Why we are publishing a number our own score cannot
We have said plainly, in earlier posts, that our risk bands are populated by construction: roughly a quarter of counties land in the top two bands every single month, whether the whole country is thriving or struggling, because the score measures relative position, not absolute distress. That is an honest limit, not a flaw to hide — but it means the score alone cannot answer a question people reasonably ask: is it actually getting better or worse out there? We can answer that from the same table, just not from the percentile.
What the raw averages show
Across all 3,221 counties, the average unemployment rate was 4.45 percent in July 2025 and 4.27 percent in July 2026 — down about 0.18 points. The median county moved from 4.3 percent to 4.0 percent. Comparing the same calendar month a year apart keeps normal seasonal patterns from contaminating the comparison, which is the same reason we compare July to July rather than July to April.
The breadth of the move matters as much as its size. In the year to July 2025, 1,414 of 3,221 counties — 43.9 percent — had higher unemployment than twelve months earlier. In the year to July 2026, that fell to 1,122 counties, 34.8 percent. Fewer counties are seeing their own unemployment rate climb than was true a year ago. Both the level and the breadth of deterioration moved the same direction: down.
What stayed exactly the same
In both of those months, our risk bands held 1,610 counties LOW, 806 MODERATE, 483 HIGH, and 322 CRITICAL — identical counts, both times, because the bands are sized by rank, not by level. If every county in the country had improved by the same percentage, the band counts would look identical to what actually happened. That is the whole point of the earlier warning: band counts are a statement about our ranking method, not a standalone economic reading. The raw average is the reading. The bands are the map of where each county sits relative to everyone else that month.
A small, real improvement, not a trend line
One year of same-month comparison is one data point, not a trend. We are not projecting this forward, and a 0.18-point move in an average built from 3,221 very different local economies can mask real deterioration in specific places even while the average eases — our own count of 199 counties that newly crossed into HIGH or CRITICAL this same year is proof that both things are true at once: the national average improved slightly, and a meaningful number of individual counties got noticeably worse. Neither number contradicts the other; they are different questions.
How to use both numbers together
Use the national average and median as the backdrop: is the overall tide going out or coming in. Use the county-level percentile score for the thing it is actually built for: finding where your specific counties sit against that backdrop, and whether they have moved. A backdrop without a map tells you nothing about any one place. A map without a backdrop lets you mistake normal churn in a fixed-size ranking for a national story it was never built to tell.
Look up where your counties sit against the national picture