Team affective tone and Psychological safety: Relating
shared leadership to decision making in new venture teams.
Master Thesis
Maarten van den Berg
Amsterdam, June 24, 2016
Student number: 10264426
MSc. in Business Administration – Strategy Track University of Amsterdam
Supervisor: Alex Alexiev Semester 2, block 3
Statement of originality
This document is written by Student Maarten van den Berg who declares to
take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original
and that no sources other than those mentioned in the text and its
references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the
supervision of completion of the work, not for the contents
Abstract
Research on new ventures has often focused on the role of a single entrepreneur and its effect on performance. However, most new ventures are founded and led by a team and therefore several entrepreneurship scholars shifted their focus more towards a shared leadership style. Unfortunately, its effect on emergent states and especially affective constructs have received little attention in the context of new venture teams (NVTs). Based upon the data collected from a sample of 61 Dutch new ventures, this study attempted to reveal how shared leadership and the emergent states team affective tone and psychological safety affect the decision making process in new ventures. We find that shared leadership in NVTs not only directly enhances decision making comprehensiveness, but also indirectly leads to more
comprehensive decision making through the mediator psychological safety. Furthermore, positive team affective tone leads to a higher pace of decision making.
Table of Content
1. Introduction ………....1
2. Literature review……….4
2.1 New Venture teams………..4
2.2 Decision making comprehensiveness and ………...5
2.2.1 Decision making comprehensiveness……….6
2.2.2 Decision making speed………...8
2.3 Shared leadership in new ventures………...9
2.4 Shared leadership, decision making comprehensiveness and decision making speed………...11
2.5 Shared leadership, Team affective tone and decision making………12
2.6 Shared leadership, team psychological safety and decision making ………...14
3. Method………18
3.1 Research method ……….18
3.2 Data collection ………...19
3.3 Sample ………... ...20
3.4 Variables and measurements………....21
3.5 Statistical procedure and qualitative analyses method ….………...25
3.6 Reliability and common bias analyses ….………26
3.7 Factor analyses ……….27
4. Results………...31
4.1 Correlations………..………...31
4.2 Hypothesis testing ………35
4.2.1 Direct effects of shared leadership on decision making comprehensiveness and speed (H1,H2)……….……..…………...35
4.2.2 Hypothesis testing of shared leadership on psychological safety and team affective tone (H3,H6)………...36
4.2.3 Hypothesis testing of psychological safety and team affective tone on decision making comprehensiveness and decision making speed (H4,H5,H7,H8)……..36
4.3 Taken together ……….38
4.3.1 Decision making comprehensiviness………..38
4.4 Results qualitative research - interviews ………...40
5.4.1 Shared leadership………40
5.4.2 Psychological safety………41
5.4.3 Environment uncertainty and decision making ………..42
5. Discussion………...45
5.1 Findings and theoretical contributions..………..45
5.2 Managerial implications ………48
5.3 Limitations and suggestions for future research ………49
References ……….53
Appendix A: Tests of Normality ……….67
Appendix B: Statistical model process macro Hayes………67
Appendix C: Measurement scales ………..67
Appendix D: Exploratory Factor Analyses………70
Appendix E: Definitions of labels………73
Appendix F: Schematic overview complete qualitative research...77
1
1. Introduction
New ventures play an important role in today’s economy. Their impact on economic growth is substantial in many countries (Sternberg and Wennekers, 2005). Although there relative growth is higher than that of more established ventures, startups also have a high failure rate (Dosi and Lovallo, 1997; Quimet and Zarutskie, 2014). Therefore, entrepreneurship scholars are interested in why just a small amount of new ventures succeed and what factors are from importance to be successful as a startup (e.g. Cooper et al., 1994; Chrisman et al., 1998; Li and Zhang, 2007).
At first, the focus of entrepreneurship scholars has mainly been on the role of the leading founder because entrepreneurs play a central role in the creation of a new venture. Therefore much attention is given to the behavior and cognitive aspects of the entrepreneur (Baron, 2007). However, most new ventures are founded and led by a team (Lechler, 2001; West, 2007). As a consequence, the attention of the field of entrepreneurship shifted also towards new venture teams (NVTs) (e.g. Amason et al., 2006; Ensley et al., 2006). The study of Carland & Carland (2012) even found that sole entrepreneurs are typically less often involved in high growth new ventures. This also made scholars look further on how entrepreneurs should lead their ventures because there is not necessarily a single leader in NVTs. Consequently, a single traditional leadership style might not be most suitable. It is acknowledged that leadership goes beyond a single person and that leadership behavior can also be found in teams (Avolio & Bass, 1995; Sivasubramaniam et al.,2002). Therefore, shared leadership might be more useful for ventures that are led by a team (Pearce, 2004). Leadership has proven to be relevant to strategic decision making because leaders must built trust within their team to sustain a stable climate. This should lead to confidence within the venture which allows leaders to make timely decisions without the need for consensus (Bourgeois and Eisenhardt, 1988; Korsgaardet et al.,1995). Additionally, leaders
2
are responsible for developing the right structures in a team and for collecting resources, which are needed for effective strategic decision making (Pearce, 2004). Although leadership in new ventures is proven to be associated with new venture performance (Ensley et al., 2006), this direct relationship might not be an ideal predictor. The focus of studies on such direct relationships stem from the normative rational model, which suggests that
entrepreneurs and high quality decisions enhance venture performance (Korsgaardet et al., 1995). However, contingent relationships might provide a better explanation of venture’s performance compared to direct ones (Lyon et al., 2000). Simply put, more knowledge of the underlying process is preferable to really understand how leadership influences decision making and eventually performance. Therefore this study will not try to link shared leadership directly to new venture performance but to the decision making process in NVTs.
The focus will mainly be on two specific elements of decision making. Pearce and Zahra (1991) argues that decision making comprehensiveness and speed are important elements to assess the quality of decision making. Decision making has to be complete and exhaustive to be of high quality. Speed of decision making is also critical because only timely made decision can result in a competitive advantage, especially when operating in dynamic environments (Talaulicar et al., 2005).
Additionally, integration of theories from team performance and organizational behavior into the entrepreneurship field has given interesting and important new insights into how certain aspects of NVTs can lead to improved venture performance (e.g. Sy et al., 2005; Cole et al., 2008). For example, the team processes have received attention by
entrepreneurship scholars (e.g. Ucbasaran, 2003; Boeker and Wiltbank, 2005). However, emergent states and especially affective constructs have received little attention in the context of NVTs (Klotz et al.,2014). This might be particularly interesting because affective
3
(Korsgaardet et al., 1995). Taken together, a more elaborate view of the causal chain of leadership to venture performance will help managers to get a better understanding on how their leadership style and behavior influence strategic decision making and ultimately venture performance. Depending on the strategic goals set by the venture, for example either growth or profit, it will help managers to develop the right decision making process to achieve these set goals. This knowledge for managers is currently lacking in the literature with respect to NVTs.
