a very basic point about jobs in the tech sector

Since this is a conversation I just can’t drag myself out of– this brief post by Alan Jacobs, about yet another person claiming that teaching kids to code will be our salvation, inspires me to highlight a point I’ve made a few times before. Many people are making some version of the argument that, in order to solve our unemployment problem specifically and our economic woes generally, we need to push students to pursue majors that are attractive to employers in the technology sector. There’s a lot you could say about that. I’ve argued at length that, in the short term at least, there’s no reason to believe these employers are actually having a hard time finding skilled workers. I’ve also pointed out that not everyone has the talents or temperament to succeed in these fields. Plenty of people go to college fully intending to get degrees in computer science or electrical engineering or similar and find it too hard to finish. To make matters worse, the big firms don’t want the median graduate. They want the stars. Great for the stars, but bad from a macroeconomic perspective.

But here’s a more important point: major tech firms are such a financial success in part because they employ very few people. When people talk about startups as being “lean,” when the culture celebrates how easy it is to get a tech startup off the ground, when the idea of a few college kids building an app in their dorm and becoming millionaires is a common trope, each of these things are indicative of the low labor costs involved in the tech industries. Yes, individuals can become fabulously wealthy in these firms. But they can in part because the pot is being divided between a much smaller number of people. Even the tech firms that we think of as giant companies have far fewer employees, relative to their revenues, profit, and power, than their older analogs. Last time I checked, Google had a market cap six times that of GM but employed less than a quarter as many people.

Now there’s a whole conversation we could have about capital-biased technological change and automation and the future of work. But in the most limited sense: it seems perverse, to me, to be pushing tons of kids to pursue work in an industry that is so dedicated to keeping labor costs low. And the push to further automate is only going to intensify; while I’m a skeptic towards the more outlandish claims of a “Singularity” and think people broadly overestimate the current state of artificial intelligence, there’s little question that more and more tasks are going to be performed by computers rather than coders. In other words, it’s quite reasonable to believe that the very job-killing disruption that computer scientists have created will come to affect computer science as well. That’s simply taking the logic of automation to its natural ends.

There is no connection between a field’s media visibility, or our vague intuition that a field represents “the future,” and the future job prospects of people in that field, and there’s never been any rational reason to assume that such a connection exists.

10 Comments

  1. Even the tech firms that we think of as giant companies have far fewer employees, relative to their revenues, profit, and power, than their older analogs.

    Partially it’s the nature of the businesses they’re running. Web stuff like Google and Amazon is pretty centralized – if Walmart could somehow remotely operate their stores from a single hub in Arkansas, they’d probably employ a lot less people too. GM and the car manufacturers would probably employ less people if they didn’t have to build local plants in other countries to get around tariffs as opposed to simply exporting cars there (same thing for Japanese and Korean firms in reverse).

    And the push to further automate is only going to intensify; while I’m a skeptic towards the more outlandish claims of a “Singularity” and think people broadly overestimate the current state of artificial intelligence, there’s little question that more and more tasks are going to be performed by computers rather than coders.

    I don’t think we’ll have employment issues due to automation, but I do agree that it’s a little perverse to be pushing kids in the direction of certain kinds of jobs. The US has no lack of educational institutions with strong science and technology programs – if the demand for those jobs is there, then they’ll re-adjust their offerings to reflect that like in the 1990s. It’s sort of like how lawyer programs finally started dropping off in enrollments due to relative lack of demand in the job market for attorneys (showing up as reduced attorney pay and job security).

    Long-term, I suspect we’ll see a ton more administrative/robot shepherd and legal compliance monitoring jobs in office. Even with increasing automation and office software, those jobs have been creeping upwards with the increasing complexity of our economy.

