Steve discusses 10 key graphs related to meritocracy and university admissions. Predictive power of SATs and other factors in elite admissions decisions. College learning outcomes - what do students learn? The four paths to elite college admission. Laundering prestige at the Ivies.


CLA and College Learning outcomes:

Harvard Veritas: Interview with a recent graduate

Defining Merit - Human Capital and Harvard University:

Chapter markers:

0:00 Introduction 
1:28 University of California system report and the use of SAT scores admissions
8:04 Longitudinal study on gifted students and SAT scores (SMPY)
12:53 Unprecedented data on earnings outcomes and SAT scores
15:43 How SAT scores and university pedigree influence opportunities at elite firms
17:35 Non-academic factors fail to predict student success
20:49 Predicted earnings
24:24 Measured benefit of Ivy Plus attendance
28:25 CLA: 13 university study on college learning outcomes
32:34 Does college education improve generalist skills and critical thinking?
42:15 The composition of elite universities: 4 paths to admission
48:12 What happened to meritocracy?
51:48 Hard versus Soft career tracks
54:43 Cognitive elite at Ivies vs state flagship universities
57:11 What happened to Caltech?

Music used with permission from Blade Runner Blues Livestream improvisation by State Azure.


Steve Hsu is Professor of Theoretical Physics and of Computational Mathematics, Science, and Engineering at Michigan State University. Previously, he was Senior Vice President for Research and Innovation at MSU and Director of the Institute of Theoretical Science at the University of Oregon. Hsu is a startup founder (, SafeWeb, Genomic Prediction, Othram) and advisor to venture capital and other investment firms. He was educated at Caltech and Berkeley, was a Harvard Junior Fellow, and has held faculty positions at Yale, the University of Oregon, and MSU.

Creators & Guests

Stephen Hsu
Steve Hsu is Professor of Theoretical Physics and of Computational Mathematics, Science, and Engineering at Michigan State University.

What is Manifold?

Steve Hsu is Professor of Theoretical Physics and Computational Mathematics, Science, and Engineering at Michigan State University. Join him for wide-ranging conversations with leading writers, scientists, technologists, academics, entrepreneurs, investors, and more.

Welcome to Manifold. Today, it's just me. I am going to be talking about meritocracy and university admissions, and I've prepared about 10 slides, 10 key figures that inform in a data driven way discussion about what is really happening in our universities today. I think you'll find this interesting. If you are listening to the audio only, I would suggest downloading the slides and looking at them as I give the presentation so that you'll have a better idea of what I'm talking about. But I'll try to make the conversation as understandable as possible, even for people who don't have the slides in order to, and in order to do that, I'll just summarize the main inferences that one makes from the data at the beginning before I discuss the fine details of what's presented in each figure.

So hopefully even without the slides, it'll be understandable to people who are listening to this discussion.

So let's start with the first, figures. The figures I'm about to show you are from a report which was formulated by a committee of both faculty members and some professional statisticians that are employed by the University of California system and this task force, this committee was asked to make a recommendation to the UC system as to the use of standardized tests like the SAT in admissions.

And as you probably know, the faculty recommended based on the results that I'm about to show you that the university should continue using the SAT for admissions. However, in what was widely described as a political decision, the regents of the University of California decided to stop using the SAT.

So let me show you the first figure here. So what is shown here is performance in college versus high school GPA and SAT score. They've broken down the data by ancestry group, and they've used a variety of performance metrics. So some are freshman performance metrics, such as non retention rate, i. e., how many kids fail to complete the freshman year; freshman grade point average; non graduation rate, i. e., what percentage of or what fraction of students fail to graduate; and grade point average at graduation. So I think these are fairly good metrics for describing the performance of each individual subset of students at the university.

There's a huge amount of data which goes into making these figures, many years of academic records and admissions files for the tens of thousands of students, actually hundreds of thousands of students attending the University of California system at any given time and color coding and the, in each graph represents different time, Subsets of the students in terms of their high school GPA, and then as you move from left to right on the figure. It goes from the, the, the group characterization goes from relatively low SAT score to very high SAT score.

And the inference you can take from this, these results is that for every ancestry group, whether it's whites or Asians or blacks or Hispanics, there is an improvement. In performance in all of these metrics, for students at the university with increasing SAT scores and independently of that, there is improvement with high school GPA.

But the data show very clearly that SAT score is an important separate factor that can be used in considering the academic strength of the applicant. And this is not surprising. I think most people would say that the SAT loads more on general cognitive ability, whereas high school GPA loads a bit more on conscientiousness or work ethic. And by combining these two metrics, you get a much more accurate predictor for how the student is likely to perform in college.

For those of you who can see the graph, I suggest looking at the far right column, graduation GPA. So that's the, in a sense, a summary of how the student did in their entire college career.

And you can see very clearly the gradient as we increase SAT score for any particular value of the high school GPA as you increase the SAT score, there's a pretty strong increase in the performance of this in the expected performance of the student as the SAT score gets higher and higher.

