In my last post, I plotted trends in various college majors, comparing how the numbers of degrees awarded in a range of subjects have grown or shrunk over the decade 2011-2021. I compared my university, the University of Oregon, and other R1 (“Very high research activity”) universities.
This post focuses on computer science and closely related majors. Its essence is encapsulated in this Facebook exchange with a friend of mine who’s a professor at UC Berkeley (where I was an undergraduate):
Me: How much did the number of computer science degrees at UC Berkeley increase between 2011 and 2021? Take a guess.
Professor X: 30%
Me: [spits out coffee]
The answer: 1100%
Yes, that’s a factor of 12 increase. Here’s a graph of the numbers, lumping computer science and related Bachelor’s degrees into one category:
The gray lines are for the other 98 R1 universities. The fit is to exponential growth, exp(growth rate x time), with a growth rate of 0.26 year-1, or +30% per year if expressed as an annual percentage increase. If the trend continues, the number will match the Berkeley undergraduate population, about 32000, in 2032, the population of California in 2059, and the Earth’s population well before the end of the century, though saturation may kick in before then.
Though Berkeley is near the upper end of growth in computer science degrees, the average of all the R1 schools’ annual percentage growth rates is still stunningly large: +18%, or a doubling time of 4 years.
You’re probably wondering (and some of you have asked me):
- How does the number of Computer Science degrees awarded compare to the total number of degrees awarded?
- What does the graph look like for other schools? Are some universities particularly strange? (I already looked at the University of Oregon, which is unusual.)
- What if I didn’t lump all Computer-Science-related degrees into one category?
- Do doctoral degrees have a similar trend?
- What does this mean for how we structure undergraduate education? Should it matter?
Let’s look at these questions. First, some details about the data that you can feel free to ignore:
- As noted in the last post, the data come from the Integrated Postsecondary Education Data System (IPEDS). I generated and downloaded csv files for each year from 2011 to 2021, and wrote a program to extract numbers from the tables, fit curves, and make plots.
- As noted, I amalgamated several degrees into this “computer science” category, as I did for all categories noted in the last post, combining nearly everything in the “Computer and Information Sciences and Support Services” Section 11 of the NCES CIP codes. The majority of degrees awarded, though, are “computer science” itself.
- Here are some examples of what the raw data look like. I’ve pasted cells from the IPEDS csv files that my analysis code analyzes, picking UC Berkeley degrees from 2021, 2016, and 2011. Note that I’m including second degrees (double majors) as well as first — I’ll return to this shortly. The last column is the number of degrees awarded.
How does the number of Computer Science degrees awarded compare to the total number of degrees awarded?
“Total” is not straightforward to define, since some students have more than one major. IPEDS notes “first” and “second” degrees. We could consider just the total “first” degrees as a measure of how many students there are, or the total first plus second degrees as a measure of how many degrees there are. Note that in my prior post, I gave a relative measure of degree growth based on “first” degrees. Does the distinction matter? Not really. Again, here’s UC Berkeley (magenta) and the other R1s (gray with two colorful mysteries), with both normalization schemes:
The growth rate in comp. sci. degrees relative to total “first” degrees awarded is similar to its absolute growth rate, a staggering +27.5% per year, with the R1 average being +15.6 % per year. The graph shows some other striking features: There are two schools (yellow and green) at which more than 30% of degrees awarded are computer science or some variant! I hope to make a quiz about degree trends, but for now here’s just one question: Guess the universities! (The answer is at the end of the post.)
I’ve made plots of number of degrees in each category relative to the total, combining the numbers for all R1s — in other words, the fraction of CS degrees, physics degrees, journalism degrees, etc., over the past decade, but I’ll keep this post’s focus on Computer Science. Let me know if you’d like to see these other graphs — there’s certainly enough material for another post!
What does the computer science graph look like for other schools?
Here you go — a zipped folder with graphs of the Computer Science (and closely related) degrees for each of the 99 R1 universities, from 2011-2021.
I make no guarantees of accuracy — see the note in next item, about Princeton, for example.
As another example, here’s the University of Chicago:
What if I didn’t lump all Computer-Science-related degrees into one category?
