Tuesday, October 21, 2014

Women In Computer Science

This NPR story, When Women Stopped Coding, has been making the rounds in tech circles.  Its main thrust can be summed up with the following graph:

The given explanation by NPR is cultural. As they see it, computers started being marketed towards boys, driving the girls out. This is the pretty standard explanation whenever someone discusses the disparity between the sexes in computer science.

However, the data they use for the graph is essentially "percent of a percent". As you know, I loathe this type of data. So I went to see what the raw data from the National Science Foundation says:

In my view, the raw data tells a different, and perhaps more interesting, story.

Essentially, the story of computer science for both men and women is that there were two bubbles. One around the year 1980, and the other around the year 2000.[1] Which maps to what happened historically. Although women are always less represented, both curves follow somewhat similar shapes. The major difference though, is what happens around the peak of the bubble.

First, in the two or three years right before the peak, a lot more men jump into the program than women. More women enroll, but there are even more men who look to join in. Women seem a little less inclined to flock to the newly "hot" programs.

Second, and more importantly, the bust right after the peak absolutely devastates female participation in computer science. The first time, in the 1980s, female degrees drop to less than half of their high point, while male enrollment only falls by 30%. The second time is even more destructive for women. All the gains from the boom are wiped out, and female participation falls back to the steady state before the boom. Male participation drops, but again, it doesn't drop all the way back down.

My interpretation of the data is that less women participate in computer science not because of cultural reasons, but because of economic ones. My hypothesis would be that more women avoid industries that are perceived to be "unstable", or have significant economic downturns, even if the industry is lucrative during boom times.

To be fair, that's a reasonably sensible position. I was in university when the last tech bubble burst, and it was a terrible time to hunt for work. I can only imagine what a high school student is thinking when they see the obliteration of large tech companies like Nortel on the news.

[1] You choose your program roughly four years before you get your degree.


  1. I would point to the fact that much of the computer-related stuff happens in non-profit or crowdsourced projects. Unless you code for banks, you are likely in the lowest income quarter among people with university degree. You are likely lower paid than high-school degree craftsmen.

    These jobs are mostly done by "nerds": enthusiasts working 60+ hours for peanuts. A man can much more easily donate his youth for such project than a woman. If he turns 40 with nothing next to his name, he can still fix his life. A 40 years old woman is likely unable to start a family, she must not throw away her youth.

  2. I think it's partly due to a less risk taking profile, yeah. What we learned during the last tech bubble was that being offered shares means nothing when the majority of companies are going to fold and not make you a millionaire. So it may well be a way more reliable strategy to go into something like nursing where the demand is steady and increasing. IT is a bit of a crap shoot and the more reliable jobs (maintenance, build manager) aren't very stimulating.

    Also it's actually pretty hard as a female engineer to get opportunities for management compared with the guys. I don't blame any female technies for noticing this and heading to a field where their skills are more likely to be recognised.

  3. I think there are also cultural reasons in play here, as the graphs don't precisely mimic the corresponding bubbles in tech.

    The two bubbles follow after the 80's Game Crash and the 2000 Dot Com busts. They don't align perfectly because people go into those degrees based upon what they were encouraged to follow in high school, and that crash took a while to percolate down to the colleges and high schools.

    But why is that the case? Why did women crash harder than men? I suspect it's because the women who did graduate into the years right after the bust had it rough enough that they communicated their problems to their younger peers more effectively. It could also be due to other cultural factors, such as gaming participation, but the fact that each degree bust echos a corresponding tech bust implies there are cultural factors at play.

    Tech in the US still hasn't recovered from the Dot Com bust, if you judge salaries and participation from grads. Other factors, such as cheap offshored labor, have contributed to those problems, and women have pursued other degrees (such as medicine) in greater numbers than men have.

  4. Spinks, I'd argue that it's hard for any techie to get into management without an MBA these days. It seems that tech has moved totally in the direction of using the MBA as the gateway to management, which means any techie who wants to move up the management chain has to submit to the soul sucking process prior to becoming eligible for a management position.

    I don't know what the participation rate is for women in MBA programs, and I'll freely admit that the company I work IT as a contractor has a much larger proportion of women in management than other major corporations, so my perception is skewed a bit. But after having spoken with some of my overseas colleagues, there is a lot of cultural bias against women in tech and especially in management. Here in the US, I see it in my (supposedly progressive) workplace. Hell, I see a lot of latent racism in the tech world, and I'm not talking the low hanging fruit of railing against foreigners "taking them good ol' 'Murican jobs!" either.

