Day 4: How much is a college degree worth?

A series following my new class: The Economics of Higher Education

Apr 30, 2025

On Day 3 we talked about a set of readings that offered three reasons why college costs so much.

Colleges employ a lot of people with college degrees, and college graduates get paid more than high school graduates. Therefore, college costs a lot.
In order to teach complex things, colleges need complex equipment, and that costs a lot of money.
Students expect an increasing variety of services from their colleges (e.g. career centers, counseling centers, etc.) and that all costs money.

The authors of those readings really think the reason college is expensive is because employing college graduates is expensive. But how much more do college graduates get paid exactly? That is the topic of conversation today!

Updated to add: a friend wrote to say that I really need to put some different papers on my syllabus for this day next year. And I think that’s great! That’s precisely why I’m writing this Substack, so I can get you all to do some of my work for me. My friend suggests you all should read:

Updated again to add: this newly updated working paper by Zachary Bleemer and Sarah Quincy title “Changes in the College Mobility Pipeline Since 1900” looks very interesting and very relevant to this conversation. Long story short: they find that the college wage premium has been increasing less for low-income students than for high-income students for three reasons: 1) public universities have been declining in funding, retention, and economic value; 2) low-income students have been diverted to community colleges and for-profit colleges; 3) higher-income students are majoring less in humanities and more in computer science which has increased their returns to college.

On to what I did chose this year!

Today’s readings

Ma, J. & Pender, M. (2023) Education pays 2023. College Board. (Highlights and browse the charts)

Let’s start with how most people think of the value of a college degree: the simple difference between how much the average college graduate earns and how much the average high school graduate earns.

People who complete a bachelor’s degree earn $29,000 (in 2021 dollars) more than people who only have a high school degree.

This is what I think most people believe about the power of a college education: it should get you ~$25-30k extra per year.

However, the earnings premium varies across groups (I made this little table; oddly the report didn’t show these numbers clearly)…

… and of course earnings vary a lot by college major and experience in career.

But still, college graduates make a lot more money than high school graduates. (Right?) No matter their major. (Right?) And they will forever. (Right?)

Before we get too far ahead of ourselves, are we even sure that the difference in earnings between college and high school grads is actually because of college?

Webber, D. A. (2016). Are college costs worth it? How ability, major, and debt affect the returns to schooling. Economics of Education Review, 53, 296-310.

The title pretty well explains this paper. It’s a good one.

Big ideas to highlight

Graduation rates, risk, and expected value
This author makes a big point of the fact that not everyone who starts college finishes college. This is a huge and really good point. Most news reports and politicians (and that College Board report we just reviewed) highlight the simple difference between the earnings of college grads and non-college grads. This makes it seem like getting that college wage premium is a sure thing. But actually only 60% of people who start college finish it. So if you’re a high school senior and you’re trying to predict how much of wage bump you will get, you should count on only a 60% chance of whatever earnings premium you hear about from your guidance counselor.
Selection due to ability
Another fact that most guidance counselors, news reports, and politicians gloss over is that the people who graduate from college are different from the people who don’t, and people who graduate from college probably would earn higher wages even if they hadn’t finished their degree. Call it ability, call it the privilege of socioeconomic status, call it knowing how to win at the grading game, call it whatever you want, but we know there’s selection into completing college. There’s also selection into different majors. If we don’t account for the fact that STEM degree holders would probably make more money than Humanities majors even if they had chosen different degrees, than we will overestimate the boost from a STEM degree. This paper is quite careful to account for selection issues, both into college at all and between types of degrees.

Methods

First the author does regressions of log earnings on demographic characteristics and education categories, excluding ability measures.
Then he runs those same regressions, but with a measure of ability (measured primarily by the Armed Forces Qualification Test captured in the National Longitudinal Survey of Youth). The difference between the regression coefficients in model 1 versus model 2 allows for an estimate of the effect of ability on earnings.
The author then uses some more advanced methods (ordered logits and multinomial logits) to estimate and then correct for the selection effects of ability into a) completing college and b) major choice.
The author puts all of these estimates together in a simulation model (this is mostly just a fancy way of saying “I added up which coefficients applied to which groups,” with some noise added to the simulations to account for natural variation and luck) to report and plot the returns to education after correcting for risk, selection, and college costs.

