Understanding (non)leadership phenomena in collaborative interorganizational networks and advancing shared leadership theory: an interpretive grounded theory study

We study leadership at the interactional level in collaborative interorganizational networks which are understood as intentionally created social entities of three or more organizations in which legally autonomous members participate to achieve overarching objectives (Powell 1990; Provan et al. 2007; Miles et al. 2010). This definition implies that members have a perception of the network’s overall purpose that is, however, usually unrelated to specific goals. The term ‘legally autonomous’ means that member organizations’ relationships are characterized by the absence of any a priori existing formal guidance of a lead firm (cf. Miles et al. 2010; Provan et al. 2007; Huxham and Vangen 2000; Weibler and Rohn-Endres 2010). However, this definition does not imply any assessments of the actual interaction quality of the individual network actors, which we refer to here as individual representatives of the organizations constituting the network.

Our understanding of the term network is related to the interorganizational or network governance perspective, which basically understands interorganizational networks as a substantive form of organizing (Provan et al. 2007; Jones et al. 1997; Powell 1990). In the literature, this perspective is usually contrasted with the structuralist or social network analytical perspective, which refers to networks in a more formal sense using social network analysis. Here, networks are basically understood as an abstract notion and referred to as “a set of actors connected by a set of ties” (Borgatti and Foster 2003: 992). In contrast, networks from a network governance perspective are understood as a distinct form of coordination and exchange between usually autonomous firms or organizations. The term network governance refers to a form of inter-firm or inter-organizational coordination that is characterized “by informal social systems rather than by bureaucratic structures within firms and formal contractual relationships between them” (Jones et al. 1997: 911). This form of coordination thus relies more on socially committing as contrasted to legally binding contracts. The members can be linked by many types of connections and exchange flows (e.g. information, social support, services, materials) (Provan et al. 2007; Jones et al. 1997; Powell 1990). Such networks are characterized by an “inherent pluralism embedded in their structures and modes of working” (Sergi et al. 2012: 406). Hence, a key challenge for network management is to channel preexisting pluralism in order to advance the collaborative network agenda. In this context, the literature points to the complex and ambiguous nature of collaboration in networks that generates paradoxes and tensions for their management (Ospina and Saz-Carranza 2010; Vangen and Huxham 2003).

The networks selected for our study had a relatively similar collaborative form, i.e. the member organizations’ intention is to work together and exchange knowledge and experience on specific themes. In line with Huxham and Vangen (2000: 1159), we use the term “collaborative” to describe organizations (rather than just individuals) that are working together. Further, the networks had a similar thematic orientation (i.e., health, occupational and environmental protection, and industrial safety) and were primarily composed of member firms from the profit sector (cf. Sect. 3.2).

We understand leadership basically as a dynamic socially constructed influence process (Endres and Weibler 2017; Alvesson 2017; Uhl-Bien 2006; Hannah et al. 2014). The social constructionist leadership perspective has gained significant in importance more recently, since there is an increasing recognition of the need for a less individual-centred, more dynamic, processual view of leadership in organizations (cf. Uhl-Bien and Ospina 2012; Endres and Weibler 2017 for an overview and conceptual synthesis). More recently, for example, Wellman (2017: 614) similarly points to the dynamic realities of leadership in contemporary organizations which involve significant challenges for leadership theorizing, and he currently observes an emerging “need for new leadership models that acknowledge that leadership is socially constructed between the members of a group and is shaped in fundamental ways by the group context”. Approaching leadership as a socially constructed phenomenon is in line with our study’s interpretive methodological foundation, and involves the following basic assumptions and key characteristics: first, the basic mechanism of leadership is social construction, i.e. “processes of intersubjectively creating social realities through ongoing interpretation and interaction” (Endres and Weibler 2017: 222; cf. Uhl-Bien 2006; Ospina and Saz-Carranza 2010; Astley 1985; Morgan and Smircich 1980). Second, the social constructionist leadership approach seeks to deepen understanding of the intersubjective day-to-day experience and practices of people and how they are co-constructing leadership processes, instead of identifying ‘objective’ laws of human behaviour. In consequence, third, leadership as such is a process constructed in interactions between individuals via ongoing interpretations. This ontological process perspective of leadership implies that leadership is not (necessarily) tied to a formal (or hierarchical) leadership position. Therefore, social constructionist approaches explicitly avoid conflating leadership with supervision or formal guidance, and pay attention to potentially emerging organization wide (informal) leadership processes, as highlighted by Endres and Weibler (2017: 227; cf. Alvesson 2017; Uhl-Bien 2006). Finally, and fourth, a necessary component of leadership then is influence, i.e. “emerging flows of influence at the interpersonal interaction level or the collective level that represent the leadership manifestation” (Endres and Weibler 2017: 226; cf. Alvesson 2017). Hence, this process-oriented leadership approach takes into consideration that from the social construction process among individuals, both single leadership and shared (or collective) leadership, as well as hybrid forms, may emerge. In the literature, the term “shared leadership” is usually used interchangeably with “collective leadership”, and describes a plural form of leadership (cf. Avolio et al. 2009: 431; Denis et al. 2012; see Sect. 5 for more details).

From an identity-related perspective, which will gain significant relevance in our conducted study, DeRue and Ashford (2010) argue that leadership only occurs when others grant or ascribe leadership, and that individuals construct and internalize a leader or follower identity and that these identities become recognized by other members. In related fields, there have also been attempts to explain how and why a priori equal group members gain credit based on status ascriptions that allow them to exert influence and to emerge as a leader of the group (e.g. Hollander 1974; Paunova 2015; Stone and Cooper 2009). This approach also fits in with our approach in that it is open enough to examine emerging leadership inductively and embedded in context-specific interactions.

