Perceived diversity, conflicts and team performance : the mediating role of intra-group conflict in entrepreneurial teams

Perceived diversity, conflicts and team performance

Perceived diversity, conflicts and team performance

Student: Bastiaan van Leeuwen

Thesis: Joint MSc Entrepreneurship UvA / VU Student number UvA: 10909540

Student number VU: 2504484 Thesis supervisor: Dhr. Dr. J. Sol Date & place: 01-07-2017, Amsterdam

Index

6.1 Implications and suggestions for further research...41  

6.2 Limitations...42  

References... 44  

Appendices index... 48  

Appendices... 49    

Abstract

This study examined the relationship between perceived diversity and team performance, mediated by four different types of conflict including: relationship, task, process and status conflict. The sample of this research, gathered with questionnaires, includes 25 members of project teams, and 25 members of entrepreneurial teams. Although that only validated questionnaire items were used, the principal component analysis indicated different groupings of items than the original study design. The main results indicated that intra-group conflict could not be differentiated into four types. Furthermore perceived diversity consisted of multiple dimensions. A significant relation was found between general perceived diversity and intra-group conflict. However, the validity of this construct was questionable due to the inclusion of only two items. Additionally, it was found that intra-group conflict was significantly related to subjective performance in entrepreneurial teams. Due to the use of a cross-sectional study design reversed causality seems plausible in the relation between intra-group conflict and subjective performance. Intra-intra-group conflict did not meet all the criteria for a significant mediation effect in the assumed relation between dimensions of perceived diversity and subjective performance.

1.0 Introduction

Diversity in teams is a broadly discussed topic in research on conflict and performance. Currently, the focus in diversity research shifted from surface- and deep-level diversity towards perceived diversity, defined by Shemla, Meyer, Greer and Jehn (2014, p. 3) as: “the

degree to which members are aware of one another’s differences, as reflected in their internal mental representations of the unit’s composition.” Using a meta-analysis on the effects of

perceived diversity and group outcome Shemla et al. (2014) conclude that perceived diversity has both positive and negative effects on team performance. Lawrence (1997) explains that the inconsistent results in diversity research as deriving from the differences in perception of members of a team. A negative perception of team members in a team can result in a breakdown of the team (Foo, 2011). Zhou (2015) points to the importance of identifying conditions under which team diversity benefits entrepreneurial team performance. Because the relations between perceived diversity and team outcomes are unclear, it seems promising to test whether perceived diversity in entrepreneurial teams can be beneficial for team performance. Diversity in teams is associated with a wider set of knowledge and skills. However, transferring this knowledge and these skills by communication can influence team performance. An increase of diversity in teams can cause difficulties in communication between members, resulting in discussions and conflicts (Lawrence 1997).

Conflicts are unavoidable when individuals work together to accomplish a common goal. The question whether conflict is beneficial or unfavourable for team performance is broadly discussed in the current literature on conflict. However no conclusive results have been presented. De Dreu and Weingart (2003, p. 478) concluded that: “conflict may have

positive consequences under very specific circumstances, and we need to detect those circumstances in new research.” Review studies that investigate the influence of team

diversity, conflict types and performance produce mixed results (Zhou, 2015). A distinction is made in literature between the advantageous or harmful effects of different types of conflict and the positive and negative associations of team diversity on team performance.

The focus of this research is to explain the relation between perceived diversity and team performance in entrepreneurial teams, mediated by different types of conflict. Past research confirmed the success of entrepreneurial teams in comparison to individual entrepreneurs (Lecher, 2001; Harper, 2008; Foo 2011; Schjoedt & Kraus, 2009). Also Zhou (2015) pointed to the fact that in the establishment and growth of new ventures more often entrepreneurial teams are involved compared to individual entrepreneurs. This research aims to contribute to understand how perceived diversity and conflict can be beneficial for team

performance under the specific circumstances of perceived diversity in entrepreneurial teams. This study was focused on the variables representing teams’ perceived diversity and intra-group conflict, and their influence on team performance in entrepreneurial teams. This creates the specific context De Dreu and Weingart (2003) pointed out to be promising in generating empirical results.

The purpose of this research is to gain more insights in the mediating role of conflict types in the relation between teams’ perceived diversity and entrepreneurial team performance. This research contributes to the entrepreneurship and conflict literature by addressing the main question:

“To what extent is the relation between perceived diversity and subjective conflict mediated by relationship, task, process and status conflict?”

After clarifying the focus of this study, firstly the theoretical framework presents the consulted research literature to find underlying theories explaining the processes occurring in teams. Secondly, in the methodology section the methods of measurement of this research will be explained followed by the process of data analysis. Thirdly, in the result section the stated hypotheses will be tested interpreted and explained. In the final section a conclusive discussion will provide an overview of the results that will finally be discussed to highlight potential implications, limitations and to provide recommendations for further directions in the differentiated field of research on entrepreneurial teams.

2.0 Theoretical framework

This chapter reflects upon the current state of literature on diversity and conflict in relation to performance. Firstly, diversity will be discussed based on the existing literature with a focus on perceived diversity and the influence on performance of entrepreneurial teams. Secondly, different conflict types will be discussed as indicated to be important in the relation between perceived diversity and team performance. Understanding the underlying theories helps to explain why some effects occur in certain contexts and how to deal with perceived diversity and conflict in relation to performance with the focus on entrepreneurial and project teams. Entrepreneurial and project teams show some overlap in their behaviour considering group processes around diversity, conflict and performance. This comparison will be further elaborated in the methodology section of this research.

2.1 Diversity

The focus of diversity research is to find relations between team member diversity, group processes and performance (Knippenberg & Schippers, 2007). Generally defined, diversity refers to the differences between individuals on any attribute that may lead to the perception that another person is different from the self (Jackson, 1992; van Knippenberg, De Dreu & Homan, 2004; Williams & O’Reilly, 1998; Roberge & van Dick, 2010).

