From the abstract:
We study the role of homophily in group formation. Using a unique dataset of MBA students, we observe homophily in ethnicity and gender increases the probability of forming teams by 25%. Homophily in education and past working experience increases the probability of forming teams by 17% and 11 % respectively. Homophily in education and working experience is stronger among males than females. Further, we examine the causal impact of homophily on team performance. Homophily in ethnicity increases team performance by lifting teams in bottom quantiles to median performance quantiles, but it does not increase the chance of being top performers. Our findings have implications for understanding the lack of diversity in entrepreneurship and venture capital industry.In other words, teams sharing common bases such as education attainment, experience, gender, or ethnicity increase performance. It would be interesting to untangle whether ethnicity is simply a proxy for cultural attributes.
I have mentioned in the past that I think that much of the Diversity advocacy is essentially just affirmative action under another name. Diversity on its own as a variable has very little predictive power in terms of outcome - and that which it has is very context dependent. Demonstrated past performance and capabilities are the real variables you are interested in.
For example, one would not expect material differentials in outcomes based on the gender composition of a cardiac surgical team or a coding team for an information system, ceteris paribus. But the gender composition of a marketing team for a new consumer product might be predictive of a desirable outcome. Similarly with race, education attainment, class, religion, etc. The value of diversity within the team is entirely contingent on the nature of the task of the team.
If the task is unknown in advance, then you are playing with a clear trade-off. Greater homogeneity based on any variable will likely be associated with an increased capacity for intra-group communication and cooperation. Greater diversity based on any variable will be associated with an increased capacity to tackle a broader domain of issues. Breadth of experience versus team effectiveness is the trade-off and the answer will be based on the nature of the task(s) which the team must tackle, i.e. context.
Or, as they put it in the paper:
Heterogeneous teams benefit from more diverse pools of skill and knowledge, but at the same time, differences in ethnicity, culture, and mother language hinder efficient communication among team members, thus potentially lowering productivity.And that is what they found in their experiment.
We find homogeneity in ethnicity increases team performance by lifting teams from the bottom quantiles to median performance quantile, potentially because it increases communication efficiency and lowers the probability of conflict within the team. However, homogeneity does not increase the chance of being top performers. We do not find homogeneity in gender, education, or past work experience is an important factor in determining team performance.Improved communication has a material impact on team effectiveness up to a given point. Top tier performance is dependent on more than simply good communication.
The other findings, beyond the role of effective communication, include:
First, we estimate the relative economics magnitudes of homophily in ethnicity, gender, education and working experience. The first central question we address is: what are the strongest homophilic forces in forming social networks? Using a novel dataset of HBS MBA students, we find ethnicity and gender are the two strongest homophilic forces in social networks. Individuals are 25% more likely to form groups with people of the same ethnicity or gender relative to randomly matching within a set of students who choose the groups that they work with on real microbusinesses. Homophily in education and working experience is weaker than demographic homophily, but they are still economically significant. School ties and shared working experience increase the probability of forming social networks by 17% and 11% respectively. Further, we find homophily in education and working experience is stronger among males than females.In sum:
Third, our unique dataset allows us to explore the dynamics of entrepreneurial team formation. Gompers and Wang (2017) document the homogeneity in gender and ethnicity in US start-up teams. Female and non-Asian minorities have been underrepresented in the innovation sector for the past 20 years and the progress to achieve diversity has been slow. One possible explanation is the biases of hiring people with similar background. Despite large volumes of research on homophily in various settings, only a few studies have explored the effect of homophily in entrepreneurship. Ruef, Ruef, and Carter (2003) survey 830 entrepreneurs on their founding team composition. They find that the probability of a team with the same gender or with the same ethnicity is higher than a random matching process would predict. In our setting, we observe MBA students tend to form entrepreneurial teams with people who have similar social and demographic backgrounds. Given a significant portion of students will be working at start-ups and venture capital firms after graduation, our results have implications for understanding start-up team diversity, recruitment process in start-ups and venture capital firms, and deal selection in venture capital.
Homogenous teams outperform heterogenous teams, with homogeneity in ethnicity, gender, education and experience being the most important in declining order.All of this has to be taken with more than a pinch of salt.
The value of homogeneity operates through the mechanism of effective communication.
Given the choice, teams self-select based on homogeneity.
The subjects are the 3,600 Harvard MBA students across four years organized in 630 teams. Harvard MBA students are typically in the 1% of IQ and are also in the 1% of familial income. They are not representative of the population at large. Further, Social Class is a critical variable not addressed in this study which is unfortunate as I suspect it is probably the most influential variable in the diversity spectrum. In many respects, a 1% Catholic Frenchwoman from a Grande Ecole will have more in common with a 1% Protestant American man from Stanford than either will have with either a French or American High School graduate.
This paper is an interesting contribution to a relevant subject but the issues are far more complex than the ideological "diversity is our strength" would suggest. Some types of diversity are indeed our strength but only under some particular circumstances. What I found interesting was the empirical support for the notion that the crux of the issue is communication effectiveness.