Despite its widely acknowledged importance, individual actors’ leadership and interaction-related leadership processes within networks have been rarely assessed empirically (cf. Jack 2010; Provan et al. 2007; Connelly 2007). For example, in a systematic review of interorganizational leadership, Müller-Seitz (2012) found only four empirical (qualitative) studies on the individual level of leadership (i.e., Beyer and Browning 1999; Martin et al. 2009; Ritala et al. 2009; Saz-Carranza and Ospina 2011). A further exploration of these sources shows that Beyer and Browning (1999) in their study of charismatic leadership in a consortium described the development of a joint vision through “cultural leadership” that influences how “followers collectively think and act” (1999: 485). Martin et al. (2009) pointed to the need for distributed leadership forms for achieving change in public service networks (see also Currie et al. 2011). Ritala et al. (2009) investigated individual skills and organizational capabilities and their interplay in successful innovation network management. Finally, given the identified research gaps, Müller-Seitz (2012) traced the “incoherence” of the literature on leadership in interorganizational networks and also pointed to the need for becoming “better engaged with classical research on interpersonal leadership” (2012: 438) and on interpersonal issues for developing a more comprehensive understanding of leadership in interorganizational networks.

Indeed, only a few additional empirical studies address this momentous topic. For example, in an early longitudinal study in two rule-making groups, Feyerherm (1994)—based on an interaction-oriented approach—revealed the informal, emergent nature of leadership in such collaborations. Similarly and based on a qualitative action research design, Huxham and Vangen (2000) emphasized the emergent nature of leadership and conceptualized leadership as “the mechanisms that ‘make things happen’ in a collaboration” (2000: 1165). In this view, leadership occurs through “leadership media,” comprising structures, processes, and participants, as well as through “leadership activities,” carried out by participants to promote the “collaborative agenda” (Huxham and Vangen 2000: 1165ff.). In line with this research, but focusing on individual network managers, Vangen and Huxham (2003) revealed the contradictory demands on “partnership managers” who become involved in both facilitative and directive leadership roles while organizing the collaboration’s activities. Tensions and dilemmas were also an important topic in Sydow et al.’ (2011) study of cluster management. They identified a “silent cry paradox” which should be “managed” by “organizing for leading (in) clusters” based on “reflexive structuration” (Sydow et al. 2011: 339). Ospina and Saz-Carranza (2010) addressed another paradox—namely, the “unity and diversity paradox” in network management—and concurrently described a set of practices as a response to the paradoxical and complex requirements of network collaboration. In a similar vein, Weibler and Rohn-Endres (2010) focused on interaction and practices, and finally portrayed leadership in interorganizational networks as occurring through the interplay of structures, individuals, and the collective—and developed the notion of shared network leadership as dialogue practice. Focusing more on the outcomes of leadership, Davis and Eisenhardt (2011) conducted a comparative case study in eight technology collaborations and found that what they termed “rotating leadership” was superior in eliciting innovation compared to both “dominating leadership” and “consensus leadership” (2011: 159). Denis et al. (2001) also found a relation between specific leadership processes and outcomes. They highlighted the importance of shared leadership in promoting change and pointed to the fragility of such leadership constellations.

While all these studies are informative and relevant taken together, they still do not clarify what leadership types, behaviors, or practices might be required in interorganizational networks in which individuals participate of their own accord and without formal fiat or based on employment contracts. This implies providing rich context-sensitive descriptions of team and leadership phenomena in practice and moving beyond narrow theoretically (pre-)framed or prescriptive accounts (cf. Hannah et al. 2014; DeRue and Ashford 2010; Blom and Alvesson 2015). Similarly, Endres and Weibler (2017) showed that there is a significant need for research that does not start with ‘leadership’, or what is assumed to be leadership, but rather with interaction dynamics and practices as they occur in a specific setting.

Interpretive grounded theory (Charmaz 2014; Suddaby 2006; Glaser and Strauss 1967) is usually contrasted with a positivist approach of qualitative research (cf. e.g. Gephart 2004 on the distinction between positivist and non-positivist/interpretive qualitative research). From an interpretive perspective, “truth” is considered as provisional, and social life as well as leadership as processual (Charmaz 2014: 17; Charmaz et al. 2018; cf. Endres and Weibler 2017: 217f.; cf. Astley 1985; Morgan and Smircich 1980). Hence, research “seeks interpretive understanding rather than a variable analysis that produces abstract generalizations” (Charmaz et al. 2018: 417; Suddaby 2006). Through attending to our participants’ perceptions, language, and experiences, our study sheds light on their views and practices of network leadership, and provides new theoretical insights on (non)leadership phenomena in collaborative interorganizational networks and on shared leadership development. We developed our findings through analytic induction by iteratively shifting between data-inspired induction and theory-inspired deduction for imaginative practice-oriented theorizing (Suddaby 2006; Charmaz 2014). In doing so, we rely on informed grounded theory, which is neither abstract (or naive) empiricism nor pure (subjectivist) relativism. Hence, grounded theory “is not an excuse to ignore the literature” (Suddaby 2006: 634; Charmaz et al. 2018: 419; Charmaz 2014). In line with this understanding, we applied “constant comparison” and “theoretical sampling”, which are considered core principles of grounded theorizing (Suddaby 2006; Charmaz 2014; Glaser and Strauss 1967). Constant comparison fundamentally implies an iterative cyclical research process in which the researcher shifts between collecting and analyzing data (see below for more details). Theoretical sampling means making decisions on what data to collect next on the basis of themes and concepts that emerged with theoretical relevance for the theory that is being constructed (cf. Charmaz 2014; Suddaby 2006: 634). This involves primarily “seeking and collecting data to elaborate the properties of the researcher’s theoretical categories but also to define variation within a category and to specify relations between categories” (Charmaz et al. 2018: 424). Applying this strategy helps researchers to stay focused, and prevents them from becoming overwhelmed by the amount of data and potentially relevant concepts (Charmaz et al. 2018). Hence, when applying these principles researchers break with “the myth of a clean separation between data collection and analysis” and replace the positivist notion of testing a hypothesis by the notion of ongoing interpretation and shifting between data and emerging theoretical categories (Suddaby 2006: 634; cf. Charmaz 2014).

