The distribution
For the period ending July 2026, across 3,221 counties: 1,610 LOW, 806 MODERATE, 483 HIGH, 322 CRITICAL. The top two bands together are 805 counties — 25.0%.
A quarter of counties sounds like a crisis headline. It is not, and we would rather explain why than let the number travel on its own.
The score is a ranking, not a verdict
Our disruption score is built as a percentile. It places each county against every other county in the same month. That design has a consequence people miss: the bands are populated by construction. If every county in America improved by the same amount tomorrow, roughly the same number would still sit in the top quarter, because the score measures relative position, not absolute distress.
So "a quarter of counties are HIGH or CRITICAL" is close to a statement about our banding, not a discovery about the economy. Anyone who quotes it as the latter is quoting it wrong, including us if we ever do.
Where the score is genuinely useful
A percentile is most informative at the extremes and least informative in the middle. The gap between the 2nd percentile and the 98th is real and large. The gap between the 48th and the 52nd is mostly noise and will reshuffle next month with no change in anyone's circumstances.
That is why the useful question is never "is my county HIGH?" — it is "where in the tail is my county, and has it moved?" A county that has climbed thirty percentile points in six months is telling you something. A county sitting steadily at the 55th is telling you almost nothing.
Use the tails. Treat the middle as unranked.
What the score does not include
It is built from labour-market series — unemployment rate and its trend, acceleration, state claims, labour-force participation change, and a sector layoff signal. It does not know about a single employer's plans, a pending acquisition, or a plant closure that has not yet reached a public filing. For those, the earlier signal is a WARN notice, which an employer must file ahead of a qualifying layoff.
It also inherits the lag of its inputs. The newest month we can score describes July 2026, because county unemployment data publishes behind the calendar.
How to read any risk score, ours included
Three questions. Is it absolute or relative? A percentile cannot tell you the country got worse. What is in it, and what is missing? A score built only from labour data cannot see a corporate decision. Which period does it describe, and when was it computed? Those are two different dates.
If a provider will not answer those, the bands are decoration.