The catalysts of exploration : the relation of team diversity to explorative team behavior

(1)

Master’s Thesis

The catalysts of Exploration

The relation of team diversity to explorative team behavior

Laura K. Wehrheim

Student number: 10825665 Date of final version: June 29, 2015

Master’s programme: Business Administration Specialisation: Strategy

Supervisor: Pepijn van Neerijnen

Abstract

Regardless of the extensive research on workgroup diversity and exploration, few studies have examined the opposing effects of team diversity simultaneously. In this thesis an attempt is made to consolidate the different perspectives on the diversity-exploration-relationship, thereby explaining the inconsistent findings of the empiri-cal literature. The influential effects of three types of team diversity on team explo-rative behavior are tested: Diversity in social categories, Diversity in knowledge and expertise and Diversity in cognitive styles. To test the moderating influence of a teams’ information processing capabilities, the moderators TMS and Team identity are included in the research design. The hierarchical regression analysis, using 480 employees on 96 workgroups, indicates that there is no statistically significant rela-tionship between Diversity in social categories and cognitive styles and Exploration. The relation of Diversity in knowledge and expertise and Exploration has shown to be statistically significant and positive.

(2)

Statement of Originality

This document is written by Laura K. Wehrheim 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.

Acknowledgements

I would like to thank my supervisor Pepijn van Neerijnen for his enthusiastic supervision. His feedback and the group sessions helped me to stay motivated during the writing process of this thesis. He further encouraged me to constantly improve my writing and analytical skills, which increased the overall quality of this thesis.

I thank my family, Stefan, Gabi and Viktor, for enabling me to do my Master abroad. I am grateful for their unconditional support of all my decisions and I appreciate us keeping such close contact throughout this chapter of my life.

I further like to express my appreciation for the support of all of my friends, the ones back home and the new ones I found here in Amsterdam. The long evenings spent at the library, the support with Latex, the coffee breaks and the mutual struggle with SPSS are memorable. It made my year in Amsterdam unforgettable.

(3)

Introduction

Diversity appears to be a double-edged sword, increasing the opportunity for creativity as well as the likelihood that group members will be dissatisfied and fail to identify with the group.

- Milliken & Martins, 1996: 403

To cope with environmental changes of fast moving customer needs, technologies and compe-tition; companies need to engage in exploring new possibilities to renew themselves continuously (Danneels, 2002; Mom, Van Den Bosch & Volberda, 2009). Exploratory product innovations have been found to be the primary instrument to achieve corporate renewal and long-term orga-nizational survival (Dougherty, 1992; Levinthal & March, 1993; Gino, Argote, Miron-Spektor & Todorova, 2010). In order to generate innovative products, organizations require internal variety (Ashby, 1952), which in turn is associated with explorative behavior (March, 1991). McGrath (2001) emphasizes that the major difference between a firm that is successful in adapting to its environment and one that is not, is its engagement in explorative behavior (Levinthal & March, 1993; March, 1991).

Diversity is assumed to be one of the most important influential factors on explorative behavior (Hoever, Van Knippenberg, Van Ginkel & Barkema, 2012; Jehn, Northcraft & Neale, 1999). Teams would not be able to produce different outputs, if its members approached problems in the same manner, sharing the same opinion and solutions (Williams & O’Reilly, 1998). Workforces nowadays have become and will continue to become increasingly diverse

(6)

on a large number of dimensions (Hollenbeck, Beersma & Schouten, 2012; Van Knipppenberg & Schippers, 2007; Van Dijk, Van Engen & Van Knippenberg, 2012). Scholars propose that the major challenge managers are facing in the twenty-first century is the diversity of their workforce (Williams & O’Reilly, 1998). Though prior research suggests that organizations find it difficult to realize the potential of diversity in teams (Van der Vegt & Bunderson, 2005). A growing body of literature engages in identifying factors influencing exploratory behavior in teams (Pearsall & Ellis, 2006; Ilgen, Hollenbeck, Johnson, & Jundt, 2005; Kozlowski & Bell, 2003). Still, the question remains: Which types of team diversity affect a team’s engagement in exploratory behavior, and which factors enhance this process?

Research on diversity in the past has been focused on problems which might arise from diverse work groups. As the research field on diversity evolved, attention focused on potential values of team diversity (Shore, Randel, Chung, Dean, Ehrhart & Singh, 2011). Diversity offers both great opportunity and challenges for an organization. On the one hand it sparks integrative insights, creativity and innovation through a greater range of perspectives, which is suppose to provide high-quality solutions (Harrison & Klein, 2007; Milliken & Martins, 1996; Shin, Kim, Lee & Bian, 2012). Workgroup diversity is assumed to have a positive impact on team creativity through an increase in ideas by a wider range of backgrounds and perspectives of the team members (Wang, Rode, Shi, Luo & Chen, 2013; Shin, Kim, Lee & Bian, 2012). The improvement of group performance via an increased information density is called the “Value-in-diversity hypothesis” (Cox, Lobel & McLeod, 1991). On the other hand “Value-in-diversity might foster conflict, division and dissolution leading to higher levels of dissatisfaction and employee turnover (Harrison & Klein, 2007; Milliken & Martins, 1996; Paulus & Nijstad, 2003). Representatives stress the performance suppressing effect of diversity (Bell, Villado, Lukasik, Belau & Briggs, 2011; Brewer, 1979; Guzzo & Dickson, 1996). Differences within a group may lead to an increase in conflict and communication difficulties, which in turn foster emotional and task conflict and resistance to participate in group processes (Paulus & Nijstad, 2003; Jehn, 1995; Camacho &

(7)

Paulus, 1995; Wang, Rode, Shi, Luo & Chen, 2013).

Although plenty of insights have been generated so far, empirical research on workgroup di-versity still generates inconsistent results (Harvey, 2013; Hoever, Van Knippenberg, Van Ginkel & Barkema, 2012). Research examining group diversity has reported positive relationships be-tween diversity and performance in some cases and negative or non-existing relationships in other cases (Van der Vegt & Bunderson, 2005). Reason for these mixed findings might be that the relationship of creativity and diversity is more complex than constructed in empiri-cal tests (Shore, Randel, Chung, Dean, Ehrhart & Singh, 2011; Bell, Villado, Lukasik, Bela & Briggs, 2011). Another reason might also be the varying typologies of diversity in the sev-eral research designs (Van Knippenberg & Schippers, 2007; Hollenbeck, Beersma & Schouten, 2012). Different kinds of diversity are supposed to influence explorative behavior in different ways. The prevalent theoretical perspectives on team diversity, namely Social Categorization, Similarity/Attraction paradigm and Information- and Decision-making theory, cannot explain these mixed findings, as they take one-sided perspectives. Therefore, a more holistic picture of team diversity is tried to be established by splitting it into the three sub-levels: Diversity in social categories, in knowledge and expertise and in cognitive styles.

