Business Cycle Indicators
Forecasting recessions has been called the dismal science's dismal task. Yet a consistent set of indicators leads, coincides with, or lags the business cycle. Yield curve inversion has been the most reliable single leading indicator. Coincident indicators (employment, industrial production) confirm what's happening now. Lagging indicators (unit labor costs, inflation) reflect past dynamics.
Key insights
Yield curve has been the canonical leading indicator
Inversion of the 10-year minus 3-month Treasury yield curve has preceded every US recession since 1969 — 8 of 8. False positives: 1966 inversion didn't lead to recession (most disputed); the 2022-24 inversion has not produced a recession through October 2026, the longest gap after an inversion on record. Whether the indicator is still working or has been broken by Fed balance sheet operations is the main current debate.
PMIs are real-time global pulse indicators
Purchasing Managers' Indices survey procurement managers about new orders, production, employment, supplier deliveries, inventories. The composite reading above/below 50 separates expansion from contraction. Manufacturing PMIs lead; services PMIs are more coincident. Global PMI is published monthly within 5-7 days of month-end — far faster than GDP statistics.
The Sahm Rule usually fires at recession onset
Claudia Sahm's rule: a recession is essentially confirmed when the 3-month average of unemployment rises 0.5pp or more above its trailing 12-month low. Before 2024 the rule had signalled every US recession since 1970 with no false positives in the modern record. It then triggered in mid-2024 and no recession followed, the first clear miss, which Sahm herself attributed to post-pandemic labour supply. Used by Fed and FOMC members for real-time recession monitoring.
US 10y-3m Treasury yield spread 2000–2024
Percentage points
Show data table
| 10y-3m (pp) | |
|---|---|
| 2000 | -0.5 |
| 2003 | 2.5 |
| 2006 | 0.2 |
| 2007 | -0.3 |
| 2010 | 3 |
| 2013 | 2.5 |
| 2016 | 1.8 |
| 2019 | 0 |
| 2020 | 1.5 |
| 2022 | -0.5 |
| 2023 | -1.8 |
| 2024 | -0.4 |
Key Finding: Negative (inverted) values have preceded recessions. The 2022-23 inversion was the deepest since 1981.
Global manufacturing PMI 2018–2024
S&P Global Manufacturing PMI, monthly
Show data table
| Manufacturing PMI | |
|---|---|
| 2018 | 53 |
| 2019 | 50 |
| 2020 | 46 |
| 2021 | 55 |
| 2022 | 52 |
| 2023 | 49 |
| 2024 | 49 |
Key Finding: Drops below 50 mark slowdowns; 2020 COVID drop was deepest since 2008. Currently hovering around contraction.
Methodology & caveats
Conference Board LEI
The Conference Board's Leading Economic Index combines 10 indicators: average weekly manufacturing hours, initial jobless claims, new orders for consumer goods, ISM new orders, non-defense capital goods orders, building permits, stock prices, a leading credit index, the interest-rate spread and consumer expectations. A 6-month decline often precedes recessions; the indicator has produced false positives but few false negatives.
Why leading indicators fail
Leading indicators can fail when: structural changes alter relationships (yield curve and central bank balance sheets), data series are revised after the fact, the signal-to-noise ratio is low at small inversions, real-time data is mismeasured. The trick is using multiple indicators that fail under different conditions — diversification of signals reduces false positives.
Real-time vs revised data
Many recession-call methodologies use revised data that wasn't available at the time. Real-time monitoring requires using vintage data — what was published when. ALFRED (Archival Federal Reserve Economic Data) captures real-time vintages. Many backtests of recession indicators look better than they performed in real-time.
Why the two yield-curve spreads disagree
Analysts quote both 10-year minus 3-month and 10-year minus 2-year. The first reflects the current policy rate, and is the one the New York Fed uses in its recession model; the second reflects expected policy over two years. They can diverge for months, and the 10y-2y gave extra false signals in the past. On 30 October 2026 the 10y-3m spread is +78bp and the 10y-2y +38bp, so neither is warning now.
What would change the picture
Sentiment surveys and hard data have split since 2022: PMIs and confidence fell without output following. A real signal would be hard data turning together, with payrolls, industrial production and real income all falling. Watch temporary-help employment and initial claims, which tend to move before payrolls.
Cite this page
GlobeChart. “Business Cycle Indicators.” Last updated 2026-10-31. https://globechart.com/economy/business-cycle-indicators/
Please also credit the original publisher named under each chart: Federal Reserve, NBER recession dates; S&P Global PMI surveys. Spotted an error? Tell us.