I read this post and it got me thinking about the value of college in terms of tribalism / homophily—i.e., whether universities function partly as “in-groups” that shape who gets opportunities later. So I asked ChatGPT, and here’s the summary it generated of our discussion:
“There’s pretty strong evidence that hiring is “school-networked,” not random.
• Startups/founders: Large administrative-data work finds founders disproportionately hire people who share the founder’s educational background—often around ~10% of hires overall and higher early on (e.g., ~14% in year 1), consistent with alumni ties shaping who gets pulled in.
• Established firms: In elite U.S. law firms, law-school concentration is far above population baselines. Example: two random lawyers from the overall sample share a law school about ~1.6% of the time (and ~0.7% in the broader population), but in the median firm it’s ~5.6%—i.e., several times higher than “chance,” implying strong school-based pipelines.
• Mechanism evidence: Multiple datasets show graduates are more likely to join firms where their university peers/alumni already work, and those hires often have better outcomes (wages/tenure), consistent with referrals, information flow, trust, and coordination advantages.
• Big caveat: This doesn’t always prove “taste-based bias” (i.e., irrational favoritism). The same pattern can come from referrals/private information and recruiting pipelines. But in markets where applicants are already pre-screened and “similarly qualified,” research on elite hiring suggests similarity/‘fit’ can become a decisive tiebreaker—so the result can look like favoritism even if partly rationalized as risk reduction.
Net: yes, people/firms hire from their own schools at rates well above random mixing, especially via alumni/peer networks, with the strongest evidence coming from linked employer–employee datasets and industry studies.”
Yes, strong agree! I think of universities as creating multiple layers of in-groups—e.g. your fraternity, your alumni network, and finally all university-educated people.
The "in-group" phenomenon is I think the thing requiring sociological explanations, because it's pretty hard to explain from an economic perspective.
I always thought Scott Alexander’s post on back scratchers clubs did a good job summarising the usual rational choice explanations for in group favouritism because after all, in most in groups, if you don’t indulge in this favouritism, other people will not indulge in this favouritism towards you, and they might even go out of their way to punish you. Of course there are also obvious evolutionary reasons for such preferences, even without getting into rational choice theory so I’m not saying the rational choice explanation is the end all and be all of possible explanations, but I’d be interested to know why you think economics does a bad job understanding this phenomenon.
In addition to sociological and economic causes, we also have to ask why people are still going to college in an era that promises, or threatens, AGI or ASI. If the current generative-AI studies on the usefulness and effectiveness of AI turn out to be robust, and if the scaling laws continue, then much of knowledge work is slated for massive disruption. We may not even need systems like ChatGPT to trigger that disruption; simpler tools such as checklists, step-by-step procedures, and standardized processes have repeatedly been shown to outperform humans at the tasks they target, assuming comparable information and typical circumstances. Even as an AI supporter, perhaps even an AI optimist, I can grant that in genuinely novel or exotic edge cases humans may retain some advantage, although that advantage may shrink if scaling trends continue. In that sense, it is hard not to think of Paul Meehl’s 1954 work on clinical versus statistical prediction. Metaculus, as of the time of writing, suggests that strong AGI could arrive by 2035. Given that possibility, why hasn’t college enrollment fallen? Scott Galloway has argued that enrollment isn’t down. One explanation is inertia: the same forces that made education attractive to students and employers may simply persist. But that answer ignores the possibility that AI is a structural shift, one that could change what credentials and training mean in the first place.
The obvious explanation is that they either don’t realise how transformative AI will be, which is almost certainly true of the wast majority of the population or are uncertain about the exact effects of transformative AI on the job market and which professions it will impact, so don’t think giving up on college is a smart idea given the available information.
This entire essay doesn't include the phrase "disparate impact". That's like having an essay about why ice cream sells better in the summer without mentioning temperature.
Hard to see how it could be a crucial factor when it's much less important in Europe and yet European countries have even higher college attendance rates.
"Traits like employees’ appearances, political correctness, and ability to intuit social norms don't help much with the object-level work involved in most jobs."
I think you undervalue the importance of people-to-people interactions within the companies. My take is that these traits are very highly correlated with "soft skills" that tend to make a technically competent employee a very useful one for the company.
Another more meta point I want to make is that I think there is still value in Bryan's theory as university is likely the easiest way to signal all three parts of the trinity together. As someone who wasn't born in the US, I would also add that figuring out other ways to credibly signal these things to future employers is very much non-trivial, and could very well incur excessive cost of acquiring necessary information, at least in the pre-LLM world. My hunch is that this would be even true for a large fraction of people born in the US.
