CS Degrees Are Mostly Just Signaling — An Interview With Economist Bryan Caplan

BY CHARLES TREICHLER

The main thing I’d say about computer science is that [programmers] have a self-concept of being totally skills-driven, but if you actually look at employment in computer science, that’s not how it works. Degrees from leading schools really do seem to matter. They really do seem to open doors.

In his controversial book, The Case Against Education, Dr. Bryan Caplan, Professor of Economics at George Mason University, uses statistical analysis to argue that our eduction system is a big waste of time and money. But he isn’t suggesting that you should drop out. Data shows that education pays big dividends (the so-called education premium), and Caplan does not dispute that fact.

What he does question is WHY education pays. Rather than providing practical skills, Caplan argues that education serves primarily as a signal to prospective employers that a given candidate is likely to be a productive employee. That’s why education can be both a “big waste” for society (because it’s an inefficient way to filter job applicants), while also being extremely lucrative for the individual (because it IS the way our society filters job applicants).

We recently spoke with Dr. Caplan and asked him to explain his thesis and to discuss how it does (or does not) apply to practical fields like computer science and engineering. We also got into topics like changing child labor laws (at least in CS) and finding more efficient ways to filter candidates.

Note: This interview has been edited for clarity and style.

Why Education Is a Waste of Time and Money

Why is most of education a big waste of time and money?Most of what you study in school you are highly unlikely to ever use in a job — if you even remember it. A lot of what I do in the book is go through the high school curriculum and compare it to what people actually use in the real world and show that the connection is much weaker than propaganda would have you believe.

I do the same thing for college majors to show how few of the classes you take are actually relevant to a job that you’re ever likely to get. [This is true] even for many college majors that sound vocational on the surface — like psychology or communications. When you look more closely, you see that there are so few jobs in those areas that the vast majority of people who get majors in those fields have to find a job someplace else because it just doesn’t arithmetically work out for them to actually end up in the field that they studied.

Then why do people with more schooling earn more money?There are two very different ways that education could cause earnings to go up. One is that it actually teaches useful job skills. And, of course, education sometimes does that, and for that there’s no real intellectual puzzle. It’s pretty obvious why someone who can read is a better worker than someone who can’t.

But what’s interesting to me is, why does all of the other stuff that doesn’t seem useful also pay? And there, my explanation (which is not original to me — it’s been around for a while, but I push it much harder than almost anyone else) is that even studying the most useless subject is a good way of persuading employers that you’re a good employee. It’s a way of convincing them that you’ve got the brains, the work ethic, and just the sheer conformity to get through the whole process. One way I like to describe it is that most education is not job training, rather, it’s a passport to the actual job training, which occurs on the job.

So you should still get that piece of paper?Yeah, the whole idea of signaling is that you want to get in with a group of people. You want to join a successful group, the most successful group that will have you as a member, so you can say, “Hey, look at me! I’m part of this group.” If you’re a highly able person you don’t want to go into low status things because then you get stigmatized. Or, if not true stigma, then at least a relative stigma: “Sure that’s good, but it’s ‘pretty good.’” And if you’re awesome, then you don’t want to be in the “pretty good” group, you want to be in the “awesome” group.

Educational Signaling in CS

Does your argument apply to fields like Computer Science where you need to know some fairly concrete things to be successful?The main thing I’d say about computer science is that [programmers] have a self-concept of being totally skills-driven, but if you actually look at employment in computer science, that’s not how it works.

Degrees from leading schools really do seem to matter. They really do seem to open doors. You can’t simply go and drop-out of high school and take a test of programming competency and expect to get hired at Google or Microsoft or Facebook or some other prestigious place.

I want to push back on that a little bit. We have a number of people on our own engineering team who dropped out of CS programs to take jobs with us. Also, our platform has done a pretty good job placing software engineers who dropped out, even at major tech companies.To me, the fact that you have a bunch of people who dropped out of their CS programs working for you isn’t that much more surprising than an insurance company with a bunch of people who didn’t finish their English degrees. These people have to find jobs somewhere.

The important thing to remember is that drop-out rates are really high in general and just because you dropped out doesn’t mean you’ll never get a job. In my mind, the question is just how much better off would that 25% of people you’re talking about be if they had stuck with it? On average, it seems pretty likely that having a degree would have opened up more doors.

[You could ask,] isn’t it possible that [the tech] field is changing so rapidly that the usual rules don’t apply, and maybe we can actually break out of this equilibrium? I hope you’re right, but my default is always that there isn’t going to be any big change. The education system as we know it has been around for about 1,000 years. If something has been around for 1,000 years, and someone says, yeah, but it’s breaking down right now, that’s the kind of thing [that makes me think] like, is it really? I’d like to see it break down, but we gotta separate our wishes from our predictions.

