Bachelor Thesis
The effect of gender diversity on team performance
Author: Anne van der Klugt -‐ 10610936
Programme: Economics and Business -‐ Finance and Organization Supervisor: Adam Booij
Date: June 29, 2016
ABSTRACT
The purpose of this thesis is to examine the effect of gender diversity on team performance. Data for this study were collected at the University of Amsterdam. The subjects are undergraduates in the second-‐year course ‘International Money’. They were distributed over 379 self-‐selected teams of 2-‐3 students that differ in the share of female. The teams made three group assignments during the course: an essay, a presentation and a discussion. This quasi-‐experiment examines if mixed-‐gender teams perform better than single-‐ gender teams and in addition to that, in which kind of assignments gender diversity has the biggest impact on team performance. The results did not support the hypothesis that diverse teams outperform single-‐gender teams. The overall result shows that there is no significant difference between the performance of mixed-‐gender teams and that of single-‐ gender teams with the same proportions.
Statement of Originality
This document is written by Anne van der Klugt 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.
TABLE OF CONTENTS
1. Introduction 4
2. Related literature 6
2.1 The synergetic effect of teams 6 2.2 Determinants of team performance 7 2.3 The effect of gender diversity on team performance 8
2.4 Main hypothesis 10
3. Data & Methodology 11
3.1 Participants 11
3.2 Procedures and quality of the data 13
3.3 Model setup 14
3.4 Summary statistics and balancing 16
3.5 Sample selection 20
4. Results 20
5. Discussion and conclusion 24
REFERENCES 27 APPENDICES 29
1. Introduction
For a long time, the use of teams was very exceptional, but nowadays many organizations assign tasks to teams instead of individuals (Hamilton, Nickerson & Owan, 2003; Hertel, 2011; Robbins & Judge, 2012). This shift in organizational behavior reflects the expectation of firms that teams could have advantages compared to individual workers (Hertel, 2011). Therefore, managing team performance has become a topic of interest. Due to the recent globalization and changes in the labor market, team diversity has become an important determinant of performance (Robbins and Judge, 2012). The concept of diversity is very broad. This thesis focuses on one specific aspect of diversity namely: gender diversity.
Kösters, den Boer and Lodder (2009) analyze the employment rates for men and women in the Netherlands. Figure 1 illustrates the importance of managing gender diversity. The employment rate of women (dashed line) has risen by more than 30% since 1970. Although the employment rate for men (dotted line) is still higher, it has not changed that much over the past few years.
Figure 1
Gross employment rate in the Netherlands in 1970-‐2008
Note: For this graph, Kösters, den Boer and Lodder (2009) used data of the CBS
The increase of female on the labor market could cause a change in gender composition of teams in organizations. To reach an optimal performance level, managers should know how to compose a team that provides the best outcomes. Another related issue is the underrepresentation of women in higher functions (Adams & Ferreira, 2009). Governments try to increase the share of females in
the board by introducing regulations that require a fixed percentage or amount of female in the board (Ahern & Dittmar, 2012; Masta & Miller, 2013). It remains unclear however, whether mixed-‐gender teams would outperform teams consisting of men or women.
To examine the effect of gender diversity on team performance, this study looks at quasi-‐experimental evidence of teams. The teams that were used are teams of students in the course ‘International Money’1 at the University of
Amsterdam. In this course, the students divide themselves into teams of two or three students that differ in share of female. Although the assignment is non-‐ random, the students in the different teams don’t differ much in observed characteristics. The core assumption is that this carries over to unobserved differences as well. The students should make three different group assignments together: an essay, a presentation and a discussion. The groups receive a different grade for each assignment, which is used as the measure of performance.
The overall results show that gender diversity does not affect team performance. There is some suggestive evidence however, that single-‐gender teams outperform teams of one female and two male students. This effect is only significant for the discussion assignment. In the other assignments, in which exactly the same students and group compositions are analyzed, no significant relationship was found.
This thesis contributes to literature because three different performance measures are studied, which allows us to examine in which type of assignment gender composition has the biggest impact. Fenwick and Neal (2001) provide evidence that there is no relation between the share of women in a team and their performance in writing a business report. However, Goldin, Katz, and Kuziemko (2006) show that women do have better verbal skills than men. These findings suggest that the impact of gender diversity could differ among the assignments; the influence of gender diversity should be larger for the presentation or discussion assignment than for the writing assignment.
