Reflections on an Education

on processing eight years at universities

I recently finished graduate school (plus two years of work in academia) and, as I head back into the “real world,” have started to process the experience. Some unsolicited thoughts.

The big vs. small debate

Against big

I did my undergraduate degree at the University of Virginia, by all accounts a good state school and a relatively large one. The comparisons have always been to the University of Michigan, or maybe UCLA. But one of the things I struggled with then was the idea that the school was big, that I would be lost in a sea of people. Common advice on campus was to “find the school within the school.” I found this sort of annoying—if we’re paying so much money to be here, shouldn’t it be easy to get attention and resources? But a fight over resources was exactly how it felt. My freshman year advisor was a graduate student in the Religious Studies department (I was hoping to become a chemical engineer at the time). I got some horrible guidance and ended up in a math class I wasn’t prepared for, the most-failed course in the undergraduate course register. Unsurprisingly, I had a bad time. Luckily, I had a good tutor (my now-wife). But it was really a hassle. A lot of time was spent in 300-person lecture halls, which resulted in a learning experience akin to Khan Academy. Some of the negatives of that experience were on me: I should have done my research more, and I was ultimately to blame for enrolling in the wrong courses. But it felt like I was swimming upstream.

Fast forward a few years when I landed at the Jackson School of Global Affairs at Yale and the experience was the polar opposite. With only 100-odd students in the entire school (both classes and dual-degree students) and an immense amount of resources, the resources- and faculty-per-student ratio was quite high. Finding part-time work, research opportunities, and grants was so easy it felt like the default. In two years, not once did I fail to get into a class I wanted. Paid trips to New York, DC, and abroad abounded. Conferences were funded. Even the catering was abundant. Things just felt . . . easier. Part of that is just the Ivy effect, but I felt that I could get my arms around the school in a way that even peers from other bigger schools within Yale didn’t report. From this experience, I concluded that smaller schools had some real value proposition that bigger environments couldn’t quite replicate.

In defense of big

But there’s nuance to this. It’s hard to control for the Ivy effect, and I have a sample size of two. I also saw some serious small-school limitations. Breadth is hard to replicate. My now-wife then-girlfriend at UVA ended up in the architecture school. She commented a number of times that, had we been at a tiny liberal arts school, that simply wouldn’t have been an option for her. And she’s right about that. If a large part of the value proposition of an undergraduate education is to figure out what you want to spend your life doing, then there is serious value to diversity of offerings, and if you want to see “the world,” then school should be as close to a microcosm of that world as possible. Smaller environments will, by definition, struggle to offer that.

The Jackson School managed well in this respect, but only because of a unique arrangement it had made that was somewhat idiosyncratic to Yale. At Jackson, you could enroll in course from other schools. Students had no trouble taking courses at the schools of business, law, environmental science, public health, or the College. This was a serious privilege: business school is expensive, but we were permitted to take whichever courses we wanted under the same Jackson tuition/scholarships. Professors at those other schools could cross-list their courses with Jackson. But it was also a near necessity: course offerings would be too limited at a small school like Jackson, so to get the educational experience the school aimed for, it essentially had to lean on the other schools. In this respect, characterizing Jackson as “small” is a little disingenuous: we benefitted hugely from the big university we also called home.

This model also has drawbacks. Because Jackson isn’t sufficiently large on its own, to attract talent, the school has to mostly offer joint appointments with other departments. So, someone might get a joint appointment at the Department of Economics and the Jackson School. Since tenure tracks are mostly in those “secondary” departments, there is an incentive to spend more of your time in those departments than at Jackson, since they will afford you more publication opportunities. In other words, the “small school within the big school” really needed the big school for it to exist. There are economies of scale and resources that smaller institutions cannot generally replicate, but which are necessary for producing high-caliber research and teaching.