Williams and O’Reilly (1998) distinguished three underlying theories that are used to explain whether diversity is negative or positive for team performance. Firstly, starting with

the social categorization perspective as theorized by Turner, Hogg, Oakes, Reicher and

Wetherell (1987) hypothesizing that between people who perceive themselves as different from others, categorization between individuals in groups is expected to appear, generally resulting in harmful group outcomes (Knippenberg, De Dreu & Homan, 2004; Knippenberg & Schippers, 2007). The underlying effects of social categorization are caused by the assumption that individuals want to maintain a high level of self-esteem. In a social comparison with others, individuals define themselves thru social categorization based on outward demographic characteristics. The creation of a social identity described by Tajfel and Turner (1986) permits a person to categorize themselves and others into specific groups. Self-categorization affects the perception, attraction and interpersonal relations towards people who are not seen as part of their social identity. This creates a negative bias towards persons who are perceived as dissimilar (Williams & O’Reilly, 1998). The effects of social categorization are also applicable in heterogeneous or diverse groups, resulting in more

conflict as a result of a split between members influencing the overall performance of the group (Williams & O’Reilly, 1998; Knippenberg & Schippers, 2007).

Secondly, the similarity/attraction paradigm explained by Byrne (1971) considers that people who are similar in attitudes, values and demographic characteristics feel attracted towards each other. The assumption is made that people who share common experiences and backgrounds find it easier to interact with each other because of the experienced trust generated by similarity. The underlying theory for the similarity attraction is explained by the term homophily, the behaviour of people to feel more confident with people who are similar in actions, appearance and thoughts (Williams & O’Reilly, 1998). Both social categorization and similarity/attraction perspective forecast that groups wherein people perceive themselves as similar to other members perform better than groups where the perception of dissimilarity is prevalent (Hentschel, Shemla, Wegge & Kearny, 2013).

Thirdly, the information/decision-making perspective assumes that diversity in teams provides a wider set of knowledge, abilities and skills helpful for problem solving and creativity (Williams & O’Reilly, 1998). Although people tend to link to similar others, diversity in groups can contribute to improve performance thanks to the opportunity to solve problems from a multidimensional way by the opportunity of access to the wider set of capabilities and perspectives of the group (Knippenberg & Schippers, 2007). The

information/decision-making perspective associates group diversity with positive effects on

group outcomes (Williams & O’Reilly, 1998).

In theory, diversity refers to a broad range of dimensions that can be divided in two main categories: surface-level diversity and deep-level diversity (Harrisson, Price & Bell, 1998). Surface-level diversity includes the differences among group members of undisguised, biological characteristics that are commonly noticeable by physical appearance (Harrisson, Price & Bell, 1998). In diversity research the most used measurable observable surface-level characteristics include age, gender, and nationality. Diversity at the deep level refers to the differences among members’ attitudes, beliefs, values, personality traits, knowledge and skills. These characteristics are not directly observable and become noticable when individuals work together (Harrisson, Price & Bell, 1998).

Although the existence of demographic diversity is noticeable in teams, the effects of diversity among members become only relevant when individuals perceive this diversity and act upon this perception (Allen, Dawson, Wheatley & White, 2007). The way people react on perceptions is supported from a social constructionist point of view with the words: “If men

572). Allen et al. (2007) support the idea that perception becomes reality concerning perceived diversity.

Homan, Greer, Jehn and Koning (2010) also argue that diversity effects are dependent on how members perceive their teams’ composition. Homan et al. (2010) consider diversity beliefs as a moderator to explain the relation between objective diversity and perceived diversity. Homan et al. (2010) found that people with positive beliefs towards diversity are more likely to be open to dissimilar others, preventing a split among members by individuals. A comparison study between real-life teams and student teams by Homan et al. (2010) revealed the positive relations between diversity beliefs and perception of diversity. The results of the research of Homan et al. (2010) show the importance of perceived diversity in contrast to the actual diversity level based on demographic characteristics. Homan et al. (2010) explains that diversity beliefs shape how the teams’ objective diversity is interpreted. Psychological safety is seen as a positive mediator according to De Dreu and Weingart (2003) and Bradley et al., (2012). Perceived diversity is associated with psychological safety and therefore might create the context for potential positive outcomes on team performance.

Following Allen et al. (2007) there is no clear definition of perceived diversity because each individual team member might have a different perception of the diversity present in the team. If a person perceives the level of diversity in their team in different ways in comparison to another member, this perception is valid independently of how they would define diversity (Allen et al., 2007). To identify perceived diversity the definition as proposed by Shemla, Meyer, Greer and Jehn (2014, p. 3) was used for this study: “as the degree to

which members are aware of one another’s differences, as reflected in their internal mental representations of the unit’s composition.” Perceived diversity impacts group processes and

behaviours of individuals. Hobman, Bordia and Gallois (2003) found that perceived diversity significantly influences the level of experienced conflict by individuals in teams. The following paragraph will expose the effects of intra-group conflict on team performance.

2.2 Intra-group conflict

‘Intra-group conflict is broadly defined by De Dreu and Gelfand (2008, p. 6): “As a process

that begins when an individual or group perceives differences and oppositions between itself and another individual or group about interests and resources, beliefs, values, or practices that matter to them.” This definition is commonly used in research between the relation

Intra-group conflict is multidimensional and distinguished in research literature by several types of intra-group conflict, including relationship, task, and process conflict (Jehn, 1995; Jehn & Mannix, 2001). Bendersky and Hays (2012) introduced status conflict as an important type of conflict affecting team performance. Each type of conflict affects team performance in a different way. The following paragraphs will provide definitions of relationship, task, process and status conflict concluding with pros and cons for each type in relation to team performance.

Relationship conflict

Jehn and Mannix, 2001, p. 238) defined relationship conflict as: “An awareness of

interpersonal incompatibilities, includes affective components such as feeling tension and friction. Relationship conflict involves personal issues such as dislike among group members and feelings such as annoyance, frustration, and irritation.” In the current literature

relationship conflict is overall perceived as unfavourable for team performance (De Dreu & Weingart, 2003; Jehn, 1995; Jehn & Bendersky, 2003; Jehn, Greer, Levine & Szulanski, 2008). The negative consequences of relationship conflict on performance originate from the belief that groups with high levels of relationship conflict are hampered in the intercourse needed to create a safe environment necessary for open idea sharing, communication and positive beliefs towards objectives (Jehn, 1995). It was hypothesised by Jehn (1995) that relationship conflict impedes team performance also because relationship conflict disturbs the focus on the task and can be a drain for resources, like time and assets, needed to solve the relationship issues. Jehn and Bendersky (2003) explain that relationship conflict decreases the willingness of members to cooperate and communicate to achieve a common goal.