In the next section, literature that is related to this topic is discussed. After that, the data used for this study is described in more detail and the empirical
model is explained. In the fourth chapter, the results are presented. The last section of this paper discusses the results and sums up the main conclusions.
2. Related Literature
2.1 The synergetic effect of teams
The rapid increase in assigning tasks to teams instead of individuals did not happen by accident. Many researchers acknowledge the advantages of teams compared to individuals. According to Hertel (2011), this phenomenon is called the synergetic effect. This means that teams could obtain a higher performance level compared to the accumulated individual performance of all the members of the team (Robbins & Judge, 2012). As a result, an organization can create greater outputs by using the same inputs as before.
The positive synergy of teams is caused by a couple of processes. According to Hertel (2011), teams are more creative and can provide a more diverse view compared to individuals. Additionally, teams are better in providing solutions for problems. The reason for this is that teams view problems, statements or ideas from all different kinds of perspectives. The effectiveness of teams is also strengthened by the competition between the members. Hertel (2011) says that due to intragroup competition, members exert more effort. Other researchers also studied the difference in performance of teams compared to individuals. Gigone and Hatsie (1997) found that the accuracy of a team is higher than the average of the individuals of that team. On top of that, teams could give support to other team members, which could cause a higher motivation and improves coordination (Hüffmeier & Hertel, 2010).
There are also papers that emphasize the drawbacks of using teams. Bennett (2004) examines the concept of social loafing. This means individuals exert less effort in a team compared to individually. Social loafing often happens when a team becomes too large. In that case, team members are unable to check how much effort each individual exerts because individual performance is not measurable anymore (Robbins & Judge, 2012). According to them, this causes a lack of responsibility, which will result in free riding of individuals. Another drawback of the use of teams is the need for conformity. Individuals feel the pressure of the group to act the same as them and adjust their behavior in line
with that of their group members (Asch, 1956). This could hurt the teams’ creativity and diverse view.
To achieve a positive synergy by using teams, a manager should know how to compose the most effective team.
2.2 Determinants of team performance
According to Maznevskis’ basic model of group processes2 (1994), there are
three main factors that influence group performance: member characteristics, group characteristics and group processes. For this thesis, it is important to describe a couple of these member-‐ and group-‐characteristics because it helps to understand why certain variables are included in the empirical model.
2.2.1 Group characteristics
A first determinant of group performance is group size. Seijts and Latham (2000) argue that smaller groups are faster in completing tasks and that the performance of individuals is higher in a smaller team. A reason for this could be that social loafing hurts team performance when a team becomes too large (Bennett, 2004). Larger teams could be preferable for decision-‐making tasks (Robbins & Judge, 2012) because a larger team views statements from different perspectives and therefore could provide a more diverse output (Hertel, 2011). Furthermore, group cohesion could be an important determinant of team performance. Members in teams with a high level of cohesiveness are attracted to each other and want to stay in the group (Robbins & Judge, 2012). Dorfman (1984) discusses the effect of cohesiveness on team performance in his paper. He observed students in small groups that participate in a business game. Performance was measured in terms of profit. The results suggest that teams that were high in cohesiveness outperform teams with a low level of cohesiveness.
2.2.2 Member characteristics
Besides group characteristics, individual characteristics are important. The skills and abilities of the team members could influence the total team performance
(Robbins & Judge, 2012). According to them, high-‐ability teams perform better than low-‐ability teams. In addition to that, high-‐ability teams are more flexible.
Another determinant of team performance is the average age of the team members. Streufert, Pogash Piaseck and Post (1990) studied the effect of age on team performance by analyzing teams in different age-‐categories. The results suggest that teams consisting of older people made fewer decisions. These teams needed more time to make decisions and ignored relevant information for making these decisions. Furthermore, older teams scored lower on all the planning and strategy performance measures.
2.3 The effect of gender diversity on team performance
This paper focuses on the influence of gender diversity on team performance. Despite the many literature written about the effect of gender diversity in teams, there is no clear answer about which kind of team performs best.