As work in team settings gets increasingly important in the information age (Van Knippen-berg, Van Ginkel & Homan, 2013; Hollenbeck, Beersma & Schouten, 2012; Shin, Kim, Lee & Bian, 2012), this thesis addresses the interaction of diversity and exploration at the team level. Teams are complex, adaptive, dynamic systems and need to be approached differently than just a gathering of individuals (McGrath, Arrow & Berdahl, 2000). While some groups are able to recognize opportunities and engage in exploratory innovation, others fail to do so (Alexiev, Jansen, Van Den Bosch & Volberda, 2010). There still exists a lack of understanding about the antecedents of exploratory behavior and what enables and motivates exploration at the group level (Jansen, Van Den Bosch & Volberda, 2006; Gupta, Smith & Shalley, 2006; Hoever, Van Knippenberg, Van Ginkel & Barkema, 2012). There seems to be consensus around the link of

(8)

exploration and innovation, which seems to lack when it comes to the relation of exploration and team characteristics, environmental factors to exploration, or how organizations coordinate the development of exploration (Gupta, Smith & Shalley, 2006; Jansen, Van Den Bosch & Volberda, 2006). The ability of a team to process information appears to be an important antecedent of explorative behavior as it enables team members to utilize their common knowledge. The trends of more reliance on self-management by teams, fewer supervision, increased immigration, glob-alization of firms and an aging workforce increase the importance of understanding the effect of group diversity on explorative behavior (Williams & O Reilly, 1998; Shalley, Zhou & Oldham, 2004). With the awareness of how team creativity is stimulated, managers are able to compose teams in an exploration-enhancing way. This fosters the long-term unique competitive advan-tage of an organization, as innovative team output is evidentially linked to firm performance and organizational survival (Amabile, 1988; in Shalley & Gilson, 2004; Nystrom, 1990; in Shalley & Gilson, 2004).

Factors that may moderate the relationship of team diversity and exlorative behavior are Team identity and Transactive Memory System. Both moderators enable better information processing, which is a psychological process enabling team members to access the broader range of information and perspectives in diverse teams (Sommers, Warp & Mahoney, 2008). The identification of the single members with their team (Team identity) might reduce perceptions of dissimilarity and in-group competition, which in turn fosters the psychological acceptance of diversity. This enables a better transmission from group diversity into team exploration. The existence of a TMS, which enables an effective information exchange between the team members, is supposed to increase the creative outcome quantity as team members are even more exposed to different perspectives.

The inconsistent findings on the effects of workgroup diversity are addressed through the three-dimensional assessment of team diversity. In doing so, this thesis contributes to the literature on team exploration by distinguishing opposing effects of diversity on exploration

(9)

and by helping to consolidate the different perspectives on the relationship of diversity and exploration.

In the next section the theoretical background of this paper is elaborated. Within, the conceptual model and hypotheses including possible moderating variables are developed. The method and result sections are following. Finally, a discussion part on the findings and possible limitations to the study as well as theoretical and managerial implications are attached.

(10)

Chapter 2

Theory and hypotheses

2.1 Team Exploration

Exploration is defined through “processes by which organizations create variety in experience through experimentation, trailing, and free association” (Marengo, 1993, in Holmqvist, 2004: 70). The concept includes “terms such as search, variation, risk taking, experimentation, play, flexibility, discovery, innovation” (March, 1991: 71). Explorative behavior is defined by the development and adaptation of new operational and administrative routines (Crossan, Lane & White, 1999; Zollo & Winter, 2002; McGrath, 2001), the search for new approaches to technologies, businesses, processes and products (McGrath, 2001), as well as the modification of existing values and decisions (Floyd & Lane, 2000). It is further associated with the creation of new competences (Floyd & Lane, 2000), the creatively development of radical innovations (Beugeldijk, 2008) and variety increasing learning (McGrath, 2001).

Exploration has become an underlying theme in various related studies, of which all acknowl-edge the importance of exploratory behavior for organizations operating in dynamic environ-ments (Jansen, Van den Bosch & Volberda, 2006; Mom, Van Neerijnen, Reinmoeller & Verwaal, 2015). Studies on innovation emphasize the creation of new knowledge and the departure from existing capabilities in order to develop the innovations needed to deal with market and tech-nological change (Alexiev, Jansen, Van den Bosch & Volberda, 2010; Benner & Tushman, 2003; Danneels, 2002). Similarly, studies on organizational learning stress the importance of

(11)

vari-ety increasing learning activities, by which a person renews and broadens his or her existing knowledge base. This experiential-based learning enables the creation of variety in experience, essential for an organization to adapt effectively (McGrath, 2001; Holmqvist, 2003; Levinthal & March, 1993). Studies on strategic management address the experimentation with new skills or market opportunities and the utilization of new knowledge to overcome existing, inertial compe-tencies (Floyd & Lane, 2000). To conduct a strategic renewal successfully and diverge from the existent strategy, organizational members need to experiment with the selection of products, the market segments, the business model and current practices (Augier & Teece, 2008). These processes are supposed to be facilitated by the existence of diversity.

2.2 Team Diversity

Diversity is a unit-level construct (Harrison & Klein, 2007). The derivation from the Latin word diversus, meaning contrary, implies that diversity is a relative concept. Thus, individuals can only be diverse in relation to other individuals (Austin, 1997). Team diversity refers to the degree to which group members differ from each other (Van Knippenberg & Schippers, 2007). When one says a work group is diverse, it means that its independent members differ in respect to one or more personal attributes (Milliken, Bartel & Kurtzberg, 2003; Jackson, Joshi & Erhardt, 2003).

In most studies on diversity, the construct is examined through readily observable attributes. Scholars differentiate between information diversity, such as educational and functional back-grounds, and social category diversity like gender, ethnicity or age (Van Knippenberg & Schip-pers, 2007; Guillaume, Brodbeck & Riketta, 2012; Jehn, Northcraft & Neale, 1999). The concept of diversity though has not been assessed in a consistent manner as some scholars made use of demographic data (e.g. Pelled, Eisenhardt & Xin, 1999; Bell, Villado, Lukasik, Belau & Briggs, 2011), while others included perceived value diversity, for example a self-rating by team members (e.g. Jehn, Northcraft & Neale, 1999).

(12)

Scholars stress that besides these observable and easily accessible diversity attributes, ones that are not readily observable, for example attitudes and values, should receive more scholarly consideration (Jehn, Northcraft & Neale, 1999; Van Knippenberg & Schippers, 2007). Further, Van Knippenberg and Schippers (2007) point out, that there is still a need to access diversity at more than one single dimension and to conceptualize diversity as a combination of several dimensions of differentiation.

Diversity can be assessed by three underlying processes: the Social categorization theory, the Similarity and Attraction paradigm and the Information and Decision-making perspective, all identified by Williams & O’Reilly (1998). Depending on the perspective taken, the effects of diversity are proposed to be either optimistic or pessimistic (Williams & O’Reilly, 1998; Van Knippenberg, Van Ginkel & Homan, 2013). Scholars arguing from the Value-in-diversity per-spective claim that diversity in group composition drives explorative team behavior (Williams & O’Reilly, 1998; McLeod, Lobel & Cox, 1996). The Information theory confirms that the more information and variety of approaches available in a work group, the more likely it is that some elements provide a novel approach to a particular problem (Milliken, Bartel & Kurtzberg, 2003; Nijstad & Paulus, 2003). The sharing of information may enable members to success-fully recombine old ideas and create something novel by applying it to the current tasks. This promotes explorative behavior and innovative outcomes (Austin, 1997; Bantel & Jackson, 1989; McLeod, Lobel & Cox, 1996). Recent research shows that the more diverse an organization, the more likely its financial returns are above its national industry medians (Hunt, Layton & Prince, 2015).

The Similarity-attraction paradigm and the Social categorization theory both claim that in-dividuals prefer to base their work processes on similarities in values and demographics, leading to a negative mindset towards exploration (Byrne, 1971, in Jehn, Northcraft & Neale, 1999; Tajfel, 1981, in Williams & O Reilly, 1998; Guillaume, Brodbeck & Riketta, 2012). Thus, indi-viduals in diverse groups are predicted to generate fewer ideas, because they are less interested

(13)

in the different ideas of others (McLeod, Lobel & Cox, 1996). Scholars disagreeing with the Value-in-diversity hypothesis further argue that diversity harms exploratory behavior of a team. Differences among group members may be a source of conflict and frustration, especially in the early formative phases of group interaction (Paulus & Nijstad, 2003; Jehn, 1995). This results in members paying less attention to each other’s contributions (Milliken, Bartel & Kurtzberg, 2003). Paulus and Brown (2007) even presume that the exposure to other ideas has a distract-ing or inhibitdistract-ing effect on team members’ idea generation. Still, fosterdistract-ing explorative behavior through team diversity is critical for an organization in order to stay innovative and to cope with its fast changing environment (Amabile, 1983; Shalley, 1991).

