that:

H4: SLAN moderates the indirect effects of SLIN on (a) knowledge sharing and (b) knowledge hiding through fear of losing power. Specifically, these indirect effects are stronger under higher (vs lower) levels of SLAN.

Method

Sample and Procedure

This study initiated a two-wave survey in Chinese knowledge-intensive teams. To ensure our samples were appropriate, we set two important team selection criteria: First, the invited teams should have a clear output which ensured that team members shared a common goal. Second, their teamwork was expected to be highly interdependent and required selective exploitation of members’ unique expertise, which served as a foundation for sharing leadership responsibilities.³⁵ Next, we employed the relevant resources of a leadership project in China to contact Human Resources managers or team leaders from different organizations. We described the purposes of the research, sample selection criteria, and survey procedures of this study to them. With their assistance, we recruited 98 teams from different industries to participate in our study. These industries included IT (24.36%), finance (16.01%), scientific research and development services (31.79%), and the pharmaceutical industry (27.84%). Sample teams were selected from information technology operation and maintenance, research and development (R & D), marketing, consultation, and scientific research.

The procedures were as follows. With the help of Human Resources managers or team leaders, we gathered a directory of all members ahead of data collection, providing each participant with a unique code to match them with their team. Meanwhile, a matching social network survey was designed to assess the leadership network structure of each team.¹⁰ Next, two time-lagged surveys were administered in the field setting with a one-month interval. At Time1, 534 members from 98 teams were invited to participate in the first survey. In this phase, we collected data regarding their leadership network structure (SLIN and SLAN) and emotions (pride and fear of losing power). At Time 2, the second survey was distributed to all team members again, who this time were asked to rate their knowledge sharing and knowledge hiding behaviors. After the two time points, 464 questionnaires from 79 teams were returned. After excluding invalid questionnaires that had incomplete responses and data had obvious problems (eg the same rating score for all measurement items), 455 questionnaires from 79 teams were obtained. We then excluded seven sets of team data (24 members in total) because social network analysis requires a minimum 80% within-group response rate.⁵⁹ This left us with 431 members’ data from 72 teams for the SLAN calculation. The effective response rates for teams and members were 73.47% and 80.71%, respectively. Furthermore, we identified 350 network stars by using leadership network sociograms and retained data from these 350 members for individual-level analysis.

The final sample comprised 431 team members from 72 teams, which also included the 350 emergent leaders. The average size of each team was 6.31 (SD = 1.56). The sample of the leader respondents had an average age of 27.62 years (SD = 3.95), 54.0% were female, and the average team tenure was 1.21 years (SD = 1.10). All were well-educated, with 77.43% holding a bachelor’s degree, 10.0% a master’s or doctor’s degree, and 12.57% had a lower degree.

Measures

Similar to Hassan et al,⁶⁰ all materials were translated from English into Chinese using a back-translation procedure. The final draft of the surveys was sent to two experts in the field of organizational behavior and human resource management to ensure accuracy and clarity.

SLAN and SLIN (Time 1)

We employed a social network approach to assess leadership network structure. The process was divided into three steps. First, team members were asked to carefully read the definitions of the four leadership roles. After each definition, they were asked to rate their peers, ranging from 1 (not at all) to 5 (to a great extent). There are five items in total, four of which reflect the leadership functions,^51,61 and the other is the overall rating of leadership.³⁴ A sample item for the leadership roles is,“Navigator – Helps to establish the team’s purpose and direction. A list of Navigator’s actions: Initiates and energizes team action in pursuit of its purpose and goals; Communicates and reminds team of its overall purpose; Helps develop specific goals towards achieving team’s purpose. Please rate to what extent each team member has engaged in the Navigator role.” A sample item for overall leadership is, “To what degree does your team rely on this individual for leadership?” The five-item scale had a reliability alpha of 0.95. Second, the responses from the five items were aggregated and then collated into an overall leadership network matrix.²⁶ Finally, we use Ucinet 6.186 to compute density and in-degree centrality, to indicate SLAN and SLIN, respectively. Agreement among members was as follows: mean r_wg = 0.95, ICC₁ = 0.33, and ICC₂ = 0.75.