We pursued our research in three collaborative interorganizational networks with member firms from the for-profit sector. We pseudonymized the networks as RedNet and BlueNet1/BlueNet 2 (referred to as “BlueNet” because we clustered these networks for empirical reasons). Given our research interest in interactions and emerging (informal) leadership activities, the fundamental criteria for inclusion in the sample were that all networks (1) have an ongoing activity, including regular face-to-face interactions; (2) have comparable organizing processes based on their similar collaborative non-hierarchical structure based on a lack of (a priori given) guidance structures, and (3) have a similar thematic focus (i.e. occupational/environmental protection, industrial safety/health, see Table 1 for more information about the networks’ descriptive characteristics).

We included a wide range of data types gathered from different sources. Table 2 contains a detailed description of data sources—interviews, observations of network meetings, document analysis—and includes information on how we used the data in our knowledge generation process.

The interviews, which included partly open-ended questions and partly focused questions, on the one hand were open enough to capture our participants’ experiences, and on the other hand permitted us to become more focused on emerging theory. The combination of “open-ended inquiry” and “focused attention” is a key characteristic of intensive interviewing (Charmaz 2014: 85ff). This strategy is recommended for grounded theory research, because it enables the researcher both to learn from participants what happens in the field and to generate “focused data” for developing theoretical enriched categories (Charmaz 2014: 87). In particular in the early research phase, and throughout the whole research process, we included open questions at the outset of each interview. Although our list of issues changed constantly to accommodate both emerging themes and our increasing understanding, we included similar questions on some issues to evaluate and better compare the themes. We started with a general request that participants recount experiences with network participation. We were particularly interested in concrete inter/actions and events, and commonly included questions about our participants’ perceptions of (1) network relationships and activities (e.g. meeting procedures, decision-making processes), (2) results achieved, (3) ways of organizing, and (4) network members’ motivations or goals for network participation. Similarly, coordinators were asked about network processes, their tasks, and activities. We further included a general request to comment on leadership if the subject had not been previously addressed. However, we did not explicitly ask about “identity”, but we were interested in our interviewees’ motivations, orientations, and beliefs about their network participation. We also conducted feedback dialogues (cf. Table 2).

We moved from empirical data to grounded theory by shifting between emerging empirical themes and theoretical conceptualizations (Suddaby 2006). We used three main types of grounded theory coding procedures: initial, focused, and theoretical (Charmaz 2014). During initial coding, we attempted to understand our participants’ views and (inter-)actions from their perspective. Whenever possible, we labeled our codes in our informants’ own language to give voice to their views. To remain open-minded and reflective, we used memo writing that included constructing maps and figures. Through several cycles reexamining initial codes, we generated first-order concepts (Van Maanen 1979) that captured our participants’ experiences with and interpretations of their network engagement. For example, we categorized according to the descriptions of network meetings, collective actions, perceptions about network leadership, coordinators’ functions and tasks, expectations toward coordinators’, and participants’ orientation toward and motivation for their network engagement.

Through focused coding, we sorted, synthesized, and categorized the large number of diverse initial codes. We compared participants’ experiences, actions, and interpretations across interviews and observations. For example, we compared observations of events (different workshop meetings and its sub-events such as factory tours or informal talks, or procedures during the meetings, e.g. setting the agenda, delivering a protocol) at different times and places (network meetings taking place in different locations/member firms). We further compared our own understanding of observed processes and events with those of our participants, in particular, when they differ. As our analysis progressed, we increasingly used theoretical coding steps to build increasingly abstract categories with enriched explanatory power. We specified already revealed relationships between codes through coding for process, which involves conceptualizing relationships between experiences and events, and finding out what actions or conditions affected the occurrence of identified processes (Charmaz 2014). We were interested in how the observed network activity or process emerged (e.g. developing a concrete project, collective decision-making), and how the participants’ emotions or actions (e.g. a member’s expert input, way of organizing by coordinators) worked to construct and change such activities and processes. We were sensitive to the meanings that different participants attributed to these processes, how they talked about the network, and what they emphasized (e.g. rejection of status-differentiation and not granting leadership to individual members). We attempted to be cognizant of possible idealization of, or hidden, and taken for granted assumptions about networks and network leadership. To understand the data at a more abstract level and to position it in the theoretical literature, we wrote memos about themes and emerging theoretical concepts. During theoretical coding, we further compared and integrated our memos and searched for relationships between categories and our emerging themes (cf. Charmaz et al. 2018; Charmaz 2014). We searched for theoretical concepts from the relevant literature that advance our understanding and might further develop our emerging categories, for example, defining variation within a category and specifying relations between categories, which represents a key idea of theoretical sampling (Charmaz 2014). That is, the grounded theory core principles of theoretical sampling and constant comparison that were introduced above were applied not only for choosing the network cases but also throughout the whole research process as we decided whom to interview next, what questions to ask, and which incidents, as reported in interviews, might be critical for our attention in subsequent observations.

