Temporal dynamics of shared leadership, team workload, and collective team member well-being: a daily diary study

ABSTRACT In this diary study, we consider shared leadership and team workload as antecedents of team mental health. We draw on conservation of resources theory to theorize how linear change trajectories of shared leadership are related to change trajectories in team members’ shared well-being and emotional exhaustion. Furthermore, we investigate the interaction between change trajectories of shared leadership and team workload, predicting that change in shared leadership will be more strongly related to change in team mental health when team workload increases. 265 team members nested in 77 teams completed a daily diary survey over five consecutive workdays. As hypothesized, an increase in shared leadership was associated with an increase in team well-being and a decrease in emotional exhaustion over time. Further, shared leadership interacted with team workload, such that an increase in shared leadership was more strongly associated with a decrease in shared emotional exhaustion when team workload increased. However, team member well-being was not affected by such an interaction. These findings address the missing link between shared leadership and team well-being and exhaustion, establish shared leadership as an important team resource, and contribute a temporal perspective on shared leadership as a dynamic team phenomenon.

Mental health impairment among employees is a critical challenge for organizations (e.g., Bliese et al., 2017;Christian et al., 2011;Swider & Zimmerman, 2010).For example, in Germany, where the sample for this study is based, mental health impairment is the number one reason for the inability to work (Krankenkasse, 2020).Research has provided several insights into which individual-level factors are related to mental health (e.g., Bakker & Demerouti, 2017).However, teamwork becomes increasingly prevalent in organizations and working in teams is impactful in shaping team members' attitudes and behaviours (e.g., Mathieu et al., 2017), and, thus, might be highly relevant for their shared mental health.Hence, there is a high need to consider mental health from a team-level perspective and as a shared phenomenon among team members (e.g., Bakker & Demerouti, 2017;Hobfoll et al., 2018;Maslach et al., 2001;Ryan & Deci, 2001;Sonnentag, 2015).In this regard, we define team mental health as the shared experience of mental health of the team members resulting from working and interacting in a shared team environment.
Prior research shows that the phenomenon of mental health is highly dynamic and changes over time (for an overview, see Sonnentag, 2015).Such changes can occur due to short-term developments such as increasing or decreasing resources at work (e.g., Bakker, 2014;Hobfoll et al., 2018;Sonnentag et al., 2023).At the same time, teams are also highly dynamic entities (e.g., Collins et al., 2016;Kozlowski, 2015), and possible teamlevel antecedents of team mental health should be subject to change.Therefore, we need to account for temporal dynamics in mental health states as well as its team-level antecedents in considering the key question what team members can collectively do to enhance their shared mental health.
Addressing this question is highly important for both theory and practice.From a theoretical perspective, this study complements the understanding of individual-level antecedents of mental health with a team-level perspective and explains what teams can actively do to contribute to their mental health (cf.Sonnentag et al., 2023).Additionally, teams themselves may be sensitive to short-term changes in their environment (e.g., decreasing resources), where an active, immediate response from within the team may be needed (cf.Hobfoll, 2011;Stoverink et al., 2020).From a practical perspective, the possibility of team-level antecedents of employee mental health, as well as the notion of collective team mental health, are important in light of the high prevalence of teamwork as an organizing principle (e.g., Mathieu et al., 2019).Furthermore, a team-based perspective of dynamic relationships between mental health and its antecedents implies that not only single individuals, but the team as a whole may benefit from increasing shared mental health.In particular, we focus on shared leadership as a potential antecedent of team mental health, given its high relevance for team functioning (e.g., D'Innocenzo et al., 2016;Zhu et al., 2018).
Shared leadership has been defined as a "set of interactive influence processes in which team leadership functions are voluntarily shared among internal team members in the pursuit of team goals" (Nicolaides et al., 2014, p. 924).Hence, team members can actively, informally, and voluntarily exert leadership influence or participate in team leadership functions (e.g., structuring and planning tasks, providing feedback, solving problems, and supporting the team climate; Morgeson et al., 2010).As theoretical background, we draw on conservation of resources theory (COR theory;Hobfoll, 1989).COR theory states that individuals and groups (i.e., teams) strive to acquire and conserve resources (Hobfoll et al., 2018, p. 106).Shared leadership has been conceptualized as a team resource (Day et al., 2004), as it can facilitate collaboration and team goal achievement (Day et al., 2004;Pearce et al., 2003).This resource-based view of shared leadership can move beyond the predominant focus on performance-related outcomes in past research (Zhu et al., 2018) and broaden our understanding of its consequences for team mental health.
In considering how and why shared leadership is related to team mental health, we pursue a temporal lens, as both have been considered as highly dynamic (for shared leadership, see e.g., Day et al., 2004;Klasmeier & Rowold, 2022;Pearce et al., 2003; for mental health, see e.g.;Bakker, 2014;Sonnentag, 2015).These dynamics are also emphasized by COR theory in terms of gain and loss cycles of resources (Halbesleben et al., 2014), which may manifest as linear change trajectories indicating a decrease or an increase of resources.This view advances our understanding of shared leadership dynamics with temporal precision (cf. McCormick et al., 2020) in terms of how and over what period shared leadership may change over time: Considering the work week as a repeating temporal structure (cf.Ancona et al., 2001;Arrow et al., 2004), a linear increase in shared leadership may evolve as a result of an improving interaction between the team members, as they accomplish tasks and achieve team goals during the week (cf.McClean et al., 2019).Applying this perspective in a daily diary study (Gabriel et al., 2019), we can also provide an empirical test of COR theory's assumption of resource cycles.Zhu et al. (2018) assumed that a higher engagement in shared leadership may be less appropriate under demanding conditions indicating a resource loss.In contrast, COR theory states that engaging in resource gains (e.g., increasing shared leadership) should become most salient in the context of resource loss (Hobfoll et al., 2018).To resolve these divergent views, we consider team workload as a boundary condition to understand when the additional engagement of the team members (i.e., increasing shared leadership) can be beneficial or detrimental for their mental health.Specifically, an increasing team workload can be critical for teams by triggering a resource loss cycle due to a growing need to coordinate and to reflect on teamwork (Dietz et al., 2017), which may decrease team mental health (Hudson & Shen, 2018;Maslach et al., 2001).To counterbalance this loss cycle, increasing shared leadership (i.e., resource gains) may become most salient (cf.Hobfoll et al., 2018).Thus, we aim to study the dynamic interplay between shared leadership and team workload in relation to team mental health.
In sum, this study offers the following contributions.First, we bridge the gap between the well-being literature and the team literature by considering team-level demands and resources, as well as their interaction in relation to team mental health.In connecting these streams of literature, we shed light on the underdeveloped but growing area of research on shared mental health in teams, move theory from a strong focus on work conditions to the active contribution of collectives on their mental health development (cf.Sonnentag et al., 2023), and extend empirical evidence on the applicability of COR theory beyond the individual level (cf.Hobfoll et al., 2018).Second, we expand the nomological network of shared leadership.Specifically, our theorizing and corresponding empirical findings can expand the predominant focus on performance outcomes of shared leadership in prior research (for an overview, see Zhu et al., 2018) and underscore the relevance of shared leadership as a team resource with benefits for mental health outcomes as well.Additionally, considering the interaction between shared leadership and team workload can provide meaningful insights into the boundary conditions that delineate when shared leadership may be particularly helpful for teams.Third, we develop a dynamic perspective on shared leadership to examine temporal changes in shared leadership and associated team mental health over time in a daily diary study.This allows us to contribute to the emerging literature on the dynamics of shared leadership and mental health by advancing theory with temporal precision regarding how and when shared leadership and team mental health change over time (cf. McCormick et al., 2020).In doing so, we can provide empirical insights into the yet underdeveloped research on resource gain and loss cycles of COR theory.

