Data through: September 2026 · updated Nov 2 · Updates: Monthly · Next: Dec 1
MThe three bars show the last 3 months for this indicator, oldest left to newest right. Bar height reflects each reading relative to the other two — the tallest is the highest of the three, the shortest the lowest.
Historical context: Currently in the 65th percentile of all historical readings.
What this means right now: The model puts the probability that the economy is currently in recession at 10% or below. Its four coincident inputs — payrolls, industrial production, real income and real sales — are not moving together the way they do in recessions. The model is coincident, not leading: it says nothing about the next 12 months, and it read 10% or below in the year before five of the eight recession starts since 1967.
Smoothed US Recession Probability · Monthly · 1967–2026
NBER recessions
Scroll to read · zoom and pan in Expand
Monthly values
Smoothed US Recession Probability — the last 12 months
Every month's reading, with the band the heatmap gives it. The chart above shows the same series in full.
Month
Value
Band
Sep 2026
0.4%
Very strong reading
Aug 2026
0.2%
Very strong reading
Jul 2026
0.1%
Very strong reading
Jun 2026
0.1%
Very strong reading
May 2026
0.1%
Very strong reading
Apr 2026
0.1%
Very strong reading
Month-end readings of the same series as the chart. Values are the current record, not point-in-time: a month that has since been revised shows its revised value.
What is the Recession Probability?
The Smoothed US Recession Probability is a monthly probability of the US economy currently being in recession, derived from the Chauvet-Piger dynamic factor Markov-switching model. The model is applied to four monthly coincident indicators: nonfarm payroll employment, industrial production, real personal income excluding transfers, and real manufacturing and trade sales.
The model was developed by economists Marcelle Chauvet and Jeremy Piger and is published by the Federal Reserve Bank of St. Louis on FRED as series RECPROUSM156N. It is a coincident model rather than a leading model, rising to high probabilities in most NBER-dated recessions (it reached 100% in 1974, 2008–09 and 2020, and peaked at 28.1% in 2001). But its real-time signals have historically aligned closely with the NBER's eventual recession dating.
How We Color-Code the Recession Probability
Our heatmap colors this indicator against historically significant thresholds:
10.0% or below
The model reads no recession under way — all but 11 of the months outside NBER recessions since 1967 read 10% or below
10.0% – 20.0%
Low — reached outside recessions in only 6 months since 1967 (April 1975, January 1978, January 1980, December 1982, January 2001, September 2005)
20.0% – 35.0%
Elevated — seen outside recessions only in April 1979 and February–March 2001 since 1967
35.0% – 50.0%
High — outside recessions only in November–December 1969 since 1967, each in the three months before a recession began
Above 50.0%
Recession reading — every one of the 52 months above 50% since 1967 fell inside an NBER recession
Historical Extremes — What Happened Next?
When this indicator reaches extreme levels, history shows consistent patterns:
Sep 2008Global Financial Crisis peak
100.0%
The model read 100.0% in September 2008, the month Lehman Brothers collapsed and credit markets froze. It first passed 50% in March 2008, after the recession began, and stayed above 50% through the recession's last month, the NBER's trough.
Apr 2020COVID-19 recession
100.0%
The model immediately hit 100.0% in April 2020 as payrolls, industrial production, and personal income all collapsed simultaneously. The recession was the sharpest on record but also the shortest as of August 2026 — lasting only two months (Mar-Apr 2020).
Jun 2009Post-GFC recovery signal
60.8%
As the four coincident indicators began recovering in mid-2009, the model retreated from its peak: 60.8% in June, then 2.8% in July 2009. That provided an early quantitative signal that the recession was ending, consistent with the NBER's eventual trough date, the last month above 50%.
Frequently Asked Questions
Has the model ever been wrong?
In the sense of a false alarm, no: every month above 50% since June 1967 — 52 of them — fell inside an NBER recession. The probability rose above 50% in seven of the eight recessions since then (in 2001 it peaked at 28.1%). But it is a coincident model: the first month above 50% came 0–7 months after each of those recessions began, never before one. Each month's reading has come about a month after the month ends since 2019 (a median 32 days from 2019 through 2025; about nine weeks, a median 62 days, from 2012 through 2018). In 2022–23 it read no higher than 1.2%. The series is also revised as the model is re-estimated with each release, so past months can change; this site keeps today's figures, not the ones first printed.
What are the four inputs?
Nonfarm payroll employment, industrial production, real personal income excluding transfer payments, and real manufacturing and trade sales — four monthly measures of current activity. The model extracts their common movement and estimates, each month, the probability that the economy is in its recession state.
What is the difference between this and other recession models?
This model uses four coincident series (payrolls, industrial production, real income, real sales) and estimates whether a recession is under way now. Yield-curve models, such as the New York Fed's, use the Treasury term spread to estimate the probability of a recession within the next 12 months — a leading estimate. The two answer different questions.
How does the recession probability model work, and what inputs does it use?
It is the Chauvet-Piger dynamic-factor Markov-switching model. The four coincident series share one unobserved factor that switches between an expansion state and a recession state. The model reports the smoothed probability of the recession state each month. It is published by the Federal Reserve Bank of St. Louis on FRED as RECPROUSM156N, from June 1967, and is revised as the model is re-estimated with new data.
How does this read alongside the site's other recession gauges?
It is a coincident gauge: it registers a recession once payrolls, production, income and sales are falling together — 0–7 months after the start in the seven recessions since 1967 in which it passed 50%. The Sahm Rule, also built on coincident data, has triggered within three months of every recession start from 1960 to 2020, though it also fired twice with no recession through September 2026, in 1976 and 2024. The yield curve is a leading gauge. A low reading here means no recession has yet registered in the coincident data, not that none is coming.