Is a recession coming? Not yet, judging by the signs that warned before past ones. We track three warning lights here and update them after each monthly jobs report. As of early November, with unemployment at 4.1%, none of them is flashing.

*Updated Nov. 8, 2026, with the October jobs report.*

That may surprise anyone who has heard for years that a recession was due. The best-known warning sign, the gap between 10-year and 2-year Treasury yields, inverted in the spring of 2022, 55 months ago, and no recession has started. So rather than lean on one sign, we look at three that measure different things.

All three warning lights are off.

Source: FRED; calculated by MKViz.

What do the three lights measure?

Each light answers a different question.

Light	Latest	What it measures	Lights up when
Sahm rule	Zero (October)	Whether unemployment is climbing	It reaches 0.5 point
Jobless claims, 4-week average	194,750 (week of Oct. 31)	Layoffs, every week	It’s 20% above its low of the past year
Chance of recession within a year	9% (Nov. 6)	What bond prices say about the year ahead	It reaches 20% to 30%

Source: FRED; calculated by MKViz.

The Sahm rule comes from economist Claudia Sahm. It compares the last three months of unemployment with the low of the past year. A rise of half a point has come early in every recession since the 1950s.

The rate was 4.1% in October, down from 4.2% in September, and employers added 165,000 jobs. That leaves the rule at zero. It has false alarms, though, most recently in 2024, when unemployment rose for a while and then leveled off.

Jobless claims count people filing for unemployment benefits for the first time each week. They come out every Thursday, so they move first. The 4-week average in the table smooths out noisy weeks. The latest single week, at 189,000, was lower than all but 1% of weeks since 1967. That tells you layoffs are rare. It doesn’t tell you whether companies are hiring.

The chance of recession is our own estimate, built the way the New York Fed builds its own. It asks one question: when long-term Treasuries paid this much more than short-term ones in the past, how often did a recession follow within a year? Today the answer is about 9%.

How often has a jump in jobless claims meant a recession?

Less often than you might think. Since 1968 the 4-week average of claims has risen 20% or more above its low of the past year 17 times. A recession began within a year after only five of them.

Only five jumps, in orange, came within a year of a recession.

Source: FRED; calculated by MKViz.

The two blue bars at the bottom are jumps with no recession since: one after the 2020 lockdowns, one in 2023.

Claims jump for other reasons too. Hurricanes in 2005 and 2017 both pushed them up. So did the months right after past recessions, including 2020, as people kept filing even though the downturn had officially ended.

That’s why the lights work better together than alone. Unemployment and claims describe the job market today. The yield curve looks a year or more ahead. One light alone has often been a false alarm, at least within a year.

Why do economists put the odds higher?

Economists surveyed by The Wall Street Journal in October put the chance of a recession in the next year at 24% on average, well above our 9%. Our number comes from one input, the yield curve, and the curve is positive again. Economists also weigh things the curve can’t see: inflation still at 3.7% in September, the Fed raising rates again that month, and energy prices.

That last one matters. In March, when oil spiked during the war with Iran, traders on the betting site Kalshi put the odds of a 2026 recession above a third, CNBC reported.

Still, both numbers say a recession is less likely than not. Right now I put a bit more weight on the economists. But survey averages have been late at turning points before: in 2022 many forecast a recession that never came. What would settle it for me is the claims average starting to rise. Until then, the survey’s number doesn’t change what I do. It just sends me back to check my cash cushion and my fixed-rate loan.

How bad does a recession usually get?

If one does come, history gives a rough sense of size. Here are the 12 recessions since 1948, measured two ways.

The rise is measured from the month before each recession to the jobless rate’s peak.

Recession	How long	How much unemployment rose
Typical (median)	10 months	3.45 points
1980, the mildest rise	six months	1.5 points
2008, the Great Recession	18 months	five points

Source: BLS and NBER via FRED; calculated by MKViz.

For most households, that rise in unemployment is the part that matters, not the official dates. The first thing I hold the 10 months up against is how long my household could go without a paycheck. In my case that’s a few months of cash, then a credit line I keep for emergencies, then selling part of my investments.

What would make me start preparing?

I look at jobless claims first, because they come out weekly and move first. I don’t watch the level so much as how far the 4-week average has climbed from its recent low. Then I check the unemployment rate and the Sahm rule, and the model last.

When the lights disagree, I don’t average them. They measure different things. My rule is not to change my view until two different kinds of signals move the same way. The yield curve alone is a reason for caution. If weekly claims start climbing too, I’d treat it as a real warning.

These are my own lines, not an official standard: the Sahm rule turning on, the claims average staying about 20% above its low for several weeks, or the model’s chance climbing into the 20% to 30% range. Your own industry often shows it first, so watch the job postings and company news where you work.

Before any light turns on, the useful checks are the same ones as always. How many months of bills could you cover without a paycheck? Are your loans fixed or variable? Does your household depend on a single income?

What to watch next: jobless claims every Thursday, the November jobs report in early December, and the Fed’s December 9 decision, its second since rate hikes came back in September. We’ll update this page after each jobs report.

Sources

Method: Sahm rule as defined by Claudia Sahm and FRED, from the monthly unemployment rate. Jobless claims: the 4-week average of weekly initial claims compared with its lowest point in the prior 52 weeks; we marked every time it rose 20% or more above that low since 1968 and measured the months to the next recession start in the NBER monthly dates. Recession chance: a probit model of a recession within 12 months on the monthly 10-year minus 3-month Treasury spread, fitted on data since 1982. Recession size: for each NBER recession since 1948, its length in months and the rise in the unemployment rate from the month before it began to its highest point before the next recession or within 24 months after it ended. Data: BLS, Department of Labor, US Treasury and NBER via FRED; calculated by MKViz. Code and data

6 more sources