Gender and tenure diversity in GitHub teams19 Jan 2015
Diversity in teams arises from any attribute that people use to differentiate themselves from others. The most obvious attributes are arguably demographic (age, gender, culture, ethnicity), but they can also be related to pretty much everything else (for example, role, tenure, expertise, or even personality). In general, diversity in teams is viewed as a double-edged sword. On the one hand, increased team diversity results in more varied backgrounds and ideas, which provide the team with access to broader information, enhanced creativity, adaptability, and problem solving skills. On the other hand, due to greater perceived differences in values, norms, and communication styles in more diverse teams, members become more likely to engage in stereotyping, cliquishness, and conflict.
Diversity in teams has been studied for a long time in offline groups, but different studies still disagree on the effects. Instead, we focused on distributed (online) software teams, such as those in Open Source Software (OSS). OSS teams are much more fluid, therefore much less tangible, than their offline counterparts. In OSS teams are naturally very diverse, consisting of contributors from all over the world, typically a mixture of volunteers and professionals, coming from varied cultural and educational backgrounds, with different interests and skills.
We focused on two diversity attributes that are prominent in OSS: gender and tenure (experience). Women are underrepresented in programming, and especially so in OSS. Moreover, the "hacker" culture is said to be male-dominated and unfriendly to women, with reports of active discrimination and sexism. OSS teams are inherently diverse with respect to experience, since they often rely on a steady influx of new contributors.
Then, we carefully extracted data from more than 23,000 active collaborative projects on GitHub, the largest and most popular online collaborative coding platform. Each observation in this data set contains the composition, characteristics, and outcomes of a project's team of contributors for each quarter (90-day period) in the evolution of the project. Using regression analysis on this data, we modeled:
- productivity (the number of commits by team developers recorded in either the main repository or any of its forks in a given quarter), and
- turnover (the fraction of the team in a given quarter that is different with respect to previous quarter)
as functions of gender diversity (measured using the Blau diversity index) and tenure diversity (measured using the coefficient of variation). We controlled for many confounds:
- team size (larger teams are likely associated with increased productivity and increased turnover);
- overall project activity (total number of commits);
- project forks (activity in forks is more likely to be limited, both in time and in amount, compared to a project's main repository);
- time (has a moderating influence on the effects of diversity: as￼ group members continue to interact, they have more time to adjust to differences between them);
- project age (projects that started later and their teams may have experienced a different GitHub culture);
- tenure median (to distinguish between more and less experienced teams, regardless of how diverse these teams are with respect to tenure);
- comments (number of GitHub comments during a given quarter on commits, pull requests, and issues, as a reflection of a project's social activity).
Our models show that both gender and tenure diversity are positive and significant predictors of productivity, together explaining a small but significant fraction of the data variability.
The paper presenting these results (co-authored by Bogdan Vasilescu, Daryl Posnett, Baishakhi Ray, Mark van den Brand, Alexander Serebrenik, Prem Devanbu, and Vladimir Filkov) has been accepted for presentation at the 2015 ACM CHI Conference on Human Factors in Computing Systems, in Seoul, South Korea, in April 2015. A preprint containing more details is available here.
This is the first academic study to consider effects of gender diversity on productivity and turnover in OSS communities. On a larger, economic and societal scale, these findings suggest that added investments in educational and professional training efforts and outreach for female programmers will likely result in added overall value.