One of the great promises of higher education is that it acts as a social ladder—one that allows students from low-income backgrounds to climb up and reach a higher social and economic status. No one, I think, ever believed it was a guaranteed social leveler, or that children from wealthier families didn’t have an easier time succeeding after college because of their own, and their family’s, social and cultural capital. But most people, in America at least, believed that on the whole it played a positive role in increasing social mobility.
Over the past couple of decades, though, particularly as student debt has increased, people have begun to wonder if this story about social mobility through college is actually true. That’s a hard question to answer definitively. Data sets that track both student origins and outcomes are few and far between, and it’s also difficult to work out what social mobility used to look like in a quantifiable sense.
However, this summer economist Sarah Quincy of Vanderbilt University and Zach Bleemer of Princeton University released a paper called Changes in the College Mobility Pipeline Since 1900. This paper overcame some of those data limitations and took a long, more than century-long, look at the relationship between social mobility and college attendance.
What they found was sobering. Not only is higher education no longer helping poor students catch up with wealthier ones, but in fact the sector’s role as a social elevator actually stopped working almost 80 years ago. This seemed like a perfect story for the podcast, and so we invited Zach Bleemer—who you may remember from an episode on race-conscious admissions about two years ago—to join us to discuss it.
This discussion ranges from the methodological to the expositional. Where does the data come from? What does the data really mean? And are there alternative explanations for the paper’s surprising findings? But enough from me—let’s hear from Zach.
The World of Higher Education Podcast
Episode 4.4 | The Widening Gap: Income, College, and Opportunity with Zachary Bleemer
Transcript
Alex Usher (AU): Zach, you wrote, with Sarah Quincy, a paper called Changes in the College Mobility Pipeline Since 1900, which looks a long way back. And you argue that the relative premium received by lower-income Americans from higher education has fallen by half since 1960. Take us through what you found—give us the 90-second elevator pitch.
Zachary Bleemer (ZB): Consider kids who were born in 1900 and were choosing whether or not to go to college in the late 1910s and early 1920s. What we were interested in was that choice, and in particular, following people for the next 20 years after they made it. Some people graduated high school but didn’t go to college, while others graduated high school and chose to go.
We wanted to compare the differences in early 1930s wages between those two groups—both for kids from lower-income backgrounds and kids from upper-income backgrounds. Now, you might be surprised to learn that there were lower-income kids going to college in the U.S. in the early 1920s, but there were. About 5 to 10% of people from the bottom parental income tercile even then were attending college.
What we found, when we linked together historical U.S. census records and followed kids forward, is that whether you were low-income or high-income, if you went to college your wages went up a lot. And the degree to which your wages went up was independent of whether you were low-income or high-income—everyone benefited similarly from going to college.
If you compare that to kids born in the 1980s, who were choosing to go to college in the late 1990s and early 2000s, you see a very different story. Everyone still gains from going to college, but kids from rich backgrounds gain a lot more—more than twice as much as kids from poor backgrounds. And that’s despite the fact they’re making the same choice. They’re going to different universities and studying different things, but when it comes down to the 18-year-old making a decision, those from poor families are just getting less from American higher education now than they did in the past—or compared to kids from rich backgrounds.
AU: I want to make sure I understand this, because it’s a crucial part of your argument. When you talk about relative premiums—premium compared to what, and relative compared to what?
ZB: What we always have in mind is the value of college for rich kids, and then asking: how much of that value do poor kids get too? In the early 20th century, and as late as the 1960s, those values were very similar. Lower-income kids were getting somewhere between 80 and 100% of the value of going to college as higher-income kids.
AU: And by “value,” you mean…
ZB: That just means how much your wages go up. So, the wage bump for lower-income kids was very similar to that of higher-income kids. Today, though, it’s more like half—or even a little less than half—of the economic value of college-going that lower-income kids receive compared to higher-income kids.
AU: So in effect, higher education is acting as an engine of greater inequality. That’s what you’re saying?
ZB: I guess it’s worth saying that lower-income kids who go to college are still getting ahead. But it’s not as much of a pipeline as it used to be. Higher education used to accelerate lower-income kids—not to the same level of income as their higher-income peers; they were never going to catch up—but at least they got the same bump, just from a lower starting point.
AU: So the gap widens now. But how do you make a claim like that over 120 years? I mean, I sometimes have a hard time getting data for just one year. How do you track college premiums across a period of 120 years? How sound is the empirical basis for this? You mentioned something about linking data to census records, which obviously go back quite a way. So tell us how you constructed the data for this.
ZB: The first-order answer is that I called up and worked with an economic historian who had much more experience with historical data than I did. Like you said, it’s hard in any period to get high-quality data that links students in high school—especially with information on their parental income—to wage outcomes 10 or 15 years later.
What we did was scan around for any academic or government group over the last 120 years that had conducted a retrospective or longitudinal survey—where you either follow kids for a while, or you find a bunch of 30-year-olds and ask them questions about their childhood. We combined all of these surveys into a comprehensive database.
In the early 20th century, that meant linking kids in the 1920 census, when they were still living with their parents, to the same kids in the 1940 census, when they were in their early thirties and working in the labor market. That link has been well established by economic historians and used in a large series of papers.
