Macro Labor Data

Every headline jobs number lands with an implicit story attached — recession, boom, resilience, crisis. The actual May–June 2026 data tells a story none of those words quite fit: an economy where almost nobody is losing a job, and almost nobody is getting a new one.

What It Is

This is the flow data underneath the paradox — the numbers that show employers and workers are both describing something real. JOLTS (the Job Openings and Labor Turnover Survey) measures the stocks and flows of the labor market monthly, with about a one-month lag; the Employment Situation report measures payrolls and unemployment. Together they’re the closest thing to ground truth this article has.

How It Works

The hires rate touched 3.1% in February 2026 — the lowest since April 2020, and before that, 2011 (CNN, March 2026). December 2025 came within half a point of the June 2009 Great Recession trough of 2.8%. The quits rate has sat at or below 2% for nearly a year, down from a roughly 3% peak during the 2022 “Great Resignation” — falling in nearly every sector, including leisure/hospitality (5.8% → 4.0%) and information/tech (1.9% → 1.1%). Low quits mean low confidence in outside options, which means fewer of the backfill vacancies that voluntary job-switching normally creates. Indeed Hiring Lab’s own characterization: the labor market “is definitely not broken, which is good news, but it’s also not really moving” (Indeed Hiring Lab, May 2026 JOLTS Report).

Economists have started calling this the low-hire, low-fire equilibrium: employers aren’t shedding workers, but they aren’t adding them either. It’s a meaningfully different diagnosis than a recession — unemployment claims and layoffs aren’t spiking, matching has simply stopped (St. Louis Fed, “The Effects of a ‘Low-Fire, Low-Hire’ Economy on Workers,” March 2026). June’s payroll print reinforces it: nonfarm payrolls rose just 57,000, well below expectations, with April and May both revised down and health care alone (+22,000) carrying nearly the entire gain while leisure and hospitality shed 61,000. The unemployment rate ticked down to 4.2%, but labor force participation fell to 61.5% and the employment-population ratio fell to 59.0% — a falling unemployment rate alongside falling participation means part of the “improvement” is people leaving the labor force, not finding jobs (BLS Employment Situation, June 2026; CNBC).

The clearest single signal of a matching failure, rather than a demand collapse, is duration: 1.9 million people have now been unemployed 27 weeks or longer, up 286,000 year over year and equal to 27.3% of all unemployed people. Rising duration alongside a historically low layoff rate means people aren’t losing jobs at unusual rates — the ones who do lose them simply can’t get rematched quickly.

The small-business side

Fifty-one percent of small business owners — and 84% of those actively trying to hire — reported few or no qualified applicants in June 2026, the highest reading since September 2024 (NFIB Small Business Economic Trends, June 2026). Nineteen percent now name “labor quality or availability” as their single biggest business problem, up six points from May’s reading, which itself had been the lowest since December 2016. NFIB is a self-selected member survey skewing toward smaller firms, so treat the specific levels with caution — but the underlying trend, a multi-decade time series moving sharply worse, is directionally reliable.

Hiring itself is expensive enough to make caution locally rational even before any skills mismatch enters the picture. Average cost per hire runs $4,700–$4,800 in 2026, up from $4,129 in 2019, with executive-level hires averaging roughly seven times the non-executive rate. SHRM-derived estimates put the cost of a bad hire at 50–150% of annual salary for most roles and 200%+ for executives. That asymmetry — a diffuse, unowned cost for a slow hire against a concentrated, attributable cost for a bad one — is a structural incentive problem, not a moral failing of any individual manager: rejecting 100 good candidates to avoid one bad one is, for that manager, the rational move.

It would be convenient to blame job boards and staffing agencies for profiting off this dysfunction, but their own numbers argue against it. ZipRecruiter’s FY2025 revenue fell 5% year over year, following a 27% decline in 2024; the Staffing Industry Analysts track staffing-industry revenue down 14% (2023), 12% (2024), and 3% (2025), with only a fragile 1–2% recovery forecast for 2026–27. The intermediary layer is deteriorating in step with everyone else’s experience, which is worth stating plainly as a check against easy villain-casting.

The compression point sits inside the recruiting function itself: 61% of recruiters report burnout and 64% report increased workload, driven mostly by a roughly 48% year-over-year rise in application volume — even as average application quality falls, thanks to AI-assisted “lazy apply” tools. Recruiters using generative AI screening report a 20% average workload reduction, but the tools themselves are strikingly unreliable: running the same AI screening tool twice on identical candidate data produces only 14% overlap in the resulting shortlist. That single statistic may be the most damning number in the entire research set — it validates both the applicant’s sense that rejection feels arbitrary and the employer’s failure to get anything reliable for what they’re paying for.

Why It Matters

Ghost jobs deserve one specific correction, because a genuinely wrong argument about them circulates constantly: the claim that subtracting monthly hires from monthly openings reveals the “true” number of fake postings. This is a stock/flow category error — openings is a stock measured on the last business day of the month, hires is a flow summed across the whole month, and the gap between them exists in every JOLTS release ever published, including the tightest labor markets on record. It proves nothing about ghost jobs on its own. The methodologically stronger estimate, from Greenhouse’s 2025 platform-derived study of actual posting and fill data, puts ghost postings at 18–22% of listings — real, but far more contained than the “90%+” figures that circulate from small self-report surveys.