More people speak freely on more topics to bigger audiences than at any time in Western history. There is censorship, self- and otherwise, on some important topics, and it has ever been so. At no point in Western history was there a civil libertarian utopia, and much of the repression from when the "Christian substructure" was at its zenith far surpassed anything we see today. And one of the key factors for combatting repression from the past is in abundance today: loud, dissenting voices crying out for more freedom. And cultures retain their power to evolve new "substructures" for promoting and retaining practices conducive to freedom. Perhaps more so with all the new technologies and means and styles of communication.
1) The obnoxious sign language person in every public event is probably an unreasonable accommodation. They take up at huge portion of the screen and are very distracting. It can't possibly pass a cost benefit test. I guess you could say the same about a fair bit of ADA rules. It doesn't take long on Google to find massive evidence of a kind of abusive ADA lawsuit industry.
Pretty much anytime you accommodate some really small group its wrong.
2) "But deafness should not become an aspiration."
I think Richard raises a point. If something IS INFERIOR and you say IT ISN'T INFERIOR then you are already lying. And once you're lying a little, why not lie a lot. There is no natural stopping point for a lie. If a little lie will make gays feel better, a bigger lie will make them feel even better. And haven't they suffered enough!
3) Homosexuality is mental illness. Trans is just that same mental illness x10. There are a lot of the same drivers (narcissism, perversion, hedonism, etc).
4) "it does not propose any constructive steps to take"
What steps do you proposed to take? Perhaps in a red state debating a law there is something to do right now, but here in Virginia I know there is a 0% chance of school choice passing the legislature. During COVID there were huge parent protests at school board meetings but mostly the school board just kept doing what it wanted to do until the governor was able to narrowly pass a bill getting rid of masks (he needed a slim slice of democratic legislatures to pass it, but those same people would never pass school choice).
It's not exactly clear what regular people supporting school choice are supposed to do besides elect republicans if they feel the school choice issue is more important then other issues that might make them vote democrat, but that doesn't appear to be enough to shift legislatures.
Awhile back the governor proposed that our entire school board be up for election that year, given all the scandals. However, he needed the state legislature to pass it. The legislature is pretty evenly divided with the DEMS having a slight edge. My district has a DEM.
So I contacted the state rep and made my case why I wanted to have a chance to elected a new school board. His reply was boiler plate progressive pablum.
The era of 1990s style bipartisan bills are over. If you want to do anything you are going to need a solid majority of your side, but no one individual can really being that about. It's basically set in stone by demographics. So it all feels hopeless.
From NS Lions: “Maybe liberalism could only ever really run on an operating system of specifically Christian foundations.”
This is something I’ve come to believe, and from what I’ve read of old books, the idea used to be a common assumption until it was killed off by militantly secular academia in the 2nd half of the 20th century. Good to see that this is being talked about more these days.
I agree that the Saudi's need to develop and economy not based on oil at some point, but I'm not sure they can.
Your average Saudi citizen is low IQ. They aren't going to develop a modern economy. They will need to import westerners to do that. But while a place like Dubai can do that, Saudi Arabia has 40M people. That means they need to import enough western knowledge workers to generate enough economic value to buy off 40M people. This isn't some tax haven city state living off the scraps of the west because it doesn't have to divide it amongst too many people.
I appreciate the desire to built something but we've seen countries with lots of natural resources waste them on vanity projects they claimed were productive before. All that seems different now is that they are allowing bread and circuses.
I remember watching a documentary about some island in the South Pacific that was ultra rich in some bird poop that made good fertilizer. For a couple of decades it was one of the richest islands in the world. They "invested" the money in all sorts of things. Hotels, broadway plays, etc. Of course all these investments ended up being retarded money losing ventures. At the end of the day they were just some dumb Pacific Islanders on an island of shit. Eventually they mined all the shit and went bankrupt. Now they are one of the poorest countries in the entire world.
> My father’s First Iron Law of Social Science: sometimes it’s this way, and sometimes it’s that way.
This is what the natural experiment evidence reads like to me. Arteaga (2018) looks more like the human capital story, while Hussey (2012) looks more like signaling. And both apply to a particular context in ways that make extrapolation to other settings difficult.
Re: "I present ten reasons to doubt that the education system is a waste of time and money." — Infovores, "Attention Caplanites: School is Less Wasteful Than You Think!", at link embedded in Arnold's blogpost.
Let me reply, quickly, tersely, point by point, to Infovores' stimulating critique of Bryan's case.
1. Misdirects policy issue. We should experiment, here and there, with alternatives to academic model, reductions in subsidies, vouchers, etc., instead of arguing about burden of proof (and standard of proof) for funding or defunding the whole system.
2. Conflates dropping out and flunking out. Ignores mismatch of curriculum and job skills/knowledge. Assumes dropout has learned less than all grads.
3. Doesn't establish that schools impart conformity (rather than filter non-conformity). (Treatment effect versus ongoing selection effects.)
4. College students are hamstrung by a collective-action problem around punishment of academic dishonesty. Detection usually would require tips and testimony by fellow students. Everyday bonds in campus residential total institution make students reluctant to report fellow students. Given low detection rate, admin ratchets up punishments (deterrence model). Severe punishments (e.g., expulsion) violate even honest students' norms of fair punishment, overriding collective interest in clear signal of academic standards. Many students reason, 'It could be me, in a pinch.' Compare Akerlof, Yellen, Katz, "Gangs, Law Enforcement, and Community Values" (1994), critique of Becker bricks-and-sticks model of deterrence when detection is low, community cooperation is needed for tips and testimony, and punishments exceed community norms of fairness.
5. Ignores mismatch of curriculum and job skills/knowledge. Maybe college calculus, which fades, nonetheless durably reinforces prior learning of maths, but few grads use maths on the job.
6. Bryan acknowledges that his 20/80 average HC/signaling ratio is a guesstimate, and that the ratio varies by context. PS: Does a 3rd causal factor — "ability bias" — figure in Huntington-Klein's model? Bryan carefully address also ability bias.
7. Miguel Urquiola makes this critique of Bryan, using examples of SCOTUS reliance on law clerks from Yale Law and Harvard Law. Bryan replies that the social costs of current education system far outweigh benefits of ever finer signaling. See debate at link below, Q&A portion:
8. Doesn't establish that attentive mentoring in mismatched curriculum in school is more productive than less-attentive mentoring in job skills/tasks in the firm or org. Assumes ceteris paribus, but firms and orgs would focus more on mentoring (even with risk of employee defection), or other training mechanisms would emerge, if subsidies were withdrawn from education system.
9. Straw man. Although a gen ed course (e.g., foreign language) might have slight value, it doesn't pass cost-benefit test.
10. Ignores that college is also a country club that delays adulthood for many. Students quickly learn to cut corners. Country club exacerbates collective-action problem of achieving cooperation for high standards. Compare Jacob et al. "College as Country Club" (JLE): https://www.journals.uchicago.edu/doi/abs/10.1086/694654
Thank you for this John A, I always appreciate your thoughts!
