Graduate unemployment reaches the high-school rate
Will BLS report, for any month, an unemployment rate for people aged 25 and over with a bachelor's degree or higher that is at or above the rate for high-school graduates with no college?
Chance by Q4 2031
4%
Cumulative probability
- Latest forecast
| Year | Q1 | Q2 | Q3 | Q4 |
|---|---|---|---|---|
| 2026 | <1% | |||
| 2027 | <1% | <1% | <1% | <1% |
| 2028 | <1% | <1% | <1% | <1% |
| 2029 | 1% | 1% | 1% | 2% |
| 2030 | 2% | 2% | 2% | 3% |
| 2031 | 3% | 3% | 3% | 4% |
Model reasoning
I put this at about 4% over five years. College graduates aged 25 and over still have much lower unemployment than high-school graduates: 2.5% versus 4.4% in September 2026. In 34 years of monthly data the gap has never fallen below 1.2 points, and past white-collar slumps widened it. What would most change this forecast is a sharp, sustained jump in graduates' unemployment, for example from AI replacing office work, while trade and service hiring stays strong.
Where things stand
The first month that can count is September 2026, and it doesn't qualify. In BLS Table A-4 (released Oct 2, 2026), the seasonally adjusted rate for people 25 and over was 2.5% for bachelor's and higher (LNS14027662) and 4.4% for high-school graduates with no college (LNS14027660), a gap of 1.9 points. I pulled both full series (Jan 1992 to Sep 2026, 416 months; October 2025 is missing because of the shutdown) from the BLS API (https://api.bls.gov/publicAPI/v2/timeseries/data/) and computed:
- The degree rate has never been at or above the high-school rate. The smallest monthly gap was 1.2 points (Dec 2025). The highest degree-to-high-school ratio was 0.70.
- On a 12-month average, the ratio rose from 0.50 (Dec 2022) to about 0.64 (mid-2026), a record high. The 12-month average gap is about 1.6 points. The narrowing stalled in 2026, though: the degree rate fell from 3.0% in January to 2.5% in September.
- Noise: from 2022 to 2026, the monthly gap scatters around its 13-month centered average with a standard deviation of about 0.24 points (month-to-month changes about 0.32). Sampling noise alone therefore can't close a 1.6–1.9 point gap; that would take a move of more than six standard deviations. A published tie or crossing only becomes plausible once the underlying gap falls to about 0.5 or less. At 0.6, the chance is about 0.6% per month; at 0.4, about 5% per month.
Reference class and mechanism
The base rate is zero in 34 years of monthly data. Downturns centred on white-collar work have widened the gap rather than closed it. In the 2001–03 dot-com bust the degree rate doubled (1.5% to 3.2%), but the high-school rate rose from 3.5% to 5.7%, and the ratio peaked at only about 0.61. When graduates lose jobs, they take jobs that don't need a degree and push high-school workers out; displaced older graduates often leave the labour force instead of showing up as unemployed. A YES needs a large structural break: AI displacing degree holders at Great-Recession scale while blue-collar demand stays strong. For example, the degree rate rising about 1.3 points or more relative to the high-school rate, as with the degree rate near 4% and the high-school rate flat at 4.4%.
Paths to YES and NO
YES: AI cuts heavily into professional and entry-level white-collar work by 2028–2031. Physical, trades and service jobs stay protected because robotics lags behind software, and tight immigration and retirements keep non-degree labour scarce. Once the underlying gap is about 0.3–0.5, monthly noise gives repeated chances. Linearly extending the 2022–26 ratio trend (about +0.04 a year) only reaches about 0.85 by 2031, a gap of about 0.6. That gives at most a few percent, so YES needs the trend to speed up.
NO: This is the default. Current evidence points to gradual AI effects:
- Layoffs are at multi-year lows. Challenger counts 573,000 announced job cuts through September 2026, about 120,000 of them attributed to AI.
- Most experts, as reported in El País (Oct 11, 2026), describe AI's labour impact so far as "more qualitative than quantitative."
- The Metaculus/Renaissance labour forecasting hub projects only a 0.7–1.9% fall in employment from AI by 2030.
- Survey-based forecasts (Karger et al.) expect AI to show up as lower labour-force participation, not higher unemployment.
- A general recession, a live risk with the Fed raising rates and 10-year yields above 5%, would widen the gap.
