Does diversity matter? Explaining the effects of team diversity on team performance using data from a student’s business simulation game

Abstract

Purpose

The purpose of this study is to determine the impact of different diversity criteria (age and gender) on team performance using a business simulation game, where performance is operationalized by market valuation (Tobin’s Q) and profitability (cumulative net income).

Design/methodology/approach

The authors collected data of a business simulation game at a German public university (128 student groups and 645 individuals). Using multiple linear regression models, they investigated the impact of the different diversity criteria on team performance.

Findings

The authors found no significant effects of gender diversity on market valuation and profitability. However, the results of this study indicate a positive relationship between age diversity and both performance variables.

Originality/value

This paper contributes to the debate on the relationship between diversity and team performance by using data from a business simulation game. This study adds value by considering different diversity factors, objective performance indicators and the endogeneity of team formation (the seminar leaders randomly assign students to teams) which has not been applied in similar studies yet. Because of the setting of a simulated business game, the results could also be applied to the real economy where we observe working teams every day.

Citation

Zumkley, M., Sage, F., de Dios Panal, S. and Prinz, J. (2025), "Does diversity matter? Explaining the effects of team diversity on team performance using data from a student’s business simulation game", Team Performance Management, Vol. 31 No. 1/2, pp. 1-12. https://doi.org/10.1108/TPM-01-2023-0005

Introduction

For several decades, scientists have been examining the impact of team constellations on team performance (Hoffman and Maier, 1961; Rosenberg et al., 1955). One potential determinant on the efficiency of a team is diversity (Kukenberger and D’Innocenzo, 2019). Even though a high number of studies on this subject already exists, literature shows conflicting results on the relationship between diversity and performance. There are studies that postulate a positive impact on performance (Kukenberger and D’Innocenzo, 2019; Terjesen et al., 2016), whereas other studies find a negative (Adams and Ferreira, 2009; Shrader et al., 1997) or even no impact at all (Carter et al., 2010). Furthermore, when analyzing the literature, it becomes clear that past diversity research has often examined only one dimension of diversity (e.g. gender diversity) and neglected other possible diversity influences. Besides, some researchers who investigated the diversity–performance relationship (Apesteguia et al., 2012) did not consider the endogeneity and self-selection bias of team formation. In these research settings, team members selected themselves into the teams (Hoogendorn et al., 2013). Our study addresses these issues. In the context of a business simulation game for students, we investigate different dimensions of diversity and measure not only gender but also age diversity in teams. To account for the issue of endogeneity and self-selection bias of team formation in prior research (Hoogendorn et al., 2013), we conduct a business simulation game in which students are randomly assigned to teams.

Literature analysis shows that performance can be measured in multiple ways and often the calculation of a performance indicator is subject to a complex procedure. One of the benefits of using data from a business simulation game is that it provides the researcher with a quantitative, objective performance measure. This means that our research enables a measurement of diversity in the team and its direct impact on team performance, which is also a deficiency, primarily in the simulation literature (Brown et al., 2020). In summary, we contribute to the diversity–performance literature by analyzing the impact of different diversity factors on team performance in a controlled but realistic setting with considering a randomized group assignment and using objective performance indicators.

Theoretical framework

In the literature, various theories can be found that aim to explain the correlation between diversity and team performance from an organizational, economic, psychological or sociological perspective (Byrne et al., 1971; Tajfel and Turner, 1986; Terjesen et al., 2016; van Knippenberg et al., 2004; Williams and O’Reilly, 1998). The impact of diversity has been empirically examined in terms of two different types of outcomes in particular – one in terms of social functioning and the other in terms of performance, which is subject of this paper. Both the social categorization theory and the similarity-attraction theory describe the negative effects of diversity. These approaches assume that it is more difficult to get along with dissimilar compared to similar individuals. This consideration can be applied to both the group level (social categorization) and an individual level (similarity-attraction) (Carter and Phillips, 2017). The social categorization approach describes that similarities and differences between individuals lead to categorizing themselves and others into groups. This results in a categorization between one’s own (in-group) and one or more out-groups. Members of these out-groups are trusted less than members of one’s own group, resulting in preference for one’s own group (Tajfel et al., 1971; Brewer, 1979; Tajfel and Turner, 1986; Ferguson and Peterson, 2015). The similarity attraction theory describes that homogeneous individuals on an interpersonal level are socially attracted by similar characteristics, such as attitudes or values, which can lead to an improvement in cooperation (Newcomb, 1961; Byrne, 1971; Byrne et al., 1971). However, we focus on diverse teams in which exactly this aspect can rather lead to repulsion (Rosenbaum, 1986). Building on the social categorization and in conjunction with the similarity-attraction approach (Williams and O’Reilly, 1998), this means that the more diverse a group is, the less successful cooperation between individual group members is, and conflicts can arise as a result. This ultimately has a negative impact on team performance (Carter and Phillips, 2017).