Let me first say for the record that I am totally on board with making some big changes to the education system at the margin. I think The Case Against Education is a great argument for opposing “free college for all” and making some significant cuts (5, 10, maybe up to 20% even) but not for cutting 80% like Caplan wants to do. I would also support significant privatization of the education system, for reasons Hanania has outlined recently.
The reason I wrote this article the way I did, emphasizing (briefly, due to both reader attention and email limit constraints) points that don't tend to get noticed by Caplan fans anymore, is that I felt the consensus in the blogosphere had become overly ideological and dogmatic to the point of greatly exaggerating what proponents of the signaling model should reasonably expect those without their priors to believe based on evidence alone.
As I think the Ives Parr essay illustrates, many have even gone far beyond what Caplan himself argues for in the book (they talk as though everything were signaling, rather than 80%) and MUCH beyond what Caplan says follows from the evidence itself rather than philosophy. Here is the full quote from Caplan in the section "What I Really Think":
"Yet political philosophy is ultimately unavoidable, because philosophy sets presumptions. Some philosophies have a presumption in favor of education; others, in favor of the status quo. From these perspectives, the burden of proof rests on advocates of cuts. This burden is surmountable. Social return estimates amply justify spending 20% less on education given a presumption in favor of education or the status quo. But such presumptions still block radical reforms, because there is minimal concrete experience with radical reforms. We can speculate that cutting spending by 80% would be a great boon, but when the burden of proof is against you, speculation can’t surmount it. I favor radical reforms nonetheless. Philosophically, I am staunchly libertarian. While not absolutely opposed to taxpayer support for education, I have a strong moral presumption against taxpayer support for anything."
My intuition, from the trenches, is that it's not a 10% problem in higher ed. Not by a long shot. And I wonder why anyone who believes that it's merely a 10% problem would bother challenging the entrenched establishment to get elbow room for radical experiments.
I think we may be thinking about two different questions.
If the education system is 80% wasteful signaling, then we can cut that 80% and the education system will be improved.
My claim is that this estimate is incorrect and by a large margin. That doesn’t mean that there aren’t big problems in education, only that Bryan’s proposed solution (just cut spending) is not going to have the effect he thinks it will.
Just because I only favor cutting, say, 10% doesn’t mean I think the size of the problem is only 10% because cutting is not the only solution available. To me, privatizing the education system would be as big of a change as some of things Bryan proposes and I’m highly sympathetic to that.
Private colleges and public colleges, alike, have massive waste, and misallocation, and plenty of signaling. The main difference is that private colleges have less taxpayer subsidy per student than public colleges do. Nonetheless, not-for-profit private colleges, too, have a large structural subsidy — tax exemption — and some taxpayer tuition subsides (Pell Grants and the like). Public colleges have, additionally, large taxpayer subsidies for tuition for all classes of students and for facilities. Although private colleges must please donors to garner more philanthropic subsidies, whilst public colleges must please lawmakers to garner more taxpayer subsidies, it turns out that private colleges and public colleges operate very similarly — curriculum. faculty policy, growth in admin, country club — if we sort colleges by selectivity, size, degree type, residential vs commuter.
Among selective colleges, it happens that not-for-profit colleges and public colleges do compete for enrollments. Among non-selective colleges, there is some head-to-head competition among not-for-profit colleges, public colleges, and for-profit colleges.
Given the evidence from the current institutional mix, why would privatization, *with the same (or 90% same) level of taxpayer subsidies* (legislative appropriations and tax exemption), radically change college operations? (Again, we must sort the comparisons roughly, by institutional selectivity, size, degree type, and residential-total institution vs commuter institution.)
What am I missing?
PS: I did address, although too telegraphically, your 10 points against Caplan's case, which points out low HC/signaling ratio, negative social externalities of zero-sum credential game, and wasteful taxpayer subsidies.
No you're right, I am speaking about privatizing schooling prior to college, along the lines of what Hanania wrote recently (link below, I should have specified).
I believe Caplan supports cutting subsidies for masters programs most, then bachelors programs, and high school the least. This sounds roughly right to me and if you were to cut 10% of overall spending (~100 billion excluding loan subsidies), but focus it almost entirely on higher ed then you would be cutting in the ballpark of 20%. But take that with a grain of salt, as the policy particulars were not really the focus of the post.
Of the ten points you mentioned, some of them are addressed in other comments here. Sheepskin effect is probably the most important one if you want to CTRL-F for it, since Bryan uses that measurement directly to get his 50% lower bound for wasteful signaling and it is really not very defensible.
Regarding conformity, see Robin Hanson's post I link to and combine it with the overlearning insight—if conformity in K-12 is really "a full-time job" as Bryan complains elsewhere then it is hard to believe students are not becoming more conformist by sitting in school (isn't that the precisely what libertarian rhetoric tends to suggest elsewhere?).
On cheating, it doesn't really matter if there are other reasons why schools don't crack down on it because Bryan says that cheating erodes the signaling value of the degree. If college is all signaling, why does such a large earnings premium remain?
Happy to talk more, though it may be better to continue in the comments section on my post—I worry I'm crowding Arnold's page too much with this debate!
I'll set aside policy ideas for grade school. (For what it's worth, I favor radical vouchers for teens, valid at any training program, internship, apprenticeship, school, etc, with loose accreditation anywhere.)
Perhaps we're talking past each other about Caplan and college. To rephrase:
The challenge to Caplan's interpretation of the sheepskin effect seems to conflate dropouts and flunkouts, and to assume that dropouts learn less than all grads do.
Re: conformity. At least for teens, it would be much better to learn conformity in work or apprenticeship or training contexts, than via school curriculum mostly useless for job knowledge.
Re: overlearning as reinforcement of prior learning. I grant the phenomenon, but say that overlearning doesn't improve human capital if the prior learning (which is reinforced) isn''t useful job knowledge. Everyone (or most?) must take trig in HS, and perhaps the subset of students who then take and pass calc 1 in college thereby retain more trig, but very few of these students will ever use trig on the job.
Re: Academic dishonesty. It does erode (but doesn't eliminate) the value of the degree. If students could solve their collective-action problem (around tips and testimony), then the degree signal would be stronger, clearer, less noisy — and the return to the degree would be greater than it is. (I wonder if one might somehow check this via a well-designed study of the return to a degree at, say, Washington & Lee University — strict honor code — and at otherwise similar colleges, with same SAT/ACT level matriculants, without a real honor code. Probably elusive.)
I've replied here, to keep the thread intact. Thx for hearing me out.
For sure the US recent past was a mix of authoritarianism and totalitarianism but are you suggesting what we have today is worse than restaurants that won't serve blacks, women fired when they get pregnant, men given job preferences because they are the "breadwinners," government programs that are whites only, etc.?