Conceptualizing the team-level constructs
When employees collaborate in a team, they interact on a daily basis, and they are exposed to the same environment and shared conditions.As a result, they tend to develop similar moods, perceptions, beliefs, and behavioural patterns (e.g., Lehmann-Willenbrock et al., 2011;Myers et al., 2004;van Yperen & Snijders, 2000).Methodologically, this is reflected in the use of multilevel data (i.e., individual perceptions of team members working in teams).As our study has a particular focus on the team level, it is necessary to explain how we conceptualize, define, measure, and form (i.e., by describing the process of emergence) our focal constructs (see e.g., González-Romá & Hernández, 2022;Klein & Kozlowski, 2000).In short, the emergence process is a composition (i.e., aggregation of the individual-level scores to the team level) approach (Chan, 1998).Particularly, we use a referent-shift consensus model for shared leadership and team workload, whereas we used a directconsensus model for team mental health.In the following, we will explain these choices in more detail.
Regarding shared leadership, Zhu et al. (2018) identified three key characteristics to conceptualize this form of team leadership.First, the source of leadership concerns the team members who exert voluntary, lateral leadership influence among each other and take on leadership functions.Second, the unit of analysis when investigating shared leadership is the team level, as shared leadership is the collective leadership influence of all team members.Third, the distribution of leadership influence is widely dispersed across all team members and is not centralized in a single person.Hence, leadership influence stems from the collective engagement of all team members and is, thus, a collective property of the team (Carson et al., 2007;Klasmeier & Rowold, 2020;Zhu et al., 2018).Accordingly, we used a referent-shift consensus model to measure shared leadership (cf.Chan, 1998).
A high level of workload represents the experience of too much work to do in relation to the time available (Sonnentag et al., 2010).Workload has mostly been investigated as an individual-level job demand, but theoretical and empirical work shift the point of view to a team-level conceptualization of workload (cf.Maslach et al., 2001).According to Funke et al. (2012), team workload is not just the sum of each team members' individual workload, but is a collective property of the team that is influenced by the team's interaction patterns, team cognition, and team environment.Hence, we defined team workload as the collective experience of too much work to do in relation to the time available in teams.Based on this, we used a referent shift-consensus model (Chan, 1998;Klein & Kozlowski, 2000), which has also been used in previous studies about team-level demands (Hudson & Shen, 2018).
In contrast to the former, team mental health is not a collective team property in terms of a collective experience or perception, but it is the shared experience of mental health among the team members resulting from the experience of working and interacting in a shared environment.This rationale builds on the assumption that inner-psychological states or experiences are most adequately captured by the focal individuals (cf.Gabriel et al., 2019).Whereas individuals can likely have a valid and accurate representation of their own individual mental health (i.e., referent is the individual), it may be very difficult to have such a representation of the collective experience of mental health among all team members (i.e., when referent is the team as a whole).These individual representations of mental health are partly shaped by the shared work experience (i.e., shared conditions and environment).Hence, team members may develop a similar or shared experience of mental health, which can be aggregated to the team level.Accordingly, we used a direct-consensus model to measure team mental health as the shared experience of mental health among the team members (cf.Chan, 1998).Following metaanalytical insights and conclusions (Montano et al., 2017), we considered positive (i.e., subjective well-being) as well as negative (i.e., emotional exhaustion) aspects of teams' mental health to cover a wider range of mental health states.Whereas emotional exhaustion as a core dimension of burnout refers to the state of physical and emotional depletion (Wright & Cropanzano, 1998), subjective well-being goes beyond the mere absence of (mental) illness and is defined as the optimal psychological functioning (Ryan & Deci, 2001).

