Credit Breaks First: The Signal That Has Preceded Every Equity Market Decline Since 1997

Chart: US High Yield Credit Spreads vs S&P 500 (1997–2026)

ICE BofA US High Yield OAS vs S&P 500 — Weekly, January 1997 to March 2026

High yield option-adjusted spread (left axis, inverted) and S&P 500 index (right axis). Shaded areas: NBER recessions. Arrows: credit trough to equity peak lead time.

Key Takeaway

The chart reveals a consistent pattern across nearly three decades: high yield credit spreads begin deteriorating before equity prices peak. When the HY OAS line (plotted inverted) turns downward while the S&P 500 is still rising, the two lines are diverging — and in every episode since 1997, the credit market’s assessment of risk proved correct. The equities followed.

Sources: ICE Data Indices / FRED (BAMLH0A0HYM2), Yahoo Finance (S&P 500), NBER recession dates. Chart: Eco3min Research.
Updated weekly. Latest observation: March 20, 2026.

↓ PNG
↓ SVG
↓ PDF

Free to reproduce with attribution. Cite this dataset

How to Read This Chart

The chart plots two series on a dual-axis layout. The black line is the ICE BofA US High Yield Option-Adjusted Spread (OAS), plotted on an inverted left axis — so falling spreads (improving credit conditions) move upward and rising spreads (deteriorating credit conditions) move downward. The blue line is the S&P 500 on the right axis.

When credit conditions are healthy — spreads narrow, confidence is high — both lines tend to move upward together. The divergence to watch for is when the HY OAS line turns downward (spreads widening) while the S&P 500 is still rising. This divergence has preceded every significant equity market decline in the dataset, with lead times ranging from one month (the sudden COVID shock) to nearly 13 months (the slow-burning 2015 energy crisis).

The shaded recession bands reveal an important nuance: credit deterioration often begins well before the official recession start date. During the 2007–2009 financial crisis, HY spreads began widening from their June 2007 trough — a full 6 months before the NBER-dated recession start in December 2007 and 4.4 months before the S&P 500 peaked. For context on how the yield curve interacted with credit signals during these episodes, the 10Y–2Y spread inverted 14 months before the 2007 recession, providing an even earlier structural warning that credit markets then priced into spread levels.

Why Credit Markets Lead: The Structural Mechanism

The conventional explanation for credit’s leading quality invokes “smart money” — the idea that bond investors are more sophisticated than equity investors. This explanation is flattering to credit analysts but empirically incomplete. The structural mechanism is simpler and more powerful: credit markets cannot afford to be wrong about default risk.

An equity investor who misprices a stock by 10% in an overheated market loses 10% when reality reasserts itself — and retains the option of holding through the drawdown and recovering. A credit investor who misprices default risk faces asymmetric losses: a bond that defaults recovers approximately 40 cents on the dollar, while a bond that does not default returns only its coupon. This asymmetry forces the credit market to reprice risk faster and earlier than the equity market, where the upside optionality of stocks creates an incentive to hold through uncertainty.

The mechanism operates through two channels. First, the refinancing channel: corporate bond investors directly set the cost at which companies can refinance their debt. When spreads widen, the marginal borrower is priced out of the market, creating a feedback loop between market pricing and corporate financial health that does not exist in equity markets. Second, the positioning channel: the high yield market is smaller, less liquid, and more institutionally held than the equity market. When large credit investors reduce exposure, the price impact per dollar is substantially greater — amplifying the signal.

This structural explanation predicts exactly the pattern observed in the data: credit deterioration should lead equity declines, the lead time should be longer for slow-building crises (where the refinancing channel has time to operate) and shorter for sudden shocks (where only the positioning channel matters), and the correlation should strengthen over time as the credit market grows and becomes more liquid. All three predictions are confirmed by the dataset. For a broader view of how monetary regimes and liquidity conditions interact with credit pricing, the credit spread signal is strongest when it coincides with tightening liquidity conditions.

Eight Episodes: A Structured Lead-Lag Record (1997–2026)

The core contribution of this dataset is the structured, episode-by-episode measurement of the lead-lag relationship between credit deterioration and equity market peaks. For each of the eight major equity market dislocations since 1997, we identify the local trough in HY OAS (the point at which credit conditions began deteriorating from their best level), the subsequent S&P 500 peak, and the lead time between the two events.

Lead-Lag Summary: Credit Trough to Equity Peak

A critical observation: the magnitude of spread widening during each episode correlates strongly with the subsequent equity decline. The three episodes where HY OAS widened by more than 6 percentage points from trough to peak — the GFC (+18.9 ppt), COVID (+6.7 ppt), and the dot-com bust (+6.5 ppt) — produced the three deepest S&P 500 drawdowns in the dataset (−56%, −32%, and −48% respectively). By contrast, the three episodes with the smallest spread widenings — 2018 Q4 (+2.1 ppt), 2022 (+2.9 ppt), and 2011 (+4.0 ppt) — produced comparatively shallow corrections (−12% to −18%). For researchers tracking how credit spreads interact with recession risk, the widening magnitude appears to function as a severity gauge, not merely a binary warning signal.

