In a lot of ways, it doesn't make sense that women are disproportionately left out of the STEM workforce. Consider the following: Only about one in seven engineers is a woman, yet women earn 50 percent of all science and engineering bachelor's degrees. In high school, women take math and science classes in similar numbers as men. And historically, throughout all levels of education, girls perform the same or better than boys in math and science.
So what counts for the extreme lack of gender parity? A large part of that might be discrimination in the workforce.
Researchers at Columbia Business School recently published a paper that shows a clear bias against women in mathematics, but also suggests the variables that can lessen the impact of that bias.
"In our setting" the authors write, "women were only half as likely to be hired as men, because they were (erroneously) perceived as less talented" on a math test. "Both men and women expected women to perform worse."
Here's how they found it out.
The researchers had study participants take a math test, and then randomly chose two of the participants to play the role of a job "candidate." The rest of the participants then acted as "employers." From there, the task was simple: The employers were asked to choose the candidate who would be the best for future math tests. If they chose correctly—that is, actually chose the person with the better score—they'd receive increased compensation for the study.
Outright, with no more information than gender, the employers were more than twice as likely to choose a male.
This effect was not easily weakened with more information. "When we allowed candidates to self-report their [expected test] performance," as one might do in a real-world interview, "women were chosen at equally low rates," the authors found. The only time the gap in hiring was reduced "but not eliminated" was when the employers were provided with the candidates' actual scores on the math test. But even in that condition, females were only chosen 43 percent of the time. Keep in mind, the women in the study performed no better or worse on the math test than the men.
And it didn't matter if the employer was male or female—the male hiring bias persisted all the same.
Sure, this was a contrived lab scenario (these were not actual job seekers or people who work as hiring managers). But the laboratory setting means that the researchers can come to a causal conclusions, and dial up or down the variables to see how the outcomes change. At the very least, the study indicates research participants discriminate against their female peers when it comes to math abilities.
It also adds to a body of research that finds a persistent and malignant bias against women in science. In 2012, Yale produced a paper that asked actual scientists, both male and female, from multiple universities to rate an application for a lab-manager position. Everything about the applications was the same aside for the name on top—half of the scientists reviewed a male candidate (John) and the other half, a female (Jennifer). The scientists rated the female application as being less competent, less hireable. The scientists even indicated they were less willing to mentor the female applications.
Furthermore—and perhaps most importantly—when the scientists were asked to state a starting salary for the applicants, they (hypothetically) offered $4,000 less to the females. Like the current experiment, both male and females presented this bias.
This bias is a huge hurdle for the sciences, but the current research also gives clues on how to erase the hiring bias. The more objective information for qualified female candidates, the better. Self-reported information doesn't help as much, the researchers find, because "men tend to be more self-promoting than women in these reports," the author's write, "but employers ... do not fully appreciate the extent of this difference." Objective information is better. It's harder to turn down a woman for a job when she presents a near-perfect transcript. But even then, as these studies show, there are still hurdles to overcome.