  2. There are 140,000 HB1s doing middle-class jobs that are only medium-skilled. Just about any reasonable bright person could do them. But, we import them from other countries while smart history majors sling coffee…

    1. Well, first, despite your assertion, there isn’t consistent or compelling evidence that history majors (or other humanities majors) suffer compared to the median graduate or even to STEM graduates. When you make claims of fact you are obliged to show your evidence. Second, we import HB1s because tech firms want to put downward pressure on wages. H1Bs are not about a shortage of workers. They’re about lowering the labor costs of domestic workers.

  3. Your words have certainly opened my eyes to pushing every child into a STEM field, but in regards to automation affecting the computer science field, I think the opposite is true to a degree. Computer software rarely stays static. Just about every application in use is shipped with known bugs or missing features slated for a later release. Computers only do what humans program them to do. As systems become more complicated to handle more complicated task, even thinking on their own, they need larger teams of people to build and maintain them. The larger an application gets, the harder it is for one man to maintain.

      1. And I say “to a degree” because eventually we will have computers that program computers(we sort of already have it). When we get to the point to where a human is no longer needed to develop the code to run chip or processor, or other low level type of software architecture, is when software engineers may start to see those same affects of automation.

        1. we sort of already have it

          What do you mean? Meta-programming? That hardly substitutes for people; it just makes certain people more productive (or at least look more productive).

  4. The past couple weeks have involved some bad news for techie friends, so I continue to come around to your point of view.

    That said, I do think that the tech economy is such that low-level coding skill is more of a general purpose asset and not something specific to the tech industry. Albeit still with restrictions to those with the right talent and temperament as you mention. The ability to fiddle with a website can help in a lot of places much like the ability to right cogent and interesting summaries of events and meetings or any number of other widely applicable skills.

    In some cases, this may just be a zero or even negative-sum asset, from an employment perspective, as knowing automation tricks can just decrease the labor intensity of the field in question. I think the argument to learn to code makes the most sense if you’ve got a talent and temperament alignment and if there’s something in your field of interest that’s in high demand that has only arcane or kludgey technical support. That sort of connection could increase the demand for labor, albeit moderately, but requires knowledge of the domain in question that a “code day” wouldn’t help with. That said, I think like the ability to do a good summary, coders do at least have the advantage that if they don’t learn the right software package for the problem they’ll face, they’ll still have an edge in transferring their skills to a new coding task. I’d argue that for simple scripting jumping from C++ to SQL is still a lot easier than even switching between romance languages.

  5. my understanding of that world is very limited.

    People who know what they’re talking about do make the same point as you. Eg:

    http://blogs.law.harvard.edu/philg/2013/12/12/why-do-people-who-chose-not-to-study-science-and-math-opine-on-the-virtues-of-studying-science-and-math/#comment-200713

    http://www.computerworld.com/s/article/9245494/What_STEM_shortage_Electrical_engineering_lost_35_000_jobs_last_year

    (or you can find twitter.com/zentree is a science-sceptic and code-sceptic …)

    It’s got to be obvious even to an outsider that the reasoning underlying “growth = tech = computers = jobs” is very weak. Each of those =’s is doing a ton of conflating and you won’t see this reasoning worked out with numbers.

    (Like: how many children are going to become software engineers; how many jobs out there; how many new are they supposed to be generating; how much demand for iOS apps could reasonably be expanded…. Or some other quantitative argument that would take into account anything other than weak associations.)

    ((This comment also relates to your statistical hands, rhetorical hearts theme. The quantified arguments I’m talking about don’t involve any higher maths at all. Just where numbers come from, how they relate to the discussion at hand, and reasons why in the future they might be constrained to eg do no more than double—or whatever. This is spreadsheet maths = regular person maths; not “real maths”. No Greek symbols, just the kind of logic I’m sure humanities people already possess.))

  6. To make matters worse, the big firms don’t want the median graduate. They want the stars. Great for the stars, but bad from a macroeconomic perspective.

    On the other hand, many non-stars would benefit from knowing a bit of a scripting language, version control, and a few other programmery things.

Leave a Comment

Your email address will not be published. Required fields are marked *