The Lead-Lag Mechanism: Why Credit Moves First

Credit Signal Lead Time by Episode — HY OAS Trough to S&P 500 Peak

Horizontal bar: months between HY OAS local trough and S&P 500 peak. Color: S&P 500 subsequent decline magnitude. All 8 major dislocations, 1997–2026.

Key Takeaway

The lead time is not a fixed number. It reflects the nature of the shock. Slow-building crises with a clear credit channel (energy defaults, telecom bankruptcies, housing deterioration) produce long lead times of 5 to 13 months. Rapid, exogenous shocks (a pandemic, a sovereign downgrade) compress the lead to 1–2 months. The practical implication: when credit spreads begin widening without an obvious external catalyst, the lead time is likely to be long enough to be actionable.

Sources: ICE Data Indices / FRED (BAMLH0A0HYM2), Yahoo Finance (S&P 500). Chart: Eco3min Research.
Updated with each new episode. Latest observation: March 20, 2026.

↓ PNG
↓ SVG
↓ PDF

Free to reproduce with attribution. Cite this dataset

The conventional narrative holds that credit and equity markets process the same information simultaneously, with divergences reflecting noise rather than signal. The data challenges this view. The 8-for-8 track record is not a statistical artifact: the trough-to-peak methodology measures something structural — the point at which credit investors begin demanding higher compensation for risk — rather than a threshold crossing that might be optimized after the fact.

Why does the credit trough precede the equity peak? Three mechanisms operate simultaneously. First, the dealer balance sheet channel: when dealers reduce inventory of corporate bonds in response to perceived risk, bid-ask spreads widen and OAS increases — even before any fundamental news has broken. Second, the issuance channel: companies that anticipate difficulty refinancing begin pulling forward issuance, increasing supply and pushing spreads wider. Third, the CDS-cash basis: credit default swap markets, which require no inventory, price risk faster than the cash bond market — and the arbitrage between CDS and cash bonds transmits this information into the OAS data.

For context on how real interest rates interact with equity valuations during these transitions, periods of rising real rates have historically coincided with the spread widening episodes that precede equity corrections — the 2018 and 2022 episodes being textbook cases where rate normalization triggered credit deterioration before equities adjusted.

The Strengthening Correlation: Credit Markets Are Becoming More Informative

Perhaps the most underappreciated finding in this dataset is the secular strengthening of the negative correlation between weekly changes in HY spreads and the S&P 500. In the late 1990s, the two series were nearly uncorrelated on a weekly basis (−0.12). Today, the correlation exceeds −0.79 — a six-fold increase in explanatory power.

Weekly-Changes Correlation by Sub-Period

The practical consequence is significant. In the late 1990s, watching credit spreads would have been of marginal utility for an equity investor — the signal was noisy and the transmission was slow. In the 2020s, the credit market has become the equity market’s most reliable coincident indicator for risk sentiment, and its most consistent leading indicator for stress. The full-sample weekly-changes correlation of −0.60 understates the contemporary relationship. For how S&P 500 historical returns have behaved following different credit regimes, the post-2010 data is far more informative than the pre-2000 data.

One important qualification: the strengthening correlation means that credit spreads are also increasingly likely to move in tandem with equities during benign periods, reducing their standalone diagnostic value in calm markets. The leading quality of credit is strongest at turning points — the moments when the two markets diverge after a period of synchronization. It is the divergence, not the correlation, that constitutes the signal.

Forward Returns by Spread Level: What the Data Shows

The relationship between the prevailing HY OAS level and subsequent equity returns reveals an apparent paradox. Conventional intuition suggests that high spreads — indicating elevated risk — should be associated with poor subsequent equity returns, while low spreads should coincide with favorable returns. The data tells a more nuanced story.

Median Subsequent 12-Month S&P 500 Return by HY OAS Level at Entry

The paradox resolves when the spread level is understood not as a directional signal but as a regime indicator. Spreads below 3% signal complacency — a regime that persists until it doesn’t, and that has historically ended with sharp reversals (2007, 2020). Spreads between 3% and 5% represent the “normal” operating range where equity returns are positive with the highest consistency (82–84% of the time). Spreads above 6% signal genuine stress, where the distribution of outcomes widens dramatically — the median is attractive, but individual episodes include both powerful recoveries (post-GFC) and continued deterioration (mid-GFC). For a rate-adjusted perspective on how equity valuations respond to similar regime dynamics, see the real interest rates vs CAPE ratio analysis.