Two of our participants, who held double network roles, also helped to provide a comparative perspective from their extensive experience in multiplex network engagements. Furthermore, our analysis benefited both from feedback from peers (i.e. from knowledgeable but detached colleagues) and from our interviewees, which enabled us to assess how our emerging theory made sense to our participants’ experience. Our analysis was supported by a qualitative data software program that helped us in storing our data, and codes in a stable format, as well as search for overlapping themes or differences in the data. However, it neither provided tools for automatic coding nor eliminated paper and pencil from the coding procedures. Figure 1 shows our data structure and summarizes how we moved from first-order concepts to second-order categories and to the theoretically aggregated main themes (cf. Clark et al. 2010, who inspired the formal presentation of our data structure).

Given our methodological approach, it is neither possible nor intended to describe all aspects of the empirical context (Charmaz 2014; Suddaby 2006). Rather, and by using the grounded theory key concept of contrasting and comparing, we focused on emerging empirical main themes and grounded theoretical key differentiation features. Because it makes the analysis more concrete and explanatory, a comparative approach has been acknowledged to be particularly useful for advanced qualitative leadership theorizing (cf. e.g. Denis et al. 2012; Chreim 2015).

The core network phenomenon that contributes to better understanding the described different leadership development is network identity. It emerged in a distinct manner in our networks, which were rather similar with regard to other aspects (e.g. overall purpose, general network structure, cf. Table 1). As illustrated in Fig. 2, two different identities emerged: a task-based network identity (BlueNet) and a joint-motivational network identity (RedNet). Table 3 summarizes the key differentiation features of BlueNet’s and RedNet’s network identity and provides representative illustrative interview quotations.

As our data showed, BlueNet members’ task-based network identity involves “individualistic network identity”, “single achievement motivation”, and “largely instrumentalist orientation towards network participation” (see Table 3, Figs. 1, 2). BlueNet members appreciated the open, respectful exchange on an equal footing in the meetings. Participants perceived their network as a collaborative entity, and thus developed an understanding of the network purpose. In part, however, a kind of wait-and-see orientation dominated the meetings. The following quote sheds light on how many of the BlueNet participants perceived and appreciated the collaboration in the workshop.

“A subject is discussed, everything is discussed intensively, but there’s no conclusion [in contrast to usual meetings that I’m familiar with in the company] … Or, what goes beyond this [beyond the actual meetings of workgroups], … And then the participants exchange opinions. Where someone starts by asking … Have you got any experience here? This is a group that I trust, where I know that can ask questions. That you just develop a network.”

BlueNet members often emphasized that they enjoyed taking part in the meetings because they can find out what other specialists think, and because they can use this for their own work in their companies. Hence, they focused on results for their “own company” —an attitude that points to a single achievement motivation. The largely instrumentalist orientation towards network participation, for example, appears from the fact that obviously only external pressure provided the motivation for using environmental management systems (cf. Table 3). Overall, the BlueNet members’ orientation toward network participation tended to be economic and rationalist (rather than idealistic), and primarily directed to achieving results for member firms only. Moreover, the BlueNet participants’ dominant perception of their network engagement was a “taking part” attitude that was related to the perception of “I as part of the network” (cf. Figure 1). This more of an entity view, which separates the individual member and the network as a whole, was a main feature of the BlueNet members’ individualistic network identity.

In contrast, RedNet members perceived themselves as part of a joint endeavor and displayed "collectivistic network identity”, “joint network motivation”, and a “largely value-laden attitude towards network participation" (see Table 3, Figs. 1, 2). A subtle clue that revealed the member organizations’ collectivistic identity was the tendency to refer to themselves as “we.” Because RedNet members developed a clear understanding about their common network purpose, they demonstrated a well-evaluated agreement on basic network positions in terms of self-commitment laid down in a “code of behavior, in which specific directions are stipulated, with regard to our subjects,” as one RedNet actor recalled. The participants’ agreement and shared understanding of overarching “subjects” was evident in their references to the subject in terms of an understanding of this theme that is taken for granted and of what the network stands for. In addition, they used the term “our subject,” indicating that RedNet members perceived themselves as a “large community”, bound together through their common concerns. RedNet actors metaphorically described their common ground and respective interdependencies: “We’re all in the same boat.” This “we feeling” accompanied the perception of increased solidarity. Concurrently, the network provided a platform “to bring our philosophy out externally much better through the network work,” as this RedNet actor stated. As we saw, the network became a means to transport values and express a specific identity for the members. On this basis, the members evaluated their network in terms of “how good we are as a collective.” Overall, they perceived RedNet as “a very good network” and that it is successful in transporting members’ values and philosophy (e.g. on environment management and sustainability).

Joint network motivation—the second component of RedNet’s identity means to be collectively motivated for common achievement and involves perceiving themselves as part of a joint endeavor. Hence, joint motivation is a collective level phenomenon: It captures the notion that the collective wants something and is motivated to achieve conjointly as a network. One indication that pointed to joint motivation was the frequent use of the phrase “we want” (e.g. “we want to promote”; “we want to see”). One participant emphasized that they wanted to pursue their concerns through joint endeavors in terms of “just the way we’re used to: network building, joint work on a joint subject and joint development of instruments and methods.” Another interviewee stated that the intention in publishing the yearbook (in terms of a documentation of the network’s activities, achievements, and purpose) was to develop a joint performance: “… but then really as joint companies, I mean in a joint performance and not each one for itself.” Thus, our interviewees were highly motivated to do something together, as the representative quotation in Table 3 shows. This attitude was related to the rejection of potential attempts to pursue micro-political or particularistic goals. As this participant recalled:

“What an individual wants, well, he can do that at home. And in the network, when people arrive at a common motive, they try to realize it.”