Speaking to high schoolers

When I’m asked from time to time to speak with high school students about their futures, my advice about school size then is basically to not index on size too much. Go where you feel you’ll thrive and where you feel like you’ll be a good fit. I generally gravitate towards smaller organizations, but I found my school within a school at UVA. And if you think you’re a “big school person,” there are ways to broaden your reach within a smaller school. Among the many considerations (not least cost and job prospects), I don’t think size should be highest on anyone’s list of importance. Just go in clear-eyed about the costs and benefits of both models.

What to take and when

Like many Foreign Affairs types (yes, chemical engineering to foreign affairs), I had a pretty generalist education in undergraduate. I specialized after that, during my working time, to focus on the financial system. But even in graduate school, I found myself one semester learning about finding fixers in war zones in the morning and learning about DSGE models in the afternoon. Policy school is by construction pretty generalist, or at least as specialist as you make it. I think I had a pretty standard “policy” education. I looked back over the total six years of school and added up the courses; it came out to roughly this:

• 35% STEM + Quantitative Economics and Finance (e.g., math, computer science, statistics, securities analysis, macroeconomics, etc.)

• 30% Humanities (e.g., history, American politics, international relations, area studies)

• 20% Language Instruction

• 15% Other Stuff (mostly political science, financial history, business, and leadership)

I got the job I wanted, so I guess I can rationalize this approach as having worked, but I don’t think it’s replicable or suitable for everyone. In graduate school, I did tailor my courses pretty considerably to my field. My advice to folks starting out in school is as follows:

Take the “harder” stuff in school, with some caveats

By “harder” I don’t mean difficulty, I mean “hard” as in “hard science” or “hard skills.” This is not because I think they’re more important (actually, in the working world, I think being able to communicate effectively, both verbally and written, is probably the most important determinant of success). It’s because of the marginal value of in-classroom instruction for some fields versus others. Essentially, the question to ask yourself is, “how much easier would X subject be to learn in a formal classroom setting versus on my own by reading or taking online courses?”

I have always enjoyed learning languages, so naturally wanted to take some at Yale. An advisor talked me out of it, and his advice stuck with me (paraphrasing): “You only get 16 classes here; language pedagogy has come a long way, you’d probably get better instruction in a language here than you would at a community college or on Duolingo, but not by that much. In contrast, you can only take a finance class with Jeffrey Garten here.” He had a point: there are certain things that cannot be replicated well outside of a school, and some that we’ve made much more headway with. Consider history for example. If you take a history class at a university, you’ll find that roughly 70% of what you learn is learned through intensive readings, probably another 15% through guided lectures discussing those materials, and another 15% through your writing assignments. From this, it follows that you could probably get roughly within a 70% striking distance of the knowledge from that class by getting a good reading list and going it alone (although the quality of a university reading list is absolutely something of value). Try learning calculus on your own, though, and you’ll find it’s much more difficult. That’s not to say you should only take “quant” classes in school (I clearly didn’t!), but it is to say that you should be mindful about the marginal value-add of a university setting when selecting courses.

The caveats. Caveat one is exploration. This is why I believe liberal arts education is still of tremendous value. If you only ever took “hard skill” classes, we probably would have fewer inspired lawyers and academics (or even just human beings), and that would be very sad. Some of the most inspiring courses I’ve taken have been soft skill small-group classes where you speak with incredible people and make incredible connections; that in and of itself is a marginal value-add of a university that is very hard to replicate.

Caveat two is expertise and connections. If you are at a university with some of the best historians around, you should probably take a class from them, because that marginal value-add is much bigger.

Caveat three is “time in the saddle,” or subjects in which immersion and repetition are hugely important and school provides a straightjacket forcing you to spend that time. This is why I’d still defend taking language courses at a university. They’re often five days a week, one hour a day, plus lots of homework. That’s a bummer for your schedule, but plenty of language pedagogy research tells us that the primary determinant of success in a foreign language is simply time on task. You aren’t likely to be able to devote that amount of work in such a concentrated manner to a language outside of school, but maintenance is a lot cheaper than construction, so learning a language in school and maintaining it after is a lot easier than the other way around.