Furthermore, it's worth noting that the slope, the rate of increase of the met, whichever metric you choose. In this case, its graduation GPA and SAT score is about the same. And this reflects the observation. That's also widely recognized by experts that SAT is about as good a predictor of student performance in college, regardless of the race or ethnicity of the students involved.

So it predicts about as well for black students as for Asian students as for white students and as for Hispanic students. And in fact, it's even true that it predicts about as well, regardless of the family income or socioeconomic status. So if two kids admitted to the university come from very different family backgrounds, one is a rich family, one is a working class family, nevertheless, if they have similar SAT scores and say, similar high school GPAs, the prediction for how they'll perform in college is going to be about the same. And so it isn't, it isn't that the student is doing well in college because they come from a wealthy family, they're doing well in college because they have a strong preparation reflected in their high school GPA and a strong ability to do academic work reflected in their SAT score.

So, what's visible right now is the results for Asian and black students in the system. And let me just toggle the slide so you can see the results for Hispanic students and also white students. And you can see again, the slopes are pretty similar.

In every case, there's a very strong effect. There's a very strong gradient as you increase high school GPA. And if you control, if you fix high school GPA and you increase SAT score, there's also a very strong gradient.

So I think these are the best. Easy to understand graphs, which show to someone who's a complete non expert that SAT score is really essential for making good admissions decisions, i. e. for identifying the most academically able students from a large group of applicants.

So let me go on to another set of graphs. This one you may have seen before. This is from the study of mathematically precocious youth. This is a cohort of gifted children that were identified, already at age 12 or 13.

They were asked as part of their gifted program in school to take the SAT at an early age and a subset of that population, which was scored, scored in the top 1% level. So within the top 1% of kids in their age group. were inducted into the study, and then that population is further broken down into four quartiles, which are indicated on the figure as Q1, Q2, Q3, Q4.

And as you go from Q1, which are kids, which are just above the 1% threshold. So 99th percentile threshold. As you go all the way to Q4, you're talking about kids that are roughly at the 1 in 10, 000 level of ability. So even within the 1%, one can make a differentiation between kids that are gifted versus, say, what they call profoundly gifted. So that would be quartile four.

And what the figure shows is longitudinal data. So they followed these kids and in this study, SMPY, the participants now are in their 50s or possibly even older. And what is shown here is certain milestones or life accomplishments of the kids in each category, in each quartile. And the fraction of kids, the proportion of kids in each quarter, Q1, Q2, Q3, Q4, that achieved any particular milestone.

And you can see from the graph, if you could see it you'll see that as you move from Q1 to Q4, the proportion of students within that quartile that have that particular achievement goes up very drastically. So, just, I'll just read off what these different curves are. One is The proportion of students that have earned a doctorate degree, so Ph. D., M. D., or J. D. The proportion of each quartile that have published a peer reviewed paper in a scholarly journal. The proportion that have a peer reviewed publication in a STEM journal, so science, engineering, medicine, mathematics. The proportion that have a doctorate in a STEM field. The proportion that have been issued a patent by the U. S. Patent Office. The proportion that are very high income. And the proportion that achieved tenure in a STEM subject at a top 50 U. S. university.

So each of the milestones that I listed is obviously something that is a pretty significant accomplishment. In the general population, the base rate of people accomplishing these things is really low. But, in this cohort of gifted kids, the proportion is much higher than in the general population. And furthermore, as you go into that 1% further and further out in the tail, say to the 1 in 10, 000 population, you see a continuous rise in the proportion of kids, the per capita rate at which kids are earning later in life, these important achievements.

Thank you very much.

So, what this says is that the SAT is administered to 12 year olds and therefore it has a very high ceiling. It's difficult for a 12 year old to max it out. It's very much, much less likely for a 12 year old to get an 800 on the SAT math section than say a high school senior. The test therefore then functions as a high ceiling measure of ability.

And we see, to use the psychometric terminology, we see that this high ceiling test has validity. It does predict that it is a good predictor of the probability with which an individual is going to accomplish one of these important milestones.

So for people who say the SAT doesn't measure ability or doesn't have predictive power in the real world, or is only a reflection of family wealth, results like this strongly contradict those assertions.

Let me go on to the next set of slides.

These are taken from a paper by Raj Chetty and collaborators. Chetty is a Harvard economist who rather amazingly managed to get access to IRS data on essentially all Americans and also access to college admissions records at the most elite universities in the country and also at many state flagship universities.

And so he is able to make very, very granular statements about, for example, the predictive power of various components of the admissions portfolio that elite schools and flagship universities use to decide which students to admit. And so it's, it's very, very interesting data. Relatively recently, you can see this paper came out in July of 2023. So just a month ago. And I'm going to walk through what I think are some of the most interesting figures that appear in this paper.