In contrast, the information and decision-making approach highlights the positive effects of diverse groups on performance (Williams and O’Reilly, 1998). Here, group diversity leads to a larger stock of knowledge, perspectives, approaches, skills and available networks and, thus, increases the cognitive resources. Provided that these resources are used reasonably to solve tasks, this can lead to positive effects, especially in tasks that require a high degree of creativity and innovation as well as in terms of decision-making efficiency (Gruenfeld et al., 1996; van Knippenberg et al., 2004; van Knippenberg and Schippers, 2007). This can lead to an improvement in team performance despite coordination problems (Mannix and Neale, 2005). Accordingly, the negative effects caused by the social categorization and the similarity-attraction approach, and the positive effects caused by the information and decision-making approach are opposed to each other. This makes it difficult to derive clear statements about the effect of group diversity on performance. The current state of studies exemplifies this finding (Guzzo and Dickson, 1996; Milliken and Martins, 1996; Williams and O’Reilly, 1998; Homroy and Soo, 2020; Azmat and Petrongolo, 2014). Based on these approaches, we assume that differences in decision-making patterns and attitudes arise on the basis of individual characteristics. To verify this, the following observable diversity variables of the students were collected and are examined in the empirical analysis. Drawing on this theoretical framework and the results of previous studies presented, we developed two different hypotheses that describe the impact of the diversity variables on team performance.

Age diversity

We argue that the more diverse a team is in terms of age, the more different perspectives and experiences are brought together. This generates synergy effects and leads to different perspectives when solving a problem. Older team members bring in more experience, routine and knowledge, while younger team members may develop innovative ideas use unconventional approaches or current methods to overcome this challenge (Wegge et al., 2008; Li et al., 2020). There are more than 70 scientific studies and several meta-studies which have empirically investigated the effects of age diversity for a total of more than 10,000 teams from different industries and countries (Kunze and Reinwald, 2017). For this purpose, team performance was operationalized as team innovation for example (O’Reilly et al., 1998). In terms of overall team performance, minimal negative effects (Joshi and Roh, 2009) or no effects at all (Bell et al., 2011; Van Dijk et al., 2012) can be found. However, based on the research of Kearney et al. (2009) and Kilduff et al. (2000) and the assumption that age diversity promotes critical debate and reflective thinking because of a broader range of experience inside and outside the organization, leading to better problem-solving, we expect a positive effect on team performance. To test this in the setting of a business simulation game in a university context, we formulate the following hypothesis:

H1.

Age diversity has a positive effect on team performance.

Gender diversity

With regard to gender diversity, we also argue that a gender mix results in different perspectives on problems and, therefore, a broad spectrum of solutions. Moreover, men and women have different skills which, when combined, create synergy effects and lead to more creative, innovative and ultimately superior outcomes. Previous studies do not provide a clear picture of the impact of gender diversity on team performance. As mentioned above, while some studies show a negative relationship (Adams and Ferreira, 2009), others show a positive (Østergaard et al., 2009) or no significant relationship (Stewart and Johnson, 2009). Particularly with respect to leadership positions in companies such as on boards of directors and supervisory boards (Schuh et al., 2014), as part of corporate governance, gender parity as well as gender diversity is addressed. Various regulatory options such as quotas for women, minimum shareholdings, target size obligations or diversity recommendations, of which some have already been codified, attempt to counteract these challenges (Rentsch, 2022). In a field experiment, Hoogendorn et al. (2013) used small teams consisting of business students to show that teams with a gender mix perform better than teams which predominantly consist of men and women respectively. By using business game data as well, Apesteguia et al. (2012) show that teams solely consisting of women perform worse than mixed teams or teams exclusively consisting of men. While the aforementioned studies indicate that teams consisting predominantly of women perform the poorest with regard to team performance, some studies (Campbell and Mínguez-Vera, 2008; Joecks et al., 2013; Wooley et al., 2010) demonstrate that more women in a team lead to a better team performance. Despite the contradictory evidence, it is clear from the findings that mixed teams or a more balanced proportion of women and men in teams (diversity) are positively associated with team performance. With the additional consideration that some of the cited studies also analyzed student teams and consequently a similar context, we also assume a positive impact of gender diversity on team performance:

H2.