Density captures the magnitude of reciprocal influence among team members, with higher levels indicating higher levels of SLAN. To compute the SLAN score, we aggregated all members’ ratings and then divided them by the total number of ratings. This approach has been previously adopted and tested in numerous studies.^8,26,34 In-degree centrality captures the importance of an emergent leader and their influence on other team members.²⁹ Members with higher in-degree centrality (ie, the strength of the incoming ties) might be recognized as leaders.¹⁰ Hence, we employ in-degree centrality to measure SLIN and identify leaders in leadership sociograms. To control the influence of team size, we standardized the in-degree centrality.

Pride (Time 1)

Pride was measured using a seven-item scale developed by Tracy and Robins,³³ which assesses positive emotion experiences triggered by achievement events or social relationships. Team members were asked to self-rate their authentic pride using a five-point Likert scale ranging from 1 “strongly disagree” to 5 “strongly agree”. A sample item is, “When working with others in my group, I feel self-worth.” (Cronbach’s α = 0.96).

Fear of Losing Power (Time 1)

Fear of losing power (FLP) was assessed using a five-item derived from Fang.¹⁸ It measures the extent to which employees fear the loss of power, status, and personal value. Team members were asked to rate their FLP using a five-point Likert scale ranging from 1 “strongly disagree” to 5 “strongly agree”. A sample item is, “I am afraid of being replaceable (losing my position) if I share all my relevant knowledge.” (Cronbach’s α = 0.95).

Knowledge Sharing (Time 2)

Knowledge sharing was measured using a five-item scale developed by Huang et al,⁶² which captures the existing knowledge sharing behaviors of team members. The scale included two dimensions, namely, tacit knowledge sharing (three items) and explicit knowledge sharing (two items). Participants’ responses ranged from 1 “strongly disagree” to 5 “strongly agree”. Sample items for these two dimensions included “I always provide my know-where or know-whom at the request of other team members” and “I share my work reports and official documents with members in this team frequently.” (Cronbach’s α = 0.92).

Knowledge Hiding (Time 2)

Knowledge hiding was measured using a three-item scale developed by Serenko and Bontis,³⁸ which captures the extent to which individuals intentionally conceal or withhold work-related knowledge from other team members. Participants’ responses ranged from 1 “strongly disagree” to 5 “strongly agree”. A sample item is, “I often twist the facts to suit their needs when communicating with other members within the team” (Cronbach’s α = 0.75).

Control Variables

Control variables included team size, gender, age, and education, as the literature has noted their impact on knowledge behaviors.^63,64 Team job tenure was also controlled as it captures the experience of members working, which could influence their knowledge behaviors⁶³ and affect team effectiveness.²⁶

Analytic Strategy

The hypothesized model included five level-one variables (ie, SLIN, pride, fear of losing power, knowledge sharing, and knowledge hiding) and one level-two variable (ie, SLAN). The multilevel model was tested in Mplus 7.4 using testing procedures as follows. To test Hypotheses 1 and 2, we built a mediational model.⁶⁵ We also applied Monte Carlo simulations (replication = 20,000) to estimate bias-corrected confidence intervals for the indirect effects.⁶⁶ To test Hypotheses 3 and 4, we built four cross-level moderated mediation models. We standardized the independent variable (SLIN) and the moderator (SLAN) with the grand mean to pay attention to the real moderating effect between groups.⁶⁷ We also calculated bias-corrected confidence intervals for the conditional indirect effects.

Results

Table 2 presents the means, standard deviations, and correlations of all variables in this study.