So apologies if it's a little hard, if it's a little hard to read this 1, what's shown here is on the left hand side, that greenish curve is showing you the increase in the proportion of students that achieve a certain outcome, as a function of SAT score. So the SAT score varies from the left hand side. I think the value I can read off is about 1200. And on the right hand side, it's the maximum value 1600.

And so these outcomes, for example, being in a very high earning job, attending an elite graduate school, or working at a highly prestigious firm, you can see that all of those outcomes, the probability of all of those outcomes are continuously rising with SAT score.

Whereas if you look at the yellow line, which is the other column of results, having a higher high school GPA does not actually increase the probability of these outcomes. So this is an interesting finding that actually SAT is a stronger predictor of outcomes post college than high school GPA. So if you're going to keep one component of academic strength in the application, insofar as you're trying to predict accomplishments post college, it's actually better to keep the SAT than GPA and not vice versa. So that's a very interesting finding.

Okay. In this next figure, what is shown is the fraction of each subset of students who are working at an elite firm. So for each grouping of SAT scores, what fraction of students are working at an elite firm post college at age 25 and these results have in these results, they've controlled for family income and also the effect of attending an elite college.

And so just reading off the results. So there are 4 quartiles of SAT score range here ranging from scores below 1390, 1390 and below on the far left, but going up to scores which are close to 1600. And you can see that the kids that are scoring close to 1600 have a very high, something like 24% chance of working at an elite firm at age 25.

Whereas for students that scored 1390, which is still a high score, but not not anywhere near maxing out the SAT. For students in that range, the probability of them working at an elite firm at age 25 is only about 13%. So it's significantly less.

And so again, this shows the validity of SAT scores even after controlling for parent income and also for controlling. After controlling for the effect of the particular college that that particular individual attended, you can see that the remains as a separate factor, with which one can predict success in this case, success means working in an elite firm at age 25 post college.

So this is another Chetty figure. This figure shows the proportion or the impact, the probability of either attending an elite graduate school or working at a prestigious firm post graduation. And it looks at the effect of four different factors, which are part of the admissions portfolio. So they're, they're generally used.

They're generally factors considered in admission to an elite university here that the set of people considered are the students that have graduated from an elite university that that group of universities is defined as Ivy plus by Chetty. And the interesting part of this is that so, so I'm sorry. So the four are different factors which are taken into account in college admissions, which are studied here independent of whether you're a legacy. So whether your parents attended that elite university, whether you are an athlete. So recruited athletes say to play Ivy League squash or Ivy League football. Whether you have a high non academic rating. So that would be an evaluation of yours based on your essay, okay. Based on what the teachers say about you in a non, non academic sense and activities, extracurricular activities. So, a high nonacademic rating. And then the fourth bar in this bar chart is high is the effect of having a high academic rating.

And you can see that the only significant positive impact on the outcome of attending an elite graduate school or working at a prestigious firm after graduation from an elite university is the academic factor. So, if you have, if you are in the subset of applicants that have a high academic rating, you have a substantially higher probability of attending elite graduate school or working at a prestigious firm after graduation from one of these Ivy Plus universities.

Interestingly, and I think very surprisingly for most people who think they're well informed about college admissions. And in this category, I would put the actual college admissions committee members, the people who actually make the decisions about who is admitted to the universities. I would suspect almost all of those people are surprised to find that the non academic rating doesn't have any predictive power for whether the student later ends up attending an elite graduate school or ends up working at a prestigious firm. I think most people would say, oh, that's a separate independent factor that predicts success later in life for the student.

But, in fact what Chetty finds is that it doesn't have a signal. Being an athlete looks slightly negative. Actually implies a slightly lower probability of attending elite graduate school or working in a prestigious firm after graduation and being a legacy doesn't seem to have any impact one way or the other.

Okay, so here's another, a bar chart. It's showing the post outcome differences after controlling for college value add. So the amount of advantage it gives you to attend a particular college. And again, what is evaluated here is the impact of being a legacy, being an athlete, having a high non academic rating and having a high academic rating.

And you can see that this particular outcome metric is, working at a firm or in a job, which is likely to lead to being in the 1% in terms of top earnings. And so basically you know, an extremely good job with very strong economic prospects. So being a legacy had essentially no impact, did not increase your chance of being in a high earning job.

I think, in fact, in this case, it's specifically at age 25. So the probability of being in a high earning job at age 25 is not high, not any higher for legacies than for other students. It's not any higher for athletes than for other students, and it's not any higher for students with high non academic ratings.

However, it is significantly higher for students with high academic ratings. So just to give you the specific numbers, if you look at, Ivy Plus matriculants, so students that went to Ivy League and a few or a few other highly elite universities, say Stanford. They have about a 12% chance at age 25 of working in a firm, which puts you on a career trajectory to be in the top 1% in terms of earnings in annual income. So 12% is the baseline for all of the students from those Ivy plus schools and having a high academic rating raises that to something like 16%. So roughly 12 plus another 4%. So it's really a significant increase in the probability. So once again, it's shown that the academic factors have nontrivial signals.