This study will use the inputs-mediators-outcomes framework to give more insight into how shared leadership influences decision making in new ventures and what role
affective states have in this relationship. The research question central in this paper is: what is
the effect of shared leadership in NVTs on new venture decision making and how do the affective states of team member’s psychological safety and team affective tone influence this?
This study uses mainly a quantitative approach combined with a qualitative part to go more into depth. It will expand the literature by further integrating the entrepreneurship field with team performance and organizational behavior research. Moreover, it will expand research on emergent states in the context of NVTs.
The rest of this paper will be structured in the following way. The second section contains the literature review and the hypothesizes. In the third section the methodology will be discussed which will be followed up by the study results in section four. In the last section, a general discussion of the study will be given together with the suggestions for future
4
2. Literature review
For over 2 decades, scholars have increasingly paid attention to the performance of new ventures. The aim of most scholars is to explain why certain new ventures perform better than others. Several variables are tested over time and knowledge in the field about the
performance of new ventures is increasing. Especially, attention to new venture team (NVT) performance has risen. However much remains understudied about new ventures and NVTs (Klotz et al.,2014) . In this literature review will be discussed what research has been done so far, followed up by what is still missing.
2.1 New venture teams
New ventures play an important role in today’s economy. Their impact on economic growth is substantial in many countries (Sternberg and Wennekers, 2005). The aim of most scholars is to explain why some new ventures perform better than others. The definition of a new venture in general is unclear. Whereas it is pretty clear what a new venture is, the definition about when a venture is still being regarded to as a startup is not that straightforward. There is already disagreement about the definition for several years (Reynolds & Miller,1992; Vesper, 2000; Amason, Shrader, & Tompson, 2006). Several authors tried to define new startups on basis of their age or size (Zahra, Ireland, & Hitt, 2000; Amason, Shrader, & Tompson, 2006). However, this can better be avoided because several boundary conditions are context-specific, such as technological intensity. A broader definition is more appropriate and therefore the definition given by Klotz et al. (2014) is used: “a new venture is a firm that is in its early
stages of development and growth” (p.227).
The focus of research has been on the leading founder/CEO for quite some time and in practice much credit is given to leading founders when a new venture is successful. According to Baron (2007), entrepreneurs play a central role in the creation of new venture and therefore much attention is given to the behavior and cognitive aspects of the entrepreneur. Whereas the
5
role and influence of the sole entrepreneur can be relevant, most new ventures are founded by and consequently also led by a team (Lechler, 2001;West, 2007). For instance in the research sample of Beckman (2006), which consists out of 160 new high-technology ventures in Silicon Valley, 90% of the ventures were founded and led by a team. Additionally, Carland & Carland (2012) argue that sole entrepreneurs are less involved in high growth startups
compared to NVTs. So looking at the TMT in new ventures is likely to be more relevant than looking at just one individual. This is in line with the UE perspective, which has been the basis of lots of research. Developed as a model by Hambrick and Mason (1984), UE states that venture performance is partially predicted by the characteristics of its top management team (TMT). The UE perspective proposes that psychological characteristics such as cognitive based values and the more observable ones such as age, education and previous experience can all influence venture’s strategic choices and consequently performance (Hambrick and Mason, 1984).
Therefore, the unit of analysis in this research is the NVT which can be defined as ‘the
group of individuals that is chiefly responsible for the strategic decision making and ongoing operations of a new venture” (Klotz et al., 2014, p. 227). Due to the fact that NVTs are
responsible for the strategic decision making, this unit of analysis is highly relevant when trying to extent our knowledge on the strategic decision making process of new ventures. 2.2 Decision making comprehensiveness and speed
Both established and new ventures are often confronted with major changes. Therefore, the decision-making process has received much attention for a long period. Especially several elements of decision-making quality have been studied over the years. Pearce and Zahra (1991) argue that decision making comprehensiveness and speed are important elements to assess the quality of decision making. Decision making has to be complete and exhaustive to be of high quality. When this not the case it will have negative consequences for the
6
performance of the venture (Hambrick and Mason, 1984). Speed of decision making is also critical because only timely made decision can result in a competitive advantage, especially when operating in dynamic environments (Talaulicar et al., 2005). Therefore, these elements are especially relevant for new ventures
2.2.1 Decision making comprehensiveness
Decision making comprehensiveness can be defined as: “the extent to which an organization
attempts to be exhaustive or inclusive in making and integrating strategic decisions”
(Fredrickson and Mitchell, 1984, p. 402). Generally, it has been proposed that decision making comprehensiveness and speed of decision making should be seen as a tradeoff. Making a thorough decision is time costly because of the extensive analysis and debate that are part of this (Mintzberg, 1973; Janis,1982). Together with the fact that managers are bounded rational (March and Simon, 1958), it is nearly impossible to coordinate all the internal made decisions, obtain all the needed information and put it all together on time for the right moment (Quinn, 1980).
Building upon the assumption that comprehensiveness and speed should be seen as a tradeoff, several propositions has been made. The role of environment dynamism has been proven to be from importance (Fredrickson, 1984). For example, it has been proposed that in an unstable environment, decision making comprehensiveness is negatively related to venture performance. After all, in an unstable environment decisions have to be made more frequent and faster. So a tradeoff in the favor of comprehensiveness at the cost of speed can have negative consequences for the venture’s performance in an unstable environment. The opposite will happen for decision making in a stable environment, so that decision making comprehensiveness will be positively related to venture’s performance (Fredrickson, 1984). These propositions would later be tested in a study of Eisenhardt (1989). The results showed that in the high-velocity microcomputer industry, both decision making speed and decision
7
making comprehensiveness are positively related to performance. This showed that the propositions made in earlier works are not fully correct and that decision making
comprehensiveness and speed are not necessarily mutually exclusive (Talaulicar et al., 2005). The finding that the concepts are not mutually exclusive led to extensive research on what other antecedents could help to explain these results. According to Simons et al. (1999), group diversity with respect to demographics and cognitive aspects has shown to be important antecedents. Hambrick and Mason (1984) described that the drive for consensus is negatively related to the amount of alternatives produced by a team which can have negative
implications for a venture . A lower amount of alternatives can imply that not all possible solutions to a problem are considered which prevent teams from being able to pick the best solution. Furthermore, Simon et al. (1999) showed that a decision making process involving an open debate and cognitive conflict is closely associated with decision comprehensiveness. With respect to new ventures, constructs such as affective and cognitive conflict and cohesion in TMTs were all found to be important and complex with respect to the decision making process (Amason and Sapienza, 1997). Cohesion was found to be particularly interesting for new ventures because these TMTs operate in a complex and ambiguous environment (Ensley et al., 2002). Team cohesiveness positively influences teams because as described by Smith et al. (1994): ‘‘top management teams that work well together react faster, are more flexible, use superior problem solving techniques, and are more productive and efficient than less integrative teams’’ (p. 432). Furthermore, cognitive and affective conflict has shown to be closely related to TMT cohesion. Cognitive conflict can be seen as judgment differences about how best to achieve common objectives (Amason, 1996). These judgment differences are likely to increase the amount of alternative solutions to problems, leading to more comprehensive decision making. This confirms the findings of Simon et al. (1999) that open debate and cognitive conflict is closely associated with decision comprehensiveness and
8
that its hold in the context of new ventures as well. Taken together, cohesion is positively related to cognitive conflict, which in turn positively influences venture’s performance (Ensley et al., 2002; Higashide and Birley,2002). It should be noted that although cognitive conflict can be beneficial for decision making comprehensiveness (Amason, 1996), without trust it can also lead to affective conflict (Amason 1996, Ensley et al., 2002). Affective conflict can lead to lower decision quality and also lowers the satisfaction level and team affect (Ensley et al., 2002; Ensley and Pearce, 2001). Additionally, team affective conflict leads to members leaving the team (Vanaels et al., 2006).