"My take is that these traits are very highly correlated with "soft skills" that tend to make a technically competent employee a very useful one for the company."
Yes, I agree. But I think it's very interesting that *these* are what Brian highlighted, rather than the actual soft skills people use. In other words, there's an open question about whether the things people call soft skills are more like skills or more like identity markers, and the way Brian wrote this list suggests he's implicitly thinking about the latter.
"figuring out other ways to credibly signal these things to future employers is very much non-trivial"
As an individual, yes. As a society, it wouldn't take much work. And so the question is less "why does an individual not break the equilibrium?" and more "why is *this* the equilibrium rather than something else?"
I found a large part is option value assuming you'll outperform, but it gets affected as larger parts of the populace also go get the same signal and make it ubiquitous.
Glad you're writing this and looking forward to reading the next installment. I always found Caplan's theory persuasive and struggled to find any critique of it that was remotely convincing. This is the first critique I've seen that made me start to question it. Though I suspect what you're going to argue is just a more sophisticated version of his model, or in any case something very different from the human capital account, which would ultimately lead to the same implication of education being a waste of time and money.
On the IQ thing -- I don't think the fact firms don't explicitly test for intelligence doesn't mean they aren't trying to measure it with degrees. I reviewed the literature on IQ and productivity, and I inevitably had to answer why firms didn't use IQ testing if it was effective. My answers:
- Sometimes it doesn't make sense to do so, if the job is unskilled, or if the skill is highly legible -- like in musicians.
- About 40% of the American population either doesn't think IQ tests measure intelligence well, or are not sure whether they do. Even if incentives are aligned, many managers, executives, and founders may choose not to select based on IQ.
- Beyond being controverisal, IQ tests are gauche; stigmatised on grounds of being racist, classist, and ageist. They assign people a single number that is assumed to measure intelligence, something considered to be innate and valuable, a dynamic which creates resentment.
- IQ tests are also boring, cold, and lazy. Some peope might choose to do reviews or holistic evaluations because they feel better, even if they don't have the same cash out.
So there is no need to explain why firms don't use IQ testing in the signalling model of education. I do, however, think that alternative theories (social, ability signalling) are drowned out in the academic debate. Ability-confounding in particular is also underrated and underdiscussed as an explanation for the education~earnings association, with at least 50% of the variance being explained by that alone.
Regarding models for education returns, a different question: if university didn't pay, would people still go?
I think some people definitely would. Not as many as there are right now, but it gives people an opportunity to regularly see people their age, of both sexes, who are highly intelligent. It makes learning easier, extends childhood, and gives people a sense of direction and credibility.
I think the prestige signalling value of the degree isn't just about getting a job, thankfully there are few startups, agencies that rely more on proof of skills and portfolios but where the signalling comes in handy is to anchor your identity in team work, in social circles and gatherings. You can be saying absolute non-sense and yet people will be more receptive to your ideas and take you seriously purely because of the prestige credentials, so because of that the person has to spend less time in fighting to get heard and just do the actual work. People just assume you know shit and don't do the mental calculus of whether to trust you or not, so in that way people with the credential have less bitter experiences in forming and maintaining relationships with all sorts of people whereas those without the credentials
I agree with most of what you said though I would add that people don't seem to be appreciating the value of college as a consumption good. It's a lot of fucking fun and helps you make friends and find romantic partners in ways that really improve the rest of your life.
I'd suggest that a huge part of what is going on here is just people consuming something they enjoy. Sure, it's a lot of money but ask people to rate the monetary value of having more friends, meeting a compatible partner, meeting enough people to know who that is etc and I think they would give you astonishingly large numbers. Also the weirdness really matters here because you connect with others through the rest of life via common experiences.
The problem is that we need to play this game of justification in terms of training etc because of the sense (by both parents and taxpayers) that it's inappropriate for others to fund such frivolous fun.
---
But I think it's a bit hard to say that Caplain is wrong and you are right -- seems the real difference ultimately comes down to whether the resulting explanation that involves all sorts of preferences about looking normal stretches the rational market model beyond the point of usefulness. And here I think it's relevant that Caplain seems to have a general issue appreciating that the utility of the rational model stems from implicit assumptions about the kind of preferences that are relevant. I mean in full generality you can basically explain any behavior with some preferences but see his attempt to treat mental illness inside that model.