The non academic or soft factors don't seem to have any signal, neither does being an athlete and neither does being a legacy. Now, remember what we found earlier that within this high academic rating, it looks like the cognitive ability measurement, which is the SAT, is the part of the academic rating that is contributing more to this increase in a better outcome as opposed to the high school GPA.

It's not the high school GPA that's causing this. It's probably the SAT score. Maybe other signals in there, like having done some research project or published a paper or having very good teacher recommendations evaluating your academic capabilities. Maybe those have signals as well. Because those could contribute to having a high academic rating.

But again, I think this kind of outcome is surprising to most people. Most people have this model that, the kids who are, you know, participating in lots of extracurricular activities in order to get that high non academic rating are going to be successful later in life. But at least as far as these outcome metrics are concerned that's not reflected in the data.

Okay. I think this is the last Chetty graph that I'm going to talk about. This is the treatment effect of IV plus admissions on income. And in this case, the metric that's used is the fraction of that particular group that has earnings in the top 1%. At age 33. So it's pretty far after college.

It's pretty long after college, 10 years to 11 years after graduating from college, who is in the top 1% in terms of earnings. And the method that they've used here is they've looked at students that were on the waitlist for a particular school, and then they compared the ones who did not get off the waitlist.

And then attended a state flagship university versus the ones who got off the waitlist and actually did attend their Ivy Plus school. Say they got off the waitlist and they attended Yale or they attended Princeton. And so they compare, they're trying to get at what did the kid get out of going to Princeton by getting off the waitlist versus a kid who didn't get off the waitlist and went to Rutgers. And the difference between the yellow bar and the green bar here on the graph is an estimate of what the kid got in terms of better prospects of being a top 1% earner at age 33 a decade later after graduation by attending Princeton instead of attending Rutgers.

So it's an interesting method to estimate the positive benefit of attending, say, Princeton instead of Rutgers.

And what is found here is an interesting result. So there is an advantage. It's roughly something like maybe a, you know, 1. 3 or 1. 5 times more likely for a kid who gets off the waitlist and gets to go to the elite school versus the kid who didn't get off the waitlist.

Now, that's pretty significant. So maybe it increases by 30% or almost 50% the chance that you're going to be a high earner, the same kid getting into Princeton as opposed to not getting into Princeton and then attending a public university.

Now, some people have criticized this method. Ideally, this would be a kind of random method where you have kids that are sort of good enough to go to an elite university, and then you just roll a, you flip a coin to decide whether they get to go there or else they have to go to the public school.

And so you, you compare the outcomes later, later in life from that coin flip. Here you could say, well, there's some systematic effect, like the kids that are getting off the waitlist are actually stronger in terms of human capital than the ones that didn't get off the waitlist. But I think Chetty had all run some tests to see whether that was true, and it didn't seem to be true.

And one of the ways they checked, I believe, was that they found that getting off the waitlist and getting into a particular one of the schools that you're waitlisted by is kind of a random event. Then you wouldn't expect that, conditional on you getting off the waitlist to say, go to Princeton, that the probability that you also got off the waitlist to go to one of the other Ivy Plus schools was higher. Indeed, they found that.

So it wasn't it wasn't true that conditional on getting off the waitlist at one of the schools you applied to increased the probability that you got off the waitlist at the other one of the other schools. And so that's consistent with the assumption that it's essentially random who gets off the waitlist, or at least close enough to random that these two populations of kids, the yellow colored bars on the bar chart and the green colored bar are roughly equivalent in terms of human capital.

So again, a very unique data set that lets you do things like try to actually estimate the so-called treatment effect of having been admitted to an elite private university as opposed to going to a public university.

Okay the last set of graphs I'm going to show you are from this very long paper called Test of Validity of the College Learning Assessment. And let me just say a few things about what the College Learning Assessment is. There's a lot of interest in trying to figure out what is the value add from attending university.

And so in the previous slide, we were just talking about what, how much greater is the value add. for, say, going to Princeton as opposed to going to Rutgers, right? So that's what Chetty was trying to get at in the last graph. You could go further and say, like, well, let me just try to measure what skills the kid has after graduating from college versus what skills they had when they entered college as freshmen.

And so that would be a kind of direct measurement of the things that college is supposed to give you, like some, some particular skills. The College Learning Assessment was created, through a collaboration between large corporations and, researchers on testing researchers.

And so the tests that comprise the college learning assessment are practical. They're measuring skills that employers, large corporations, actually say they want from their white collar knowledge workers. So, for example, it could be a task like reading an article and summarizing it, or looking at these graphs and summarizing what's in the graphs. Write a report summarizing some information. So very practical things, not a strange abstract IQ test. Rather really it tries to be representative of the kinds of things that quote knowledge workers do in our modern economy.