Adams and Ferreira (2009) analyze the effect of gender diversity on boardrooms in US-‐firms. They use firms of the S&P 500, S&P MidCaps and S&P SmallCap. The effect of gender on observable measures of board inputs and board level governance characteristics was analyzed. The researchers found that female directors are less likely to experience attendance problems and that the CEO turnover increases with the share of female directors. Additionally, women participate more in monitoring committee meetings than men. The overall performance, measured in Tobins’ Q and return on assets, is worse when there is greater gender diversity. The cause for this could be that women increase the monitoring intensity of the board. Too much monitoring would decrease the shareholders value (Adams and Ferreira, 2007). They conclude that, for firms with strong shareholder rights, extreme monitoring is counterproductive.
On the other hand, Carter, Simkins and Simpson (2003) found a positive relationship between gender diversity and performance, measured in Tobins’ Q. To study this, they analyzed 638 publicly traded Fortune 1000 firms. The researchers made a distinction between low-‐women firms, which means the firms have no female in the board, and high-‐women firms, which means the firm had two or more women in the board. Besides an increase in performance they
also found that firm size increases when the share of female is larger. However, the researchers did not study what could be the cause of these relationships.
Another commonly used method for studying the effect of gender diversity on performance is analyzing performance of business teams. The participants in these games are mostly undergraduate students. Fenwick & Neal (2001) studied a business game that consists of 65 self-‐selected teams of 4-‐7 individuals. The teams were ranked into five groups on basis of their profit. In addition, the researchers include a second measure of performance; the quality of a business report that the teams have to write. No relationship was found between the writing skills of the teams and the share of female. However, the results indicate that the share of female students in a team is positively related to profit. 88% of the groups that had a high ranking (1 or 2) were consisting of two or more female students. Additionally, teams with a share of female of at least 0.4 obtained higher profits over the ten periods of the game. The researchers argue that women are more cooperative and people-‐oriented compared to men. On the other hand, men are more competitive. This combination explains why mixed-‐ gender teams perform better.
Later on, Apesteguia, Azmat, and Iriberri (2012) studied another business game with teams consisting of three members. They classified the teams into four groups: teams without female, teams with only one female, teams with two female and lastly, a team consisting of only female students. The experiment provides evidence that all the other teams outperform teams consisting of three women. The reason for this could be that these teams were less aggressive than others. In addition to that, the researchers found suggestive evidence that a team of two men and one woman performs best.
Hoogendoorn, Oosterbeek and Praag (2013) did also observe the performance of undergraduate students in business teams. They analyzed 45 teams that participated in an entrepreneurship program. The students should manage a business in small teams. The share of women in the team varies between 0.1 and 1.0. Performance was measured in terms of sales, profits and earnings per share. They conclude team performance is optimal with a share of woman between 0.5 and 0.6. They provide evidence that the graph of performance has an inverse U-‐shape. First, performance goes up when the share
op female increases. When the share of female reaches 0.5-‐0.6, including more female in the team could only hurt the teams’ performance. According to the researchers, part of this relationship could be explained by the fact that diverse teams are more extreme monitors. They found that the level of monitoring is highest in teams with a share of female of 0.5 and that the relation between performance and monitoring is again inverse U-‐shaped. In contrast to the findings of Adams and Ferreira (2009), monitoring is positively related to performance.
This thesis differs from former literature by using other measures of performance. Not the performance of boards or business games is analyzed, but the performance of student-‐teams in a course at university. Where other papers focus on firm value or profit as a measure of performance, this paper examines the effect of gender diversity in teams on writing skills, presenting skills, and discussion skills of the group. The weighted average of these three performance measures is also considered for reasons of accuracy. On top of that, it is possible to examine if gender diversity of teams has a greater impact on performance in one particular assignment. Furthermore, the sample that is used for this study is larger than in most literature in which undergraduates are studied.
2.4 Main hypothesis
To examine the effect of gender diversity on team performance, the following hypothesis follows from literature:
H1: Gender diversity has a positive effect on team performance.
Because this quasi-‐experiment is most comparable to the papers in which they use undergraduates (Apesteguia et al., 2012; Fenwick and Neal, 2001; Hoogendoorn et al., 2013), the expected effect of gender diversity is positive. To confirm this hypothesis, the coefficients of the mixed-‐gender groups should be positive and significantly different from zero.
