Degrees, Jobs and the 2026 U.S. Labor Market: Testing the “Blue-Collar Boom, College Slump” Thesis

Review period: 1990-01-01 to 2026-09-19 Scope: Current and historical U.S. labor-market outcomes by education, age, occupation and industry, with emphasis on 2019–2026, entry-level white-collar work and possible AI effects. Source policy: Primary reliance on BLS, Census Bureau and Federal Reserve data and research; academic and private hiring-platform evidence is used with explicit methodological caveats.

Executive summary

Verdict: the claim is partly true but substantially overstated. There is a genuine and historically notable deterioration in the market for recent college graduates and entry-level workers in selected knowledge-work fields, especially technology, information, finance and other remotely performed occupations. But the broader proposition—that workers without degrees are enjoying one of their strongest labor markets in years while college graduates generally face a significantly weaker one—is not supported by the full set of national data.

In August 2026, unemployment among people age 25 and older was 2.7% for those with a bachelor’s degree or more, versus 4.4% for high-school graduates with no college and 4.7% for people without a high-school diploma. The 1.7-percentage-point high-school/college gap was only slightly wider than the 1.5-point gap in August 2019. The less-than-high-school/college gap was actually narrower than in 2019. College graduates therefore still have a large absolute employment advantage, and the current education gap is small by recessionary historical standards. (bls.gov)

Moreover, falling unemployment for the least educated partly reflects declining participation rather than exceptional job creation. Between August 2025 and August 2026, the unemployment rate for adults without a high-school diploma fell from 6.7% to 4.7%, but their labor-force participation rate fell from 47.5% to 44.7% and their employment-population ratio from 44.3% to 42.6%. High-school-only workers also experienced declines in participation and employment-population ratios. (bls.gov)

The important exception is recent college graduates, ages 22–27. Their unemployment rate was about 5.6% in the second quarter of 2026, versus 3.6% in March 2019, while their underemployment rate reached 42%. Those outcomes are clearly weak, although not unprecedented: recent-graduate unemployment exceeded 7% after the Great Recession, and underemployment near or above 40% has appeared repeatedly since 1990. (newyorkfed.org)

The best description of the overall market is “low hire, low fire, with unusually weak entry-level white-collar hiring.” July 2026 had 7.3 million openings but only 5.1 million hires, a 3.2% hiring rate and a 1.9% quits rate. Hiring fell sharply during the month in professional and business services. Information and finance had hiring rates of only 2.5% and 1.8%, compared with 3.9% in construction and 4.3% in retail. (bls.gov)

On AI, there is now credible early evidence of a contribution at the margin, particularly through reduced hiring of young workers in highly exposed occupations. Stanford researchers find a 19% relative employment shortfall for workers ages 22–25 in AI-exposed occupations, while two Census working papers find adverse post-ChatGPT outcomes for early-career workers and graduates from highly exposed majors. But these studies also document pre-existing trends, exposure-measure limitations and alternative explanations. Other research finds deterioration beginning before ChatGPT, and New York Fed research attributes much of the earlier rise in young-graduate unemployment to remote work and reduced mentorship. The evidence does not yet establish that AI is the principal cause of the broad white-collar slowdown, much less an economy-wide displacement shock. (census.gov)

1. Unemployment by educational attainment

Current position

Seasonally adjusted CPS estimates. (bls.gov)

These data reveal neither a broad noncollege boom nor a general college recession:

• Bachelor’s-degree unemployment was unchanged over the year and remained far below every lower-education group.
• High-school-only unemployment was slightly higher than a year earlier.
• Some-college/associate unemployment rose by half a point.
• The striking improvement was confined to the least-educated group—and was accompanied by a steep participation decline. (bls.gov)

Employment levels tell a similar story, although monthly CPS level comparisons are affected by sampling variation, demographic shifts and updated population controls. On a seasonally adjusted basis, bachelor’s-plus employment was about 259,000 higher than in August 2025, while employment was lower in all three groups below the bachelor’s level. Rates are therefore more reliable than raw group totals, but neither measure supports a sweeping noncollege boom. (bls.gov)

