US Recession Probability (Smoothed)

RECPROUSM156N is the smoothed US recession probability from the Chauvet-Piger dynamic-factor Markov-switching model, published monthly via FRED. The model uses four monthly macro indicators (nonfarm payrolls, real personal income less transfers, real manufacturing and trade sales, industrial production) to estimate the probability that the US economy is in a recession in any given month.

Why it matters: this is a rigorous econometric recession probability that has tracked NBER-defined recessions with high reliability since the 1970s. Probabilities above 80% historically coincide with NBER-dated recessions (with a 1-3 month lag for the NBER's official call). The model is used by the FRBSF Economic Letter, by academic researchers, and by institutional macro investors as a quantitative complement to leading indicators like the yield curve, Sahm Rule, and Conference Board LEI.

RECPROUSM156N is at low single digits (typically below 5%) in March 2026, far below the 80% recession-confirmation threshold. The model has not signaled recession through the 2024-2026 period despite the historical Sahm Rule trigger in August 2024 and the 26-month yield-curve inversion 2022-2024. The four input indicators (NFP, real personal income less transfers, real manufacturing-trade sales, industrial production) have all held up well in 2024-2026, which the model interprets as expansion.

The dynamic-factor methodology is conservative and lagged: it requires sustained deterioration across multiple indicators before flagging recession, and it typically aligns with NBER calls rather than leading them. The current low reading is consistent with the broader 2024-2026 soft-landing narrative. The risk is rapid deterioration: model probabilities can move from <5% to >80% in 2-4 months once recession begins. Watch for sustained declines in NFP (negative MoM prints) and industrial production.

Modern RECPROUSM156N range: 99% in late 2008 (GFC peak), 100% in May 2020 (COVID lockdown), low single digits during expansions. The model has correctly identified every NBER-dated recession since 1970 with a 1-3 month lag and minimal false positives. The drivers are the four input series: NFP (employment), real personal income less transfers (consumer income), real manufacturing-trade sales (industrial activity), and industrial production. Sustained YoY declines in any two of these typically trigger probability moves above 50%.

First, the next monthly print. Sustained moves above 20% would signal the model is detecting deterioration even as the headline narrative remains constructive.

Second, the four input series independently. Negative MoM industrial production combined with negative NFP for two consecutive months historically triggers the model.

Third, comparison with the Sahm Rule realtime indicator. When both fire together, the recession signal is high-confidence; divergence (Sahm fires but recession-prob doesn't, as in 2024) signals possible false positive.