Coding for Girls: What Actually Changes the Odds

Most writing on this subject is either a recruitment poster or a list of famous women, and neither helps a parent decide anything on a Tuesday evening. What follows is the research on where girls actually leave computing, which is earlier and for different reasons than most parents assume.

I have been teaching this subject long enough to have watched the pattern happen in front of me, and the useful finding is that the decision point arrives years before anyone is choosing subjects. Every number below comes from a source you can open and check.

Is there any evidence that boys are better at coding than girls?

No, and the data points the other way among children who actually take the subject. In England in 2022, girls sitting GCSE Computer Science outperformed boys at the top grades.

The Raspberry Pi Foundation's foundational evidence review on gender in computing education, published in April 2023, reports that "female students continue to outperform their male counterparts in the subject, with 40.6% of girls earning a 7/A grade in 2022 compared to 32.3% of boys."

That gap matters because of what sits next to it in the same review. Summarizing research by Kallia and Sentance, it notes that "boys feel more confident and predict their programming performance better than girls," while "girls tend to underestimate their abilities."

So the honest summary is uncomfortable. Among children studying computing, girls do better and believe they are doing worse. Whatever is removing girls from this subject, it is not their performance in it, and any explanation that starts with aptitude is arguing against the grades.

When do girls actually start opting out?

Around age six, which is roughly a decade before anyone thinks the decision is being made. This is the finding that should change how a parent talks at home, and almost nobody mentions it.

In four studies with 400 children, Lin Bian, Sarah-Jane Leslie and Andrei Cimpian found that by the age of 6 "girls are less likely than boys to believe that members of their gender are 'really, really smart'," which was their child-friendly way of asking about brilliance. At age 5 the researchers found no statistically significant difference between boys and girls. It appears at six.

The second half is the part that should worry a parent more. "Also at age 6, the girls in these studies begin to shy away from novel activities said to be for children who are 'really, really smart'." The belief does not stay an opinion. It immediately changes what a child will agree to try.

What makes this finding solid rather than merely alarming is the control the researchers ran. They also asked children who gets the best grades, and there the drop did not appear: older girls were actually more likely than younger ones to pick girls as top of the class. Girls at six were not reporting that girls do worse. They had absorbed something specific about brilliance, and only about brilliance.

Why does the way an activity is described matter so much?

Because in the same research, the effect vanished when the activity was described differently. Girls were less interested than boys in a game presented as being for "really, really smart" children, and that gender difference was not statistically significant for a game presented as being for children who try hard.

Sit with that for a moment, because it is the single most practical thing on this page. The activity was identical. Only the sentence introducing it changed, and the gap opened or closed accordingly.

The application at home is immediate. "You have to be really clever to do this" is an advertisement aimed away from your daughter, even when it is meant as encouragement, and even when it is true. "This is fiddly and people get it wrong a lot at first" is not. The second framing is also a more accurate description of programming, which is convenient.

This extends to how you praise. Telling a child she is brilliant at something attaches her performance to a fixed quality she has been taught to doubt she has. Telling her she worked out a hard bug attaches it to something she did, and she has no reason to doubt that.

How big is the drop-off by the time it shows up in the numbers?

Large, and it moved in the wrong direction for twenty years. The share of computer science bachelor's degrees in the United States going to women fell over two decades, even while the number of women earning them rose.

The National Science Foundation reports that among computer sciences bachelor's degree recipients "the share receiving bachelor's degrees declined, from 27.0% in 1998 to 19.9% in 2018."

The two facts in that sentence need holding at once, because quoting either alone misleads. More women study computer science now than in 1998, and the number with bachelor's and doctoral degrees more than doubled. The field simply grew faster around them. This is a story about share of a growing field, not about women leaving in absolute terms, and anyone presenting it as a collapse is overstating it.

It is still a real problem. But it is worth knowing exactly what the number says, because your daughter will eventually be told a dramatic version of it, and a dramatic version of this statistic is itself discouraging.

Do girls-only coding classes actually work?