Task conflict

According to several researchers (Bradley et al., 2012; De Dreu and Weingart, 2007; Jehn & Bendersky, 2003; Pelled et al., 1999) task conflict may be beneficial for team performance under specific circumstances. To test the influence of conflict on team performance the following definition of task conflict is used to fit the purpose of this research. Jehn and Mannix (2001, p. 238-239) defined task conflict as:

“An awareness of differences in viewpoints and opinions pertaining to a group task. Similar to cognitive conflict, it pertains to conflict about ideas and differences of opinion about the task (Amason & Sapienza, 1997). Task conflicts may coincide with animated discussions and

personal excitement but, by definition, are void of the intense interpersonal negative emotions that are more commonly associated with relationship conflict.”

Jehn (1995, 1997) indicate that a moderate level of task conflict can be beneficial for group outcome dependently of the type of task. When complex cognitive tasks are perceived as challenging for the team, task conflict can contribute to problem solving by stimulating members to think differently and help to create a broader perspective about the task. This argument by Jehn (1995) is in line with the information/decision-making perspective as proposed by Byrne (1971). Task conflict can be beneficial for team performance under certain specific circumstances (De Dreu & Weingart, 2007). According to Jehn (1995) timing of task conflict is an important denominator to determine the influential effects. Task conflict in the middle of cooperation is seen as fruitful for performance because members can share their own perspectives about the challenge of the known, but unsolved task (Jehn, 1995). De Dreu and Weingart (2003) suggest on the basis of empirical evidence that task conflict negatively relates to team performance and allocate this difference to the overlap between relationship and task conflict. De Wit, Greer and Jehn (2012) also relate the mixed results of research on conflict and performance to the relatedness between the constructs of relationship and task conflict. De Dreu and Weingart (2003) and Jehn and Bendersky (2003) both explain that differences in the effects between relational and task conflict might derive from the existence of possible mediators.

Process conflict

Process conflict is experienced in the situation when members argue about the responsibility to complete a specific duty, defined by Jehn and Mannix (2001, p. 239) as: “An awareness of

controversies about aspects of how task accomplishment will proceed. More specifically, process conflict pertains to issues of duty and resource delegation, such as who should do what and how much responsibility different people should get.” According to Jehn and

Mannix (2001) process conflict positively influences team performance. According to Jehn and Mannix (2001) the effects of process conflict on team performance are seemingly different between high- and low-performing groups. High-performing groups experience lower levels of process conflict during the beginning but process conflict increases till the end of the group interaction in comparison to low-performing groups (Jehn & Mannix, 2001). Jehn and Mannix (2001) explain that the occurrence of process conflict at the beginning of a project helps to create guidelines for the members on how to accomplish the work. When new

tasks arise during the end stage of a project it is helpful for the members to reorient on which member would perform the objective the best as possible. Contrary according to Jehn, Northcraft and Neale (1999) process conflict inhibits people to effectively do their work because of the distraction from disagreements about work responsibilities.

Status conflict

Goncalo, Polman and Maslach (2010, p. 13) emphasize that status conflict is an important but understudied form of conflict. Bendersky and Hays (2012, p. 323) define status conflict as:

“disputes over people’s relative status (i.e., respect) positions in their group’s social hierarchy.” Following Bendersky and Hays (2012) there are indicators showing how process

conflict is related to group performance in a negative way. Bendersky and Hays (2012) explain the harmful effects of status conflict on performance by the negative impact on information sharing, more than other types of intra-group conflict. Status conflicts are seen as structural in nature and influence group performance constantly due to the high status holders in groups who try to perpetually negotiate their position in a competitive way (Bendersky & Hays, 2012). Furthermore Bendersky and Hays (2012, p. 324) add that status conflict has a moderating effect between task conflict and performance. When task conflict is at a high level and status conflict at a low-level group performance is the highest. Bendersky and Hays (2016) argue that people use diversity characteristics of members to develop differences in perception of the status of others. The results of Bendersky and Hays (2016) show that the influence of status conflict on team performance depends on preliminary status agreements. When preliminary status agreements are strongly settled, status conflict is harmful for team performance. On the contrary, when preliminary status agreement is almost absent, status conflict can be favourable for team performance (Bendersky & Hays, 2016).

Based on the literature review the following hypotheses are proposed to test for mediation of different conflict types between perceived diversity and subjective performance:

H1: Higher levels of perceived diversity, decreases the level of subjective performance. H2: Higher levels of perceived diversity, increases the level of relational conflict. H3: Higher levels of perceived diversity, increases the level of task conflict. H4: Higher levels of perceived diversity, increases the level of process conflict. H5: Higher levels of perceived diversity, increases the level of status conflict.

H6: Higher levels of relational conflict, decreases the level of subjective performance. H7: Higher levels of task conflict, decreases the level of subjective performance. H8: Higher levels of process conflict, decreases the level of subjective performance. H9: Higher levels of status conflict, decreases the level of subjective performance.

H10: The effect of perceived diversity on the level of subjective performance mediated by the level of conflict in the team.

3.0 Methodology

In order to test the hypothesis in this quantitative research, suitable methodology was applied in the form of questionnaires to collect the necessary data for analysis. The collected data will contribute to generate results and propose concluding remarks surrounding the relations between perceived diversity, intra-group conflict types and team performance. The cross-sectional nature of this research provides the advantage to measure different variables from a large group of respondents at once with little to no expense. The variables of interest central in this section are; perceived diversity (independent variable), team performance (dependent variable) intra-group conflict types (mediating variables between perceived diversity and team performance). Furthermore several control variables were included that might influence the relations between the main variables. Firstly, a description of the research setting, participants, and research procedure will be provided to enhance reproducibility and verifiability of this research. Secondly, the used variable measures will be explained and tested for reliability and validity. Subsequently regression analysis was conducted, concluding with a mediation analysis.

Research setting

For the first sample a student setting was used to provide the opportunity to minimize external influences. Horwitz and Horwitz (2009) argue that effects of team diversity on outcomes might be stronger in real organizations where stakes are higher as compared to student samples. For external validity purpose entrepreneurial teams were included in this research to test whether the results from the project team setting were also applicable to entrepreneurial teams. Conducting research in this manner increases the generalizability of the results from this study. Although the notion needs to be made that the results of this study are only valid for this sample, because of the inclusion of two specific samples of project and entrepreneurial teams.