Historical unusualness

In August 2019, unemployment was 2.1% for bachelor’s-plus workers and 3.6% for high-school-only workers, a gap of 1.5 points. In August 2026, the corresponding gap was 1.7 points. That is a modest change, not a historical break. The current 2.0-point gap between workers without a high-school diploma and bachelor’s-plus workers is smaller than the 3.3-point gap in August 2019. (fred.stlouisfed.org)

During and after the Great Recession, educational unemployment gaps were far larger: high-school-only unemployment was around 9–10% during parts of 2011, while bachelor’s-plus unemployment was roughly 4–5%. Historically, recessions usually hurt less-educated workers more, widening rather than reversing the education gradient. (fred.stlouisfed.org)

Assessment: The aggregate educational divergence is not historically unusual. What is unusual is the combination of low overall unemployment, very low hiring and elevated unemployment among young graduates.

2. Employment and labor-force participation

All four education groups had lower employment-population ratios in August 2026 than a year earlier:

• less than high school: −1.7 percentage points;
• high school only: −0.8 point;
• some college/associate: −2.0 points;
• bachelor’s-plus: −0.5 point. (bls.gov)

Labor-force participation also fell over the year for every group. The decline was largest among people without a high-school diploma and those with some college or an associate degree. This is crucial because a falling unemployment rate can coexist with a weak market if potential workers stop searching and leave the measured labor force. (bls.gov)

Bachelor’s-plus employment-population ratios have also not recovered to their pre-pandemic level: the rate was around 72% during much of 2019, compared with 69.2% in August 2026. Some of that decline reflects aging and compositional changes, so it cannot automatically be attributed to weaker labor demand or AI. (fred.stlouisfed.org)

The broad conclusion is therefore more nuanced than unemployment rates alone suggest: employment utilization has weakened across education levels, not just among college graduates.

3. Recent graduates and entry-level white-collar hiring

The market for recent graduates is the strongest evidence for the thesis. The New York Fed estimates that unemployment among non-enrolled bachelor’s-plus workers ages 22–27 was about 5.6% in the second quarter of 2026, with 42% of employed recent graduates in occupations that typically do not require a degree. (newyorkfed.org)

The unemployment rate was roughly two percentage points above March 2019. It also exceeded the 4.1% overall unemployment rate in August 2026—an inversion that is psychologically and economically significant even though it does not apply to experienced college graduates. (newyorkfed.org)

This outcome is weak but not unprecedented. Recent-graduate unemployment surpassed 7% after the Great Recession, and the New York Fed’s historical series shows that underemployment around 40% is a recurring early-career phenomenon. College graduates often require time to move from noncollege jobs into occupations that fully use their training. (newyorkfed.org)

Private-platform data reinforce the picture, although they are not nationally representative. LinkedIn reported that U.S. hiring in early 2026 was about 23% below February 2020 and that degree-requiring entry-level roles had weakened particularly sharply. Its national hiring index was still 6.5% lower year over year in August 2026. Handshake reported that employers were posting roughly half as many jobs for the Class of 2026 as when that class entered college near the post-pandemic hiring peak. (linkedin.com)

Applications per posting have risen sharply, meaning much of graduates’ difficulty comes from a worsening ratio of applicants to available starting positions. This is consistent with a market in which layoffs remain limited but new labor-market entrants cannot get onto the first rung. (linkedin.com)

4. Job creation by industry and occupation

August 2026 payroll pattern

Total payrolls rose by 162,000 in August, but the average monthly increase over the preceding 12 months was only 31,000. Most August growth was concentrated in food services and local-government education rather than spread across the economy. (bls.gov)

(bls.gov)

This is a real sectoral divergence, but it is not equivalent to an education divergence. Health care, education, construction and manufacturing employ mixtures of degree and nondegree workers. Information, finance and professional services are degree-intensive, but they also experienced extraordinary pandemic-era hiring, remote-work expansion and interest-rate sensitivity.

Occupational unemployment

Not seasonally adjusted. (bls.gov)

The occupation data strongly reject an all-purpose “white collar bad, blue collar good” narrative. Professional unemployment worsened, but office-support unemployment improved. Installation and repair remained tight, while construction and extraction unemployment increased.