Shared leadership: a dynamic perspective
Shared leadership is a team phenomenon that develops and changes over time, based on team members' day-to-day interactions (e.g., Kozlowski et al., 2016;Pearce et al., 2003), and therefore needs to be considered as highly dynamic (e.g., Carson et al., 2007;D'Innocenzo et al., 2016;Klasmeier & Rowold, 2022).However, most research is based on crosssectional designs that cannot speak to dynamics in shared leadership.To address this issue, diary studies offer the opportunity to examine short-term temporal changes in focal constructs (e.g., Gabriel et al., 2019;Ohly et al., 2010) and to study team dynamics (Kozlowski, 2015).
Diary studies found that day-specific dynamics account for a high amount of variance of leadership behaviour (Johnson et al., 2012).As the team is a typical source of leadership behaviour (Morgeson et al., 2010), it is plausible that shared leadership may also display short-term dynamics across days (see Klasmeier & Rowold, 2022).Dynamics in leadership behaviour can be unsystematic (i.e., day-level fluctuations) or systematic (i.e., linear trajectories; McClean et al., 2019).While the former may be more spontaneous or due to unpredicted events, the latter may occur as continuous adaptation to longer-lasting events like different stages of project work, task cycles, teamwork and team effectiveness episodes, changes in the team environment (e.g., Arrow et al., 2004;Collins et al., 2016;Lorinkova & Bartol, 2021;Marks et al., 2001), or due to accumulative experience (McClean et al., 2019).Theories of entrainment assume that behaviour and experience are aligned with temporal schedules or repeating cycles that occur due to social norms or organizational processes (Ancona et al., 2001).A typical and impactful repeating temporal cycles is the work week structuring the work and shaping workrelated behaviour (e.g., Ancona et al., 2001;Dust et al., 2022).Thus, we focus on systematic short-term dynamics as linear change trajectories (i.e., continuous rate of change indicating a decrease to an increase over a work week).This perspective is also supported by COR theory's notion of resource gain and loss cycles (Hobfoll et al., 2018).As a dynamic theory, COR assumes that initial resource gains or losses can lead to subsequent resource gains or losses, resulting in a temporal sequence of accumulating resource gains or losses (Halbesleben et al., 2014).
Shared leadership can be utilized as a strategic means by the team to allocate their resources and effort while facing current demands and satisfying the team needs (Morgeson et al., 2010).Depending on the (un)successful adaptation and resulting engagement in different leadership functions by the team, we assume that teams may differ in their change patterns across a work week (i.e., have variability in change over time; cf.Collins et al., 2016;Ployhart & Vandenberg, 2010).For example, a team might start the week by planning tasks and anticipating team needs throughout the work week.Based on this, the team members' interactions evolve, positive collaborations accumulate, and the team shows progress in attaining common goals across the week, resulting in a linear increase in shared leadership.Another team is faced with more routine tasks during the week, so that shared leadership stagnates (i.e., shows no linear changes).Finally, a third team experiences disagreement about the current needs and weekly goals, inhibiting a successful collaboration and resulting in emerging conflicts over the course of the week.In this case, shared leadership should decrease.Beyond the team level, individual-level factors may also shape the temporal development of shared leadership across the week.For example, recent research found that motivation and performance significantly changed in a linear fashion across the work week (Dust et al., 2022).This may shape the interactions among team members, with implications for the development of shared leadership over time.
In sum, some teams might show a linear increase in shared leadership, while other teams may stagnate or show a linear decrease in their level of shared leadership over time (cf. McClean et al., 2019; see also Kozlowski, 2015).These differences in linear change trajectories between teams can be used to predict change in focal team outcomes (Chen et al., 2011;Ployhart & Vandenberg, 2010).Hence, it is reasonable to consider a defined and continuous period of time (i.e., a work week; cf.Ancona et al., 2001;Arrow et al., 2004) to better compare how different change trajectories relate to changes in our focal outcomes.Additionally, a work week (i.e., from Monday to Friday) is also a reasonable period of time for investigating temporal dynamics of team mental health, as recovery processes over the weekend can influence their dynamics (e.g., Fritz et al., 2010).Next, we elaborate why change trajectories in shared leadership may be linked to dynamic changes in team mental health.

