9 Recession Indicators Economists Actually Watch (Latest 2026 Readings)
No single indicator predicts recessions reliably. These nine, tracked together, form the best real-time picture economists have. Current readings sourced from FRED, ISM, the Census Bureau and the Conference Board, reviewed October 2026 (each series shows its own release month, which is not the same thing as this review date).
10Y-2Y Yield Curve Spread
Threshold: Inversion (below 0) precedes 7 of last 7 recessions
Lead time: Leads recession by 12-24 months
Current context: Inverted for roughly two years before re-steepening. The spread narrowed through mid-September, bottoming at +0.20 pp on 21 September after the Federal Reserve's 16 September rate increase to 3.75-4.00% lifted short-end yields, then recovered to +0.41 pp by 30 September, back in line with its early-September range of +0.40 to +0.43. The un-inversion itself is historically watched - recessions have sometimes begun as the curve re-steepens.
Caveat: False signal: 1998 inversion led to no immediate recession. 2019 inversion was followed by 2020 COVID recession, but COVID was arguably the cause, not the curve.
Sahm Rule Recession Indicator
Threshold: Recession signal at 0.50 pp rise in unemployment 3-month average above 12-month low
Lead time: Real-time signal, designed to trigger at recession start
Current context: Was elevated in 2024 as unemployment rose, then receded through 2025-26 and turned slightly negative, easing further to -0.07 in August 2026 (from -0.03 in July) as the unemployment rate held at 4.1%. A renewed sustained rise in unemployment would be needed to approach 0.50 again.
Caveat: Created by Claudia Sahm in 2019. Has triggered in every US recession since 1970 in real time. The current reading of -0.07 sits far clear of the 0.50 trigger.
Initial Jobless Claims (4-week avg)
Threshold: No publisher sets a fixed recessionary level, and a fixed level does not survive the record: at the eight recession onsets since this series began in 1967 the 4-week average ranged from 210,000 to 438,000.
Lead time: Leads formal recession declaration by 3-6 months when rising sharply
Current context: Fell from COVID-era highs of 900k+ to sub-200k by 2022. At a 200,000 four-week average for the week ending 26 September 2026 it has eased through September from around 207,500 in late August; the latest single week was 197,000, down 1,000 from the prior week's 198,000, while continuing claims eased to 1,701,000 (week ending 19 September) with the insured unemployment rate historically low at 1.1%. That 200,000 is 0.118% of a 169.8 million labour force, below all eight recession onsets since 1967 on that basis - though only just below February 2020, when claims stood at 209,650 and gave no warning of the recession that began that month.
Caveat: Highly volatile week-to-week. Focus on 4-week moving average. Holiday weeks and seasonal anomalies distort single-week readings significantly. The level alone is also a poor cross-era comparison, because the labour force has roughly doubled since 1967 - scaled to the labour force, the eight onsets span 0.12% to 0.40%.
ISM Manufacturing PMI
Threshold: Expansion above 50. ISM puts the overall-economy breakeven at 47.5. Sustained sub-45 historically precedes recession.
Lead time: Leads recession by 0-6 months when in deep contraction
Current context: Registered 54.5 in September 2026, down 0.1 point from August's 54.6 and a ninth consecutive month above 50. New orders strengthened to 55.3 (from 53.7) and backlogs jumped to 56.4 (from 51.8), while production eased to 56.7 (from 58.3) and the employment sub-index improved to 52.7 (from 51.2). The outlier is prices, up 6.8 points to 77.9, the sharpest input-cost reading of the cycle. ISM puts the September headline as consistent with real GDP growth of 2.4% annualised.
Caveat: Manufacturing is 9.7% of US GDP on a value-added basis (BEA, 2026Q2), so the signal carries most weight when manufacturing and services move together; an isolated manufacturing swing has limited macro reach. Watch the Services PMI separately.