2.2.2 Decision making speed
Whereas decision making comprehensiveness has received more attention than decision making speed, it is acknowledged to be closely associated with it. As mentioned earlier, comprehensiveness is generally assumed to reduce decision making speed because it leads to an increase in alternatives which need more time to be analyzed elaborately (Talaulicar et al., 2005). In some cases, a high level of analysis might event lead to decision making paralysis (Langley, 1995). To prevent this from happening, a more autocratic leadership style by a CEO can help increase speed. However, even autocratic CEOs need to gather advice from its TMT to make complex decisions in which trust is an important aspect (Eisenhardt, 1989). Also team composition has been found to influence decision making speed. New ventures TMTs with high diversity with respect to tenure, functional background and education have a greater propensity for action. Additionally, homogeneous teams react faster to competitors moves (Hambrick et al., 1996). Lastly, past performance of a venture is also likely to influence the speed of decision making. Ventures that performed well over the years have built more organizational slack. The consequence of this is that manager tend to be less risk taking because the level of organization slack will become higher than the aspiration level (Singh, 1986).
9
2.3 Shared Leadership in new ventures
For over several decades, leadership has been from great importance in management and organizational behavior studies. Especially leading CEOs have received much attention (Finkelstein and D’aveni, 1994; Waldman et al., 2001; Carmeli et al., 2011). These studies looked at formally appointed leaders and their effect on teams, so a type of vertical leadership. The entrepreneurship field also integrated this stream of research. After all, entrepreneurs play a central role in the creation of new ventures (Baron, 2007).
However, it is acknowledged that leadership goes beyond a single person and that leadership behavior can also be found in teams (Avolio & Bass, 1995). This is especially interesting for entrepreneurship scholars because it is most likely that new ventures are founded by a team instead of a single individual. As argued by Sivasubramaniam et al. (2002): “We believe the team can influence each member just as the individual leader can influence his or her followers” (p.67). Researchers have also characterized leadership as representing a collective influence process (Astin & Astin, 1996; House & Aditya, 1997; Sivasubramaniam et al.,2002). According to Pearce (2004) shared leadership can be defined as: “a simultaneous, ongoing, mutual influence process within a team that is characterized by “serial emergence” of official as well as unofficial leaders” (p. 48). In practice, this often implies that the authority will mainly shift to the person that has the most skills and knowledge about the particular issue that is faced by the team (Pearce, 2004). As a
consequence team members experience similar emotions, feel highly involved in the team and feel that organizational outcomes are due to their collaborative efforts (Gronn, 2002). Shared and vertical leadership differ in several ways. Shared leadership can be seen as a process in which leadership is executed by a whole team. On the contrary, vertical leadership can be seen as an influence on this process by a single designated person (Ensley et al., 2006). Additionally, a venture with team based leadership can built upon their collective knowledge
10
instead of relying mainly on just that of a single person. Lastly, vertical leadership is a top-down influence process and shared leadership should be seen as a collaborative process. (Ensley et al., 2006).
To get more insights into how these types of leadership influence performance, several studies have been conducted. Pearce and Sims Jr (2002) looked at how both vertical and shared leadership influence team effectiveness. They found that shared leadership is a better predictor of team effectiveness than vertical leadership. Ventures could therefore decide to put in a place a what they call ‘conscious strategy’ of distributing leadership to team members. Additionally, Ensley et al. (2003) proposed that in the context of new venture teams, shared leadership and team effectiveness is partially mediated by cohesion level and shared vision. This all does not imply that vertical leadership has no value in a team at all. For example, the level of vertical transformational leadership was also positively related to team effectiveness. (Pearce and Sims Jr, 2002; Pearce et al., 2004). Whereas the research of Pearce and Sims Jr (2002) was done inside an established venture, Ensley et al. (2006) extended their research to the context of new ventures. Both shared and vertical leadership were good predictors for new venture’s performance. However, shared leadership showed to be a better predictor. The concepts of vertical and shared leadership do not necessarily have to be mutually exclusive. A vertical leader can be crucial for shared leadership in two ways (Pearce, 2004). First of all, a team leader is mainly responsible for designing a team. The decisions made in this process can set expectations for interaction among team members and contribute to the level of shared leadership. Secondly, the team leader is also often responsible for securing resources and maintaining positive constituents (Pearce, 2004). This sets the boundaries for the team as a whole. Besides, well developed shared leadership can work as a buffer for stress that could impact a team leader (Lovelace, Manz and Alvezprovide, 2007).
11
2.4 Shared leadership, decision making comprehensiveness and decision making speed As mentioned before, shared leadership can be seen within a team when it is characterized by serial emergence of official as well as unofficial leaders (Pearce, 2004) and this implies that the authority will mainly shift to the person that has the most skills and knowledge about the particular issue that is faced by the team (Pearce, 2004). Additionally, this type of leadership in a venture results in a process in which leadership is executed by a whole team and its collective knowledge. So this should be seen as a collaborative process (Ensley et al.,2006). By relying upon the collective knowledge of a team, a higher amount of alternatives will be produced (Hambrick and Mason, 1984). Furthermore, a groups’ information processing capacity is positively influenced by shared leadership (Pearce and Sims, 2002). This will result in a more elaborate process of considering all the possible alternatives and more debate on which one should be chosen (Simon et al., 1999; Talaulicar et al., 2005). Therefore, the following hypothesis is proposed:
Hypothesis 1: Shared leadership has a direct positive relationship with decision making
comprehensiveness.
In general, researchers have characterized leadership as representing a collective influence process (Astin & Astin, 1996; House & Aditya, 1997; Sivasubramaniam et al.,2002).
Although this is expected to lead to more comprehensive decision making, it might also slow down the pace of decision making. Shared leadership will lead to decision making based upon the collective cognitive resources of the team and it therefore will take more time than a decision taken individually (e.g. Goodstein et al., 1994; Talaulicar et al., 2005). After all, to benefit from the collective knowledge of the team, advice and information has to be gathered from the team before making the final decision. Therefore, the following hypothesis is set (H2):
12
Hypothesis 2: Shared leadership has a direct negative relationship with decision making
speed.
