It’s orientation time (in my house at least), which translates into chaos and lots of takeout so let’s jump right in. In this week’s issue, I look at some competing theories on why the unemployment rate for recent graduates is so high. Spoiler alert: we’re not 100% sure or even in agreement.  

What’s Causing High Unemployment for Recent Grads? 

From Many recent grads say AI is making it harder to get a job. Economists aren't so sure | NPR 

NPR looks at competing economic arguments for why recent grads are having a harder time finding a job.  

Our Thoughts  

Before we even dive into some of the arguments of this piece, I want to make one thing clear. According to data from the Federal Reserve Bank of New York, the unemployment rate for recent grads (22-27 years old) was 5.7% in June. This is higher than the overall unemployment rate of 4.1%. Now, for the fair warning. I’m not an economist by training, so I’m doing my best to untangle some of the studies reported in this NPR piece because they are important. Whether or not AI is impacting the labor market for entry level roles is certainly a question all of the newly arriving students have on their minds as they hope to find a job upon graduation. So, let’s jump in and see what the data says.  

NPR talked to three economists for this piece. They all agree that the job market for new graduates is bad. However, they have differing opinions and research-supported ideas about why this is the case. First up is Erik Brynjolfsson at Stanford who makes the case that AI is at least part of the story. He supports this conclusion with payroll data showing a 19% relative employment decline since late 2022 among workers aged 22 to 25 in AI-exposed occupations like software development, while older workers in those occupations held steady or grew. Next is David Deming from Harvard who says that timing doesn’t work because, by his calculation, junior hiring began falling roughly six months before ChatGPT was released. His likely candidate? Remote work. A New York Fed analysis estimates that remote work explains 64% of the increase in unemployment among young college graduates between the 2017 to 2019 period and the 2022 to 2024 period. Finally, Anders Humlum at Chicago doesn’t say what the cause is, but argues that it’s not AI. A study from Ramp and Revelio Labs found that the entry-level headcount was up 12% at the companies with the highest levels of AI adoption, meaning that if companies were really substituting AI for junior workers, you would probably expect them to have fewer entry-level hires.  

So, who’s right? Honestly, no one knows quite yet. Each of the papers has challenges that make it unlikely to be the full explanation of what’s happening (the economy is complicated, right?). Brynjolfsson and colleagues are the most careful of the three, but even they admit that once they controlled for the education level of an occupation, their estimate gets cut in half. They also report that their ADP payroll data shows an effect size about six times larger than Census data covering the same years. Meanwhile, Deming’s comparison window closed before the return-to-office pushes really took hold, meaning if remote work is the cause, but it’s receding, the recent graduate unemployment shouldn’t keep climbing. Finally, Humlum’s comment is based on the Ramp and Revelio Labs study, which found the growth at AI heavy firms. The problem—these firms had an average entry-level headcount of 9.9 people, meaning a 12% increase is...one person. The employment gains noted in the report also only reach statistical significance in one sector—Information, which covers software, internet, and media. The one area where you might expect to see explosive hiring growth for new AI companies. However, I’m not sure if it truly matters which one is the likely culprit because for us, settling it wouldn’t change much. 

Look at what Brynjolfsson and the New York Fed researchers are actually describing because underneath their disagreement about causes, they land on something frustratingly similar. Brynjolfsson’s argument is that AI substitutes for the kind of knowledge that gets written down and taught, while complementing the judgement that only comes from experience. This is why the people AI hasn’t touched are the ones who have been doing the work for a decade. The New York Fed’s argument is that employers don’t want to hire fresh graduates onto distributed teams because teaching them what they need to know from a distance is harder. Both of those are descriptions of an employer looking at a new graduate, calculating what it will cost to turn that person into a useful employee, and then deciding the investment isn’t worth making. One story says the alternative got cheaper; the other says the conversion got more expensive. Either way, the recent graduate gets the same rejection letter.  

So, we’re back to the thing that Brynjolfsson says AI complements rather than replaces—experience. That sounds obvious until you ask where experience comes from. It comes from entry-level jobs. A new graduate shows up with information about their new field and spends two to three years watching how the work actually gets done, transforming their textbook information into field-specific knowledge that an AI model can’t replicate. Yes, that process is expensive, and someone more senior pays in time and patience, but that’s how companies transform a new employee into a mid-career professional. I suddenly feel a strong sense of déjà vu. Probably because I raised a version of this in Issue 102, putting AI squarely in the driver’s seat as the cause. Several months and three studies later, I’m less certain about the cause, but considerably more certain about where it ends up.  

So, what do we do with this on campus? First, no matter which economist turns out to be right, the message is the same. Experiences which build judgement have become more valuable. Co-ops, clinical placements, student teaching, undergraduate research, practicums and internships are still high impact practices. They’re just absorbing the work that employers used to pay for themselves. Even with that said, institutions cannot absorb that training function. An internship is no substitute for two years working 8-5, Monday-Friday. But what we can do is build more employer partnerships to help ensure that any student who wants that opportunity can find it. Oh, and if someone hands you the Ramp finding as evidence that the worry about AI is overblown or the Stanford paper as evidence that the sky is falling, you can now help them dig a little deeper and see that no one has the answers...yet.  

Sparks 

• What New Research Reveals About Student Success (Inside Higher Ed) -  A look at recent research about mental health, financial vulnerability and career literacy and how they impact student success. After enrollment, student success is probably the second most talked about thing on campuses, so having four studies summarized in one spot? Sign me up.
• Education Department proposes accreditation overhaul (Higher Ed Dive) - The Department of Education has released a proposed rule to change the accreditation system. I know accreditation seems abstract (and perhaps a bit boring), but these changes could have lasting impacts for higher education.
• Degree wage premium in decline since 2010, finds study (Times Higher Education) - A study out of Australia finds that the college degree wage premium has been in decline since 2010. People have started questioning it in the US, so it’s interesting to see what international data says.