Gender diversity is positively related to team performance.

Method

Description of the final sample (participants)

In total, 128 student teams, consisting of 645 individuals, were part of our data set. The students were enrolled at the Faculty of Business Administration at a German University. Student teams consisted of four to six members with an average of 5.03 (SD = 0.63) members. The average age was 23.19 (SD = 2.66) and 52% of the sample were female.

Business simulation game

The business simulation game was carried out at a German university within the scope of a student seminar. A game was played once a semester and was held as a block seminar all day every day for one week. Approximately 5–7 (fictitious) periods were played per game, depending on the semester and the faculty requirements. On average, 32 groups took part in each game with each seminar leader or team of seminar leaders supervising around five groups at the same time. In the business game, student teams represented a fictitious manufacturing company. Over several (fictious) periods, they had to implement the company’s goals and strategy, compete against rival companies as well as face and counter market changes and their complex interactions. At the beginning of each period played, the economic and financial situation as well as the strategy of one’s own company was analyzed. The teams received a market research report and an economic forecast. Based on this information, all teams made operational and strategic decisions for their company. In the course of the game, students had to make an increasing amount of decisions as the complexity rose. The main goal of any team was to maximize its profits. Although the setting was virtual, the game aimed to reflect the real-world business reality as closely as possible. For example, it considered exogenous shocks and as the simulation was also designed for the long term, the periods represented fiscal years, which in turn were affected by previous periods and also had an impact on subsequent periods.

Design and procedure

We used data from the computer-based business simulation. The data of the original values of the dependent variables were based on a cloud version of the business simulation game described above. Here, the values calculated by the simulation software resulted from the decisions made by the teams in the respective periods. During the entire simulation, one or more seminar leaders supervised the students who were randomly assigned to teams before the start of the simulation [1].

All teams received the same instructions from the seminar leader, orally, for example, in the form of an introductory session and, in writing, in the form of manuals and scripts. There was no supervisor within the group, helpful advice and support came much more from the seminar leaders. Decisions for each period were made jointly in the group in the business areas of sales, research and development, purchasing, manufacturing, human resources and finance and accounting. An example of a decision is how many sales employees the company would like to hire in the upcoming period, which consequently influences sales, but also personnel costs. However, each student was responsible for one of these business sectors. This means that even if decisions were made jointly, the main task of each student was to analyze the data and the situation in the respective area for which they were responsible and to process and consolidate the information as a basis for joint decisions. There was also no compensation for participation, but the module grade of the mandatory seminar was tied to the success of the student groups, so that all students had a clear incentive to perform as best as possible. The collected data of the student groups refer to a period of four semesters, that is, two years but one simulation game (seminar) only refers to one semester. The diversity variables were consolidated as part of the seminar held at the university over the four terms. Hence, at the end of the four semesters, the collected data from the econometric model were recoded accordingly. The authors of this study were present as seminar leaders as well. All data was compiled at the group level.

Measures and descriptive statistics

Independent variables.

The age of the students at the time of the business simulation and the gender of all students were used as the basis for the operationalization of the diversity variables. This data was provided by the faculty. Based on diversity literature (Campbell and Mínguez-Vera, 2008; Lee-Kuen et al., 2017), this paper addresses gender diversity by the Blau index. By using the Blau index, the number of gender categories and the distribution of group members across these categories were taken into account. Thus, it can be analyzed what effect an increase in this index, and consequently, an increase in gender diversity has on team performance (Campbell and Mínguez-Vera, 2008; Lee-Kuen et al., 2017). The Blau index (∅ 0.42, Table 1) is calculated as follows where p refers to the proportion of group members in ith category and k indicates the number of categories for an attribute of interest (Prinz and Wicker, 2016; Solanas et al., 2012) [2]:

1-∑i=1kPi2

Age diversity is represented by the coefficient of variation of age as a measure of spread or dispersion which can be calculated by dividing the standard deviation of the mean of the age of all team members by the mean of the age of all team members. The higher the coefficient of variation of age, the more diverse the team is regarding age (Prinz and Wicker, 2016). One could assume that student age differs only slightly. However, on average, the individual age differs from the average age of all students (23.19) by 2.66 (standard deviation) and, thus, by about 3 years, which in Germany represents six semesters and, hence, an entire undergraduate degree program in business administration at universities. For this reason, we argue that there is sufficient variance in the age of students to represent age diversity. The average coefficient of variation of age is 0.10, which means that on average the standard deviation of the individual ages is 10% of the average age.