And the thought was that by creating something like the CLA, College Learning Assessment, one could identify students who say had not been very impressive coming out of high school, went to a not very prestigious university, but learned a lot while at university. And by the time they come out, they have really good skills. They have really good skills that make them valuable to a big corporate employer. And if you could administer this CLA to all the graduating seniors who apply for a job at your company, you might find many diamonds in the rough. Even though they don't have an elite pedigree, they might have a very high CLA score and you might then be comfortable hiring them into a very sought after position at your company.

So that was kind of the thought for CLA. Now, people who are familiar with psychometrics would say, well, it's probably going to turn out to be the case that something like the SAT, which already measures general cognitive ability at age 17 or 18 and, you know, general cognitive ability is pretty stable once you reach adulthood.

The thinking might be that, well, CLA is actually not measuring anything independent of SAT, in fact, it's probably measuring some general factor of intelligence plus some narrower skills, like if you try to break CLA subtests into, say, more math type stuff or more verbal type stuff.

In fact, that's what the researchers found. So there was a very big study done. There were 13 universities that participated in this study and ranging from at the most elite end MIT and the University of Michigan, but then also many less well known directional state colleges, much less a prestigious, much less elite, university. So there's a huge range of universities that participate even also, I think, some historically black colleges as well.

So it was found that indeed the CLA can be decomposed into a kind of general factor of capability with some sub factors. So not surprising. And the most interesting aspect of this research is what it shows about the degree to which students actually improve In their knowledge worker skills or generalist skills or general analytic skills, critical thinking skills. How much do students actually improve through 4 years of college education? And the answer, which I think is pretty discouraging if you look at it carefully. The answer is that actually there isn't much improvement.

So, the way this is done is ideally you would do a longitudinal study where you, you, you test some incoming freshmen, you wait until they graduate from college, and then as they're graduating from college, you test them again. So that any, for any particular individual, you can see the increase in their critical thinking skills, writing skills, et cetera.

The researchers were not able to do that. So they did it all in one year. So basically they, what they did is they tested the cohort of incoming freshmen and they tested the cohort of graduating seniors and then they compared them and the seniors did better on the CLA than the freshmen, but not really that much better.

And, typically the effect sizes would be something like less than half the standard deviation. Sometimes the effect size is actually negative because in high school, students are required to take math. They actually know some math when they take the SAT and apply to college. But in many majors in college, you're not required to do any math. And so some students actually are worse at math when they graduate from college than when they come in. So the delta, CLA delta on the really math loaded stuff is actually negative. For due to the impact of four years of college, which is sad.

So, first of all, the inference here is that college is not actually teaching you, at least not very much, critical, not improving very much your critical thinking, your generalist capabilities.

Even your writing skills are not really improving that much for most kids through college. And so, I think this is a shocking result for most professors and higher ed administrators. Although, actually, to be honest, most, most, most of them are not aware of these results. But if they were aware of these results, they'd probably be shocked.

Now, just to be careful here, we're not saying that college fails to teach narrow skills. So, in other words, If there were a test of accounting, freshmen, very few freshmen probably know anything about accounting, but the seniors who majored in accounting would probably have improved their score tremendously.

If there were a school leaving test on chemistry or electrical engineering or C++ programming, you would probably find the people that majored in those subjects perform much better on that subject test than the incoming freshman. And indeed, that's the case. That's why we have G. R. E. subject tests in mathematics, computer science, physics, chemistry. That's for measuring the mastery of the student of that particular subject area when they graduate from college.

And indeed, there is dramatic improvement in narrow focused knowledge. But the claim which has become popular among humanities professors and some social science professors that study of their subject builds critical thinking skills or builds generalist capabilities or writing skills or reading comprehension skills that is not found in these results.

Now, one subtle aspect of the analysis that was done in these, in this paper, which kind of reveals something very interesting here is that one of the problems here is that, the seniors that are tested in a given year, if you try to compare them to the incoming freshmen who take the CLA again, this is across 13 different school, the universities, major universities.

The seniors generally are a more select population because there's a, there's a, a lot of these schools, there's a fairly high attrition rate. So the students that make it through college to their senior year are generally somewhat stronger than the freshman. So only a subset of freshmen survived to the senior year and generally it's the more able ones.

And so the researchers had to find a way to correct for that factor because they didn't do a longitudinal study. They were just comparing, you know, 1 senior of a given year versus a freshman of a given year. And so they found that they could correct for this selection effect by adjusting for the SAT scores of the individuals, because the higher SAT score students entering the university were more likely to persist until their senior year. So they found an average, a delta in average score between the seniors and the incoming freshmen. And so they use that to correct their results so that they could directly compare the average scores among seniors versus average scores among freshmen.

But, of course, that reveals the fact that that is a reasonable method for making the correction. That reveals that all of this work to build the CLA was in some sense not capturing that much more information than what was already known about the particular student from their SAT score, which was obtained when they graduated when they were leaving high school.

