Getting Girls Into Engineering: What Actually Works at Home

Three things have real evidence behind them: contact with women who do the work, unshared hours with actual tools, and a parent auditing their own expectations. Everything else in this space - posters, slogans, pink-coloured kits - ranges from harmless to counterproductive.

The framing that helps most is narrow. The problem isn’t that girls can’t do engineering; internationally, the performance gap in mathematics is small. The problem is a collapse in expectation between ages 11 and 15, compounded by a hands-on experience gap that starts with what adults buy and who they hand the screwdriver to. This guide covers the numbers, the mechanism, and the specific interventions with the strongest evidence - including what not to bother with.

Where Things Actually Stand

The headline figures depend heavily on how you draw the boundary around “engineering.”

• EU: Women were 40.5% of scientists and engineers in 2024 - 7.9 million people, up from 3.4 million in 2008 - according to Eurostat. But the distribution is lopsided: 22.4% in manufacturing versus 45.1% in knowledge-intensive services, with national figures ranging from Latvia’s 50.9% down to Finland’s 30.7%.
• US: Women make up roughly 35% of the STEM workforce (NSF NCSES, Diversity and STEM, 2023) but only about 16% of architecture and engineering occupations (BLS data cited by ASME, 2024).
• Graduates: Women earn roughly 25% of ICT degrees and 24% of engineering degrees across OECD countries.

Read together: women are well represented in science broadly, and thinly represented in the hardware end of engineering specifically. That’s the gap this article is about.

It Is Not an Ability Gap

This part is settled well enough to state plainly. In PISA 2022, boys outperformed girls in mathematics by 9 score points across OECD countries - a small difference, and one that reverses in some countries. Meanwhile girls report more mathematics anxiety and lower belief in their ability to complete maths and science tasks.

The expectation gap is far larger than the performance gap:

• 10.7% of girls versus 15% of boys expect to work as science or engineering professionals, on similar science scores (OECD PISA).
• Among students reaching PISA proficiency Level 4 in reading, maths and science, 49% were girls - versus roughly 28% of STEM degrees going to women.
• Within girls, self-efficacy predicts performance dramatically: girls in the top quarter of the OECD’s maths self-efficacy index scored 147 points higher than girls in the bottom quarter.

If self-belief moves scores by 147 points and gender moves them by 9, the intervention target is obvious.

The 11-to-15 Window

Microsoft’s 2017 study with the London School of Economics - 11,500 women aged 11-30 across 12 European countries - found that girls’ interest in STEM peaks around age 11½ and begins waning by 15. Interest in humanities dips at the same age but recovers; STEM interest doesn’t. Only 42% said they’d consider a STEM career, and 60% said they’d feel more confident if they knew men and women were equally employed in the field.

The US picture is consistent. Junior Achievement’s 2023 survey of 1,012 teens found 10% of girls wanted a non-medical STEM job versus 21% of boys - essentially unchanged from 2018 - and identified parents as the single biggest influence (28%) on a teen’s dream job. Earlier Girl Scout Research Institute work found 74% of teen girls interested in STEM and 81% of those open to a STEM career, but only 13% naming it as their first choice. Notably, STEM-interested girls were far more likely to enjoy hands-on science projects (83% versus 56%) and understanding how things work (88% versus 65%).

So the window is roughly ages 9 to 13, before the drop-off, and the lever is hands-on experience.

The Tinkering Gap - and Where It Comes From

Here’s the mechanism, and it’s uncomfortable because it implicates adults rather than schools.

Adults buy building toys to boys. A Purdue University analysis of 1,069 real customer reviews for STEM toys found gifts going to boys versus girls at a 2.5:1 ratio overall - but the split by category is the finding that matters: engineering and construction toys went 59.5% to boys and 8.5% to girls, physics toys 57.9% versus 8.5%, while maths and general science toys were near parity (30.3% versus 29.4%). Around 65% of gender-identified purchases were parents and grandparents buying for boys. The bias isn’t against educational toys. It’s specifically against giving girls things to build.

Packaging changes adult behaviour, not just children’s. In a Child Development study (Coyle & Liben, 2020), 61 mother-child pairs were given the same mechanical toy, randomly packaged as the girl-coded “GoldieBlox” or a boy-coded “BobbyBlox.” Mothers given the boy-packaged version built more with the pieces; mothers with daughters read the narrative instructions more while mothers with sons built more. Girls learned the underlying belt-drive principle better from the boy-packaged version. The adult’s behaviour, cued by the box, shaped what the child learned.

The gap shows up years later as confidence. A study of engineering undergraduates who persisted to graduation (Chachra & Kilgore, ASEE 2009, drawing on NSF’s Academic Pathways Study) found male students reporting higher confidence in maths and science (p<0.034) and in open-ended problem solving (p<0.0005) across all four undergraduate years - with no performance or preparation difference explaining it. The authors’ warning is the important part: a confidence gap turns into a real experience gap, as women take documentation and coordination roles on team projects while men take the technical build.

Parents’ expectations diverge even at equal performance. OECD’s ABC of Gender Equality in Education found parents in every surveyed country more likely to expect a STEM career for a son than a daughter at identical mathematics performance - in some countries about 50% for sons versus under 20% for daughters.