Dependent variables.

Measuring performance is one key to success of organizations (Austin, 2013). Literature analysis shows no consensus on how performance is measured the best (Carton and Hofer, 2006). After analyzing 138 articles on “overall organizational performance,” Carton and Hofer (2006) show that there is no right or wrong decision regarding the choice of one or more performance measure(s). However, following Hoch and Seyberth (2021) based on established literature, we operationalize team performance using different measures to account for not only market valuation but also profitability. As often used in literature to measure the market performance of organizations (Post and Byron, 2015; Zhang, 2020), the market valuation of the companies is represented by Tobin’s Q as an indicator how the market views the long-term value of a company (Zhang, 2020). It is calculated by dividing the total market value of the firm by the total asset value of the firm (Lewellen and Badrinath, 1997). For profitability, we use a key figure that, to our knowledge, has not yet been used to measure team performance. In this study, we use the cumulative net income to measure team performance, as it represents the overall performance of the teams and, consequently, represents the entire performance timeframe (the net income from each period accumulated over all periods played). Hence, this key figure serves as a longer-term performance measure as well. Because of the skewed distribution of the cumulative net income and the fact that the absolute numerical values in millions of euros for a business simulation game tend to have less explanatory power with regard to the relations of these numbers, we use the natural logarithmic value (ln) of the cumulative net income. Alongside Tobin’s Q, it serves as a dependent variable (Wooldridge, 2013). On average, the natural logarithm of the cumulative net income is 4.79, while the average value of Tobin’s Q is 0.98.

Because of the metric scale level of the dependent variables, multiple linear regression models (OLS) are used to analyze the impact of diversity on team performance (Backhaus et al., 2018). However, as the cumulative net income in its original form contains negative numerical values, direct logarithmization makes little sense. For this reason, a constant (c) was added to all values of this dependent variable, where c equals the sum of the largest negative value of the dependent variable + 1 to calculate the natural logarithmic value (ln) of the cumulative net income with only positive values of the dependent variable (Wicklin, 2011). This value can be transformed so that this dependent variable now represents the natural logarithmic value (ln) of the cumulative net income:

ln(Cumulative net income+c)=β0+β1x1+…+βixi+ε

Results

The results of the multiple linear regression models (OLS), which analyzes the impact of diversity on team performance, are presented in Table 2 [3]. Regarding H1, we assumed that team performance will increase with an increase of age diversity. Although the coefficients have the expected signs, they can only marginally be interpreted as significant (^†p < 0.1). Nevertheless, the direction of the hypothesis with respect to both the impact on profitability (p-value: 0.079) and the market valuation (p-value: 0.082) is conclusive [4]. Analogous to the first hypothesis, we assumed in H2 that team performance increases with increasing gender diversity. As both coefficients are not significant, the second hypothesis must be rejected.

Discussion and implications

At least in our setup, the standard deviation of approximately three years, which represents the duration of a bachelor’s degree in Germany, seems to be sufficient to confirm an effect of age diversity on team performance, which led to a confirmation of H1. For example, if team members complete the simulation game quite early in their studies, then on average, they also have members in the team who complete the game at the end of their studies and, thus, have significantly more experience, theoretically and probably professionally as well. In return, the younger team members may have more “open-minded” ideas and approaches. This is in line with the information and decision-making approach and could lead to the positive performance effects. With regard to the rejection of H2 and against the background of the social categorization theory and the similarity attraction approach, teams that are diverse in terms of gender tend to perform less effective. Compared to conventional didactic learning approaches, business simulations enable economic theory to be put into practice and ensure that the participants in the simulation (e.g. students) develop skills such as time management, teamwork and communication (Brown et al., 2020). We believe, that the findings of the simulation game can be applied to teams in general, but especially to start-ups. The purpose of start-ups is to develop innovative ideas and, in combination with management skills, to make them business-ready. We observe comparable teams, which are often diverse in terms of age and gender. As mentioned previously and with respect to the diversity variables, we consider one advantage of our study to be the random assignment of students into groups and, thus, take into account the endogeneity and self-selection bias of team formation (Hoogendorn et al., 2013). This allows us to compare whether randomly assigning students to teams produces different findings with regard to the diversity–performance relationship to, for example, similar studies that did not account for the random assignment like Apesteguia et al. (2012). To our surprise, despite the different study settings, the tendencies are relatively similar. Implications for higher education can also be derived from these results. Rather than letting the groups form themselves or making a purely random distribution of students into groups, professors, lecturers and, in this case, seminar leaders should compose the student teams.