On the other hand, there’s about a 1.9% unemployment rate amongst programmers, and lots of talk about a “developer shortage”[1]. Do you think tech companies are just so desperate for people that they might be willing to overlook a lack of credentials?I mean it’s just a question of, will they overlook it only in the employment decision or in wages as well? It is true that it’s easier in Computer Science to go and get a job just because you know the stuff, but there are a lot of reasons why someone who’s great at coding could still make a bad employee, and these are the things that employers use credentials for in order to weed people out. Which makes sense if employers are worried that, even if the person is a competent programmer, [the fact that they didn’t graduate means] that they’re not a team player, or they’re unreliable, or have a problem following orders.

That said, dropping out of a CS program to work in tech is almost part of the mythology in Silicon Valley. So I wonder if the social stigma is less here? Some early-stage startups might even want people who are a little non-conformist?Maybe. Really I would just like to see the data. Is it true that people who have the same tested skills but very different credentials wind up doing just as well in terms of earnings? Or do people who have the skills but not the credentials have to settle for a lower career path? That’s the main thing that I’d be looking at.

It seems hard to believe that, on average, you don’t wind up making more money and having a better career if you just stick with the program and finish it, although, of course there can be stand-out exceptions to that rule. It’s just a matter of whether you should let those exceptions make you forget there’s a rule.

The Wage Premium in CS

Do you think that the percentage of the wage premium that can be attributed purely to signaling goes down in more technical fields like CS?That’s a good question. I’d say my honest answer is, I think so, but I’m not completely convinced. So here’s the thing: the total wage premium in tech fields is higher than for other ones. It’s still possible that the ratio of the signaling share to the skills share could be the same, but you’re just getting more skills. For example, if a CS degree has twice the percentage effect on your earnings as a business degree, say, it could be that he’s learning more skills in CS, but it could also be that CS sends a better signal, it’s more convincing in terms of the percentage of the pay-off from just saying, “Oh wow, look at me!” versus, “I know skills.”

People have often noted that one of the best ways to profit from your physics degree is to go to Wall Street and get a job doing finance. They work in an area that has absolutely nothing to do with physics, but the physics degree opens doors.

Coding Bootcamps

There’s a lot of focus right now on coding bootcamps. I know you’re a proponent of vocational education, so I wonder what your view is?[Bootcamps are] very good for society. This is something that tax-payers don’t have to pay for, and people are acquiring real job skills.

For the individuals concerned, I think a lot of it depends. If you are the kind of person who just couldn’t stand to go through a regular, conventional education, then it’s probably a great option for you, and you’ll at least do a lot better than you would have without it. On the other hand, if there were kids saying, “I got into Princeton — should I do Princeton or bootcamp?” I think I’d say, “From your point of view, yeah, go to Princeton because I think Princeton will open up a lot more doors for you than the bootcamp will.”

Innate Ability and Motivation

Do you think some people have the innate ability to be programmers and other people don’t?There’s two things. [First] there’s a difference in ability. There’s work saying that besides general IQ, there’s also specific differences in quantitative versus verbal abilities. Think about the SAT. Math and verbal correlate — we have a high positive correlation — but it’s far from perfect. So there are people who are very quantitative, but not so verbal. For computer programmers, it’s clearly more important to have good quantitative skills than good verbal skills.

Motivation is also a big deal. If you don’t like something, the fact that you would be good at it if you worked hard at it just isn’t all that relevant because you’re not likely to work all that hard at something that you just don’t like.

Finding More Efficient Ways to Filter Candidates

Are there other, more efficient ways to filter prospective employees are there than educational signaling? For example, Triplebyte does background-blind, skills-based assessment, and companies pay us because we find good developers (some with non-traditional backgrounds) who they wouldn’t have found on their own.My guess is that you know more about it than I do. An important thing to remember, is that I can still be right about how the world generally works, and yet you could run a very successful business working with the one or two or three percent of the CS workforce [who don’t fit the norm]. Something can be rare, and yet a business can make billions of dollars going and serving that rare thing.

Doing online skills-based testing is great as far as it goes, but there’s the question of, if there’s someone who doesn’t have any conventional credentials but does well on the test, should you think that they are comparably good employees as someone who has the same score on the test but has the credentials? It’s hard. My strong guess is, in general no, because, especially in this society, when a person drops out of school, there’s normally something deviant about them. Not in every case, but it’s something that, as an employer, I would be nervous about. Why didn’t he go to college, what happened? What’s his problem?