Results

Table 2 shows total lifetime earnings for (col 1) high school graduates, (col 2) people who have completed some college, (col 3) people who completed a college degree in STEM, (col 4) business, (col 5) social sciences, or (col 6) arts and humanities.

Row 1 shows results that don’t account for selection based on ability into going to college or major choice, but it does account for the fact that 40% of people who start college don’t graduate. Row 2 does account for selection based on ability.

This table shows that if you assume high school graduates and STEM majors have the same underlying abilities and getting a STEM degree is purely random, then getting a STEM degree is worth $841,437 in additional lifetime earnings. (All of these calculations are in 2014 dollars, so if you want to account for inflation and put it into 2025 terms, multiply by 1.3). However, when you account for the fact that people who get STEM degrees are different from people who only complete high school, and people who get STEM degrees would earn more with or without the degree, that estimate shrinks to $645,026.

That’s assuming that college is free. If you take into account typical college costs, the STEM bonus is $626,199. If you assume high college costs, the STEM bonus is $425,590. If you were a family that paid full price for a Colorado College education, the lifetime earnings STEM bonus is $355,031. (That assumes that a CC education is no more valuable than an education anywhere else, which, hey, I like to think we do better than average. You make your own assumptions.)

You can see in the table that getting a degree in Arts and Humanities doesn’t payoff as well. At a high cost college (but lower cost than full pay at CC), an Arts and Humanities degree is worth only $16,095 over the course of a lifetime.

Table 2 does calculations for a student in the smack dab middle of the ability distribution. But of course this is not Lake Wobegon, we’re not all above average, and we’re not all at the median either. Table 3 shows the simulated returns to various degrees with various costs, but for the student whose cognitive skills are better than only 25% of their cohort. For that student, a STEM degree at a high cost college is worth $215,170. Over a 40 year career, that’s an extra $5,379 a year. Recall that College Board report promised more like $25-30k extra per year.

I’m starting to understand better why the vibes are so down on college these days. Table 4 really brought it home for me. Table 4 shows the age at which the investment in college pays off, or when the increase in earnings overtakes the amount it cost to get the degree. For a middle ability student who paid average college expenses, that age is 31 for a STEM degree and 41 for an Arts and Humanities degree. For the middle ability student who went to a high cost college, those ages are 42 and 64. For the low ability student who went to a high cost college, a STEM degree doesn’t pay for itself until age 45 and an Arts and Humanities degree never does.

All of these calculations are on average, essentially assuming that all students in the same ability-degree-major-expense bucket will have exactly the same outcome. When the author accounts for natural variation and luck, he finds that a STEM degree at a high cost college for a middle ability student is a net positive investment 73% of the time. In Arts and Humanities, it’s a positive investment only 50% of the time.

Phew, that does seem dire. The author is good to point out that these calculations only account for the financial returns to college, and there are plenty of non-monetary benefits too. Like, for example, people with college degrees tend to like their jobs better (that College Board report talks about this). And people with college degrees tend to marry people who also have college degrees (though this trend seems to be waning), and tend to stay married more often (probably in part because they are wealthier), which is all good for household finances, kids, and life satisfaction.

Even still, it’s notable to me how different these numbers are from what you would think just listening to the zeitgeist. I think young people believe going to college should fundamentally change their life. Like, at least earn them an additional $25k a year. Instead, the bump is at best $16k annually for a typical STEM student, it’s often only a couple thousand bucks a year, and sometimes the return is zero or even negative.

No wonder there are so many cranky people available to talk to NPR.

Bengali, L., Valletta, R. G., & Zhao, C. (2025). Explaining Stagnation in the College Wage Premium (No. 2025). Federal Reserve Bank of San Francisco Working Paper. (Abstract, Introduction, and Figures)

Next question: has the college earnings premium been changing? And if so, why? This is a brand new, and pretty technical, working paper from economists the San Francisco Fed. I think the ideas are really interesting. Let’s go.

First these authors document that the college earnings premium has been changing over time. It increased a lot from 1980 until about 2000, and since then it’s held steady and maybe even started declining. They’re wondering what’s behind that stagnation and maybe decline.

In general economists think that the college earnings premium is determined by the interaction between supply and demand (duh). Supply meaning: how many college educated workers there are. Demand meaning: how many college educated workers firms want to hire.

This figure shows that in 1960 there were about 4 high school educated workers for every 1 college educated worker. In 2020, it was basically a 1:1 ratio.