  5. @Redbeard, the lag is because you choose your degree 4 years before you get it. So a bust in 2001 means the high schoolers who would have chosen comp sci in 2001 choose something else, but that only shows up in the degree stats in 2005.

    The people who graduate in 2001-2004, when the bust actually happened, chose comp sci during the boom times.

  6. Rohan--

    That's exactly my point. And that the crash happened four years after each other recognizable tech crash means that grads didn't abandon their degree programs once things started to look grim, perhaps thinking that the crash wasn't going to affect themselves personally.

    In the US, it's quite common to change your majors if you find things aren't working out, but apparently the data is for only those who received degrees, not started them. I'd like to see what the ratios are of starting to finishing a tech degree over these time periods.

    Back to my original point...

    When the inevitable crash did happen, it was women who moved into other degree programs more completely than men did. That is where the cultural aspect comes into play.

  7. Stick to tackle Titans Gevlon: http://www.forbes.com/sites/susanadams/2013/04/15/college-degrees-with-the-highest-starting-salaries-3/

    A BS in CompSci isn't (shouldn't be) for coding. Coding skills are cheap, replaceable, and dead when a language goes out of fashion. Understanding technology and being able to guide a company with that knowledge is why people with CompSci degrees who position themselves correctly (Tech Lead, BA, SME) earn as much as they do, with basically zero risk of being outsourced.

    Workers in India/Brazil have an easy time learning coding, and will do so for less than someone in the US. They basically can't learn how to interact with Business people in such a way as to properly represent IT.

    That's the model today in most major companies with any significant need for IT, and those careers fit a woman just as well as a man (good pay, flexible hours, job security)

  8. ???

    Of course people overseas can learn to represent IT in businesses well. I know quite a few who do.

    A lot of it has to do with what hiring the right people, which is 95% of the struggle with any job. You don't hire a helpdesk person into a solutions architect job unless they demonstrate they can perform the task at hand, just like you wouldn't hire any old advertising person to work in a biochemistry lab.

  9. I think you missed the point.

    Coding is a cheap skill; there are plenty of people all across the globe who will do it for less than someone in the US, and so long as the code itself is solid, you generally don't care who wrote it.

    Having the mix of skills required to be an effective TL, BA, or SME, or other positions a CompSci major fits into isn't easy. You can't take a month-long course and hit the ground running like you can with coding.

    Someone with that skillset is in demand, because it's fairly rare (especially compared to coding). It's also somewhat region-specific. Someone with a heavy accent or who isn't completely fluent in English is at a disadvantage here, because a large part of the job is communication and 'selling' people on the IT perspective.

    Now, lets say someone in India has all of the above (we work with a few, though they are more at the junior level); they don't come cheap, and that cost is increasing. It's debatable, but some would argue we are already at the point where the minor cost savings doesn't justify outsourcing wholesale. A lot of companies are bringing a lot of these positions back in-house now, or changing where they go to outsource (certain EU countries or South America vs India/Asia), or how they do it (use a 3rd party for the coding/development, keep the rest of IT in-house).

    My point, now with a lot of words behind it :), is that someone with a CompSci degree, male or female, is in a good position right now so long as they use it correctly (which is true for any degree of course). If you pay $200k+ for a CompSci degree from a decent University and then become a QA drone in the gaming industry, yea, life is going to suck for you, but that's more on you than the limits of the degree.

  10. I would like to see the equivalent graphs for the other professions, I might do that after work. But either way, there had to be a shift in the other distributions (doctors etc), no? What is your theory as to why CompSci not have the equivalent shift? Is it because the other fields were perceived as less economically risky choices?

    1. A shift in the other distributions to make up for people going into CS? It's probably not noticeable because the number of people in CS is likely much smaller that the people going for law or medicine. The NPR graph shows percentages, so the absolute scale is masked.

  11. Also, I don't quite get the linked elaboration of percent-of-percent data because I'm not familiar with the WoW terminology used for examples, can someone explain in terms more widely known?

    1. It was just a rant about people choose "unusual" values, which are technically true, to give a false impression of their argument.

      For example, in the NPR graph above, it looks like the maximum in 1984 is important, and the critical turning point, because that's what we usually look for in a graph.

      But because it's a graph of percentages, the important point is not the maxima, but the times where the slope of the line is the steepest, because those are the times where the ratio changed the fastest. And if you look at the NPR graph, the steepest slope occurs around the year 2000.

      In general, though, using a percentage where the denominator changes wildly is poor form. It is likely to give misleading results, because the change in the denominator is probably more important than the change in the numerator.