“Every LGB I've known has been pretty vanilla and normal."
More often than not we don't know what issues others have. With that said, I've known Ts who were as normal as anyone too. If someone comes well with depression, addiction, etc., does that mean it isn't an issue?
“no mental-illness issues for LGBs“
It would seem that T is more likely a mental health issue than LG (and B might be even less likely) but I don't think we know for certain whether any of these are mental health issues or not.
Whether chicken or egg, LGBTs have more diagnosed mental illnesses. If everyone were more accepting this might decrease but I'd be surprised if it ever went away completely.
With respect to NS Lyons. I'm guessing you could have written the same article about China 20 years ago. Certainly many of us, myself included saw great promise in the liberalization going on. Now we watch as Xi clamps down on disent and threatens Tiawan. I'm afraid there isn't some magical formula for freedom. Democracys can become a dictatorship of 51% or even just a loud minority. So too can the most benevolent dictatorship turn evil. I wish the Saudia's good fortune, but I'm not planning on moving there anytime soon.
Overall, I think Nick H-K's work is highly motivated reasoning, possibly mixed with analyzing a different educational system than the US has. (That is the charitable interpretation of his description of what he describes as literally what people do.) Many of his critiques of Caplan raise the question of whether he actually read the book carefully, as his concerns are already addressed, at length.
Unfortunately most of the education economics literature reads a lot like "I would like this to be true, so I am going to claim it is", and only a little bit like "I would like this to be true, but damn... it looks like it isn't!" I trust the latter group a lot more!
I really like Infovores, and would recommend he not lean on someone like Nick H-K.
I don't think this is accurate of NHK, certainly doesn't come across that way in his writing (see medium article linked in my post for a dated, but very accessible treatment of his thoughts).
My sense from reading your essay DH (which I appreciated, thank you for taking the time!) was that you dislike a lot of the rhetoric schools use around teaching that make what they do sound more pure and wonderful than it really is. I sympathize with that to an extent while emphasizing that evaluating HC vs. signaling is largely a separate question from what schools say they do. It's possible for school to help students learn even if the rhetoric greatly overstates its impact or if it neglects to mention components of the schooling bundle that actually matter more (e.g. teaching conformity, punctuality, etc.). It may even be the case that schools' idealized rhetoric itself helps students to learn (I know I personally am much more motivated when I focus on the intrinsic value of learning something instead of the attendant conformity or the signaling aspects).
More directly to the point of whether NHK is reliable/correct— can you tell me where the econometrics is wrong in his Empirical Economics publication? The empirically unresolvable aspects of this debate are very important.
"What schools actually *do* when they allow you to continue in your education is, effectively, measure what you’ve learned and see if it passes some minimum standard. If you don’t, you drop out. We end up with those failing the (lax) minimum populating the dropout years. They’ve learned little, so they earn little. In the final year, you see everyone who passes the minimum, whether they learned just enough or WAY MORE than enough. The final year contains a wide range of big learners, so on average there’s a big jump in earnings that year.
Intuitive, based on literal actions people perform, and a totally-HC explanation of sheepskin effects"
That is an absolutely obvious error in regards to the US school system, whether K-12 or college level. Obvious in the sense that everyone who is at all familiar with the US system has to know it does not work at all like that. At all.
So why does H-K claim it?
Is he ignorant of how the US system works, perhaps being an education economist that only studies a foreign country with a different system? That's a possibility, and it is fine certainly. But why does he present this as a counter to Caplan's book? Caplan is explicit that he is referring to the US system, and focuses on that system. Does H-K not realize this somehow? Perhaps because he didn't read it carefully? Why is he arguing about the ratio of human capital to signaling in the US system if he doesn't study the US system?
Or does Nick H-K know how the US system actually works, and makes this obviously incorrect statement because he wants to argue for a particular conclusion and will make claims about reality that support that conclusion regardless of their relation to reality?
I bring that up because I think you can make the econometrics say damned near whatever you want. No one is going to get your data and do a reproduction. Other people are going to do studies and probably find the opposite results, and there won't be a huge "OMG we need to figure this out!" response. It will just be published, and life will go on. I am not accusing Nick H-K of fraud as such, just that the rigor necessary to trust data arguments like this is not reached.
Much more readily available is to see the quality of the arguments they make outside the black box of their math. If they are making inconsistent arguments in prose, one can feel confident they are going to make the same mistakes when strolling the garden of forking paths in their data analysis, when it is much easier to make those motivated errors without even realizing it.
If I get some time this week I will try and dig through his econometrics for signs of sin, but I doubt there will be glaring obvious errors.
I think NHK goes to great pains to make his work accessible to a broad audience. He posted a less math heavy version of his published article for people to read (see my footnotes), and has a youtube channel dedicated to explaining causal inference used by economists in simple terms.
If there is an error in his medium article excerpt, it's possible that the error is partly mine as he speaks more precisely in the later articles he prepared over several years more of thinking on the issue in order to publish. But at least how I understood NHK's medium piece, what he means by "effectively measure against a minimum standard" is simply that schools require students to pass a set of requirements to get a degree and if you fail some of the required credits or leave early without completing them then you don't meet the minimum standard they consider adequate.
This observation is not that much different to my eyes from what Zvi says in an old Less Wrong post:
"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."
As a former high school teacher, I am not at all sure that the last year gets the student up to a minimum standard. Or perhaps more rigorously, the minimum standard is way below the published standards.
My middling suburban school desperately wanted students to graduate. It was both official policy and teacher mindset. I heard more than once, "I can't bear to fail X. You can't do anything without a diploma."
The school had systems to see who was "on track" to graduate. If a student wasn't, they could get tutoring, some times special classes. This was very much geared to "pass the tests".
Not too many people wound up in that situation. New teachers learned very quickly that failing more than a few students meant you would not be retained. There were lots of ways to keep that from happening, and still feel like you were keeping your integrity. If a test showed lots of students hadn't learned, you make an easier test. Before the unit test, you can give a practice test that has all the questions of the unit test, just phrased differently or in a different format--and make it clear that what has been tested is what they have to know. You can review the day before, being careful to mention everything on the exam and little else. You can even review in the first half of the period and test in the second. If all else fails, you can "scale" the test, giving everyone an extra five or seven points. You could then pretend (to yourself and to the outside world) that the students had learned the material. The fact that they would forget almost all of it within a few months was never faced.
Thank you for your thoughts, Roger, I think this provides a useful illustration of a situation that can be interpreted under each of the competing explanations of the sheepskin effect.
(1) Ability bias says that students who graduate are simply more capable than non-graduate students with nearly the same number of years in school. Researchers try to account for ability bias by only comparing students who are otherwise very similar, but of course this is hard to do perfectly and so probably this story still explains part of the earnings premium measured by sheepskin studies.