Pedagogical quality is real, but metrics lag

There is a now-tired debate in academia about whether we should firewall teaching and research. The logic for separating them goes something like this: some people are really good researchers but really bad teachers (or at minimum very disengaged teachers), and there might even be a negative correlation. Even if there’s not a correlation, if someone takes a job for the research, they’ll view teaching as a chore, and that’s unfortunate for students. Let’s self-segregate and have a separate “tenure”-type track for people who are just really good instructors. This is just an application of the specialization of labor principle: don’t force everyone to do everything, let them specialize (teaching vs research) and the result is greater than the sum of its parts. Cards on the table, I have long been pretty persuaded by this argument, but I am tempering my view a bit now.

I have seen first-hand that there are some truly extraordinary people who are both extremely good researchers and extremely good teachers. I ran into a lot of them at Yale, but they’re everywhere. I maintain that they are probably few and far between, but I do have a friend in academia who holds the “latent good” view: that if you’re so good at learning a subject as to become an excellent researcher in it, you’re probably good enough at learning stuff that you can learn to become an excellent instructor, even if you’re not one by nature. My counterpoint would be that there are still institutional incentive reasons that a very good instructor would actively try to avoid teaching (e.g., to rack up publications for tenure), but I take his point. I think there are also important distinctions between undergraduate and graduate schools: you need great researchers available to PhD students, so you can’t really firewall it at that point. That isn’t as applicable to undergraduates, who in any case probably are more sensitive to quality of instruction given their age. I therefore prefer some binning of those two disciplines at least at the undergraduate level, but I’m not partisan about it.

Setting the debate aside, though, one thing that is certainly true is that some people are really good at explaining stuff, and other people aren’t as good at explaining stuff. I don’t know if that’s an innate or a learned thing, but it’s a thing. Over the past couple of years, I have watched two different instructors teach the same topic, and I have seen wildly different levels of success. I watched one instructor spend a confusing ten minutes explaining a topic I saw another instructor explain flawlessly in two minutes with a pithy analogy.

Another observation is that being very good at doing a thing in general does not necessarily correspond to being good at teaching it. The Jackson School is famous for its Senior Fellows, people who are “professors of the practice”: not academics, but decorated practitioners. You can think of former heads of state, or retired generals or admirals or diplomats. Some of them are very successful instructors. Many of them are not.¹ This probably points to some amount of support for the hypothesis that teaching is a learned skill.

The challenge for students past and present is that transparency about pedagogical skill isn’t where it should be, and varies considerably from university to university. Both at UVA and at Yale I have seen student- or school-sponsored ranking tables, in which student reviews get aggregated by class and sometimes by professor. I think this is probably the right approach, but in the case of core required classes probably doesn’t do much for you. Outside of those, I strongly endorse using these tools. Universities themselves should take them a bit more seriously in my view, but again, tenured professors have a bit of a “competitive moat,” if you will. My advice to students, though, is to take those rankings very seriously. Even if the class content looks great, if the consensus is that the instructor is terrible, avoid it at all costs.

Is it even about pedagogy?

This perhaps gets at the bigger question of why we have higher education at all. Is it to prepare people for the workforce? If so, via education or via connections? Is it to inspire people and create a thoughtful voting populace necessary for a functioning democracy? Is it not about teaching students at all, but rather about fixing a market failure in provisioning high social-value research (in this case, student fees are essentially a form of revenue that reduces, but does not completely replace, subsidies from the state)?

I’m not in the hard sciences, but I think there’s more to the last question than most Americans give it credit for. Recall that academic research (almost always intimately tied to the government, sometimes directly so) is either the birthplace of or laid the groundwork for nuclear power, nearly every vaccine ever created, the internet, and so on. This is the subject of lots of research (see, for example, research by Mariana Mazzucato, who argues that state-funded academic research and investment has been the engine of American innovation and dynamism). We shouldn’t dismiss the possibility that this is the single largest value-add of higher education institutions in the country. In fact, I’m not even sure we should call them “higher education” institutions, since that name implies that most of their value is in the education rather than the research.

