Diversity in Data Science: Does Gender Bias Exist?
23rd December 2019
Historically, the labour force has been a bit of a boys’ club. In the United Kingdom, for instance, it wasn’t until the late 1980s that the percentage of working women reached over 60 per cent. But even today, less than half of all women globally participate in the workforce, an amount significantly less than the 75 per cent figure of their male counterparts.
And the technology sector is no different. Women hold just one-quarter of computing jobs. And more shockingly, they represent a meagre 13 per cent of leadership roles in technology companies.
For one reason or another, there’s a glaring gap in the technology sector between where women currently are and where they should be.
Unfortunately, data science shows a clear gender bias, as well.
According to 365 Data Science, men make up 69 per cent of the data science workforce, meaning women represent just 31 per cent. So while this percentage is higher than the technology industry average, there’s still plenty of room for improvement.
There are numerous potential causes for this disparity.
As you can imagine, there isn’t just one issue causing the gender gap in data science roles. Instead, it’s the combination of inherent biases, hiring processes, and systematic problems. Let’s take a look at some of the most impactful causes.
Overall industry bias prevents women from applying.
Discouragingly, many people today still perceive women in technology fields to be a social stigma. And although their work is generally equal to, if not better than, that of men, they’re commonly seen as inferior workers.
For example, a 2017 study examined nearly five years of open-source Github data to analyse the possibility of gender bias in the acceptance rate of pull requests. In the study, researchers discovered that the pull requests of women who included gender-identifiable information on their Github profiles were 10 per centless likely to be approved than the pull requests of women with non-gendered profiles.
This bias is likely the reason that men holding technology roles in the UK earn, on average, 25 per cent more than women and why 65 per cent of technological women worldwide feel discriminated against in the workplace. With lower relative pay and a more hostile work environment, it’s no wonder that women are shying away from technology and data science professions.
The “motherhood penalty” hinders hiring and leadership development.
Women in data science don’t only face gender bias regarding their abilities. They also experience unfair assumptions about the amount of time they spend at work.
Because women are historically seen as caregivers in the family, many managers assume that mothers (or potential future mothers, for that matter) are less devoted to their work. They buy into the stereotype that women with families need more flexible schedules, a greater amount of time off, and work less productively than men.
This false assumption causes hiring managers, either purposefully or inadvertently, to favour hiring men over women. Additionally, it prevents working mothers from advancing their careers at the pace that they otherwise would have.
A study by the Institute for Fiscal Studies revealed that the “motherhood penalty” causes working mums to see their hourly wage drop to about a third below their men counterparts by the time their child is 20. Unfortunately, this issue expands beyond data science, affecting most other industries as well.
University degree choices reduce the female candidate pool.
Around half of all data scientists have received a university degree in mathematics, engineering, or computer science. However, only about 15 per cent of engineering graduates in the United Kingdom are female. And, representation in computer science is even less, at 13 per cent (as of 2014).
Therefore, we can assume that another component causing the gender gap in data science is a limited pool of potential female applicants from which employers can hire.
Regrettably, the lack of female representation in data science and the small number of women in associated majors feed off one another, creating a negative feedback loop. Female university students don’t have data science mentors or role models to look up to, so they choose other majors. Because female students are choosing other majors, fewer data science role models and mentors emerge. It’s a classic catch-22.
A lack of female mentors and role models are the two most common issues women cite for the gender gap in technology. | Source:Adeva
Unstructured application processes contain inherent human bias.
The most impactful source of gender bias in data science, though, likely resides in the application process for data science roles. From the job description to the final interview, the sheer amount of human factors make gender bias challenging to combat.
A 2011 study in the Journal of Personality and Social Psychology, for instance, found that job descriptions that include masculine wording like competitive, deter more-than-qualified women from applying because they feel as if they wouldn’t fit in that type of environment. Therefore, an already small candidate pool is shrunk further due to careless wording.
Additionally, being able to determine gender from resumes and holding unstructured interviews with applicants encourages the development of the issues we outlined above, consciously or not. Hiring managers generally took to hire someone similar to themselves. So if men are the ones doing the hiring (which is likely the case in data science), then it’s going to be more difficult for women to break the mould and join the field.
How can we reduce gender bias in data science?
Women have been instrumental in the success of the technology sector since the dawn of the first computer. But their achievements don’t stop there. Fortune 500 companies with at least three female directors have increased their return on: capital invested by 66 per cent, sales by 42 per cent, and equity by 53 per cent. By succumbing to gender bias, we only work to impede innovation in all aspects of business and technology, especially data science.
In the second part of this three-part series, we dive further into the inherent gender biases of current hiring processes while outlining ways in which AI could help remove them.