Science, technology, engineering and mathematics — the STEM fields — are where much of the modern economy’s value and power are being created. They are also among the most stubbornly male-dominated corners of the professional world. Despite decades of effort, women remain a distinct minority in STEM work, and the reasons have less to do with ability than with a pipeline that leaks at every stage and a set of barriers that can be identified and, with the right interventions, dismantled.
The numbers, stated precisely
The often-quoted “28% of STEM workers are women” figure is real but depends entirely on definitions, and it is worth being exact about what is being counted.
In the United States, the Census Bureau found that women made up about 27% of the STEM workforce as of 2019, up dramatically from just 8% in 1970 — even though women constitute roughly 48% of the total workforce. The commonly cited “28%” traces to this Census measurement.
Globally, UNESCO reports that about one in three researchers — 33.3% — are women, a figure covering more than a hundred countries. In the education pipeline, women earn roughly 35% of STEM degrees worldwide, a share that has been essentially flat for a decade.
The single most important thing these aggregate numbers hide is horizontal segregation — the fact that women’s representation varies enormously between fields. In the US in 2019, women held only about 15% of engineering jobs and roughly 25% of computing jobs, but around 47% of mathematics roles and 45% of life- and physical-science roles. Because engineering and computing together account for the large majority of all STEM jobs, their low female shares drag down the overall average. Women are close to parity in biology and mathematics and vanishingly rare in software engineering and mechanical engineering. “28% of STEM” is an average that conceals this crucial variation.
The leaky pipeline
The metaphor researchers use is a “leaky pipeline”: at every transition — from school to university, from degree to job, from junior to senior roles — a disproportionate share of women drops out. The evidence for the leak is direct. US data show that around 60% of women with science and engineering degrees end up working outside science and engineering occupations, compared with 41% of men. The problem is not only that too few women enter STEM, but that many who train for it never work in it, or leave mid-career.
The leaks have identifiable causes at each stage. Early on, girls encounter stereotypes — often subtle, often from adults who mean well — that STEM is a masculine domain, eroding confidence and interest before university is even chosen. In higher education, women in male-dominated STEM courses can face a chilly climate, few female peers, and a shortage of role models. In the workplace, they meet bias in hiring and promotion, isolation as one of very few women, inflexible cultures poorly suited to those with caregiving responsibilities, and harassment. Each factor is survivable alone; stacked together, they push women out at every stage.
The glass ceiling within STEM
Even women who stay in STEM hit an additional barrier at the top. Technical fields have their own glass ceiling: women are better represented among junior engineers and scientists than among the managers, principal investigators, senior architects and executives who lead them. The attrition compounds — a field that starts at 28% women at entry may be down to single digits in its senior technical and leadership ranks. This is why the aggregate number, discouraging as it is, actually understates the concentration of the problem at the top.
What actually works to break the ceiling
The encouraging finding from the research is that the barriers are not immovable, and that some interventions have real evidence behind them while others — however well-intentioned — do not. Drawing on reviews by the US National Academies of Sciences, Engineering, and Medicine and others, the interventions with the strongest support cluster around a few themes.
Structured mentorship and sponsorship. General encouragement helps little; structured programmes that pair women with mentors — and, more powerfully, sponsors who actively advocate for their advancement — improve retention and persistence. Sponsorship matters more than mentorship at senior levels, because promotion decisions are made by people with power, and women need advocates in those rooms.
Visible role models. Seeing women who have succeeded in a field measurably raises girls’ and young women’s interest and sense of belonging. Near-peer role models — someone only a few years ahead — are especially effective, because they make success feel attainable rather than exceptional.
Bias interruption in hiring and evaluation. The evidence here contains an important nuance: one-off implicit-bias training sessions have weak and often short-lived effects. What works better is redesigning the processes in which bias operates — using structured interviews with consistent questions, evaluation rubrics with predefined criteria, and blind or anonymised review of applications and work where feasible. Changing the system beats trying to change individual minds one workshop at a time.
Active-learning pedagogy. In university STEM courses, replacing passive lectures with active, collaborative learning has been shown to disproportionately improve outcomes and retention for women and underrepresented students — closing gaps rather than merely raising all boats.
Pay transparency and family-friendly policies. Transparent pay structures narrow the gender pay gap by removing the secrecy in which discrimination hides, and flexible working and decent parental leave reduce the mid-career exodus that drains women from technical fields precisely when they would be moving into leadership.