By the middle of the 20th century, sociologists were conducting very large-scale longitudinal surveys. The biggest of these was called Project Talent, put together by the American Institutes for Research in 1961. They randomly sampled over 400,000 American high school students, collected a ton of information, and then re-surveyed them between 1971 and 1974 to ask what had happened in their lives.
In more recent years, there’s been a large set of governmental surveys, primarily conducted by the Departments of Labor and Education. Some of these will be familiar to education researchers—like the National Longitudinal Survey of Youth (NLSY). Others are less well known, but there are lots of them. All we did was combine them all together.
AU: I noticed in one of the appendices you’ve got about nine or ten big surveys from across this period. I guess one methodological limitation is that they don’t all follow respondents for the same amount of time, and you’d also be limited to questions where the surveys provided relatively similar answers. You never get your dream data, but those would be the big limitations—you’ve got to look for the similarities, and that restricts you.
ZB: I’d add another restriction. You’re right that, as we filtered down which datasets we could use, the key variables we needed were: parental income when the student was in high school, level of education by age 30, and how much money they made at some point between ages 30 and 35. All of our surveys had those variables.
We also looked for information about what college they attended and what their college major was. Ideally, the surveys also included some kind of high school test—like the SAT or an IQ test—so we could see what kinds of students from what academic backgrounds were going to college.
But there was another key limitation. In most of the data before 1950, it was really difficult to get a direct measure of parental income. Instead, we usually had proxies like parental occupation, industry, or level of education—variables that are highly predictive of income, but not income itself.
So, a lot of the work of the paper was lining up these measures of varying quality from different surveys to make sure the results we report aren’t just noise from mismeasurement, but instead reflect real changes on the ground in American higher education.
AU: So you ran the data and noticed there was a sharp inflection point—or maybe not sharp, but certainly things started to get worse after 1960. When you first saw that, what were your hypotheses? At that point, you’ve got to start looking at whatever variables you can to explain it. What did you think the answer was, and what did you think the confounding variables might be?
ZB: My expectation was that two things would primarily explain the change. My background is in studying undergraduate admissions, so I thought the first explanation would be rising meritocracy in admissions. That might have made it harder for lower-income and lower-testing kids to get access to high-quality education. I also thought changes in affirmative action and in access to selective schools for kids from different backgrounds, along with rising tuition that made it harder for lower-income kids to afford those schools, could have played a big role. That was one possible story.
The second possible story is that it had nothing to do with the causal effect of college at all. Instead, maybe the poor kids who go to college today aren’t as academically strong as they were in the past. Perhaps in the past only the brilliant poor kids went to college, while all the rich kids went regardless of ability. So it could have looked like poor kids were getting a big benefit from college, when in fact those few who made it would have done well anyway.
It turns out neither of these explanations is the primary driver of rising regressivity. On the test score story, it’s always been the case that rich kids who go to college have relatively higher test scores than rich kids who just graduate high school—and that poor kids who go to college have relatively lower scores compared to their peers. That hasn’t changed since 1960.
And on the access story, it’s always been the case that rich kids dominate the schools we now think of as “good”—the fancy private universities and the flagship public universities. But over the last 50 years, poor kids have actually slightly increased their representation at those schools, not the other way around. Rising meritocracy hasn’t pushed poor kids out. If anything, the variety of admissions programs universities have implemented to boost enrollment among racial minority and lower-income students has relatively increased their numbers compared to 1950 or 1960.
AU: You were just making the case that this isn’t about compositional change in where poor students went. I heard you say there are more lower-income students at Harvard, Yale, and MIT than there were 50 or 60 years ago—and I have no doubt that’s true. But as a percentage of all poor students, surely that’s not true. The vast wave of lower-income students, often from minority backgrounds, are ending up in community colleges or non-flagship publics. Surely that has to be part of the story.
ZB: Yes. It turns out there are three primary trends that explain this rising collegiate regressivity, and you just hit on two of them.
The first is exactly your point: lower-income students primarily go to satellite public universities, basically all the non–R1 publics. Higher-income students, if they attend a public university, tend to go to the flagship, research-oriented universities.
I’ll skip talking about Harvard, Yale, and Princeton—almost no one goes to those schools, and they’re irrelevant to the overall landscape.
AU: Because they’re such a small piece of the pie, right?
ZB: Exactly. Fewer than 1% of students attend an Ivy Plus school. They don’t matter when we’re talking about American higher education as a whole. The flagships, though, matter a lot. About a third of all four-year college students go to a research-oriented flagship public university.
What’s happened since 1960 isn’t that poor kids lost access to those schools—it’s that they never really had access in the first place. Meanwhile, those schools have gotten much better over time. If you look at simple measures of university quality—student-to-faculty ratios, instructional expenditures per student, graduation rates—or even our own wage “value-added” measures (the degree to which each university boosts students’ wages), the gap between flagship and non-flagship publics has widened dramatically since the 1960s.
The flagships have pulled away. They’ve gotten more money—both from higher tuition and from huge federal subsidies, in part for research—and they’ve used that money to provide much more value to the students who attend. And those students tend to be higher income.
