The actual numbers
Our county table holds 3,221 counties across 36 monthly periods, running from July 2023 to July 2026. We last recomputed it on 18 September 2026.
Read those two dates together. The newest month of data available describes July 2026, and the freshest version of it we could build was assembled in mid-September — a gap of about 79 days. Today, in October, that same newest period is roughly eleven weeks behind the calendar.
Nothing is broken. County-level unemployment comes from the Bureau of Labor Statistics Local Area Unemployment Statistics programme, which publishes county estimates on a lag, and then revises them. A vendor cannot have July's county numbers in July.
Why this matters more than it sounds
If you are watching for a downturn in a specific county, an eleven-week lag means the earliest you can see it in this series is nearly three months after it began. Any decision that needs a faster signal — staffing, lending, site selection — cannot rest on this series alone, however precise it looks.
It also makes year-on-year comparisons easy to get wrong. Compare "the latest month" across two providers who refreshed on different dates and you may be comparing different vintages of the same month, because revisions land after first publication. The honest fix is to compare a named period, not "the latest".
What to ask a provider
Ask three questions, and accept only specific answers. Which period does your newest row describe? Not "we update monthly" — the actual month. When was it computed? The compute date and the data period are different facts and both matter. Do you keep revisions, or overwrite them? A provider who overwrites cannot tell you whether a number changed because the economy moved or because BLS revised it.
We keep the period and the compute date on every row for exactly this reason, and we will tell you both. A series that cannot answer those three questions is not a time series — it is a snapshot with a confident label.
The faster signals, and their own limits
For earlier warning, the useful sources are the ones that file on a legal clock rather than a statistical one: WARN layoff notices, which employers must file ahead of qualifying layoffs, and state unemployment-claims series, which publish weekly. Both are narrower than LAUS — WARN only catches layoffs above a threshold, and claims are state-level rather than county-level — so they are a complement, not a replacement.
The right structure is a slow, complete series for level, and a fast, partial series for change. What gets people into trouble is using one of them as if it were the other.