Thus, a fundamental logic behind these observations was the motivation to develop something collectively, as one RedNet actor’s quote showed: “In RedNet it’s all about growing something together.” In essence, our data showed that the actors perceived themselves as part of a joint endeavor. The competing environment was not seen as providing opportunities for joint achievement outside of the network. Further, we found an overall positive orientation toward the network in terms of a largely value-laden attitude towards network participation, which represents the third component of joint motivational network identity. The underlying attitude towards network participation went beyond an instrumentalist orientation. RedNet participants showed a significant commitment to the network theme. They were interested in doing something together, and they transcended single interests. The collective engagement was largely based on the conviction to do “something reasonable”, and many participants emphasized ideals and values, and thus were not guided by economic and shareholder value in the first place, as the representative quotation in Table 3 illustrates. Network members were fundamentally interested in value-related and ethical themes, as this interviewee noted:

“We make sure that we always invite speakers who work through the topic in depth. I mean, that you take this step into the world behind the material. That you include philosophical and moral topics as well.”

Many participants illustrated this topic by reporting on an enthusiastic engagement for value-laden, concrete projects (e.g. fair trade) and that were considered to have a certain “societal relevance” (e.g. ecological thinking and sustainable development). This focus on the common good, or at least the welfare of the greater collective, emerged as a core motive for establishing network relationships rather than building intensive dyadic cooperation among the members.

The key differentiation feature in the network coordinators’ role and practices relates to the way in which they shape their members’ network identity. It should be noted that—in line with our methodological approach—it is neither our intention nor is it necessary to provide a description of the whole set of functions and practices that are necessary for ensuring the networks’ functioning (e.g. general administrative tasks). Rather, we focus on the functions and practices that were related to the emerging leadership themes and identity pattern (i.e. either the development of a task-based identity or a joint-motivational identity). As depicted in Fig. 1, in BlueNet the coordinators’ particular main practices were guided by legitimization efforts. This involves creating concrete (practical) value for member firms to make these network advantages visible to the member organizations. Such practices were considered as key facilitators of the members’ task-based network identity. BlueNet participants emphasized, for example, the need to develop a knowledge data base (e.g. on the basis of “worksheets”, which were regarded as very useful for working on potential or actual existing problems) to evidence the network benefit for member firms and to justify network participation. Further, the BlueNet coordinators stressed the importance of making the network as such perceivable for their members by developing a corporate identity (e.g. a network logo).

In contrast, RedNet coordinators’ activities facilitated joint-motivational network identity by the preparation of context and processes for identity enactment (cf. Fig. 1). This involves an inclusive, open-minded, and heedful moderation without “prefabricated opinions” to give “everyone an opportunity to keep up”, and “to meet people halfway as well” (RedNet actor).The coordinators summarized different knowledge and meanings, promoted collective sense making, and were engaged in sense giving in that, for example, they tried to show “cross-connections” of subjects, or provided input from an overarching perspective. However, such inputs were provided more as consultation than as compulsion, as this coordinator recalled: “Often a vote isn’t taken; there’s more of a formation of opinion.” The coordinators noted that they tried to include space for identity expression and forged the network as an identity enactment scene to support the emergence of collectively shared stories and ideals. In doing so, coordinators sensibly shifted between moderating open-ended, collective learning processes and forging consent or decision-making. Interestingly, however, decisions were usually not made through a democratic vote, as observed by this coordinator: “Voting by show of hands is very rare.” Rather, to capture emerging consent or still existing diverging individual opinions, the coordinator relies on the perceived mood or atmosphere: “It’s more of a feeling if we’ve addressed a topic, discussed it. There is an unspoken consensus, accordance, without things having to be voted on.” In the meetings, the coordinator looked around the group, made eye contact with those present, and attempted to capture the mood to extract a line that allowed him to continue his work. Overall, the coordinators’ practices were related to expert knowledge and relational skills, such as mindful listening and sensitivity to the mood within network settings. Given the crucial tasks of coordinators, it became apparent that networks were not “self-organizing”. As one coordinator reported, he often thought about a stronger form of “self-organization”, but also realized that such an approach would not work—an observation that he also reflected and confirmed in the feedback dialogue. Finally, a pleasant ambiance and an emotionally positive climate appeared to be relevant, as for example this interviewee observed: “Everything was combined with a certain amount of cheerfulness and ease.” Another participant saw the positive climate of cooperation as stemming partly from “this step into the world behind the material, that you include philosophical and moral topics as well,” and finally noted, “In the RedNet, we have a philosophical component as well.”

Our findings contribute to better understanding of leadership phenomena in collaborative interorganizational networks, in particular concerning leadership at the neglected (interpersonal) interaction level without disregarding relevant conditions related to the network level (e.g. Huxham and Vangen 2000; Feyerherm 1994; Denis et al. 2001; Ospina and Saz-Carranza 2010; Weibler and Rohn-Endres 2010; Miles et al. 2010; Müller-Seitz 2012; cf. our literature review). An initial finding of our fieldwork is that the networks either developed shared leadership or cooperated “leaderless” (i.e. established a leadership void). Our fieldwork shows that embeddedness in a peer-like work setting based on the lack of a priori existing formal guidance structures may significantly affect participants’ views on how to be influential and how to “lead” or “be led” in such a way that single leader emergence—in terms of direct or individual leadership—is thwarted in all of our network cases. For example, we found a sense of network embeddedness among the participants that involved rejection of status differentiation. Regarding leadership, this means that even an individual’s potential higher status was not seen as providing the source of accepted influence necessary to grant leadership to such an individual. These findings are striking and counterintuitive, because it might be expected that a person who—for good reason, such as significant expertise and engagement—“stands out objectively” (as emphasized by a network participant), would emerge as a leader (Stone and Cooper 2009). Indeed, general research on leader emergence showed that an individual emerges as a leader based on “credits” ascribed to him or her by the other group members, for example, the individual’s high competence or outstanding contribution to the achievement of group goals (e.g. Hollander 1974; Stone and Cooper 2009 for an overview). In contrast, our work shows that this is not the case and suggests that organizational contexts actually exist in which participants, such as those in our networks, successfully strive to maintain the informal peer-like setting, and collaborate in a way that no single individual is granted a leading influence.

