Note on the recession risk score (2026-03-24 – 2026-07-29)
The recession risk score is built from four components. One of them, the Sahm indicator, is derived from US labour data. Between 2026-03-24 and 2026-07-29 that data did not update: four groups of series were declared in configuration but never assigned to a scheduled fetch, and 17 of them fell behind by 120–151 days, holding their 2026-02 values while the rest of the dashboard refreshed normally.
The Sahm indicator stayed at 0.27 — its peak — while the published series fell month after month to 0.07. The score read 46.4; with the data restored it reads 41.1.
The displayed category was affected, not only the number. On 53 days, beginning 2026-04-30, the dashboard showed Moderate where the data supports Low. Every difference ran the same way: the dashboard overstated risk.
We found this by asking whether the data behind the headline numbers was actually current, and checking each series against its source rather than against our own pipeline status. The comparison above applies a 35-day publication lag, so it never credits MID with data the source had not yet released.
Corrected on 2026-07-29. The fix is not the backfill. Scheduled jobs now assert that every declared data group is fetched by at least one of them — the same gap had been repaired twice before, one series at a time, without anyone asking which groups were still unassigned. The series behind the regime read and the recession risk score are now covered by the freshness guard, which previously watched only policy rates and the EUR curve.