Slight note about Dan Wang picking 2005: That was the peak of CS degrees awarded because it's 4/5 years after the height of the dot-com bubble. So the upward bump in the mid-2000's is somewhat explainable as an anomaly.
I think his point 1 is underrated. CS degrees are flat because aptitude is flat.
You can compare CS degrees to other degrees over time at nsf.gov:
We have more grads than ever, but they are dumber than ever (we have the data to prove this), getting less difficult degrees.
I have a bad feeling that we are running up against some diminishing returns on education and hiding it with numbers like the total number of grads. The number of grads for difficult degrees and the quality of grads seems to be another story.
> In 1970s 1-in-2 college grads aced Wordsum test. Today 1-in-6 do. Using that as a proxy for IQ of the median college grad, in the 70’s it was ~112, now its ~100.
Re: the choice of 2005 as origin, if you rebase on 2009, then CS degree growth looks on par with other STEM fields. Thus I think is's misleading to say CS majoring is "flat". At a certain end-to-end view it is, but that obscures that it crashed, and is now rapidly growing, not that it's been stagnant and unchanging.
In the early 2000s it was widely perceived as the economically smart thing to go into a technical field other than CS, especially petroleum engineering, chemical engineering, and similar areas, or maybe even better, a non-technical field like law. They had higher pay and were seen as more stable employment options. CS employment was seen by many people as unlikely to ever fully recover from the dotcom crash, partly because of mass outsourcing. So unsurprisingly, new enrollment numbers were low, and some departments at smaller schools even closed. Then from around 2005, CS began to be seen as a lucrative again, and enrollment has been steadily rising every year since then. Universities have responded likewise by reopening CS departments, hiring more faculty in existing ones, and increasing class sizes (so enrollment growth is actually considerably outpacing faculty growth).
This does not look to me like a market that fails to respond to economic incentives, but exactly like one that does respond. Maybe it ought to respond even faster, although if it did in general, the CS-degree crash of the 2000s might've been even deeper than it was.
>if you rebase on 2009, then CS degree growth looks on par with other STEM fields. Thus I think is's misleading to say CS majoring is "flat".
No, what you described is called "misleading with data" at best and "lying" at worst. There is a clear dip in CS on that plot that is suffered but none of the other points. Moreover, the growth is below the rate of increase in college graduates.
The clear dip in CS is what the rest of my comment is about. The point is that the entire premise of the original post, which claims CS enrollment is "flat" by cherry-picking a specific benchmark year and ignoring huge movement between that year and the present, is indeed "'misleading with data' at best", because CS enrollment is not "flat". If you pick an earlier or later year, you get different narratives.
The post then spends a bunch of time musing on why CS enrollment supposedly doesn't respond to economic incentives, even though responding to economic incentives, i.e. rising and falling together with tech booms and busts, is precisely what the data shows it doing.
Did you read the post you are accusing of "misleading with data"? He describes that exact phenomenon the sentence after you quoted. It's completely ridiculous to cut out context and then him out for not mentioning the context you just cut out.
Yes, I know it seems silly now, but this was a very real scare that CS would be a worthless degree in the future. I was dissuaded from majoring in CS during the height of this, but ultimately changed my major back to CS when I realized that I was actually really good at it.
In that time (early 2000's), even if you had a strong interest in computers, you'd be pushed toward Computer Engineering instead of CS. Business school was a popular alternative with MBA enrollments peaking in the early 2000s.
Where I went to college, UMBC, Computer Engineering covered enough of Electric Engineering and Computer Science that it set you up for either a career or further education in either. In my opinion, it was the best value.
I did in fact have considerable pressure from family to change my major from CS to some form of Engineering in this time frame, ideally some "hard-science" engineering like Chemical Engineering. CS was seen as a dead-end, easily exportable coder job, and well-meaning relatives really wanted me to get an engineering degree instead.
> The number of grads for difficult degrees and the quality of grads seems to be another story.
This is why I never believed the "There is no STEM shortage" stories. They were mostly based on the assumption that all STEM grads were qualified to work in STEM. That's not true in any field where you can complete a degree with a C or D average.
I find it unfortunate that some of the larger software companies filter out applicants based on the school they graduated from, but I understand it. The companies have to have some sort of filtering mechanism, because many colleges clearly do not.
>The companies have to have some sort of filtering mechanism
There's no one to hold companies accountable for having a unjustifiably high or low filter. And frankly I don't think anyone outside of CS has gotten those filters right any more than CS has. An incorrect filter turns into "Well, there's no one qualified to work for us" or "Well, all of these candidates failed to do the job" which leads to incorrect assumptions about the labor pool, which is what's going to bias your hiring decisions.
There are companies out there hiring for A+ students that would be just as well off with a C or D average student because what the company is doing day-to-day is too far from what university was like.
Do you actually need millions for a software consultancy? You're hiring people who wouldn't be employed elsewhere. You could pay them less and cut prices to your clients and everyone would come out better.