Shared leadership as a resource for team mental health
We draw on the central tenets of COR theory, which proposes that individuals and collectives (i.e., teams or organizations) strive to acquire and conserve resources (Hobfoll, 1989(Hobfoll, , 2011;;Hobfoll et al., 2018).Based on this basic principle, stress and impaired well-being result when resources are threatened by loss, an actual resource loss or outstretch of resources occurs, or in case of a failed resource investment (e.g., Halbesleben et al., 2014).COR theory explicitly focuses on the social context and incorporates a multilevel thinking of resources (Hobfoll, 2011).Thus, gain or loss of collective resources can influence the shared experience of well-being in teams (Hobfoll, 2011;Junker et al., 2021;Maslach et al., 2001;Sonnentag, 2015).Accordingly, we consider mental health as a team-level outcome, as the members of a team should be affected by collective resource gains or losses similarly (Hobfoll, 2011).
The definition and value of a resource depends largely on the context (see e.g., Hobfoll et al., 2018), which is team mental health in this study.To enhance team well-being and to reduce emotional exhaustion, team resources should include the shared experience of control, competence, community, support, and relatedness while acting autonomous, as especially autonomy-related resources can improve wellbeing (Halbesleben et al., 2014;Hobfoll, 2001;Maslach et al., 2001).Based on this premise, shared leadership can be considered as an important team resource (cf.Day et al., 2004).According to the definition of Nicolaides et al. (2014, p. 924), a key aspect of shared leadership is the team's voluntary engagement in leadership functions to reach for team goals.Which leadership functions are actually shared within a team, is explained by the functional leadership approach (Morgeson et al., 2010).In general, the functional leadership approach proposes different leadership functions, which are necessary for team need satisfaction and team effectiveness (Morgeson et al., 2010).Specifically for shared leadership, as an internal and informal source of team leadership, the team collectively engages in the leadership functions of structuring and planning tasks, providing feedback, performing team tasks, solving problems, and supporting the social climate within the team (Morgeson et al., 2010).Accordingly, we refer to shared leadership as shared functional leadership.
While practising shared leadership, teams autonomously work together for the achievement of common goals (e.g., Kukenberger & D'Innocenzo, 2020;Nicolaides et al., 2014;Pearce et al., 2003).Achieving common goals can stimulate collective mastery and growth, for example in the shape of higher collective efficacy beliefs (see Nicolaides et al., 2014).Thus, this shared experience of competence and control while acting autonomous in a strongly related unit is likely to protect a team against exhaustion and can enhance the team's wellbeing (cf.Halbesleben et al., 2014).
Teams utilize shared leadership to create a social context that facilitates cooperation and mutual support (Aubé et al., 2018;Liu et al., 2014;Zhu et al., 2018), reduces conflicts (Sinha et al., 2021), and builds a strong shared sense of purpose (Mathieu et al., 2015), which helps to reduce ambiguities and provides clear goals and expectations for the team.This can be critical to prevent exhaustion and to support well-being (Bakker & Demerouti, 2017).In addition, sharing different leadership functions within the team involves collectively structuring and planning tasks, providing feedback, solving problems, and supporting the social climate within the team (Morgeson et al., 2010).These behaviours cover collective resources which are related to team member well-being (e.g., Bakker et al., 2006;Costa et al., 2015;Torrente et al., 2012), and, thus, render shared leadership as a collective resource for the whole team.
Importantly, our dynamic perspective of shared leadership in teams extends to its team-level outcomes as well.Specifically, increasing shared leadership may relate to an increase in well-being and decrease in exhaustion, as an increase in autonomy-based resources (i.e., shared leadership; cf.Pearce et al., 2003) should have a strong impact on well-being (Halbesleben et al., 2014).In line with COR theory (e.g., Hobfoll et al., 2018), resource gains over time (i.e., gain cycle) may initiate a growth in team well-being, as it reflects acquisition of resources beyond the previous level of available resources.Hence, increasing shared leadership over time may be of importance for team mental health beyond a high initial or general level of shared leader, as temporal changes in resources are more likely to predict changes in mental health than stable job conditions do (cf.Hobfoll et al., 2018;Sonnentag, 2015).In contrast, a continuously decreasing shared leadership over time would indicate a resource loss cycle.This may coincide with less mutual feedback and reduced efforts to collectively structure tasks that may foster negative social interactions (Bergman et al., 2012), inhibit task accomplishment (Carson et al., 2007), and impair cooperation and teamwork (Aubé et al., 2018), with likely decrease in team well-being and increase in emotional exhaustion along with these developments.Summarized, we predict: Hypothesis 1a: Change in shared leadership is positively related to change in team well-being over time.
Hypothesis 1b: Change in shared leadership is negatively related to change in team emotional exhaustion over time.

Team workload as a challenge for team mental health
We defined team workload as the collective experience of too much work to do in relation to the time available in teams.According to COR theory (Hobfoll et al., 2018), team workload can be demanding and resource depleting for the team, which will manifest in the shared experience of impaired well-being and higher exhaustion (cf.Halbesleben et al., 2014;Maslach et al., 2001).When team workload increases, the experience of being overwhelmed by the increasing amount of work accumulates over time among the team members.Dealing with this increasing overload may be associated with psychological costs (Bakker, 2014) and may involve the risk of a resource exhaustion (Hobfoll et al., 2018): Across the week, ad-hoc coordination, short-term changes in the task strategy, or conflict management may become increasingly important over time to deal with the team's increasing workload (cf.Dietz et al., 2017).Additionally, increasing team workload can trigger a resource loss spiral: Resource loss and resulting exhaustion increase the vulnerability to future loss as existing resources may be outstretched (Hobfoll, 2001;Hobfoll et al., 2018).Thus, over time, a downward spiral may develop between increasing team workload and team-level well-being and exhaustion (cf.Hobfoll et al., 2018).Hence, we assume: Hypothesis 2a: Change in team workload is negatively related to change in team well-being over time.

Hypothesis 2b:
Change in team workload is positively related to change in team emotional exhaustion over time.