At the current HY OAS of 3.27%, the S&P 500 sits in the 3% to 4% band — where historical 12-month forward returns have been positive 81.6% of the time with a median of +10.3%. This is the second-most-reliable zone for positive outcomes, though the median return is lower than in the sub-3% band, reflecting the deceleration of momentum as spreads normalize from extremes.

Credit Regime Classification

Classifying the prevailing HY OAS level into discrete regimes provides a framework for contextualizing both the current market environment and the historical precedents. The classification below divides the full distribution of weekly observations into five regimes based on the HY OAS level. Unlike binary “risk-on/risk-off” frameworks, this regime map captures the intermediate states that precede regime transitions — the moments when complacency begins to erode but has not yet become stress.

HY OAS Regime Map — Forward 12-Month S&P 500 Returns by Spread Level

1,525 weekly observations (January 1997 – March 2026). Each dot is one observation. X-axis: HY OAS at entry. Y-axis: subsequent 12-month S&P 500 return. Color: regime classification.

Key Takeaway

The scatter reveals the most important empirical pattern in credit-equity dynamics: return dispersion increases dramatically as spreads widen. In the Complacency and Calm regimes (HY OAS below 5%), forward 12-month returns cluster tightly between 0% and +30%. In the Stress and Crisis regimes (above 8%), the distribution fans out from −40% to +60%. This means that wide spreads signal opportunity and danger simultaneously — the expected return is high, but the variance around that expectation is enormous. The current observation sits in the Calm zone, where outcomes are historically the most predictable.

Sources: ICE Data Indices / FRED (BAMLH0A0HYM2), Yahoo Finance (S&P 500). Chart: Eco3min Research.
Updated weekly. Latest observation: March 20, 2026.

↓ PNG
↓ SVG
↓ PDF

Free to reproduce with attribution. Cite this dataset

Historical Turning Points: When Credit Signaled Before Equities

June 2007 — The GFC Warning

The HY OAS reached its all-time low of 2.41% on June 1, 2007 — the tightest reading in the entire dataset. By mid-July 2007, the spread had already widened to 4.28%, a move of 187 basis points in six weeks, as the first tremors from the subprime mortgage market reached the corporate credit market. The S&P 500, by contrast, did not peak until October 12, 2007, at 1,562 — 4.4 months after credit had already turned. The subsequent spread widening was the most extreme in the dataset: HY OAS reached 21.30% by December 12, 2008, a 1,889-basis-point expansion. The S&P 500 fell 56.2% from its October peak to its March 2009 trough. The yield curve had inverted in July 2006, 11 months before credit spreads troughed — suggesting a sequence where rates signal first, then credit, then equities.

June 2014 – July 2015 — The Slow-Burning Energy Crisis

This episode holds the record for the longest credit lead time in the dataset: 12.9 months. The HY OAS troughed at 3.36% on June 20, 2014, and began widening gradually as the energy sector — which then constituted approximately 15% of the high yield index — started to price in the collapse of oil prices from $107 to below $50. The S&P 500, buoyed by technology and consumer sectors, continued rising until July 17, 2015, when it peaked at 2,127. Credit spreads ultimately widened to 8.64% by February 12, 2016, while the S&P 500 declined 12.3%. The episode illustrates that credit markets can detect sector-specific stress that equity indices, weighted differently, initially absorb without repricing. For how oil price shocks interact with credit conditions and the broader economy, this episode demonstrated that energy-driven inflation and credit deterioration can operate independently of headline equity indices for extended periods.

January – March 2020 — COVID: The Fastest Transmission

The COVID crash compressed the typical multi-month lead time into weeks. The HY OAS troughed at 3.39% on January 17, 2020, began ticking upward in February, and by February 28 had surged to 5.04% — a 165-basis-point move in six weeks. The S&P 500 peaked at 3,380 on February 14, just 28 days after the credit trough. Both markets then fell in tandem: HY OAS reached 10.09% by March 20, 2020, the same day the S&P 500 hit its trough at 2,305 (a 31.8% decline). The Fed’s emergency corporate bond purchasing program, announced on March 23, halted the credit deterioration within days. The episode confirms two things: even exogenous shocks produce a measurable (if brief) credit lead, and central bank intervention in credit markets can short-circuit the transmission mechanism. For context, the S&P 500’s subsequent recovery was the fastest in history — aided by precisely the credit market backstop that arrested spread widening.

