The degree premium is fraying. Graduates are the first to feel it

THE PROMISE that a university degree is a ticket to a comfortable middle-class life is fraying faster than anyone expected. For five consecutive years, the unemployment rate for recent college graduates has exceeded the overall national rate-something almost unheard of between 1990 and 2018. The class of 2026 inherits not a temporary downturn but a structural mismatch: the labour market has changed while the advice to pile on credentials has not.

The numbers are stark. The unemployment rate for college graduates aged 22 to 27 reached 5.8% last year, the highest reading outside a pandemic since 2012, according to the Federal Reserve Bank of New York. Nearly half of recent graduates-42%-are underemployed, working in jobs that do not normally require a degree, the highest share since 2020, says Kelly Services, a staffing group. More than half of all graduates are in non-degree jobs. The trouble is not that graduates are idle. It is that they are idled in the wrong places.

The deeper question is not whether this generation is working hard enough. NACE, a group that tracks campus recruiting, found that the class of 2025 started job searches earlier, submitted more applications and accepted offers at a higher rate than the year before. The problem lies on the employer side. Hiring has slowed to levels last seen after the Great Recession, and the industries that traditionally absorbed degree holders have been shrinking for years. From 2023 to 2025, the information, finance and professional-services sectors-once the primary destinations for graduates-shed an average of 9,000 jobs per month. Before the pandemic, those same sectors added 44,000 per month. The on-ramp has been dismantled.

To be sure, much of the public anxiety about graduate employment is pinned on artificial intelligence. It is tempting to think that generative AI is replacing entry-level white-collar work at scale. The evidence does not yet support this. A study examining five different measurements of AI exposure found no meaningful impact on the labour market between 2022 and 2025, regardless of how the data were analysed. Employers cited AI as a reason for fewer than 55,000 of the 1.2 million job cuts announced in 2025-less than 5%. The freeze in hiring predates the current AI cycle. Software automation and the long tail of the Great Recession eroded traditional graduate jobs well before ChatGPT arrived.

The risk from AI is not absent. A February 2026 paper from the Federal Reserve Bank of Dallas found that AI is simultaneously reducing entry-level hiring and raising wages for experienced workers in the same occupations. College graduates work disproportionately in jobs where a large share of tasks could eventually be automated. The mechanism is not immediate displacement but a narrowing of the first rung: firms that can automate routine analytical work no longer need to hire and train as many juniors. The long-term effect may be to compress the career ladder, making it harder for new graduates to enter professions where experience is a prerequisite for advancement. That would be an unfortunate irony: a technology that should raise productivity ends up entrenching incumbents who own the skills that cannot yet be automated.

Even before the AI question, the path to a first job runs through an internship, and that path is narrowing. Competition for internships nearly doubled in a single year, with Handshake, a campus job platform, recording an average of 109 applications per posting in 2025, up from roughly 55 the year before. In tech the figure reached 273 per posting. The squeeze is a symptom of a broader hiring freeze. Firms face the same calculation in every sector: hiring young people is an investment whose payoff is uncertain when the outlook is murky. If you are unsure about the future, the first thing you stop doing is hiring the inexperienced.

The financial consequences for borrowers are beginning to catch up with them. Student loan debt totals $1.86 trillion, having resumed annual growth after a brief decline in 2023 and 2024, according to data compiled by Educationdata.org. Defaults have returned with force after the pandemic-era pause on repayments. Between April 2025 and March 2026, defaults surged by 4.2 million, leaving 9.5 million borrowers in default, reports EdSource. The New York Fed estimated that roughly 1 million borrowers defaulted in the fourth quarter of 2025, with an additional 2.6 million in the first quarter of 2026. The average borrower entering default is now nearly 40 years old, but the problem is not confined to older graduates. The share of all student loan balances past due has risen to just over 10%, nearing pre-pandemic levels. The on-ramp period, during which missed payments were not reported to credit bureaux, has ended. A second wave may be coming from the roughly 7 million borrowers who were on the now-defunct SAVE repayment plan and were pushed into forbearance by litigation.

All of this casts a harsher light on the advice to keep going for more education. Graduate-school marketing pitches rest on a genuine statistic: master's degree holders earn, on average, 20% more than those with only a bachelor's. But the enrolment growth that headlines celebrate is largely an illusion. Domestic graduate enrolment among the core age group-Americans aged 25 to 44 with a bachelor's as their highest degree-has grown by a mere 1% since 2013, once international students are stripped out, according to analysis by Eduventures. The graduate enrolment rate among this population has fallen by 30% in 13 years. Total graduate numbers look healthy only because the pool of degree-eligible Americans has grown by nearly 30% and international enrolment has surged. The signal is clear: Americans who have already sunk four years and tens of thousands of dollars into a degree are increasingly reluctant to invest further. They have been paying attention to the job market.

What should follow? The first task is for employers and policymakers to recognise that the entry-level labour market is broken in a way that a recovery in macroeconomic headline numbers will not automatically fix. The decline in white-collar hiring is structural, not cyclical. Firms have found ways to do more with fewer analysts, and they are unlikely to reverse course simply because GDP growth picks up. The result is a pipeline problem: if fewer juniors are hired today, there will be fewer mid-level managers tomorrow and fewer seniors the day after. The cure for scarcity is not to hoard existing talent but to invest in the next cohort.

A more serious answer lies in the education system itself. Universities have an incentive to enrol as many students as possible and graduate them as quickly as possible, because their revenue is front-loaded and tuition-driven. They have less incentive to ensure that degrees lead to jobs that justify the debt. The rise of non-degree certificates, bootcamps and microcredentials offers a partial correction, giving students shorter, cheaper routes to skills that employers actually want. But the main problem is not the absence of alternatives. It is the absence of accountability. If a university's graduates are routinely underemployed, that should matter to the institution's finances as much as to the borrower's credit score. Tying federal aid and institutional subsidies to employment outcomes, rather than simply to enrolment, would force a harder conversation about which programmes are worth the investment.

The broader lesson is not that college is worthless. The wage premium still exists. But a degree is no longer an automatic insurance policy against economic hardship. It is an investment whose return depends on the same variables as any other: the field of study, the institution's reputation, the macroeconomic environment and whether the borrower's expectations are calibrated to reality. The advice to "just get your degree" was always a simplification. It is now a dangerous one.

For the class of 2026 and those that follow, the choice is not whether to try harder. It is whether to invest more in a system whose returns are declining. The answer, for many, will be to demand better evidence before spending another four years. That is not cynicism. It is arithmetic.