What Works, Ranked by Evidence

• Contact with women engineers (Strength of evidence: Strongest - randomised trial) Find one real person: a family friend, a university open day, an engineering-society school visit, a factory tour. One hour of contact beats a year of encouragement.
• Hands-on build hours the girl controls (Strength of evidence: Strong - correlational + experimental) Buy a kit with real technical content and let her do 100% of the assembly. Sit beside her, hands in your lap.
• Auditing your own language and expectations (Strength of evidence: Strong - OECD, Junior Achievement) Stop saying “you’re so smart” and start saying “that took you three tries and you got it.” Notice whether you’d have bought this kit for a son.
• Naming the mechanism out loud (ages 11+) (Strength of evidence: Moderate) Tell her about the confidence gap. Teens respond well to being told the game is rigged in a specific, fixable way.
• Girls-only build groups (Strength of evidence: Unresolved) Worth trying if she’s the only girl in a mixed group; the research on single-gender STEM settings is genuinely mixed.
• Gendered “for girls” product lines (Strength of evidence: Weak to counterproductive) Judge the technical content, not the colour. If the same product exists un-gendered, buy that one.

The randomised trial is worth the detail. Breda, Grenet, Monnet and Van Effenterre (The Economic Journal, 2023) sent 56 female scientists into classrooms reaching about 20,000 French high-school students. Enrolment of Grade 12 girls in selective, male-dominated STEM programmes rose from 11% to 14.5%, and stereotypes about scientific careers and gendered ability declined. A one-hour visit moved real enrolment decisions by a third. Nothing else in this literature has that clean a causal result.

Choosing Kits Without the Pink Tax

The buying rule follows directly from the Purdue and Coyle-Liben findings: pick for technical content, and prefer products that aren’t gender-coded at all. A kit that a kid assembles and then programs passes the test regardless of what colour the box is.

A few concrete paths that work well in the 9-13 window, with current prices:

• Wearables: The CircuitMess Clockstar 2.0 ($99, ages 9+) is a build-and-code smartwatch - about an hour of screwdriver assembly, then coding custom watch faces and a gyroscope-driven game. Wearable tech tends to convert interest into ownership quickly because the result gets worn all day.
• Music technology: Synthia 2.0 ($149) is a buildable synth, sampler, sequencer and drum machine; Jay-D 2.0 ($119, ages 9+) is a DJ mixer with real DACs and amplifiers. For a girl whose declared interest is music rather than robots, these are engineering projects in disguise - and the Music Bundle 2.0 ($318) pairs them.
• AI and robotics: Wheelson 2.0 ($169, ages 9+) is a self-driving robot car with a camera and object recognition, programmable in CircuitBlocks, then Python and C++.
• Space. The NASA Mars Perseverance Rover ($349, ages 11+) is a roughly 20-hour hand-soldered build for a teen who wants a serious project.

None of these are marketed as girls’ products, which is the point. For a wider survey by age, see our best STEM gifts for kids by age guide.

Four Things Not to Do

Don’t take the tool. The single most common failure mode is a parent “just doing the fiddly bit.” The Coyle & Liben result is a warning: whoever holds the pieces is doing the learning.

Don’t over-praise ability. “You’re a natural” makes a bad session evidence that she isn’t. Praise the specific effort and the debugging.

Don’t make it about representation. A 10-year-old doesn’t want to be a statistic correction. She wants to build something that works. Save the structural conversation for 12+, and even then frame it as information, not a mission.

Don’t let her opt out of the hard part. Girls are frequently allowed to skip the frustrating debugging stage in a way boys aren’t - and that stage is where the confidence comes from. Our piece on how STEM projects build resilience and grit covers why the struggle is the mechanism, not the obstacle.

Frequently Asked Questions

Why do girls lose interest in STEM around age 12?

Interest peaks at roughly 11½ and starts declining by 15, according to Microsoft’s 2017 study with the LSE of 11,500 women across 12 European countries. The decline isn’t tracked by ability - the OECD PISA maths gap is about 9 points, and girls are 49% of students reaching high proficiency across all three domains. What changes is expectation: girls report more maths anxiety and lower self-efficacy, and adults around them expect STEM careers for sons more than daughters at identical performance.

What actually increases girls’ persistence in engineering?

The strongest causal evidence is for contact with women who work in the field: a randomised trial reaching about 20,000 French students raised girls’ enrolment in selective male-dominated STEM programmes from 11% to 14.5% after a single classroom visit. Beyond that, hands-on building experience the girl controls herself, and parents monitoring their own expectations and language, have the best supporting evidence.

Are building toys really given to boys more often?

Yes, and specifically building toys. A Purdue analysis of 1,069 STEM-toy reviews found engineering and construction toys went to boys 59.5% of the time versus 8.5% to girls, while maths and general science toys were roughly at parity. About 65% of gender-identified purchases were adults buying for boys.

Should I buy STEM kits marketed specifically for girls?

Judge the technical content, not the marketing. Experimental work (Coyle & Liben, Child Development, 2020) found that girl-coded packaging changed how mothers played - reading narrative instructions rather than building - and that girls learned the mechanical principle better from the boy-packaged version of the identical toy. If the same product exists without gender coding, that’s the one to buy.

What age should a girl start hands-on engineering?

As early as she’ll sit for it, but the critical window is roughly 9-13, before the documented interest drop-off. Screwdriver-assembly electronics kits are typically recommended from ages 7-9, and soldering projects from about 11 - so there’s a suitable step at every point in that window.

Does it help if she’s the only girl in a robotics club?

It can go either way, and the research on single-gender STEM settings is genuinely mixed rather than settled. What is documented is the team-role problem: mixed teams often drift toward girls taking documentation and coordination while boys take the technical build. If that’s happening, either fix the role rotation explicitly or find a group where she’ll be one of several.

Conclusion

The gap in engineering isn’t in ability or even in stated interest - it’s in hours spent with hands on hardware, and in what the adults around a girl expect. Both are fixable at home, this month: find her one real engineer to talk to, buy a kit with genuine technical content, and then keep your hands off it while she builds. If you want somewhere to start, the un-gendered build-and-code kits at CircuitMess - a smartwatch, a synth, an AI robot car - are all projects where the finished device belongs entirely to whoever assembled it.