3. Data & methodology 3.1. Participants
The subjects of this quasi-‐experiment are students in the course ‘International Money’. The course is a second year course of the bachelor in Economics and Business at the University of Amsterdam. For students that have chosen the track Economics, Finance or Economics & Finance this is a compulsory course in their programme. The course is focused on the monetary relations: in theory, practice and policy. Important topics in this course that will be studied are the balance payments and the foreign exchange market (UvA, 2014).
Data of four past academic years is studied, namely: 2011-‐2012 till 2014-‐ 2015. The number of participants varies across the years and is, after cleaning the data, 146, 194, 196 and 308 respectively. In total the information of 843 students is used. The fraction of female students in this sample is 31.1% (n= 262) and 68,9% (n=581) are male students. Only data of the Dutch version of this course is analyzed.
In each workgroup the students are divided into self-‐selected groups of two or three individuals. The teacher of the course is not involved in the group making process, unless students are unable to find a group by their selves. The students are distributed over 378 groups. 23% (n=87) are groups of three students and 77% (n=291) consists of two students. Participants could not be part of two different groups in the same academic year, but there are a couple of students that need more than one attempt to pass this course. As a result, the same students could participate in this study for more than one year.
Table 1
Distribution of students, teams and share of female over academic years
Year Students Teams Share of female
2011-‐2012 145 68 0.27 2012-‐2013 194 85 0.32 2013-‐2014 196 89 0.28 2014-‐2015 308 136 0.34 Total 843 378 0.31
During the course, there are three types of group assignments, namely: (I) writing an essay, (II) giving a presentation and (III) discuss another groups’
essay in presentation form. The group assignments are called skills activities because they test three different skills.
First of all, the writing skills of the teams are tested. The students should write an essay of 1200 words. The grade obtained for the essay will count for 10% of the end-‐grade of the course. Students write about a self-‐chosen topic that is related to the curriculum of the course. The students need to write a short analytical and critical text based on academic sources. In the essay, they answer an explanatory research question. The students are assessed on spelling and academic writing, and should only give information that is necessary and important. If the students hand in their essay too late, it is graded as a zero.
Second, the group has to present their essay findings. The presentation has duration of ten minutes and counts for 5% of the end-‐grade. The students are evaluated on content, non-‐verbal, and verbal skills. Giving the right answers to questions of fellow students is also taken into account. When the students are not present during the workgroup they have to do their presentation, the assignment is again graded as zero.
The last assignment is the discussion. The discussion counts for 5% of the end-‐grade too. The purpose of this assignment is to give critical comments on another groups’ essay. First, a very short summary should be given. After that, the essay is evaluated and finally a conclusion is provided. Only giving comments on misspellings and the layout of the essay is not sufficient. Again if there are students not present, their grade is zero.
The presentations are distributed over several weeks. There are a couple of presentations per workgroup and after each presentation, the discussion about that same essay follows. The teachers differ among the workgroups and the grades they give could vary between 0 and 10. These grades are announced after all the presentations and discussion have taken place. For the essay, the deadline is the same for all groups. The remainder 80% consists of an individual grade for the written final exam that counts for 70% and a group presentation about the theory discussed in class that counts for 10%. Sometimes students are not picked out to give a presentation. If not, their end-‐term counts for 80%.
3.2 Procedures and quality of the data
First of all, with the approval of the coordinator of the course, administrative data regarding students’ ID, group composition and group grades were collected. After that, the information specialist of the Faculty of Economics and Business collected additional information about the students’ gender, age, credits obtained in the first year, and grades obtained in the first year. Before the data could be used for research, the student ID numbers of the participants were replaced for random numbers, so the subjects in this study are as anonymous as possible.
The advantage of collecting data in this way is that there is no problem with low response rates or inaccurate answers. Additionally, it is possible to use a large sample, which makes the quasi-‐experiment more precise.
Nonetheless, some observations were left out because these were not useful. First of all, some student ID numbers and group grades were missing. According to the explanation of the coordinator, these missing observations were due to students that signed up for the course, but never showed up. The students without a student ID are students that decided to participate in this course after the official registration. It was impossible to track down these students, which means there is too little information about their demographics. Both groups are ignored.