These inconsistent predictions on the effect of team diversity are probably based on the different approaches scholars took in defining the construct Diversity (Jehn, Northcraft & Neale, 1999; Van der Vegt & Bunderson, 2005; Hollenbeck, Beersma & Schouten, 2012). Literature on group diversity suggests that different dimensions of diversity have opposing effects (Harrison & Klein, 2007; Dahlin, Weingart & Hinds, 2005; Shin, Kim, Lee & Bian, 2012). Contrarians to the cognitive value perspective contest that diversity in values and cognition has a negative impact on group collaboration (Williams & O’Reilly, 1998; Kurtzberg, 2005), whereas diversity on information- related dimensions has a positive effect (Jehn, Northcraft & Neale, 1999; Pelled, Eisenhardt & Xin, 1999). In order to test if all types of diversity have a similar effect on team behavior and to integrate the optimistic and pessimistic views on diversity, there is a need to conceptualize diversity in a more complex manner (Williams & O’Reilly, 1998; Harrison & Klein, 2007; Dahlin, Weingart & Hinds, 2005). Furthermore, recent research has focused on meso-level effects of group composition by investigating the distribution of multiple attributes simultaneously instead of investing single individual attributes such as age in groups (Thatcher & Patel, 2012).

Following the recommendation of Van Knippenberg and Schippers (2007), diversity is defined by three facets in this thesis. Factors that contribute to the first facet are readily observable,

(14)

demographic attributes that are not very job related, namely gender and age. The second facet is more job-related attributes such as educational or functional backgrounds. And third, diver-sity is conceptualized by readily visible but not always job-related attributes like personality, attitudes and values. Concluding, diversity is being examined through three variables in this thesis, namely Diversity in social categories, Diversity in knowledge and expertise and Diversity in cognitive style (see Figure 1, p. 24). It is important to access all facets of diversity in order to detect possible opposing effects and to understand the main effect of team diversity.

2.2.1 Diversity in social categories

Diversity in social categories refers to differences between individuals regarding variables such as age, sex and race (Mannix & Neale, 2005). It is the most commonly analyzed facet of diversity and is included in this thesis for reasons of completeness. The proposed effect of Diversity in social categories is exploration-suppressing.

Diversity in social categories can be assessed through the Theory of Social categorization. The theory states that individuals tend to categorize themselves and others based on surface-level information like age, sex and race (Williams & O’Reilly, 1998; Bell, Villado, Lukasik, Belau & Briggs, 2011). Individuals use these easily observable, physical attributes instead of less salient attributes to categorize themselves and other group members into social groups (Bell, Villado, Lukasik, Belau & Briggs, 2011; Williams & O’Reilly, 1998). This act of social categorization leads group members to perceive in-group and out-group members differently and to rank individuals that are not included in their group as less trustworthy or honest (Mannix & Neale, 2005; Williams & O’Reilly, 1998). Although these attributes are of low job-relatedness, they are still considered to influence team performance negatively through social processes (Bell, Villado, Lukasik, Belau & Briggs, 2011; Williams & O’Reilly, 1998). As a consequence the advantages from diversity cannot be extracted, as the collaboration amongst diverse team members is dysfunctional.

(15)

Scholars predict that individuals in teams, which are diverse in social categories, are more likely to feel discomfortable, experience greater psychological distance and communicate and cooperate less (Bell, Villado, Lukasik, Belau & Briggs, 2011; Wang, Rode, Shi, Luo & Chen, 2013). It leads to a lower integration of individuals within the group and a higher likelihood of employee turnover (Bell, Villado, Lukasik, Belau & Briggs, 2011). These interpersonal processes hinder the collaboration between diverse team members. In this thesis, Diversity in social categories is accessed through diversity in gender and age, as these diversity attributes are most common in team compositions.

It is suggested that individuals of different gender than the majority of their group members, are less attached to the group (Milliken & Martins, 1996; Tsui, Egan & O’Reilly, 1992). This has an increasing effect on social categorization and in- and out-group perceptions. Consequently, levels of conflict and interpersonal tension are higher, levels of friendliness are lower, both leading to lower levels of explorative collaborative behavior (Kramer, 1991, in Wiliams & O’Reilly, 1998). Individuals with a similar age have developed similar perspectives and shared experience. Members of different ages lack these commonalities. This is supposed to complicate communi-cation within a team and to decrease the possibility of teams making use of diversity (Williams & O’Reilly, 1998). It is suggested that diversity in age tends to drive employee turnover, as members are less committed to their organization and less invested into team work (Williams & O’Reilly, 1998; Jackson, Brett, Sessa, Cooper, Julin & Peyronnin, 1991). The assessment of both diversity in age and gender leads to the overall hypothesis:

Hypothesis 1a (H1a): The variable Diversity in social categories is negatively related to exploratory team behavior.

(16)

2.2.2 Diversity in knowledge and expertise

Diversity in knowledge and expertise refers to diverse educational, functional, occupational backgrounds or a diverse range of industry experience (Milliken & Martins, 1996). There are several arguments for an effect of Diversity in knowledge and expertise on explorative team behavior, although inconsistent opinions about the direction of this effect exists. The camp supporting the Value-in-diversity hypothesis, argues in favor of an exploration-increasing effect by accentuating the extension of the group’s knowledge pool. This fosters task-related capa-bilities, specialization and learning processes (e.g. Mannix & Neale, 2005; Jehn, Northcraft, & Neale, 1999; Williams & O’Reilly, 1998). The opposing scholarly camp argues that prior knowledge just reinforces old routines, makes team members insecure and leads to an informa-tion overload, which suppresses explorative behavior (e.g. Gino, Todorova, Miron-Spektor & Argote, 2009; Van der Vegt & Bunderson, 2005; Milliken & Martins, 1996). The presumption for a positive relationship is that the knowledge and expertise of the group is relevant to the group task (Paulus & Brown, 2007). Diversity in knowledge and expertise is examined in this thesis to reveal which of the two effects is predominant. It is proposed that the effect of diversity in knowledge and expertise is exploration-stimulating.

Diversity in knowledge and expertise can be assessed best through the Information and Decision-making perspective. It addresses how the composition of a group effects its informa-tion and decision-making processes (Williams & O’Reilly, 1998; Gruenfeld, Mannix, Williams & Neale, 1996). The theory proposes that added information enhances group performance, independent from possible negative side effects on group processes (Williams & O‘Reilly, 1998). Consequently, Diversity in knowledge and expertise has a positive impact as skills, abilities, information and knowledge of group members increase. Therefore knowledge diverse groups are supposed to generate a broader range of knowledge than homogenous groups (Williams & O‘Reilly, 1998; Mannix & Neale, 2005). Teams can access a larger knowledge pool than just

(17)

their individual one and possess more task-relevant capabilities that help them to deal with non-routine problems (Williams & O’Reilly, 1998; Van Knippenberg & Schippers, 2007). The resulting increase in ideas and perspectives affects exploration and innovation positively (Jehn, Northcraft, & Neale, 1999; Wang, Rode, Shi, Luo, & Chen, 2013).

