Workers aged 22 to 25 in AI-exposed jobs sit 19% below a counterfactual, and hiring looks like the channel

Stanford's gap for 22- to 25-year-olds widened from 15% to 19% in 11 months. Payrolls kept rising, so AI's first mark may be missing hires.

AI's first labor-market shock may be the junior vacancy a company never posts, not the worker it fires. Stanford's Digital Economy Lab finds employment of 22- to 25-year-olds in highly AI-exposed occupations about 19% below a counterfactual as of June 2026. The July 2025 vintage showed 15%. The gap widened 4 points in about 11 months, with no comparable gap for experienced workers.

The missing rung is the entry-level job a firm may stop filling even as its experienced employees become more productive. If AI lets experienced employees absorb junior work, today's hiring restraint could become tomorrow's shortage of experienced workers. The evidence points to a hiring mechanism. It does not yet show that AI caused the gap.

The gap shows up in hiring, and layoff counts hide it

Stanford's researchers describe the adjustment as occurring mainly through fewer hires, rather than more departures. The paper says it "appears to operate primarily through reduced hiring of young workers rather than increased separations". That channel keeps layoff counts quiet, so the signal worth watching is reduced junior recruitment, not a surge in dismissals. It fits an economy where Fed Governor Christopher Waller says the labor market has been "low hire, low fire" for over a year.

His 10 November speech gives the aggregate picture: unemployment of 4.1% in October and payrolls up 165,000. The 12-month average is about 80,000 a month. The share of new labor force entrants among the unemployed has "reached close to the highest levels of the past twenty years".

Five Fed Districts report AI-linked hiring restraint, while Fed voices split on size

The Beige Book gives the district view. The October edition says five Districts reported firms that cut staffing through attrition, delayed hiring or expected less job creation after investing in AI. The same report finds "little or no net change in overall employment levels",.

Governor Michael Barr, speaking for himself at Yale on 12 November, says "there has been relatively little displacement of workers by AI". He adds that AI may be a labor substitute in some sectors, "especially for younger, less-experienced workers". The views are his own, he says, and not necessarily the Board's or the FOMC's.

Governor Michelle Bowman, in her 10 November speech, puts adoption at about 25% of businesses and roughly 10% of workers using AI daily. She calls displacement a risk, even as others see wages rise.

The causal link to AI is unproven, and the Stanford authors say so

The objection is fair: the gap does not show that AI caused it. The most direct caveat comes from the Stanford authors. They say the result "raises legitimate questions both about how much of the pattern is actually caused by AI". They also question whether the ADP payroll sample generalizes. Their summary is plain: "We do not see widespread, economy-wide job displacement associated with AI".

Waller lists AI beside tariffs, immigration and general uncertainty as sources of caution. A weak entry-level market could have several parents.

The Beige Book shows employers describing AI as a reason for restraint at some firms. Stanford's gap is consistent with that and cannot prove it. The evidence supports AI-related hiring restraint at some firms, not a demonstrated economy-wide employment effect. Fewer hires differ from a weakening junior-to-senior pipeline. The unresolved question is the size of the offset, because no source here prices the extra jobs that cheaper production might create.

The ATM precedent says a 19% gap could still be a transition cost

The best precedent for an offset is the ATM. Economist James Bessen reports that tellers per average urban branch fell from 20 to 13 between 1988 and 2004. Urban branches rose 43%, and teller jobs did not decline overall.

On Bessen's account, cheaper branches created demand that absorbed the lost work. If AI does the same for firms, the 19% gap could be a transition cost. The critical countercase is that lower production costs could expand demand enough to create more jobs than AI suppresses through reduced junior hiring. An AI-driven demand boom could also offset lost employment overall while still leaving a persistent shortage of entry-level openings.

The analogy breaks at the bottom of the ladder. Tellers kept relationship roles, and nothing in Bessen's account says ATMs removed the junior work that trained people. AI may target entry-level tasks directly. The source covers total teller jobs, not hiring by age.

The China-shock research shows that a stable aggregate can hide concentrated damage. It estimates net US job losses of 2.0 to 2.4 million from import competition, 1999 to 2011, larger at the local level than the industry level. That is trade, not AI, and no equivalent causal estimate exists for AI here.

Barr cites a record low. The labor share of national income hit 52.8% in the second quarter of 2026. His postwar range is 60% to 65%. The record implies a share 7.2 points below the bottom of that range and 12.2 points below the top.

Barr dates the fall to 2001, which predates today's AI tools. His footnotes add that part of the post-pandemic drop reflects a shift toward less labor-intensive industries. He also says "some of this reflects measurement issues." The share is context for his capital-bias concern and a poor test of the missing rung.

The next test is the Federal Reserve's 25 November Beige Book. A rise above five Districts would suggest broader AI-linked hiring restraint. But testing the missing rung also requires evidence on junior hiring itself.

The BLS November jobs report on 4 December will show whether payrolls still hold. The evidence supports weaker young hiring where AI exposure is high. Whether that thins the supply of experienced workers is not shown, and AI as the cause remains unproven.