Interesting essay and look forward to the follow-up. I'm not fully following, though. Showing your SATs as a signal plausibly works for an occasional 1550 kid. A little weird but fine. Would it work for the 1020 SAT, the median kid? Wouldn't that be more of a harmful disclosure? So the 4 years of college is more of a "laundering" mechanism?
> So although doing well at college signals more conscientiousness than lazing about, an even better signal of conscientiousness would be acquiring all the same knowledge without attending college at all! In fact, people who are capable of learning college-level material independently should be going out of their way to avoid college lest they be confused for those who can only do it within a motivating social structure.
A lot of weight rests on this short dismissal of conscientiousness, and I don't think that weight is supportable. I'm about halfway through the book and Caplan points out that getting a GED provides no benefit to high school dropouts, so even on purely selfish grounds they shouldn't bother. Your theory that learning outside of school should be a better signal doesn't match the reality. In his blog posts Caplan emphasizes that being willing to sit through years of boring school is ITSELF the signal for conscientiousness, which cannot be substituted with some knowledge test. The world of work is not simply one of accumulating academic knowledge to pass a test, it is WORK. A brilliant autodidact speed-reader who can accumulate the necessary knowledge to pass a test in a short time is not necessarily someone who is going to come into work day after day without slacking off.
How would you design a test for conscientiousness? If you did, you would literally revolutionize personality research. I don't think you can do it in less than several years. If it was shorter, say one week, the signal would be very weak because non-consciountious people can push hard for one week if the stakes are as high as their entire careers.
Is there anyone for whom higher education is not a waste of time and money? Do you mean it's a waste of time and money only for the top percentiles of intelligence and motivation?
One chapter of Bryan's book calculates the selfish value of higher education. He concludes that college is a good deal for good students, but a bad deal for poor students who are less likely to graduate and less likely to have lucrative majors. His larger argument on signalling is that it's SOCIALLY wasteful, and thus we should cut subsidies to it.
I read this post and it got me thinking about the value of college in terms of tribalism / homophily—i.e., whether universities function partly as “in-groups” that shape who gets opportunities later. So I asked ChatGPT, and here’s the summary it generated of our discussion:
“There’s pretty strong evidence that hiring is “school-networked,” not random.
• Startups/founders: Large administrative-data work finds founders disproportionately hire people who share the founder’s educational background—often around ~10% of hires overall and higher early on (e.g., ~14% in year 1), consistent with alumni ties shaping who gets pulled in.
• Established firms: In elite U.S. law firms, law-school concentration is far above population baselines. Example: two random lawyers from the overall sample share a law school about ~1.6% of the time (and ~0.7% in the broader population), but in the median firm it’s ~5.6%—i.e., several times higher than “chance,” implying strong school-based pipelines.
• Mechanism evidence: Multiple datasets show graduates are more likely to join firms where their university peers/alumni already work, and those hires often have better outcomes (wages/tenure), consistent with referrals, information flow, trust, and coordination advantages.
• Big caveat: This doesn’t always prove “taste-based bias” (i.e., irrational favoritism). The same pattern can come from referrals/private information and recruiting pipelines. But in markets where applicants are already pre-screened and “similarly qualified,” research on elite hiring suggests similarity/‘fit’ can become a decisive tiebreaker—so the result can look like favoritism even if partly rationalized as risk reduction.
Net: yes, people/firms hire from their own schools at rates well above random mixing, especially via alumni/peer networks, with the strongest evidence coming from linked employer–employee datasets and industry studies.”
Yes, strong agree! I think of universities as creating multiple layers of in-groups—e.g. your fraternity, your alumni network, and finally all university-educated people.
The "in-group" phenomenon is I think the thing requiring sociological explanations, because it's pretty hard to explain from an economic perspective.
I always thought Scott Alexander’s post on back scratchers clubs did a good job summarising the usual rational choice explanations for in group favouritism because after all, in most in groups, if you don’t indulge in this favouritism, other people will not indulge in this favouritism towards you, and they might even go out of their way to punish you. Of course there are also obvious evolutionary reasons for such preferences, even without getting into rational choice theory so I’m not saying the rational choice explanation is the end all and be all of possible explanations, but I’d be interested to know why you think economics does a bad job understanding this phenomenon.
