April 17, 2014
Let me first outline the signalling theory of education. Everyone who has taken a standard micro-economics course has examined this model, but for most of us it has remained as a curiosity, being superseded by more 'persuasive' human capital models (in the tradition of Becker) and our own pre-conceived idea that "education is good".
Spence's theory of education
In the early 1970's, economist Michael Spence examined the role of signalling in labor markets. He considered the problem faced by employers: employees vary in quality, but the employer has no easy way to ascertain quality prior to hiring. Because of their higher productivity, meanwhile, high quality employees could command higher wages, should they be able to convincingly signal their quality to prospective employers.
While one could imagine that over time, workers would develop individual reputations for skill or the lack thereof, Spence focuses on the "one-shot" game, in which no reputation information is available nor is there any incentive to worry about future reputation. Instead, employers rely on indices (unalterable characteristics such as age or sex) and signals (alterable characteristics, particularly education level) to decide what wages to offer each worker.
Under these circumstances, Spence argues that a signal can be used to distinguish workers only if "the costs of signalling are negatively correlated with [worker skill]" (p.358). He constructs a simple model in which there are two types of worker: high productivity and low productivity. The former can acquire education at a lower cost (in terms of time, money, annoyance, etc) than the latter. Spence shows that in this model, there will be a signalling equilibrium in which high productivity workers will seek more education than low quality workers, and employers will offer higher wages to those with higher levels of education. In Spence's model, education – rather than increasing worker productivity – merely serves as a reliable signal of it. [Source]
But after I heard Bryan Caplan's talk on Econtalk I realised Bryan is right and it is clear that there is more formal higher education than there should be. Regardless of what one may think about it, there is no case for government subsidisation of higher education [This may have the effect of slightly changing my recommendations re: higher education in BFN].
I've spent considerable time shrinking the transcript of Bryan's interview. [Word version here]. I'll read further and provide more detailed thoughts later, but this extracts of the transcirpt is a MUST-READ for anyone interested in higher educaiton policy. Here's a (separate) long podcast that I've not heard.
Russ: How much more do college graduates earn relative to high school graduates, and how has that return changed over time?
Bryan Caplan: People who go to college are not the same [as others] even when they start. So, the kind of person who goes to college is different at the beginning. And there are a lot of reasons to think that people who are different in the beginning would have made more money even if they hadn't gone to college. The most obvious one is that people who go to college are generally smarter. They were smarter before they started.
I take a look at what people actually study in school, and I see very little evidence that most students are acquiring any technical skills. And also, surprisingly, when I read educational psychology, there is a lot of question about whether college students are actually learning much in the way of thinking skills, either.
Russ: What about for someone who has attended college but not graduated? What's the premium for that?
Bryan Caplan: That is about 10%. A lot less. … A lot of what's going on is that for many people, graduation is hard. It's actually hard to pass the classes, hard to keep doing it. The kind of people who write about education are people who have had a fantastic educational experience. They've always been doing well in class; they've been getting their heads patted from kindergarten on. And not only did they have no trouble finishing, but the people they know had no trouble finishing. But most people have a lot of trouble finishing. So, right now, the 5-year graduation rate for 4-year college is only about 55%. So like 45% fall by the wayside during this period. As to why exactly it is they don't finish, part of it is probably just the material is too difficult. A lot of it is that people get really bored. And then, another point is that even if you wouldn't actually fail out, it's just very discouraging to constantly receive Cs or worse. Even if you could get your 4-year college degree with Cs and get something out of the labor market, it's hard to spend four more years being told that you are at the bottom of the barrel.
Bryan Caplan: If you were to go and put in 3 or 4 years and then not finish, you might only be getting some of this 10% premium, and that is not very good for all those years you are putting in. It really doesn't pay very much. Here's the key thing to keep in mind. When we say that there's an average graduation rate of like 55%, a lot of that is predictable. So if you have great SAT (Standardized Admission Test) scores, if you are great in high school, then your graduation rate is probably closer to 90 or 95%. On the other hand, if you were in the bottom quarter of your high school class then you are talking about maybe a 10% chance of finishing, something along those lines. So even if you aren't a great student, doing the math actually eventually becomes fairly easy—yeah, I can get an 83% if I finish but I only have a 10% chance of doing that. Even if you don't think in such complex terms or think so quantitatively, just noticing that people like you almost never finish probably does discourage a lot of people from trying.