5. Major nondegree sectors

Construction and skilled trades

Construction remains relatively healthy. Industry unemployment was 3.1% in August 2026, the July hiring rate was 3.9%, and nonresidential specialty contractors continued to add jobs. Long-run BLS projections call for about 587,000 construction and extraction openings annually through 2035, largely from replacement as well as growth, with a 2025 median wage of $59,540. (bls.gov)

But it is not unambiguously booming: BLS described August payroll growth as little changed, and unemployment among construction and extraction occupations rose year over year. Projected replacement openings are evidence of continuing demand and retirements, not proof of exceptional current job creation. (bls.gov)

Manufacturing

Manufacturing has improved from its December 2025 low, led recently by machinery and fabricated metals. Yet its July 2026 hiring rate was only 2.3%, well below the economy-wide rate, and payroll employment remains characterized by a modest recovery rather than an extraordinary reshoring boom. (bls.gov)

Health care

Health care remains a major source of jobs, especially in home health and hospitals, although its 2026 pace has slowed. Because the sector contains physicians and nurses as well as aides, technicians, clerical workers and support staff, health-care growth cannot be classified simply as nondegree employment. (bls.gov)

Hospitality and food service

Food-service hiring was exceptionally strong in August, but leisure-and-hospitality unemployment was 6.6%, and average weekly earnings were only about $603—far below other major sectors. Strong gross hiring in hospitality often reflects turnover and seasonal demand, not necessarily exceptional job quality or bargaining power. (bls.gov)

Transportation, repair and other trades

Transportation and material-moving unemployment improved to 5.7%, while installation, maintenance and repair unemployment fell to 3.1%. These are among the clearer areas of relative strength for nondegree workers, but they do not represent the entire noncollege labor force. (bls.gov)

6. Job openings, hiring and labor-market churn

The U.S. market is no longer characterized by the extraordinary worker leverage of 2021–22. In July 2026:

• openings were 7.3 million, or 4.4% of employment plus openings;
• hires were 5.1 million, a 3.2% rate;
• quits were 3.1 million, a 1.9% rate;
• layoffs and discharges were 1.7 million, a 1.0% rate. (bls.gov)

This is a low-hire, low-fire market. Incumbent workers are generally secure, but unemployed people, career changers and new graduates face fewer transitions into jobs.

Sector hiring rates reinforce the divergence: construction and retail were above average, while manufacturing, information and finance were below average. Professional and business services recorded a 188,000 monthly decline in hires. (bls.gov)

Low hiring can raise recent-graduate unemployment even without mass layoffs. Employers facing uncertainty commonly delay positions that require training and instead retain experienced workers. That mechanism fits both the JOLTS data and the concentration of weakness among younger rather than older college graduates.

7. Wages by education, occupation and income level

The wage evidence is less supportive of a dramatic education divergence than the hiring evidence.

Average hourly earnings rose 3.1% over the year through August 2026. The Atlanta Fed’s median individual Wage Growth Tracker was 4.1%; job switchers received 5.0% growth and stayers 3.6%. These figures indicate continued nominal wage growth, but not a renewed economy-wide wage boom. (bls.gov)

Current industry wage levels remain highest in degree-intensive sectors. Average weekly earnings were about $2,066 in information, $1,878 in financial activities and $1,682 in professional and business services, compared with $1,646 in construction, $1,495 in manufacturing and $603 in leisure and hospitality. These averages reflect hours, occupational composition and seniority, so they should not be read as causal returns to education. (bls.gov)

The latest headline releases reviewed here do not establish a sustained 2025–26 acceleration in wage growth for high-school-only or bottom-quartile workers relative to bachelor’s-plus or high-wage workers. The Atlanta Fed tracker permits education, occupation and wage-quartile filtering, but the latest public headline is an overall measure. A strong claim of education-specific wage divergence therefore requires a dedicated CPS-microdata calculation rather than inference from sector hiring. (atlantafed.org)

Assessment: The thesis is stronger for access to new jobs than for current wage growth. The education wage premium remains large, even though selected trades offer strong earnings without a bachelor’s degree.