Okay, so the nightmare message from the CLA study is that college doesn't really improve your general cognitive abilities or skills relevant to the modern world, generalist skills relevant to the modern workplace. There is some kind of stable, relatively stable quantity of mental ability or general cognitive ability, and it's fairly well measured by the SAT.

So kind of a negative outcome in the minds of many college administrators and professors, but for people who have studied psychometrics, they realize this is actually, the reality that we live in. And in fact, they could have predicted this before the whole exercise of designing a college learning assessment and then doing this huge study with 13 colleges. And I think there's a huge number of kids tested, like 100, 000 kids tested or something like that. So anyway, the results were somewhat predictable to people who understand cognitive ability and psychometrics.

But if you're a parent, and you're considering borrowing, allowing your child to borrow hundreds of thousands of dollars to attend university, you might want to think very carefully about what the kid is going to get out of university.

If they're not really going to improve their critical thinking skills, and if they're not likely to actually improve the generalist skills that corporations want in their employees, then why is the kid going to college? There might be other reasons why it's good for the kid to go to college. But I think many people mistakenly believe that the deltas here in capabilities as measured by CLA are going to be larger as a consequence of college than they actually turn out to be.

And again, just to emphasize again, if you're going to college to learn how to be an engineer or learn how to be a chemist or learn how to be an accountant, none of that is impacted by these particular results. To get at that, you would look at something like the G.R.E. subject exams. And there, of course, there is significant learning that goes on that can be measured by those assessments.

So, that concludes the graphs and data that I wanted to discuss with you. Again, I recommend to anyone who wants to think seriously about college, either from a self interested standpoint, like a parent or a child, a student, deciding where to go to college or whether to go to college, or a policy person trying to make recommendations for how society should handle higher education, invest in higher education.

Everyone should be aware of these results and the inferences that follow from the data that I just showed you. Unfortunately, I would say after 30 years of being a college professor and almost a decade being a high level administrator at a Big Ten university, I would say very, very few professors and very, very few administrators understand the content of what I just shared with you. But I believe these research results are strong. They're all based on extremely large data sets. The analysis I think has been done properly and the inferences that I discussed are actually correct.

So we have a very big knowledge gap between the best social science not inference about meritocracy in college and what the general opinion, general consensus opinion tends to be even among educators and college administrators.

So let me close with just some final remarks which I've summarized here on the slide, or I've outlined here on this slide. So what, what is the model of the elite university? So Chetty spent all this time studying what they called Ivy Plus universities. So Ivy league universities plus Stanford plus some other prestigious universities.

MIT. Maybe Caltech. Duke. Well, what is the model? I think one thing that people should realize is that admissions to those universities, or at least primarily Ivy like universities, really proceeds along roughly four different tracks. So it's not as if you can just say, like, there is a type of Harvard student or there is a type of Brown student. Actually, there are really four, in a sense, types of students attending those schools. Okay, so, and, and on the slide here I have, I've described the four categories as smart, rich, interesting with a question mark, and diverse, diversity. Okay, so those are the four components.

Smart means the academic admits. The kids they're admitted primarily for their academic strength, and that might only be something like a quarter of the students in the class.

If you look at the 75th percentile achievement level for admits to these Ivy Plus universities, they're pretty strong. They're scoring, you know, there's their scores getting close to 1600 well above 1500. They've taken many AP courses. They probably have done some science research project or something like this, published an article or something.

They're, they're quite strong. But that's only a small portion of the class. There's another group of the class that's there because literally because they're rich. Because their families have a lot of money. They are potential donors and those students might be substantially weaker than the ones that were admitted just on academic merit.

Okay. Now, overall, if you just look at the numbers, there aren't a lot of academically weak kids at these Ivy Plus universities in absolute terms. They're all well above average in terms of relative to the average in the general population. But the groups other than the smart group are far less able than what the university could get.

on campus if they only cared about academic merit. That's one of the main points I would like to make. Okay, so the first group is the academic admits, which I call smart. There's another group called rich. There's another group I call interesting with a question mark. Interesting means the kids that have high non academic ratings, that somehow the nice lady on the admissions committee fell in love with because the kid was in a lot of clubs and maybe went to El Salvador to build a school and the teachers really spoke highly of this kid and consequently that kid got in with probably significantly weaker academic credentials than the kids who got in in the smart category.

But this is a very large group in Ivy Plus, on Ivy Plus campuses, kids that were really there because they're interesting, they're academically much stronger than the average high school graduate in the country, but they're nowhere near the very top group, okay?

And what's interesting about Chetty's work is that his work suggests that these criteria that the nice ladies on the admission committee are choosing to single out their quote interesting kids to give them a high non academic rating or personal rating is what Harvard calls it in their admissions process. Those kids with a high personal rating, they don't necessarily do that well later. There doesn't seem to be a lot of evidence in, or any evidence in Chetty's work that those kids are actually more likely to say, have high paying jobs, get into top graduate programs etc. Okay, so that's the third category.