Table 2 Variable Means, Standard Deviations, and Correlations

Confirmatory Factor Analysis

SLAN and SLIN can be methodologically separated from other variables.²⁶ Thus, we conducted a confirmatory factor analysis for the remaining variables (ie, pride, fear of losing power, knowledge sharing, and knowledge hiding). The four-factor model exhibited an excellent fit: χ² = 375.82, df =164, χ²/df = 2.29, CFI = 0.96, TLI = 0.96, RMSEA = 0.06, SRMR = 0.04. In contrast, the other alternative models all showed a worse fit (pride and fear of losing power as one factor: χ² = 2270.36, df =167, χ²/df = 13.59, CFI = 0.64, TLI = 0.59, RMSEA = 0.19, SRMR = 0.20); two-factor model (pride and fear of losing power as one factor, and knowledge sharing and knowledge hiding as another factor: χ² = 2584.31, df =169, χ²/df = 15.29, CFI = 0.59, TLI = 0.54, RMSEA = 0.20, SRMR = 0.21); and one-factor model (all variables as one factor: χ² = 3710.22, df = 170, χ²/df = 21.82, CFI = 0.40, TLI = 0.33, RMSEA = 0.24, SRMR = 0.24). The results indicate discrimination between research variables.

Common Method Bias

This study adopted both procedural and statistical remedies to minimize common method bias.⁶⁸ First, we employed a social network approach to measure SLIN and SLAN, so they could be methodologically separated from other aggregation-based variables.²⁶ Further, we collected ratings of emotions and knowledge behaviors at different times, which provided temporal separation to mitigate the influence of common method bias.^68,69 Finally, statistical methods were applied to detect the magnitude of common method bias, including Harman’s single-factor test and control for unmeasured latent method factor. Harman’s single-factor test has been used by Hassan et al.⁷⁰ The results of Harman’s single-factor test showed that the cumulative variance explained 77.45% of the total variance, with the first principal component explaining 33.39% of the variance (less than 50% and no more than half of the total variance). It indicated that common method bias had a limited impact on our study. In addition, following the recommendation by Podsakoff et al,⁷¹ we controlled for an unmeasured latent method factor. The results showed that the fit indices between the controlled model with latent common method factor and the above four-factor model were not significantly changed (Δχ2/df = 0.15, ΔCFI = 0.001, ΔTLI = 0.003, ΔRMSEA = 0.002, ΔSRMR = 0.002). It alleviates the concern for common method bias.

Tests of Mediation Effects

Figure 2 and Table 3 show the results of the mediation effects. Hypothesis 1 proposed that pride mediates the effect of SLIN on knowledge behaviors (knowledge sharing and knowledge hiding). Figure 2 shows that SLIN was positively associated with pride (B = 0.33, SE = 0.08, p < 0.001), and that pride significantly predicted knowledge sharing (B = 0.37, SE = 0.07, p < 0.001). Table 3 shows that pride mediated the effect of SLIN on knowledge sharing (Effect = 0.13, SE = 0.03, 95% CI [0.06, 0.21]), supporting Hypothesis 1a. However, the results indicated an insignificant relationship between pride and knowledge hiding (B = −0.02, SE = 0.04, p > 0.05), and pride did not mediate the effect of SLIN on knowledge hiding (Effect = −0.01, SE = 0.01, 95% CI [−0.03, 0.02]), failing to support Hypothesis 1b.

Table 3 Results of Indirect and Conditional Indirect Effects of SLIN Through Emotions

Figure 2 Results from multiple mediation model. *p < 0.05, ***p < 0.001. Notes: Presented are nonstandardized model coefficients. Control variables and covariances between shared leadership in networks and knowledge behaviors are not shown.

Hypothesis 2 proposed that fear of losing power mediates the effect of SLIN on knowledge behaviors. Figure 2 shows that SLIN was positively associated with fear of losing power (B = 0.21, SE = 0.09, p < 0.05), which was in turn significantly associated with knowledge sharing (B = −0.23, SE = 0.05, p < 0.001) as well as knowledge hiding (B = 0.19, SE = 0.05, p < 0.001). Supporting Hypotheses 2a and 2b, fear of losing power significantly mediated the effects of SLIN on knowledge sharing (Effect = −0.05, SE = 0.02, 95% CI [−0.10, −0.01]) and knowledge hiding (Effect = 0.04, SE = 0.02, 95% CI [0.01, 0.08]).

