## How girls and boys differ in STEM subject choice

11th October 2021 by Timo Hannay

Gender balance in STEM (science, technology, engineering and maths) is a perennial
topic of concern and campaigning. For our part, SchoolDash is proud to be associated with Ada Lovelace Day, the global celebration of women in
STEM, which this year kicks off on 12th October, and with Maths4Girls, an
initiative that introduces budding British mathematicians to role models with whom they can identify.

This post digs into some of the numbers that explain why such initiatives are
necessary and the nature of the problem they're seeking to address. Specifically, we're going to look at
gender differences in STEM subject choice at A-level using data from
the Department for Education (DfE) that covers the cohorts in England that sat their A-levels in 2018 and
2019 (ie, before COVID-19 struck).

In summary, we find that:

- Computing shows the most skewed gender ratio, with only 3 girls for every 20 boys, but Maths and Physics
show the biggest notional shortfalls, with over 20,000 girls a year 'missing' in each subject. If
anything, these gender gaps widened slightly between 2018 and 2019, particularly in Maths
- Despite this, Maths remains the second most popular STEM subject among girls (after Biology), with 17%
of female A-level students taking it in 2019. This compares to only 4% for Physics
- Gender disparities exist in all parts of England and at every kind of school. Indeed, they tend to be
*bigger* at schools with low levels of disadvantage, where STEM subjects are more popular across
the board but gender differences tend to be even more pronounced
- Single-sex and independent schools show somewhat less gender-stereotyped subject choice, but it's
unclear how much of this is down to the schools themselves rather than simply the characteristics of their
intake
- On a more positive note, we see relatively little evidence for gender-specific cohort effects in which
very small numbers of girls or boys wishing to take a subject at a particular school result in none of
them doing so

For more on this topic, including school-level information, see also the data and research
assembled by the Advanced Mathematics Support Programme.
They have a useful factsheet for teachers too.

For more on our analysis, read on...

The data presented here use two distinct and complementary ways of measuring
differences between girls and boys in A-level subject choice:

**Percentage point gap:** If a subject is taken by, say, 20% of boys but only 10% of girls
then we can say that there is a 10 percentage point gap. One benefit of this approach is that it provides
an indicator of roughly how many students we're talking about (on the basis that 1 percentage point
corresponds to 1,400-1,500 girls or 1,100-1,200 boys each year).
**Gender ratio:** This shows how many girls take a subject for each boy that does so. It is
important to understand because sometimes small percentage-point gaps can hide very large gender
disparities. For example, a subject taken by 1% of boys and 0.1% of girls would have a gap of less than
one percentage point but a huge boy:girl gender ratio of 10:1.

**Mind the gap**

We will start by looking at *percentage-point gender gaps*. Figure 1 shows these for all six A-level STEM subjects covered by the DfE data. They
come from exams sat by students in England in 2018 (blue columns) and 2019 (red).

Biology has more girls than boys, while Chemistry is almost balanced. Other STEM
subjects show relative shortfalls for girls, with Maths having the biggest gap of all. Note that this is in
part because overall participation rates in Maths are relatively high (roughly 17% of girls and 35% of boys
in 2019). In four of the six subjects, percentage-point gender gaps increased between 2018 and 2019.

(Hover over the columns to see underlying values and cohort sizes. Click on the
figure legend to hide or display years.)

Figure 1: Difference in proportions of girls and boys taking
A-level STEM subjects

Figure 2 breaks these national results down by school
location and type. Looking at Maths participation by region, disparities are present everywhere, though the Midlands and the
East of England have larger gaps than most. So does London, but note that overall Maths participation there
is high (22% of girls and 40% of boys in 2019).

Some outcomes are a slightly counterintuitive. For example, schools with
*lower* deprivation (as indicated by the proportion of students
eligible for free school meals) have *larger* gender gaps. So do those with higher Ofsted ratings or greater proportions of students with English as an additional language (EAL; usually a positive indicator of
educational outcomes). Once again, this is largely because they have higher overall entry rates in these
subjects. As we shall see below, these don't always translate into worse gender ratios.