Prior expertise further improves the capacity of each member to create a new product as teams are able to understand task requirements better and learn from prior mistakes (Gino, Argote, Miron-Spektor & Todorova, 2010). Furthermore, task expertise enhances the team members’ ability to quickly specialize and contribute to new tasks in a useful way (Gino, Todor-ova, Miron-Spektor & Argote, 2009). Empirical findings have shown that diversity in expertise translates into a faster execution of creative ideas and that groups with more heterogeneous knowledge are higher in productivity (Stroebe & Diehl, 1994). A number of studies already proved a stimulating effect on creativity by an increase in idea generation (Dugosh, Paulus, Roland & Yang, 2000; Nijstad, Stroebe & Lodewijkx, 2002). The combination of individual knowledge, skills and abilities allows a group to engage better in explorative behavior than just its individual members (Nijstad & Paulus, 2003).

Another argument in favor of a positive relationship is the promotion of individual and collective team learning processes and search processes by Diversity in knowledge and expertise. Team learning processes are “activities by which team members seek to acquire, share, refine, or combine task-relevant knowledge through interaction with one another” (Argote, Gruenfeld & Naquin, 1999: 370, in Van der Vegt & Bunderson, 2005). An increase in learning and search processes fosters the collaboration between diverse, dissimilar team members. This exposure to new opinions and perspectives promotes team members’ explorative behavior through the cross-fertilization of ideas (Van der Vegt & Bunderson, 2005). The opportunity of novel combinations of knowledge elements initiates the creation of innovative ideas.

Levitt and March (1988) take a more pessimistic view on Diversity in knowledge and exper-tise. They argue that prior expertise can be counterproductive to explorative behavior because

(18)

team members might be locked up into old routines and are less likely to deviate from well-known practices. Team members might also take a narrower focus on past strategies or practices instead of searching for new ones (Gino, Todorova, Miron-Spektor & Argote, 2009). Individu-als with a high level of expertise are tempted to tap into their prior learnings. This leads to exploitive behavior instead of explorative one, which is in turn is associated with innovation. Scholars already proved that exploiting past knowledge instead of exploring new knowledge domains typically leads to general incremental innovation than to radical innovation (Benner & Tushman, 2003; Gupta, Smith & Shalley, 2006). Individuals with more expertise might create a higher number of ideas, but because they are sticking to strategies and practices that have been successful in the past, their ideas are incremental in nature and not radical (Audia & Goncalo, 2007).

Further, according to the Theory of Social comparison, individuals use the performance level of other group members as a comparative basis for their own appropriate level of performance (Paulus & Dzindolet, 1993). Team members tend to compare themselves with less fortunate others and converge towards the performance level of low performers in the group when there is no strong incentive for high performance, called the “Theory of downward comparison” (Wills, 1981). This leads to the tendency of group members to converge to each other and become more similar over time (Paulus & Dzindolet, 1993; Ziegler, Diehl & Zijlstra, 2000; Festinger, 1954). The lower-level performance through social matching and resulting uniformity within the group, might offset possible effects of diversity and lead to productivity losses (Paulus & Dzindolet, 1993; Festinger, 1954). Consequently, the downward-matching tendency of individuals has a reducing effect on team exploration (Camacho & Paulus, 1995; Paulus & Dzindolet, 1993).

Additionally, the various perspectives from highly diverse functional backgrounds might lead to an information overload on the team members (Van der Vegt & Bunderson, 2005). The diverse knowledge pool further increases the complexity of team problem solving (Ancona & Caldwell, 1992; Milliken & Martins, 1996), hindering explorative behavior.

(19)

Besides the arguments for a negative impact of diversity in knowledge and expertise, the major reason for collaboration in teams is still the additional access to expertise and knowledge resources of other team members. This is assumed to offset the negative effects:

Hypothesis 1b (H1b): The variable Diversity in knowledge and expertise is positively related to exploratory team behavior.

2.2.3 Diversity in cognitive styles

Cognitive diversity refers to differences in terms of how team members access problems (Paulus & Nijstad, 2003). A cognitive style is defined as “a consistent approach to organizing and pro-cessing information” (Messick, 1984: 1; in Riding & Sadler-Smith, 1997). Ziebro and Northcraft (2009) refer to the relationship of diversity in cognitive styles and exploration as the “Paradox of Creativity”: Team members are primarily attracted to similar group members and thus are less likely to exchange information with dissimilar ones. However, the probability of producing highly novel outputs is more likely when different ways of approaching problems are combined. By examining diversity in cognitive styles in this thesis, its main effect is tried to be revealed in order to resolve the paradox. In spite of the opposing proposed effects of cognitive diversity, the proposed main effect is exploration suppressing.

The Similarity and attraction paradigm implies production losses for work processes in cognitive diverse teams (Van Knippenberg & Schippers, 2007). The Similarity/attraction per-spective suggests that group members are naturally unlikely to share information and interact with dissimilar people (Byrne & Nelson, 1965; Byrne, 1971; Festinger, 1954). The attentional focus of individuals varies with the characteristics of the group member. Team members are more interested in individuals that are similar to them and pay less attention to those that are different (Byrne, 1971). This intergroup bias of favoring the opinion of a few members may have an isolating effect on some group members with a different approach to problems, which

(20)

overall might reduce the additive effect of cognitive diversity (Van Knippenberg & Schippers, 2007).

Furthermore, Diversity in cognitive styles might result in misunderstandings among team members (Paulus & Nijstad, 2003). Different cognitive styles result in different thought worlds of the team members with different perceptions (Dougherty, 1992; Boland & Tenkasi, 1995). This might create communication barriers between the individuals, which may make it difficult for team members to interpret the input of others (Bechky, 2003; Cramton, 2001). The differ-ent cognitive styles may hinder team members in following the argumdiffer-entation of each other as their assessment of problems differ. This confusion complicates the information sharing pro-cess. Overall it decreases both team satisfaction and the members’ impression of their own creative performance (Kurtzberg, 2005). Studies indicate that heterogeneity in cognitive styles has a negative effect on social integration and communication, leading to less engagement in explorative behavior by the team members (Mannix & Neale, 2005).

Although, few studies suggest that cognitive diversity might be beneficial for the objective functioning of a group as a greater number of ideas gets generated (Kurtzberg, 2005; Shin, Kim, Lee & Bian, 2012). The exposure to different perspectives decreases the probability of infor-mation stagnation within a group (Ziebro & Northcraft, 2009; Hoever, Van Knippenberg, Van Ginkel & Barkema, 2012). Especially unfamiliar situations require team members to reconsider their own assumptions and how these relate to the assumptions of others, to question existing differences and dependencies (Dougherty & Tolboom, 2008; Skilton & Dooley, 2010; Majchrzak, More & Faraj, 2012). The resulting increase in knowledge fosters the numbers of possible novel information recombinations, enhancing explorative behavior and rising the probability of more innovative outcomes (Barron & Harrington, 1981). The overall effect is an increase of both quantity and quality of generated ideas.

It is suggested that the communication barriers, which are created by diverse cognitive styles, hinder the increase in novel information recombinations. More, as the arguments for

(21)

a negative relation between Diversity in cognitive styles and Exploration are more sound and outnumber the argument in favor of an exploration-enhancing effect, it is hypothesized:

Hypothesis 1c (H1c): The variable Diversity in team members’ cognitive styles is negatively related to exploratory team behavior.

2.3 Moderating effects

Scholars stress that studies testing the relationship of diversity and team outcomes dominantly focused on “main effects”, while leaving possible moderating variables aside (Van Knippen-berg & Schippers, 2007; Jackson & Joshi, 2004). As the main effect propositions could not be completely empirically supported in the literature on team diversity, it is important to take moderators into consideration when explaining the diversity-exploration relationship (Van Knippenberg & Schippers, 2007; George & Chattopadhyay, 2008; Van Dijk, Van Engen & Van Knippenberg, 2012).