Sometimes, for some children, and the evidence is weaker than the marketing. Girls report a stronger sense of belonging in single-sex settings, but the broader evidence on separating students by gender does not show the academic benefits people expect.

The belonging finding is real. The Raspberry Pi Foundation review cites work by Leonard and colleagues finding "mixed-sex high schools having lower scores for feelings of belonging than single-sex schools," and belonging is not a soft outcome here, given that the problem is confidence rather than ability.

The counterweight is a large meta-analysis. Erin Pahlke, Janet Shibley Hyde and Carlie Allison analyzed 184 studies covering 1.6 million students in grades K to 12 across 21 nations. Studies without controls for selection effects showed modest advantages for single-sex schooling, but controlled studies "showed only trivial differences," and in some cases small differences favoring coeducation. Their conclusion is that results from the highest quality studies do not support the view that single-sex schooling provides benefits over coeducational schooling.

Two honest caveats on using that against girls-only coding clubs. It studied schooling broadly rather than computing specifically, and a voluntary weekend club is a different thing from a segregated school. But it does tell you that the burden of proof sits with the claim, and that when a girls-only program appears to work, the most likely explanation is selection: the families who sign up were already committed.

The practical read is that this is a fit question, not a principle. A girl who has already decided coding is a boys' activity often does better in a room that quietly disproves it. A girl with no such idea may not need one, and may prefer her existing friends.

What can a parent actually influence at home?

Three things, none of which require you to know how to code: what you say about difficulty, what your daughter sees you struggle with, and whether quitting is treated as information or as failure.

The difficulty language is covered above and is the highest-leverage. The second is nearly as strong. A child who only ever sees adults being competent learns that competence is a state you arrive in rather than something built, and a parent visibly failing at something and continuing anyway teaches more than any encouragement.

The third is where most well-meaning households go wrong. When a girl says she wants to stop, the instinct is either to accept it immediately, which confirms the exit, or to insist, which makes it a fight. The better move is a question: is this boring, or is this hard right now? Those need opposite responses, and children are usually accurate when asked directly.

It also helps not to make coding a gendered event. A daughter enrolled with fanfare about girls in tech has been told the thing is unusual for her before she starts. Our guides on getting a child interested in coding and what age to start apply here without modification, which is rather the point.

Does it matter who teaches the class?

It helps, and it helps less than people assume. The evidence review is direct that similarity to role models matters for attracting and retaining women in computing, while noting that sharing the same gender may not be sufficient on its own.

That second half is the useful correction. A woman teaching the class is not a strategy by itself, and choosing a worse class for a female mentor is a bad trade. What the research keeps pointing at is not the mentor's gender but whether the child's confidence is being addressed directly, which is a teaching behavior rather than a demographic one.

What to look for instead is specific. Does the teacher name what the child did rather than what she is? Does the class treat a bug as normal rather than as evidence about the student? Is there someone who would notice her going quiet, which is how this usually starts?

We have more than 20 expert instructors, and the honest reason that number matters here is matching rather than choice for its own sake. A child who has decided she is bad at this needs a mentor who will not confirm it, and with one teacher there is no second try.

What should you do if your daughter says coding is not for her?

Ask her what gave her that impression, and listen for whether the answer is about the subject or about herself. "It is boring" is a curriculum problem you can solve. "I am not good at it" at age eight is almost never a true report.

The research above says what to expect. She is more likely than a boy to underestimate her own ability, she may have absorbed a message about who is clever before she could read, and her actual performance is not the thing driving the conclusion. None of that means pushing harder. It means not accepting her self-assessment as evidence, while entirely accepting her preferences as preferences.

There is a real distinction and it is worth getting right. A girl who has tried coding properly and does not enjoy it has learned something useful, and the correct response is to let it go. A girl who never really started because she decided in advance that it was not for her has learned nothing, and the correct response is a low-stakes second look at something she can do in an afternoon.

If you want to see whether the second look lands before committing to anything, a free trial class is one live session with a mentor, no card details and no commitment. Watch how she is spoken to when she gets something wrong, because that is the part that decides whether she comes back.