A telling moment at a recent Thanksgiving dinner captured the changing gender landscape of technical education. A first-year engineering student at Cornell University was happily sharing details about her semester project—a computer game she was designing with another female classmate, reminiscent of the 1980s arcade classic Asteroids. She suddenly broke into an even broader smile. “Half of the class is women!” she exclaimed, beaming with pride. “It’s so cool!” Her experience reflects a remarkable decadeslong transformation at institutions like Cornell, where engineering programs are at or approaching gender parity.
This progress exemplifies a broader trend at postsecondary institutions serving students with high math achievement. Over the past two decades, the universities serving students with the highest math achievement (over 750 for an average math SAT score) have made impressive strides in attracting, retaining, and graduating women in physics, engineering, and computer science (PECS) majors. At these institutions, on average, men are still more likely than women to graduate with degrees in these majors. However, the male-to-female relative ratio shrunk from about 2.2:1 in 2002 to 1.5:1 in 2022.
But this progress runs in contrast to a troubling phenomenon at institutions serving students with lower math achievement. In these institutions, the gender representation gap has dramatically widened. Among the institutions serving the lowest math-achieving students (average math SAT around 450), the male-to-female relative ratio climbed from about 3.5:1 in 2002 to 7.1:1 in 2022.
These findings come from an analysis that we recently published in the journal “Science.” It uses a near-census of over 34 million bachelor’s degrees awarded in American colleges and universities to reveal this stark and growing divide. We classify institutions based on the average math SAT scores of their students during the period from 2002 to 2022. We focus on average math SAT scores because, relative to other institutional characteristics such as tuition, admissions rates, student-faculty ratios, or expenditures, this measure was the most predictive factor of the gender gap in PECS.
These findings matter because the institutions struggling the most with gender equity are precisely those serving most American students, particularly those serving most students of color and students from lower-income families. When researchers talk about the “STEM gender gap,” we’re usually talking about research, policies, and programs focused on institutions serving students with the highest math achievement. But by overlooking the widening chasm at other institutions, we’re failing millions of women who could benefit from these high-paying technical careers.
The financial stakes for these women are very real. Even at institutions serving students with lower math achievement, graduates with degrees in PECS command higher salaries than their peers in other fields. Research suggests women may actually receive a larger earnings premium from PECS degrees at less selective institutions compared to more selective ones. In fact, while men tend to see greater financial benefits from PECS degrees at highly selective schools, there is no such gender disparity in the earnings advantage at less-selective institutions. These majors lead to well-paying jobs, and these schools often serve students who need or prefer to stay close to home. The solution isn’t to limit these opportunities for men, but rather to extend them to women who could equally benefit from these career paths.
Making sense of the divergent patterns—and what to do about them
This divergence in trends is clear in our analysis—and reflected in Figure 1, below. What explains the divergence isn’t quite as clear.

The different patterns across institutions serving high scorers and low scorers can’t be chalked up to the usual explanations of men and women having different interests and confidence in math and science, aspiring to different occupations, or having different academic preparation. For example, men scoring in the lowest percentiles of math achievement are as likely to major in PECS as women scoring in the 80th percentile. This isn’t about inherent math ability or men clustering in the upper tail of achievement. Even so, when we accounted for all these (and more) student-level factors, the pattern of bigger gaps at institutions serving students with lower math achievement remained.
This suggests that something is happening at institutions serving students with lower math achievement that is not happening at institutions serving students with higher math achievement—something that is keeping men in PECS majors, but not women.
Our research shows that institutions serving students with lower math achievement face a double challenge: they increasingly struggle both to recruit women into PECS majors and to retain them once enrolled. The pipeline isn’t just leaky—it’s barely flowing. Interestingly, this polarization doesn’t exist in other STEM fields like biology, chemistry, and even mathematics, where gender ratios remain more balanced across all types of institutions. This suggests there’s nothing inevitable about the PECS gender gap—it’s a product of specific barriers we can address.
It is tempting to think that institutions serving students with high math achievement are improving the gender balance in PECS majors because they tend to have the institutional financial resources to devote to it or a larger applicant pool to choose from. While those factors do matter, they are dwarfed by the importance of the average math achievement of students. This suggests it’s not just about having money, but more likely, environmental factors and supportive programs, which are no doubt easier to implement with funding, but are not synonymous with it.
So, what can we do about it?
First, while institutional resources alone don’t determine success, external support targeted specifically at promoting gender equity can make a difference. For example, the National Science Foundation has invested hundreds of millions of dollars in promoting gender diversity in STEM faculty, but half that funding has gone to institutions whose students are at or above the 80th percentile of average SAT math achievement—precisely where the gender gap is already narrowing. While this investment has helped increase female representation among faculty and likely contributed to women’s persistence in PECS at these institutions, it’s time to experiment with greater support for institutions facing the biggest challenges.
Second, we should translate proven interventions like undergraduate research opportunities and peer mentoring programs to less-resourced institutions. These strategies have shown particular promise for retaining women and students of color in STEM fields. While well-resourced research universities often have the means to provide these experiences, we need creative partnerships to extend similar opportunities to all institutions. Expanding these opportunities may also require additional resource and service investments. We have ample evidence that related programs focused on belongingness and other social-psychological factors for student success that work at institutions serving students with high math achievement may not be as effective elsewhere, often because other supports are not in place.
External partners and organizations have a crucial role to play in addressing these disparities. While organizations like Girls Who Code conduct valuable outreach work, their impact could be amplified by strategically focusing their programs on schools and communities that feed into institutions where gender ratios are most imbalanced. Industry partnerships and internship programs should similarly prioritize building pathways for women at these institutions. This targeted approach should extend to community colleges as well, particularly since they serve as key feeder institutions to many less-selective four-year schools. Current research suggests that community college PECS programs often create unwelcoming environments for women, suggesting an urgent need for intervention and support at this critical entry point.
The narrative of progress in closing the STEM gender gap masks a troubling reality: We’re making gains at institutions serving students with the highest math achievement while losing ground everywhere else. The good news is that we know change is possible—the success of institutions serving students with high math achievement in narrowing their gender gaps, combined with the continued gender balance in other STEM fields, shows that these disparities aren’t fixed in stone. But achieving similar progress across all institutions will require acknowledging where we’re falling short and committing to change. Until we reckon with this disparity and allocate resources accordingly, claims of progress ring hollow for the majority of women who could benefit from technical education.
The Brookings Institution is committed to quality, independence, and impact.
We are supported by a diverse array of funders. In line with our values and policies, each Brookings publication represents the sole views of its author(s).