2.5 Shared leadership, Team affective tone and decision making
Recently, a study of Hmieleski et al. (2012) also took a shared perspective of authentic
leadership. They showed that team affective tone mediates the effect between shared authentic leadership and new venture performance. Additionally, cognitive conflict can be beneficial for decision making comprehensiveness (Amason, 1996) but without trust it can also lead to affective conflict (Amason 1996, Ensley et al., 2002). Affective conflict can lead to lower decision quality and also lowers the satisfaction level and team affect (Ensley et al., 2002; Ensley and Pearce (2001).This makes clear that theories such as a affective event theory (AET) also seems to play an important role with respect to decision making.
AET originates from the psychological field (e.g. Sy et al., 2005; Cole et al., 2008) and is being integrated into the disciplines of organization behavior (e.g. Tsai et al., 2012; Chi et al., 2011). According to Marks et al. (2000), the two most important mechanisms that link inputs and outputs are behavior-based processes and the so called emergent states. The team processes have already received attention by entrepreneurship scholars. The team processes is about how a team as a whole (e.g. through activities) tries to convert resources into outcomes (LePine et al., 2008). Emergent states have received less attention in the literature. These are about the overall climate of a team, which include aspects such as creativity and trust (Klotz et al.,2014). According to Barsade and Gibson (2007), emergent states can be separated into affective and cognitive constructs. The literature on NVT has progressed in gaining more insights of the cognitive constructs. The main focus has been on NVT collective cognition (Chowdhury, 2005; West, 2007; Vissa and Chacar, 2009) and NVT cohesion (Foo et al.,2006; Vissa and Chacar, 2009).
13
attention to date in the context of NVTs (Klotz et al.,2014). Affective Events Theory (AET) with respect to work experiences has been introduced by Weiss and Cropanzano (1996). They argue that affective reactions are important to take into consideration because it can be an important mediator between work events and the following behaviors. Consequently, team research has used the affective events theory as the basis of several mediating affect
constructs (Klotz et al.,2014). George (1990) defined group affective tone as: “consistent or
homogeneous affective reactions within a group”(p. 108).
Several studies have been taken affective events into consideration when doing research on team performance. Pirola-Merlo et al. (2002) showed that overall team climate, hugely influenced by affective events, is strongly related to team performance. This is in line with earlier research on other affect-related concepts as mediators (e.g., Wilson-Evered et al., 2001; Smith-Jentsch et al., 2001. More interestingly, team leadership was proven to indirectly influence team performance through a mediating effect of team climate. So leadership showed to be closely related to the team affective tone. This was again confirmed by research of Chi et al. (2011). Leader positive moods showed to directly affect team performance.
Interestingly, leader positive moods also had an indirectly effect on team performance through team affective tone. All in all, in it clear that AET is relevant in team organized settings.
As mentioned before, A few studies already showed that leadership and affective event theory are closely associated (e.g. Chi et al., 2011; Hmieleski et al., 2012). Especially certain types of team designs are considered to be relevant. Self-managing and autonomous team designs are those in which team members have greater responsibility compared to manager led teams (Hackman, 1987). Carson et al. (2007) showed that self-management and autonomy are important antecedents of shared leadership. Additionally, the extent to which team
members feel autonomous and empowered positively relates to team affective tone (Kirkman and Rosen,1999). Therefore, hypothesis three is set as the following (H3):
14
Hypothesis 3: Shared leadership has a positive relationship with team affective tone.
Furthermore, Gibson and Earley (2007) argued that positive group affective tone will lead to more focus on positive information. Consequently, this leads to team members being more confident and certain in achieving their goals. More importantly, Rhee (2007) showed that a positive group affective tone leads to higher levels of energy, higher cooperation and more information and idea sharing. This particularly interesting because higher idea and
information sharing lead to a higher production of alternatives ,which in turn leads to higher decision making comprehensiveness (Hambrick and Mason, 1984). With regard to decision making speed, little is known about its relationship with team affective tone. However, a positive group affective tone increases morale-building communication, helping cooperation and the level of supporting each other’s ideas (Rhee, 2007). Together, these things could help to increase the efficiency of the decision making process. Building further upon the third hypothesis (H3), the following hypothesizes are proposed:
Hypothesis 4: Shared leadership has an indirect relationship with decision making
comprehensiveness mediated by team affective tone, so that shared leadership will increase decision making comprehensiveness through an increase in team affective tone .
Hypothesis 5: Shared leadership has an indirect relationship with decision making speed
mediated by team affective tone, so that shared leadership will increase decision making speed through an increase in team affective tone .
2.6 Shared leadership, team psychological safety and decision making
Besides team affective tone, psychological safety can also be an important mediator in the context of NVTs. Founders are likely to be more emotionally involved (Gimeno et al., 1997). Together with the fact that these founders have more legitimate power in the group, this may lead to challenges for the formation of favorable psychological safety (Klotz et al.,2014).
15
The construct of psychological safety has been introduced by Edmondson (1999). He defined psychological safety as: “ a shared belief that the team is safe for interpersonal risk
taking” (p.354). The concept has to be clearly distinguished from group cohesiveness that can
lead to a reduction of the willingness of challenging each other’s ideas, resulting in lowering interpersonal risk taking (Edmondson, 1999). Psychological safety is based upon the trust and respect among team members which results in safe overall team climate.
The concept has been picked up by many scholars over the years in which it showed to be particularly interesting in many contexts. Edmonson (1999) showed that psychological safety is closely associated with learning behavior in work teams. Besides, it functions as a mediator for antecedent such as team leader coaching and context support. These finding are confirmed by Carmeli and Gittell (2009) which also showed the importance of shared goals and knowledge. Later, the construct has been extended to an organizational climate for psychological safety (Baer and Frese, 2003). This means the overall climate in which team members feel free to speak up without any negative consequences. Such a safe climate turned out to be positively related to venture performance. Additionally, a safe psychological climate positively moderates the relationship between process innovation and venture performance (Baer and Frese, 2003). Leadership behavior is also closely associated with psychological safety. The construct functions as a crucial mediator between the relationship of leader inclusiveness with quality work improvements and creativity (Nembhard and Edmondson, 2006; Carmeli et al., 2010) and leadership behavior with employee voice behavior (Baer and Frese, 2003; Walumbwa and Schaubroeck, 2009). More recently, Bradley et al. (2012) used the construct of psychological safety to describe a specific context in which a team operates. Task conflict was positively related to team performance because psychological safety facilitates the performance benefits of task conflict. All in all, psychological safety seems to be a relevant mediator in many relationships in several contexts. However, it has not yet been
16
tested in the context of NVTs. (Klotz et al., 2014).