And the fourth category, as we all know, is diversity. So universities are trying very hard to have a kind of, Representative balance of different ancestry groups. And so, as we know from, say, the Harvard SFFA versus Harvard Supreme Court case, students with much weaker academic credentials are getting into these Ivy Plus schools simply on the basis of their race or ethnicity. And now the Supreme Court has said that that is unconstitutional, that that is illegal. So something may change there. Although the stated strategy of Harvard, which was released, sent to the alumni right after the Supreme Court decision was they would continue to try to make the campus diverse and they would do so by allowing students to make to, to show that somehow their ancestry is related to their overcoming some big obstacle which they could describe in their essay and therefore in a way they're, they're going to push all those diversity admits through the quote, interesting category. They're basically going to give them a high, effectively a high personal or non academic rating because they're, they come from some particular ancestry group that's desired by social engineers.

Whether that stands up in court, I think SFFA will be watching and many other lawyers will be watching very closely, how this works out in the coming years. Maybe that will work, maybe it won't work for these for schools like Harvard.

Now, point number two here, has to do with laundering prestige or laundering merit. And so what these schools are really doing is that they are trying to get the employer or the average person who thinks about, say, a Yale graduate. They're trying to associate all of their graduates with the academic smarts of the smaller group of academic admits. So it's a kind of bait and switch. If you don't know which of the kids was admitted to Yale for being brilliant, then you might just assume that all the kids coming out of Yale are brilliant. Even though that isn't really true, if you for a sufficiently high definition of brilliant.

And over the last 20 years, people I know who do hiring literally people like Goldman or McKinsey or hedge funds or venture funds or tech firms, people more my age, have said many people have said to me over the years, well, I just don't see consistent quality now, even in Harvard grads. So literally Harvard grads, people will say, I can't count the number of times someone has uttered a sentence like this to me. Hiring a kid from Harvard is a crapshoot. You're not sure what you're getting. You could get a kid who's really brilliant or you could get a kid that's just okay.

And I think that's literally a consequence of the admissions model that I just described in point number one. You have this weird mixture of kids with very, a very large range of abilities. If you doubt this claim, you should listen to this earlier, earlier episode of Manifold where I interviewed a recent graduate of Harvard who had been really an academic admit, I think primarily, and is now doing a PhD at an elite graduate program in, I believe, statistics or machine learning, something like this. And he recounts, this is an hour long, hour plus long interview with him, he recounts how surprised he was at the vast range in academic strength of kids on the Harvard campus. And so employers are starting to detect this, elite employers, elite firms are starting to detect this. And you'll notice the elite firms are trying to filter for the ones that are, where the academic admits. That's probably why, in Chetty's results, being a high academic rating kid within the admitted pool at Harvard does increase substantially the chances that you're at one of these prestigious firms when you're 25 or 33. So the employers are kind of onto it and they're no longer just trusting the Harvard admissions office to do their filtering for them.

They have to then do further filtering among the Harvard grads to figure out which ones are the real good ones. And so we don't know what the long term viability of this is. So, the long term viability, you know, it could work out that people gradually say, well, okay, Harvard is not doing a good job filtering human capital anymore because of their other goals like social justice. And so the, the boost, the leg up used to apply to every Harvard grad is now only going to apply to the subset of Harvard grads that we've scrutinized very hard and decided, yeah, this kid was one of the high academic ability admits.

In discussing this kind of thing. Sometimes I refer to firms as hard versus soft and by hard, like either firms or job categories. I mean, particular jobs that require actual technical skills and for which it becomes evident right away that you don't have these technical skills.

So if you're hired to trade derivatives and you don't understand the Black Scholes model, you're going to be in trouble right away. If you're hired to develop software, and you don't understand the basic, you know, algorithms for sorting, you're going to be caught right away. Et cetera, et cetera. Right.

And so, those firms I think are going to be less and less enamored of, say, the Harvard pedigree over time. But what about the soft firms that are looking for generalists? People who don't, aren't being asked to do something very specific and technical. They're, they're being asked to be glib and clever, maybe charm clients, maybe analyze something and give a kind of very subjective opinion.

There, of course, you might say it'll be a long time before people figure out who are the talented ones and who aren't. In fact, if you're cynical, you could say in a lot of these generalist jobs, the main task for you is to get along well with your peers, charm your peers and your boss, and that's how you get ahead because they don't actually ever test. You're not ever actually tested on real performance. The whole point of the job being soft is that there's no hard measurement of your performance that happens with any frequency.

I think what really distinguishes a lot of these Ivy kids from the kids who go to the state flagship universities is that they generally have more ambition. They just have more naked ambition, more drive to get to the top, more will to power, so to speak. And, of course, those qualities are useful. Those qualities in a way might be more useful than being good at some technical subject.

And so for that reason, I think still, I think in generalist career tracks, say at soft firms it will be a long time actually before the prestige or the halo associated with being in one of these top, having graduated from one of these top universities goes fully away.