I mean you've spotted what you claim is a clear inefficiency. Your guys could be taking the jobs that Infosys and gang are taking.
You are assuming the companies care enough. Their current strategy is petition the government to open offshoring quotas. That's much easier and cheaper than actually figuring something out.
> This is why I never believed the "There is no STEM shortage" stories. They were mostly based on the assumption that all STEM grads were qualified to work in STEM. That's not true in any field where you can complete a degree with a C or D average.
This is true IFF we assume that gpa accurately reflects qualification to work in industry.
There is a difference between getting a C or D average because you don't care about school, and getting that because you can't do any better. Unfortunately the two are hard to separate, and even if the grad didn't care about school - it doesn't mean they will care about work.
I was in the boat of not caring about school (although I did get a B) and it worked out - but anyone following that path should understand its an up hill climb.
> NYC and SF are small islands compared to the rest of the country.
"Salary" is the wrong metric to consider. The "interesting" metric is "salary minus cost of living". And this is stagnating even in NYC and SF, since the costs of living are rising.
I think the point is a huge number of tech jobs are within these locations. The SF bay area has something stupid like the top 3 cities when it comes to tech jobs. Throw out all the locations with the most jobs and surprise, salaries suck.
Well, I wouldn't say salaries "suck" outside of well-known tech centers like SF/Bay Area and NYC. They aren't great, though. Even in the Bay Area, at least, the salaries of engineers still pale in comparison to the salaries of managers and executives.
Maybe not suck as compared to a lot of jobs, but the problem with software development seems to be that salary growth rises rapidly up to a plateau and then basically stops. The opportunities for salary growth past a certain point require risk-taking such as starting a company, consulting in some very specific niche, or what have you. It's still better than the lack of wage growth in other fields, I'm sure, it's just not something that was explained to me AT ALL when I started down this long road many years ago.
Absolutely. Even the so-called "parallel track" technical equivalents to management generally have a lower pay scale and smaller bonus pool. In terms of career development and growth it's almost always better to be on the non-technical career path once you're at Lead Engineer or higher. Even director-level people on the technical side of the tree are generally viewed as (relatively) more expendable and lower-status than their equivalents on the non-technical track.
> Even in the Bay Area, at least, the salaries of engineers still pale in comparison to the salaries of managers and executives.
Depends on where you work. If you work at an actual technology company that "gets it", they'll have parallel management and individual contributor tracks where you can make as much as a director-level person as an IC, or more.
Emphasis on "can make" is due in this case. It's still a different dynamic with respect to status and a different pay scale in general (that is to say, exceptions of course exist).
Re: the Wordsum test, I can't believe it is seriously used to assess IQ. If we assume that all participants know all of the words used, then maybe it would have some merit. But so many of the words used are either archaic, little-used, or culturally specific to North America that you could only reasonably expect a well-read and/or well-educated (and therefore wealthy) American person to know them.
It still likely correlates well with IQ. That's the whole point of IQ. A bunch of different cognitive tasks like solving puzzles and memory tests tend to correlate very well with each other. I'd be very surprised if wordsums didn't predict IQ very well.
It's hard to believe a person of high IQ is not going to be well-read and/or well educated.
The only point about this test is it's specific to native speakers of English, but if you were born in North America than it surely applies to you even if your parents speak another language.
> It's hard to believe a person of high IQ is not going to be well-read and/or well educated.
What of high IQ individuals who are born into low- or middle-class families, who didn't have their parents read them 19th century British novels, and therefore don't know what a dowager is?
That doesn't match what I see in the ground. I get 10-20 interns a year on my teams from big state universities and liberal arts schools with CS and stem degrees.
They tend to be awesome. The best ones outperform most of the consultants you find.
Nothing you say relates to IQ stats about the median grad. If CS enrollment was up, it would push the needle I'm sure. It's not. You may as well be saying, "What do you mean most wine is bad? I drink good wine all the time."
I'm getting a random lot of 20 kids from average schools. Surely, if IQ is dropping significantly amongst the general population of CS students, I'd be seeing that?
The author's point was that college IQ was dropping, and this explained the lack of rise of CS enrolemt. Simce CS is objectively harder, only a lower portion of college students can now do it.
That's their hypothesis. So their hypothesis isn't that CS student IQs are dropping.
Is your company actually selecting 20 kids at random, or selecting 20 from the best of the ones applying?
Are you also comparing this over a long timeframe? 112 in the 70's -> 100 now means less than 1 point every 3 years. You probably can't notice a difference that small year over year.
Of course, "the plural of anecdote is not 'data'," is the first thing I would say. I know someone who just went through the process to hire a half-dozen interns, some from top schools, who (frankly) would argue with you. There are a thousand possible reasons for this but I think the best course of action is to dispense with that, personal, line of reasoning.
Interesting that in the nsf data it shows that during the 2000-2004 time frame the number of CS grads rose and then peaked in 2004, while the number of CS women grads during the same time frame was flat, and then dropped in 2004.