(2) Signaling says that students who graduate differ only because the diploma successfully communicates information about their traits (smarts, conscientiousness, and conformity, among others) to employers and that the student with nearly as much literal time in school is otherwise identical. No doubt this is part of what's going on too.
(3) Human capital says that the teachers tried their very hardest to teach the student who failed to graduate but they never quite met the standard required. For a high school student, this minimum amount of learning would likely consist of basic habits of conformity (such as their willingness to attend class or accept teacher help, show up to makeup sections, turn in 1st 2nd or 3rd chances on assignments, etc.) rather than the kind of knowledge that is typically measured in tests of retention. This is also a plausible.
[Note also that as it gets easier and easier to graduate from high school this poses a challenge for the signaling model because only a costly signal is worth anything to employers. There is no information value to a degree that is truly being given away, as Noah Smith has pointed out, also in my footnotes.]
And again I would stress that it is crucial to remember that the measurement of the sheepskin effect is very imperfect. Regression analyses most likely fail to isolate students with a degree who are truly comparable to students without a degree and as a result you could get an estimate that shows a 50% difference when the true number is much smaller. See work by Thomas Sowell (the shrinking residual problem) or Emily Oster (observational studies) for some general background on the methodological issues here.
"Note also that as it gets easier and easier to graduate from high school this poses a challenge for the signaling model because only a costly signal is worth anything to employers."
Which makes for a perverse incentive for the educational system as a whole. As the high school diploma becomes useless as a screen (because everyone has one), a college degree takes its place. This makes schooling more important. More money spent and more people employed (including the researchers who purport to show how useful school is or isn't).
It also means that high school teachers can say, "Sure he can't write very well, but they'll fix that in college." Middle school teachers say the same thing but kick the can to high school. In fact, elementary school teachers ... Modern schools try very hard not to break up age groups.
That matches my experience as well. I have never understood why teachers curve grades the way they do, straight up just adding points so not too many students failed, other than as making sure those students don't fail.
Many people reason, "In a just world, blacks would do as well as whites. Therefore, blacks must be given special treatment if that is required to equalize outcomes." Similarly, many teachers believe, "In a just world, everyone would get a diploma. Therefore, it is just to give extra points if that is required to pass the course and eventually get a diploma."
Ok finished the HC vs Signaling article. I wasn't really impressed, although I generally agree with his overall point at a high level.
The math is fine, whatever, it hardly matters.
The real thrust of his article is in classifying things.
Measuring the returns to education is hard. Amusingly he assumes this can even be done in the mathy section, despite constantly referencing measures of the returns that are non-equivalent in the rest of the text. In this he matches the general rhetoric of the pro-status-quo education system crowd: schooling does everything nice we can think of, not just teach you specific skills that you can test and measure, and that is why more school leads to more money.
Determining whether any observed effect, such as the sheepskin effect, is all human capital or signaling is hard. I agree, but not to the extent he makes it out to be, which is generally to bend over backwards to show how all the signaling effects could also be human capital. (Rarely the other direction, I notice...) Yea, sure, we probably can't tell in any given case whether it is 66.5% or 67% one or the other. We might be able to reasonably get it down to 50 vs 60, if we are careful, or at least "more or less than half".
Determining the mediating effects are hard. I kind of agree, but "mediating effects" seems to be a lot of "all the crap that could be either signaling or human capital but aren't really affected by schooling" which sort of makes the point vague to the point of uselessness. I think he is not even trying here.
One can't help when reading the article but think he really wants to say "Look, we can't tell exactly how much of the returns to education are due to actual human capital formation vs signaling of extant traits, so let's just pretend it is almost all human capital and move on." He clearly doesn't want to admit that we should be able to measure just how much human capital is being accumulated, or even identify what is supposed to be being accumulated.
In other words, I think he falls into the trap of "We can't measure this exactly, so I can believe whatever I want." A better way to look at that is to say "We can't measure this exactly, but it is a mix of A and B, and in roughly this range." That still rules out "whatever" here and there; Nick is only willing to rule out 100% of one and 0 to the other, and seemingly not much else, although he really implies that any high numbers for signaling must be wrong. I don't remember his ever pointing out that a high number for human capital must be wrong, but I might have missed it.
I am reading H-K's Human Capital vs Signaling (2020) paper now. I will have comments on it shortly.
To the point on the Medium article and quote, you are being WAY too charitable. He says that sheepskin effect, the well studied and highly tested effect that a huge portion of the returns to education come from getting the degree only, and do not scale commensurably with completion of the steps to get the degree, comes from the fact that getting the degree happens after you are measured to have passed some minimal bar. That only makes sense if the measurement only happens right before they give you the degree, and not every step along the way e.g. after every class, or every year, etc. In other words, his claim only makes sense if no one has much idea of what you might have learned in all that time between starting a college program and before you take that final exam that then reliably measures all you have learned, and having passed, you get the degree.
The American system works in exactly the opposite way: there is no final exam you must take, but rather a slew of small exams for every class. Nick's mental model of how the US education system works that supports his human capital story of the sheepskin effect is not at all how the system works despite his claim to the contrary.
And Nick knows this. He knows it perfectly well. How am I so sure? Well, he teaches at UC Santa Barbara, which might have a wildly different system, but to my knowledge does not.
Zvi is more on point, but is still a little off. He makes the assumption that college courses get more difficult as they progress. This is not always the case, and in fact it is often the case that failing courses gets more difficult as one progresses. Why? Because poor student evaluations are death for non-tenured faculty, especially from students within the major. In effect most students will not fail upper level courses unless they blatantly fail to complete course work, in the sense of "didn't show up for exam" or "failed to hand in term paper." However, from my understanding we do see an increasing return to years of education in college, where dropping out in the early years gives less return per year than dropping out in the third year.
I will note, however, that Zvi's short explanation there does not consider the empirically obvious issues of grade inflation or the like, and to be fair to Zvi almost no HC theorists' arguments do either. Nick H-K's paper brings up a south american study where the college stopped offering some courses, and the students saw salaries drop. Turns out the hiring firms had exams for incoming workers, exams that tested on those subjects. That's pretty funny, and pretty illegal here in the states for most jobs. I am curious how that sort of thing being normal here would affect the observed signals. About the only field I am aware of where that sort of thing is normal here is in programming, and as it turns out that is also a field where college degrees are much less valuable.
More people speak freely on more topics to bigger audiences than at any time in Western history. There is censorship, self- and otherwise, on some important topics, and it has ever been so. At no point in Western history was there a civil libertarian utopia, and much of the repression from when the "Christian substructure" was at its zenith far surpassed anything we see today. And one of the key factors for combatting repression from the past is in abundance today: loud, dissenting voices crying out for more freedom. And cultures retain their power to evolve new "substructures" for promoting and retaining practices conducive to freedom. Perhaps more so with all the new technologies and means and styles of communication.
I’m sure the people in Hong Kong are delighted that they have been able to put all of this new technology to use in being loud, dissenting voices.