Single-sex schools, which are often considered
to reduce gender stereotyping among teenagers, have somewhat lower disparities in Biology and Computing, but not in
Maths or Physics. Note, however,
that overall entry rates in Maths and Physics are higher at single-sex schools, so it is also important to
look at gender ratios (see below). Much the same is true of grammar
schools, which overlap considerably with single-sex schools.

(Use the menus below to explore other school categorisations and STEM subjects.
Hover over the columns to see underlying values and sample sizes. Click on the figure legend to hide or
display the different years.)

Figure 2: Difference in proportions of girls and boys taking
A-level STEM subjects by school type

Figure 3 shows the same 2019 data, but by local authority.
This uses a cartogram in which each area is scaled according to its pupil population, making it easier to
see small, densely populated urban locations. Red areas have low levels of female participation, blue areas
have high levels and maroon ones are roughly in balance. Predictably, these patterns are very different by
subject: compare Biology, Chemistry, Computing, Further Maths, Maths and Physics.

Note that sometimes a large imbalance, or even gender parity, is simply a
consequence of low overall entry rates. To give an extreme example of the latter, Knowsley in the North West
achieved a perfect balance by reporting no girls or boys taking A-level Maths
in 2019.

(Explore further using the menu below. Hover over the map to see corresponding data
and cohort sizes for each region.)

Figure 3: Difference in proportions of girls and boys taking
A-level STEM subjects by local authority (2019)

The results shown above come from state schools (which are much easier to slice and
dice by type because data about them are plentiful). But independent schools are a significant part of the
mix too. As shown in Figure 4, they tend to do better at getting girls into STEM
subjects, notably in Maths and Computing, where the gender gaps are considerably smaller than at state
schools.

Of course, this doesn't demonstrate that independent schools themselves are *causing*
the effect since the characteristics of their intakes are very different to state schools. Indeed, one could
even argue that if girls who are attracted to STEM subjects are disproportionately likely to attend an
independent school then this would tend to increase gender gaps in the state sector. The data shown here
demonstrate that there is a difference between the sectors, but not, unfortunately, anything about its
underlying causes.

(Hover over the columns to see underlying values and cohort sizes. Click on the
figure legend to hide or display individual sectors.)

Figure 4: Gender gaps for A-level STEM subjects at state and
independent schools (2019)

**Rational numbers**

This section provides roughly the same analysis as above, but using *gender
ratios*. These are presented as girl:boy, so 0.5 means half as many girls as boys, while 2.0 means twice
as many girls.

Figure 5 shows a summary by subject. In this case, Computing
shows the greatest imbalance with a national ratio of 0.15 in 2019 (ie, 3 girls for every 20 boys). Maths
has a ratio of about 0.6 (3 girls for every 5 boys), while Biology shows the inverse, with a ratio of 1.7
(roughly 5 girls for every 3 boys).

Figure 5: Ratios of girls : boys taking A-level STEM subjects

Here, too, disparities in Maths were evident in
every region, as shown in Figure 6. Once again,
the Midlands and the East of England showed the largest disparities. On this measure, London does relatively
well.

Consistent with the analysis of percentage-point gaps described above, schools with
lower deprivation tend to have worse gender ratios for Maths. But the same is not true of schools with higher Ofsted ratings or greater proportions of EAL students, which have more balanced gender ratios despite having
large percentage-point gaps. The same is true of single-sex schools and
grammar schools, which have more balanced gender ratios in Maths and the other STEM subjects covered here.

It is interesting to observe that single-sex schools also have different proportions
of female and male teachers. As a whole, 64% of teachers in single-sex schools are female, which is very
close to the proportion for mixed-sex schools (62%). But at girls' schools fully 75% of teachers are female
while at boys' schools only 47% are. How these proportions vary by subject, and whether or not they have any
direct effect on A-level subject choice, is unfortunately not possible to determine using the public data
currently available.

(Use the menus below to explore other school categorisations and STEM subjects.)