My non-scientific observations of higher education in the United States is that universities are service-providers with a pretty simple revenue model and provide three services. The revenue model includes, to vary degrees, student tuition and fees, state subsidies, investment income, and philanthropic donations. Universities then provide three services, more or less all public goods, all expected to grow the economy in some way:

1. Education: Universities educate students. Most of those students go on to be in the professional labor force, but some become full-time researchers and instructors (PhDs). This is the obvious one. Essentially the social value proposition is more aggregate human capital for the economy (bigger L—human capital—in the Cobb-Douglas sense).

2. Research: Universities produce vast amounts of research with large but hard-to-quantify social value. Some of that research directly spins out into start-ups and private ventures (probably a lot more frequently than many Americans realize). Sometimes, it lays the groundwork for future private ventures many decades later (e.g., mRNA research). Sometimes, it’s only social value (e.g., it’s important to understand how ranked-choice voting impacts democratic accountability and voter preference aggregation, but there’s no marketplace for that knowledge). The social value proposition here is higher aggregate productivity for the economy (higher TFP in the Cobb-Douglas sense).

3. Social Aggregation: Universities bring together interested, bright, ambitious people from different fields in one physical location. In my view, this is probably the least appreciated role of universities and the hardest to quantify. In private markets, we’re kind of sold on why this matters (for example, this is a sizable portion of the value proposition of startup incubators like Y Combinator). The “cross-pollination” phenomenon is also what explains the network effects that drive innovation hubs like Silicon Valley. There is some sorting and matching service here, similar to why people pay (implicitly or explicitly) for professional networking or dating applications. This is the “power of soft connections” that drives so much career progression and future innovation. There is also some indirect talent-acquisition service going on here (e.g., the visiting professor who has a firm and hires some of her best students). Then there’s the post-graduation network effects and signaling efficiencies whereby a university inducts graduates into a professional network that can recruit internally, knowing that some standard has been met by all members. The social value proposition here is also a productivity-enhancing one, mostly about efficient matching and innovation (higher TFP in the Cobb-Douglas sense).

My sense is that the degree to which each service matters varies considerably based on the school and discipline. The production of a doctor or lawyer is almost certainly a human capital improvement gain: someone didn’t know stuff to produce a thing, and now they do. For a post-doc biotech researcher, though, it’s mostly the research component that matters. In policy school, I think it’s probably mostly social aggregation. Find a mid-career highly successful military officer looking to make a pivot and introduce him to leaders in AI, and you’re reasonably likely to get a new entrepreneur in defense tech (I’ve seen this one before). I think that would probably happen regardless of academic courses, although those help. My half-baked take (impression, really) is that universities the world over are pretty good at the education portion, but that American universities are unusually good at the research and social aggregation functions, and that this explains some nontrivial amount of American exceptionalism in innovation.

Higher ed has a marketing problem

Americans seem less and less enthused with higher education. According to Pew Research, seven in ten Americans think higher education is going in the wrong direction. Lots of this appears to be about cost, which, fair enough. But some of it is deeper, which is of no surprise to anyone following the news the past few years. A lot of it is about campus free speech questions.

But these questions are, once again, centered mostly around questions of education, and it seems the media attention and debate writ large mostly set aside the huge social value of university research. For those who are serially bothered by the “bad news” bias (i.e., fear sells better than hope), I have good news for you from an unlikely place: university newsletters. Here is a small smattering of headlines from Yale’s daily newsletter from the month of May:

• “Yale study finds that exercise helps slow tumor growth in mice, and may do the same for humans”

• “New Yale research uses observational data to challenge existing theories about cold dark matter — and may prompt a fundamental rethinking of this critical, unseen cosmic substance.”

• “Tetracyclines are a versatile and popular class of antibiotics, but how they actually work has been poorly understood. Researchers in the lab of Yale’s Christopher Bunick have now discovered how they attack bacteria — findings that could lead to better, safer antibiotics.”

• “In a new study, Yale researchers discovered that the virus responsible for COVID-19 doesn’t linger in the placenta of women who become infected during pregnancy — a finding that offers important insight into the condition known as long COVID.”