Another complication of this sample is that some students obtained a zero for all three assignments. This means the students never handed in any
Table 2
Summary of the assignments in the course International Money
Essay Presentation Discussion
Evaluated on: -‐ Spelling -‐ Academic writing -‐ Giving only necessary and important information Evaluated on:
-‐ Non verbal skills -‐ Verbal skills -‐ Answering questions of fellow students Evaluated on: -‐ Critical comments Weight:
assignment and were not present in the workgroups when they had to present or discuss an essay. An explanation for this could be that the students gave up on the course in an early stage. Under assumption the students did not leave the course because of their team composition, the students were left out.
A third complication is that there were some groups where the teacher individually graded the members of the team. In addition to that, there were team members that decide to not show up for their presentation or discussion but the other team members did. Both lead to different grades within one team. To tackle this problem the dummy ‘intact’ was created. With a regression for binary outcomes it was tested if the group composition had an effect on the group be intact or not. If not, sample selection is not a problem.
Because this paper ignores the observations mentioned above, some ‘teams’ were left with only one person. These observations measure the performance of an individual, rather than team performance. Therefore, these observations are ignored in this study.
3.3 Model setup
Performance is the dependent variable in the model. To measure group performance, the grades for the three different kinds of assignments are used: the essay, the presentation, and the discussion. All the members of the group receive the same grade for their assignment. For every kind of assignment, a separate regression is done so that the degree of influence on performance among the assignments is comparable. There are also two regressions done for the weighted average of the three assignments in which the essay counts twice as much as the presentation and discussion assignment, just like in the course. Share of female is the variable of interest. The share of women in the groups could be 0, !!, !!, !! or 1. The distribution is 57.7% (n=218), 4.8%(n=18), 14.8%(n=56), 2.11%(n=8) and 20.6% (n=78), respectively. These five different groups are distinguished in the model with dummy variables: F=0, F=𝟏
𝟑, F= 𝟏 𝟐, F=
𝟐 𝟑 and F=1. The two homogeneous groups (F=0 and F=1) are used as a reference point. The most interesting team is the team consisting of one female and two
male students because this team best represents the share of female in the total sample (31.1%).
Control variables that are included in the regression are group size, average age, average credits, and average GPA. There are also dummies included for special students that finished their first year under certain circumstances.
Size: As discussed in the literature section, size could be an important
determinant of performance. The groups in this quasi-‐experiment consist of two or three students, which means groups are small and do not vary a lot in size. The teams where the share of female is 𝟏
𝟑 or 𝟐
𝟑 are always groups of three students
and teams that have a share of female of !! are always consisting of two students. The single-‐gender teams could vary in size.
Age: The age of the students could also influence group performance. The
age that is included is the age of the students at the beginning of the course. This means the students that need more than one attempt to pass the course could have different ages over the academic years.
Credits: The average credits are calculated over the first year only,
because ‘International money’ is a second year course.
GPA: GPA is also calculated only over the first year. In fact, it is not the real
GPA but an approximation. It is measured as an average of all the end-‐term grades in the first year. Because of the administration system of the University of Amsterdam, insufficient grades are not taken into account by calculating the GPA. Therefore, it is measured in this way.
Table 3
Frequency of share of female Share of female N % 0 218 57.7% 1 3 18 4.8% 1 2 56 14.8% 2 3 8 2.11% 1 78 20.6%
Note: ‘N’ denotes the number of groups observed for the given share of female
0 50 100 150 200 250 0 1/3 1/2 2/3 1 Figure 2
Distribution of teams
NE, FS and HC: These control variables are all dummies. The dummies
have the value 1 if there is a student in the group that finished the first year suffering from one of the following circumstances: (I) negative study advice = NE (II) ’februari-‐stakers’ = FS or (III) ’hardheidsclausule’ = HC. The dummies are included because these students have a very low number of credits, and often a GPA below the average. By negative study advice the University of Amsterdam means students that have not reached the mandatory amount of 42 credits. They have to leave the faculty but could try it again after three years. In this sample there are four of these students. The second special group consists of three students who are called ‘februari-‐stakers’. These students quit the study in their first year before February. In this way, students get part of their tuition back and can sign up again for a study at the Faculty of Economics and Business next year. Their credit points are between 0 and 6. Lastly, there were students that had not obtained enough credits the first year, but could stay because of a special personal reason, which is called the ‘hardheidsclausule’. This was the case for 25 students. Their credit points are below 42.