To test the likelihood of reverse causality, we compared the values of Akaike’s Information Criterion (AIC) and Bayesian Information Criterion (BIC) for our mediation model and reverse causal model, based on recent work.^72,73 The results show that the hypothesized model (AIC = 14263.61, BIC = 14568.384) had lower AIC and BIC than the reverse causal model (AIC = 14938.50, BIC = 15247.14), indicating that our mediation model provided a better fit to the data.

Tests of Moderated Mediation Effects

We first tested moderation effects of SLAN. The results show that SLAN significantly interacted with SLIN to predict pride (B = 0.50, SE = 0.23, p < 0.05) and fear of losing power (B = −0.74, SE = 0.24, p < 0.01). These results provide initial support for Hypotheses 3 and Hypotheses 4. Then, we tested four moderated mediation models (see Table 3). In support of Hypothesis 3a, the indirect effect of SLIN on knowledge sharing (via pride) was stronger when SLAN was high (Effect = 0.14, SE = 0.05, 95% CI = [0.04, 0.24]) compared to low (Effect = −0.01, SE = 0.05, 95% CI = [−0.11, 0.08]), and the difference between these two conditions was significant (Effect = 0.15, SE = 0.07, 95% CI = [0.01, 0.30]). However, SLIN affecting knowledge hiding via pride was not significant either when SLAN was high (Effect = −0.01, SE = 0.02, 95% CI = [−0.05, 0.02]) or low (Effect = 0.001, SE = 0.01, 95% CI = [−0.01, 0.01]), rejecting Hypothesis 3b. Similarly, the conditional effect of SLIN on knowledge sharing (via fear of losing power) was stronger when SLAN was high (Effect = 0.06, SE = 0.02, 95% CI = [0.01, 0.10]) compared to when it was low (Effect = −0.03, SE = 0.02, 95% CI = [−0.07, 0.02]). The difference between two conditions was significant (Effect = 0.09, SE = 0.04, 95% CI = [0.01, 0.16]), supporting Hypotheses 4a. Moreover, the conditional effect of SLIN on knowledge hiding via fear of losing power was stronger when SLAN was high (Effect = −0.07, SE = 0.03, 95% CI = [−0.14, −0.004]) compared to when it was low (Effect = 0.03, SE = 0.03, 95% CI = [−0.02, 0.09]), and the difference between the two conditions was significant (Effect = −0.10, SE = 0.05, 95% CI = [−0.21, −0.001]), supporting Hypotheses 4b.

Discussion

In this research, we discussed the multilevel nature of shared leadership (SLAN and SLIN). We also investigated the relationship between SLIN, emotions, and knowledge behaviors, as well as the moderated mediation effects of SLAN. Our results show that emergent leaders (SLIN) may feel pride and fear of losing power, which in turn can influence their knowledge behaviors. The results also indicate that SLAN is essential in increasing knowledge sharing and mitigating knowledge hiding when individuals take on emergent leadership roles. Surprisingly, pride strengthened knowledge sharing but failed to reduce knowledge hiding. One possible explanation for this could be that knowledge sharing and knowledge hiding are driven by different factors. Knowledge sharing may be driven by altruism, while knowledge hiding may be motivated by egoism via a series of tradeoffs between costs and values.³⁸ As such, pride may enable individuals to share knowledge to improve their self-esteem. However, they may also engage in knowledge hiding if there is no adequate return to justify their costs. This explains the non-significant results regarding Hypotheses 1b and 3b.