Our study concretizes this phenomenon by adding that networks are not necessarily “leaderless” or a “leadership-free zone”. Rather, leadership is vitally needed; however, it emerged only in the collective form. Our fieldwork showed that, given embeddedness in a peer-like networked context, participants developed context-sensitive alternative (i.e. non-individualistic and more relational) conceptions of being influential, which—under certain circumstances (as described in our model)—led to collective efforts by the network as a whole (i.e. network drive and shared leadership development). These aspects of our findings echo the relevance of plural leadership forms, such as collective or shared leadership, in collaborative interorganizational networks (Denis et al. 2001; Huxham and Vangen 2000; Ospina and Saz-Carranza 2010; Weibler and Rohn-Endres 2010; White et al. 2016).

Moreover, we have identified distinct types of network identity as a significant condition for the emergence of the observed different network leadership phenomena. As described in our model, a task-based network identity is related to a leadership void, whereas joint-motivational network identity is related to shared leadership development. Since identity emerged as a key concept for better understanding network leadership phenomena in our context, we draw on literature on identity and developed two network-specific identity conceptions. Before we continue to discuss our theoretical contribution in this regard, we provide information on the concept of identity relevant for our study’s context and contribution.

In keeping with our social constructionist leadership approach, we subscribe to an understanding of identities as socially constructed phenomena that are dynamic and fluid in nature (Ashforth et al. 2008, 2011; Gioia et al. 2013). In line with Ashforth et al. (2008: 327), we refer to identity as “a self-referential description that provides contextually appropriate answers to the question” Who am I as individual? (individual identity) or Who are we as organization? (organizational identity) (Ashforth et al. 2008; Corley et al. 2006). According to Corley et al. (2006: 87), organizational identity refers to a “self-referential meaning where the self is the collective” comprising self-definition, values, and affect as its core. From a social constructionist perspective, an organizational identity is seen as being defined by the members of an organization “to articulate who they are as an organization” and, thus, primarily involves “the labels and meanings that members use to describe themselves and their core attributes” (Gioia et al. 2013: 127). In our study’s context, the network’s identity also involves a collective-level motivational concept (i.e., “joint motivation” as contrasted with “single achievement motivation”). We, therefore, labelled this newly developed identity concept a “joint-motivational network identity”.

Our study inductively linked a collective or network level identity (i.e. in terms of an identity held by the network; cf. Corley et al. 2006; Ashforth et al. 2011) with (collective) leadership emergence. In contrast, research on leadership identity still focuses on the individual level of identity (i.e. in terms of an identity held by an individual). A leader identity can be defined as a “sub-component of one’s identity that relates to being a leader or how one thinks of oneself as a leader” (Day and Harrison 2007: 365; cf. Epitropaki et al. 2017). Because of the emerging network-specific nature of the identity constructions found in our investigation, we add to this understanding the notion of leader-and-follower identities as contextualized phenomena emerging from social interaction processes (Collinson 2006; DeRue and Ashford 2010). However, even this interaction-oriented leader-identity perspective usually conceptualizes single leader emergence (i.e. the development of either a leader or a follower identity). In this context, leadership identity researchers traditionally emphasize that (effective) leadership emergence requires clear role differentiation and clarity in individuals’ identities as leader and follower. For example, DeRue and Ashford (2010) highlight that strong and effective leader–follower relationships imply that there is clarity in the individuals’ identities as leader and follower. They describe a process by which individuals jointly co-create their respective identities as leaders and followers—via processes of claiming and granting leadership among each other—and thereby construct a leadership relationship. Overall, and although considering relevant dynamics of the social construction of leadership, they remain at the individual-level of identity as the source of leadership, in that they focused on the differentiation of either a leader or a follower identity. Similarly, Lindgren and Packendorff (2011) referred to individual-level identity concepts when studying the emergence of shared leadership.

Hence, even when plural forms of leadership are studied, such as collective or shared leadership, existing leadership identity approaches neglect organizational or collective level identity conceptions (i.e. in terms of an identity hold by an organization or network) as a (potential) source of leadership emergence (cf. Epitropaki et al. 2017). That is, existing leadership identity approaches usually do not consider the possibility that neither a leader nor a follower identity will emerge. Hence, DeRue and Ashford (2010: 642) finally called for further research on leadership identity that extends their work “by considering the process [of identity construction] at the group level” and on what happens when “less emergence of a well-defined leadership identity” might probably occur in certain contexts. This is indeed what we empirically found in our contexts, in that our data did not bear out an identity differentiation between leader and follower. Moreover, we were able to move beyond that by revealing that the network participants developed a joint-motivational network identity. This network-identity fueled the cultivation of alternative conceptions and practices of being influential as a network collective. On this basis, network participants developed network drive in terms of collective efforts and thus finally enacted shared leadership. Overall, we advance the study on collaborative interorganizational networks by elucidating relevant network internal functioning mechanisms related to (non)leadership phenomena, and in particular by revealing the collective nature of network leadership, and the joint-motivational network identity-based energizing source of its emergence.