- Office, sales and customer-service jobs held by high-school graduates are also exposed to AI.
Estimate
I used a hazard rate that is near zero for the next few releases. The Oct and Nov 2026 data (out Nov 6 and Dec 4) can't plausibly close a 1.9-point gap, so I leave only a residual for data errors. The hazard rises over time as possible AI disruption builds up. A Monte Carlo of the trend gap, with monthly noise of about 0.24, recession shocks, and random AI shocks that close the gap, gave 2–5% by end-2031 under reasonable settings. It reached about 10% only if I assumed a sustained narrowing of 0.1±0.15 points a year, which I think is too aggressive given that the narrowing stalled in 2026. I land at about 0.3% by end-2027, 1.6% by end-2029 and 3.6% by end-2031.
Strongest case against my number: if AI capable of doing most remote work arrives around 2027–28 and companies adopt it fast, degree holders' unemployment could rise far more than any history suggests. In that world, YES could be 20–40% likely, and giving it about 10% weight adds about 2–4 points. That is already roughly reflected in the tail I have included. Data errors and changes to the survey method add only negligible probability.
Sources
- BLS Table A-4, September 2026: https://www.bls.gov/news.release/empsit.t04.htm
- BLS release schedule: https://www.bls.gov/schedule/news_release/empsit.htm
- CNBC, September jobs report (2026-10-02): https://www.cnbc.com/2026/10/02/jobs-report-september-2026.html
- NY Post, summarising WSJ and Burning Glass (2026-09-07): https://nypost.com/2026/09/07/lifestyle/us-job-market-is-booming-for-people-with-no-college-degrees/
- Challenger, Gray & Christmas, September report (2026-10-01): https://www.challengergray.com/blog/job-cuts-fall-in-september-hiring-plans-up-3-over-2025-on-weak-early-seasonal-hiring/
- Metaculus/Renaissance labour forecasting hub (2026-04-20): https://www.globenewswire.com/news-release/2026/04/20/3277174/0/en/renaissance-philanthropy-launches-forecasting-hub-to-predict-ai-s-impact-on-the-workforce.html
- El País (2026-10-11): https://english.elpais.com/economy-and-business/2026-10-11/the-ai-labor-apocalypse-is-on-hold-for-now.html
- AIMultiple summary of AI job-loss forecasts: https://aimultiple.com/ai-job-loss
Resolution criteria
Question. By the last day of each quarter, has BLS reported, for any month, an unemployment rate for people aged 25 and over with a bachelor's degree or higher that is at or above the rate for high-school graduates with no college?
Definitions.
- Rates: the monthly, seasonally adjusted Current Population Survey unemployment rates for the civilian noninstitutional population aged 25 and over, both sexes combined: BLS series LNS14027662, “Bachelor's degree and higher,” and LNS14027660, “High school graduates, no college.” Both rates must refer to the same month and the same BLS publication vintage.
- Bachelor's degree and higher includes bachelor's, master's, professional, and doctoral degrees. High school graduates, no college includes a high school diploma or equivalent but no college attendance.
- At or above: the bachelor's-and-higher rate is greater than or equal to the high-school-no-college rate using the published rates to one decimal place; a published tie counts. Do not recalculate rates from rounded population counts.
Resolves YES if a qualifying BLS publication reports such a comparison for at least one month. Initial estimates and subsequently revised estimates both count. The Observation period convention applies: the observation month must end after the opening date, so September 2026 is the first eligible month. A qualifying publication must fall within the Resolution window and be published by the relevant quarter-end. Once YES, always YES, even if later revisions reverse the comparison.
Resolves NO if no qualifying comparison has been published by the quarter-end. Comparisons using not-seasonally-adjusted rates, annual averages, different months, different publication vintages, other age groups, bachelor's-only or advanced-degree-only subgroups, or all workers without a college degree do not count. No attribution to AI or particular occupations is required.
Resolving source. BLS's monthly The Employment Situation, Table A-4, “Employment status of the civilian population 25 years and over by educational attainment”, its archived releases, or BLS's official CPS database for the two specified series. A renamed table or successor series counts only if it preserves the specified population, educational categories, monthly frequency, and seasonal adjustment. BLS publications only; third-party reproductions are not independently resolving sources. Apply Resolution date (A3), the Resolution window, and the Observation period conventions.