Participants

Two samples of teams were surveyed for this research. The first sample was collected from twenty-five of the 34 students (74% response rate) who participated in the course ‘Enterprising Skills’. The 34 students were divided into 8 teams of four or five members. The supervisor of the course pre-selected the teams based on characteristics like study direction, nationality, place of residence and gender to create comparable teams. Pre-selection was based on the following characteristics. Each team included if possible at least one MSc

Entrepreneurship student, an equal mixture of nationalities, at least one student who lives in Amsterdam, and a balance in gender differences. In the sample 68% were women and 32% male, with an average age of 25.88 years (SD = 4.93, range 22 – 46 years). During the course the teams accomplished the ‘48h Entrepreneurial Team Challenge’. This challenge consisting of several objectives required innovative and entrepreneurial behaviour in order to win the challenge. The purpose of the challenge was to complete all objectives in 48 hours. Performance at the end of the challenge was measured by means of a voting system where the students and lecturer voted for each team based on their final presentation of the challenge.

The second sample consisted of a similar amount of twenty-five of the 41 contacted members (61% response rate) operating in entrepreneurial teams. The 25 respondents were from 10 different entrepreneurial teams located in the Netherlands, except one team operating in Germany. The teams were selected based on the criteria that they include three or more members and they have to be part of the directing core team. In the sample 12% were women respectively 88% male, with an average age of 26.76 years (SD = 3.57, range 22 – 36 years). The entrepreneurial teams operate in different sectors ranging from music industry, technology, financial sector, consumer product segment, employee service and food and beverage retail.

Both samples were checked for comparable means and standard deviations for each variable of interest (Appendix D). It was found that both groups responded on a similar way indicated by the comparable values of the means and standard deviations. Analysing the samples separately would be harmful for the results considering the sample sizes. Because of similarities in the responses both samples were treated in the analysis as one total sample representing entrepreneurial teams. The two groups were combined (n = 50, 18 teams), resulting in a total response rate of 67% and a division between women and men of 60% respectively 40% with a total average age of 26.32 years (SD = 4.29, range 22 – 46 years). Twelve different nationalities were included in the total sample with a majority of Dutch respondents of 62% followed by 18% German, furthermore were included in the total sample 2% Russian, 2% Greek, 2% Mexican, 2% Spanish, 2% Singaporean, 2% Australian, 2% Surinamese, 2% Cypriot, 2% Italian, 2% Belgian.

The inclusion of both project teams and entrepreneurial teams in this research seems odd at first hand, a closer look at the actions of both groups indicate many similarities in behaviour. An overview in similarities between the samples will be discussed in the section about entrepreneurial and project team performance.

Research procedure

A survey methodology is applied for this research since there already existed appropriate and pre-tested measurement tools for perceived diversity, intra-group conflict types and performance. The survey was conducted after the completion of the presentations of the ‘48H Entrepreneurial challenge’. The survey as shown in Appendix A was handed to the students in class by their supervisor to increase the response rate. Because some of the respondents was not able to complete the paper survey 11 respondents (18% response rate) were invited to complete an online survey as presented in Appendix B, two students completed the survey. The choice to approach one part of the respondents online derives from the problem that they didn’t have classes together after the final presentations of the challenge.

The second data collection round among entrepreneurial teams was completed with the help of an online survey as presented in Appendix C. The survey for the second sample was slightly adapted to fit the sample in terms of control variables and small moderations in the framing of questions due to the different context were entrepreneurial teams operate. Members of pre-selected teams were contacted by mail with the request to participate in a research questionnaire. In the invitation the purpose of the research was explained and because the survey included questions about confidential information anonymity was guaranteed to increase the chance of participation and honest answers. One week after the initial invitation a reminder was mailed to the members of interest to increase the response rate. The decision to collect data with the use of an online survey originated to overcome the problem of accessibility of entrepreneurial teams in real life due to their often-unstructured ways of operations. The quantitative data from all respondents collected by surveys will be analysed with a programme called ‘SPSS’. After the completion of data collection a data set was constructed in ‘SPSS’ wherein all variables were defined, labelled and coded. Data from the workgroup teams were first entered from the paper survey to an Excel file and pasted in SPSS. The data from the entrepreneurial teams who completed the survey in an online survey programme called Qualtrics was directly converted into an SPSS file and added to the existing data file of students.

Research strategy

The selection to include respondents was made in advance of data collection and based on several criteria. First, a survey needed to be filled in completely. Secondly a team needed to include three or more members in order to measure the group processes of interest. This last criterion is used following Moreland (2010) who differentiate dyads from teams. Moreland

(2010) considers dyads as more transitory compared to teams and can be formed and broken down faster. Furthermore dyads in relation to teams show more and different affection towards each other than teams. Moreland (2010) explains that teams show specific characteristics that result in certain processes and dynamics that are absent in dyads.

The data was examined for errors and missing values as follows: the responses were screened for non-response and out of range values on questions, which did occur. The question representing objective performance for the project teams was not used for analysis due to missing values. This was caused by the circumstance that the project teams did not all had a grade for their achievements by their supervisor at the moment the survey was conducted. Because the surveys are anonymous it was not possible to collect this data later on. Also the questions about time to complete the challenge for workgroups and months of collaborations for the entrepreneurial teams were excluded due to incomplete data. Furthermore the question on challenges completed was only applicable for the project teams and therefore not included for further analysis. Other errors or missing values did not occur in the data. Subsequently values of variables were checked to have no out-of-range values. Based on the preset range of the scales no suspicious values were found.

After controlling the entered data, psychometric analysis was conducted to check the reliability and validity of the scales. Factor analysis on all items was conducted followed by a reliability test for each scale to check for scale reliability. Factor analysis was done to make sure the variables discriminate and measure different dimensions of the constructs. To test the dimensionality of the constructs on convergent and discriminating items, confirmative factor analysis was done because the scales were pre-defined. The conducted factor analysis (Appendix F) and reliability analysis (Appendix G) based on several criteria will be further elaborated in the analysis paragraph.

