Is the “labor shortage” number a big-business number?

JOLTS job-openings data gets cited constantly to argue employers can’t find domestic workers — the same argument used to justify sending jobs overseas and importing labor on guest-worker visas. We tested whether the survey behind that number is actually built from large-firm responses. Sources: BLS JOLTS establishment-size research series (Dec 2000–May 2026) and DOL OFLC H-1B/H-1B1/E-3 LCA disclosure data (FY2026 Q1–Q2).

Share of job openings from 5,000+ employee establishments

3.2%

May 2026, Total Private. Barely moved from 2.0% in 2019 and 1.9% in 2015 — a stable feature of the data, not a recent shift.

JOLTS survey response rate, 2020 → 2024

58% → 33%

And per BLS/Federal Reserve research, response is lower in the largest size classes — those estimates are the least reliable, not the most.

“Certainty stratum” in the sample design

100% sampling

Every establishment above a size threshold is included with certainty by design — it’s the small ones that get randomly subsampled.

Where the nation’s job openings actually sit

JOLTS classifies every sampled unit into one of six employment-size classes and publishes job-openings levels for each. This is establishment size — a single physical location — not firm size. Read at face value, the headline “can’t find workers” number is overwhelmingly a small-and-mid-business number: locations under 250 employees account for ~77% of measured openings.

Response rates cut the other way too: BLS and Federal Reserve researchers have both flagged that the largest size classes have the lowest survey response, meaning the already-small 5,000+ slice is also the noisiest one in the published data.

Establishment ≠ firm: why “large employer” doesn’t mean “large establishment”

A firm is one or many establishments under a common EIN. A retail or restaurant chain with 50,000 employees shows up in JOLTS as thousands of separate small locations — most falling into the “under 250” buckets above. A company that concentrates its U.S. workforce into a handful of huge campuses shows up as a few units in the certainty stratum instead. Using this project’s own H-1B/LCA disclosure data, distinct worksite addresses per employer vary enormously — illustrating just how differently “large employer” maps onto establishment size class from one company to the next.

Distinct worksite street addresses named across each employer’s FY2026 LCA filings. Not a headcount measure — DOL’s position-count fields turned out to be batch-filing ceilings (e.g. Qualcomm’s LCAs are routinely filed in round lots of 100), not confirmed hires, so we didn’t use them for anything requiring real magnitude.

The gap nobody outside BLS can check: JOLTS never publishes an industry × establishment-size cross-tabulation — not for tech, not for any sector. So the sharper version of the “big offshoring-heavy employers drive the shortage number” claim can’t actually be tested against public JOLTS data, by us or anyone else. That’s a real transparency limit in the dataset, not evidence either way.