In addition to sociological and economic causes, we also have to ask why people are still going to college in an era that promises, or threatens, AGI or ASI. If the current generative-AI studies on the usefulness and effectiveness of AI turn out to be robust, and if the scaling laws continue, then much of knowledge work is slated for massive disruption. We may not even need systems like ChatGPT to trigger that disruption; simpler tools such as checklists, step-by-step procedures, and standardized processes have repeatedly been shown to outperform humans at the tasks they target, assuming comparable information and typical circumstances. Even as an AI supporter, perhaps even an AI optimist, I can grant that in genuinely novel or exotic edge cases humans may retain some advantage, although that advantage may shrink if scaling trends continue. In that sense, it is hard not to think of Paul Meehl’s 1954 work on clinical versus statistical prediction. Metaculus, as of the time of writing, suggests that strong AGI could arrive by 2035. Given that possibility, why hasn’t college enrollment fallen? Scott Galloway has argued that enrollment isn’t down. One explanation is inertia: the same forces that made education attractive to students and employers may simply persist. But that answer ignores the possibility that AI is a structural shift, one that could change what credentials and training mean in the first place.
The obvious explanation is that they either don’t realise how transformative AI will be, which is almost certainly true of the wast majority of the population or are uncertain about the exact effects of transformative AI on the job market and which professions it will impact, so don’t think giving up on college is a smart idea given the available information.
This entire essay doesn't include the phrase "disparate impact". That's like having an essay about why ice cream sells better in the summer without mentioning temperature.
Hard to see how it could be a crucial factor when it's much less important in Europe and yet European countries have even higher college attendance rates.
Caplan discusses things like Griggs vs Duke Power in the book, and doesn't find it a convincing explanation.
"Traits like employees’ appearances, political correctness, and ability to intuit social norms don't help much with the object-level work involved in most jobs."
I think you undervalue the importance of people-to-people interactions within the companies. My take is that these traits are very highly correlated with "soft skills" that tend to make a technically competent employee a very useful one for the company.
Another more meta point I want to make is that I think there is still value in Bryan's theory as university is likely the easiest way to signal all three parts of the trinity together. As someone who wasn't born in the US, I would also add that figuring out other ways to credibly signal these things to future employers is very much non-trivial, and could very well incur excessive cost of acquiring necessary information, at least in the pre-LLM world. My hunch is that this would be even true for a large fraction of people born in the US.
"My take is that these traits are very highly correlated with "soft skills" that tend to make a technically competent employee a very useful one for the company."
Yes, I agree. But I think it's very interesting that *these* are what Brian highlighted, rather than the actual soft skills people use. In other words, there's an open question about whether the things people call soft skills are more like skills or more like identity markers, and the way Brian wrote this list suggests he's implicitly thinking about the latter.
"figuring out other ways to credibly signal these things to future employers is very much non-trivial"
As an individual, yes. As a society, it wouldn't take much work. And so the question is less "why does an individual not break the equilibrium?" and more "why is *this* the equilibrium rather than something else?"
I looked at the value of college in detail a few years ago: https://www.strangeloopcanon.com/p/should-you-go-college
I found a large part is option value assuming you'll outperform, but it gets affected as larger parts of the populace also go get the same signal and make it ubiquitous.
Glad you're writing this and looking forward to reading the next installment. I always found Caplan's theory persuasive and struggled to find any critique of it that was remotely convincing. This is the first critique I've seen that made me start to question it. Though I suspect what you're going to argue is just a more sophisticated version of his model, or in any case something very different from the human capital account, which would ultimately lead to the same implication of education being a waste of time and money.
On the IQ thing -- I don't think the fact firms don't explicitly test for intelligence doesn't mean they aren't trying to measure it with degrees. I reviewed the literature on IQ and productivity, and I inevitably had to answer why firms didn't use IQ testing if it was effective. My answers:
- Sometimes it doesn't make sense to do so, if the job is unskilled, or if the skill is highly legible -- like in musicians.
- About 40% of the American population either doesn't think IQ tests measure intelligence well, or are not sure whether they do. Even if incentives are aligned, many managers, executives, and founders may choose not to select based on IQ.
- Beyond being controverisal, IQ tests are gauche; stigmatised on grounds of being racist, classist, and ageist. They assign people a single number that is assumed to measure intelligence, something considered to be innate and valuable, a dynamic which creates resentment.
- IQ tests are also boring, cold, and lazy. Some peope might choose to do reviews or holistic evaluations because they feel better, even if they don't have the same cash out.
So there is no need to explain why firms don't use IQ testing in the signalling model of education. I do, however, think that alternative theories (social, ability signalling) are drowned out in the academic debate. Ability-confounding in particular is also underrated and underdiscussed as an explanation for the education~earnings association, with at least 50% of the variance being explained by that alone.
