Diversity and Performance in Entrepreneurial Teams

Diversity and inclusion have become an increasingly central issue in many workplaces, but what are the performance implications of diversity-promoting efforts for firms and workers?

In recent years, policymakers have proposed and implemented gender (or race/ethnicity) quotas in many settings. For instance, the Norwegian government enforced a gender quota on corporate boards. More recently, California has become the first state in the U.S. to mandate racially or otherwise diverse directors. In principle, a diverse team can perform better because of its members’ broader perspectives and knowledge base. Still, at the same time, differences in backgrounds may hinder communication or trust, thus lowering overall productivity.

In this study, we examine the role of diversity and performance in an entrepreneurial setting by examining a unique setting where Harvard Business School MBA students proposed and worked on new business ventures as a team. We have three main findings: First, when students are free to choose their teammates, we find strong tendencies, up to 25% more likely, to select someone who shares the same gender or race/ethnicity. Second, when students are randomly assigned to teams with balanced demographic characteristics by a computer algorithm, we find that more diverse teams performed worse than homogeneous teams. Nonetheless, when students form teams voluntarily, diverse teams performed just as well as homogenous teams. Lastly, we find that when a faculty advisor is a woman, teams with more women performed significantly better. We do not find any differential performance effect on teams when a faculty advisor is a man.

Here is the setting: First-year MBA students from 2012 through 2016 were required to take a field course in their first year. Throughout the course, students were required to design and launch a real business. At the beginning of the semester, students formed teams of 5-7 people with their classmates. At the end of the semester, a faculty advisor and a panel of judges from venture capital firms rank all the projects based on the quality of the ideas and teams’ performance. When the field course was first introduced to the students in 2012, the school randomly assigned each student to the teams based on a computer algorithm. After 2013, the school changed the team formation policy and started to have students choose teammates themselves. The school did not impose any restrictions on how students formed their teams.

We first estimate the relative economic magnitudes of homophily in race/ethnicity, gender, education, and work experience. We find that individuals are 25% more likely to form groups with people of the same race/ethnicity or gender relative to randomly matching. Selection based upon education history and work experience is weaker than endowed demographic attributes, but it is still economically significant. Having gone to the same college or worked at the same company increases the probability of team formation by 17% and 11%, respectively. Interestingly, we also find team selection effects of shared education and work experience are stronger among male students than female students.

Next, we examine the effect of horizontal diversity (i.e., diversity within a team) on team performance. Because teams were randomly assigned by a computer algorithm in 2012, we can estimate the causal impact of diversity on performance for this cohort of students. We find exogenously diverse teams performed worse than exogenously formed homogeneous teams. Specifically, a one standard deviation increase in race/ethnicity diversity decreases performance by as much as 16%. Further, this negative effect of diversity is driven by the joint homogeneity of both gender and race/ethnicity. In other words, teams with homogeneity in terms of both gender and race/ethnicity performed the best. We hypothesize that randomly assigned diversity hampers communication efficiency and increases the probability of conflict within the team. Strikingly, the negative effect between gender-race/ethnicity diversity and performance is eliminated when teams are formed organically among the 2014-2016 cohorts. In other words, when teams are free to choose their teammates, diverse teams perform just as well as homogenous teams. These results suggest that mandated diversity may not always bring about performance benefits because biases against certain groups could remain.

Lastly, we look at the performance effects of diversity in the mentorship relationship between faculty advisors and team members (i.e., vertical diversity). Because the gender of the faculty advisor for the section is exogenous, we can also infer the causal impact of student-faculty tie on performance. We find a significant effect of gender ties for teams with more female students. When the faculty advisor is a woman, teams with more female students performed better. Conditional on having a female faculty advisor, the probability of a team being in the top three in the section increases from 8.3% to 38.1% when the number of female team members increases from the lowest quantile to the highest quantile. When the faculty advisor is a man, there are no differential impacts on performance based on the gender makeup of the team. These results are consistent with the importance of mentorship, and the positive performance impact women may experience when mentored by women.

Our results on the diversity and its performance implications are important beyond the context of our research setting. First, the main criteria for evaluating these start-ups were related to the actual business concept and the ability to attract real customers. Second, a significant share of these start-up businesses continued to operate after the semester ended, and many proceeded to raise external funding, including from venture capital firms. Third, over 30% of 3,864 MBA students in our sample have gone on to work in venture capital or technology-related areas after graduation, representing a sizable labor inflow to the entrepreneurial ecosystem.

Our results shed important light on the lack of diversity in entrepreneurship. We also call for more careful policy interventions in addressing inclusivity in workplaces. Forced diversity, such as gender or racial quotas, may not bring the performance benefits that diverse teams engender, especially when such policies do not consider other relevant factors or when biases persist within a purportedly diverse team.

The complete paper is available for download here.