Collaborative interorganizational networks have recently been considered to be particularly fruitful contexts for examining emerging shared leadership—not at least given their lack of formally (a priori given) guidance structures. For example, Sergi et al. (2012) argued that these settings provide “fertile grounds for exploring the ins and outs of collectivistic approaches of leadership” (2012: 406; cf. Denis et al. 2012; Endres and Weibler 2017). Since, in our context, shared leadership processes emerged empirically, we draw on literature on shared leadership study and relate our emerging findings to this field. Shared leadership is a plural form of leadership which involves leadership being collectively shared among different people, and constructed in interaction (Avolio et al. 2009; Denis et al. 2012; Nicolaides et al. 2014). Shared leadership fundamentally has been defined as “an emergent state where team members collectively lead each other” (Avolio et al. 2009: 431; cf. Pearce and Conger 2003; Day et al. 2004). Hence, shared leadership may not be localized in any individual who conducts a superior role; rather it is a collective-level phenomenon which is embedded in a web of dynamic relationships (Avolio et al. 2009; Endres and Weibler 2017; Fitzsimons et al. 2011). Shared leadership (and related plural leadership forms such as collective and distributed leadership) is among the most momentous emerging themes in leadership research (cf. for recent overviews Denis et al. 2012; Fitzsimons et al. 2011; Dust and Ziegert 2016; Nicolaides et al. 2014; Endres and Weibler 2017). Given these instructive overviews, the discussion on the various forms and perspectives is not repeated here. Instead, and in keeping with our methodological/paradigmatic foundation, we draw on perspectives of shared leadership that study leadership as socially constructed through interactions, and as embedded in a web of relationships and context (Endres and Weibler 2017; Sergi et al. 2012; Fitzsimons et al. 2011; Uhl-Bien 2006; Hosking 1988 cf. our theory section). Pursuant to extant literature, we use the terms shared leadership—as characterized above—and collective leadership interchangeably (cf. Avolio et al. 2009: 431; Nicolaides et al. 2014); however, we point to their socially constructed and relational basis (Endres and Weibler 2017; Fitzsimons et al. 2011; Sergi et al. 2012).

Despite the considerable amount of insightful contributions, a key problem in shared leadership study is still how to grasp and conceptualize emerging shared leadership processes that are embedded in interaction and relationship dynamics among individuals, and thus on how to distinguish such plural leadership forms from other interaction related collective level dynamics, such as cooperative teamwork or similar (Denis et al. 2012; Endres and Weibler 2017; Sergi et al. 2012; Crevani et al. 2010). Hence there is a need to better understand how shared leadership actually occurs. Therefore, Endres and Weibler (2017: 231) called for “more explorative studies that approach leadership phenomena inductively, and thus do not start with ‘leadership’ or what is assumed to be leadership, but rather with interaction dynamics and practices as they occur in a specific setting”, in particular, when trying to study plural forms of leadership. That is, the emergence of those demanding collective level leadership forms should be empirically evidenced. Based on their critical interpretive synthesis of research on social constructionist leadership, Endres and Weibler (2017) argue that the consideration of influence processes in terms of leadership manifestation may help to reduce the risk of diluting the distinctiveness of leadership. Basically, this means, conceptualizing a phenomenon as leadership involves the identification of emerging flows of influence processes at either the interpersonal or collective level (or both). This, however, is particularly challenging for the collective level, i.e. for identifying shared leadership phenomenon, a constellation in which no single individual is identifiable as the source of influence, rather the collective is. Hence, a key question is how to make such elusive dynamics such as emerging (informal and collective) leadership tangible if no single individual’s influence emerges as network leadership (Blom and Alvesson 2015; Denis et al. 2012). For that reason, in order to qualify general collective interaction related dynamics as leadership, in place of (traditional) individual level influence components, collective level components are needed to capture a collective’s influence and to understand the generative mechanism of its emergence. We address both of these problems. Our emerging concepts’ main contribution here is twofold: first, the cultivation and enactment of network-sensitive conceptions of being influential, and in particular the finally emerging concept of network drive in terms of collective efforts as a network, capture the influence component at the collective level in place of the missing leadership-related influence activities from single individuals. These emerging flows of influence represent the collective “leadership manifestation”, which is a conceptual prerequisite for collective or shared leadership (Endres and Weibler 2017: 222). Second, joint-motivational network identity with its components collectivistic network identity, a largely value-laden attitude towards network participation, and joint network motivation, captures the underlying energizing source of shared leadership development. On the basis of our study’s findings, we offer advancements in a more precise theoretical framing and study of plural leadership forms, such as collective or shared leadership, than existing literature has done (e.g. Crevani et al. 2010; Lindgren and Packendorff 2011; Huxham and Vangen 2000; Denis et al. 2012). We suggest setting a high bar for conceptualizing collective phenomena in terms of shared or collective leadership (cf. Endres and Weibler 2017; Sergi et al. 2012; Blom and Alvesson 2015)—that is, to identify collective level influence processes and the joint-motivational identity-based (energizing) source of its emergence in the respective context. This implies taking the non-emergence of leadership more seriously into consideration. Our study provides a rich description of, and new insight into, non-leadership (i.e. a leadership void in one of our networks in which no single participant emerged as a leader, nor did the collective become engaged in leadership processes). Therefore, our findings echo the relevance of non-leadership phenomena for understanding leadership in contemporary organizations more comprehensively (cf. Alvesson and Sveningsson 2003; Alvesson and Spicer 2012; Blom and Alvesson 2015; Chreim 2015).