Theoretical Implications

Current findings suggest that we should shift our focus to shared leadership in networks (SLIN), as these individuals are the core of teams¹⁰ and their knowledge behaviors are critical to team functions.^7,21 These individuals deliberately hiding crucial information that could have strategic value to the team can be detrimental to the team’s ability to attain its goals.⁷⁴ Losses caused by knowledge hiding have been reported to amount to $31.5 billion in Fortune 500 firms.⁷⁵ Therefore, it is imperative to include individual-level analyses when investigating the effects of shared leadership on knowledge behaviors. However, individual-level knowledge in shared leadership contexts is largely undeveloped.⁸ In this study, we consider both team-level shared leadership (SLAN) and individual-level shared leadership (SLIN) from a dual network lens. This provided insight into multilevel shared leadership and broadens research understandings on individual perspectives for future studies. Further, our novel findings confirm the double-edged sword effect between SLIN and knowledge practices. Amidst the “ongoing affective revolution”,¹⁵ this research can help academic researchers and practitioners understand the crucial discrete emotions that shape employees’ knowledge behaviors. Moreover, our work identifies the contingent factor of SLAN that can contribute to organizations/teams promoting the positive side of SLIN in order to avoid impacts of its more problematic aspects. We list three specific contributions of this study below.

First, this study furthers the knowledge of the multilevel nature of shared leadership and specifically clarifies the conceptualization and operationalization of SLIN in the literature. While previous research has considered this multilevel structure,^8,54 there are several critical questions that remain unanswered. For instance, how can this structure be interpreted through an integrated theoretical lens? How does this structure develop from various definitions and measurements of shared leadership? Thus, we reviewed the existing literature and summarized the terms, definitions, and measurements related to this structure. We then defined and distinguished between SLAN and SLIN from dual network lenses, enabling the possibility to develop shared leadership literature. More importantly, by clarifying SLIN, our work expands the research focus, from the entire leadership network to the focal leaders in the network, enabling future research to focus on the outcomes of SLIN at different levels.

Second, this study investigated both the positive and negative effects of SLIN on leaders. By doing so, we extend existing research on the positive effects of shared leadership on the team knowledge process.^7,21 By shifting the perspective from the whole team to specific members, our findings identify preliminary evidence about the more problematic side of shared leadership.¹¹ The focus of the current study also moved away from the implicit assumption that shared leadership activities are perceived and experienced the same way. Liang et al³⁵ suggest that individual differences may result in distinct individual-level outcomes. We found that there are indeed differences in individuals’ appraisals of taking on leadership roles, resulting in a diverse range of emotional experiences and knowledge behaviors. Our work was a response to Sweeney et al²² who have suggested that attention to individuals is necessary to gain insight into how they interact within a leadership network. More importantly, these findings highlight the importance of pride and fear of losing power as behavioral motivators. These discrete emotions have unique implications for knowledge behaviors, and in supporting subsequent team success. By targeting specific discrete emotions, our findings complement the work by Hmieleski et al¹⁵ by enriching the understandings of the mediating mechanisms of shared leadership, as well as advancing the general understandings of team leadership, discrete emotions, and knowledge management practices.

Third, our work determined a boundary condition for the effects of SLIN on knowledge behaviors, which provides solid support for preventing the dysfunctional impact of SLIN. Specifically, SLAN was shown to be an important factor in explaining the different reactions of emergent leaders. As shown in Table 3, the indirect effects between SLIN and knowledge behaviors were stronger at a high level of network density, while these indirect effects were not significant at a low network density level. This indicates the significance of network density in strengthening one’s pride and knowledge sharing behaviors, as well as weakening their FLP and knowledge hiding behaviors. These discoveries not only add significant value to the shared leadership and knowledge management literature, but also provide theoretical implications for practitioners to intervene in counterproductive knowledge behaviors. In addition, this research responds to Carter et al⁹ who suggested developing theoretical extensions of leadership and undertaking multilevel leadership network research. We have extended shared leadership theory by considering two aspects of leadership – leadership as relational and leadership itself – and employed a social network approach to investigate how this multilevel structure affects individuals.