Why the stagnation is the real alarm
Perhaps the most important recent data point is that the share of women among STEM graduates has been flat for ten years. This is the alarm bell. It means that current efforts, whatever their good intentions, are not shifting the pipeline at the entry point. ### Where the leak begins: confidence, not competence
One of the most important and counterintuitive findings in this literature concerns when girls disengage from STEM, and why. Careful studies of school performance find that girls perform as well as or better than boys in mathematics and science through most of their schooling. The divergence is not one of ability but of confidence and belonging: girls, on average, report lower confidence in their mathematical ability than boys of identical measured performance, and this confidence gap — not a skills gap — predicts who pursues STEM further. Research has traced part of this to the subtle transmission of stereotypes by adults, including teachers and parents who, often without any conscious intent, convey that mathematics and computing are boys’ territory. By adolescence, many girls who are entirely capable have already ruled STEM out, not because they cannot do it but because they have absorbed a message that it is not for them. This is why interventions that arrive at university are, for many, already too late: the decisive filtering has happened years earlier, in classrooms and living rooms.
The computing paradox
Computer science offers a striking and instructive anomaly. In the 1980s, women earned a substantially larger share of computer-science degrees in the United States than they do today — the female share actually peaked around the mid-1980s and then declined for decades, even as women surged into medicine, law and the biological sciences. This history matters because it refutes the notion that women’s underrepresentation in computing reflects some fixed, natural preference. If it were natural, the share would not have fallen precisely as personal computers became mass consumer products marketed heavily toward boys, and as the cultural image of the programmer hardened into a male “geek” archetype. The computing case demonstrates that field-level gender composition is a cultural artifact, responsive to marketing, media images and institutional signals — which is discouraging in that it shows how easily progress can reverse, but encouraging in that it shows the composition is not destiny and can be deliberately reshaped.
The business and innovation cost
Beyond fairness, there is a growing recognition that the underrepresentation of women in STEM produces worse technology. The examples are numerous and increasingly well-documented: crash-test dummies modelled on male bodies that left women at higher risk of injury; medical research conducted predominantly on male subjects that missed sex differences in drug responses and disease presentation; voice-recognition and other systems trained on male-dominated data that worked poorly for women; and, more recently, concern that artificial-intelligence systems built by homogeneous teams can encode and amplify gender bias at scale. When the people who design technology do not reflect the people who use it, the technology serves some users better than others. This reframes the diversity question in STEM not merely as a matter of who gets good jobs, but as a matter of whether the products that increasingly govern daily life actually work for half the population.
The role of organisational culture in retention
For women who do enter STEM careers, the evidence identifies workplace culture as the decisive factor in whether they stay. Surveys of women who left engineering and technology roles consistently cite not a loss of interest in the work but the surrounding environment: isolation as one of very few women, exclusion from informal networks where influence and opportunity flow, a lack of advancement despite strong performance, inflexible expectations incompatible with caregiving, and, in too many cases, harassment. The remedies that retention research supports are concrete — building critical mass so that no woman is the “only one” on her team, ensuring transparent and equitable pathways to promotion, establishing genuine accountability for harassment, and offering the flexibility that keeps mid-career women from being forced to choose between the work they trained for and their families. Retention, in short, is not about “fixing” women to fit an inhospitable culture; it is about fixing the culture so that the women already there are not driven out.
The counter-intuitive “gender-equality paradox”
No honest survey of this field can ignore a genuinely puzzling finding that critics sometimes raise: the so-called gender-equality paradox in STEM. Several studies have observed that some of the most gender-equal countries in the world — the Nordic nations in particular — have lower shares of women among STEM graduates than some less gender-equal countries. On its face this seems to suggest that when women are freest to choose, fewer choose STEM, implying the gap reflects preference rather than barrier. The finding is real and deserves engagement rather than dismissal, but its interpretation is hotly contested. Some researchers argue it reflects that in wealthy, secure societies people feel freer to follow intrinsic interests, while in economically precarious societies the high pay of STEM fields pulls women in out of necessity. Others have challenged the paradox on methodological grounds, noting that the result is sensitive to exactly how “share of women in STEM” is measured and that alternative measures weaken or reverse it. The most defensible reading is that both preferences and barriers matter, that the two are deeply entangled (preferences themselves are shaped by the stereotypes and cultures discussed above), and that the paradox does not license the lazy conclusion that the gap is natural and nothing should be done. Even in the most gender-equal societies, the barriers of stereotype, culture and workplace climate persist — and where they have been actively dismantled, women’s participation has risen.
Breaking the glass ceiling in STEM will require moving beyond awareness campaigns to the structural changes that the evidence actually supports — redesigned hiring and evaluation, active pedagogy, real sponsorship, and workplaces built to retain women through the career stages where they now leak away. The stakes extend beyond the women themselves to the quality of the science and technology a society produces, which is only as good as the range of minds allowed to shape it. The talent exists in abundance; the field’s failure to keep and elevate it is a design problem, and design problems can be solved — deliberately, measurably, and starting long before the university gate.
Sources: US Census Bureau (2021, 2019 data); UNESCO, Women in Science and UIS data; US National Science Foundation / NCSES, The STEM Labor Force; National Academies of Sciences, Engineering, and Medicine, Promising Practices for Addressing the Underrepresentation of Women in Science, Engineering, and Medicine; UNESCO GEM Report analyses.