Conference Board Leading Economic Index (LEI)
Threshold: 12-month change below -4% typically precedes recession
Lead time: Leads recession by 3-9 months on average
Current context: The LEI slipped 0.1% in August 2026 to 99.5 (2016=100) after rising 0.2% in July, its first monthly decline since March. Four of the ten components fell, with consumer expectations the persistent drag and building permits also down across nearly all regions. That tipped the six-month growth rate back to -0.1% (February to August 2026), from -0.6% over the prior six months - a much shallower decline than 2024's, but negative again rather than the positive turn the July reading had suggested. The Conference Board's reading is that the economy is still expanding but growth is set to slow, with real GDP forecast up 1.9% in 2026.
Caveat: The LEI is a composite of 10 forward-looking indicators, making it more robust than any single series. However, it has given false signals - notably in 2015-16 and 1995-96.
Consumer Confidence Index
Threshold: The Conference Board publishes no recession threshold for the headline index. Its 80 line applies to the EXPECTATIONS sub-index, which is at 63.6.
Lead time: Coincident to 3-month lead
Current context: Fell sharply from 140+ in 2021 during the inflation period, recovered toward 105 in mid-2025, then eased into the low 90s by 2026. It then dropped 6.7 points to 81.9 in September 2026 (1985=100, from 88.6 in August), its weakest in over a decade, in a release published 29 September. The Present Situation Index fell 7.9 points to 109.3 (from 116.2) and the forward-looking Expectations Index fell 5.9 points to 63.6 (from 69.5), a third straight monthly decline. It is the Expectations Index that carries the Conference Board's below-80 recession-ahead marker, and it has been under that line since early 2025 with no recession following - which is the caveat to hold against this whole series rather than a signal to act on.
Caveat: Consumer confidence is a survey of expectations, not actual spending behaviour. It can remain elevated even as financial conditions tighten, and can crash on news events without affecting spending. Note also that the headline index and the Expectations sub-index sit on different levels, so the 80 line cannot be read across to the headline.
Unemployment Rate (3-month trend)
Threshold: Rising trend is the concern. Sahm rule quantifies the rise needed to signal recession.
Lead time: Lags recession - unemployment typically rises after recession begins
Current context: Troughed at 3.4% in January 2023, among the lowest readings in 50 years. Rose to 4.3-4.5% through 2025 and eased to 4.2% in June, then held at 4.1% in July and August 2026. The soft summer now looks softer in the first telling than in the data: July payrolls were first reported as a 23,000 fall and have since been revised to a 21,000 gain (FRED PAYEMS). August added 162,000, well above the roughly 53,000 expected. The steady jobless rate is why the Sahm rule has stayed slightly below zero at -0.07.
Caveat: The level of unemployment (currently 4.1%) is less important than the direction and speed of change. The concern is the direction of travel, not the absolute level.
ICE BofA US High Yield Credit Spread
Threshold: Recessionary when sustained above 600 bps. Distressed credit above 900 bps.
Lead time: Leads recession by 3-9 months when rapidly widening
Current context: Reached 1100 bps during the 2008 crisis. At 312 bps on 30 September 2026 the spread is still far below the danger zone, but it is the one series on this dashboard currently moving the wrong way: it bottomed at 266 bps on 21 September and has widened every session since, 46 bps in nine days. On the page's own standard that is too short a run to read as a signal, so it is logged as a watch item, not a warning.
Caveat: Credit spreads reflect market pricing of default risk. They can widen dramatically on a single news event and compress just as quickly. Sustained widening over weeks is the signal, not a single spike.
Housing Starts (annualised)
Threshold: Below 1.2M sustained typically coincides with recession.
Lead time: Leads recession by 12-24 months (one of the longest leading indicators)
Current context: Peaked at 1.8M annualised in early 2022 before the rate-hiking cycle. Starts ran at a 1,275,000 annual rate in August 2026, 2.6% below the revised July rate of 1,309,000 and 1.2% below August 2025, holding above the 1.2M mark. The composition is the more useful read and it cuts both ways: single-family starts, the economically sensitive component, rose 7.6% to 918,000, their strongest month of 2026, while building permits eased 2.7% to 1,394,000 (still 3.5% above a year earlier) and completions dropped 11.9%. The housing signal stays mixed rather than clearly weak.