Some people suggest work-trials as a better way to evaluate people.The main problem with just doing on-the-job testing or just giving people a trial run is that there’s a limit to how many people you can give that to, especially for highly skilled jobs.

In your ideal world, how would we filter candidates?In an ideal world, there would be so little government funding for education that people would just regard it as an extreme luxury good and then people would take into consideration a lot of things that we currently ignore. I’d say that the system we have is largely a product of the easy accessibility of education.

In a world where it was just a lot less accessible, people would have to look at cheaper routes. What I always tell my students is, “If governments listened to my policy advice, a lot of you couldn’t afford to be here. But, you also wouldn’t need to be here any more. You’d be able to go and get the job that you are going to get right now.” Whenever you are thinking about the effects of policy change, think about how it changes the whole system. Don’t just think about how it affects you individually.

Degree Inflation and Upward Mobility

That’s the whole idea of degree/credential inflation, right?Yeah, precisely.

Could you just explain that briefly?There’s been a lot of work on how the education and training of workers has changed over time, and one of the strongest results is that as the education level of American workers has gone up by about four years over the last 70 years, the amount of education you need to get a job has gone up by over three years — but for one and the same job. It’s not that the economy has modernized and now you need college degrees to do modern jobs, rather we see that even for jobs that we think of as not really being college-type jobs (like being a waiter), still the average education of those workers has gone way up. The key point is that education levels have changed a lot more than job skills.

Would you rather be a high school drop-out looking for a job today, or a high school drop-out looking for a job in 1950? In 1950, the stigma would have been a lot less because there were so many other people in the same boat as you. It just didn’t say such a bad thing about you, and now it does. There is this treadmill effect where you have to have more education now to get the same job that your parents or grandparents could have gotten.

I wonder about upward mobility. Is there some truth to the idea that universal education opens up doors for people?I’d say that increased access to education both opens and closes doors. For one individual, it opens doors. If a poor kid gets a scholarship, then it has opened some doors for him. But it’s important to remember what we lost. We’ve lost the ability to get a good job right out of high school. Because the more accessible it is, the more education you need in order to impress employers.

Really, we have expanded opportunity for the brightest kid from a poor family at the same time the siblings of that brightest kid probably have worse opportunities, at least in terms of their chances of getting a more highly-skilled job. There’s this nice meme that says, When everyone has a bachelor’s degree, no one does. That’s really the idea.

Child Labor (in Software Development)

One of the controversial things you talk about in in your book is changing child labor laws. A lot of programmers start programming super young — should we hire more high school drop-out programmers?Absolutely, yes. What I’m saying in the book is this: we have these negative associations with child labor because we are picturing a 19th century coal mine or a steel refinery or something, and that has almost nothing to do with modern employment. These freakish scenarios of kids being dropped into vats of molten steel have nothing to do with child employment.

Right now, it’s up to parents to decide whether or not most activities are good for their kids, and so parents can go and take you to dangerous countries, they can take you mountain climbing, and they can also go and deprive you of a childhood if they want to. They can say, “Hey, you’re gonna be an Olympic ice skater.” Or, “You’re going to spend all of your spare minutes learning the piano.” And for all of these things, we rely upon parents to make the decision (even though we know that some parents make bad decisions) because we think parents are the best guardians of the interests of their children, and it’s not really appropriate for outsiders to interfere unless they’ve got a really strong reason to do so, such as physical abuse. I say we should have the same standard for child employment. Except in freakish situations, it should be up to the parent to decide whether or not it’s a good idea for their kid [to work] or not.

If you feel this is terrible — that a kid gets a job when he’s 12 years old programming computers — then explain why that kid’s life is so much worse than a kid who was made to spend every day practicing violin? I’d rather be the kid making money programming computers than be the kid stuck with the violin lessons.

Dr. Caplan, thanks so much for your time!

Bryan Caplan is Professor of Economics at George Mason University and a blogger for EconLog. He is the author several widely read books including The Case Against Education and The Myth of The Rational Voter. Dr. Caplan has been published in the New York Times, Washington Post, Wall Street Journal, The Atlantic, American Economic Review, Economic Journal, Journal of Law and Economics, and Intelligence, and appeared on ABC, Fox News, MSNBC, and C-SPAN. Dr. Caplan is a proponent of Effective Altruism, and he describes himself as “an openly nerdy man who loves role-playing games and graphic novels.” Caplan lives in Oakton, Virginia, with his wife and four kids.

[1]https://www.bloomberg.com/news/articles/2018-03-08/demand-for-programmers-hits-full-boil-as-u-s-job-market-simmers ↩