For Berkeley, note from the snippet of the CSV file shown above that there were 117 computer-science-related degrees awarded at UC Berkeley in 2011 and 1411 in 2021; of the latter, 797 were in computer science itself and the rest were “other.” I would guess the others are degrees in the new Data Science program, but I haven’t looked into this. Even ignoring the “other,” computer science itself has increased 600% in a decade. Here’s the graph, again showing all R1s and keeping Berkeley highlighted:
The annual percentage increase: 25.0 +/- 2.1 % for UCB; +18.3 +/- 6.8 % for all R1s.
The drop in 2021 is, I suspect, Berkeley’s new Data Science degree cannibalizing Computer Science.
Conversely, my categorization may include too little. As noted above, a few universities’ trends look strange. For example, Princeton shows up as having zero computer science degrees before 2019. Looking at the CSV files, this is technically correct, but digging more I find degrees with CIP code “14.0901 Computer Engineering, General” that, in my bins, fall into the “Engineering” category rather than “Computer Science.” Was this degree equivalent to Computer Science? I could count it, but there are schools with sizeable number of majors in both “14.0901 Computer Engineering, General” and “11.0701 Computer Science,” implying that in many places the former is not the same as the latter, as the name would also suggest.
Do doctoral degrees have a similar trend?
No.
Here’s the graph for all R1s, again with Berkeley highlighted. There’s a slight increase (3.9 %/year); at all R1s: +3.4 +/- 4.4 %/year.
What does all this mean for how we structure undergraduate education?
What does a Computer Science department do when its majors grow by an order of magnitude in a decade? (This is a problem many other departments would love to have!) I would guess that it puts a lot of strain on teaching and leads to very large classes. I don’t really know, though. I interact a lot with Biology and Chemistry faculty at my own and other universities, in addition to my own Physics colleagues, and I also run into a good number of faculty from other departments in settings like committees. However, I hardly ever talk to computer science faculty. (I’ve met fewer than 10 in 16 years.) I don’t have a good sense of what the culture of computer science higher education is, therefore.
It certainly is the case that the number of undergraduate majors is, and should be, a determinant of a department’s size and influence. What form this relationship should take isn’t obvious, though. Should a 10x increase in undergraduate degrees lead to 10x the number of faculty? What timescales should trends be averaged over? What other factors should set faculty hiring?
These questions are often raised, but I have yet to hear a clear answer to them, nor do I have one myself. (My university just announced three new associate dean positions. Surely they’ll tackle these tough questions. They are, in case you’re wondering, “an Associate Dean for Diversity, Equity, and Inclusion; an Associate Dean for Graduate Studies; and an Associate Dean for Research and Scholarship.” Note that we already have large apparatuses for Diversity, Equity, and Inclusion, Graduate Studies, and Research, but the only thing whose growth rate exceeds computer science’s is university administration.)
Returning to our story: The number of students who want to major in computational topics is so striking that it leads me to wonder if it could allow us to turn other enrollment issues on their heads. We would like to train more physics students, for example; what if rather than trying to recruit more physicists to the university, to study physics per se, we enticed more computer science students to study physics. Or, similarly, if we got the computer science students with a fondness for history to branch out into historical studies. The fraction that could be grabbed may be small, but the denominator is so large that the net result could well be greater than efforts to directly grow the ranks of physicists or historians. Something like this could be an outcome of Data Science degree programs, including a new one here at the University of Oregon. For this goal, however, and for the more standard data science goals of training students to handle real-world information, I and others think there needs to be more integration with Physicists and other non-computer-science fields. (My image analysis course could have been useful for this goal, but I only managed to attract one Data Science major.) We’ll see what happens…
Though I consider myself fairly aware of the landscape of modern higher education, I admit that I was shocked to discover exactly how enormous the rise of computer science degrees is. It marks a shift whose magnitude I didn’t really grasp, and I don’t think I’m the only one!
Today’s illustration
The coast at Heceta Head. Mediocre, but quick.
— Raghuveer Parthasarathy. February 11, 2023
Notes
Quiz answer: CalTech (yellow) and MIT (green).