Regarding models for education returns, a different question: if university didn't pay, would people still go?
I think some people definitely would. Not as many as there are right now, but it gives people an opportunity to regularly see people their age, of both sexes, who are highly intelligent. It makes learning easier, extends childhood, and gives people a sense of direction and credibility.
See
https://www.technotheoria.org/p/does-selecting-employees-for-iq-work
https://www.technotheoria.org/p/why-does-everybody-hate-the-white
For sourcess.
> IQ tests are an example of a very cheap but very credible signal
It is definitely possible to study for IQ tests
https://www.technotheoria.org/p/do-iq-tests-measure-intelligence
Just 3 retakes is enough to boost IQ scores by 7 points. Anecdotally, both my SAT/GRE scores increased by 1 SD after studying harder.
I think the prestige signalling value of the degree isn't just about getting a job, thankfully there are few startups, agencies that rely more on proof of skills and portfolios but where the signalling comes in handy is to anchor your identity in team work, in social circles and gatherings. You can be saying absolute non-sense and yet people will be more receptive to your ideas and take you seriously purely because of the prestige credentials, so because of that the person has to spend less time in fighting to get heard and just do the actual work. People just assume you know shit and don't do the mental calculus of whether to trust you or not, so in that way people with the credential have less bitter experiences in forming and maintaining relationships with all sorts of people whereas those without the credentials
I agree with most of what you said though I would add that people don't seem to be appreciating the value of college as a consumption good. It's a lot of fucking fun and helps you make friends and find romantic partners in ways that really improve the rest of your life.
I'd suggest that a huge part of what is going on here is just people consuming something they enjoy. Sure, it's a lot of money but ask people to rate the monetary value of having more friends, meeting a compatible partner, meeting enough people to know who that is etc and I think they would give you astonishingly large numbers. Also the weirdness really matters here because you connect with others through the rest of life via common experiences.
The problem is that we need to play this game of justification in terms of training etc because of the sense (by both parents and taxpayers) that it's inappropriate for others to fund such frivolous fun.
---
But I think it's a bit hard to say that Caplain is wrong and you are right -- seems the real difference ultimately comes down to whether the resulting explanation that involves all sorts of preferences about looking normal stretches the rational market model beyond the point of usefulness. And here I think it's relevant that Caplain seems to have a general issue appreciating that the utility of the rational model stems from implicit assumptions about the kind of preferences that are relevant. I mean in full generality you can basically explain any behavior with some preferences but see his attempt to treat mental illness inside that model.
Interesting essay and look forward to the follow-up. I'm not fully following, though. Showing your SATs as a signal plausibly works for an occasional 1550 kid. A little weird but fine. Would it work for the 1020 SAT, the median kid? Wouldn't that be more of a harmful disclosure? So the 4 years of college is more of a "laundering" mechanism?
> So although doing well at college signals more conscientiousness than lazing about, an even better signal of conscientiousness would be acquiring all the same knowledge without attending college at all! In fact, people who are capable of learning college-level material independently should be going out of their way to avoid college lest they be confused for those who can only do it within a motivating social structure.
A lot of weight rests on this short dismissal of conscientiousness, and I don't think that weight is supportable. I'm about halfway through the book and Caplan points out that getting a GED provides no benefit to high school dropouts, so even on purely selfish grounds they shouldn't bother. Your theory that learning outside of school should be a better signal doesn't match the reality. In his blog posts Caplan emphasizes that being willing to sit through years of boring school is ITSELF the signal for conscientiousness, which cannot be substituted with some knowledge test. The world of work is not simply one of accumulating academic knowledge to pass a test, it is WORK. A brilliant autodidact speed-reader who can accumulate the necessary knowledge to pass a test in a short time is not necessarily someone who is going to come into work day after day without slacking off.
How would you design a test for conscientiousness? If you did, you would literally revolutionize personality research. I don't think you can do it in less than several years. If it was shorter, say one week, the signal would be very weak because non-consciountious people can push hard for one week if the stakes are as high as their entire careers.
Yes, I thought there was too much handwaving away of that angle.
Is there anyone for whom higher education is not a waste of time and money? Do you mean it's a waste of time and money only for the top percentiles of intelligence and motivation?
One chapter of Bryan's book calculates the selfish value of higher education. He concludes that college is a good deal for good students, but a bad deal for poor students who are less likely to graduate and less likely to have lucrative majors. His larger argument on signalling is that it's SOCIALLY wasteful, and thus we should cut subsidies to it.