We're nowhere near, society deciding that meritocracy is completely dead and, and these hyper elite universities are not as prestigious as, they don't deserve to be as prestigious, the prestige that they currently have. I think we're still quite far from that fading away.

One thing I did notice because my kids are high school college age and a lot of their friends and kids of my friends are also in that age group.

I'm kind of aware of what's going on right now in college admissions and college applications. And one of the interesting things that's happened is that because so many kids want to study computer science and certain fields related to computer science or technology it's getting very, very hard to be admitted.

For example, at state flagship schools, say, University of Illinois or University of Michigan Berkeley, you have to be admitted directly to the CS program or to the engineering program. And the kids that are admitted directly to those programs out of high school are actually very strong academically.

I would say they're comparable to the top 25, you know, top 25% of Ivy Plus kids. So the kids that are above the 75th percentile at these Ivy Plus colleges, the ones that I would refer to as academic admits.

So it could be that more and more kids of that quality level. And we're talking really about maybe the top 5, 000 high school graduates a year in the country, something like that, who have ordered thousands of maybe less than 10 kids each year, the top 5, 000 most talented kids out of high school every year. More and more of them are ending up at state flagship schools and not at Ivy Leagues because, as I mentioned earlier, the Ivy Leagues are not really trying to fill their whole class with kids at that ability level. They're only trying to fill a fraction of their class with kids at that ability level.

So that's an interesting phenomenon for me, and we'll see how that plays out. We may go back to the state of affairs that held when I graduated from high school, which is that most families were perfectly happy to send their kid to the local flagship university, and they weren't really gunning for the Ivy League.

All of that happened from long after I went to college, and now we're still at sort of peak frenzy of applicants, kids applying to like, because of the common app, applying to like 30 schools or something like that, and really hoping to get into one of these Ivy Plus schools. But in reality, I think a very large chunk of talented kids end up going to their state flagship schools, particularly to study technical subjects.

Final comment about Caltech and MIT. MIT was one of the schools that for a while made the SAT optional. I think it was because of COVID. But then they did a careful study and concluded like the UC study that I talked about at the beginning of this presentation. They concluded they really needed the SAT to figure out which students were really able to handle the MIT curriculum.

And so that now they've gone back to requiring the SAT they are one of very few schools in the country right now that requires the SAT for admission. So MIT I would say did the right thing.

MIT did what the faculty committee recommended that the UC should do, University of California should do, but the University of California regents who are political appointees did the opposite. Okay, so that's the situation. MIT is doing the right thing.

But what happened to Caltech? My alma mater. Caltech seems to have been taken over by a bunch of woke leftists.

The, the, the single individual who probably did the most to make Caltech a world class institution for science and technology is a guy called Robert Millikan. Millikan won the Nobel Prize in physics for measuring the charge. He was the first person to directly measure the charge of an electron. It's a pretty fundamental accomplishment in the early 20th century, right?

He did this. And the big library in the middle of the Caltech campus used to be called Millikan Library. And in fact, in some of the official histories of Caltech, it's sometimes the nickname for the school, or at least at one time the nickname for the school was Millikan School because he did so much to build it up. He was the president of Caltech.

Well, these wokesters decided that because Milliken was a, quote, eugenicist, they would strip his name off the library. And now he, you know, is no longer an honored, you know, founder and builder of Caltech, past president of Caltech. So, pretty alarming.

And consistent with that, Caltech no longer, not only does not require SAT scores, they actually tell applicants not to mention anything about their SAT scores in their application.

For alumni like myself this is exactly opposite to the whole ethos of Caltech. So the whole ethos of Caltech, you know, for the last gosh, almost 100 years now has been a very strong adherence to meritocracy. And generally they're very numbers driven. They, they, they want to actually see how you perform on either bespoke tests that they used to administer or standardized tests like the SAT.

So it's just shocking where Caltech has ended up. I don't really know what the internal dynamics are at the school, but it's not looking good.

A lot of alumni like myself refuse to give them money now because I think the administration there is very wrong headed. And I fear for the future reputation of Caltech.

And by the way, the GRE subject tests, the GRE exams, which I said earlier were useful for figuring out which kids had learned a lot of chemistry in undergrad, as undergrads, or learned a lot of physics or a lot of math as undergrads. Caltech is now, for their graduate program, not requiring those GRE tests the way they did in the past. They've made the GRE test optional. And I just can't understand how they can ensure a high quality, general quality of a student submitted without the data that's included in those GRE results.

So, anyway, I think the situation is quite bad for Caltech. I congratulate MIT for doing the right thing. It seems like the MIT administration right now knows what they're doing. And the Caltech administration does not.

So, I hope this discussion will be interesting to you. As I said, I will make all these graphs, these slides available. To you, and you know, please feel free to share them with any friends or colleagues that you think are interested in the real state of meritocracy and university admissions in the United States.

Thanks a lot for your time.