A favorable team psychological safety climate is one in which team members feel free to speak up without any negative consequences (Baer and Frese, 2003). Besides
self-management and autonomy, social support is an important dimension of the internal
environment that facilitates shared leadership (Carson et al., 2007). This consist out of all the behavior of team members to provide emotional and psychological strength to each another. Accordingly, this results in an environment where members feel that their efforts and ideas are valued which makes it more likely that members work cooperatively (Kirkman & Rosen, 1999; Marks et al., 2001). Hence, hypothesis six is set as the following (H6):
Hypothesis 6: Shared leadership has a positive relationship with team psychological safety. A psychological safe climate can have positive effects on the decision making process. When there is a safe environment in which there is much support and there are no negative
consequences for team members, it is more likely that they want to take the risk of proposing new ideas (West, 1990). Therefore, such an environment will positively influence learning behavior and creative potential of team members, which in its turn positively influences the level of decision making comprehensiveness. This is in line with the findings of Carmeli et al. (2013), which showed that when there is a high level of openness and generatively, a team is able to thoroughly process information. This eventually would have a positive effect on decision making comprehensiveness. In a psychological safe environment, other mechanisms through which a team can perform better are better team learning (Edmonson, 1999), higher job involvement (Brown & Leigh, 1996) and more efficient problem solving (Carson et al., 2007). Especially the last can be beneficial for the pace of decision making. Taken this all together, the following hypothesis are proposed (H7,H8):
17
Hypothesis 7: Shared leadership has an indirect relationship with decision
comprehensiveness mediated by team psychological safety, so that shared leadership will increase decision comprehensiveness through an increase in team psychological safety .
Hypothesis 8: Shared leadership has an indirect relationship with decision speed mediated by
team psychological safety, so that shared leadership will increase decision speed through an increase in team psychological safety .
(H8) (H7) (H6) (H3) (H5) (H2) (H1) Decision Making Comprehensiveness Psychological Safety Shared Leadership
Decision Making Speed
Figure 1: Conceptual framework Hypothesis overview
Team Affective Tone
18
3. Method
In this section the overall method of this study will be described. The general research design will be discussed and is followed up by how the data has been collected. Furthermore, it will be explained what the sample characteristics are and what measurements have been used. Afterwards, the statistical procedure is discussed. The final sub section contains the results of the reliability, common bias and factor analyses of the measurements.
3.1 Research method
This study is focused on getting more insights how shared leadership influences decision making through the emergent states team affective tone and psychological safety. Therefore, an explanatory approach is used because it intends to establish a causal relationship among the variables. The research should be described as mainly quantitative in nature because of the usage of surveys. However, a qualitative part is also added following up the survey. This consists out of four interviews with participants of the survey. This has been done to be able to get more insights into how and why certain answers are given in the survey, serving also to check if the theoretical reasoning used in this study is accurate. Taken together, this makes it a mixed method research.
So to give an answer to the earlier described research question and hypothesis, both surveys and interviews are used . The unit of analysis in this research is the NVT which can be defined as ‘the group of individuals that is chiefly responsible for the strategic decision
making and ongoing operations of a new venture”. (Klotz et al., 2014, p. 227). The
population out of which the sample has been subtracted is the Dutch new venture population. The data collection process is a joint effort of two master students from two Amsterdam universities. This is done to increase the reach in the Dutch new venture landscape. The sample procedure used in this study is purposive heterogeneous sampling, meaning that the
19
judgment of the researcher is important. However, it can be expected that getting access to and cooperation of new ventures is difficult. Consequently, a relatively low response rate is expected due to the fact that new venture teams are likely to be busy with managing their ventures in its early growth stage. Therefore, convenience sampling is likely to be present but is limited to the lowest level as possible. Through the use of multiple researchers in the data collection phase, the sample is expected to be diverse. Furthermore, the data collected through the survey is considered to be cross-sectional. Although longitudinal studies are preferred to infer causal relationships, in the timeframe available for a Master Thesis this is not feasible.
3.2 Data collection
The data has been collected through the use of digitally administered questionnaires. Collection started at the 26th of April, 2016. Planned was a collection period of about four weeks. Eventually, the survey was closed at the 25th of May, 2016. Participants had the
opportunity to complete the questionnaire in either English or Dutch. To find ventures that are in their early growth phase, several methods were used. As mentioned earlier, defining new venture on basis of age is not ideal because several boundary conditions are context-specific (Klotz et al., 2014). Consequently, searching for new venture only on basis of foundation date is not sufficient. The most important requirement for ventures to be allowed to participate in the survey is their growth phase. Ideally, the new venture should still be in its early growth phase. A specific industry in which the ventures operates has not been part of the criteria. First of all, the personal networks of the researchers were used. Both work for large ventures and their networks have proven to be valuable. Secondly, family members across the country have been asked if they knew any new venture in their personal network that might be suited for this questionnaire. However, the most important strategy was to use existing
20
all promote their startup landscape with extensive information and also have an online database with most of the new ventures in their city. One of the criteria included in these databases is the growth phase of the venture. Therefore, many new ventures have been found through these databases. Additionally, several institutions and communities helped to spread the message. Think of Startup Iamsterdam, Rockstart and Leiden Startup.
The new ventures that fulfilled the requirements were contacted by email and/or telephone. In return for the participation in the survey, they were given the opportunity to leave their contact information at the end of the survey. A document containing the results of both studies and its implications for the new ventures will be sent towards them upon
completion of the research. One week after the first invitation, a reminder has been sent to the ones that either did not complete the survey or did not respond to the invitation at all. For the ones that still did not responded after the first reminder, a second and last reminder was sent. In this reminder was clearly stated that the reminder would be the last one. From the
beginning to the end, participating in the survey was voluntary and the invitation could be ignored for any reason.
After closing the survey, the researchers moved towards the qualitative part of the data collection. Several participants in the survey were contacted with the request if they were prepared to answer some questions with regard to their completed questionnaire. In total, four interviews were conducted to obtain more insight into why certain answers were given and if these are in line with the theoretical reasoning used for this research.
3.3 Sample
In the data collection process, approximately 120 Dutch new ventures were contacted through either email or telephone. The new ventures are mainly located in the area of the cities of Amsterdam, Rotterdam, Den Haag, Hilversum and Ede. The amount of respondents that
21
started the survey is 124. Of this group, 62 respondent actually completed the survey. Of this group, one respondent had to be taken out because the venture did not met the criteria with respect to growth phase. Consequently, the total sample left consist out of 61 respondents. These 61 respondents belonged to 50 different NVTS. The dropout rate of 50% is relatively high but was expected. About 30% dropped out after the first general questions. It is assumed that the reason for this is that the respondents decided that they did not wanted to participate after all due to time issues. A look at the other points at which respondents dropped out and stopped their survey gives no clear clarification of why this is the case. The points of dropout vary substantially. It is believed that it is most likely that respondents dropped out by the same reason mentioned before, due to time issues. It can be expected that respondents had little time to fill in this survey next to their work obligations.
