Leading vs coincident signals

Recession indicators split into two fundamentally different jobs. Leading signals fire months before a downturn — they forecast. Coincident (nowcast) signals confirm a recession already in progress — they timestamp. Confusing the two is the most common methodological error in recession watching.

Why both are needed

A pure forecast panel is blind to unmodeled shocks: no yield curve predicted COVID-19. A pure nowcast panel gives no runway — by the time it fires, the recession has started. This monitor keeps them as separate blocks: forecast signals drive the composite score, nowcast signals confirm or veto whether a recession may already be underway.

Examples

Leading: yield curve (6–18 months early), building permits (often 12+ months), credit spreads, bank lending standards. Coincident: the Sahm rule and Chauvet–Piger probabilities — fast, real-time, but describing the present rather than the future. Initial claims sit on the border: high-frequency layoff counts that confirm onset within weeks but offer only a very short lead.

Fact: The Sahm rule's perfect record is for onset detection — economists calling it 'a recession predictor' are describing it wrong, which is exactly the kind of category error a nowcast/forecast split prevents.

See also

Personal alerts in the bot
This page updates when the site rebuilds — the bot messages you the moment a signal fires. Alerts and the weekly digest are free; Plus adds custom thresholds, quiet hours and point-in-time episode replays.
Get alerts in Telegram — @us_recession_chance_bot

Glossary · ← back