On 1 June 2026, three economists at the Federal Reserve Bank of New York published an analysis that reframes a problem usually pinned on artificial intelligence. Writing in the bank's Liberty Street Economics series, Natalia Emanuel, Emma Harrington and Amanda Pallais argue that the sharp rise in unemployment among young college graduates owes less to AI than to the spread of remote work. Their estimate is direct: remote work can account for 64 per cent of the recent increase.
The figures they set out are precise: unemployment among college graduates under 29 averaged 3.1 per cent in the years 2017 to 2019, then rose to 3.7 per cent in 2022 to 2025, an increase of about a fifth. Over the same span the rate for more experienced graduates barely moved, slipping from 1.9 per cent to 1.8 per cent. The cost of the change landed on those at the start of their working lives.
Where the gap opens
The authors separate jobs that can be done remotely, such as software engineering, from those that cannot, such as mechanical engineering. In remotable occupations, the unemployment rate of young graduates rose by almost a full percentage point between 2017 to 2019 and 2022 to 2024, while the rate for older workers in the same occupations edged down. In non-remotable work, the gap between young and old widened briefly in 2020 and then returned to its earlier level.
That divergence is the heart of the finding. The rise in young graduate unemployment traces almost entirely to occupations where the work moved to a distance. The timing reinforces the point: the increase began with the pandemic, before generative AI spread widely, and it persists even when the authors hold an occupation's exposure to AI constant.
What proximity was doing
To understand the mechanism, the authors turned to proprietary data from a single large company. When people worked beside their colleagues, they received more feedback on their output and more mentorship. When even a short distance separated them, that feedback fell away sharply, and the loss was greatest for younger workers, who rely most on the comments that shape their development.
The firm's hiring behaviour told the same story: while its offices were closed, it hired fewer inexperienced people and more who could work without much guidance. When the offices reopened, it returned to hiring younger staff, except on its distributed teams, where it continued to favour experienced workers. The company, in effect, was willing to train juniors when they were close enough to teach and reluctant when they were not.
If the route into a profession has always run through proximity to more experienced colleagues, what happens to a generation hired, where it is hired at all, into roles where that proximity no longer exists?
The structural reading is clear. The difficulty facing young graduates is not a shortage of talent or effort; it is a property of how work has been arranged. The conditions through which judgement is built, close contact, regular feedback, the informal correction that happens when people share a room, have been thinned out, and the people who needed them most are the ones being turned away.
Opinion: The On-Ramp Was Never Automatic
Early-career development has always rested on something that rarely appears in a job description: the steady, low-stakes exposure to experienced colleagues through which a novice becomes capable. That exposure was a feature of the working environment, not a trait of the individual. Remove it, and the individual is left carrying a cost the structure created.
The New York Fed's findings describe exactly this transfer. Firms that once absorbed the cost of training juniors by working alongside them have, in distributed settings, declined to take it on. The young worker is not told that the door has closed because the mentorship model broke; they are simply not hired, and left to read the outcome as a judgement on themselves.
Return-to-office mandates are now being defended partly on these grounds, and the research lends them some support. But proximity is a blunt instrument. The question is not only whether people share a building, but whether the structures that turned presence into development, attention, feedback, the deliberate teaching of the inexperienced, are being rebuilt or merely assumed.
What the figures leave open is who owns the on-ramp now. If the conditions that form professional judgement are no longer a by-product of how work happens, then someone has to choose to create them. The alternative is a labour market that asks each new graduate to arrive already experienced, and quietly penalises the distance it designed.
Declaration of Generative AI and AI-assisted technologies in the writing process:
The author made use of Generative AI or AI-assisted technologies in the preparation of this post.
Sources
Natalia Emanuel, Emma Harrington and Amanda Pallais, "Remote Work Leaves Younger Workers Sidelined," Federal Reserve Bank of New York, Liberty Street Economics, 1 June 2026
Natalia Emanuel, Emma Harrington and Amanda Pallais, "The Power of Proximity to Coworkers," NBER Working Paper No. 31880
The contents of this article are for informational purposes only and do not constitute professional, legal, or financial advice.
