Recession Indicators: Which Ones Actually Worked?

Five recession indicators were tested against the nine recessions that the National Bureau of Economic Research (NBER) dated from 1960 to 2020. The yield curve signalled 8 to 23 months before the peak in 9 of its 10 tests, with no completed false signal. Weekly jobless claims signalled before the peak in 4 of 8 recessions and gave 4 false signals.

The Sahm Rule, payroll declines and the Chicago Fed National Activity Index signalled after the peak in 25 of their 26 tests. The medians were 3 to 4 months after the peak. This guide sets out the tests, how far our own thresholds move the results and the signals since 2020, which it reports separately.

The NBER dates a recession by its peak month, when activity was highest. We explain the process in what a recession is and who dates it. The NBER has changed no announced turning point since 1978. We used its nine peaks from April 1960 to February 2020.

A test is one rule scored on one recession. A signal is a run of consecutive months in which an indicator's rule was met. The lead time runs from the first month of the signal to the peak month.

A signal before the peak has a positive lead. A signal after it is late. We count it as a detection but not as a warning.

A signal is false if the next peak came more than 24 months after it began and it did not begin during a recession. The 24-month horizon is our choice. A signal after the April 2020 trough, the month activity was lowest, has no later peak to measure against, so we report it separately.

Source: FRED; Chicago Fed; YX Insights

We set every threshold except two. The Sahm Rule's 0.50 is Claudia Sahm's. The −0.70 for the activity index is the Chicago Fed's.

The two leading indexes we found on FRED have stopped. The Philadelphia Fed's Leading Index ends in February 2020. The Organisation for Economic Co-operation and Development (OECD) composite leading indicator ends in January 2024.

We used the activity index, a weighted average of 85 monthly indicators of national activity. It measures activity that has been reported, so we expected it to signal late.

The Sahm Rule compares the 3-month average of the Unemployment Rate with its low of the previous 12 months. FRED's real-time version uses the rate as first published.

Payrolls, claims and the activity index include all revisions made to date. Payroll employment is one of the series the NBER weighs in dating a peak, so payroll signals and peak dates are not independent.

The chart plots each signal that matched a recession, in months from the NBER peak month. A dot left of zero came before the peak. The top row gives the months from each peak to the NBER's announcement of it, for the six recessions from 1980 to 2020.

Source: FRED; NBER; Chicago Fed; YX Insights

The yield curve led the peak in 9 of 10 tests. The dates are in our guide to the inverted yield curve. The 10-year minus 2-year spread gave no signal before the February 2020 peak.

The Sahm Rule signalled in all nine recessions, a median of 4 months after the peak. The exception was October 1969, 2 months before the December 1969 peak. Its one false signal began in November 1976.

Payroll declines signalled in all nine recessions, from 1 to 12 months after the peak, with a median of 4. The one false signal began in November 2002, after the 2001 recession had ended. The activity index signalled in all eight recessions it covers, from 1 to 11 months after the peak, with a median of 3.

Initial claims had the widest spread, from 16 months before the peak to 3 months after it. The median was 2 months before. The four false signals began in February 1977, October 1984, September 2005 and September 2017.

The NBER announced the six peaks from 1980 to 2020 between 4 and 12 months after they happened. All 33 signals that matched those six recessions came before the announcement.

Counted to the announcement month, the Sahm Rule came 2 to 8 months before, payroll declines 1 to 8, the activity index 2 to 8 and claims 3 to 18. The yield spreads came 13 to 35 months before. These are not lead times, because they are counted to a different date.

The chart gives the false signals under each threshold tested for claims, the Sahm Rule and payroll declines. The right-hand column gives the median timing in months, before or after the peak. Looser thresholds bought earlier signals and cost false ones.

The claims rule moved most. At 10% above its low, it signalled a median of 16.5 months before the peak and gave 21 false signals. At 30%, it gave 1 false signal and a median of 1 month after the peak.

Two other choices moved the results. A 60-reading minimum on the yield curve misses two of the four recessions on the 10-year minus 3-month spread. A 5-reading minimum adds two false signals on the 10-year minus 2-year spread.

A 36-month horizon, in place of 24, cuts the claims rule's false signals from 4 to 1 and lifts its median lead to 12.5 months before the peak. None of the thresholds is settled.

The NBER's last trough is April 2020. Its table, last updated on 14 March 2023, lists no peak after February 2020. The table gives every signal that began after the April 2020 trough.

Source: FRED; NBER; YX Insights

All six signals have ended. The real-time Sahm indicator printed 0.00 in September 2026.

The months run from the start of each signal to October 2026. They are not lead times, false signals or a verdict. They do not belong beside the completed lead times above.

Our rule defines a false signal by the date of the next peak. The NBER has dated no peak since February 2020, so we leave these six unscored.

The record supports an order of timing and a count of false signals. The yield curve came first and claims next. The Sahm Rule, payroll declines and the activity index came after the peak.

It cannot put the indicators on one scale. Each covers a different span. Three of the five rules are ours. Nine recessions since 1960 is a small record. We chose the indicators after seeing the peak dates.

The 10-year minus 3-month spread signalled before the peak in all four recessions it covers, with no completed false signal. Four recessions is a small base. Its signal from October 2022 ended in December 2024. No other indicator in the set signalled early in every recession it covered without false signals or misses.

This explainer is part of the YX Insights Academy. Each one takes a single idea and checks it against real data.

The same approach runs through everything else we publish:

Good places to start on the website:

No indicator in this test gave early warning in every recession it covered with no false signals or misses, except the 10-year minus 3-month yield spread, which covers only four. The 10-year minus 2-year spread led the peak by 10 to 23 months in five of six. The Sahm Rule and payroll declines signalled in all nine recessions. Together they signalled after the peak in 17 of 18 tests.

In this record, an inverted yield curve came before the peak in 9 of 10 tests, by 8 to 23 months. The inversion that began in July 2022 ended on 26 August 2024. We report it separately, because the NBER's table, last updated on 14 March 2023, lists no later peak to score it against. We cover the dates in our guide to the inverted yield curve.

The Sahm Rule signals a recession when the 3-month average of the Unemployment Rate is 0.50 percentage points or more above its lowest 3-month average of the previous 12 months. It signalled in all nine recessions since 1960, a median of 4 months after the peak. It gave one false signal, in November 1976.

Yes. At our thresholds, the Sahm Rule gave 1 false signal, payroll declines 1 and initial claims 4. The yield curve and the activity index gave no completed false signal. The count depends on the threshold. Claims gave 21 false signals at 10% above their low and 1 at 30%.

The NBER announced the six peaks from 1980 to 2020 between 4 and 12 months after they happened. It announced the December 2007 peak on 1 December 2008 and the February 2020 peak on 8 June 2020. We explain how it dates recessions in what a recession is.

DISCLAIMER: This article is strictly educational. Any information or analysis in this note is not an offer to sell or the solicitation of an offer to buy any securities. Nothing in this note is intended to be investment advice and nor should it be relied upon to make investment decisions. Any opinions, analyses, or probabilities expressed in this note are those of the author as of the note's date of publication and are subject to change without notice.