• “A newly approved breast cancer drug — developed by New Haven-based biotech firm Arvinas and based on the pioneering research of founder and Yale chemist Craig Crews — will soon be available as a once-daily oral therapy.”

• “A cohort of 15 Ukrainian veterans recently visited campus [Jackson School] for an intensive leadership program designed to build on their battlefield experience — and help them chart their country’s future. Though designed for the Ukrainians, lead instructor Jimmy Hatch ’24, a U.S. Navy SEAL veteran, said the program ‘ended up teaching us.’”

• “How much is the war in Iran hitting Americans’ bottom line? Martha Gimbel, director of the Yale Budget Lab, weighs in”

That’s a lot of social goods for one month of headlines! I imagine that if you compiled these types of newsletters all throughout the country, you might have a different view of the value of universities. It’s a shame they don’t do a better job marketing this.

The great GPA debate rages on

One thing you hear a lot about on Ivy League campuses is lots of angst about GPAs: angst from students about theirs being too low, and angst among faculty and outsiders about them being too high (read: grade inflation). I want to offer some annoyances about the whole concept. (The following is pretty much limited to undergraduate education, since many if not most graduate schools don’t do GPAs anyways.)

Comparing GPAs from curved and uncurved classes is apples-to-oranges. This is a constant source of tension among students, I think reasonably. Some courses are curved, and some are not. Grades, then, are measuring two distinct things: in the curved context, they measure relative performance (and relative to whom is another issue related to selection bias and institutional requirements); in the uncurved context, they measure nominal performance. Being “90% right” then means two very different things—you’re either pretty good at knowing the stuff versus . . . actually, I couldn’t tell you what that means in the context of a curve, because it depends on the curve! Which leads me to . . .

Intra-university departmental variation is a problem. Some universities have different GPA averages by department. This is bad for two reasons. First, it makes GPA as a relative-ranking yardstick less useful: a GPA of 3.5 might be really good for engineering and really bad for political science. Second, it’s distortive: students take into account which courses, or even departments, have lower expected GPAs when choosing classes. Pre-law students are famous for this: they know that they need at least a 3.7 to get seriously looked at by the best law schools, so take the highest GPA courses they can find and use the extra time to prep for the LSAT, which is 100% rational given their incentives, but obviously not optimal in the aggregate.

Inter-university variation is also a problem. One complaint that you’ll hear from non-grade-inflation school students is that, since the Ivies are well-known for their tendency to grade inflate, non-Ivy students are doubly punished: they have a degree from a less prestigious brand and they face lower expected GPAs for the same courses. That is, again, distortionary and suboptimal. Give me a student with a 3.8 from Harvard and a student with a 3.8 from Michigan and I wouldn’t really be able to tell you how to compare them.

GPA does not adjust for difficulty-of-class. Even setting aside expected GPA or curved vs. uncurved, some courses are simply known to be harder than others, in some cases because they’re thorny topics, in other cases because the professor is known to assign a ton of work. Anyone who has been in school recently knows that there are “light-lift” classes, known for minimal work requirements and easy assignments. Student A could load up on these, while Student B chooses to take the most challenging courses on offer. Even if the curves are exactly the same and there’s no intra-university departmental variation in GPAs, the two GPAs wouldn’t really be comparable: Student A spent a lot of time tossing a frisbee and Student B was burning the midnight oil to achieve the same GPA.

I don’t mean to imply that grade inflation is justifiable because there are issues with GPA, nor do I have a brilliant scheme to resolve the issues. But I do mean to say that I would not index very highly to GPAs when assessing a student or job candidate.

The challenges ahead

Universities face two massive challenges in the coming years, both of which will reshape how universities carry out their core functions: AI and a trust deficit.

There are two big AI debates about higher education. The first is about whether we even need higher education (or education at all?) in an era of AI, or AGI. This is highly speculative, obviously. The second is, stipulating that we do still need education, how should university education change to serve the needs of the future. I don’t believe I am sufficiently knowledgeable about AI to talk about it, so I won’t really.