Practical Implications

In today’s dynamic, complex and ambiguous market environment, shared leadership is indispensable for organizations to maintain competitive advantages. The effectiveness of shared leadership relies on teams’ ability to obtain, coordinate, and integrate unique knowledge among their emergent leaders.²⁷ However, given that knowledge is embodied in particular individuals, team knowledge processes are only facilitated when members and leaders contribute their knowledge voluntarily. Our study focused on individuals in leadership networks, highlighting how SLIN engages in both productive and counterproductive knowledge behaviors through discrete emotions. Further, our work provides meaningful insights to potential practitioners regarding ways to mitigate the problematic side of SLIN, and offers the following managerial implications for organizations/teams.

Considering the potential costs and benefits of SLIN, organizations should take action to reduce its negative aspects while amplifying its positive aspects. To do this, organizations must be mindful of leaders’ emotions. In terms of knowledge hiding, the question of how to alleviate individuals’ fears of losing power is critical. Fear of losing one’s position or advantage encourages individuals to engage in knowledge hiding.¹⁸ Some research has indicated that the absence of reasonable payoffs may lead to counterproductive knowledge behavior.^76,77 Therefore, organizations could combine HRM policies to reward or promote SLIN who lead their teams effectively. Attention to pride seems to be a desirable practice for knowledge sharing. Teams could adopt social recognition programs⁷⁸ to recognize the contributions of emergent leaders, which would help to strengthen their pride and knowledge sharing.

Moreover, our findings highlight the significance of SLAN for one’s emotional experiences and knowledge behaviors. Carson et al³⁴ suggested that several internal factors can increase the density of SLAN, specifically: shared purposes, social support, and voice. Hence, managers can promote goal orientation and leadership motivation of SLIN by shaping collective work goals. They might also develop team support mechanisms that allow team members to provide emotional strength and assistance to one another. Furthermore, creating an environment where team members can express their ideas freely might be another potential way to increase the likelihood of team members taking on leadership roles. Based on these findings, we recommend organizations build denser leadership networks through staffing and training. For example, organizations could increase network density by designing leadership mechanisms that allow for a greater number of members to take on leadership roles or responsibilities, such as rotating team member leadership responsibilities over time.¹¹ Or, organizations could consider mentoring and developing team members to enhance their shared leadership competencies. The elevated leadership competency might contribute to the effectiveness of mutual influence among team members and therefore increase network density.

Limitations and Future Directions

There are some limitations of the current study that may provide opportunities for future research. First, despite collecting data from multiple sources and phases, this study was limitated in its ability to explore emotional stability. Therefore, future studies could adopt a longitudinal research design to capture changes in discrete emotions.

Second, consistent with most studies, we utilized density to capture leadership network structure.^26,34 However, networks with the same density may have different dispersions of influence among peers.¹¹ Thus, using centralization and density together would be desirable for future research.

Third, our study examined the double-edged sword effect of SLIN on knowledge behaviors at the individual level, but this finding cannot be inferred to the team level. We encourage future research to examine these effects at the team level.

Conclusion

Effectiveness in shared leadership teams depends on the voluntary knowledge contribution of emergent leaders. To provide insights into the relationship between emergent leaders and their knowledge behaviors, this study defines SLAN and SLIN and examines how and when SLIN can be a double-edged sword in organizations. The significance of this study is in enriching leadership network research and providing an integrated framework for the evolution of multilevel shared leadership. Our findings contribute to shared leadership, discrete emotions, and knowledge management literature by demonstrating that SLIN with high in-degree centrality can lead to the formation of distinct discrete emotions (pride and fear of losing power) and behaviors (knowledge sharing and knowledge hiding). Further, a high-density leadership network (SLAN) was shown to help trigger the positive side of SLIN. In practice, our work provides evidence of valuing the emotions of SLIN and developing network density to manage the dark side of SLIN. We hope that this study will inspire future research to expand the understanding of multilevel shared leadership and the interactions among emergent leaders and their team dynamics.