The protective influence of shared leadership: interactions with team workload
Resource gains become salient in the context of resource losses (Hobfoll, 1989;Hobfoll et al., 2018).High resources generally provide more coping opportunities (e.g., Bakker, 2014).However, a resource loss due to increasing team workload can lead to an exhaustion of existing resources, as the temporal change in team workload may disturb the stable balance between accessibility of team resources (i.e., the general level of resources) and team demands (cf.Hobfoll, 2011).Initial loss of resources can lead to further loss, resulting in a loss cycle (Hobfoll, 2001;Hobfoll et al., 2018).Hence, a resource loss needs to be counterbalanced though resource gains beyond the general resource level.Thus, to overcome negative consequences of resource loss, teams may respond to changes in their situational requirements (specifically, increasing team workload) by increasing their level of shared leadership throughout the week.
As the collective work is more stimulating when leadership is shared (Aubé et al., 2018) and shared leadership can foster the shared experience of mastery and growth, increasing team workload may be perceived as less threatening to resources and therefore may not attenuate team well-being as much when teams show more shared leadership behaviours (see also Hooker & Csikszentmihalyi, 2003).Additionally, increasing engagement in shared leadership across the week can result in a steady growth of positive interactions, social exchange relationships, and social support among team members (Drescher et al., 2014).In line with COR theory, these collective resource gains are important for tackling the challenges associated with an increasing team workload (cf.Hobfoll, 2011).Conversely, when team workload increases but shared leadership stagnates or even decreases, a team may not have sufficient resources to deal with this change in team workload as increasing resources (i.e., resource gains) are most effective in protecting against diminished well-being when demands increase (Hobfoll et al., 2018).More specifically, the lack of collective goal striving, as well as continuously decreasing engagement in problemsolving and support for the social climate may exacerbate the negative consequences of increasing team workload.Hence, a team may feel more and more overwhelmed, resulting in a resource loss cycle which gains momentum over time (e.g., Hobfoll, 2001;Hobfoll et al., 2018).Thus, we predict: Hypothesis 3a: Change in shared leadership interacts with change in team workload, such that the positive relationship between change in shared leadership and change in team wellbeing is stronger when team workload increases over time.

Hypothesis 3b:
Change in shared leadership interacts with change in team workload, such that the negative relationship between change in shared leadership and change in team emotional exhaustion is stronger when team workload increases over time.

Procedure and participants
We applied a daily diary design (Gabriel et al., 2019;Ohly et al., 2010).As a correlational study, it exempts from IRB approval.In line with recent diary studies in the context of formal leadership (e.g., Bormann, 2017), we collected data over five consecutive working days (from Monday to Friday).Data collection started in May 2019 and ended in December 2019.With support from research assistants, we invited teams with a five-day working week from different organizations in Germany.Participating teams were informed about the procedure and purpose of data collection.Teams could receive a feedback report on request as an incentive for their participation.
All members of a team had to participate in the same week.At the beginning of the study, participants completed a general survey assessing their demographics (e.g., age, education), team characteristics (e.g., team tenure, contact time), trait measures of the constructs (i.e., ratings on the average extent of shared leadership, team workload, well-being, and emotional exhaustion), and their end of work times in the next week.Based on their individually reported end of work time for each day, daily surveys were sent by email two hours before the end of work in the subsequent week.If the participants did not fill out the daily survey within 90 minutes, a second email reminded them to complete the questionnaire.Access to the daily survey was only possible on that given day.To guarantee the anonymity of data, the survey tool generated an internal code for all participants that was used to match the daily surveys and the general survey.Team affiliation was coded similarly.
We invited 295 team members from 79 teams to the general survey.From those, 274 team members (response rate = 93%) from 79 teams completed the general survey.As a research assistant informed the first author that two teams did not provide valid data (i.e., some team members did not consider all of their team colleagues in their ratings), these two teams were excluded from the survey.The remaining 265 team members were then invited to the daily survey.On average, participants filled out 4.42 daily questionnaires (66% filled out all daily surveys) resulting in a final sample of 1184 person-days and 377 team-days.
About half of the team members were female (49%) and on average 34 years old (SD = 11.21years).Most of them held a university degree (57%) and worked about for 3.60 years in their team (SD = 4.39 years).The average team size was 3.49 (SD = 1.50).Team members were in direct contact with their team colleagues for approximately 13 hours per week (SD = 11.90 hours).The sample consisted of teams from the IT sector (20%), chemistry and science sector (11%), trading sector (10%), and public administration sector (9%).

Measures
Before describing our measures, we first explain the multilevel nested structure of the data.Our data had a three-level structure (see Figure S1 in the Online Supplement).The lowest level represented daily measures of the individual team members (Level 1 -daily individual level).These were nested in daily measures at the team level (Level 2 -daily team level).Hence, we aggregated the daily observations of the members of a team for every day (e.g., all team members' reports on Monday were aggregated to a team-level value for that day).These team-day data were nested in teams (Level 3 -team level).

Shared leadership
We used a shortened version of the team leadership questionnaire (Morgeson et al., 2010) that was adapted for daily measurement.The voluntary engagement in leadership functions of the team members is a key aspect of our underlying construct definition of shared leadership (Nicolaides et al., 2014, p. 924).Hence, we rely on the functional leadership approach and assessed those leadership functions that are best suited to be fulfilled by teams themselves (see Morgeson et al., 2010): Structure and plan (e.g., "Today at work, we defined and structured our own work and the work of the team"), provide feedback, perform team task, solve problems, and support social climate.Each of these five leadership functions was measured with two items on a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree).Cronbach's alpha ranged from .87 to .91 over the five days.We calculated intraclass correlation (ICC1) for the daily team level and the team level.The ICCs for the daily team level refer to the proportion of dayspecific variance in the team member ratings that can be due to their team membership (i.e., shared variance between the members of a team for each day).Furthermore, the ICC1 for the team level refers to the proportion of variance that can be due to their team membership in general (i.e., shared variance between the members of a team across all days).The ICC1 was .16 for the daily team level and .32 for the team level.The average degree of interrater agreement (r wg(j) (James et al., 1993), was between .77and .82for the daily measures (SDs r wg(j) = .29-.31).

