Sexism in higher education
Some time ago I wrote a post with empirical evidence for sexism in computer science. I’ve since realised that the data I used then is part of a much larger data set maintained by the US National Center for Education Statistics: here are some more pictures of their data.
What I want to show
The ratio between male & female1 students gaining degrees in various subjects has often changed dramatically over a fairly short period of time (generally about 50 years here: about two generations). Such a dramatic, rapid change shows that these ratios are not due to innate ability but to women being discouraged from studying some subjects.
By plotting pictures of this data these changes become much easier to see, compared with looking at large tables of numbers.
The data and how it was plotted
The data I’m using comes from NCES and in particular it comes from these tables. These are, I am sure, updated fairly frequently: the data I am plotting here was fetched in November 20232.
The plots are simple-minded: they just look at the male/female ratio without taking any account of the total number of students. There is no smoothing. For fields which started very small there is therefore sometimes quite a lot of variability, especially for higher degrees.
The ratios were computed from the numbers of students in the tables, and in particular I didn’t use their precanned figures. Where I checked, mine are the same.
The data does not contain any gender information other than male & female: in particular it takes no account of trans people or anything like that. It’s unlikely that that data was even gathered until quite recently, of course, but in any case this is just missing from the pictures because it’s not in the data.
The tables contain data for three levels of degree: batchelor’s (BSc or BA I presume), master’s (MSc, MA, MPhil I presume) and doctor’s (PhD). Those have been plotted separately.
I’ve used the starting year of any given academic year: 1984–85 turns into 1984. I’ve not generally plotted the data before 1970 so all the graphs have the same start date: where there is data before 1970 in the tables it is generally decadal. The end dates are the end date in the data which is slightly variable.
The plots have a y-axis which runs from 0 to 50% or to the nearest multiple of 5% above the maximum female percentage.
I have looked, rather casually, for sources of similar data for the UK. I haven’t found anything. I have asked the UK Office for National Statistics though, and if they have anything useful I will write another post.
The pretty pictures
This is necessarily just a fairly arbitrary selection of plots of subject areas I thought would be interesting: there is a lot more data there that I have not plotted.
Computer and information sciences
This is table 325.35, and is an updated version of what I plotted previously. This plot runs from 1970 to 2020.
CS & IS graduate sex ratio, US, 1970–2020
You can see from this that the ratio has gone up since about 2010, but it is still far lower than it was in about 1984. The data for higher degrees shows far less of a bump than the data for first degrees: presumably whatever drove women out of CS courses had less effect for higher degrees. The data for doctor’s degrees is quite bumpy because the numbers are rather low: this is common in all the graphs.
Mathematics and statistics
This is table 325.65. This plot runs from 1970 to 2020.
Mathematics & statistics graduate sex ratio, US, 1970–2020
This shows pretty much no sign of a 1980s peak followed by a decline. What it does show is a slight decline after about 2000 which perhaps is visible in all three lines.
Engineering and and engineering technologies
This is table 325.45. Plot from 1970 to 2020 again.
Engineering & engineering technologies graduate sex ratio, US, 1970–2020
Well, there were essentially no women studying engineering in 1970 (who knew?), but there are a lot more now. A woman studying engineering is now more likely than a man to pursue a higher degree in the subject. There might be a 1980s effect, but there definitely is something after about 2000, although it seems to have gone away now. This is a pretty dramatic picture, I think.
Physical sciences and science technologies
This is table 325.70. The plot runs from 1970 to 2017, which is the most recent data.
Physical sciences & science technologies graduate sex ratio, US, 1970–2017
There’s no 1980s effect. There is quite a strong post–2000 effect.
English language and literature/letters
This is table 325.50. The plot runs from 1970 to 2017, which is the most recent data.
English language & literature/letters graduate sex ratio, US, 1970–2017
More women than men do first degrees in this area, and this ratio has been pretty stable for a long time. Once upon a time not many women went on to do doctoral degrees, but the ratio has mostly caught up now.
Health professions and related programs
This is table 325.60. Plot from 1970 to 2020.
Health professions & related programs graduate sex ratio, US, 1970–2020
This has always been dominated by women. What is probably hidden in this data is that most of this dominance was nursing and related degrees, while the number of female (medical) doctors was rather low. However it has climbed steadily and dramatically since 1970, and in 1920 about 60% of new doctors were female.
Social sciences and history
This is table 325.90. The plot runs from 1970 to 2017, which is the most recent data.
Social sciences & history graduate sex ratio, US, 1970–2017
Unfortunately there is no table for economics degrees: this is the best proxy I could find. It’s not terribly interesting although it does show some signs of both the 1980s and post–2000 dips.
Visual and performing arts
This is table 325.95. Plot from 1970 to 2020.
Visual & performing arts graduate sex ratio, US, 1970–2020
Again, women have always slightly dominated first degrees, and have been increasingly likely to do higher degrees. There is perhaps some fall in the ratio for doctoral degrees after 2000. Is there a 1980s dip here?
What can you conclude
First of all and most obviously: there is no evidence for differences in innate ability between men and women here. If you look particularly at the graphs for CS & IS, engineering and physical sciences, you will see enormous changes in the percentage of women graduating in these areas within two generations. In the case of Engineering the change between 1970 and 2020 was by a factor of 28, from 0.8% to 23% female. The change between 1949 (the first data in the table) and 2020 was by a factor of nearly 77. It is absolutely impossible that such changes should be due to changes in innate ability, and it is equally impossible to even glimpse any possible differences in innate ability in the presence of this vast socially-driven change.
Secondly, as before, something happened in CS & IS which drove out nearly half the women who studied it in a little more than a generation. This was probably the period when achieving a degree in this area was most lucrative.
And this time I’ll come out and say it: I think that what did this was just obviously white male tech bros who started arriving on CS courses after the home computer revolution of the early 1980s. These are people who clearly behave just as badly towards women as you would think they would. Who would have guessed it?
Thirdly there may be some evidence of a decline of women in some science & engineering subjects after 2000. I don’t know what is causing that, if it’s even real.
Finally, things are now more equal — better in fact — than they were in 1970, although CS & IS has not recovered from its infestation of tech bros.
I haven’t yet looked for tables like the ones I’ve used here categorised by ethnic group. I imagine they will tell exactly the same story: that there is no indication at all of any innate difference in ability between ethnic groups.