Early-warning labor indicators

United States: who is hurting, and where

Unemployment rate and 12-month change are 3-month averages to September 2026; the other columns are the single latest month (September 2026). Not seasonally adjusted at source, so the 12-month change compares like months. The employed share is shown only for groups defined by age and education; the occupation groups are defined by a current or most recent job, so everyone in them has worked and the share would read near 100%. Small groups are noisy: graduates 22–27 have roughly 87 unemployed survey respondents in the latest month, so read a single month with care.

Is anyone a canary in the coal mine?

The data do not show a clean lead: young graduates and early-career white-collar workers tend to move with the wider market, not ahead of it, in the Jan 2010–Sep 2026 history.

Method: for each month, the 12-month change in the 3-month-average unemployment rate of the candidate group is correlated with the same measure for the comparison group 0 to 12 months later. A group is called a leader only if the best lag is at least three months, the correlation is 0.4 or higher, and it beats the same-month correlation. March 2020 through February 2022 is left out, because the pandemic shock and rebound would otherwise dominate every comparison. The remaining history has no sharp downturn, so a lead that only appears in one would not show here. Correlation is not causation; treat this as a screen, not a forecast.

Job cuts that cite AI, by country

Layoff announcements NextChapter tracks where news coverage ties the cuts to AI, automation or machine learning, grouped by the country the company is based in.

This is what the coverage says each company claimed, counted once per company per month. It does not verify that AI was the cause: companies sometimes cite AI for cuts driven by cost or demand, and some say outright that AI is not the reason. Country is the company’s home country, not where the jobs are lost; a global cut counts at headquarters. It covers only announcements on NextChapter’s tracked list, so large announcements not on it are missing. Collection began in October 2026, so there is no earlier history. The AI-washing check on SEC restructuring filings in the monthly report tests the claim directly.

Compared with ten peer economies

Young workers everywhere are unemployed at a higher rate than the overall workforce. What matters for the canary question is whether that gap is widening, and whether it is widening more where AI adoption is higher or growth is weaker. Sorted by how much the youth-to-total ratio grew over the past year.

• Youth and all-ages rates are seasonally adjusted (OECD); the latest month differs by country because statistical offices publish on different schedules.
• Graduate gap = unemployment rate of tertiary-educated 25–34 year-olds minus that of 25–34 year-olds with upper-secondary education (latest year, OECD). A negative number means graduates fare better; a number near zero or positive means the usual degree advantage has gone.
• “Jobless 12+ months” is the share of all unemployed people out a year or longer (OECD, latest year).
• “Firms using AI”: Eurostat for the six EU countries (share of enterprises with 10+ staff, 2025 uses a revised survey so is not strictly comparable with earlier years); for the U.S., the Census Bureau’s Business Trends and Outlook Survey (24.6% of businesses of all sizes in the latest two-week period). The two measures differ in who is counted and are not directly comparable. No consistent figure is available here for the UK, Canada, Australia or Japan.
• “Jobs cut citing AI” counts only layoff announcements NextChapter tracks, by the company’s home country and not where the jobs are lost; see the AI section above. Zero means none tracked, not none happened.
• GDP growth is real annual growth for the latest full year (IMF World Economic Outlook; recent years may be estimates).
• With eleven countries and few matched AI figures, this table is descriptive. It cannot show that AI adoption causes any of these differences.