8. What is causing the divergence?

8.1 Monetary tightening and weak aggregate hiring

Higher financing costs and uncertainty have discouraged expansion in technology, venture-backed companies, real estate, finance and business services. The broad decline in hiring and quits predates much of the measurable AI effect and affects both AI-exposed and non-exposed entry-level jobs. Census research estimates that monetary-policy shocks may explain up to one-quarter of the relative early-career employment decline through the second quarter of 2025, although they do not explain all of the weakness at highly AI-exposed firms. (census.gov)

8.2 Post-pandemic normalization and overhiring

Technology, e-commerce, information and remote-work-friendly businesses expanded unusually rapidly during the pandemic. Subsequent hiring retrenchment, reduced digital demand growth and the end of near-zero interest rates would have weakened college-entry jobs even without generative AI. The current information-sector contraction and low finance hiring rate are consistent with this normalization. (bls.gov)

LinkedIn’s analysis of software-engineering hiring found that entry-level and overall software hiring broadly tracked the wider technology sector and national hiring slowdown, cautioning against treating the lack of an entry-level rebound as proof of an AI effect. (economicgraph.linkedin.com)

8.3 Remote work and the reduced incentive to train juniors

New York Fed researchers estimate that remote work could explain 64% of the increase in young-college-graduate unemployment between 2017–19 and 2022–24. Their mechanism is that remote and distributed teams provide less informal mentoring, feedback and apprenticeship, making experienced workers more attractive than recent graduates. Their firm-level evidence shows reduced hiring of inexperienced workers on distributed teams. (libertystreeteconomics.newyorkfed.org)

This estimate should not be treated as definitive for the entire economy: the firm-level evidence comes from a particular Fortune 500 setting, occupation-level remotability may correlate with other shocks, and the aggregate exercise is not a randomized experiment. Still, its timing—weakness beginning before mass generative-AI adoption—makes remote work a credible part of the explanation. (libertystreeteconomics.newyorkfed.org)

8.4 Demographics and replacement demand

An aging workforce can create persistent replacement demand in construction, maintenance, transportation and health care without generating rapid net employment growth. The large projected number of construction openings is therefore compatible with modest payroll growth: many openings replace retirees or workers leaving the occupation. (bls.gov)

8.5 Immigration

Immigration can increase labor supply in construction, hospitality, transportation and other nondegree-intensive sectors, potentially restraining wage pressure even while employers fill vacancies. But the August 2026 CPS showed the foreign-born civilian population at about 48.5 million, essentially unchanged from August 2025, so immigration alone is not a sufficient explanation for the latest monthly education pattern. Causal assessment would require industry-, geography- and skill-specific analysis. (bls.gov)

8.6 Reshoring, infrastructure and industrial investment

Federal infrastructure, semiconductor, energy and manufacturing investment plausibly supports selected trades and nonresidential construction. But if this were driving a broad nondegree boom, manufacturing and construction hiring should be exceptionally strong. Instead, manufacturing’s hiring rate was below average, construction payroll growth was characterized as little changed and the strongest manufacturing signal was a modest rebound from a recent low. The data support localized industrial strength, not a nationwide blue-collar supercycle. (bls.gov)

8.7 Supply of college graduates and credential competition

More workers competing for a limited number of traditional graduate-entry positions can raise recent-graduate unemployment without reducing demand for experienced professionals. It can also make employers more selective, shift hiring toward candidates with internships or prior experience, and reduce the signaling value of a bachelor’s degree. This is consistent with elevated recent-graduate unemployment alongside a 2.7% unemployment rate for the broader bachelor’s-plus population. (newyorkfed.org)

9. The AI hypothesis

Evidence consistent with an AI effect

Evidence against AI as the dominant explanation

• New York Fed research finds that young-graduate weakness was already concentrated in remotable occupations and that much of the increase predates rapid generative-AI diffusion. Holding AI exposure constant did not eliminate the age pattern. (libertystreeteconomics.newyorkfed.org)
• Research using unemployment-insurance records and LinkedIn profiles finds deterioration in AI-exposed jobs beginning in early 2022, before ChatGPT’s November 2022 launch, and no clear post-launch acceleration in many groups. (arxiv.org)
• Stanford’s more stringent firm-time controls push the start of the measurable divergence into 2024, underscoring that the apparent 2022 break partly reflects other shocks. (digitaleconomy.stanford.edu)
• The Stanford study finds no evidence of widespread economy-wide displacement, while Census research on business AI adoption reports that labor reductions remain uncommon and that adoption frequently augments existing tasks. (digitaleconomy.stanford.edu)