Teams operating in dynamic environments need to be able to access and use each member’s resources to reach the team’s full potential (Gardner, Gino & Staats, 2012). The likelihood for teams to perform well is dependent on the degree to which the group can access and make use of the resources required for the team performance (Hackman & Katz, 2010). The knowledge utilization in teams is therefore critical for enabling the assessment and processing of intra-group information (Gardner, Gino & Staats, 2012). Teams need to transform their individual specialized knowledge into an integrative knowledge that complements the knowledge of other members, in a “process of mutual influence and collaborative emergence” (Majchrzak, More & Faraj, 2012: 951) (Majchrzak, More & Faraj, 2012; Hargadon & Bechky, 2006).

Consequently, the moderators Transactive Memory System and Team identity are included, as they stimulate the process of knowledge utilization. The existence of a Transactive Memory System is supposed to moderate the translation of diversity effects on work processes through

(22)

the mechanism of communication (Van der Vegt & Bunderson, 2005). Further, the moderating variable Team identity is integrated as a motivational force enabling the interaction between team members (Van der Vegt & Bunderson, 2005). Both variables allow better information processing within a team by an increase in communication and interaction, as the willingness of group members to attend to and be influenced by information from others in the group is increased (George & Chattopadhyay, 2008).

2.3.1 Transactive Memory System (TMS)

The concept of Transactive Memory originates in psychological research and is defined as “the cooperative division of labor for learning, remembering and communicating team knowledge” (Lewis, 2003; Hollingshead, 2001; Wegner, 1986). As a system, the TMS represents a “set of individual memory systems in combination with the communication that takes place be-tween individuals” (Wegner, Giuliano & Hertel, 1985, in Wegner, 1986). A TMS consists of meta-memories of each group member, the location of knowledge within the team and the com-munication processes which connect the single members. Team members, which are supported by a well-developed TMS, are able to access further knowledge existent within the team through a better knowledge integration (Gino, Todorova, Miron-Spektor & Argote, 2009; Wegner, 1986). Explorative team behavior is considered to be moderated by a TMS as it affects the way team member communicate and share information, which influences the performance of the group (Ren & Argote, 2011). A TMS has rarely been tested in regard to its effect on explorative behavior. Still it is relevant to include a TMS as a possible moderator as team members are supposed to be enable to exchange ideas smoothly and recombine knowledge elements, leading to an increase in explorative behavior. Therefore the existence of a TMS can be a major reason of why teams are able to benefit from their diversity with different degrees of effectiveness (Gino, Argote, Miron- Spektor & Todorova, 2010).

(23)

Moderating effect of TMS

If information is not sufficiently spread within a group, information remains underutilized, resulting in a process loss (Miner, 1984). Access to a larger knowledge pool unfolds creative potential within a team (Ziebro & Northcraft, 2009). The core logic of a TMS is that team members rely on and complement each other as sources for knowledge in areas of learning and expertise (Heavey & Simsek, 2014). Teams with a well-functioning TMS are able to effectively share and recombine the single knowledge elements, which results in explorative team output (Gino, Todorova, Miron-Spektor & Argote, 2009). Teams with highly developed communication systems are supposed to gain the greatest benefits from functional diversity (Jackson, Joshi & Erhardt, 2003; Simons, Pelled & Smith, 1999; Tjosvold, Hui, Ding & Hu, 2003). A TMS is predicted to promote the benefits of functional diversity and improve the ability of a team to create innovative products through advanced communication processes. Exploratory behavior within the group increases as the combination of individual contributions to a collective product is fostered (Gino, Todorova, Miron-Spektor & Argote, 2009). Studies confirm that teams with highly developed TMS have a higher performance level than teams without one (Gino, Todorova, Miron- Spektor & Argote, 2009; Lewis, 2003; Austin, 2003).

Next to translating more information into group processes, a TMS enables the creation of novel combinations between single knowledge elements by facilitating communication processes (Gino, Todorova, Miron-Spektor & Argote, 2009). Different to behavioral integration, in which team members engage in similar behaviors, the transactive memory emerges through processes of compilation, in which team members engage in producing unique knowledge in nonlinear ways (Heavey & Simsek, 2014). The detailed knowledge possessed by single team members is accessable for the whole group and the range of possible information combinations is expanded (Gino, Todorova, Miron-Spektor & Argote, 2009; Kozlowski & Ilgen, 2006). Consequently, a TMS facilitates unique and nonoverlapping knowledge rather than just shared knowledge

(24)

(Heavey & Simsek, 2014; Zajac, Gregory, Bedwell, Kramer & Salas, 2014). This gives rise to novel cognitive structures, which differ from the ones of the individual members (Heavey & Simsek, 2014). Through the increase in perspectives and knowledge variety, new concepts can be developed by rearranging existing knowledge elements (Mom, Van Neerijnen, Reinmoeller & Verwaal, 2015; Katila & Ahuja, 2002; Tsoukas, 2009). Consequently, a well-functioning TMS is considered to moderate innovative team outcomes.

Furthermore, team members need to be able to implement their prior experience within the common group processes, which is also suppose to be moderated by a TMS (Gino, Todorova, Miron-Spektor & Argote, 2009). Prior task experience is an important source of innovative team outcomes as it improves the task-relevant capabilities of the team members (Conti, Coon & Amabile, 1996; Gino, Todorova, Miron-Spektor & Argote, 2009). The TMS moderates this knowledge transfer from one task to a related one and therefore enables the increase in creation of novel solutions (Lewis, Lange & Gillis, 2005). As knowledge gets widely distributed in the team, knowledge exchange and coordination is effectively enabled. Thus, the need for extensive exchanges is reduced and team members can focus their attentions on the members, who hold the task relevant knowledge (Heavey & Simsek, 2014). Concluding it is hypothesized:

Hypothesis 2 (H2): The positive relationship between Exploration and Diversity in knowledge and expertise is moderated by a TMS, so that this relationship is stronger with the existence of a TMS.

2.3.2 Team identity

Team identity is defined as the “knowledge of membership in a social group (or groups) together with the value and emotional significance attached to that membership” (Tajfel, 1978: 63, in Van der Vegt & Bunderson, 2005). Though a strong identification is more difficult to achieve in diverse teams (Van der Vegt & Bunderson, 2005), the moderating effect of a collective team

(25)

identity is analyzed. Team identity is supposed to have a different moderating effect on Diversity in knowledge and expertise than it has on Diversity in cognitive styles.

Moderating effect on Diversity in knowledge and expertise

Regardless of the opposing opinions on the moderating effect of a strong team identity, the restraining effect on the positive relation between Diversity in knowledge and expertise and explorative group behavior prevails.

Besides the proposed overall negative effect, some scholars argue in favor of a positive mod-erating effect. They stress that the trusting, knowledge-sharing environment is supported by a strong team identity (Edelman, Bresnen, Newell, Scarbrough & Swan, 2004; Hansen, 1999). Research on social networks indicates that stronger relationships between team members create mutual trust (Hansen, 1999). Trust is an expectation about the future behavior of others and the extent to which individuals are confident in the reliability of future actions of their peers (Maurer, 2010; Mom, Van Neerijnen, Reinmoeller & Verwaal, 2015). As trust rests on predic-tions of team members’ good intent and behavior, it needs time to develop and is built on prior experience (Maurer, 2010). A strong team identity can facilitate the confidence into the future behavior of other team members. Farther, a high level of trust increases the commitment of individual members to share knowledge and to help each other in understanding new external knowledge (Hansen, 1999; McEvily, Perrone & Zaheer, 2003; Lane, Salk & Lyles, 2001). Conse-quently, a strong team identity increases team members’ efforts to collaborate behavior which enables an effective exploitation of newly acquired knowledge (Van Wijk, Jansen & Lyles, 2008; Hansen, 1999).