1) The obnoxious sign language person in every public event is probably an unreasonable accommodation. They take up at huge portion of the screen and are very distracting. It can't possibly pass a cost benefit test. I guess you could say the same about a fair bit of ADA rules. It doesn't take long on Google to find massive evidence of a kind of abusive ADA lawsuit industry.
Pretty much anytime you accommodate some really small group its wrong.
2) "But deafness should not become an aspiration."
Everything not forbidden is mandatory.
https://twitter.com/RichardHanania/status/1665720633110437888?cxt=HHwWgICx-eSF6p0uAAAA
I think Richard raises a point. If something IS INFERIOR and you say IT ISN'T INFERIOR then you are already lying. And once you're lying a little, why not lie a lot. There is no natural stopping point for a lie. If a little lie will make gays feel better, a bigger lie will make them feel even better. And haven't they suffered enough!
3) Homosexuality is mental illness. Trans is just that same mental illness x10. There are a lot of the same drivers (narcissism, perversion, hedonism, etc).
https://emilkirkegaard.dk/en/2023/01/homosexuality-is-a-mental-illness/
4) "it does not propose any constructive steps to take"
What steps do you proposed to take? Perhaps in a red state debating a law there is something to do right now, but here in Virginia I know there is a 0% chance of school choice passing the legislature. During COVID there were huge parent protests at school board meetings but mostly the school board just kept doing what it wanted to do until the governor was able to narrowly pass a bill getting rid of masks (he needed a slim slice of democratic legislatures to pass it, but those same people would never pass school choice).
It's not exactly clear what regular people supporting school choice are supposed to do besides elect republicans if they feel the school choice issue is more important then other issues that might make them vote democrat, but that doesn't appear to be enough to shift legislatures.
School choice isn't the only option, but outlawing public employee unions is probably just as difficult to enact.
Awhile back the governor proposed that our entire school board be up for election that year, given all the scandals. However, he needed the state legislature to pass it. The legislature is pretty evenly divided with the DEMS having a slight edge. My district has a DEM.
So I contacted the state rep and made my case why I wanted to have a chance to elected a new school board. His reply was boiler plate progressive pablum.
The era of 1990s style bipartisan bills are over. If you want to do anything you are going to need a solid majority of your side, but no one individual can really being that about. It's basically set in stone by demographics. So it all feels hopeless.
From NS Lions: “Maybe liberalism could only ever really run on an operating system of specifically Christian foundations.”
This is something I’ve come to believe, and from what I’ve read of old books, the idea used to be a common assumption until it was killed off by militantly secular academia in the 2nd half of the 20th century. Good to see that this is being talked about more these days.
I agree that the Saudi's need to develop and economy not based on oil at some point, but I'm not sure they can.
Your average Saudi citizen is low IQ. They aren't going to develop a modern economy. They will need to import westerners to do that. But while a place like Dubai can do that, Saudi Arabia has 40M people. That means they need to import enough western knowledge workers to generate enough economic value to buy off 40M people. This isn't some tax haven city state living off the scraps of the west because it doesn't have to divide it amongst too many people.
I appreciate the desire to built something but we've seen countries with lots of natural resources waste them on vanity projects they claimed were productive before. All that seems different now is that they are allowing bread and circuses.
https://en.m.wikipedia.org/wiki/The_Line,_Saudi_Arabia - stuff like this make me skeptic
I remember watching a documentary about some island in the South Pacific that was ultra rich in some bird poop that made good fertilizer. For a couple of decades it was one of the richest islands in the world. They "invested" the money in all sorts of things. Hotels, broadway plays, etc. Of course all these investments ended up being retarded money losing ventures. At the end of the day they were just some dumb Pacific Islanders on an island of shit. Eventually they mined all the shit and went bankrupt. Now they are one of the poorest countries in the entire world.
> My father’s First Iron Law of Social Science: sometimes it’s this way, and sometimes it’s that way.
This is what the natural experiment evidence reads like to me. Arteaga (2018) looks more like the human capital story, while Hussey (2012) looks more like signaling. And both apply to a particular context in ways that make extrapolation to other settings difficult.
Re: "I present ten reasons to doubt that the education system is a waste of time and money." — Infovores, "Attention Caplanites: School is Less Wasteful Than You Think!", at link embedded in Arnold's blogpost.
Let me reply, quickly, tersely, point by point, to Infovores' stimulating critique of Bryan's case.
1. Misdirects policy issue. We should experiment, here and there, with alternatives to academic model, reductions in subsidies, vouchers, etc., instead of arguing about burden of proof (and standard of proof) for funding or defunding the whole system.
2. Conflates dropping out and flunking out. Ignores mismatch of curriculum and job skills/knowledge. Assumes dropout has learned less than all grads.
3. Doesn't establish that schools impart conformity (rather than filter non-conformity). (Treatment effect versus ongoing selection effects.)
4. College students are hamstrung by a collective-action problem around punishment of academic dishonesty. Detection usually would require tips and testimony by fellow students. Everyday bonds in campus residential total institution make students reluctant to report fellow students. Given low detection rate, admin ratchets up punishments (deterrence model). Severe punishments (e.g., expulsion) violate even honest students' norms of fair punishment, overriding collective interest in clear signal of academic standards. Many students reason, 'It could be me, in a pinch.' Compare Akerlof, Yellen, Katz, "Gangs, Law Enforcement, and Community Values" (1994), critique of Becker bricks-and-sticks model of deterrence when detection is low, community cooperation is needed for tips and testimony, and punishments exceed community norms of fairness.
5. Ignores mismatch of curriculum and job skills/knowledge. Maybe college calculus, which fades, nonetheless durably reinforces prior learning of maths, but few grads use maths on the job.
6. Bryan acknowledges that his 20/80 average HC/signaling ratio is a guesstimate, and that the ratio varies by context. PS: Does a 3rd causal factor — "ability bias" — figure in Huntington-Klein's model? Bryan carefully address also ability bias.
7. Miguel Urquiola makes this critique of Bryan, using examples of SCOTUS reliance on law clerks from Yale Law and Harvard Law. Bryan replies that the social costs of current education system far outweigh benefits of ever finer signaling. See debate at link below, Q&A portion:
https://www.youtube.com/watch?v=6j4PReevtqw
8. Doesn't establish that attentive mentoring in mismatched curriculum in school is more productive than less-attentive mentoring in job skills/tasks in the firm or org. Assumes ceteris paribus, but firms and orgs would focus more on mentoring (even with risk of employee defection), or other training mechanisms would emerge, if subsidies were withdrawn from education system.
9. Straw man. Although a gen ed course (e.g., foreign language) might have slight value, it doesn't pass cost-benefit test.