Bottom line on AI

The AI hypothesis has moved beyond pure speculation. Multiple independent administrative datasets now show a post-2022 relative decline in the hiring or employment of young workers in highly AI-exposed fields. The result appears strongest where AI substitutes for routine human tasks and weaker where it complements experienced workers. (digitaleconomy.stanford.edu)

But correlation remains stronger than causal proof. AI exposure is correlated with remote work, technology-sector overhiring, higher interest-rate sensitivity, geographic concentration and changing skill supply. Several studies find pre-trends, and the aggregate labor market shows no widespread AI displacement. The defensible conclusion is:

Generative AI is probably contributing to the entry-level knowledge-worker slowdown, but it is not yet demonstrated to be the principal cause of the overall college-graduate labor-market weakness.

10. Economic and policy implications

Consumer spending

A durable improvement in employment or wages for lower-income nondegree workers would support consumption because additional income is likely to be spent relatively quickly. Conversely, weaker hiring of young professionals can delay household formation, home purchases and spending on durable goods. These effects may partly offset each other.

Current participation and employment ratios, however, do not show a broad noncollege income boom. Federal Reserve business contacts also describe affluent consumers as continuing to spend while many other households remain cautious, suggesting that wealth, debt burdens and cumulative inflation remain at least as important as educational labor-market divergence. (bls.gov)

Wage growth and inflation

Labor shortages in health care, hospitality and selected trades can sustain wage pressure in labor-intensive services, while weak white-collar hiring restrains wage growth in finance, information and business services. The aggregate result is moderate nominal wage growth rather than a generalized wage-price spiral: average hourly earnings were up 3.1%, while the Atlanta Fed median tracker was 4.1%. (bls.gov)

Income distribution

Relative strength in trades and service occupations may compress earnings differences near the lower and middle parts of the distribution. But the large education wage-level premium persists, and weakened access to entry-level professional jobs could increase lifetime inequality by reducing career progression for younger cohorts. The latter is potentially more important than a temporary difference in starting unemployment.

Federal Reserve policy

The Fed must respond to the aggregate balance of employment and inflation, not one education group. Its September 2026 projections put median 2026 unemployment at 4.1%, PCE inflation at 3.7%, core PCE inflation at 3.4% and the median appropriate year-end federal-funds rate at 4.1%. Those projections imply that above-target inflation remains a major policy constraint even with weak hiring in some groups. (federalreserve.gov)

A widening college/noncollege divergence would matter to the Fed if it signaled hidden slack, impaired labor-force entry or falling aggregate wage pressure. Conversely, continued tightness in construction, care and hospitality could sustain service inflation even while white-collar sectors weaken.

Companies investing heavily in AI

For AI adopters, the current evidence supports potential productivity gains and selective reductions in junior hiring, but not yet wholesale labor-cost elimination. Companies that cut training pipelines too aggressively risk future shortages of experienced workers because senior talent is normally produced through junior employment and on-the-job learning.

The emerging pattern—AI complementing experienced workers while substituting for some junior tasks—could improve near-term margins but weaken internal talent development. Firms will need structured apprenticeships and AI-assisted training if the traditional first rung disappears.

The broader “AI trade” in financial markets

For investors, the labor evidence offers a mixed signal:

1. Bullish interpretation: AI may be producing real productivity and margin benefits before aggregate statistics show large displacement.
2. Cautionary interpretation: Evidence is concentrated in narrow cohorts and exposed occupations; it cannot yet justify economy-wide assumptions of massive labor substitution.
3. Macro risk: If AI suppresses entry-level income and career formation without quickly generating new occupations, consumer demand and political support for AI investment could weaken.
4. Execution risk: Heavy AI capital expenditure must ultimately produce revenue or measurable productivity; reduced junior hiring alone is unlikely to validate broad market valuations.

Thus, the labor data modestly support the microeconomic AI-adoption thesis but do not by themselves validate the strongest versions of the financial-market “AI trade.”