Team workload
Team workload was measured using a four-item version from the Copenhagen Psychosocial Questionnaire (COPSOQ, Kristensen et al., 2005;German version by;Nübling et al., 2006).Items were adapted to capture day-specific workload on the team level.Additionally, the instructions for this scale asked team members to rate these items with respect to the day-specific experience of the entire team.Team members rated the items (e.g., "To what extent did your team have to work very quickly today?") on a fivepoint Likert scale ranging from 1 (never) to 5 (always).Cronbach's alpha ranged from .82 to .86 over the five days.ICC1 was .25 for the daily team level and .32 for the team level.Mean r wg(j) was between .74 and .79 for the daily measures (SDs r wg(j) = .22-.31).

Well-being
We used an adapted five-item version of the WHO Well-Being Index (Topp et al., 2015) to measure subjective well-being.In line with previous research, we did not shift the referent to the team but rather asked all team members about their individual well-being (e.g., Hudson & Shen, 2018;Junker et al., 2021).This decision builds on the rationale that the experience of states like well-being (and emotional exhaustion) may be most accurately measured by self-ratings of the focal individuals (Gabriel et al., 2019).Team members rated the items (e.g., "Today I have felt active and vigorous") on a five-point Likert scale ranging from 1 (never) to 5 (always).Cronbach's alpha ranged from .80 to .87 over the five days.ICC1 was .12 for daily team-level and .19 for the team level.Mean r wg(j) was between .74 and .85 for the daily measures (SDs r wg(j) = .19-.30).

Emotional exhaustion
We measured emotional exhaustion with four items from the Copenhagen Burnout Inventory (CBI, Kristensen et al., 2005;German version by;Nübling et al., 2006) that have been adapted for daily measurement.An illustrative example was: "How often have you been emotionally exhausted at work today".The items were rated on a five-point Likert scale ranging from 1 (never) to 5 (always).Cronbach's alpha ranged from .85 to .89 over the five days.ICC1 was .13 for the daily team-level and .22 for the team level.Mean r wg(j) was between .74 and .83for the daily measures (SDs r wg(j) = .22-.32).

Factorial validity
We used confirmatory factor analyses (CFA) to test the factorial validity of our daily measures.Given the fact that the number of estimated parameters did exceed our sample size on the highest level, estimating a multilevel CFA resulted in serious convergence problems and was therefore not feasible.As the measurement of shared leadership covered five different leadership functions, we modelled shared leadership as a second-order construct.Accordingly, we tested a CFA with eight first-order factors and one second-order factor.The model yielded an acceptable fit to the data (χ 2 = 1264.64,df = 219, p < .001,CFI = .93,RMSEA = .06,SRMR = .06)that was better compared to an alternative model with seven first-order factors (items of daily emotional exhaustion and daily team workload loaded together on one factor; ∆χ 2 = 1538.15,∆df = 3, p < .001)and an alternative model, in which the items of daily emotional exhaustion loaded together with the items of daily well-being (∆χ 2 = 1639.80,∆df = 3, p < .001).

Analytical strategy
We used an analytical approach developed by Chen et al. (2011) which aims to examine the relationships between team-specific change trajectories in the focal constructs over time.Accordingly, we conducted the following steps to prepare our data: First, we aggregated the daily measures to the team level (i.e., aggregated all measures of the members of one team for each day).Aggregation can be justified as all variables had a notable amount of shared variance on the daily team level (Level 2) and sufficient interrater agreement (Bliese, 2000).Second, we specified growth curve models to capture change trajectories in our focal constructs using random coefficient modelling.In particular, we estimated within-team slopes of time for all constructs by modelling a random slope of time as linear change over the five working days using the lme4 package in R (Version 4.0.0;R Core Team, 2020).Formalized, we estimated the following random-slope model of time for each construct: Daily team level: In this equation, � x :jk is the observed and aggregated score of X of team k at time j, β 1k is the linear change of time which can vary between teams (specified by u 1k ).Additionally, likelihoodratio tests indicated that the teams in all constructs differed significantly in their change patterns over time.
Third, we extracted the team-specific random slopes of time (i.e., the change trajectories) for all constructs.Using SEM with the lavaan package, we modelled our proposed relationships between change in the independent variables and change in the dependent variables as well as the interaction terms, while controlling for initial values of the dependent variables (i.e., values on Monday).To test our hypotheses, we included all proposed relationships in a single statistical model.Prior to this final analytical step, we standardized the change scores and initial values.

Results
Table 1 presents the means, standard deviations, and intercorrelations of all constructs.
The results of our hypotheses testing are presented in Table 2 and additionally depicted in Figure 1, in order to provide an overview of our model at a glance.Our statistical model had an excellent fit to the data (χ 2 = 0.11, df = 2, p = .95,CFI = 1.00,RMSEA = .00,SRMR = .01).
Supporting Hypotheses 1a and 1b, the results indicated that an increase in shared leadership was significantly associated with an increase in team well-being (β = .52,p < .001)and decreasing emotional exhaustion over time (β = −.37,p < .001).Moreover, team workload dynamics were related to a decreased well-being over time (β = −.21,p = .039)and an increase in emotional exhaustion (β = .18,p = .047),which supports Hypotheses 2a and 2b.
The interaction between dynamics in team workload and shared leadership on well-being was not significant (β = −.08,p = .447).Therefore, Hypothesis 3a received no support.Regarding Hypothesis 3b, results showed a significant interaction between dynamics in team workload and shared leadership on emotional exhaustion change (β = −.19,p = .042).Thus, shared leadership dynamics over time were more strongly related with a decrease in collective emotional exhaustion, when team workload increased (β conditional = −.55,p < .001) vs. decreased over time (β conditional = −.18,p = .141;see Figure 2).For a more detailed analysis, we calculated the marginal effects using Johnson-Neyman-Technique (see Figure 3).This analysis revealed that the negative relation between change in shared leadership and change in emotional exhaustion did only become non-significant, when the standardized change score of team workload was below −.84, which would indicate a strong decrease.Thus, Hypothesis 3b was supported.
To illustrate temporal change in shared leadership and to examine the robustness of our findings, we ran several supplementary analyses, which are included in the online supplement.Using these analyses, we could replicate our pattern of results in predicting end of week levels of team well-being and emotional exhaustion while controlling for initial values of these constructs.These analyses provide additional support for the proposed temporal direction of our findings.