3.4 Summary statistics and balancing
Table 4 reports the mean sample statistics of the teams. The table shows that the grades for the three assignments vary between 0-‐9.5. This is because some teams did not hand in their assignments in time, or were absent when they had to do their discussion or presentation. Furthermore, the fraction of females among the teams varies between 0-‐1, which means there exist groups without women as well as groups without men. The average age in the groups is quite different, with a difference of almost 8 years between the team with the youngest and the team with the oldest members. The minimum amount of credits of the group is a bit confusing because normally, students need to have 42 credits or more to go to the next year. The ‘special students’ that were discussed before cause this low number. The low GPA is because in the approximation that is used, only the end-‐term grades are included. This means also insufficient grades are taken into account for the GPA.
In table 5 the mean characteristics are shown for the teams classified by the share of women in the team. The classification of the teams corresponds with
the dummies for the different team compositions in the model. Interesting is that the mean credits of a team increases when the share of female goes up. However, GPA is highest in a group with one man and two women. If there was not controlled for these characteristics, this could complicate the analysis. If the number of credits were positively related to performance, teams with a higher share of female students would always do better regardless of team composition. In addition, when the GPA is positively related, teams consisting of two female and one male would do better. The inclusion of team characteristics as control variables will reduce the selectivity of the assignment. However, it cannot eliminate it.
Table 4
Descriptive statistics of performance measures and group characteristics
Variable N Mean Standard
deviation Min Max
Size 378 2.23 0.42 2 3 Age 378 21.3 1.18 19.3 27.1 Credıt 378 53.64 6.22 24 60 GPA 378 6.25 0.79 3.85 8.31 NE 378 0.01 0.1 0 1 FS 378 0.01 0.09 0 1 HC 378 0.07 0.25 0 1
Share of female 378 0.31 0.40 0 1
Essay 378 6.76 1.18 0 9.5
Presentation 378 7.04 1.39 0 9.5
Discussion 378 7.03 1.74 0 9
Note: ‘N’ denotes the number of groups observed. Over bar denotes group average.
Table 5
Mean characteristics by gender composition of teams
F=0 F=𝟏 𝟑 F= 𝟏 𝟐 F= 𝟐 𝟑 F=1 Size 2.21 (0.03) 3.00 (0.00) 2.00 (0.00) 3.00 (0.00) 2.21 (0.05) Age 21.43 (0.08) 21.27 (0.25) 21.50 (0.15) 20.81 (0.28) 20.89 (0.10) Credıts 52.89 (0.44) 52.56 (1.44) 54.21 (0.83) 55.71 (1.41) 55.35 (0.62) GPA 6.18 (0.05) 6.03 (0.21) 6.27 (0.10) 6.67 (0.35) 6.44 (0.08) NE 0.02 (0.01) 0.00 (0.00) . . . . . .
After this, the individual characteristics of the students in the teams are analyzed. Table 6 specifies the individual characteristics of males and females among the five different teams. The average age of the women in this course is somewhat lower than the age of men. Additionally, females have collected more credits and have rounded the first year with a higher GPA than men did.
It is possible to test if the students among the teams differ in their characteristics. In other words: do the females and males in the one team have the same characteristics as females and males in other teams. Table 7 reports the results of a regression of the individual characteristics on the dummies that indicate the share of female, separately for male and female students. Each column represents a different regression. There are not many significant differences in individual characteristics among the teams but a few exceptions are shown. When you are a man with negative study advice, the chance is higher that you are in a team with only men. In addition, men that made use of the ‘hardheidsclausule’ more often work together with one female. However, when a female used the ‘harheidsclausule’ the probability that she is in a group with only female is higher. This is also the case for female that were a ‘februari-‐staker’ in their first year. Additionally, when a female student has a higher age, they more often work together with one man. To check if the characteristics are jointly significantly different from zero, an F-‐test is done. In almost all cases, the variables are jointly significantly different from zero. Which means they have an effect on team composition. On top of that, a second test is done to check if there is any significant joint effect of the characteristics. These results also indicate that there is imbalance in characteristics (p<.01).