Measurement of variables

With the use of validated measures two separate surveys were constructed existing of 44 questions for the students, and 43 questions for the entrepreneurial teams. The deviation between the numbers of questions was caused by the inclusion of a question related to the number of completed challenges for the project teams. This additional question was not used for further analysis as described in the previous section. Both questionnaires included the same measurements on the variables of interest with minor adaptations in the phrasing of questions and different control variables applicable to the samples. The questions in the survey represent items of different scales used to measure each variable. The following main

variables were included in this research: perceived diversity, relational conflict, task conflict, process conflict, status conflict, subjective performance and control variables. The included control variables were: team size, gender, age, teamwork experience, education and nationality. The original conceptual model did not met the criteria do conduct a proper analysis as explained in the following section on measurements of constructs.

Perceived diversity

To measure the independent variable three complementing scales were used to measure perceived diversity. Each scale measures a different dimension of perceived diversity: surface-level, deep-level and general. A 6-item scale, developed by Van Dick et al. (2008) was used to measure surface-level perceived diversity. The following questions were included to measure perceived surface-level diversity: “How diverse do you think your team is in

general?” and “How similar or different are the members of your team with respect to their: age, gender, cultural background, country of birth. A 4-item scale also constructed by Van

Dick et al. (2008) was used to measure deep-level diversity by asking respondents the following questions: “How similar or different are the members of your team with respect to

their: educational background, personal values, attitudes about work, learning goals?”

Answer options were given based on a five-point Likert-type scale ranging from “1 = Not at

all diverse” to “5 = Completely diverse”. The alpha coefficients for surface-level and

deep-level perceived diversity were .73 (M = 3.76, SD = .79) and .68 (M = 3.37, SD = .74) respectively. Finally a 3-item scale compiled by Hentschel, Shemla, Wegge and Kearney (2013) was included to measure general perceived diversity. The questions comprised of:

“When I am supposed to describe my work team, I automatically think about the differences among my colleagues”, “I am very aware of the differences among my colleagues”, “I sometimes think about the differences among the colleagues in our team”. Response options

were given based on a 5-point Likert-type scale: “1 = Strongly disagree” to “5 = Strongly

agree”. For the third scale the coefficient alpha was .69 (M = 3.54, SD = .72).

The results of the factor analysis indicated that the item PercDiv#3.0 of the scale measuring general perceived diversity loaded together with the items measuring surface-level perceived diversity. Therefore the item PercDiv#3.0 was moved from the scale of general perceived diversity to the scale measuring surface-level perceived diversity, as shown in Table 3. The item PercDiv#1.5 did not loaded on the right component and decreased the coefficient alpha of the scale measuring surface-level diversity. A similar question relating to cultural background was already included; therefore PercDiv#1.5 was not used for further

analysis.

The value of the Cronbach’s alpha below .7 (.68) indicated that the scale measuring perceived deep-level diversity was technically not so reliably following literature prescribing a Cronbach’s alpha of at least .7 (Pallant, 2010). Theoretically the measures are applicable to capture perceived diversity following Van Dick et al (2008) and Hentschel et. al (2013) who constructed this scale based on theoretical assumptions of phenomenon that can occur in team dynamics and found average internal consistency across the items above .7. The decision to increase the coefficient alpha of the scale in this research is done because an unreliable scale can’t produce reliable and valid results.

Psychometric analysis was conducted to check how scale improvement was possible to accommodate the technical criteria for reliability of the scale. The total Cronbach’s alpha of the scale needs to equal or exceed .7 to be considered acceptable. The internal consistency of the scale below .7 might be caused due to a relative small sample size. Another possible explanation might be that the items of the perceived diversity scales measure closely related dimensions of perceived diversity. Therefore the following values were taken in consideration. First was checked for negative values, an indicator of incorrectly scored items by controlling the corrected item-total correlation, which was not the case. Secondly the Cronbach’s alpha if item deleted should improve the current alpha value of .68. Thirdly the correted item-total correlation needs to be 0.5 or higher and the squared multiple correlation .3 or above. The final decision to include an item was based on theoretical considerations of importance for the construct.

The process of item deletion is shown in Appendix I. Based on the previous criteria it is chosen to delete the item PercDiv#2.0. Removing more items would slightly increase the internal consistency but fewer items make the total scale less reliable. A new variable was created representing the combined items that fulfilled the criteria with α ≥ .7. After the process of scale improvement the average internal consistency across the three items representing the scale of deep-level perceived diversity was .71 (M = 3.34, SD = .78) and considered to be reliable following Pallant (2010).

Based on the results of the factor analysis and reliability analysis, the following three scales were composed (Appendix H, I and J). First, a 7-item construct measuring perceived surface-level diversity with a coefficient alpha of .73 (M = 3.65, SD = .74). Secondly, a 3-item scale measuring perceived deep-level diversity with α = .71 (M = 3.34, SD = .78). Finally, a 2-item scale measuring perceived general diversity with a good internal consistency of α = .75 (M = 3.77, SD = .73).

Intra-group Conflict

The mediating variables in this research were intra-group conflict types. As previous noted in the theory section intra-group conflict could be divided in four main conflict types relationship, task, process and process conflict. Each conflict type is measured with the use of previous developed validated item scales. Most of the questionnaires used in conflict research find their inspiration in the Intragroup Conflict Scale (ICS), developed by Jehn (1995). Relationship and task conflict was measured using Jehn and Mannix (2001) validated multiple item scale. Three items measured relationship conflict on a 5-point Likert type scale anchored by “1 = None” and “5 = A lot”. Included questions were: “How much relationship tension is

there in your work group?” “How often do people get angry while working in your group?” “How much emotional conflict is there in your work group?” For relationship conflict the

coefficient alpha was .84 and considered to be reliable (M = 2.21, SD = .93).

Task conflict was measured using a 3-item scale with response options on a 5 point Likert-type scale ranging from “1 = None” and “5 = A lot”. The three items measuring task conflict included the following: “How much conflict of ideas is there in your work group?”

“How frequently do you have disagreements within your work group about the task of the project you are working on?” “How often do people in your work group have conflicting opinions about the project you are working on?” For task conflict the coefficient alpha was

.80 (M = 2.75, SD = .68). Process conflict was measured with the help of a 3-item scale constructed by Shah and Jehn (1993). Based on a 5-point Likert type scale participants had answer options shifting from “1 = None” till “5 = A lot”.