Discussion
In this study, we linked shared leadership dynamics to temporal changes in team well-being and emotional exhaustion, based on COR theory (Hobfoll et al., 2018).We could show that increasing shared leadership was related to increased team member well-being as well as decreased team-level emotional exhaustion over time.Additionally, we assumed that shared leadership may interact with team workload, as resource gains should be most important in the context of resource loss.Results supported this assumption only partially, as an increase in shared leadership was more strongly associated with dynamics in team-level emotional exhaustion when team workload increased.However, this interaction could not be found with regard to team-level well-being dynamics.

Theoretical implications
Given the correlational nature of our study design, our findings and related implications have to be interpreted with caution (i.e., regarding causality; see Antonakis et al., 2010).First, this study provides a novel perspective on shared leadership by showcasing the temporal dynamics of this team phenomenon.Considering dynamics in shared leadership is necessary to understand how shared leadership evolves over time (cf.Kozlowski et al., 2016).Indeed, dynamics are inherent in the definition and conceptualization of shared leadership (see D 'Innocenzo et al., 2016;Pearce et al., 2003).Specifically, our results show that changes in shared leadership occur not only as teams mature or develop over longer time spans (e.g., lifecycle models of team development; Drescher et al., 2014), or due to different stages of project work (Lorinkova & Bartol, 2021), but rather more quickly and dynamically.Even over short periods of time, teams may show significant differences in their temporal development of shared leadership behaviour.
As such, our results also add to the discussion about dynamics in teams at large and underscore the notion that teams can be considered as volatile and agile collectives rather than stable (cf.Arrow et al., 2004;Collins et al., 2016;Klonek et al., 2019).
Second, our research can underline the value of shared leadership for teams by theoretically and empirically linking it to team-level well-being and emotional exhaustion.As these are two previously unstudied outcomes of shared leadership (Zhu et al., 2018), we may also be able to expand the nomological network of shared leadership towards mental health outcomes.This expansion seems particularly relevant given the long existing theorizing about the social nature of mental health (e.g., Hobfoll, 2001;Maslach et al., 2001), the growing prevalence of teamwork (Mathieu et al., 2017), and the importance of such outcomes for organizational functioning at large (e.g., Christian et al., 2011;Cole et al., 2011).Our empirical insights lend support to our argument that a continuous engagement in sharing different leadership functions among the team (e.g., collectively structuring and planning tasks, or performing team tasks) can create a social context and shared experience that may be beneficial for the development of shared mental health among the team members.
Our findings can also contribute to the literature on workrelated well-being and may help to bridge the gap between this stream of literature and the literature on teams.Previous research on work-related well-being rarely took a collective or team-level perspective into account (Sonnentag, 2015).As teams are widely used as organizing principle, this view can generate meaningful insights and adds to the scarce but growing research on team-level antecedents of shared mental health (Junker et al., 2021).Moreover, our results offer opportunities for theory-building and future studies to include teamlevel constructs to predict mental health from a multilevel perspective.In this regard, COR theory offers a suitable background to understand higher-level resource gains and loss ( Hobfoll, 2011;Hobfoll et al., 2018).However, although the ICC values for well-being and emotional exhaustion indicate a medium to strong effect of team membership (i.e., a shared experience of mental health; LeBreton & Senter, 2008), most of the variance in these constructs is still related to the individual level.As a consequence, the (daily) team-level relationships between these constructs can possibly be underestimated (cf.Bliese et al., 2019;Lüdtke et al., 2008).Hence, our findings regarding team mental health have to be interpreted with caution and need to be replicated in future studies (Köhler & Cortina, 2021).
Besides this, our study can contribute to the understanding of well-being as a dynamic state (e.g., Sonnentag, 2015).Previous diary studies have mostly focused on day-specific fluctuations of individual-level well-being (Bakker, 2014;Sonnentag, 2015), but seldom examined change trajectories over time (for a notable exception, see Hülsheger et al., 2014).This perspective implies that shared well-being and emotional exhaustion do not only alter from day-to-day, but also exhibit systematic (i.e., linear) change trajectories over a continuous period of time.These findings can add to the assumption of resource gain and loss cycles of COR theory (e.g., Hobfoll et al., 2018).
Third, the interaction between shared leadership and team workload regarding team emotional exhaustion may emphasize the conceptualization of shared leadership as a collective resource for teams.A growing engagement in shared leadership can help to mitigate negative outcomes for teamwork and task work due to increasing team workload (Funke et al., 2012).Additionally, this implication contradicts the speculation of Zhu et al. (2018) that shared leadership may not be appropriate in times of high pressure, as our results indicate that shared leadership can unfold its potential especially under conditions of increasing team workload.This meets the expectation of COR theory about the importance of resource gains in the context of resource loss (Halbesleben et al., 2014).Therefore, shared leadership can be seen as a dynamic means for teams to adapt their behaviour and to counterbalance resource loss.
Finally, our finding that increasing shared leadership may buffer against increasing team workload in the case of teamlevel exhaustion, but does not interact with team workload dynamics when predicting changes in team well-being could be due to the complexity of the well-being construct.Wellbeing is defined as optimal psychological functioning and experience, and not just the absence of mental illness (e.g., depletion, exhaustion, or strain; Ryan & Deci, 2001).Hence, a reduction of the challenges associated with team workload may not suffice for increasing team-level well-being.