The venture age of the new ventures turned out be on average 23,52 months
(SD=18,66) and the average size of the venture in terms of employees is 7.97 (SD=14,62). Furthermore, the size of the venture team is on average 2,61 (SD=1,14). The respondents were also asked to mention the industry in which the new venture operates in. In total, the sample consists out of ventures of 12 different industries. The new ventures are mainly
operating in the service industry (29,5%), IT (16,4%) and hotel and catering industry (11,5%). 3.4 Variables and measurements
In collecting the data, the focus in this experiment is on five variables. These are shared leadership, team affective tone, psychological safety, decision making speed and decision making comprehensiveness. Also, additional variables were measured to control for other internal and external effects.
22 Translation
The scales used in this research are subtracted from English studies. As mentioned earlier, participants had the opportunity to complete the questionnaire in either English or Dutch. Therefore, all items had to be translated from English to Dutch. To make sure that in this process no mistakes were made and no valuable content will be lost, translations have been checked by the other researcher. Additionally, a third person also checked if no mistakes have been made. On basis of this, some minor corrections had to be made.
Independent variable Shared leadership
Shared leadership will be measured with a 8 item scale developed by (Mihalache et al., 2014). This scale has a Coefficient α of 0.87. An example item is: ‘Please indicate how much you
agree with the following statement: The new venture team jointly determines the
implementation of new business:’. Measured on a 7 likert scale ranging from 1 (strongly
disagree) to 7 (strongly agree). A high score indicates a high level of shared leadership within the new venture team.
Mediator Team affective tone
Team affective tone will be measured with a 20 item scale developed by (Watson et al., 1988). Validated by Watson et al. (1988) and Chi et al (2011) with a Coefficient α ranging from 0.84 of 0.90 and consists out of 2 subscales. There is one subscale for positive affect and one for negative affect. An example item is: ‘I have felt interested since I work for this new
venture”. Measured on a 5 likert scale ranging from 1 (never) to 5 (always). After recoding
the negative affect subscale, a high score indicates a positive team affective tone.
The original scale gives researchers the opportunity to choose the timeframe over which participants have to reveal how they have felt. There has been debate over which timeframe is ideal under certain circumstances. When a study wants to measure moods at
23
specific events and time, than a timeframe of two weeks should be chosen (Robinson and Clore, 2002). However, if the time frame is longer than these two weeks, participants have to rely on their generalized beliefs about their affect (Chi et al., 2011). For this study, the aim is not to measure moods at specific events. It is intended to measure the overall affect while working in the new venture team. Therefore, it is believed that the usage of a timeframe longer than 2 weeks is justified.
Mediator Psychological safety
Psychological safety will be measured with a 7 item scale developed by (Edmondson, 1997). Validated by Nembhard and Edmondson (2006) with Coefficient α of 0.73 (N=7) and Baer and Frese (2003) with Coefficient α of 0.82 (N=6). An example item is: ‘if you make a
mistake on this team, this is often held against you’. Measured on a 7 likert scale ranging from
1 (strongly disagree) to 7 (strongly agree). A high score indicates a feeling of being psychological safe.
Dependent variable Decision making comprehensiveness
Decision making comprehensiveness will be measured with a 5 item scale developed by (Miller et al. ,1998). Validated by Atuahene-Gima and Li (2004) with Coefficient α of 0.82 (N=7) An example item is: ‘The new venture team conducted multiple examinations of
suggested course of action’. Measured on a 7 likert scale ranging from 1 (strongly disagree) to
7 (strongly agree). A high score indicates a process of comprehensiveness decision making.
Dependent variable Decision making speed
Decision making speed will be measured with a 3 item scale developed by (Wally and Baum, 1994) which showed to have a Coefficient α of 0.70. An example item is: ‘When this company
24
ranging from 1 (strongly disagree) to 5 (strongly agree). A high score indicates that decision making speed of the new venture is high.
Control variable Environment uncertainty
A particular important control variable used in this study is that of environment uncertainty. In several studies, the level of environment uncertainty has proven to be an important moderator to relationships involving decision making comprehensiveness and decision making speed (Fredrickson, 1984; Eisenhardt, 1989; Heavey et al., 2009). So to accurately measure the effects of the 5 main variables, controlling for environment uncertainty is needed. Environment uncertainty will be measured with a 8 item scale developed by Hoque (2004) and with a Coefficient α of 0.75. An example item is: ‘Indicate the relative
predictability of the firm’s external environment: Suppliers’ action’. Measured on a 5 likert
scale ranging from 1 (very unpredictable) to 5 (very predictable). A high score on this scale means that the environment is predictable and that there is little environment uncertainty.
Remaining control variables
Besides environment uncertainty, other more simple control variables are measured. First of all, the size of the new venture is measured. This is measured as the number of full time employees. This is relevant because the amount of slack available to a venture is expected to be higher for larger ventures than for small ones. A large venture can posses greater slacks which can be beneficial in overcoming the costs of decision comprehensiveness (Atuahene-Gima and Li, 2004). In line with environment uncertainty, a venture’s industry is also taken into account because each industry demands other levels of decision making speed or ventures might have other goals in term of the tradeoffs between comprehensiveness and speed.
Secondly, the size of the new venture team is also controlled for because this can influence the level of debate and conflict. Consequently, this might influence decision making speed and comprehensiveness (Amason and Sapienza, 1997; Simon et al., 1999).
25
3.5 Statistical procedure and qualitative analyses method
The raw data needs to be checked on missing values or any other errors that might occur. If this is not done, this could have consequences for the hypothesis testing. First of all, there has been checked if there are any errors in the data by running a short frequency analysis. No errors were found. Secondly, the data was checked for missing values. Unfortunately, several of the measured items had some missing values. To solve this problem, only the cases without missing data are analyzed. As mentioned in the previous section, of the remaining 62
respondents one had to be taken out because it did not fit the criteria of a new venture. In the survey, counter-indicative items were used. These were re-coded to make them ready for further analyses. Afterwards, a reliability analyses has been conducted to test if the used scale were internally consistent. Next, the scale means of the variables are computed. These are needed to run the required analyses later on. At last, the scale means of the individuals were used to compute the scale means on the team level because this is the unit of analyses. From the measured variables, only team affective tone and psychological safety are normally distributed (see table A, appendix). This is shown by both the Kolmogorov-Smirnov and Shapiro-Wilk tests (p> 0.05). For shared leadership, decision making speed, decision making comprehensiveness and environment uncertainty the data is negatively skewed and not normally distributed (p<0.05). This means that the respondents tend to give high scores with respect to these variables. This relative high scores might be explained by the fact that entrepreneurs are in general very optimistic about their chances (Fraser and Greene, 2006). Therefore they might think that they can predict their environment well and collectively make fast and thorough decisions in comparison with competitors.