Collaborative interorganizational networks provide a challenging context for “leaders” in traditional terms, since single leadership—as the still dominant and widely spread leadership understanding in the business and organization context—obviously seems not to be the way of leadership that fits into this context. This applies even despite the finding that leadership is called for and considered as crucial for fruitful forms of network participation. Our findings suggest that network coordinators or individuals with similar roles should understand that, in the network context, participants feel significant, yet sometimes implicit, aversions against single leadership (e.g. based on the observation that status differentiation is rejected) and, more importantly, that the networks probably either develop shared leadership or established a leadership void (i.e. cooperated leaderless). Hence, network coordinators or managers should accept being sidelined, whereas network participants as a collective clearly gain center stage, and may develop shared leadership. However, this will occur only under certain circumstances, which relate to the described identity-based energizing source of its emergence. Our study’s findings suggest that the revealed processes of shared leadership development are neither self-organizing nor another type of self-leadership. We argue that shared leadership may be fostered through a context-sensitive influence on members’ network identity. For example, network identity-related processes may be forged in a way that they may transcend members’ perception of “I as part of the network” and develop a sense of “we as a network” (i.e. in terms of a joint-motivational network identity). This implies, for example, moving beyond evidencing the benefits of the network for the members in a dominant instrumentalist way (cf. the BlueNet case). Rather, coordinators should consider becoming engaged in creating a kind of open space in terms of an identity enactment scene, which allows exploring commonly held and identity enhancing ideals and values (cf. Wellman 2017; Kourti 2017). To create such a space, it is important to understand—besides meeting the structural and administrative requirements for network functioning—that interactions must be nourished to acquire the adequate dynamics within networks (cf. Larson and Wikström 2007). This implies maintaining an informal atmosphere to forge an open and unbiased exchange that is respectful and non-judgmental, and that may ease emotional expression, and to move beyond a rational and pure instrumentalist attitude towards the network. Our study’s findings concerning the importance of an informal atmosphere without status differentiation echo recent theorizing by Wellman (2017: 612), who, for example, suggests that “downplaying formal job titles” among other things may support shared leadership development among group members. Further, communication styles should relax from focused efforts to convince others. Rather, curiosity and a learning orientation should be fostered along with listening, which might include engaging in sense giving (Clark et al. 2010) and forging a learning environment (Weibler and Rohn-Endres 2010).

There may be further possibilities in framing a network coordinator’s function and roles (e.g. network coordinative and managing functions, such as acting as relationship manager, integrating and connecting participants). During our fieldwork, we intended to understand whether this may be relevant; however, we finally found that these aspects of the network coordinators’ functions were less relevant in understanding the specificity of the leadership-related phenomena in our networks. Rather, the identity-related themes seem to provide more ample scope for supporting shared leadership development. Hence we focused on these aspects, a procedure which is in line with our methodology; it is neither intended nor possible to provide a complete description of emerging empirically grounded phenomena (cf. Charmaz 2014; Suddaby 2006).

Our contributions must be viewed in the light of potential limitations of our study. First, it is possible that our participants withheld information from us, particularly information that might cast a negative light on their values and motivations for network participation or their ways of exerting influence. However, this problem has probably been mitigated, because we considered multiple perspectives from different data sources, and we looked for coherence in our data (Charmaz 2014). For instance, in the early phase of our research we were surprised by the somewhat striking “harmony” in the network settings. We tried to better understand this phenomenon and learned successively that our participants developed a certain and basically very relaxed attitude toward consensus finding and included this as a subcategory in our analysis. Because nobody is compelled to do anything in the network, the collective (leadership) processes that were revealed and appeared to be extremely harmonious are not “always present”, and not are easily achieved. Shared leadership will only emerge “voluntarily” based on participants’ ideas, values, and actual concerns. Thus, even though differences may exist, they do not necessarily always have to be fully resolved. Second, it was neither possible nor intended to provide an all-encompassing description of facilitators’ and interaction dynamics, e.g. related to network evolution or relationship development. Rather, and in line with our grounded theory approach (and its core procedures of constant comparison and theoretical sampling), which implies focused coding and selections, we focused on inductively derived themes that more accurately advance understanding the focal processes of non-leadership emergence and shared leadership development, as summarized in our model. Nevertheless, and third, we recognize that our suggestions with regard to the network level outcomes are an initial step. That is, given our network-centered perspective the assessment of outcomes focuses on the network level and neglects other possible outcomes. Therefore, future research should expand this perspective and assess (ideally various forms of) outcomes at the member organizations’ level. Fourth, although our study’s findings include individual level dimensions in terms of value orientation and (motivational) orientations towards network engagement, further individual-level/intra-psychical or intrapersonal characteristics such as our participants’ personality structures are potentially also relevant and thus deserve further investigation. Finally, concerning the transferability of findings to other contexts, we described the contextual conditions and mapped the domain of our network cases. Based on this, we argue that our suggestions for shared leadership development are limited in their applicability to traditional organization contexts; rather they apply to the core domain of collaborative non-hierarchical network contexts and similar cooperative and low structured organizational settings (e.g. self-managing and self-governing teams/organizations). Future research should further develop our key concepts in relation to other contexts. Beyond that, our grounded theory model provides ample scope for studying various relations in depth. For example, future research should explore in more detail the content of the values and orientations related to joint-motivational network identity. This should involve the study of potential “negative values”, which—in contrast to the positive values of network members in our context (cf. RedNet)—would probably not contribute to the common good. Concerning the motivational dimensions of our model, it would be promising to study in more detail how probably existing participants’ individual level motivation could be lifted to the (collective) network level.