That is a huge increase in the supply of college educated workers. Usually we would think that a huge increase in supply would result in lower prices, a.k.a. a smaller college wage premium. But instead we saw an increase in the college wage premium, at least until 2000. So what else must have been going on?

An increase in the demand for college educated workers, of course. Economists think this is due to skill-based technological change. (We talked about this on both Day 2 and Day 3.) Basically the idea is that as technology changes you need different kinds of workers, and the kinds of technology changes we’ve seen over the late 20th century led firms to want more college educated workers.

These authors want to plot how much of the increase and now stagnation and maybe decrease in the college wage premium is due to demand versus supply factors. They put together a complex model that allows for workers of different types (i.e. high school and college grad), a substitutability between those different types of workers that varies over time, and technology change that can affect the demand for different types of workers non-linearly and differently over time. Then they estimated the parameters of that model to best fit the aggregate data. It’s more complicated than an OLS regression, but the idea is the same.

I’ll stop right here to admit that I’m not a macroeconomist and I don’t totally understand how they mapped their regression results to their conclusions. But I recently hosted the Director of Research of the SF Fed as a guest speaker for my Principles of Macroeconomics class and he tells me he hired good people, so I’m going to take these authors’ conclusions at face value.

The money chart is below. This chart shows that from 1960 until about 2000, there was strong demand for college educated workers, which pushed the college earnings premium up (see the the red bars above zero). There were drags on the premium due to increasing college going rates/the supply of college educated workers (those blue bars below zero), but after the late 1970s they weren’t strong enough to pull the wage premium down overall. That changed around 2000. Around then the supply and demand factors evened out, and in very recent years, the demand for college educated workers has been falling and thus dragging down the college wage premium.

Why? These authors say it’s because of three things.

A slow down in the kind of technology advancement that requires college educated workers.
Firms are finding that they can more easily substitute between high school and college educated workers.
It doesn’t help that more people keep going to college, pushing up supply at the same time that demand is falling.

In class we had a great discussion about why the students think the elasticity of substitution between college and non-college workers may be increasing. They think it could be due to several factors. First, selection into college isn’t what is used to be (everybody goes to college now) so it’s not as strong of a signal; you have to go to grad school to stand out now. Second, the quality of education that students produce is not as high because you can AI your way through college and/or rely on grade inflation or pass/fail grades to increase your GPA. Third, the quality of education that professors produce may not be as high because professors aren’t as good as they used to be or could be now (yes, I encourage very authentic conversation in my classroom). Finally, jobs may not need as specific of skills anymore so it’s easier to substitute a college educated worker with a high school educated worker.

There were several students who did not buy that the elasticity of substitution between college and high school educated workers is increasing. In their experience applying for jobs and helping to hire people into firms, resumes without a college degree go straight in the trash can. This led to some super interesting conversation about which industries still absolutely require a college degree (finance, law, medicine), which don’t (tech), and why (liability to lawsuits, ease of firing workers, i.e. the risks associated with hiring a bad fit worker).

I think all of this is really interesting. And worrying. There’s now a full 41% of the adult American population that has a college degree, and they think it should earn them a substantial wage bump. Especially if they got a STEM degree, which an increasing fraction of them did. But the college wage premium is already smaller than people think it should be, at least after accounting for selection and the cost of college. And the last paper we read today suggests that the college wage premium is falling, in part due to changes in technology that may hit STEM majors hardest. And all of that is before GenAI has had much impact on the economy.

Rut. Roh.

This last paper also reminded me of some of the ideas in the Goldin and Katz paper from Day 2. Remember how that paper said that the reason why universities got so big at the beginning of the 20th century is because technology advanced in such a way that highly specifically trained workers became more valuable, so universities had to offer more specialized training? This last paper seems to suggest that trend is over. Now technology may be changing in such a way that doesn’t advantage highly specialized training.

What will that do to the market for institutes of higher education? Are we about to see a resurgence of interest in a general, broad, and dare I say liberal arts education?

Tomorrow’s readings

Bryan Caplan’s The Case Against Education: Why the Education System is a Waste of Time and Money

Introduction
Chapter 1: The Magic of Education
Chapter 2: The Puzzle is Real

Today we talked about what kind of a wage bump college graduates can expect in the labor market, or maybe that they can’t.

Next, for Friday of Week 1, we take on The Case Against Education.

The Liberal Arts Ledger

Discussion about this post

Ready for more?