The Sheepskin Effect

post by Zvi · 2018-05-05T12:10:00.437Z · score: 29 (5 votes) · LW · GW · 4 comments

Previously: The Case Against EducationThe Case Against Education: FoundationsThe Case Against Education: Splitting the Education Premium Pie and Considering IQ

Epistemic Status: The spirit of Local Validity as a Key to Sanity and Civilization [LW · GW]

The sheepskin effect is that completing the last year of high school, college or graduate school is much more profitable than completing any of the previous years, rivaling those other years combined. Employers seem to be paying for the degree (aka the sheepskin, which it’s printed on) rather than the human capital being built over time.

In the education chapter of Book Review: The Elephant in the Brain, I noted Robin relied on the sheepskin effect as strong evidence (along with other arguments, including impacts on national vs. personal income) that school was mostly signaling. Bryan Caplan does the same. He cites the data, seeing (on top of a 10% bonus in pay per year of school) 32% bonus pay for finishing high school, 10% for junior college, 30% for a bachelor’s degree and 18% for a masters. To those who claim this is mostly ability bias, he replies:

Ability bias explanations for sheepskin effects aren’t just hard to square with statistical evidence; they’re hard to square with the glaring fact that education spikes in degree years. If the labor market ignores credentials, why do so many college grads opt for zero graduate education? Are we supposed to believe one-third of the population has exactly the right ability to finish high school, but not advance to college? One-seventh has exactly the right ability to finish college, but not advance to graduate school?

To debunk sheepskin effects, correcting for these neglected abilities would have to drastically cut the payoff for degrees but not the payoff for years of schooling. What abilities would even conceivably qualify?

This seems like a straw man; no one thinks the labor market ignores credentials, so it’s easy to see why students act the way they do. Not only is not finishing high school severely punished as such (as the numbers show), not trying is at least sort of illegal. In addition, there’s a huge barrier to getting into college or graduate school, and large costs involved in starting, often involving relocation. Also, much of college is about being ready for the rest of college, and the early part of graduate school is largely to get you ready for the later parts.

I also notice, looking again, another instance of the mistake of assuming people are maximizing. We are definitely not supposed to believe that because a lot of people do something, it was right for them!

Having dismissed ability bias here, he then reasons:

After digesting all the evidence on the sheepskin effect, you may feel ready to channel King Solomon. Human capital and signaling come before you as litigants. They ask you to split the education premium between them. A ruling with a great ring to it: “Human capital gets credit for the payoff for years of education; signaling gets credit for the payoff for degrees.” This implies a human capital/signaling split of roughly 60/40 for high school, 40/60 for college.

Yet on reflection the Solomonic ruling treats human capital too generously. The sheepskin effect doesn’t measure signaling. Instead, the sheepskin effect sets a lower bound on signaling.

To see why, picture a world that lacks the notion of “graduation.” Can we safely declare educational signaling would vanish in such a world? Of course not.

What I wrote back in my previous review of Elephant in the Brain:

One note I would make is about the sheepskin effect, where the last year of a college degree is much more valuable than previous years. There’s been some debate about this online lately between Bryan Caplan and Noah Smith. I agree that this is largely a signaling effect, with ‘completed all eight terms’ much more impressive than ‘completed seven of eight terms’ since you don’t know how many more terms the first student could have finished if necessary.

What the discussion misses, it seems to me, is that only after graduation do you know that the first three years were real. It is easy to become ‘a senior’ through completion of a number of credits, saving the stuff they find hardest for last or even being in terrible shape to match up with graduation requirements. I strongly suspect that a lot of people who drop out in year four are much farther from finished than they would have you believe.

I’d like to expand upon that, because this effect seems huge but remains almost always unmentioned.

I’ll start with a real example.

I have a learning disability that makes it very difficult to learn foreign languages. This was bad enough to nuke my average in high school, despite having studied the same language (Hebrew) for most of a decade whether I wanted to or not (I’ve held on to maybe a hundred words?), and in college things threatened to get much worse. My college demands four terms of a single foreign language. I chose what appeared to be the easiest one for an English speaker, Italian. While it would be cool to speak Italian, I didn’t choose it for how much cooler it would make vacations and restaurants – I was fully aware that Italian was of little use. But I was desperate to get through this, ideally without my average being nuked again, and if it was marginally easier than Spanish but only 10% as useful, then Italian it would be.

When the term ended, I had spent the majority of my studying time on Italian I, and still (just barely) failed by the numbers. I managed to get the grade changed from an F to a D by promising not to take Italian II. I then managed to find a psychologist who vouched for my disability, I think on the basis of an IQ test combined with my history of failures – there really was no other explanation. So I was granted an exception, and allowed to take Asian literature (which I quite enjoyed) and Etymology (which was boring as hell but not hard) instead of the remaining three terms. The D still ended any hopes of my getting honors and crippled any hopes of a top graduate school, and after that I stopped trying that hard to get As, but I got to graduate.

Without that exception, would I have graduated? My guess is yes, because my family and I would have taken epic measures to make it work. I’d have taken a year off to live in Italy (or Israel) if I’d had to. But I can’t be sure it would have been enough.

