The Degree Still Pays. The First Job Is the Problem.

Irene Chang graduated from Georgia Tech in May with a degree in industrial and systems engineering. By the time NPR spoke to her in August, she had sent about 450 applications since the previous September, landed about 19 interviews and received no offers. She was twenty-one, held a credential that admissions officers dine out on, and was doing everything the standard path asks of a person. The path, at that moment, was not asking much back.

Her story is the kind that sells the argument this publication is sometimes accused of making: that the degree is obsolete, that the smart move is to trade, build, sell or post instead. It is worth saying early and plainly that the data do not support the strong version of that claim. They support something narrower and, for a nineteen-year-old with a decision to make, more useful. The degree is not obsolete. The assumption that it converts automatically into a first job is.

The premium is large, and it is not going away

Start with the number that the skeptics must reckon with. In 2025, according to the Bureau of Labor Statistics, workers with a bachelor’s degree had median usual weekly earnings of $1,578 and an unemployment rate of 2.8 percent. Workers with only a high-school diploma earned $966 a week and were unemployed at 4.3 percent. That is a gap of $612 a week, roughly $32,000 a year if you multiply by fifty-two, and it is a gap across all ages, not only the young. An associate’s degree sat in between, at $1,135.

Georgetown’s Center on Education and the Workforce, which has done more than anyone to put dollar figures on the question, ranked 4,600 colleges and universities by return on investment in its February 2025 update, using College Scorecard earnings data. The pattern is instructive. Institutions that mostly grant associate’s degrees and certificates pay back faster, offering a higher return at ten years. Then the relationship reverses, and by forty years, in Higher Ed Dive’s phrase, “almost all bachelor’s degree-awarding colleges” come out on top. The center holds earnings flat after the tenth year, an assumption it calls conservative, so the long-run figures are by its own account cautious.

This is the steelman, and it deserves to be stated without squirming. The degree is a bet with a long payoff and a modest failure rate in aggregate. Anyone who tells an eighteen-year-old to ignore it should have a better argument than the existence of a famous dropout.

What the premium does not say

An average is a statement about a population, not a promise to a person, and the Georgetown work is candid about this. Its own page says ROI “varies from person to person, even among graduates of the same program and institution.” Several of its design choices flatter the headline: the data cover only students who received federal financial aid, about 55 percent of students by the center’s account, and debt is not part of the calculation. The center does count earnings for everyone who enrolled, non-completers included, which is a point in its favour, but it means low graduation rates drag the figure down rather than disappear from it.

The spread across institutions is wide. Eight of the ten best-returning institutions are nursing and health-care schools; the other two are MIT and Caltech. The bottom of the ranking, Higher Ed Dive reports, includes many arts institutions and for-profit schools that award bachelor’s degrees. Jeff Strohl, the center’s director, called the institutional figure “the start of the conversation.” The degree is not one asset. It is a category, and the category contains a nursing license and a very expensive art school.

Then there are the people who never collect the premium at all. The National Student Clearinghouse counts 43.1 million Americans with some college and no credential, 37.6 million of them under sixty-five. They paid in, in time if not always in money, and drew out nothing an employer can read on a résumé. Any honest accounting of what college is worth has to put them in the denominator, and they are why the median entering student is a more modest figure than the median graduate.

Americans, for their part, have noticed. In a 2024 Pew survey, “Only 22% say the cost of getting a four-year college degree today is worth it even if someone has to take out loans.” Forty-seven percent say it is worth it only if no loans are needed, and 29 percent say it is not worth it. Even among four-year graduates, only about a third say it is worth the cost with loans. These are views about price and risk, not about whether the degree pays, and the distinction matters. Pew also found that about one in four adults says a four-year degree is extremely or very important for a well-paying job, while 54 percent of workers aged 25 to 34 did not have one in 2023. The country already lives, to a large degree, in the world the degree skeptics describe.

The squeeze at the door

The sharper problem is not the long-run premium. It is the first rung. The New York Fed’s tracking of recent graduates, those who have just left school, put their unemployment rate at about 5.6 percent and their underemployment rate at 42 percent in the second quarter of 2026, and describes conditions as “challenging.” Underemployment, which the Bank derives from the kinds of jobs graduates hold, is the quieter half of the problem. For comparison, NPR reports that the recent-graduate rate of 5.7 percent in June compares with 4.1 percent for all workers. A degree used to buy a better-than-average chance of work, and for the newest graduates it currently does not.

The reason is contested, and the contest is the most interesting part of the story. The explanation that suits the narrative of this publication, and of a good deal of Silicon Valley, is artificial intelligence. There is real evidence behind it. A Stanford team led by Erik Brynjolfsson, using payroll data from ADP, found that young workers in AI-exposed occupations fared worse than their peers elsewhere, though the authors are careful to call their results “early, descriptive indicators—canaries in the coal mine—rather than causal estimates.” Brynjolfsson told NPR that “AI is not the whole story, but it’s part of the story and the evidence is building.”