Caveat: Housing is more sensitive to interest rates than the overall economy. The 2022-24 correction was driven by rate-sensitivity, not income decline - making it a more ambiguous signal in the current cycle.
Composite Model Probability (September 2026 release)
Beyond individual indicators, the most widely watched formal model is the NY Fed yield-curve model:
- NY Fed Yield Curve Model: Applies a probit model to the 10Y-3M Treasury spread, the method set out by Estrella and Mishkin in The Yield Curve as a Predictor of U.S. Recessions (New York Fed, Current Issues in Economics and Finance, 1996). As of its September 2026 release (based on data through August 2026), the model's 12-month recession probability is 13.88%, down from the series peak of 70.9% in May 2024 - which is when the deep 2023 inversion reached the end of the model's twelve-month lag.
This output should be treated as a probabilistic signal, not a prediction, and the reading is more ordinary than it looks. Against the model's own published history since 1960 the series averages 15.3% with a median of 9.1%, so roughly 14% is an unremarkable month rather than a reassuring one. Its pre-recession record is also patchy: it never reached 30% in the year before the 1973-75, 2001 or 2020 recessions, so a reading at this level is not evidence that a downturn cannot start. Professional-forecaster surveys are also published but are subjective and vary, so only the publicly computable yield-curve model is cited here.
False Signals: What to Ignore
The most dangerous false positives in recent history:
- 1998: The yield curve briefly inverted after the Russian default and LTCM crisis. No US recession followed - the economy accelerated into the dotcom boom.
- 2019: The yield curve inverted in late 2018 and through mid-2019. The 2020 recession did follow, but it was caused by COVID - an exogenous shock unrelated to the curve inversion. Many economists consider 2019 a near-miss that external factors interrupted.
- 2015-16: Manufacturing PMI contracted for 14 months, the LEI fell, and credit spreads widened on oil-price collapse. No recession was declared.
Frequently Asked Questions
What is the most reliable recession indicator?
No single indicator predicts every recession reliably, but the yield curve inversion (10Y-2Y spread) has preceded all seven post-WWII recessions, making it the most consistent leading signal. The Sahm rule has the advantage of triggering at recession onset rather than predicting it in advance, making it more reliable for real-time confirmation. Economists generally look at multiple indicators simultaneously - when three or more signal recession, the probability is substantially higher.
Do all indicators need to flash red before a recession?
No. Recessions are called when the preponderance of evidence across multiple indicators suggests a broad-based decline. In 2020, the COVID recession was declared despite most indicators only turning negative for one month because the depth was extraordinary. In 2001, the recession was declared despite the yield curve not inverting in the traditional pattern. The NBER committee weighs the indicators holistically, not mechanically.
What is the Sahm rule exactly?
The Sahm rule, created by former Federal Reserve economist Claudia Sahm and published in 2019, signals the start of a recession when the three-month moving average of the national unemployment rate rises 0.5 percentage points or more above its low during the previous 12 months. It was designed to be a real-time trigger for automatic fiscal stabiliser policies. As of the August 2026 print, the reading is -0.07 - far below the 0.50 threshold, having turned slightly negative as unemployment held at 4.1%. It has flagged every US recession since 1970 with no false positives.
Why does the yield curve invert before recessions?
When the Federal Reserve raises short-term rates to fight inflation, short-term Treasury yields rise above long-term yields. Long-term yields are anchored by expectations of future growth and inflation, which tend to fall when the Fed is tightening. The inversion signals that markets expect lower growth (and potentially Fed rate cuts) in the future. Historically, the inversion itself tightens credit conditions and slows the economy, creating a partial self-fulfilling dynamic. The lag from inversion to recession onset is typically 12-24 months.
What does the LEI measure?
The Conference Board Leading Economic Index (LEI) is a composite of 10 forward-looking indicators: average weekly manufacturing hours, initial jobless claims, new orders for consumer goods and materials, the ISM new orders index, new orders for capital goods, building permits, S&P 500 prices, the Leading Credit Index, the interest rate spread (10-year Treasury minus federal funds rate), and average consumer expectations. The index is designed to signal economic turning points 3-9 months in advance.