The Recent Evolution of AI-Related Labor Demand (WP 26-24)
A · Strong evidence Federal Reserve Bank of Cleveland economist Kevin Rinz links Lightcast job ads and Current Population Survey flows to standard AI-exposure measures in documented event-study regressions through 2026Q2.
Summary based on the full text·Original: about 25 minutes to read
In short
Cleveland Fed Working Paper 26-24 (September 2026) by Kevin Rinz. AI mentions in US job ads have climbed since early 2024, mostly in occupations whose tasks LLMs can perform. On job-ad data alone, demand in those occupations looks stable to stronger: postings stopped declining and posted wages rose. Worker-flow data add a less comfortable picture: more churn, more people competing per opening, and falling wages for those who stay in their jobs.
Key numbers
- 3.1 percentage pointrise in the share of job ads mentioning AI per standard deviation of LLM exposure
- 3.2 percenthigher real posted wages per standard deviation of exposure by 2026Q2, vs 2022Q3
- 0.9 percentfall in real wages of continuously employed workers in exposed occupations, 2024Q3-2026Q2
- 9 percentrise in unemployed workers per new job ad per standard deviation of exposure
- 95 percentshare of the 2024Q3 'artificial intelligence' spike driven by within-employer language shifts
As reported by the source.
English·Web page·opens d.repec.org
What it reports
- AI mentions in full-time job ads spiked in 2024Q3, almost entirely the phrase "artificial intelligence" in tech roles; 95% of that spike came from employers changing wording, and it faded fast. Six quarters later about 2% of those employer-occupations used a specific AI skill term.
- Each standard deviation of LLM exposure is associated with a 3.1-point higher AI-mention rate, close to 4 points in later quarters. Older, non-LLM exposure measures show weak or negative links.
- Postings in exposed occupations stabilised after a pre-ChatGPT decline. Real posted wages were 1.7% higher per standard deviation by 2023Q1 and 3.2% by 2026Q2; the rise is clearer in ads that do not mention AI.
- Exposed occupations saw outflows and inflows rise by 3-5% of pre-period averages; unemployment transitions were almost all layoffs, with no rise in quits. Gross churn is nearly 0.6 points higher since the spike.
- New-hire wages rose about 1% (not significant); continuously employed workers' wages fell about 0.9% (significant). Unemployed per new job ad rose about 9% per standard deviation.
- Younger workers saw reduced hiring relative to older workers in exposed occupations, and minimum experience requirements rose.
How they got it
Occupation-by-quarter event studies with occupation and quarter fixed effects, using 2022Q3 as the reference quarter, applied to Lightcast online ads. Exposure is the first principal component of several published LLM-exposure indices, compared against Webb, Frey-Osborne and Brynjolfsson et al. measures. CPS month-to-month person links give flows and earnings, with three periods (pre-ChatGPT, post-ChatGPT, post-spike). Appendices test selection into wage posting.
Limits
- Descriptive correlations with exposure, not causal effects of AI adoption.
- Posted wages can be strategic and the share of ads posting a wage rose over the period.
- CPS cells are small, forcing long periods and wide intervals; the net-outflow difference is not significant.
- Short paper; most robustness checks sit in appendices.
Why it matters for companies
Job-ad volume and posted pay can look healthy while the internal market for exposed roles softens. Incumbents in those roles face less pressure for raises because outside options tightened, which shows up first in retention and pay-review data, not in hiring metrics. Generic "AI" language in ads tends to be branding that fades; firms tracking AI skills demand should look for specific skill terms. The drop in junior hiring in exposed roles is now visible in public statistics, so talent pipelines need a deliberate plan.
Summary based on the full text. Read the original for the full argument and data.Read the original on d.repec.org
- Tags
- Workforce, Job postings