The items included for process conflict were: “How often are there disagreements

about who should do what in your work group?” “How much conflict is there in your group about task responsibilities?” “How often do you disagree about resource allocation in your work group?” For process conflict the coefficient alpha was .72 (M = 2.02, SD = .71). Finally

status conflict was measured with the use of the 4-item scale developed by Bendersky and Hays (2012). Answer options ranged from “1 = None” till “5 = A lot”, the included questions were: “My team members frequently took sides (i.e., formed coalitions) during

conflicts.” “My team members experienced conflicts due to members trying to assert their dominance.” “My team members competed for influence.” “My team members disagreed about the relative value of members’ contributions” For status conflict the coefficient alpha

was .82 and considered reliable (M = 1.98, SD = .79).

Although theory indicated that intra-group conflict could be divided among four different types of conflict. In this study was found with the use of factor analysis that

relational, task, process and status conflict could not be analysed as separate variables. Pearson correlation indicated multicollinearity between the types of conflict with r = .6 and above. For further analysis a new variable was constructed representing intra-group conflict including all items of the different conflict types (Appendix K). The new 13-item scale, representing total intra-group conflict has a high internal consistency with α = .93 (M = 2.22,

SD = .69).

Entrepreneurial and project team performance

Entrepreneurial teams are more influenced by external factors and face more challenges than student teams. Venture creation is a new and mostly unstructured task. There is need to cope with the turbulent external environment meanwhile managing the internal processes of the start-up process as well (Sjoedt & Kraus, 2009). Entrepreneurial teams are advantageous compared to individuals in the context of a start-up process associated with uncertainty, risk and need for achievement. According to Salas, Cooke and Rosen (2008, p. 540): “Teams are

used when errors lead to severe consequences; when the task complexity exceeds the capacity of an individual; when the task environment is ill-defined, ambiguous, and stressful; when multiple an quick decisions are needed; and when the lives of others depend on the collective insight of individual members.” Schjoedt and Kraus (2009, p. 515) presented a

comprehensive literature-driven definition of entrepreneurial teams:

“An entrepreneurial team consists of two or more persons who have an interest, both financial and otherwise, in and commitment to a venture’s future and success; whose work is interdependent in the pursuit of common goals and venture success; who are accountable to the entrepreneurial team and for the venture; who are considered to be at the executive level with executive responsibility in the early phases of the venture, including founding and prestart up; and who are seen as a social entity by themselves and by others.”

For the student population the setting of project teams is more appropriate. The description of project teams by Cohen and Bailey (1997, p. 242) is applicable for this research. “Project

teams are considered time-limited, produce one-time output and the tasks are non-repetitive considering the performance of developing new ideas” (Cohen & Bailey, 1997, p. 242).

A comparison between characteristics of entrepreneurial teams and project teams helps to understand why these concepts can be used interchangeably for research purpose on team processes and outcomes. There is variation between the student population and

entrepreneurial team population considering the durance of the teamwork and in terms of long and short-term goals. Task-related contingencies are likely to differ in short- and long-term teams. In short-term teams, greater urgency may surround goals and missions (Joshi and Roh, 2009, p. 610). In contrast with long-term teams, task requirements may be more stable, and distribution of tasks and roles may also be more clearly defined (Joshi and Roh, 2009, p. 610). Another dimension along which these teams are likely to differ is the longevity of team membership. The members of short-term teams have likely shorter tenure than the members of long-term teams (Joshi & Roh, 2009).

A comparison between samples of students and entrepreneurial teams based on their similarities shows that the two groups show comparable characteristics. However there are a few variables that are very different comparing the two groups of interest. First, the performance indicators are different between both groups. Secondly, entrepreneurial teams know each other probably already before they decide to start as a team and have a different goal to achieve in terms of performance measurement compared to students. There are also short- and long-term orientation differences between the samples. The students are probably short-term oriented and have a clear set of goals to accomplish. Entrepreneurial teams are long-term oriented in later stages of the venture creation. In early stages of start-ups mostly short-term goals need to be accomplished before long-term goals can be anchored.

Similarity can be found in the nature of the tasks both groups have to fulfil, non-routine tasks with need for creativity, problem solving and innovativeness. The goals are somehow clear to both student and entrepreneurial teams but how to get there is rather still unknown. Another similarity between the project and entrepreneurial teams is the level of task interdependence. In both teams the members have to rely on each other in order to accomplish goals.

Subjective performance measures were used for both student and entrepreneurial teams. Growth is cited as most important objective of new ventures measured in sales growth, revenue growth and employment growth (Zhou, 2015, p. 9). Also performance is measured by financial or accounting metrics. For start-ups these measures are not effective considering the low levels of cash flow and growth in early stages. Zhou (2015) indicates the irrelevance of objective performance measurements for start-ups because new ventures often lack enough sales. The focus of entrepreneurial teams is to create the venture, managing team membership, create identity and dedication (Zhou, 2015, p. 9). Self-rated team-level performance can be used according to Zhou (2015). Subjective measures for entrepreneurial team performance seem more applicable. Recommended by Zhou (2015) is to use both objective and subjective

measures for performance in research on entrepreneurial teams. Because the focus in this research concerns the team level processes, subjective measures were used because objective measures tell more about firm level performance. An often-described motivation of entrepreneurs in research literature is to create their own firm instead of making huge profits. Profit as second order interest for entrepreneurial teams, justifies the use of subjective performance as valid performance measurements (Zhou, 2015).

Allen et al. (2007) discussed the choice to measure performance of ventures subjectively or objectively in order to gain accurate results. Objective measures limit the opportunity to gain valid results of ventures operating in different segments. In contrary subjective measures allows for comparison between industries, however it can lack reproducibility limiting the scope of generalizability (Allen et al., 2007). Table 1 shows an overview of the comparison between entrepreneurial and student teams. Based on this comparison the analysis was done with both teams included. The data between student and entrepreneurial teams were comparable based on the means, standard deviations and range of values.

Table 1.