The main analyses part will start with a correlation analyses to take a first look at if there is something going on in the dataset with respect to the variables and if the proposed relationships possibly might be present. However, correlation does not imply causality and
26
therefore a regression analyses has been done to test for direct and mediation effects. Control variables were also included in the regression analyses. For the analyses the process macro written by Hayes (2013) for SPSS is used. As mentioned before, some variables were not normally distributed. This assumption violation can be overcome by performing
bootstrapping. This method is preferred over a Sobel test because bootstrap confidence intervals make more realistic assumptions with respect to normality (Hayes, 2009). The number of bootstrap samples for bias corrected confidence intervals in this analyses was 5000 to make sure the required level of confidence was reached. This is in line with the
recommendation of Hayes (2013).
With respect to the four interviews which had been done after the closing of the survey, the following procedure has been used. First of all, the transcripts have been created manually. Secondly, the data was code manually. Both open and axial coding were part of this procedure. In general, this was a mainly deductive approach because the focus was on the variables used in quantitative part. However, an inductive approach was also integrated by looking at related variables and keeping an open mind for new approaching themes.
Afterwards, the codes were grouped and organized into an overview. At final, together with the notes taken during coding the transcripts, the search for patterns and relating ideas was conducted.
3.6 Reliability and common bias analyses
It is necessary to assess the scale reliability. The Cronbach’s Alpha is most commonly used to test for internal scale reliability. Of the total sample, the ones that did not complete the
questionnaire were removed from the computation of the Cronbach’s Alpha. The test has been run for shared leadership, team affective tone, psychological safety, decision making
comprehensiveness, decision making speed and environment uncertainty. The results of the Cronbach’s Alpha test can be found in table 1. Of the six variables, 4 of them
27
have Cronbachs alpha > .7, which indicates high level of internal consistency. Only psychological safety and environment uncertainty do not show a high level of internal consistency.
According to George and Mallery (2003) the following rules of
thumb regarding Cronbach’s Alpha are: “_ > .9 – Excellent, _ > .8 – Good, _ > .7 – Acceptable, _ > .6 – Questionable, _ > .5 – Poor, and _ < .5 – Unacceptable”
(p. 231). Environment uncertainty has a Cronbach’s Alpha of r= .460. If one of the items of the scale would be deleted, the reliability will increase to a maximum score of r=.484. Therefore, the information of the environment uncertainty scale is not reliable and cannot be used for further analyses. With respect to Psychological safety, the score of r=.682 is
questionable. If one of the items would be deleted, no increase in the Cronbach’s Alpha would be achieved. It has been decided to proceed with the data of psychological safety because the Cronbach’s Alpha is very close to r=.7 and the scale has been proven to be reliable in earlier studies.
The scales used in this study are self-reports and therefore common bias can be present. A Harman’s single factor test has been conducted. In the dataset, the maximum variance explained by a single factor is 25.48 percent. Therefore, the issue of common bias seems not to be present because the maximum variance explained by a single factor is below 50 percent.
3.7 Factor analyses
At final, a factor analyses has been conducted. For the results of this analyses, see Appendix D. A principal components factoring analyses was conducted on the scales together with a varimax rotation when more than one component was found. For all the scales, the Kaiser– Meyer–Olkin measure verified the sampling adequacy for the analysis, KMO >.6. Also the Bartlett’s test of sphericity, p < .001, indicated that correlations between items were
28
sufficiently large for all scales to fulfill the principal component factor analyses. An initial analysis was run to obtain eigenvalues for each component in the data. For shared leadership, decision making comprehensiveness and decision making speed just one component had an eigenvalue over Kaiser’s criterion of 1. The scree plot confirmed this. These components explained respectively 72.9%, 57.3 and 74.% of the variance.
For psychological safety, three components had eigenvalues over Kaiser’s criterion of 1. The scree plot confirmed this. These components explained 31.2%, 21.6 and 15.3% of the variance. To measure psychological safety, the last 3 items of the scale will be used for
further analyses because they explain the most variance and are the least extreme. Therefore it is believed that these items are the most suited as accurate predictors.
With respect to team affective tone, five components had eigenvalues over Kaiser’s criterion of 1. The scree plot confirmed this. These components explained 28.9%, 22.0%, 7.3%, 5.8% and 5.2%. of the variance. The first two components seemed to be the most relevant due to their relatively high level of variance explanations. Therefore, a principal components factoring analyses was conducted with an extraction of a fixed number of factors. This was set at 2 factors because the scale consists out of both positive and negative
measurements. Indeed, the items that cluster on the same factors suggest that factor 1 represents positive team affective tone and factor 2 represents negative team affective tone. This is line with the study of Tsai et al. (2012) that made a clear distinction between these two concepts. These two components will be used separately in further analyses.
As mentioned earlier, environment uncertainty turned out to be not internally
consistent. Also for the factor analyses, the Kaiser–Meyer–Olkin measure barely verified the sampling adequacy for the analysis, KMO >.5. The result is sufficient but not ideal. The Bartlett’s test of sphericity, p < .001, indicated that correlations between items were
29
components factoring analyses was conducted with an extraction of a fixed number of one factor. This is done with the purpose of getting more information about which items might be the most suitable to be used as a single item scale in further analyses. Using all items as a single item scale is not preferable due to the relatively small sample used in the analyses which might lead to problems with the degrees of freedom while doing regressions. Item 1 to 3 seems to go well together and explain 22,5% variance. The use of these items are with respect to suppliers’ action, customers’ demands, tastes and preferences and market activities of competitors. The four interviews conducted in this study confirm that the respondents mainly based their rating of environment uncertainty on basis of these aspects. The other issues were not mentioned or mentioned less frequently.
30
Table 2
Means, Standard Deviations and Correlations
Variables Number of Items M SD 1 2 3 4 5 6 7 8 9 10
1. Shared Leadership 8 5.29 1.38 -
2. Positive Team Affective Tone 10 4.08 .56 .286* -
3.Negative Team Affective Tone 10 4.00 .54 -.320* -.051 -
4, Psychological Safety 7 5.15 1.16 .659** .574** -.274 -
5. Decision Making Comprehensiveness 5 4.54 1.18 .493** .507** -.029 .639** -
6. Decision Making Speed 3 3.79 .95 .138 .525** .108 .346* .146 -
7. Venture Age 1 23.52 18.66 .061 -.034 -.020 .064 -.201 .068 -
8. Size New Venture 1 7.97 14.62 .116 .025 .039 -.002 -.259 .279 .090 -
9. Industry 1 4.90 2.85 .045 .062 -.032 .033 -.065 .051 .268* -.064 -
10. Size New Venture Team 1 2.61 1.14 .401** -.029 -.383** .350* .088 .118 -.030 .268* -.089 -
Note: *Correlation is significant at the .05 level (2-tailed) ** Correlation is significant at the .01 level (2-tailed)
Table 1
Variable Cronbachs Alpha
Shared Leadership .944
Team Affective Tone .835
Psychological Safety .682 Decision Comprehensiveness .803
Decision Speed .819
31
4. Results
In this section the results of the statistical procedures and qualitative research will be presented. Firstly, the results of the correlation analyses will be discussed. Secondly, the hypothesis testing will be presented by discussing the results of the regression analyses. At final, the main findings of the qualitative research will be discussed.