A good friend of mine ran into this exact problem with the same requirement, couldn’t get the waiver, has no other remaining requirements, and will probably never graduate.

More data. My mother is a professor at Columbia University, where she is in charge of undergraduate biology education. One cool effect of this is that she’d bring related dilemmas and puzzles home so we could explore them at dinner. I helped her plan exam strategies, deal with discipline issues and so on, and it was both great fun and an actual education in the way that school isn’t. 

Occasionally we looked at a series of students who wanted to graduate with a major in biology. The problem was that their transcripts were, shall we say, not so flattering. They’d ‘completed’ close to the full eight terms, but did they have a high enough average in their major? Were the poor grades (Ds and sometimes Cs) in some required courses not acceptable? We all, including the school, wanted to let students graduate when we could – that’s the business, after all, and no one wants to ruin a kid’s life – but the degree has to mean something. These were many of the students who ‘complete seven of eight terms,’ and many others were those who knew a version of this examination was coming and they wouldn’t pass.

Countless others, no doubt, simply saved all the hardest and most difficult courses for the end, possibly in a way that made the logistics impossible to solve. And, well, whoops.

If I give you eight chess puzzles to solve, and you solve seven of them, that’s a lot less impressive than if you solve all eight. If I give you thirty-two courses in ten different fields of study with varying difficulty, and you choose your order so as to solve and pass the first twenty-eight, you are not remotely 7/8ths done.

I could thus tell a human capital story, or an ability bias story, for the sheepskin effect. The final test is real, so if you built up real human capital, and learned how to learn things and remembered your lessons and persevere when the going gets tough, and all that, you win out. If you didn’t do that stuff, you fail at the end when you can’t hide it any longer. Or, for ability bias, only at the end do we learn who had the right stuff all along; same principle. If the final test is sufficiently ‘more real’ than the others, that bonus at the end makes perfect sense.

Thus, I don’t think the arguments from sheepskin are as strong as many think they are. I do think that the education premium is mostly signaling and ability bias, including (but far from limited to) the sheepskin effect. And I do think Bryan offers other much stronger evidence, such as the fact that anyone could walk into any college class and take it for free sans the degree, and actual no one ever does. But I don’t think the sheepskin effect puts a lower bound on the signaling share, or offers that much evidence, because in a world without signaling you’d see it anyway, and I’m curious how Bryan would respond.

 

4 comments

Comments sorted by top scores.

comment by ialdabaoth · 2018-05-05T12:19:28.844Z · score: 13 (3 votes) · LW · GW

Shouldn't this:

A good friend of mine ran into this exact problem with the same requirement, couldn’t get the waiver, has no other remaining requirements, and will probably never graduate.

count as strong, direct counterevidence for this:

The final test is real, so if you built up real human capital, and learned how to learn things and remembered your lessons and persevere when the going gets tough, and all that, you win out. If you didn’t do that stuff, you fail at the end when you can’t hide it any longer. Or, for ability bias, only at the end do we learn who had the right stuff all along; same principle. If the final test is sufficiently ‘more real’ than the others, that bonus at the end makes perfect sense.

Unless you're suggesting that part of the ability being measured, is entangled with the ability to get a waiver?

comment by Raemon · 2018-05-06T02:59:29.240Z · score: 6 (1 votes) · LW · GW

My own take was that there are a couple separate issues:

1) Is 7/8ths of schooling as proportionately strong signal as 8/8ths?

2) Is language requirements a signal anyone should care about?

My guess is Zvi thinks the answer to #2 is "no" (given that Zvi has a lot of bad things to say about school in the first place), but that the existence of the language requirement is a proof-of-concept that #1 is unlikely to be true. In the case of language, an employer shouldn't care about that aspect of schooling one way or another, but in the case of biology class if you're hiring a biologist, you should.

comment by lionhearted · 2018-05-06T22:51:09.868Z · score: 5 (1 votes) · LW · GW

Good post. To play devil's advocate for a moment — I basically agree with you, but for the heck of it — perhaps the ability to graduate is actually a relatively honest signal for "ability to finish stuff" which results from either not having any major weaknesses in a big confounded mess of variables and character traits, or having the problem-solving skills, political skills, system/bureaucracy-navigation skills/whatever to be able to mitigate one's weaknesses if any... just y'know, finishing a 4-year degree with all its attendant tedium and arbitrary requirements, maybe that shows the degree-holder can finish stuff.

Just playing devil's advocate, mostly agree with you. Good post.

comment by Elo · 2018-05-05T17:13:25.297Z · score: 5 (1 votes) · LW · GW

Ability to complete final year is proof that you completed n-1 year. In theory n-1 year is proof that you completed n-2 year... Etc. But if it was too linear then you would be able to walk away saying n-2 year was done without having to go to n-1 year and final year. (because it's obvious you can do those too but you want to be efficient and get to the workforce).

Yes there's a hard stuff at the end problem. But there's also an incentive to keep this structure and extract money from the students.