The second trend is what you mentioned: increasing diversion to community colleges. Interestingly, before 1980, community colleges were already well established in the U.S. and enrolled only slightly more lower-income than higher-income students. They actually enrolled a lot of high-income students, and the gap was small. Since the 1980s, though, that gap has grown substantially. There’s been a huge diversion of lower-income students toward community colleges—and those schools just provide lower-value education to the students who enroll.
AU: At some level this is a sorting story, right? You see that in discussions about American economic geography—that people sort themselves into certain areas. Is that what you’re saying is happening here too?
ZB: It’s not about sorting inside the four-year sector. It’s about sorting between the two- and four-year sectors. And on top of that, we think there’s fundamentally a story about American state governments choosing to invest much more heavily in their flagship publics—turning them into gem schools, amazing schools—while leaving the other universities in their states behind. Those flagships enroll far more higher-income than lower-income students.
AU: When I was reading this paper, one thing that struck me was how hard it is to read about American higher education without also reading something about race. The last time you were on, we were talking about SCOTUS and the Fair Harvard decision. But as far as I can tell, this paper doesn’t talk about race. I assume that goes back to our earlier discussion about data limitations—that race just wasn’t captured at some point. What’s the story there?
ZB: No—we observe race throughout this entire period. In fact, you could basically rewrite our study and ask: how has the relative value of college for white kids compared to Black kids changed over the last hundred years? I suspect you’d see very similar patterns.
The datasets we’re working with observe both parental income and race, but they aren’t large enough to separately analyze, for example, just white students and then compare lower- and higher-income groups over time. There’s a sense in which you could tell our story in terms of race, or you could tell it in terms of class—and both would be right. At a first-order level, both are happening. And within racial groups, the evidence we’ve been able to collect suggests that class gaps have substantially widened over time.
Similarly, we show some evidence that even within the lower-income group there are substantial gaps between white and Black students. So in part, I saw this as an interesting complement to the work I’d already done on race. It points out that while race is part of the story, you can also reframe the entire conversation in terms of America’s higher education system leaving lower-income students behind—irrespective of race.
AU: Right, because it strikes me that 1960 is only six years after Brown v. Board of Education. By the early to mid-1960s, you’d start to see a bigger push of Black students entering higher education, becoming a larger share of the lower-income sector. And a few years later, the same thing with Latino students.
Suddenly lower-income students are not only starting from further behind, but also increasingly made up of groups who, irrespective of education, face discrimination in the labor market. Wouldn’t that pull things down? Wouldn’t that be part of the explanation?
ZB: Keep in mind that when we measure wage premiums, we’re always comparing people who went to college with people who only finished high school. So there are Black students on both sides of that comparison, across both lower- and higher-income groups.
That said, I think your point is well taken. We don’t do any work in the paper specifically looking at changes in the racial composition of students by parental income over this period. One thing we do show is that the test scores of lower-income students who go to college aren’t falling over time. But you’re probably right: while racial discrimination affects both college-goers and non-college-goers, it’s entirely plausible that part of what we’re picking up here is the changing racial dynamics in college-going.
AU: What’s the range of policy solutions we can imagine here, other than, you know, taking money away from rich publics and giving it to community colleges? That’s the obvious one to me, but maybe there are others.
ZB: And not just community colleges—satellite publics as well. I’ve spent the last five years of my life thinking about how to get more disadvantaged students into highly selective universities, and what happens when they get there. The main takeaway from that research is that it’s really hard to get lower-income students into highly selective universities. It’s also expensive, because of the financial aid required.
But once they get into those schools, they tend not only to benefit in terms of long-run wage outcomes, they actually derive disproportionate value. Highly selective schools are more valuable for lower-income kids than for the higher-income kids who typically enroll there.
What I’ve learned from this project, though, is that the closing of higher education’s mobility pipeline isn’t fundamentally about access. It’s about investments—by state governments, by students, by donors, by all the people and organizations that fund higher education. Over time, that funding has become increasingly centralized in schools that enroll a lot of wealthy students.
So, the point you brought up—redirecting funds—is important. In California they call it “rebenching”: siphoning money away from high-funded schools and pushing it toward low-funded schools. There’s very little academic research on what happens when you do that, but our study suggests that this century-long trend of unequal investment has disadvantaged low-income students. Potentially moving in the other direction could make a real difference for them.
AU: Zach, thanks so much for being with us today.
ZB: My pleasure.
AU: It just remains for me to thank our excellent producers, Tiffany MacLennan and Sam Pufek, and you, our listeners and readers, for joining us. If you have any questions or comments about today’s podcast, or suggestions for future editions, don’t hesitate to get in touch at [email protected].
Join us next week when our guest will be Dmitry Dubrovsky, a research scholar and lecturer at Charles University in Prague. He’ll be talking to us about the slow-motion collapse of Russian higher education under Vladimir Putin. Bye for now.
*This podcast transcript was generated using an AI transcription service with limited editing. Please forgive any errors made through this service. Please note, the views and opinions expressed in each episode are those of the individual contributors, and do not necessarily reflect those of the podcast host and team, or our sponsors.