FS 0.01 (0.00) 0.06 (0.06) . . . . 0.01 (0.01) HC 0.064 (0.02) 0.11 (0.08) 0.04 (0.03) . . 0.09 (0.03)
Note: The values in parenthesis denote standard errors. ‘.’ Denotes that there are no students under this circumstance for the given share of female. Over bar denotes group average.
Table 7
Assignment of individuals separated for male and female students among teams Female in teams consisting of: Male in teams consisting of:
F= 𝟏 𝟑 F= 𝟏 𝟐 F= 𝟐 𝟑 F=1 F=0 F= 𝟏 𝟑 F= 𝟏 𝟐 F= 𝟐 𝟑 Age 0.006 (0.013) 0.044* (0.025) -‐0.003 (0.010) -‐0.041 (0.027) -‐0.009 (0.010) -‐0.001 (0.006) 0.012 (0.008) -‐0.002 (0.002) Credit -‐0.002 (0.004) -‐0.002 (0.006) -‐0.005 (0.005) 0.009 (0.007) -‐0.001 (0.003) 0.001 (0.002) -‐0.000 (0.003) 0.000 (0.000) GPA -‐0.003 (0.021) -‐0.035 (0.033) 0.049* (0.026) -‐0.012 (0.040) 0.005 (0.024) -‐0.013 (0.016) 0.014 (0.019) -‐0.006 (0.007) NE . . . . . . . . 0.182** (0.089) -‐0.055 (0.051) -‐0.112 (0.075) -‐0.015 (0.016) FS -‐0.183 (0.194) -‐0.420 (0.286) -‐0.145 (0.197) 0.745** (0.340) -‐0.343 (0.386) 0.421 (0.365) -‐0.044 (0.103) -‐0.034 (0.030) HC -‐0.114 (0.082) -‐0.157 (0.171) -‐0.099 (0.080) 0.369** (0.181) 0.054 (0.104) 0.069 (0.095) -‐0.108** (0.049) -‐0.014 (0.011) F-‐value 3.80*** 14.47*** 3.45*** 27.63*** 12.07*** 4.59*** 9.15*** 1.32 P-‐value 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.25
Note: The values in parenthesis denote standard errors. The means are measured at the individual level. ‘.’ Denotes that there are no students under this circumstance in the team. Each column represents a separated regression. ***/**/* Denotes significance at the 1%/5%/10%-‐level.
Table 6
Characteristics of individuals by gender composition of teams Female in teams consisting of: Male in teams consisting of: F= 𝟏 𝟑 F= 𝟏 𝟐 F= 𝟐 𝟑 F=1 F=0 F= 𝟏 𝟑 F= 𝟏 𝟐 F= 𝟐 𝟑 Age 20.95 (0.28) 21.30 (0.18) 20.73 (0.23) 20.87 (0.09) 21.45 (0.08) 21.47 (0.29) 21.70 (0.24) 21.03 (0.35) Credit 55.33 (1.42) 54.43 (0.82) 56.94 (1.57) 55.47 (0.62) 52.75 (0.42) 51.17 (1.84) 54.00 (1.23) 53.25 (1.68) GPA 6.36 (0.24) 6.20 (0.12) 7.03 (0.25) 6.47 (0.08) 6.16 (0.05) 5.87 (0.22) 6.33 (0.14) 5.95 (0.43) NE . . . . . . . . 0.01 (0.00) . . . . . . FS . . . . . . 0.01 (0.01) . . 0.03 (0.03) . . . . HC . . 0.04 (0.03) . . 0.04 (0.02) 0.03 (0.01) 0.06 (0.04) . . . .
Note: The values in parenthesis denote standard errors. The means are measured at the individual level. ‘.’ Denotes that there are no students under this circumstance for the given share of female.
3.5 Sample selection
Before starting to test the effect of gender diversity, first should be tested if the teams that are not intact could be deleted from the sample. With a simple logistic regression for binary outcomes, the effect of all the variables included in the model on the dummy intact was tested. The dummy variables that indicate the share of female had no significant effect on the dummy intact. A second check is done to see if the dummies had a joint effect. Again no relationship was found (p=.84). This means the teams (n=53) could be deleted3.
