Superpower Daily
Stanford Finds a 19% Employment Gap for Young Workers in AI-Exp...Alabama Subpoenas OpenAI Over Hugging Face Breach, Testing a Ne...Meta Sends MetaRoCE to OCP for Loss-Tolerant AI EthernetNVIDIA Adds BlueField-4 Scale-In for 800 Gb/s AI-Factory Infras...Microsoft Turns AI Governance Into Runtime Controls Across Nine...AWS Brings Ray Into SageMaker HyperPod With Recovery Tools and...Stanford Finds a 19% Employment Gap for Young Workers in AI-Exp...Alabama Subpoenas OpenAI Over Hugging Face Breach, Testing a Ne...Meta Sends MetaRoCE to OCP for Loss-Tolerant AI EthernetNVIDIA Adds BlueField-4 Scale-In for 800 Gb/s AI-Factory Infras...Microsoft Turns AI Governance Into Runtime Controls Across Nine...AWS Brings Ray Into SageMaker HyperPod With Recovery Tools and...
Businesspublished

Stanford Finds a 19% Employment Gap for Young Workers in AI-Exposed Jobs

The divide is concentrated in hiring, not layoffs or pay, and is strongest where AI is used to automate tasks. Economy-wide employment differences remain limited.

By 2 min read
Stanford Finds a 19% Employment Gap for Young Workers in AI-Exposed Jobs
Sign in to save and react

Listen to this story

The audio brief

About 1:33
0:001:33
Read transcript
Young workers in AI-exposed occupations are now showing a sharp employment gap: for people ages 22 to 25, employment is 19 percent lower than among peers in less-exposed fields. That divide has widened from 13 percent, according to an August update from Stanford economists. The pattern is concentrated in hiring, not in layoffs, quitting, or pay. Since 2022, employment fell about 11 percent in the 40 percent of occupations judged most impacted by AI, while it grew 10 percent in the remaining 60 percent. The researchers used anonymized, high-frequency payroll data aggregated by ADP, along with an occupation-impact gauge and the Anthropic Economic Index, which examines how workers use Claude. The clearest negative relationship appears where AI is used for automation—that is, replacing tasks previously done by people. Where it is used for augmentation, helping workers complete tasks they still perform, outcomes are more mixed and can be flat or positive. A second divide follows the kind of knowledge a job requires. Roles built around codified knowledge—learned through schooling, documents, or procedures—saw slower entry-level growth. Experience-heavy jobs built on tacit knowledge grew faster for mid- and senior-career workers. The important constraint is scope: economy-wide employment differences remain small. And this is an observed relationship, not proof that AI caused each change. The key question is whether the hiring gap among young workers persists or narrows as workplaces adapt.

Story brief

3 key points

An August update to Stanford economists’ labor-market research finds that AI exposure is translating into a sharper entry-level employment divide, especially for 22- to 25-year-olds. Employment fell roughly 11% since 2022 in the most-impacted 40% of occupations, versus 10% growth in the rest, with weaker hiring—not more exits—driving the difference. Automation-oriented Claude use shows the clearest negative...

  1. 01

    The reported gap widened from 13% to 19% between young workers in highly and less AI-exposed occupations.

  2. 02

    Economy-wide employment differences remain small, limiting the finding’s reach beyond early-career workers.

  3. 03

    Automation-oriented AI use correlates with weaker entry-level employment; complementary use aligns with flat or rising outcomes.

Workers ages 22 to 25 in the most AI-exposed occupations now have employment levels 19% below peers in less-exposed fields, according to updated research from Stanford economists. The gap has widened even as relative employment differences across the full economy remain small.

The researchers separate exposure from use

The August update of Canaries in the Coal Mine uses anonymized, high-frequency payroll data aggregated by ADP. It ranks occupations with an existing labor-market-impact gauge and the Anthropic Economic Index, which examines how occupations use Claude in everyday work.

Two types of AI use

  • Automation-oriented use means AI replaces work previously done by people. Occupations with more of this use had the weakest relative entry-level employment.
  • Augmentation-oriented use means AI helps workers perform tasks they still do. Entry-level employment outcomes in those occupations were more mixed.

The researchers say the pattern is consistent with automation-oriented AI substituting for labor, while complementary uses are associated with flat or rising employment. That is an observed relationship, not proof that AI use caused each employment change.

The young-worker gap widened

Employment gap for workers ages 22 to 25
13%→19%
percent

The reported employment gap between young workers in the most AI-exposed occupations and peers in less-exposed occupations rose from 13% to 19%.

Hiring, rather than exits, drives the pattern

Since 2022, employment for 22- to 25-year-olds fell about 11% in the 40% of occupations judged most AI-impacted. It grew 10% in the 60% judged least impacted. The difference is mainly associated with lower hiring rates, rather than increased firings or quitting, and appears primarily in employment levels rather than pay rates.

That distinction limits the finding. The study found little to no relative employment difference economy-wide between the occupations judged most and least exposed to AI; its clearest effect is among workers ages 22 to 25.

Formal knowledge shows a second divide

The researchers used required formal education in O*NET as a proxy for codified knowledge: knowledge taught through schooling, documents, or procedures. Occupations with more codified knowledge had slower entry-level employment growth. Jobs relying more on tacit knowledge, built through experience, showed faster growth for mid-career and senior workers.

Occupations with more college graduates showed a smaller gap between more- and less-exposed work. In occupations with fewer graduates, the least-exposed jobs grew while the most-exposed declined.

Sources

  1. arstechnica.comAI is hitting entry-level jobs hardest, Stanford study finds

Related coverage

Continue reading