Team performance

The dependent variable performance was measured subjectively. To measure perceived performance of the teams two complementing scales were used. First the 2-item scale as proposed by Jehn (1999) was implemented in the survey with the items: “How well do you

think your team performed?” “How effective is your team?” The response options were based

on a Likert-type scale with the options varying from “1 = Poor” till “5 = Excellent”. The coefficient alpha for subjective performance measured of this scale was .77 (M = 3.8, SD = .74). The second subjective measure used derived from the 5-item scale constructed by Ancona and Caldwell (1992) compiled by the following items: “How would you rate the

performance of your team on the following dimensions: efficiency, quality of innovativeness, adherence to schedules, work excellence, team productivity?” Preset answer options on a

Likert-type scale ranged from “1 = Poor” till “5 = Excellent”. The coefficient alpha of this scale was .78 (M = 3.64, SD = .65). The two scales combined resulted in a coefficient alpha of .87 and are considered to form a reliable scale (M = 3.68, SD = .64). Principal component analysis indicated that all items of the separate scales load on the same component. For this reason the scales are not treated as separate constructs in the final analysis.

Control variables

Individual and team characteristics are included as control variables minimizing the effect of potential external variables. The selected control variables are: team size, gender, age, teamwork experience, level of education and nationality. The average team size was 4.74 members (SD = 1.64, range 3 – 8 members). Teamwork experience of individuals in the overall sample was 4.2 (SD = .95), indicating that most of the respondents had much teamwork experience. The majority of 84% participated in University College and obtained a Bachelor (42%), Master (40%) or Ph. D. (2%) degree, 14 % did start but did not finished higher education and 2% had some other form of education.

4.0 Analysis

Psychometric analysis

In advance of hypothesis testing psychometric analysis was conducted to assess the convergent validity using factor analysis (Appendix F) and scale reliability with reliability tests (Appendix G) on the constructs and variables. First the factor analysis was conducted followed by a reliability analysis. The factor analysis was done at once including all items of each variable. The following criteria were used to check whether the data was suited for factor analysis. According to Winter, Dodou and Wieringa (2009) sample sizes below n = 50 can still generate acceptable results when doing a factor analysis. The sample size of this research (n = 50) needs to be treated with caution in order to produce reliable results. Therefore is checked in the correlation matrix for coefficients greater than .3. Many correlations of .3 and above were found in the correlations matrix, indicating factor analysis is appropriate. Furthermore the Kaiser-Meyer-Olkin value of .7 is above the threshold of .6 needed for a good factor analysis. Subsequently is checked for significance (p < .05) by looking at the Bartlett’s Test of Sphericity, which considers being significant with p < .00. The relationships between the variables were considered linear based on the results of scatterplots among the different relations. Outliers were not detected in the data.

The second step in the factor analysis entails factor extraction with the use of principal component analysis. Following Kaiser’s criterion only components with an Eigen value of 1 or above are of interest for further analysis. The first eight components show Eigen values greater than 1, explaining 31.2%, 11.6%, 7.6%, 7.3%, 5.3%, 4.5%, 3.6% and 3.2% of the variance respectively, 74.3% in total. The screeplot of Eigen values showed only three components indicated by the change in the curve of the plot between components three and four. The results of a parallel analysis (32 variables × 50 respondents) developed by Watkins (2000) showed two components exceeding the original Eigen values, which means that two components should remain based on parallel analysis, as shown in Table 2. Based on the theory there should be at least three components representing the main variables. Further analysis will reveal the final solution of how many components should remain for further analysis.

Table 2.

In the component matrix most items load above .4 on the components one till five, a few factors load on components 6, 7 and 8. This suggests that a five-factor solution is assumingly more appropriate. The pattern matrix revealed that items loading above .3 are structured around five components. These results were not in line with the expectations based on the pre-defined validated scales. Based on the predefined constructs six factors were determined: perceived diversity, relationship conflict, task conflict, process conflict, status conflict and subjective performance.

The five fixed variables explained in total 63.0% of the variance, with components one till six explaining, 31.2%, 11.6%, 7.6%, 7.3%, and 5.3% of the variance respectively. According to the communalities all items show high values above .5, indicating that all items fit well with the corresponding components. The interpretation of the data was done with the use of Oblimin rotation. The five-factor solution as shown in the pattern matrix as presented in Table 3 was used to match items to a component. The twelve items representing perceived diversity were scattered around components 2, 4 and 5. This result could be explained because these factors represent different but related dimensions of perceived diversity namely: surface-level, deep-level and general. Because one item of the general perceived diversity scale load on the component representing the surface-level perceived diversity scale, this item was considered as part of the surface-level perceived diversity scale. The occurrence that all conflict types load on component 1 can be explained by the underlying theory that all twelve items include some level of conflict measurement but from a different dimension.

Correlations among the different conflict types were high (.6 and above), indicating the problem of multicollinearity. Other researchers also found problems of high correlation among the different forms of conflict (Jehn et al., 1999). Jehn et al. (1999) also explain that many researchers did found diversification between the types of conflict. Furthermore the seven items measuring subjective performance loaded on component 3 indicating the items measure the same construct. Overall it appears from the data that the items used to measure

the variables of interest are somewhat valid. An important finding was that based on the correlations and the factor analysis no differentiation between the conflict types was noticeable. It is also important to note that the scale of perceived diversity is divided among three different dimensions representing, surface-level diversity, deep-level diversity, and general diversity.

The results of the factor analysis show there are three main variables for further analysis: perceived diversity (existing of 3 components), conflict types (1 component) and subjective performance (1 component). Separation between five different components was found. There was a weak correlation below .3 between the five components, which indicated that the five components were not strongly related. The differentiation between these variables was mainly expected based on the scale measurement tools used. Except all loadings of the different conflict types on the same component indicated that statistically the variables are congruent. For further analysis relationship, task, process and status conflict are treated as one and the same variable representing the level of intra-group conflict. Reliability analysis (Appendix G) on each scale was done after the factor analysis (Appendix F) as explained in each section describing the measurement of each variable. A new variable was created, representing the different scales measuring perceived diversity. The following five variables were used for further analysis: surface-level, deep-level and general perceived diversity, intra-group conflict and subjective performance. Table 3 provides an overview of the distribution of the items on different components based on the pattern and structure matrix.