Thereafter, some descriptive statistics about performance and characteristics are analyzed. This is done at the group-‐level as well as for the individual-‐level.
In order to examine if mixed-‐gender teams outperform single-‐gender teams, this thesis uses the ordinary least squares method. This technique is most commonly used in papers that discuss gender diversity. To confirm the hypothesis, the betas of the mixed-‐gender teams need to be significantly different form zero and positive (𝛽 > 0). This is tested with a T-‐test. For every regression, the option for robust standard errors is used. After the OLS-‐ regressions, a joint F-‐test was done. This test examines if all the betas of the mixed-‐gender teams are jointly significantly different from zero (𝛽 ≠ 0).
4. Results
In table 8 the mean performance of the three different assignments is shown. The last row shows the weighted average for all the assignments. The weighted total indicates that when the share of female goes up, performance goes up. But when the share of female becomes larger, performance goes down. After the share of female reaches 𝟐𝟑 , performance goes up again. Figure 2 reports this relation between the share of female in a team and the average of the weighted average of the performance measures.
Figure 2
Relation performance and share of female
In table 9 the correlations for the three performance measures are shown. These are all significant and positively correlated. This means that when the group obtains a high grade for the one assignment it is likely to get also a higher grade for the others. All can thus be used to measure an underlying performance concept.
Table 8
Mean team performance by gender composition of teams
F=0 F=𝟏 𝟑 F= 𝟏 𝟐 F= 𝟐 𝟑 F=1 Essay 6.72 (0.07) 6.78 (0.24) 6.48 (0.22) 7.21 (0.62) 7.00 (0.11) Presentation 6.99 (0.11) 7.11 (0.17) 7.10 (0.16) 6.91 (0.86) 7.16 (0.11) Discussion 6.92 (0.12) 6.90 (0.44) 6.85 (0.29) 7.13 (0.90) 7.48 (0.09) Weighted Average 6.84 (0.07) 6.89 (0.18) 6.72 (0.16) 7.11 (0.73) 7.16 (0.08)
Note: The values in parenthesis denote standard errors. The weighted average denotes the average of the grades weighted by the count for the end term namely: 10% for the essay and 5% for the presentation and discussion.
Table 9
Correlations between performance measurements
Essay Presentation Discussion
Essay 1.00
Presentation 0.42*** 1.00
Discussion 0.27*** 0.42*** 1.00
Table 10 provides the results of five different OLS-‐regressions. The first three columns give the results for the three different kinds of performance measures separated. The fourth column is a regression for the weighted average of the three performance measures without the inclusion of control variables. In the last column the weighted average is again the measure for performance but now control variables are included.
The results of the last two columns are a little bit different. This is because the inclusion of control variables removes the selectivity represented in table 7. Controlling for observable characteristics will reduce the selectivity of the assignment, but not eliminate it. Factors like motivation or cohesiveness are ignored but could seriously affect the results. In the regression of the weighted average measured without controls as well as the one measure with controls no significant effect of gender diversity on performance was found. This means H0 could not be rejected and mixed-‐gender teams do not perform better than single-‐ gender teams. The results indicate that gender diversity has no effect on team performance.
There is one exception to this statement. There is suggestive evidence that single-‐gender teams outperform teams consisting of one woman and two men. This coefficient is different from zero at the 10%-‐level. However, this is only the case in the discussion assignment. No systematic relation is found among the other assignments. The result is surprising because most comparable literature says gender diverse teams should outperform single-‐gender teams. In addition to that, Apesteguia et al. (2012) found suggestive evidence that a team of two men and one woman performs best.
There are some other conclusions that could be drawn from these results. In all three assignments as well as in the regressions for the weighted average, the GPA-‐coefficient is positive and significantly different from zero. This means when GPA goes up, the performance goes up. Because GPA is used as a measure for ability, this result is in line with the findings of Robbins and Judge (2012).
Another more remarkable finding is that when students have a ‘Februari-‐ staker’ in their team, their performance is higher. Reason for this could be that these students are extremely motivated. After all, they signed up for the same