Furthermore, a strong team identity enhances the alignment of individual goals and the commitment to a common team goal (Mom, Van Neerijnen, Reinmoeller & Verwaal, 2015; Moran, 2005). As the level of cohesiveness increases by a strong team identity, the agreement

(26)

on the common goals rises (Mom, Van Neerijnen, Reinmoeller & Verwaal, 2015; Adler & Kwon, 2002; Moran, 2005). Hence, the stronger the identification with the team and the closer the collaboration between the team members, the higher the level of agreement leading to a better implementation of a common explorative outcome.

However, a strong team identity can also lead to competition between the individual group members and in-group favoritism, which has a decreasing effect on team outcome (Fein & Spencer, 1997). Arguing from the social capital perspective, individuals might act defensively if they feel that they did not receive credit for their involvement in the team process (Edelman, Bresnen, Newell, Scarbrough & Swan, 2004). This affects the information processing ability of the team as affected people might avoid participating in group processes and hold back information. As internal arguing and disincentives to knowledge sharing increase, the benefits of Diversity in knowledge and expertise get reduced.

Further, a strong team identity enhances peer pressure (Towry, 2003). The stronger the relationships between the individual team members, the more intense the commitment to the common goal (Mom, Van Neerijnen, Reinmoeller & Verwaal, 2015; Moran, 2005). Consequently, team members stop investing into learning outside the scope of the common team goal as they are unwilling to deviate from the prevalent group opinion (Mom, Van Neerijnen, Reinmoeller & Verwaal, 2015; McGrath, 2001). Engaging into explorative behavior, which deviates from the common goal is therefore seen as an unnecessary waste of resources (Mom, Van Neerijnen, Reinmoeller & Verwaal, 2015; Edelman, Bresnen, Newell, Scarbrough & Swan, 2004). This results in a pressure towards uniformity (Festinger, 1954; Towry, 2003). These “unwritten norms of conformity, control and compliance” (Edelman, Bresnen, Newell, Scarbrough & Swan, 2004: 66) hinder teams in translating their diversity into innovative, explorative activities. Hence, a strong team identity reduces diversity characteristics in teams, as uniform behavior is enhanced. Further it hinders teams in making use of the advantages of diversity and engaging in

(27)

exploration by an increase of in-group competition and peer pressure. Thus, it is hypothesized:

Hypothesis 3a (H3a): The positive relationship between Diversity in knowledge and expertise and Exploration is moderated by a strong team identity, so that this relationship is weaker for higher values of team identity.

Moderating effect on Diversity in cognitive styles

A strong team identity is claimed to decrease the negative relationship of Diversity in cognitive styles and exploratory team behavior.One opposing argument to this claim is that a high team identity leads team members to act in a group-typical way. This “pressure toward uniformity” (Festinger, 1954: 130) within the group has a suppressing effect on explorative behavior. It is particularly harmful if work groups develop counter-productive norms, for example as a reaction to a conflict with the management, (Van Knippenberg, 2000).

Moreover, individuals sharing a common sense of team identity attribute group character-istics to themselves and therefore accept divergent cognitive styles amongst each other (Van Knippenberg, 2000). This strengthens the commonalities between the diverse members instead of the characteristics that differentiate them from another. The in-group identity is strength-ened, fostering a positive attitude towards team members and reducing stereotypes, which in turn motivates individuals to engage in team work (Gaertner, Dovidio & Bachmann, 1996). Re-sulting, the perceptions of social categories proposed by the Theory of Social categorization are overall increase in knowledge sharing evidently fosters individual problem solving capabilities and team exploration (Edelman, Bresnen, Newell, Scarbrough & Swan, 2004).

Furthermore, individuals, who highly identify with their team, develop a feeling of psy-chological ownership of the common team goal (Pierce, Rubenfeld & Morgan, 1991). These individuals are more committed to the team and the achievement of the common goal than to their own goals (Van der Vegt & Bunderson, 2005). Stereotypes and the refusal of diverse

(28)

cognitive styles among the different team members are reduced (Ashforth & Mael, 1989; Sherif & Sherif, 1969; in Gino, Todorova, Miron-Spektor & Argote, 2009). This enhances the individ-uals’ cooperative behavior with their team members, fostering the translation of diversity into explorative behavior. This argumentation leads to:

Hypothesis 3b (H3b): The negative relationship between Diversity in cognitive styles and Exploration is moderated by a strong team identity, so that this relationship is weaker for higher values of team identity.

(29)

Chapter 3

Method

This chapter contains the empirical part of this thesis. First, the characteristics of the research setting and samples will be outlined. Subsequently, the variable measurements and correspond-ing reliabilities are examined. Finally, the statistical approach taken to test the hypothesized relations is described. The items of the questionnaire used in this research design can be found in Appendix A (in Dutch).

3.1 Research setting

An experimental longitudinal simulation, called the “Business Strategy Game” (BSG) is used to collect the data for this study. Parallel to this 9-week simulation game, a survey has been carried out, which tracked interpersonal key determinants of team performance at five different points in time. The first wave was surveyed during the first week of the BSG, followed by wave 2 to 5 in each consecutive week. Response rates for the survey were ranging from 69% to 92.7%. 566 Business School students participated in the survey. The students were divided into 117 teams of five of which 35% were randomly assigned, while the remaining teams formed themselves. In total, 81 participants were removed because they could not be allocated to a team and 21 teams because less than 4 team members filled out the questionnaire. The final dataset therefore contains 480 participants grouped in 96 teams. Ranging from the age of 18 to a maximum of 27 years, the average age of the students was 18.5 years. The standard deviation

(30)

of the participants’ age is .98 years. The majority were male participants (68%). All students granted permission for the usage of the survey data for research purposes.

Though the realism of the business game already serves as a motivation for participation (Clark & Montgomery, 1996), participants were further motivated by using their firm’s perfor-mance as a part of their course grading. This highly motivated the participants to engage in the business simulation game and to perform well, likewise to manager in reality (Chen, Katila, McDonald & Eisenhardt, 2010).

The survey is matching the purposes of this research proposal as the data is raised in a realistic setting. The research setting has many advantages. First, it offers complete and transparent information about the actions of all participants. Instead of recording only certain moves, all actions by every team are captured. This enables researchers to gather unique data sets and to measure variables that would otherwise be impossible, too difficult or too costly to obtain in real settings (e.g. discrete R&D investments) (Chen, Katila, McDonald & Eisenhardt, 2010). Secondly, the experimental design also controls for confounding factors. This enables researcher to isolate the phenomena of interest easily (Chen, Katila, McDonald & Eisenhardt, 2010). Additionally the research setting provides a realistic view on competition as each simulation generates different outcomes resulting from interactions of the competing teams (Chen, Katila, McDonald & Eisenhardt, 2010). Finally, the elicitation of the data was conducted over a 9-week time period in consecutive years. The longitudinal nature of the experimental simulation allows the evolution of firms and industries, leading to a better understanding of causalities to be observed (Chen, Katila, McDonald & Eisenhardt, 2010).