Russ: You write—this was a very nice way of capturing it: "For students in the bottom quartile of academic ability, paying a year's tuition is almost as foolish as buying 10,000 lottery tickets." Just that the odds are stacked against them; they have a small likelihood of finishing. Do we know anything about why those people don't finish? How many of them just were getting good grades and just decided to give up, versus flunked out? Do we know anything about that. One of the major costs of college isn't tuition. It's the foregone income that you give up by not working. You don't get that back, either.
Bryan Caplan: That's absolutely right. You're right that that is more important. One of the things that I've learned writing the book is that even though list prices for tuition have gone way up, still the actual tuition for your main 4-year state universities is not that much. After you adjust for all financial aid and everything else, it's still only in the ballpark of $5000 a year in most places. So actually the foregone earnings could easily outweigh that by a factor of 3 or 4.
When people talk about education, they assume 100% completion rate. Far better to put in the 50%, although even better, of course, is to individually tailor the predictions based upon the performance of the student in high school and SATs and so on, and then you'll see, well for some people it really is very close to 80%, because they are great students, they always have been, and they almost certainly are going to graduate. And on the other hand, as we said, for other people, who have maybe a 10% graduation chance, then they are talking of a real pittance.
Even to say that college graduates get 83% is very misleading. There's a huge range. One of the main things that matters is your major, where there are your stereotypical high-earning majors, where you are getting a lot more than that. So that would be engineering, computer science, finance. And actually economics.
There's a very common view, often associated with David Card and his many students that education is just as lucrative for high-ability students and low-ability students. The way they reached this conclusion is basically saying, look, if you take a look at the graduates of college, high-ability graduates seem to get as large percentage increase in their earnings as low-ability graduates. The main problem with this, again, is that they are only looking at graduates. If you go and look at the fact of ability on whether you graduated all, that's the main way where you very clearly see that the lower-ability students don't get the same return as high-ability students, because they are less likely to get much of a return at all. So, it's one thing to say, if a low-ability student manages to get to the finish line then he'll get a big gain. But what are his chances of getting over the finish line? Pretty bad.
Russ: Let's talk about the underlying explanations for why there is any premium at all. Over the years, since I've been studying economics, the two approaches that fight against each other are the human capital model and the signaling model. Talk about what the differences between those two models are, and which you think has, if you can, more weight than the other.
Bryan Caplan: The human capital model and the signaling model are both stories about how education successfully increases earnings. They are not disagreeing about whether education actually raises earnings. Rather they are disagreeing about why. The human capital story says that you go to school; they actually teach you a bunch of useful jobs skills; you then finish and the labor market rewards you because you are now able to do more stuff. The signaling model says, no, no, no, no; that's not what's going on. What's going on is that people go to school; they don't actually learn a lot of useful stuff; however, the whole educational process filters out the people who wouldn't have been very good workers. So people who are lower intelligence, lower in work ethic, lower in conformity—those people tend to not do very well in school. They drop out. They get bad grades. And that's why the labor market cares. It's not that the school actually transforms you to a good worker from a bad worker. It's that the schooling, the school puts a little sticker on your head—you know, Grade A student, Grade B student, Grade C student.
A very simple way of explaining it is think about two different ways to raise the price of a diamond. One way is by cutting it very beautifully so that it is actually a better diamond. Another way, though, is you put on that funny monocle thing and you look at it and you appraise it. These are both ways that you can raise the price of a diamond. So, cutting the diamond can raise the price. But also a very credible appraisal can raise the price as well. And the human capital story basically says that schools take these diamonds-in-the-rough and it cuts them very nicely and then that's why they are more valuable. And signaling says, no, no, no: what's going on is students show up to school basically as well as they are going to be, and then what the school does is it puts them through a bunch of tests and it makes them jump through a lot of hoops, and then it certifies them and certifies their quality. And that's why employers actually care. Now of course, any sensible person will say: Well, there's some truth to both stories. But, so the real question is not: Is it all human capital or is it all signaling? The question is: What's the balance? The general view among most active labor economists is that signaling is basically irrelevant; it's maybe 5%, 10%; it's something that we can pretty much forget about. My view, though, is signaling is more like 80%
Russ: Let me let the listeners understand why I have a natural skepticism about it. And I think a lot of labor economists do as well. And the reason is that it's an extremely expensive signal. So, you are saying, for 4 years, I give up the chance to work; I pay this tuition, whether it's $5000 or $10,000, or $30,000, or $40,000—at a private university. And for that enormous amount of money, I prove that I am a good worker and I get a sticker on my head. Wouldn't there be an easier, cheaper way to get the sticker? If all it's doing is measuring ability, this 4-year slog that's extremely expensive? That's the best way that people have come up with to get the sticker?