10. Ignores that college is also a country club that delays adulthood for many. Students quickly learn to cut corners. Country club exacerbates collective-action problem of achieving cooperation for high standards. Compare Jacob et al. "College as Country Club" (JLE): https://www.journals.uchicago.edu/doi/abs/10.1086/694654
Thank you for this John A, I always appreciate your thoughts!
Let me first say for the record that I am totally on board with making some big changes to the education system at the margin. I think The Case Against Education is a great argument for opposing “free college for all” and making some significant cuts (5, 10, maybe up to 20% even) but not for cutting 80% like Caplan wants to do. I would also support significant privatization of the education system, for reasons Hanania has outlined recently.
The reason I wrote this article the way I did, emphasizing (briefly, due to both reader attention and email limit constraints) points that don't tend to get noticed by Caplan fans anymore, is that I felt the consensus in the blogosphere had become overly ideological and dogmatic to the point of greatly exaggerating what proponents of the signaling model should reasonably expect those without their priors to believe based on evidence alone.
As I think the Ives Parr essay illustrates, many have even gone far beyond what Caplan himself argues for in the book (they talk as though everything were signaling, rather than 80%) and MUCH beyond what Caplan says follows from the evidence itself rather than philosophy. Here is the full quote from Caplan in the section "What I Really Think":
"Yet political philosophy is ultimately unavoidable, because philosophy sets presumptions. Some philosophies have a presumption in favor of education; others, in favor of the status quo. From these perspectives, the burden of proof rests on advocates of cuts. This burden is surmountable. Social return estimates amply justify spending 20% less on education given a presumption in favor of education or the status quo. But such presumptions still block radical reforms, because there is minimal concrete experience with radical reforms. We can speculate that cutting spending by 80% would be a great boon, but when the burden of proof is against you, speculation can’t surmount it. I favor radical reforms nonetheless. Philosophically, I am staunchly libertarian. While not absolutely opposed to taxpayer support for education, I have a strong moral presumption against taxpayer support for anything."
Infovores,
My intuition, from the trenches, is that it's not a 10% problem in higher ed. Not by a long shot. And I wonder why anyone who believes that it's merely a 10% problem would bother challenging the entrenched establishment to get elbow room for radical experiments.
I think we may be thinking about two different questions.
If the education system is 80% wasteful signaling, then we can cut that 80% and the education system will be improved.
My claim is that this estimate is incorrect and by a large margin. That doesn’t mean that there aren’t big problems in education, only that Bryan’s proposed solution (just cut spending) is not going to have the effect he thinks it will.
Just because I only favor cutting, say, 10% doesn’t mean I think the size of the problem is only 10% because cutting is not the only solution available. To me, privatizing the education system would be as big of a change as some of things Bryan proposes and I’m highly sympathetic to that.
Infovores,
Thanks for your reply and clarification.
Private colleges and public colleges, alike, have massive waste, and misallocation, and plenty of signaling. The main difference is that private colleges have less taxpayer subsidy per student than public colleges do. Nonetheless, not-for-profit private colleges, too, have a large structural subsidy — tax exemption — and some taxpayer tuition subsides (Pell Grants and the like). Public colleges have, additionally, large taxpayer subsidies for tuition for all classes of students and for facilities. Although private colleges must please donors to garner more philanthropic subsidies, whilst public colleges must please lawmakers to garner more taxpayer subsidies, it turns out that private colleges and public colleges operate very similarly — curriculum. faculty policy, growth in admin, country club — if we sort colleges by selectivity, size, degree type, residential vs commuter.
Among selective colleges, it happens that not-for-profit colleges and public colleges do compete for enrollments. Among non-selective colleges, there is some head-to-head competition among not-for-profit colleges, public colleges, and for-profit colleges.
Given the evidence from the current institutional mix, why would privatization, *with the same (or 90% same) level of taxpayer subsidies* (legislative appropriations and tax exemption), radically change college operations? (Again, we must sort the comparisons roughly, by institutional selectivity, size, degree type, and residential-total institution vs commuter institution.)
What am I missing?
PS: I did address, although too telegraphically, your 10 points against Caplan's case, which points out low HC/signaling ratio, negative social externalities of zero-sum credential game, and wasteful taxpayer subsidies.
No you're right, I am speaking about privatizing schooling prior to college, along the lines of what Hanania wrote recently (link below, I should have specified).
I believe Caplan supports cutting subsidies for masters programs most, then bachelors programs, and high school the least. This sounds roughly right to me and if you were to cut 10% of overall spending (~100 billion excluding loan subsidies), but focus it almost entirely on higher ed then you would be cutting in the ballpark of 20%. But take that with a grain of salt, as the policy particulars were not really the focus of the post.
Of the ten points you mentioned, some of them are addressed in other comments here. Sheepskin effect is probably the most important one if you want to CTRL-F for it, since Bryan uses that measurement directly to get his 50% lower bound for wasteful signaling and it is really not very defensible.
Regarding conformity, see Robin Hanson's post I link to and combine it with the overlearning insight—if conformity in K-12 is really "a full-time job" as Bryan complains elsewhere then it is hard to believe students are not becoming more conformist by sitting in school (isn't that the precisely what libertarian rhetoric tends to suggest elsewhere?).
On cheating, it doesn't really matter if there are other reasons why schools don't crack down on it because Bryan says that cheating erodes the signaling value of the degree. If college is all signaling, why does such a large earnings premium remain?
Happy to talk more, though it may be better to continue in the comments section on my post—I worry I'm crowding Arnold's page too much with this debate!
https://www.richardhanania.com/p/the-old-school-reformers-case-for
I'll set aside policy ideas for grade school. (For what it's worth, I favor radical vouchers for teens, valid at any training program, internship, apprenticeship, school, etc, with loose accreditation anywhere.)
Perhaps we're talking past each other about Caplan and college. To rephrase:
The challenge to Caplan's interpretation of the sheepskin effect seems to conflate dropouts and flunkouts, and to assume that dropouts learn less than all grads do.
Re: conformity. At least for teens, it would be much better to learn conformity in work or apprenticeship or training contexts, than via school curriculum mostly useless for job knowledge.
Re: overlearning as reinforcement of prior learning. I grant the phenomenon, but say that overlearning doesn't improve human capital if the prior learning (which is reinforced) isn''t useful job knowledge. Everyone (or most?) must take trig in HS, and perhaps the subset of students who then take and pass calc 1 in college thereby retain more trig, but very few of these students will ever use trig on the job.
Re: Academic dishonesty. It does erode (but doesn't eliminate) the value of the degree. If students could solve their collective-action problem (around tips and testimony), then the degree signal would be stronger, clearer, less noisy — and the return to the degree would be greater than it is. (I wonder if one might somehow check this via a well-designed study of the return to a degree at, say, Washington & Lee University — strict honor code — and at otherwise similar colleges, with same SAT/ACT level matriculants, without a real honor code. Probably elusive.)
I've replied here, to keep the thread intact. Thx for hearing me out.
