Big Data Shows Women Engineers Downplay Coding Skills

Recruiters may overlook women engineers who do not list their programming experience

Group of people, one women and four men, sitting in a line waiting for an interview
Photo: iStock Photo

Many would like to answer a key question bedeviling Silicon Valley: How can tech giants and startups recruit more women engineers? An online recruiting platform’s data reveals that recruiters may overlook some women engineers because the latter are less likely than similarly-qualified men to list their proven knowledge of programming language skills on their profiles.

The preprint study by Raviv Murciano-Goroff, a Ph.D. candidate in economics at Stanford University, shows how women engineers with experience contributing to open-source software websites are about 10 percent less likely than men to self-report their coding skills in job candidate profiles. That matters because the study also found that recruiters are 22 percent more likely to click a button to “save” candidate profiles if candidates actively list specific technical skills—even if recruiters can already see objective evidence of candidates possessing those technical skills.

“It is not clear if this finding is particular to how the recruiters on this platform search for candidates,” Murciano-Goroff says. “In the future, I would like to investigate if other recruiters at various different types of employers also show a similar interest in self-reported skills even when presented with evidence of candidates’ actual previous experience and abilities.”

Overall, the study—which has not yet been submitted for peer review by a journal—shows that profiles for women receive about 12.37 percent less attention from recruiters than the profiles of men. This trend holds even among experienced coders who both self-report programming skills and have those skills independently confirmed as “Verified – High Experience” by the recruiting platform’s algorithms: Recruiters were about 9.49 percent less likely to save the profiles of women compared to men among those who qualified as highly experienced JavaScript coders.

Such observational studies cannot shed much light on the motivations behind why job candidates and recruiters made certain choices. But in his paper, Murciano-Goroff was still able to highlight several intriguing trends in the data—consisting of almost 4 million job candidate profiles—provided by an unnamed recruitment platform popular with Silicon Valley companies. The platform allows candidates to self-report technical skills and also verifies certain technical skills based on either open-source coding contributions or answers to questions on the website StackOverflow.

Murciano-Goroff also gauged the programming experience of certain candidates based on their contributions to open-source software websites: places where he could see how many lines of code were uploaded in specific programming languages and the popularity of each individual’s contributions among the open-source community.

By cross-referencing the recruiting platform profiles with their respective open-source contributions, Murciano-Goroff found the disparity in self-reporting among women and men engineers with apparently similar experience—based on their open-source contributions—in programming languages such as JavaScript, Ruby, C#, PHP, and Python.

Take this all with a few grains of salt: only 182,000 profiles from the online recruiting platform had corresponding open-source contributions. This subset was further skewed by the fact that women accounted for just seven percent of these profiles. Women also made open-source contributions in a smaller number of programming languages, and the average number of programming languages appearing in the self-reported skills section of profiles for women was lower than the average number for men.

Still, this self-reporting gap’s existence among candidates with a history of open-source contributions remains intriguing. A 2017 study by Josh Terrell, a software development engineer at Amazon who conducted the research while at California Polytechnic State University, showed that the open-source community of GitHub accepted women’s contributions at a higher rate than they accepted men’s contributions—but only if the women were not identifiable as women. The finding of gender bias in the open-source community suggests that women who persist despite that may be especially motivated.

“If that means that the women coders who do participate in open source are particularly committed to being involved in coding despite the gender imbalance and potential for discrimination, this would mean the lower level of self-reporting that I find is even more surprising,” Murciano-Goroff says.

Recruiters might knowingly compensate for the underreporting tendency by being more likely to save the profiles of women who possess specific technical skills based on the recruiting platform’s verification and open-source contributions. But the analysis found no evidence that recruiters on this particular platform have been pursuing such a strategy.

This was backed up anecdotally by Murciano-Goroff’s conversations with many recruiters and human resource managers: None seemed to be aware of the possibility that men and women might display different approaches to self promotion in online job candidate profiles.

One explanation for the difference in self-reporting comes from researchers such as Shelley Correll, a sociologist at Stanford University, whose work has shown how young women can experience a confidence gap in comparison with men when assessing their own skills. Work by Sylvia Beyer, a psychologist at the University of Wisconsin-Parkside, also found gender gap related to self-confidence in skills among computer science students.

But the latest study cannot definitively rule out other possible explanations. For example, women engineers may be intentionally leaving certain languages off their self-reported list if they have less interest in jobs requiring them to use that specific programming language.

On the other end, it’s very possible that recruiters are not only screening for candidates with the most experience in specific programming languages. Recruiters may also have different objectives such as looking for those candidates likeliest to accept a potential job offer. In that case, it could make sense for them to prioritize candidates who take the additional step of listing specific skills on their profile.

Murciano-Goroff hopes to expand his research beyond the early stages of passive recruiting: the point when recruiters contact individuals who may or may not be searching for new jobs. “Much more research is needed to understand how the later stages of the hiring and recruiting pipelines might be impacting the diversity of the tech workforce,” Murciano-Goroff says.

Still, even such early findings could prove of interest for tech giants such as Google, Facebook, and Twitter that want to boost their employment of women engineers along with creating more welcoming workplace cultures. Those companies had 20 percent, 17 percent and 13 percent of their respective technical staff positions filled by women in 2017.

The Tech Alert Newsletter

Receive latest technology science and technology news & analysis from IEEE Spectrum every Thursday.

About the Tech Talk blog

IEEE Spectrum’s general technology blog, featuring news, analysis, and opinions about engineering, consumer electronics, and technology and society, from the editorial staff and freelance contributors.