Bryan Caplan: Yes. Now the thing to keep in mind is, there are many different kinds of ability, and school seems to actually be weeding people out on almost all the ones you can think of. So, it's not just intelligence. But what if what you are trying to find are workers who are hard-working, and especially workers who are conformist? Now it is a lot easier to see why does it have to be so expensive? Suppose that someone came along, here's the way I'm going to certify the people who are hard-working but it only takes a month. Now, who are the first people who are going to be in line for this one-month certification? They are going to be people who aren't very hard-working. The people looking for an easy way out.
Not only can signals be expensive, but they really have to be if there is a valuable prize at the end. So, think about this. Suppose that somebody came up with a chemical process to make diamonds as cheaply as plastic. How much longer would people continue to give diamond engagement rings? This would be the end of it. Why? Because the engagement ring just has to be expensive, otherwise it does not show commitment.
So, the same thing goes with education. If someone figured out a way to make the current process happen in a year, this would mean that a whole lot of people would put in a year; they'd finish; and then it wouldn't really show that much about you. It's got to be a long, drawn-out, painful process or else it doesn't really separate you from the pack. And according to signaling, that's the key point, is to separate yourself from the pack.
Russ: I'm curious what evidence—besides armchair theorizing—you point to that labor economists ignore. That you say is in other fields, etc. For the signaling theory.
Bryan Caplan: So, in terms of the research, one very well-established fact that gets very little play is what's called the 'Sheepskin Effect'. So, we've sort of been touching on this point on how not finishing, starting college without finishing seems to raise earnings by only 10%, whereas it raises earnings by, seems to raise earnings by 83% if you do finish. So, this is actually part of a much more general fact, which is that a lot of the payoff for education comes from getting your degree. It comes from crossing the finish line. Right now, in the early decades of the signaling model, this fact was not well-established. And so there was a lively debate: Is there a sheepskin effect? Is there not a sheepskin effect? But until the sheepskin effect was well-established, when it was still in debate, almost everyone took for granted that a large sheepskin effect would show that signaling was important. Because otherwise, why would it be so important to just get over that finish line? So, in terms of the human capital model, it's really puzzling. What is it, the last class that teaches you—
Russ: The capstone. It's the capstone class. The whole idea.
Bryan Caplan: Yeah, the capstone class. So, like, why is it the person one [?] class short of graduation is only getting 10%, whereas if you finish that class you would get 83%?
Russ: Well, hang on. First of all – I don't know if this is a universally understood name – but a 'sheepskin' is another word for graduating college. 'Getting your sheepskin.' I don't know the origin of that. Do you know it, Bryan?
Bryan Caplan: Yes. Yes, I do. So, it's another word for 'diploma.' And the reason is diplomas used to be written on sheepskins, actually.
Russ: The premium. The premium over high school students is over 10% if you don't graduate. That is, attending college makes you a little more money relative to a high school graduate. Is that true if you go for one year, two years, three years? What you are claiming is, you might be claiming—if you go for 3 and a half semesters and you are 1 course short of graduation, you still only get 10%? Is that true?
Bryan Caplan: It's a little more complicated than that. So, if you go and take a very close look at the data for college, you'll see something like for the first year of college, that might increase, if you [?] essentially finish that, that might increase your earnings by 5-10%. Then year 2, maybe another 5-10%. Year 3 seems to give you nothing. And then it's year 4 that gives you the remainder. Which is huge.
And we see that's very similar for high school as well. So, like, 9th grade seems to give you a bit, 10th grade a bit; 11th grade seems to give you nothing at all; and then 12th grade, finishing that, getting a diploma, that's what gives you a very big raise over what a high school dropout would earn.
Russ: Yeah. I guess the complication is that the people who do get, say, three and a half years into their college degree or one course short, why don't they finish? And what does that tell you.
Bryan Caplan: Exactly. Now you're thinking like someone who believes in signaling. Now you're asking, why didn't this person finish? What is wrong with that person? Maybe they just had some bad luck. But also it suggests: “Look, in our society it's expected that you finish. There are a lot of different ways that you could have made up whatever problems that you had; so I'm nervous about you as an applicant”.
But let me go back to how the debate played out. So there was a long period when economists just weren't sure if there was a sheepskin effect or how big it was. During that period everyone took for granted that a large sheepskin effect would show that signaling was important and the lack of one would at least undermine that. Now, in the late 1980s, early 1990s, it became totally clear that there were huge sheepskin effects—better data came along and several papers were published and they've never been challenged. Not even challenged successfully—no one's even tried to challenge them. The data are now so clear. But almost as soon as the evidence came in very strongly that sheepskin effects were very real and very large— then labor economists moved the goal posts and said, Well, that doesn't really prove anything.