11. Indicators to watch over the next 12–24 months

Indicators that would confirm the thesis

1. Education unemployment: Bachelor’s-plus unemployment rises persistently while high-school-only unemployment remains stable or declines, materially narrowing or reversing the normal education gap.
2. Employment and participation: High-school-only employment-population ratios recover while bachelor’s-plus ratios continue to fall; falling nondegree unemployment without participation gains would not confirm strength.
3. Recent graduates: Recent-graduate unemployment remains above 5–6% after aggregate hiring recovers, and underemployment moves materially above its historical range.
4. Entry-level share of hiring: Young-worker hiring continues falling inside the same firms and occupations where experienced-worker hiring is stable.
5. AI exposure: The junior-employment gap widens specifically in occupations with observed AI adoption—not merely theoretical exposure—and survives controls for interest rates, remote work, industry, geography and pre-existing trends.
6. Sector hiring: Information, finance and professional-services hiring remains below construction, maintenance, manufacturing, health support and hospitality for several quarters.
7. Wages: CPS or Atlanta Fed microdata show sustained faster wage growth for high-school-only, low-wage or trade workers than for bachelor’s-plus and high-skill workers.
8. Vacancies and postings: Degree-requiring entry-level openings fall relative to nondegree openings even as total job postings recover.

Indicators that would contradict it

1. Recent-graduate unemployment falls as the aggregate hiring rate improves.
2. Entry-level software, finance, consulting and professional hiring rebounds following lower rates or improved business confidence.
3. AI-exposed occupations show similar recoveries for young and experienced workers.
4. Construction or manufacturing weakens while health care remains the principal source of job creation.
5. Nondegree unemployment stays low only because participation and employment-population ratios continue to decline.
6. Education-specific wage growth shows no sustained advantage for workers without degrees.

Conclusion

The United States does have a meaningful labor-market divergence—but it is best described as an age, hiring-flow and sector divergence, not a wholesale reversal of the economic return to education.

Experienced college graduates remain substantially less likely to be unemployed and substantially more likely to participate and be employed than workers without degrees. Nondegree sectors such as construction, maintenance, transportation and hospitality contain areas of strength, but manufacturing hiring is subdued, construction is mixed and hospitality remains high-turnover and low-paid.

The clearest losers are recent graduates and other first-time white-collar applicants trying to enter information, finance, professional services and remotely organized knowledge work during an unusually low-hiring period. AI probably contributes in highly exposed roles, and the 2026 Census and Stanford evidence deserves serious attention. Yet monetary tightening, pandemic overhiring, remote work, weak mentorship and sectoral normalization remain powerful competing explanations.

Accordingly, the thesis should be stated narrowly:

Recent college graduates and entry-level knowledge workers are experiencing one of their weaker labor markets of the post-2010 era, while selected nondegree occupations remain comparatively resilient. But workers without degrees as a whole are not experiencing an historically exceptional boom, and causal evidence that AI is the dominant driver remains incomplete.

Sources

• BLS, Employment Situation—August 2026, including education, occupation and industry tables. (bls.gov)
• BLS, Job Openings and Labor Turnover—July 2026. (bls.gov)
• BLS data via Federal Reserve Bank of St. Louis FRED, historical unemployment and employment-population ratios by education. (fred.stlouisfed.org)
• Federal Reserve Bank of New York, The Labor Market for Recent College Graduates. (newyorkfed.org)
• Emanuel, Harrington and Pallais, Remote Work Leaves Younger Workers Sidelined. (libertystreeteconomics.newyorkfed.org)
• U.S. Census Bureau, You’re (not) Hired: Artificial Intelligence and Early Career Hiring in the Quarterly Workforce Indicators. (census.gov)
• U.S. Census Bureau, Graduating into Disruption: Labor Market Outcomes for AI-Exposed College Majors. (census.gov)
• Stanford Digital Economy Lab, Canaries in the Coal Mine? revised August 2026. (digitaleconomy.stanford.edu)
• Frank et al., AI-Exposed Jobs Deteriorated Before ChatGPT. (arxiv.org)
• Federal Reserve Bank of Atlanta, Wage Growth Tracker. (atlantafed.org)
• Federal Reserve Board, September 2026 Summary of Economic Projections. (federalreserve.gov)
• LinkedIn Economic Graph and Handshake entry-level hiring research, used as contextual private-platform evidence. (linkedin.com)