This page mixes three BLS surveys with different frames and error bands. The unemployment and participation rates come from the household survey (CPS), a sample of about 60,000 households; payrolls come from the establishment survey (CES), which counts jobs rather than people — one person holding two jobs is counted twice — and is revised twice after first release, sometimes substantially. Openings, quits, and hires come from JOLTS, a much smaller establishment sample with a one-month extra lag and wider revisions. Initial claims are an administrative count from state UI systems: near-real-time, but covering only workers eligible for (and filing for) benefits.
The V/U ratio is our calculation — JOLTS openings divided by CPS unemployed persons, matched by month — and inherits both surveys’ limitations: openings measure advertised positions, not the intensity with which firms recruit for them, and posting a vacancy has become cheaper over time, which can inflate the numerator relative to earlier decades. Payroll month-over-month changes are computed between consecutive published observations; where a monthly release was missed (several 2025 federal releases were skipped), a single point spans two months. Percentile context lines rank the latest value against each series’ full published history and describe rarity, not desirability.
None of these measures capture the quality dimension of work — hours, benefits, security, or whether a job matches a worker’s skills. A low unemployment rate is compatible with widespread underemployment, and headline tightness can coexist with slack in particular regions, occupations, and demographic groups.