And when you know that these people making these arguments have been through the entire educational process; they finished at least three different degrees. To be a researcher on this, you finished your high school degree, you finished your bachelor's degree, your master's, probably your Ph.D. And for people like that to say, I'm totally unconvinced that it matters whether you actually get your degree and cross the finish line, to me it's just insane. Like, you know very well, you were a student, you know that if you didn't finish that would ruin your life and prevent you from getting this job. You know that. Everyone around you knew that. If you were to go and deny that to your fellow students and say, I'm not showing up for the final exam because what difference does it make? It makes a lot of difference. And it makes a lot of difference because people who don't finish are quite different from people who do, and employers will hold it against you.
Here is one fact that I've often noticed. What do students do when a professor cancels class? They are happy. They cheer. And from a human capital point of view this is bizarre. Basically, the professor is saying: “You want me to train you to be a better worker so you can do better in real life? Well, I'm going to keep your money and I'm not going to give you the training. See ya!”. That is effectively what the human capital model is saying is happening when a professor cancels class.
On the other hand, so the signaling model says, well why won't the students be happy? The employers will never know that the professor canceled class. What they are learning they are probably going to never need to know again. It's not going to show up on their transcripts. If everybody learns less then this is not going to change the distribution of grades in all likelihood. So then students get an extra afternoon off and it's not going to affect their future.
Or here's another one of my favorite debating points. Claim: Right now you can get the best education in the world for free if you want it. Well, suppose you think Princeton is the best education in the world. You don't need to apply; you don't need to get admitted. All you do is move to Princeton and start attending classes. And in my experience, no one will stop you; no one will card you. If you go to the professor and say, I'm not a student here but I'm interested in your class, most professors will get a tear in their eye: Someone actually wants to learn from me.
But if you go and get this totally free Princeton education for four years, there is one thing you won't have at the end: any proof you ever did it. Right. And if you consider—Deal A is you go to Princeton and you get a Princeton education with no record you ever did it, or you go to a much lower-ranked school where you admit you are getting a worse education but there is a record, which one is going to do more for your career? Almost everyone says, well, obviously the second one. The first one may make you an interesting person, may be a great experience, but employers aren't going to care. They won't believe you if there is no sign you were ever there. Whereas getting a bachelor's degree by the book from Podunk State on the other hand, that doesn't give you nearly as much as getting a bachelor's degree from Princeton but it gives you something that is real and tangible.
. So, I've often had this argument with Tyler Cowen about signaling. He says, Look, Bryan, you don't know what you're talking about. I actually have a job where I have to hire people and fire people and I know how things work. And he says, And I say within three months we know whether a worker is good or not, so it's crazy to think you have to go and spend all these years signaling in order to get the job.
But I say: Wait a second, you are looking at this the wrong way. What has to happen before you can evaluate someone for three months? You have to hire them. What has to happen before you hire them? You have to interview them. What has to happen before you interview them? Well, you have to go and pull their application out of a giant stack of other applications. And any time someone has 300 applications for one job, what are they going to do? They are going to go and start thinning it by throwing away applications from people that don't seem like they are worth giving an interview to. So, if you have an unconventional transcript, an unconventional background, so maybe you don't have any proof that you did it or maybe it's just something weird that you did or unusual that you did, this is something where employers quite reasonably don't give you a chance because they are not in the business of giving chances. They are trying to run a business.
Russ: Here's an interesting example. Sometimes someone turns out to have lied about their qualifications for a job. They don't have a degree from Harvard. They didn't get a Master's in such-and-such. And they are immediately fired. And they are not fired because they can't do the job well. They are fired because they lied. Not because they are incompetent. Which would seem to suggest that the piece of paper itself is maybe not so important.
Bryan Caplan: Well—there is a reason to be concerned about having con artists in your employment, right? I'm saying as a business decision, saying look, this guy is perfectly able to do the job, but he's a con artist. I would fire that person—
Russ: I didn't say it well. What I'm saying is that actually attending university didn't have any impact on whether they were good at the job or not. It's not like the people say, oh, the person didn't attend this university, doesn't have the degree—obviously they can't do the job. They've been doing the job. So it's really just the piece of paper as a signal that matters. When they are fired it's not because they don't have the skills. They are fired because they were dishonest. Which is fine. I understand that. But it seems to be consistent with your theory.
Russ: Your point is that the people who go and don't go aren't the same. Let's come back to that and talk about how that affects how people measure the return to education and why it's so important.