Preliminary analysis was done to check the variables were in line with the assumptions of normality, linearity and homoscedasticity. Normality of the variables was checked using a Shapiro-Wilk test appropriate for relative small samples of n = 50, for values of p ≤ .01 or p ≤ .05. The scales measuring surface-level perceived diversity (p = .00) and intra-group conflict (p = .04) were considered normally distributed. Not normal distributed were deep-level perceived diversity (p = .06), general perceived diversity (p = .08) and subjective performance (p = .38). Although in general the Likert type items are considered as ordinal scales not applicable for parametric tests, however the items that together represent Likert scale data can be analysed using parametric analysis (Boone & Boone, 2012; Gliem & Gliem, 2003; Norman, 2010).

Scatter plots were used to get a first impression of the relationships between the variables. It is noticeable from this rough analysis that intra-group conflict appear to have a linear negative effect on subjective performance, with intra-group conflict associated with a decrease subjective performance. Furthermore the linear relationship observable between

surface-level perceived diversity and intra-group conflict was neither positive nor negative. Deep-level perceived diversity shows a positive linear relation with intra-group conflict, associating perceived deep-level diversity with an increase of intra-group conflict.

Table 3.

A positive linear relation between general perceived diversity and intra-group was found, implying that general perceived diversity positively related to the levels of intra-group conflict. The linear relation between surface-level perceived diversity and subjective performance is slightly positive, indicating that perceived surface-level diversity is positively

related to subjective performance. The linear relation between deep-level perceived diversity and subjective performance was negative, indicating that perceptions of deep-level diversity are associated with a decrease of subjective performance. A negative linear relationship was observed between general perceived diversity and subjective performance, which implies that higher general perceived diversity negatively influences the level of subjective performance. The scatterplots do not show a clear indication for heteroscedasticity, assuming that the variability in the scores of the variables was similar. Subsequently a Pearson correlation test was done to provide information on the strength, significance and direction of the relationships by looking at the Pearson correlation coefficients. Finally linear regression analysis and mediation analysis were conducted to test the hypothesis in this research.

Pearson correlation analysis was conducted for representing the relations between the variables as shown in Table 4. Values of the Pearson correlation indicate the nature of the relationships, values > 0 associated with positive relations and values < 0 linked to negative relations. The value range of -1 till 1 represents the amplitude of the relations where -1 is associated with strong negative influence and 1 stands for a strong positive relation. Significance of the relations were checked by looking for values of p ≤ .01 or p ≤ .05, represented by the bolded figures in Table 4.

Hierarchical multiple regressions were used to investigate the predicting ability between variables, after controlling for team size, gender, age, nationality, teamwork experience and educational level. Independent variables, not included in the main regression will also be included as control variables to assess the predictive power of each variable separately while controlling for the other independent variables. As described in previous parts of the methodology section preliminary analyses were conducted to ensure the assumptions of normality, linearity and homoscedasticity were not violated.

The mediation analysis was conducted following Baron and Kenny (1986) who provided three regression equations to estimate in order to test for mediation. The following regressions were estimated. First, regression was conducted between types of perceived diversity on subjective performance. Secondly, regression was performed between forms of perceived diversity on intra-group conflict. Thirdly, regression of intra-group conflict on subjective performance was performed. Finally was tested for insignificance of the relation between forms of perceived performance and subjective performance in the presence of intra-group conflict, indicating mediation in the relation between types of perceived diversity and subjective performance by intra-group conflict.

5.0 Results

In this chapter, the results of this study are presented. First the frequencies and the correlations among all variables are shown in Table 4 and interpreted. Secondly, results of linear regression will be discussed to get an overview of the nature of the relations between the variables. Finally, the conducted mediation analysis will be displayed to expose to what extend the relationship between perceived performance and subjective performance was mediated by intra-group conflict. The results of preliminary analysis revealed high correlations of .6 and above among the predefined set of conflict variables. According to Pelled et. al (1999) high correlations are problematic for regression analysis. The five-component solution of the factor analysis was retained for further analysis. This implies that intra-group conflict will be seen as one variable and perceived diversity was divided into surface-level, deep-level and general perceived diversity. Furthermore subjective diversity was included without any alterations in the items.

The relationships among all variables were explored using Pearson correlation coefficients. There was a moderate, positive correlation between deep-level perceived diversity and intra-group conflict, r = .30, n = 50, p < .05, implying that deep-level perceived diversity increases with intra-group conflict for members in entrepreneurial teams. A moderate, positive correlation was found between general perceived diversity and intra-group conflict, r = .43, n = 50, p < .01, with general perceived diversity associated with a increase levels of intra-group conflict. Further there was a moderate, negative correlation between general perceived diversity and subjective performance, r = -.30, n = 50, p < .05, suggesting that a increase of intra-group conflict associated with lower levels of subjective performance. Finally there was a strong, negative correlation between intra-group conflict and subjective performance, r = -.54, n = 50, p < .01, indicating that intra-group conflict has a decreasing effect on levels of subjective performance.

Some control variables correlated with the main variables. Although in all cases control variables could be related, these results show that it this specific context the used set of control variables was necessary to be included in order to explain a part of the variance in the relations. There was a moderate, positive correlation between teamwork experience and surface-level perceived diversity, r = .30, n = 50, p < .05, with levels of teamwork experience associated to increase surface-level perceived diversity. There was a moderate, negative correlation between teamwork experience and deep-level perceived diversity, r = -.28, n = 50,

p < .05, indicating that teamwork experience is associated with lower levels of deep-level

and subjective performance, r = .35, n = 50, p < .05, with levels of education associated to increase levels of subjective performance. The variable nationality correlated significantly with age and teamwork experience, however these relations were not representative for this research.

Table 4.

The variables included in the Pearson correlation analysis as depicted in Table 4 were used to conduct linear regression and mediation analysis to test the hypotheses. Preliminary analyses indicated the original conceptual model was no suitable for generating reliable results because of two main reasons. Firstly, principal component analysis and high correlations among the conflict types indicated that separation between forms of conflict was technically not possible. Secondly, perceived diversity loaded on three dimensions representing surface-, deep-level and general perceived diversity. Therefore, the pre-defined hypotheses were altered to correspond with the new differentiation between variables. The following new hypotheses are stated below in order to find out: “To what extent is the

relationship between ‘surface-, deep-level, general perceived diversity’ and subjective performance mediated by intra-group conflict?”