As long a bellies are full and people think tomorrow will be better than today, government can get away with tons.
For sure the US recent past was a mix of authoritarianism and totalitarianism but are you suggesting what we have today is worse than restaurants that won't serve blacks, women fired when they get pregnant, men given job preferences because they are the "breadwinners," government programs that are whites only, etc.?
“Every LGB I've known has been pretty vanilla and normal."
More often than not we don't know what issues others have. With that said, I've known Ts who were as normal as anyone too. If someone comes well with depression, addiction, etc., does that mean it isn't an issue?
“no mental-illness issues for LGBs“
It would seem that T is more likely a mental health issue than LG (and B might be even less likely) but I don't think we know for certain whether any of these are mental health issues or not.
Whether chicken or egg, LGBTs have more diagnosed mental illnesses. If everyone were more accepting this might decrease but I'd be surprised if it ever went away completely.
Deafness - I agree.
With respect to NS Lyons. I'm guessing you could have written the same article about China 20 years ago. Certainly many of us, myself included saw great promise in the liberalization going on. Now we watch as Xi clamps down on disent and threatens Tiawan. I'm afraid there isn't some magical formula for freedom. Democracys can become a dictatorship of 51% or even just a loud minority. So too can the most benevolent dictatorship turn evil. I wish the Saudia's good fortune, but I'm not planning on moving there anytime soon.
I wrote up a point by point reply to Infovores here some folks might find interesting: https://dochammer.substack.com/p/the-wastefulness-of-schools
Overall, I think Nick H-K's work is highly motivated reasoning, possibly mixed with analyzing a different educational system than the US has. (That is the charitable interpretation of his description of what he describes as literally what people do.) Many of his critiques of Caplan raise the question of whether he actually read the book carefully, as his concerns are already addressed, at length.
Unfortunately most of the education economics literature reads a lot like "I would like this to be true, so I am going to claim it is", and only a little bit like "I would like this to be true, but damn... it looks like it isn't!" I trust the latter group a lot more!
I really like Infovores, and would recommend he not lean on someone like Nick H-K.
I don't think this is accurate of NHK, certainly doesn't come across that way in his writing (see medium article linked in my post for a dated, but very accessible treatment of his thoughts).
My sense from reading your essay DH (which I appreciated, thank you for taking the time!) was that you dislike a lot of the rhetoric schools use around teaching that make what they do sound more pure and wonderful than it really is. I sympathize with that to an extent while emphasizing that evaluating HC vs. signaling is largely a separate question from what schools say they do. It's possible for school to help students learn even if the rhetoric greatly overstates its impact or if it neglects to mention components of the schooling bundle that actually matter more (e.g. teaching conformity, punctuality, etc.). It may even be the case that schools' idealized rhetoric itself helps students to learn (I know I personally am much more motivated when I focus on the intrinsic value of learning something instead of the attendant conformity or the signaling aspects).
More directly to the point of whether NHK is reliable/correct— can you tell me where the econometrics is wrong in his Empirical Economics publication? The empirically unresolvable aspects of this debate are very important.
Regarding H-K, recall the paragraph you quoted:
"What schools actually *do* when they allow you to continue in your education is, effectively, measure what you’ve learned and see if it passes some minimum standard. If you don’t, you drop out. We end up with those failing the (lax) minimum populating the dropout years. They’ve learned little, so they earn little. In the final year, you see everyone who passes the minimum, whether they learned just enough or WAY MORE than enough. The final year contains a wide range of big learners, so on average there’s a big jump in earnings that year.
Intuitive, based on literal actions people perform, and a totally-HC explanation of sheepskin effects"
That is an absolutely obvious error in regards to the US school system, whether K-12 or college level. Obvious in the sense that everyone who is at all familiar with the US system has to know it does not work at all like that. At all.
So why does H-K claim it?
Is he ignorant of how the US system works, perhaps being an education economist that only studies a foreign country with a different system? That's a possibility, and it is fine certainly. But why does he present this as a counter to Caplan's book? Caplan is explicit that he is referring to the US system, and focuses on that system. Does H-K not realize this somehow? Perhaps because he didn't read it carefully? Why is he arguing about the ratio of human capital to signaling in the US system if he doesn't study the US system?
Or does Nick H-K know how the US system actually works, and makes this obviously incorrect statement because he wants to argue for a particular conclusion and will make claims about reality that support that conclusion regardless of their relation to reality?
I bring that up because I think you can make the econometrics say damned near whatever you want. No one is going to get your data and do a reproduction. Other people are going to do studies and probably find the opposite results, and there won't be a huge "OMG we need to figure this out!" response. It will just be published, and life will go on. I am not accusing Nick H-K of fraud as such, just that the rigor necessary to trust data arguments like this is not reached.
Much more readily available is to see the quality of the arguments they make outside the black box of their math. If they are making inconsistent arguments in prose, one can feel confident they are going to make the same mistakes when strolling the garden of forking paths in their data analysis, when it is much easier to make those motivated errors without even realizing it.
If I get some time this week I will try and dig through his econometrics for signs of sin, but I doubt there will be glaring obvious errors.
I think NHK goes to great pains to make his work accessible to a broad audience. He posted a less math heavy version of his published article for people to read (see my footnotes), and has a youtube channel dedicated to explaining causal inference used by economists in simple terms.
If there is an error in his medium article excerpt, it's possible that the error is partly mine as he speaks more precisely in the later articles he prepared over several years more of thinking on the issue in order to publish. But at least how I understood NHK's medium piece, what he means by "effectively measure against a minimum standard" is simply that schools require students to pass a set of requirements to get a degree and if you fail some of the required credits or leave early without completing them then you don't meet the minimum standard they consider adequate.
This observation is not that much different to my eyes from what Zvi says in an old Less Wrong post:
"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."
https://www.lesswrong.com/posts/tnK2tNc6fJ8PbYRkn/the-sheepskin-effect
As a former high school teacher, I am not at all sure that the last year gets the student up to a minimum standard. Or perhaps more rigorously, the minimum standard is way below the published standards.
My middling suburban school desperately wanted students to graduate. It was both official policy and teacher mindset. I heard more than once, "I can't bear to fail X. You can't do anything without a diploma."
The school had systems to see who was "on track" to graduate. If a student wasn't, they could get tutoring, some times special classes. This was very much geared to "pass the tests".
Not too many people wound up in that situation. New teachers learned very quickly that failing more than a few students meant you would not be retained. There were lots of ways to keep that from happening, and still feel like you were keeping your integrity. If a test showed lots of students hadn't learned, you make an easier test. Before the unit test, you can give a practice test that has all the questions of the unit test, just phrased differently or in a different format--and make it clear that what has been tested is what they have to know. You can review the day before, being careful to mention everything on the exam and little else. You can even review in the first half of the period and test in the second. If all else fails, you can "scale" the test, giving everyone an extra five or seven points. You could then pretend (to yourself and to the outside world) that the students had learned the material. The fact that they would forget almost all of it within a few months was never faced.