The New York Fed’s own economists disagree about the proportions. In a June 2026 post, Natalia Emanuel, Emma Harrington and Amanda Pallais argued that the timing of the surge “suggests that remote work—not generative AI—explains the bulk of the rise in youth unemployment.” They estimate remote work accounts for 64 percent of the increase in unemployment among young college graduates between 2017-19 and 2022-24, a figure they themselves describe as a back-of-the-envelope calculation. Their logic is that managers are reluctant to hire people who must be trained at a distance. Harvard’s David Deming, quoted by NPR, makes the same point from a different angle: junior hiring began falling about six months before ChatGPT was released, and so “I think it’s more like remote work.” Anders Humlum of the University of Chicago points to firm-level data showing that heavy users of AI tools are hiring more than anyone else.

Notice what both camps share. Neither says the degree has stopped working. The Fed authors find that unemployment for more experienced college graduates slipped from 1.9 percent to 1.8 percent over the same period in which it rose for the young. The shock is concentrated at the entry point. It is a queue problem, not a value problem. The people at the back of the queue are real, and so is the cost of waiting, but a graduate who clears the queue is still on the right side of a large premium. Whether AI will widen the queue is the open question; on the evidence so far, the honest answer is that nobody knows, and the people who say they do are selling either a course or a fear.

The trades, and who is actually leaving

If the college path has a worse first year than it used to, the natural question is where the people are going. The cleanest recent data are from the National Student Clearinghouse. Total postsecondary enrollment in fall 2025 rose 1.0 percent to more than 19.4 million. Growth was driven by a 3.0 percent rise at community colleges, while private four-year institutions declined. Undergraduate certificate and associate enrollment grew 1.9 and 2.2 percent, against 0.9 percent for bachelor’s programs, and community colleges now enrol 752,000 people in undergraduate certificates, up 28.3 percent since fall 2021. Enrollment in computer and information science fell across every award type and institution, by between 3.6 and 14.0 percent depending on the category.

That is a real shift, and it is a quiet one. It is not an exodus. Reporting on spring 2025 Clearinghouse data puts enrollment at trade-focused schools near 871,000, up almost 20 percent since spring 2020, but the same report relays a Quartz analysis arguing that much of this reflects Gen Z aging into eligibility rather than a genuine turn away from college, that total trade enrollment across all ages grew at only 0.4 percent a year from 2017 to 2023, and that growth was concentrated in culinary arts and health care rather than construction and electrical work. We have not independently reviewed the Quartz analysis and attribute it as the IBTimes does. The caution is nonetheless a fair one. The stories about a generation of plumbers are, so far, mostly stories.

Which brings us to the people this publication writes about. The teenage founders, the traders, the creators who skipped or left school are not evidence about the median student. They are evidence about a tail. They are selected: for ability, for appetite for risk, for a safety net that lets them fail, and sometimes for a credential they quietly hold, such as admission to the university they declined to finish. Observing that a handful of such people did well without the degree tells you very little about what would have happened to the average eighteen-year-old who attempted the same thing, any more than a lottery winner’s retirement is a retirement plan. The surviving founder is visible; the ones who tried the same route and returned to enrol are not interviewed.

None of this makes the builders irrelevant. It makes them informative in a particular, limited way. They show what can be done without permission at an age when the degree path would have had them in a lecture hall, and they show that the price of trying has collapsed. A teenager can now ship software, run a storefront or publish to an audience for close to nothing. The relevant question for the average student is not whether to be a builder instead of a graduate. It is whether to build as well, and whether the four years of structure are worth the price.

What obsolete means

The word in this publication’s title can be read three ways, and only one of them survives the data.

It does not mean that college no longer pays. The BLS, Georgetown and the Fed’s own figures for experienced graduates all say it does, by a wide margin, on average, over a lifetime. It does not mean that everyone should go: the Pew respondents who balk at the loans, and the 37.6 million adults who stopped out, are making a rational response to a product whose price is certain and whose payoff is probabilistic. And it does not mean that the trades are the new default, since the numbers that suggest so are smaller and stranger than the headlines.

What is obsolete is the bundle. For two generations the degree was sold as four things at once: skill, signal, network and a guaranteed first job. The skill and the network still travel with it. The signal is intact for those who finish and are lucky in their field. The guaranteed first job, the one piece a family could count on, is the part now failing, and failing for the cohort least able to wait for it. On the aggregate data, a graduate like Chang is still on the right side of a large premium. She is also, today, without a job.

The honest advice to a young person is therefore less cinematic than either side would like. Treat the degree as an expensive option whose value depends on field, finish and price, not as a verdict on your character in either direction. Do not drop out because a founder did. Do not enrol because a parent did. And whichever door you take, build something before you reach the first-job queue, because the shared finding of the evidence above is that the squeeze is at the entry point. The degree is not obsolete. The waiting is.