NBER Recession Watch
The official US business cycle chronology from the National Bureau of Economic Research, plus FRED's USREC monthly indicator. The canonical "ground truth" for recession dating — and famously slow to publish (median 7-month peak announcement lag, 15-month trough). Updated .
NBER's Business Cycle Dating Committee declares peaks and troughs only after extensive deliberation — protecting against false recessions, at the cost of late confirmation. The BCI nowcast on the dashboard exists precisely because of this lag: real-time risk management can't wait 7-15 months for NBER to confirm. This page surfaces NBER's view as the reference benchmark; the BCI-based regime calls live elsewhere.
Current state
USREC indicator history (1950 onwards)
USREC = NBER's monthly recession indicator (1 in recession, 0 in expansion). Shaded bands mark each recession period. Since 1950 the US has had NBER-dated recessions. The current expansion (started ) is on its ᵗʰ month — long by historical standards.
NBER chronology — every modern recession with its announcement lag
Why the announcement lag matters
The lag is the whole problem
Median NBER peak announcement comes
When NBER does call it, it's right
Despite the lag, NBER's record is essentially perfect — every modern US recession was correctly dated, and there are zero "false positive" recession declarations. The committee deliberates until the data is unambiguous. NBER's role is the ex-post benchmark; nowcasts measure themselves against this ground truth.
Most recent recession (for context)
Sources: NBER Business Cycle Dating Committee (chronology + announcement dates) and FRED USREC series (monthly indicator). Cached locally at
cache/alternative_data/nber/ and refreshed nightly via scripts/alt_data/fetch_nber_recessions.py.
Related research: the Recession Ensemble is the BCI nowcast counterpart — three-model probability of recession with published risk bands, designed to give a real-time signal years before NBER confirms. The Business Cycle Index is the underlying factor model.