Thank you for your thoughts, Roger, I think this provides a useful illustration of a situation that can be interpreted under each of the competing explanations of the sheepskin effect.
(1) Ability bias says that students who graduate are simply more capable than non-graduate students with nearly the same number of years in school. Researchers try to account for ability bias by only comparing students who are otherwise very similar, but of course this is hard to do perfectly and so probably this story still explains part of the earnings premium measured by sheepskin studies.
(2) Signaling says that students who graduate differ only because the diploma successfully communicates information about their traits (smarts, conscientiousness, and conformity, among others) to employers and that the student with nearly as much literal time in school is otherwise identical. No doubt this is part of what's going on too.
(3) Human capital says that the teachers tried their very hardest to teach the student who failed to graduate but they never quite met the standard required. For a high school student, this minimum amount of learning would likely consist of basic habits of conformity (such as their willingness to attend class or accept teacher help, show up to makeup sections, turn in 1st 2nd or 3rd chances on assignments, etc.) rather than the kind of knowledge that is typically measured in tests of retention. This is also a plausible.
[Note also that as it gets easier and easier to graduate from high school this poses a challenge for the signaling model because only a costly signal is worth anything to employers. There is no information value to a degree that is truly being given away, as Noah Smith has pointed out, also in my footnotes.]
And again I would stress that it is crucial to remember that the measurement of the sheepskin effect is very imperfect. Regression analyses most likely fail to isolate students with a degree who are truly comparable to students without a degree and as a result you could get an estimate that shows a 50% difference when the true number is much smaller. See work by Thomas Sowell (the shrinking residual problem) or Emily Oster (observational studies) for some general background on the methodological issues here.
"Note also that as it gets easier and easier to graduate from high school this poses a challenge for the signaling model because only a costly signal is worth anything to employers."
Which makes for a perverse incentive for the educational system as a whole. As the high school diploma becomes useless as a screen (because everyone has one), a college degree takes its place. This makes schooling more important. More money spent and more people employed (including the researchers who purport to show how useful school is or isn't).
It also means that high school teachers can say, "Sure he can't write very well, but they'll fix that in college." Middle school teachers say the same thing but kick the can to high school. In fact, elementary school teachers ... Modern schools try very hard not to break up age groups.
That matches my experience as well. I have never understood why teachers curve grades the way they do, straight up just adding points so not too many students failed, other than as making sure those students don't fail.
Many people reason, "In a just world, blacks would do as well as whites. Therefore, blacks must be given special treatment if that is required to equalize outcomes." Similarly, many teachers believe, "In a just world, everyone would get a diploma. Therefore, it is just to give extra points if that is required to pass the course and eventually get a diploma."
Ok finished the HC vs Signaling article. I wasn't really impressed, although I generally agree with his overall point at a high level.
The math is fine, whatever, it hardly matters.
The real thrust of his article is in classifying things.
Measuring the returns to education is hard. Amusingly he assumes this can even be done in the mathy section, despite constantly referencing measures of the returns that are non-equivalent in the rest of the text. In this he matches the general rhetoric of the pro-status-quo education system crowd: schooling does everything nice we can think of, not just teach you specific skills that you can test and measure, and that is why more school leads to more money.
Determining whether any observed effect, such as the sheepskin effect, is all human capital or signaling is hard. I agree, but not to the extent he makes it out to be, which is generally to bend over backwards to show how all the signaling effects could also be human capital. (Rarely the other direction, I notice...) Yea, sure, we probably can't tell in any given case whether it is 66.5% or 67% one or the other. We might be able to reasonably get it down to 50 vs 60, if we are careful, or at least "more or less than half".
Determining the mediating effects are hard. I kind of agree, but "mediating effects" seems to be a lot of "all the crap that could be either signaling or human capital but aren't really affected by schooling" which sort of makes the point vague to the point of uselessness. I think he is not even trying here.
One can't help when reading the article but think he really wants to say "Look, we can't tell exactly how much of the returns to education are due to actual human capital formation vs signaling of extant traits, so let's just pretend it is almost all human capital and move on." He clearly doesn't want to admit that we should be able to measure just how much human capital is being accumulated, or even identify what is supposed to be being accumulated.
In other words, I think he falls into the trap of "We can't measure this exactly, so I can believe whatever I want." A better way to look at that is to say "We can't measure this exactly, but it is a mix of A and B, and in roughly this range." That still rules out "whatever" here and there; Nick is only willing to rule out 100% of one and 0 to the other, and seemingly not much else, although he really implies that any high numbers for signaling must be wrong. I don't remember his ever pointing out that a high number for human capital must be wrong, but I might have missed it.
I am reading H-K's Human Capital vs Signaling (2020) paper now. I will have comments on it shortly.
To the point on the Medium article and quote, you are being WAY too charitable. He says that sheepskin effect, the well studied and highly tested effect that a huge portion of the returns to education come from getting the degree only, and do not scale commensurably with completion of the steps to get the degree, comes from the fact that getting the degree happens after you are measured to have passed some minimal bar. That only makes sense if the measurement only happens right before they give you the degree, and not every step along the way e.g. after every class, or every year, etc. In other words, his claim only makes sense if no one has much idea of what you might have learned in all that time between starting a college program and before you take that final exam that then reliably measures all you have learned, and having passed, you get the degree.
The American system works in exactly the opposite way: there is no final exam you must take, but rather a slew of small exams for every class. Nick's mental model of how the US education system works that supports his human capital story of the sheepskin effect is not at all how the system works despite his claim to the contrary.
And Nick knows this. He knows it perfectly well. How am I so sure? Well, he teaches at UC Santa Barbara, which might have a wildly different system, but to my knowledge does not.
Zvi is more on point, but is still a little off. He makes the assumption that college courses get more difficult as they progress. This is not always the case, and in fact it is often the case that failing courses gets more difficult as one progresses. Why? Because poor student evaluations are death for non-tenured faculty, especially from students within the major. In effect most students will not fail upper level courses unless they blatantly fail to complete course work, in the sense of "didn't show up for exam" or "failed to hand in term paper." However, from my understanding we do see an increasing return to years of education in college, where dropping out in the early years gives less return per year than dropping out in the third year.
I will note, however, that Zvi's short explanation there does not consider the empirically obvious issues of grade inflation or the like, and to be fair to Zvi almost no HC theorists' arguments do either. Nick H-K's paper brings up a south american study where the college stopped offering some courses, and the students saw salaries drop. Turns out the hiring firms had exams for incoming workers, exams that tested on those subjects. That's pretty funny, and pretty illegal here in the states for most jobs. I am curious how that sort of thing being normal here would affect the observed signals. About the only field I am aware of where that sort of thing is normal here is in programming, and as it turns out that is also a field where college degrees are much less valuable.