Math-intensive fields have a gender problem: The men are worse at math

Educators and scientists alike are trying to increase the representation of women (and especially women of color) in the male-dominated, math-intensive fields of physics, engineering, and computer science (PECS). These fields have some of the worst gender imbalances of all college majors, with about four men for every one woman.

To understand how to improve female representation in PECS, we first need to understand the nature of the gender imbalance. This post describes findings from a recent study I did with Taek Kim and Zachary McDermott, in which we used nationally representative data from a cohort of U.S. students that was in ninth grade during the 2009-2010 school year and followed into their later years at four-year colleges. I focus on four key findings.

The gender ratio in PECS subjects is largest among low-achievers, where more men enter these fields

First, we need to understand if this imbalance is the same at all achievement levels. With few exceptions, research and policy around gender and racial imbalances tend to focus on average patterns, such as the 4-to-1 male-to-female ratio in PECS on average. But these average ratios could conceal different relative representations at different achievement levels. If the relative concentrations change, then we need to question why they change as they do, as well as what this means for interventions and evaluating equity, which I’ll discuss later.

In a recent paper published in Science, we found that the 4-to-1 male-to-female ratio among PECS majors is not the same throughout achievement levels. In general, men were much more likely than women to be in these math-intensive majors despite poor STEM performance—similar to emerging patterns seen in other STEM fields and economics. If we look at college students who were among the poorest STEM achievers during high school (i.e., the bottom decile of high-school STEM achievement), the male-to-female ratio of students majoring in PECS is 10-to-1. If we look among the highest STEM achievers, that ratio is less than 2-to-1. In fact, about 10% of men scoring below the 1^st percentile of STEM achievement were in these majors. To find 10% of women choosing these majors, we need to look all the way out to the 80^th percentile (see Figure 1).

These trends make clear that the labor market demand for undergraduates with computer science and engineering degrees could be met by women, as many women who are highly performing in STEM are not choosing these fields despite many low-STEM-performing men choosing them. In other words, not only could these high-demand areas meet that demand with STEM-competent people, but they could improve the gender diversity of the fields in the process.

The higher rates of low-achieving men in these fields cannot be explained by many student factors

Next, we need to know if we can explain the gender imbalance by other student attributes. Previous research indicates that many student-level factors contribute to the gender gap in PECS (and STEM more broadly)—factors such as differences in confidence in STEM, women having a comparative advantage in English, women preferring majors that are perceived to have a larger direct social impact, and gender differences in earlier career aspirations.

In our study, these factors did matter, but they mattered most for students nearer the top of the distribution. All of these factors—as well as some others—collectively explained about 99% of the gender gap among the highest-achieving third of students. This suggests that earlier exposure to PECS courses and careers, boosting confidence, extracurricular activities like Girls Who Code, and role-model interventions may help close the PECS-majoring gap … for high-achieving students.

However, the further down the achievement distribution we looked, the less our models could predict the gender gap in PECS. Among the lowest-achieving third of students, these same factors reduced the gender gap by only 34%. In other words, low-achieving men were unexplainably majoring in PECS at higher rates than observably similar women, a finding made more perplexing by comparison to the patterns among high-achievers.

Previous studies suggest that PECS fields are more welcoming to men by creating and feeding off of masculine cultures and stronger beliefs that innate ability is required for success in these fields. Our data did not contain measures of the climates in these fields. It would be reasonable for future studies to look beyond students—to the fields themselves, and broader society—to understand why a man with an SAT math score of 350 (on a 200-800 scale) and high-school STEM GPA of 1.5 is as likely to major in PECS as a woman with a 600 SAT math score and high-school STEM GPA of 3.5.

Black, Latina, and Indigenous women choose these majors as often as comparably achieving white and Asian women

Black, Latina, and Indigenous women are less likely to major in PECS than white and Asian women. Yet, we found that Black, Latina, and Indigenous women were majoring in these fields as often as white and Asian women at the same level of STEM achievement, and high-achieving Black, Latina, and Indigenous women with expressed interest in these fields prior to college were persisting at the same rates as comparable white and Asian women. What can explain these seemingly conflicting patterns?

Although we see Black, Latina, and Indigenous women choose and persist in PECS as often as comparably achieving white and Asian women, the hurdles they face before college often prevent them from realizing their STEM potential. This leads them to appear lower in the achievement distribution, where all women are less likely to choose these majors. Factors such as segregation in education, less content coverage, and biases against girls of color create obstacles for Black, Latina, and Indigenous girls to achieve at their potential in STEM. This means that to increase the representation of Black, Latina, and Indigenous women in PECS, we need to equalize earlier educational and extracurricular opportunities.

We must consider STEM competence when evaluating progress in the STEM pipeline

Finally, when evaluating equity and progress in the STEM pipeline, these patterns remind us that we need to consider what students are bringing with them in terms of their STEM performance. Though fewer in numbers, women in PECS tend to be higher performing than their male peers. Women are much more concentrated among the highest achievers in these majors, which implies that we would expect to see women’s representation increase from undergraduate to graduate school in PECS if graduate PECS programs select from the top of the achievement distribution. Yet, we don’t see women’s representation rise. It stays flat.

The fact that roughly the same proportion of women are in graduate school in these fields as in undergraduate is not an accomplishment. Rather, it is a failure of the system that PECS fields are differentially losing more competent women in exchange for less competent men.