But I do strongly endorse reading this excellent essay by Nils Gilman on AI and the university:

The University As We Know It Is Finished

An excellent passage from the essay (my emphasis):

If the post-AI university’s pedagogic value proposition is the formation of cognitive capacity in conditions that cannot be replicated on a screen, then the function and responsibilities of faculty members must also be reconceived. It clearly no longer makes sense for professors to stand in front of a hall full (or, too often, only half full) of students delivering lectures. As a mechanism of information conveyance, AI can now provide the same at near-zero cost, tailor-made to the specific knowledge gaps of individual students. Instead, professors must reconceive of themselves as interlocutors, serving as performative models of how to calibrate uncertainty and revise frames in real time. The classroom experience should focus on helping students to understand how to constitute a goal rather than generate a text in response to a prompt provided by the professor.

And then there is the trust deficit. My own Yale University has made an honest and introspective attempt at beginning to right the ship, by publishing its 58-page Report of the Yale Committee on Trust in Higher Education. The Committee Report, the Gilman essay, and my observations here about the the roles of the university are in good conversation about the constituent roles of the university, and how we’ll have to think about parsing them.

From the Committee Report (my emphasis):

The range of topics revealed another challenge related to declining trust: widespread uncertainty about the fundamental purpose and mission of higher education. Trust is earned by doing what you say you’re going to do—and, ideally, doing it well. In recent years, however, universities have been expected to be all things to all people: selective but inclusive, affordable but luxurious, meritocratic but equitable. Rather than build public support, this diffusion of purpose has contributed to distrust. Without a clear mission and purpose, it becomes difficult to judge whether colleges and universities are living up to their fundamental commitments.

And from Gilman:

Longer term, however, we should expect the disruption caused by AI to be not just pedagogical but to the structure of the university as such. Kerr’s great insight was that the multiversity’s incoherence was not a bug but a feature—that a loosely bundled institution mirrored a loosely bundled society by providing something for everyone, from the Nobel laureate to the newbie grad student, from the NIH grant-seeker to the remedial English student. What held those disparate functions together was a social infrastructure of knowledge transmission: the laboratory, the lecture hall, the examination, the credential. Once AI can provide information delivery at near-zero cost there is no longer a compelling reason why research, teaching, and credentialing need be co-located in the same institution. What will replace the multiversity is likely to be not one thing but several: research centers that focus exclusively on the new-knowledge-production business; independent communal residence facilities that know they are in the coming-of-age business; and teaching systems that are honest about what skills they are inculcating. Even credentials from the most exclusive universities may not retain much social signaling value.

The throughline between the AI challenge and the trust deficit is that universities will need to make clear to the world—and more importantly, to themselves—what they are, and what their mission is. That will involve tackling questions of free speech, affordability, AI-tailored learning, research-production, political bias, and engagement with society writ large. It will not be an easy task, but fortunately there are some smart folks at the helm.

Concluding thoughts

In six years of school—eight of academia generally—I am leaving with fond feelings and a yearning to go back in some capacity. Perhaps what I’ve enjoyed the most about it is the essentialism of it all. I’ve avoided departmental politics mostly, so my experience, both as a researcher and a student, has been largely devoid of the administrative and bureaucratic nonsense that makes working life difficult. Simply learn the thing, publish the paper. No one cared when I worked—a paper written at 3am is the same as a paper written at noon. No one cared where I worked. There were virtually no meetings. The only thing was to understand, and to understand properly. I will forever cherish the days of professors in flip flops taking calls with senators, and teaching whatever and however they thought was best—they understood what actually mattered, and dispensed with the rest.

The challenges are real, though. From AI to concerns about equity and free speech to downright funding, universities appear increasingly in the crosshairs. But they provide a lot more than education to young people; they are engines of productivity and growth, centers for free expression and criticism, hugely important social aggregators, and perches from which to explore the greatest questions we face as a society. Don’t give up on them.

1. That does not, to be sure, mean that there isn’t value there. As discussed later, I think a lot of the value of school is not in the instruction at all, but rather in the connections, and that’s a very good reason to have successful connected people at a school, even if they aren’t great instructors.