Though caution might be appropriate regarding the generalizability of the findings as the survey was conducted among students only. Prior research states that there are no empirical differences in behavior of student teams and executive teams (Clark and Montgomery, 1996; Chen, Katila, McDonald & Eisenhardt, 2010). This suggests that concerns regarding the gener-alizability of the study results as an indication for executive behavior and industry competition

(31)

can be mitigated (Chen, Katila, McDonald & Eisenhardt, 2010). As the participants’ average age of 18.5 years is relatively low, the research results especially affect teams in technology-based ventures or start-ups, in which the workforce is particularly young (Chen, Katila, McDonald & Eisenhardt, 2010). A further limitation is the limited focus exclusively on key aspects of phenomena, which eliminates some complexity (Davis, Eisenhardt & Bingham, 2007). In the simulation game firms for example can not form alliances, make acquisitions, enter new markets or do not have to make long run expenses due to the finitude of the simulation game (Chen, Katila, McDonald & Eisenhardt, 2010).

3.2 Sample and data sources

3.2.1 The Business Strategy Game

The simulation game is placed in the global athletic footwear industry, in which the teams have to compete directly with each other. All teams start under identical conditions with a company that performs well in both revenues and profits. The purpose of the game is to react to the changing organizational environment with an appropriate strategy. The teams can select between four different geographical markets (North-America, Latin America, Europe-Africa, Asia-Pacific) and three different sales channels to market and sell their products (internet, private-label, wholesale). They can further decide upon different factors regarding e.g. op-erations, human resources, marketing, sales, or corporate finance. Following each of the nine rounds, the teams are provided with a feedback report on their internal and external environ-ment. The performance outcomes are generated and are presented by five equally weighted components (earning per share, return on equity, stock price, credit rating and image rating).

The participants form the boundary conditions of the game themselves as their collective decisions determine the level of the competitive intensity. Consequently, the teams face the same circumstances in industry settings. Performance differences between the teams are therefore a

(32)

reflection of their ability to judge on changes in the internal and external environment. This makes it easy to isolate and measure the effects of different team constructs on their performance.

3.2.2 The GIO Dataset

In addition to the BSG dataset, the GIO survey was used, which is conducted with the same co-hort of students. The two datasets are combined to collect further information about personality characteristics. Both datasets are used to examine the proposed hypotheses.

3.3 Measures

Exploration

To measure explorative team behavior, the scale by Mom, Van den Bosch and Volberda (2009) is used (3 items, Cronbach’s α = 0.802). Two of the original five items are deleted on the basis of the Confirmatory Factor Analysis (see section 3.4.). The measure was conducted in wave 5 of the survey. An example item is: “To what extent did you engage in game-related activities that can be characterized as activities requiring you to learn new skills or knowledge?”. A 7 point Likert-scale ranging from 1 (strongly disagree) to 7 (strongly agree) is used.

Diversity in social categories

The mean of the standard deviations of the items Age and Gender are used to assess Diversity in social categories. The indicators of the variable do not necessarily have to covary with each other as Diversity in social categories is a formative construct. Therefore consistency reliability is not been considered when evaluating the adequacy of the measure (Jarvis, MacKenzie & Podsakoff, 2003).

(33)

Diversity in knowledge and expertise

To assess the formative construct of Diversity in knowledge and expertise, the mean of the stan-dard deviations of the Average grade for high school graduation and of the stanstan-dard deviations of the Achieved ECTS are used.

Diversity in cognitive styles

To measure Diversity in cognitive styles, the standard deviations of Need for cognitive closure (17 items, Cronbach’s α = 0.949) and the standard deviations of Transition process (6 items, Cronbach’s α = 0.804) are included (Total Cronbach’s α = 0.913). Both basic dimensions of cognitive style, the Wholist-Analytic Style and the Verbal-Imagery Style are considered (Riding, 1997). The variable Need for cognitive closure represents the organization of information by the team members and the variable Transition process the representation of information in the team. The variable Need for cognitive closure is integrated in the GIO dataset, while the variable Transition process is conducted in the third wave of the BSG dataset. A 5 point Likert-scale, ranging from 1 (strongly disagree) to 5 (strongly agree), is used for measuring Need for cognitive closure. Nine of the 18 items of the measure Need for cognitive closure were counter-indicative items, meaning that low scores indicate high levels of need for cognitive closure. These items are reverse coded. For the Transition process a 7 point Likert-scale, ranging from 1 (strongly disagree) to 7 (strongly agree), is used. One item of each variable is removed on the basis of the CFA. In order to compute the variable Diversity in cognitive styles the items of Need for cognitive closure are transformed to a 7 point Likert-scale (by applying X2 = 1.5 * X1 - 0.5).

Transactive Memory System (TMS)

To assess the potential moderating influence of a Transactive Memory System on the relation-ship of Diversity in knowledge and expertise and Exploration, the measure scale of Lewis (2003) is used. It consists of the three factors; specialization, credibility and coordination (10 items,

(34)

Cronbach’s α = 0.84). One of the original 15 items has been reverse coded. The items were conducted in wave 3 of the BSG dataset. In total five items from different factors are removed on the basis of the CFA, because their factor loading did not exceed 0.6. An example question for the factor specialization is: “Each team member has specialized knowledge of some aspects of our project.” An example question for the factor credibility is: “I trusted that other members’ knowledge about the project was credible.” An example question for the third factor coordina-tion is: “Our team worked together in a well-coordinated fashion”. A 7 point Likert-scale is used, ranging from 1 (strongly disagree) to 7 (strongly agree).

Team identity

To assess the potential moderating influence of Team identity on the relation of all diversity variables and Exploration, four items are considered (Cronbach’s α = 0.908). One item of the original eight items has been reverse coded. Four items are removed because of a factor loading of < 0.6 in the CFA. An example item is: “I strongly identify with the other members of my team”. The items of wave 3 were used. A 7 point Likert-scale was used, ranging from 1 (strongly disagree) to 7 (strongly agree).

Control variables

The results of the current study are controlled for three control variables. According to Mannix & Neale (2005), communication and information exchange are underlying mechanism influencing the effect of team diversity. Therefore, Communication frequency (wave 3, 7 point Likert-scale) is chosen as a control variable. Next to the industry belongingness, the average team grade is also included into the control variables. The former control variables Openness to experience, Communication and information sharing and Intragroup trust are removed because their overlap with other variables has been too high.

(35)

Definition Key Reference Measure Dependent Variable

Explorative behavior

Development and adaption of new operational and administrative routines

Crossan, Lane & White, 1999

3 items, wave 5, 7 point Likert-scale, Cronbach’s α = 0.802 Independent Variable Diversity in social categories Inter-group members’ differences regarding surface-level variables such as age, sex and race.

Mannix & Neale, 2005

Formative construct: Age and Gender Diversity in knowledge and expertise Diverse educational, functional, occupational backgrounds or range of industry experience.

Milliken & Martins, 1996

Formative construct: Average grade and ECTS points

Diversity in cognitive styles

Differences in terms of how team members approach organising and processing information.

Paulus & Nijstad, 2003; Messick, 1984

23 items, wave 3, Cronbach’s α = 0.913 Moderating Variable

TMS

Cooperative division of labor for learning, remembering and communication of team knowledge.

Lewis, 2003

10 items, wave 3, 7 point Likert-scale, Cronbach’s α = 0.84

Team identity

Knowledge of membership in a social group (or groups) together with the value and emotional significance attached to that membership.

Tajfel, 1978

4 items, wave 3, 7 point Likert-scale, Cronbach’s α = 0.908

Table 1: Measurement of variables

3.4 Confirmatory Factor Analysis

A Confirmatory Factor Analysis (CFA) is performed to confirm the fit of the factor structure. To perform the CFA, the program IBM SPSS Amos 21 is used. The hypothesized measurement model includes a second-order TMS factor, indicated by three first-order factors (specialization, credibility, coordination), each indicated by five items. Diversity in cognitive styles are the second second-order factor, indicated by the two first-order factors Need for cognitive closure (18 items) and Transition process (7 items). The residual factors are Exploration (5 items), Intragroup trust (4 items), Team identity (8 items), Openness to experience (10 items) and Communication and Information sharing (3 items)(see Appendix B.1).