Practical implications
Organizations need to recognize and promote shared leadership parallel to formal or hierarchical leadership.The increase of agile teams points to the value of shared leadership (e.g., Annosi et al., 2017).Such developments towards more shared leadership in teams can be actively supported by formal leaders in an organization.Especially expressions of leader humility, empowering leadership, and transformational leadership have been identified as antecedents of shared leadership (e.g., Klasmeier & Rowold, 2020;Zhu et al., 2018): For example, formal leaders can foster collective goals and a shared purpose in order to encourage shared leadership in teams.Additionally, organizations can actively communicate that shared leadership is recognized as a desirable behaviour and encourage team members to actively engage in informal leadership processes.

Limitations and future directions
This study has several limitations that point to opportunities for future research.First, the reliance on self-report measures to capture our focal study variables may raise concerns about common method bias.However, there are some reasons that justify the use of self-reported data, and we took steps to reduce potential bias.We considered temporal change patterns at the team level over time for hypothesis testing.This means that the variables at this level reflect the change in the day-specific shared consensus of all team members across days, and thus may be less biased than individual-level ratings.Moreover, especially for affective data, self-reports are more valid than other-ratings (Gabriel et al., 2019).Nevertheless, we encourage future research efforts to consider a wider range of possible well-being indicators.
A second limitation is related to our measure of shared leadership and team workload.In the case of team workload, Funke et al. (2012) discussed several other methodological strategies to measure team workload (e.g., physiological and behavioural data or strategy shift measures).Likewise, a common approach to measure shared leadership is the use of social network analysis.Using social network analysis, however, seldom captures the content of shared leadership (Carson et al., 2007;Zhu et al., 2018).Additionally, using a social network approach to measure different team leadership functions would have been difficult to implement in a daily diary study design (Gabriel et al., 2019) due to the increased length of the questionnaire.Thus, more research is needed to compare different measures for shared leadership and team workload to provide evidence-based guidelines for the operationalization of these team constructs.
Third, we controlled for initial values of team well-being and team emotional exhaustion, for example, to account for possible ceiling effects.Following the suggestion of an anonymous reviewer, we also controlled for initial values of shared leadership and team workload in a separate model.The pattern of results remained largely unchanged and still supported Hypotheses 1a through 2b.The only difference between the two models concerned the prediction of change in emotional exhaustion by the interaction of change in shared leadership and change in team workload, which slightly failed to reach the conventional significance level (β = .18,SE = .11,p = .09).However, the marginal slope plot using Johnson-Neyman Technique indicated only a small difference in the region of significance for the marginal relationship between shared leadership change and change in emotional exhaustion starting from values of team workload change of greater than −0.78 vs. −0.84 in the model when controlling for initial values of well-being and emotional exhaustion.Nevertheless, this finding calls for a cautious interpretation of the interaction, given its possible limited robustness.
Fourth, the magnitude of the ICC1 values has to be discussed.Although, the ICC1 values for the daily team level and team level are above recommended cut-offs for aggregation (e.g., Bliese, 2000), a notable share of variance of our focal constructs can be attributed to the daily individual level.While the ICC1 values are comparable to previous research on shared leadership (e.g., He et al., 2020), and team mental health (e.g., Junker et al., 2021), they suggest that the team-level constructs represent only a limited amount of sharedness between the team members.
Finally, to the best of our knowledge, this study is among the first empirical investigations of the dynamics of shared leadership as an antecedent of team members' well-being and exhaustion using a daily diary design.This requires cautious interpretation of our results, and future studies should aim to replicate these findings to demonstrate their robustness.

Conclusion
The present study developed a dynamic temporal model to examine the relationship between changes in shared leadership and dynamics in well-being and exhaustion in teams.We drew on COR theory (Hobfoll et al., 2018) to argue how the sharing of leadership functions can serve as a team resource.We hypothesized and found that dynamic changes in shared leadership predict changes in team-level well-being and emotional exhaustion.These findings may emphasize the importance of shared leadership for team functioning, establish shared leadership as a team resource, and advance our understanding of the temporal dynamics of team leadership and its benefits for shared mental health.

Figure 2 .
Figure2.Simple slope plot.Change variables measured as team-specific standardized change trajectories.Simple slopes plotted for ±1 standard deviation around the change trajectory of the moderator.Hence, "Team Workload Increase" refers to an above average (i.e., positive) change trajectory of team workload, whereas "Team Workload Decrease" refers to a below average (i.e., negative) change trajectory of team workload.

Figure 3 .
Figure 3. Marginal slope plot using Johnson-Neyman-Technique.Change variables measured as team-specific standardized change trajectories.The plot depicts the marginal relationship of shared leadership change with emotional exhaustion change conditioned by the change in team workload.Plot created using an App by Finsaas and Goldstein (2021).

Table 2 .
SEM of dynamic changes in shared leadership, team workload, and collective member exhaustion and well-being in the team.
N = 77 teams.Results are standardized coefficients.