'04 was just post dot-com bubble, though. Sad if those numbers haven't picked up since then. We badly need people who are doers and not just rent seekers or mid-level functionaries.
institutional education is subject to diminishing returns at every level of the process. it's a fractal of diminishing returns.
the "college for everyone" experiment is grinding our society into paste and generating enormous debt burdens for an entire generation. time to end it.
those with aptitude for programming can learn programming without a college degree. let's find out just how flat aptitude really is. all we know now is that aptitude is flat in the institutional environment.
What's interesting to me about that data is the high rate of biological and agricultural sciences grads, relative to all other sciences. CS grad rates seem to fall right about where they should on XKCD's Purity Scale [1] (if CS had been included in it). 2004 was an outlier for CS, but it has reverted to its mean since.
But bio and ag sciences are beating out Psychology, which is a bit of a reversal. Biology isn't easy, especially not any program with a requirement for organic chemistry. Could that be b/c the big midwest and southwest universities (Iowa, Texas A&M, etc) have large bio/ag programs to support that industry in their region?
Although biology and organic chemistry are much harder then CS anything, they can be learned in school and from school resources only. No one expects you to already come in already knowing how to code or configure linux. If you are good students willing to work, you will learn them, period. With cs, you are expected to already know a lot that was not taught in school - meaning that all kinds of cultural and social effects play the role (who your parents and friends were).
Also, there is a lot less myths about biology around. Nobody assumes you have to be some kind of nerd to learn it, nobody assumes there is special in-born ability for it totally different from all other kinds of intelligence (like is often implicitly assumed even with things like operating system configuration), there is much less cultural bullshit about "hackers and their culture" around.
Organic chemistry looks complicated but it is, by and large, just an enormous amount of memorization. The kind of complex systems reasoning you have with any non-trivial computer science isn't there.
By contrast, even though it is the same subject matter, chemical engineering is (correctly) perceived as being vastly more difficult than chemistry because it requires reasoning about complex system-level behaviors that has no analogue in chemistry. This creates the oft-observed effect that being skilled at chemistry has surprisingly low correlation with being effective at chemical engineering despite being the same domain.
Most programming jobs are easy, it takes effort and luck to find one that is not. Things like system administration and configuration are largely about remembering thing. There are some aspects of it that are harder and require being good at math - by you dont need them on practical jobs and they are still easier then real math.
But realistically, a.) memorizing that much is hard b.) my friends who studied chemistry said that it becomes much easier when you understand how it works.
trying to learn organic chemistry by memorizing mechanisms is like trying to learn maths by memorizing derivations... it might work for an intro course but you will quickly hit a wall.
chemistry and chemical engineering has very little to do with each other, despite the name. chem eng is focused on scale up while (organic) chemistry is focused on novel mechanisms
I haven't personally taken organic chemistry, but several of my friends who have taken both organic chemistry and algorithms have said that algorithms is by far the harder of the two. So I tend to disagree that biology and organic chemistry are harder than anything in CS.
Eh, my college girlfriend majored in Chemistry while I was CS and Math. P. Chem was certainly a huge challenge, but I never got the sense that she was smarter than me, just maybe more determined in certain ways and better at memorizing shit. She got a 4.0 and was valedictorian, but she dropped any elective classes if they looked like they would not result in an "A" for her--I especially remember her dropping Calc II, for example. So she got her 4.0 by successfully gaming the system, whereas I never got overly focused on grades as the ultimate marker of successful learning.
Btw, I have done quite well in life, not sure where she ended up; probably chasing some other kind of "grade" these days, lol.
I have no particular opinion on harder/easier. (There is a lot of memorization in organic chemistry earlier on because you don't have the tools to come up with answers from first principles at that point.)
But, unlike CS, there isn't this assumption that you've played around with organic chemistry and/or chemical engineering in your spare time if you want to major in it in college.
As someone who graduated with an engineering degree from one of those large Midwestern universities, yes. There are millions of acres of farmland and tons of ag companies who are making developments to eek out another 1% yield. At my university, the ag and life sciences college was only recently overcome by engineering to be the largest college.
In Michigan the universities have a lot more focus on automotive because that's what runs the state. Agriculture runs a huge portion of Iowa, Missouri, Kansas, and other nearby states.
I think his point 1 is underrated. CS degrees are flat because aptitude is flat.
You can compare CS degrees to other degrees over time at nsf.gov:
https://www.nsf.gov/statistics/2016/nsb20161/#/report/chapte...
We have more grads than ever, but they are dumber than ever (we have the data to prove this), getting less difficult degrees.
I have a bad feeling that we are running up against some diminishing returns on education and hiding it with numbers like the total number of grads. The number of grads for difficult degrees and the quality of grads seems to be another story.
> In 1970s 1-in-2 college grads aced Wordsum test. Today 1-in-6 do. Using that as a proxy for IQ of the median college grad, in the 70’s it was ~112, now its ~100.
More stats: https://medium.com/@simon.sarris/why-is-computer-science-enr...