The model fit is evaluated by using the comparative fit index (CFI), the standardized root mean square residual (SRMR) and the root mean error of approximation (RMSEA). The

(36)

chi-square test is included to test the difference between observed and expected covariance matrices. Further indices, such as goodness-of-fit index, adjusted goodness-of-fit index, normed fit index and the Akaike Information Criterion behave erratically or not robust in smaller samples (Hu & Bentler, 1995, in Lewis, 2003), and thus are less important in this study.

To improve the overall model fit, items with standardized regression weights of < 0.60 are deleted. Consequently, the variable Openness to Experience is excluded completely from the model (see Appendix B.2).

The χ2 associated with the model is significant, χ2(718, N = 102) = 1012.22, p = 0.00, which suggests that the model is not consistent with the observed data. In complex models with many variables and degrees of freedom, the χ2 _{test is often statistically significant despite}

a reasonably good fit to the data. Furthermore, the χ2 test is strongly affected by sample size (Schumacker & Lomax, 1996). Therefore, Schumacker and Lomax (1996) recommend to use the normed χ2 test to measure the model fit, as it is less sensitive to sample size resulting in χ2/df = 1.41 (Criteria for acceptance range from < 2 (Ullman, 2001) to < 5 (Schumacker & Lomax, 2004)). According to the normed χ2 test the model therefore is consistent with the observed data. The CFA of the measurement model shows an acceptable fit, with CFI = 0.882 (Cutoff close to 0.9 for an incremental fit; Hu & Bentler, 1999), SRMR = 0.085 (Cutoff value close of 0.08, Hu & Bentler, 1999) and RMSEA = 0.066 (Cutoff value close to 0.06, Hu & Bentler, 1999) The model shows some issues regarding Reliability (Composite Reliability (CR) > 0.7; Hair, Black, Babin & Anderson, 2010), Convergent Validity (Average Variance Extracted (AVE)> 0.5; Hair, Black, Babin & Anderson, 2010) and Discriminant Validity (Square root of AVE greater than inter-construct correlations; Hair, Black, Babin & Anderson, 2010)(see Table 2). As the BSG dataset contains only empirically proven constructs and the Cronbach’s Alphas all exceed the critical threshold of 0.7, it is proceeded with the analysis.

(37)

CR AVE TI E DCS TMS Team identity (TI) 0.896 0.685 0.828

Exploration (E) 0.771 0.542 -0.125 0.736 Diversity in

cognitive styles (DCS) 0.04 0.143 -0.908 -0.393 0.378 Transactive Memory

System (TMS) 0.681 0.509 0.627 0.041 -0.518 0.714

Table 2: Confirmatory Factor Analysis Model Validation

3.5 Statistical methods

To perform the statistical analyses, the Statistical software Package for Social Sciences (SPSS) is used. After checking frequencies, rows in which interviewees did not fill in their team number and rows in which less than 4 team members filled out the survey, are deleted. Missing values are replaced according to the mean. In the next step counter-indicative items are reverse coded into different variables.

Further, descriptive statistics, skewness, kurtosis, check for multicollinearity and normality tests were computed. All variables contain non-normally distributed items, with either minor skewness or minor kurtosis issues. Though the values lie in an acceptable range. Scholars argue that with reasonably large samples, skewness does not make a substantive difference in the analysis outcome (Tabachnick & Fidell, 2001). Furthermore, in research settings with more than 200 participants the risk of a kurtosis resulting in an underestimate of the variance is reduced (Tabachnick & Fidell, 2001). As this research contains 480 individual respondents, the analysis is conducted. All variables are tested for reliability, except for the formative constructs. Every Cronbach’s Alpha exceeds the critical threshold of 0.7.

All variables are conceptualized as group level variables. To examine the consistency in rat-ings among group members reporting on the same change, the intraclass correlation coefficient ICC(1) is computed. It is used to address whether judges rank order targets in a relatively consistent manner with other judges (Lebreton & Senter, 2007). The value provides informa-tion about the stability of mean ratings for a group of k raters, indicating “the extent to which individual ratings are attributable to group membership” (LeBreton & Senter, 2007: 834)

(38)

(Le-Breton, Burgess, Kaiser, Atchley & James, 2003). The one-way intraclass correlation coefficient ICC(1) for the variables Exploration (ICC(1) = 0.734, Cronbach’s α = 0.742, p = 0.00), Team identity (ICC(1) = 0.755, Cronbach’s α = 0.846, p = 0.00) and Diversity in cognitive styles (ICC(1) = 0.913, Cronbach’s α = 0.913, p = 0.00) all exceed the traditional cutoff point of 0.70 (Lebreton & Senter, 2007). The ICC(1) for the variable TMS (ICC(1) = 0.690, Cron-bach’s α = 0.776, p = 0.00) is just slightly below the recommended cutoff point. Following, the data is aggregated to the team level. The most conservative approach of taking the average responses from an independent set of individuals at the same group-level, is chosen (Herold, Fedor, Caldwell & Liu, 2008).

For detecting multicollinearity, the Variance Inflation Factors (V IF ) is computed, with exploration as the dependent factor. The V IF indicate whether a predictor has a strong linear relationship with other predictor(s) (Field, 2013). The V IF of Diversity in social categories (V IF = 1.59, Tolerance = 0.63), Diversity in knowledge and expertise (V IF = 1.62, Tolerance = 0.62), Diversity in cognitive style (V IF = 2.20, Tolerance = 0.45), TMS (V IF = 1.16, Tolerance = 0.86) and Team identity (V IF = 1.25, Tolerance = 0.80) are all lower than the critical threshold of 4 (O’Brien, 2007), and the largest VIF is not greater than 10 (Bowerman & O’Connell, 1990). Further, the tolerances are all above 0.2 (Menard, 1995) indicating that there are no problems regarding multicollinearity.

Scale means have been created for the variables Exploration, Diversity in social categories, Diversity in knowledge and expertise, Diversity in cognitive style, Transactive Memory System, Team identity and the three control variables (see Table 3).

(39)

V ariables M SD 1 2 3 4 5 6 7 8 9 1. Exploration 4.55 0.49 (0.80) 2. Div ersit y in so cial categories 0.49 0.32 -0.04 ( -) 3. Div ersit y in kno wledge and exp ertise 2.74 1.37 0.19 0.34** ( -) 4. Div ersit y in c ogn itiv e st yles 0.97 0.31 0.02 0.59** 0.60** (0.91) 5. T ransactiv e Memory System 4.91 0.45 0.08 0.14 0.12 0.06 (0.84) 6. T eam id e n tit y 5.20 0.64 -0.08 0.28** 0.08 0.25* 0.34** (0.91) 7. Com m unication frequency 3.85 0.52 0.03 0.09 0.03 0.10 0.16 0.42** ( -) 8. Industry b elongingness 5.60 3.10 -0.001 -0.11 -0.03 -0.31** -0.06 -0.42** -0.18 ( -) 9. Av erage team grade 6.93 0.20 0.06 -0.02 -0.11 -0.06 0.21* -0.07 0.17 0.08 ( -) Note: N = 96 . Reliabilities are rep orted along the diagonal * Correlation is significan t at the 5% lev el (2-tailed) ** Correlation is significan t at the 1% lev el (2-tailed) T able 3: Means, Standard Deviations, Correlations and Reliabilities

The catalysts of exploration : the relation of team diversity to explorative team behavior

Master’s Thesis