Figure 6: Ratios of girls : boys taking A-level STEM subjects by
school type

Figure 7 shows gender ratio data by local authority area. The
picture once again varies by subject: compare Biology, Chemistry, Computing, Further Maths, Maths and Physics. Here too, look out for extreme effects caused by small numbers
of students. For example, the very girl-heavy ratio achieved for Maths
by Hartlepool in the North East is based on just 28 students (19 girls and 9 boys).

(Use the menu below to explore other subjects; hover over each local authority area
to see corresponding data values and cohort sizes.)

Figure 7: Ratios of girls : boys taking A-level STEM subjects by
local authority (2019)

Figure 8 compares the gender ratios at state schools that we
have analysed above with those at independent schools. For every subject shown, gender ratios are more
balanced (ie, closer to a value of 1) among independent schools. As already pointed out, we can't tell from
these data alone how much of this is due to something the schools themselves are doing and how much is
simply a consequence of their distinctive intake.

(Hover over the columns to see underlying values and cohort sizes. Click on the
figure legend to hide or display individual sectors.)

Figure 8: Gender ratios for A-level STEM subjects at state and
independent schools (2019)

**All together now**

Figure 9 summarises the relationship between the gender gaps
(vertical axis) and gender ratios (horizontal axis) across all six subjects. The size of each dot represents
the total numbers of students taking the subject (hover over them to see corresponding data values).

Computing shows the greatest skew towards boys, but is still a niche A-level
subject, so the absolute number of 'missing' girls is relatively modest (about 9,000). Maths and Physics
show somewhat less extreme gender ratios, but overall entry rates are higher, so the notional shortfalls in
numbers of girls are larger (around 20,000-25,000 apiece). Conversely, Biology is short of boys to the tune
of about 7,000. Chemistry – sometimes thought of as a Cinderella science caught between the profundity
of Physics and the self-knowledge of Biology – is admirably well balanced on both measures. (In fact,
Chemistry has a slightly higher participation rate among boys, but a gender ratio in favour of girls; this
is because the whole A-level cohort contains more girls than boys.)

Figure 9: Percentage point gaps and gender ratios by A-level
subject (2019)

**Birds of a feather?**

Finally, we will look for signs of cohort-size effects. These can occur when very
small numbers of students are interested in taking a subject, sometimes resulting in none of them doing so.
For example, if only one or two girls at a particular school want to take Physics A-level then they may be
dissuaded from doing so in order not to feel unusual or isolated. Overall, we don't see particularly strong
evidence for this in the data analysed here, though that is far from the same as demonstrating that it
doesn't happen.

For example, in Physics, the numbers of girls
choosing the subject (red columns) are generally low (ie, shifted towards the left of the figure), but there
are no obvious shortfalls in the numbers of schools reporting just one or two girls – the curve rises
smoothly as it approaches the vertical axis. There *is* such a shortfall for boys (blue columns)
– the curve dips near the axis – but since they represent the majority of students, this is
presumably due to the fact that schools with very few boys taking A-level Physics will often not be in a
position to offer it at all.

Maths arguably shows a hint of a shortfall
among schools with low numbers of girls, though this may also be due to low total numbers of students. Further Maths and Computing do
not show obvious shortfalls, but the numbers of girls here are so small that one might argue these subjects
as a whole are suffering from cohort-size effects. Chemistry shows
shortfalls for both sexes – which is understandable given the roughly even gender balance in that
subject. Biology shows more or less the reverse behaviour to Maths,
which makes sense given that it has the inverse gender ratio.

Figure 10: Distribution of student numbers by A-level subject in
state schools

**All askew**

There is no reason to believe that all A-level subjects should attract equal
proportions of boys and girls – and they don't (see this recent analysis by FFT Education Datalab for the broader picture). But the gender
balance in many STEM subjects is so far out of kilter that at the very least it deserves serious attention.
The potential risk is not only that girls (or boys) with genuine interest and aptitude are put off from
subjects in which they could excel, serious though that may be. It is also that the analysts, engineers and
technologists who shape our collective future will not be representative of the society they serve. While not the only kind of bias that
ought to concern us, gender is surely one aspect in which we can still strike a much better balance.