4.1 Correlations
In table 2, the correlation matrix can be found. It will give us a first idea about if the hypothesizes can be proven in this study and if their might be some other interesting things going on in the data. The most logical variables to take a look at are the main studied
variables in this study. First of all, shared leadership is positively correlated to psychological safety (r=.659, p<.01), suggesting that there might be a positive relationship between shared leadership and psychological safety as proposed in hypothesis six. On its turn, psychological safety is positively correlated to both decision making comprehensiveness (r=.639, p<.01) and decision making speed (r=.346, p<.05), strengthening the idea of a mediation effect as
proposed in hypothesizes 7 and 8. Also in line with hypothesis 3 , shared leadership is correlated to both positive (r=.286, p<.05) and negative (r=-.320, p<.05) team affective tone Furthermore, positive team affective tone is positively correlated to both decision making comprehensiveness (r=.507, p<.01) and decision making speed (r=.525, p<.01). Together, this strengthens the idea of a mediation effect of team affective tone.
As proposed by hypothesis 1 and 2, it was suggested that there is a positive direct relationship between shared leadership and decision making speed and decision making comprehensiveness. The correlations show a positive correlation of the former (r=.493,
p<.01) and no correlation for the latter.
32
new venture team size, there is a positive correlation with shared leadership (r=.429, p<.01) and new venture size (r=.268, p<.01). The latter feels logical because larger ventures are likely to have more departments and therefore more department managers. These manager are probably to at least some extent involved in decision making. A simple explanation for the first could be that in a larger new venture teams, the chance that someone is more specialized in a certain skills than others is higher. This higher variety of skills could lead to higher level of authority shifts, leading to a higher feeling of shared leadership within the team. At last, size new venture team is also correlated with negative team affective tone (r=-.383, p<.01) and psychological safety (r=.350, p<.05). Although out of the scope of this study, this might be due to the fact that when the members of the team feel that the internal climate of the team is psychological safe and there is little negative team affective tone, team member turnover will be lower. After all, the team will be more attractive and therefore results in a larger team size.
33
Consequent
M (Psychological Safety)
M (Positive Team Affective
Tone)
M (Negative Team Affective Tone)
Antecedent Coeff. SE p Coeff. SE p Coeff. SE p
X (Shared Leadership) a .45 .11 < .001 .11 .07 .13 c' -.09 .06 .15
M (Psychological Safety) - - - - b1 - - -
M (Positive Team Affective Tone) - - -
- - - b2 - - -
M (Negative Team Affective Tone) - - -
- - - b3 - - -
Constant i1 1.44 .86 .10 i2 3.30 .54 < .001 i3 4.76 .49 < .001
R =.50 R =.17 R =.24
F(8,41)=5.21, p=<.001 F(8,41)=.1,02, p=.44 F(8,41)=1.58, p=.16
Table 3: Results Regression Analyses Mediation
Note: N=50. Size new venture team, size new venture, industry, firm age and 3 single item scale environment uncertainty were included as control variables
Consequent
Y (Decision Comprehensiveness) Y (Decision Speed)
Antecedent Coeff. SE p Coeff. SE p
X (Shared Leadership) c' .25 .12 <.05 c' -.09 .13 .47
M (Psychological Safety) b1 .41 .17 <.05 b1 .14 .18 .43
M (Positive Team Affective Tone) b2 .41 .26 .12 b2 .81 .28 <.01
M (Negative Team Affective Tone) b3 .38 .24 .12 b3 .33 .26 .21
Constant i4 -.59 1.59 .71 i5 -1.72 1.69 .31
R =.65
R =.40
F(11,38)=6.53, p=<.001 F(11,38)=2.27, p=<.05
Table 3 Continued: Results Regression Analyses Mediation
Note: N=50. Size new ventue team, size new venture, industry, firm age and 3 single item scale environment uncertainty were included as control variables
34
Direct Effect
on Y (Decision
Comprehensiveness) Coeff. SE p. LLCI ULCI t
X (Shared Leadership) .25 .12 <.05 .0122 .4884 2.13 Indirect effect on Y (Decision Comprehensiveness) Total .19 .15 -.0702 .4999 Psychological Safety .18 .11 .0194 .5097
Positive Team Affective Tone .05 .10 -.0745 .3615
Negative Team Affective Tone -.04 .04 -.1710 .0153
(Constant 1) -.22 .12 -.5523 -.0311
(Constant 2) .08 .12 -.0842 .3927
(Constant 3) -.14 .15 -.4379 .1404
Total Effect X on Y .44 .12 <.001 .2077 .6787 3.80
Table 4: Total, Direct and Indirect Effects Regression Analyses Decision Making Comprehensiveness
Direct Effect
on Y (Decision Speed) Coeff. SE p. LLCI ULCI t
X (Shared Leadership) -.09 .13 .47 -.3444 .1616 -.73 Indirect effect on Y (Decision Speed) Total .12 .15 -.1475 .4329 Psychological Safety .06 .09 -.1027 .2863
Positive Team Affective Tone .09 .12 -.0815 .3944
Negative Team Affective Tone -.03 .04 -.1645 .0223
(Constant 1) -.09 .11 -.3540 .0840
(Constant 2) .12 .14 -.0911 .4335
(Constant 3) .03 .15 -.1491 .5071
Total Effect X on Y .03 .12 .81 -.2158 .2743 .24
35
4.2 Hypothesis testing
4.2.1 Direct effects of shared leadership on decision making comprehensiveness and speed (H1,H2)
To test the mediations hypothesis proposed in this paper, the process macro written by Hayes (2013) for SPSS is used. The conceptual and statistical Process model for simple mediation, which were used in the analysis are represented in Appendix B. M represents the mediator(s), X the independent variable, Y the dependent variable. Additionally, the control variables size new venture, size new venture team, industry, venture age and the three single item scales of environment uncertainty were integrated into the regression analyses as covariates. The results of the regression analyses can be found in the tables 3-5.
First of all, the direct effect of shared leadership on both decision making comprehensiveness and decision making speed are tested. The direct effect of shared leadership (c′ = .25) on decision making comprehensiveness is statistically different from zero, t=2.13, p = <.05 with a 95% confidence interval from .0122 to .4884. This direct effect of .25 means that two new venture teams who differ one unit in their reported shared
leadership are estimated to differ by .25 units in their reported decision making comprehensiveness. Therefore, the first hypothesis (H1) is supported.
The direct effect of shared leadership (c′ = -.09) on decision making speed is not statistically different from zero, t=.-.7315 p = >.05 with a 95% confidence interval from -.3444 to .1616. Therefore, no inferences can be made from the data and the second
hypothesis (H2) is not supported. Even if the coefficient was found the be significant, the effect would have been small.