Bryan Caplan: When you see that college graduates are earning 83% more than high school graduates, you have to ask, well there's one thing that's different about them and that the college graduate has—went to college and finished. But is that the only thing that's different about college graduates versus high school graduates? Of course not. So the average college graduate is probably going to be smarter; he's probably going to be harder-working. He's probably going to be more conformist, less impulsive, have many other advantages. And that 83% is capturing all the advantages that the college graduate has. Not just the single advantage of the education. It's capturing the benefits of a lot of advantages he had before he ever started school.
So if you really want to find out how much did the education raise his earnings, you need to correct for or adjust for all these initial advantages. So there is a big research literature that tries to do this in economics. If you do this the simple way, namely you try to go and measure the pre-existing advantages and then re-do the statistics, then you generally find that the true education premium is a lot less than the observed one. So, basically just putting in, just correcting for your measured IQ before you start college, that will usually bring the payoff down by about 30%. So, bring it down from like 80 to like 55, something like that.
And then once you put in that adjustment you say, well, there's a lot more going on than just intelligence. There's work ethic, conformity. Some of these things there aren't very good measures for, but you can still say it seems plausible that this actually—there are other abilities that are accounting for this.
If we go and correct for every ability we can measure, we see that the payoff for education seems to go down by 35%. Well, what about all the stuff we haven't put in? Maybe that changes it to like 45%, something like that. So after going over all of the evidence that I could find and just weighing it all, my best guess is that only about 55% of the gain that you seem to get is genuine. And the rest is what college graduates would have actually earned if they hadn't even gone to college.
If you look at earnings in countries where they have really crummy school systems, it seems like a year of education pays as much or more as it does right here.
Russ: Is that true? I'm a skeptic about that, Bryan. Tell me about that.
Bryan Caplan: Right. So there's been a lot of effort to measure the international return to education, look at different countries. And the very standard result is that in poor countries the payoff in the labor market of getting another year of education or getting more degrees seems to still be on the order of like 10% per year. In terms of what they are teaching them, it's crazy. But in terms of saying, look, people who are willing to jump through these hoops and endure this horrible process where they don't even learn anything, it still seems to say something about you as a worker and it makes employers in those countries more willing to hire you, and you do actually get an advantage. Even though you don't seem to be learning much of anything.
You can take a look at developing countries and see that people with more education make more money there, just like they do here. The premiums are very similar to here or actually a bit higher. So—again, the main thing to remember is of course employers don't know the details. So you might have gone to the school where the teacher never showed up, and maybe they wouldn't have considered that to be much of a signal, but if teachers show up two thirds of the time and yours didn't show up at all, they may still be willing to give you a big pay hike because they are just playing the odds. Which is what all employers are doing all the time, is just playing the odds.
To me, what is most impressive is that when I actually talk to people who officially don't believe in signaling, their actual experiences are so similar to mine. So, similar to what the signaling model would predict. I was talking with one very, very eminent labor economist about his experience as a bank apprentice in Germany, and I say, so did you consider just going on and being a banker? And he said, no, no, you have to get college for that. And he actually laughed at his own sentence, because he realized that what he was saying fit with my story so much better than his; and yet that was his spontaneous reaction to what he saw with his own eyes.
Russ: Well let's close with an issue that comes up now and then. Which is the case for subsidizing education. A lot of people argue that because the returns are so high, even if they are only 40 or 50%, the returns are high. And since students have difficulty accessing capital, frequently to borrow to get that, to make that investment, capital markets for human beings are not very effective. I can't promise you my future returns so it's hard for me to use that as collateral. So people use that along with the alleged externalities, positive externalities of having an educated public, to justify public subsidies to education, either in the financing of education or the provision of it. What's your opinion on that?
Bryan Caplan: Well, really what's so important about the signaling model is that it says, look, you are correct to believe that education increases the earnings of the individual who gets the education. You are wrong, though, to think that it's socially valuable for everyone to get a lot more education. Because all that will do is raise the bar for how much you need to have in order to get a job at the same level. The heart of the signaling model is saying there actually is a big negative externality of education. That every time you get more, you are making everyone else who didn't get the, didn't jump through the same number of hoops look worse. So really what the signaling model is saying—
Russ: It's an arms race.
Bryan Caplan: Yes, it's an arms race. And the fact that the private return is high is really a very bad argument for pouring more money on.
What you are looking at is return for the marginal people who are just on the edge of going or not going. The return for those people is quite mediocre. And if you adjust for ability and everything else, I would say that we have subsidized education way past the point of positive returns. So by my calculations, actually, the social return to education is now quite negative.
It would be a much better policy to drastically scale it back, so rather than encouraging more people to go, I think it's better to discourage them from going or at least to encourage them less.