Limitations and future research

With regard to the socio-cultural background of the students, we basically consider a relatively homogenous pool of participants. This is rarely the case in the real economy. In particular, with respect to age diversity and the average age of around 23 years, the comparison with, for example, employees in companies is not easy at the first instance. In this regard, we argue to not only analyze student teams in principle, but also transfer the data to a broader type of work teams because of the framework of the business simulation. However, it must be acknowledged in this context that, even if we assume enough age variance within the student teams with a standard deviation of around three years, which represents an undergraduate degree program in business administration, this will probably be significantly greater in the real economy, in which, for example, employees represent different generations, from “Baby Boomers” to “Generation Z.” The importance of three years at university compared to three years in professional life must also be taken into account. In this regard, even though the major advantage of this work is that we were able to collect data that cannot be collected in the real economy or only with considerable effort, it would be very interesting in the future if the findings of this work were tested under real conditions in a company, for example, within the scope of a collaboration. Until then, the question will remain whether the findings of a business simulation game can be transferred to the real economy. Moreover, the simulation data come from only one university. However, to present more representative results that may take into account different levels of performance across universities, it would be desirable in future research to additionally analyze data from other universities that have offered the same simulation game and match it with the data from this study to present more meaningful results.

Conclusion

The aim of this paper is to determine the effects of different diversity criteria (in terms of age and gender) in student teams on team performance using a business simulation game, where performance is operationalized by both market valuation (Tobin’s Q) and profitability (cumulative net income). So far, the existing literature does not provide clear results regarding the diversity-performance relationship. Besides trying to gain more understanding in this regard, this study adds value by analyzing the impact of different diversity factors on team performance in a controlled but realistic setting with considering a randomized group assignment and using objective performance indicators. Because of the setting of a simulated business game, the results could also be applied to the real economy where we observe working teams every day. Thus, besides the scientific relevance, general implications for the real economy and its established companies can be derived from the results and, given that the study essentially analyzes data from a business simulation game for students, for example, startups and their often still relatively young founders and employees can probably also benefit from the study results, as they are probably still partly in a similar age group as students. For the given setting, we found a statistically significant positive impact of age diversity on team performance, whereas no significant effect was observed for gender diversity.

Table 1.

Data and descriptive statistics

	Data of the business simulation game (n = 128 groups)
	Mean	SD	Min.	Max.	Scale
Dependent variables (team performance)
ln cumulative net income	4.79	0.62	0	6.45	Metric
Tobin’s Q	0.98	0.45	0	1.80	Metric
Independent variables (diversity)
Blau index (gender diversity)	0.42	0.12	0	0.5	Metric
Age variation coefficient (age diversity)	0.10	0.05	0.02	0.32	Metric

Source: Table by author

Table 2.

Multiple linear regression models regression models of ln cumulative net income and Tobin’s Q

	(OLS 1)	(OLS 2)
Variables	Team performance ln cumulative net income	Team performance Tobin’s Q
Diversity
Gender diversity	−0.082 (0.335)	−0.005 (0.371)
Age diversity	1.966† (1.111)	1.259† (0.718)
Constant	4.636*** (0.156)	0.856*** (0.183)
Observations	128	128
R-squared	0.026	0.020

Notes:

The coefficients in the OLS 1 model are presented based on the logarithm scale. Robust standard errors in parentheses; ^†p < 0.1; *p < 0.05; **p < 0.01; ***p < 0.001

Source: Table by author

Notes

1. Based on the student enrolments for the respective seminar/semester, the faculty administration determined a number of seminar leaders for the semester who supervised the student teams during the game. The enrolled students were randomly allocated to their teams by the faculty administration. This means that the groups were put together without a system or structure. Students, therefore, had no influence on who they were in a team with.

2. In addition, the Shannon Index was also calculated to operationalize gender diversity. However, because of a lack of deviation to the coefficients of the Blau Index, no new model was created to reduce complexity.

3. We would like to comply with the feedback and comments of the reviewers and only show the impact of the independent variables, as only these are derived within the theoretical framework. Models that take control variables into account will be provided if requested.

4. As the coefficient of variation of age and the average team age are correlated, leading to spurious effects, an additional estimation including the average team age was calculated. The results show unchanged slopes, but insignificant parameters (OLS 1 coefficient of variation of age: p-value: 0.188; OLS 2 coefficient of variation of age: p-value: 0.115).