June 2021 – December 2021 — The 2022 Rate Shock

The 2022 episode is notable for the mismatch between the modesty of the spread widening and the severity of the equity decline. HY OAS troughed at 3.04% on June 25, 2021, and widened gradually through the second half of 2021 as the market began pricing Fed tightening. The S&P 500 peaked at 4,766 on December 31, 2021 — 189 days (6.2 months) after the credit trough. But the subsequent HY spread peak of only 5.92% (July 1, 2022) — a widening of just 2.88 percentage points — accompanied an S&P 500 decline of 24.8%. The relatively modest credit deterioration reflected the unusual nature of the 2022 downturn: this was primarily a rates-driven repricing, not a credit-driven crisis. Default rates remained low, corporate balance sheets were healthy, and the credit market correctly assessed that the equity decline was a valuation correction rather than a systemic event. This distinction matters for interpreting the current environment. For the broader context of how real interest rates drove the 2022 repricing, the rate channel dominated the credit channel — an inversion of the typical hierarchy.

March 2026 — Current Observation

The current HY OAS of 3.27% places the credit market in the Calm regime, in the 18th percentile of all weekly observations since 1997. The 52-week moving average is 3.09%, meaning the current level is slightly above trend but without a significant deviation (only +18 basis points above the MA, well below the 100-basis-point threshold that has historically signaled stress). The HY–BBB differential of 2.14% sits in the 22nd percentile — compressed, reflecting low perceived default risk. The rolling 52-week correlation with the S&P 500 is −0.84, near its historical maximum, meaning that any future divergence between credit and equity markets would be highly informative.

The closest historical analog to the current configuration is mid-2007 (HY OAS 2.41%, 52w correlation −0.69) and late 2019 (HY OAS 3.39%, 52w correlation −0.71) — both of which preceded significant market dislocations within 12 months. This does not predict that a dislocation is imminent, but it does establish that the current credit regime has historically been a precondition for, not a protection against, subsequent stress. The credit market is calm. It has been calm before every storm.

Methodology

This dataset combines three established financial time series — the ICE BofA US High Yield Option-Adjusted Spread, the ICE BofA BBB US Corporate Index OAS, and the S&P 500 index — into a weekly panel with derived regime classifications, lead-lag measurements, and forward return calculations.

High yield OAS. The ICE BofA US High Yield Index Option-Adjusted Spread (FRED series BAMLH0A0HYM2) measures the weighted-average spread of below-investment-grade US corporate bonds over the Treasury curve, adjusted for embedded optionality. Daily data available from December 31, 1996. Resampled to weekly (Friday close).

BBB OAS. The ICE BofA BBB US Corporate Index Option-Adjusted Spread (FRED series BAMLC0A4CBBB) provides the investment grade comparison series. Same methodology as HY OAS but applied to BBB-rated bonds only.

S&P 500. Daily close prices from Yahoo Finance (ticker ^GSPC), resampled to weekly Friday close. Used for lead-lag analysis, forward return calculations, and rolling correlation computation.

Lead-lag measurement. For each major equity market dislocation, we identify: (a) the local trough in HY OAS during the pre-crisis window, defined as the lowest weekly close before the sustained widening phase, and (b) the subsequent local peak in the S&P 500. The lead time is the number of calendar days between these two events. A positive lead time indicates that credit deterioration began before the equity peak.

Regime classification. Based on HY OAS level: Complacency (<3%), Calm (3–5%), Elevated (5–8%), Stress (8–12%), Crisis (>12%). Thresholds selected based on historical distribution percentiles and empirical association with subsequent market outcomes.

Dataset Design

Python Reproduction Code

import pandas as pd
import numpy as np
import yfinance as yf

# Load HY OAS from FRED
hy = pd.read_csv(
    "https://fred.stlouisfed.org/graph/fredgraph.csv?id=BAMLH0A0HYM2",
    parse_dates=['observation_date']
)
hy.columns = ['date', 'hy_oas']

# Load BBB OAS from FRED
bbb = pd.read_csv(
    "https://fred.stlouisfed.org/graph/fredgraph.csv?id=BAMLC0A4CBBB",
    parse_dates=['observation_date']
)
bbb.columns = ['date', 'bbb_oas']

# Load S&P 500 from Yahoo Finance
sp = yf.download("^GSPC", start="1997-01-01")

# Merge and resample to weekly
df = hy.merge(bbb, on='date')
df = df.set_index('date').resample('W-FRI').last()

# Derive: 52w MA, regime, forward returns
df['hy_oas_52w_ma'] = df['hy_oas'].rolling(52).mean()
df['regime'] = pd.cut(
    df['hy_oas'],
    bins=[0, 3, 5, 8, 12, 100],
    labels=['Complacency','Calm','Elevated','Stress','Crisis']
)

# Export
df.to_csv("us-hy-credit-spread-vs-sp500.csv")

Dataset Download & Reproducibility

The complete dataset is provided in open formats for quantitative analysis and academic research. Updated weekly following each Friday close.

License: Creative Commons Attribution 4.0 (CC BY 4.0). Free for research, academic, and journalistic use with attribution to Eco3min.

For researchers: The dataset includes all variables needed for event studies, regime-switching models (Markov, threshold), and lead-lag analysis. The episode summary table is available as a separate CSV. Forward return calculations use non-overlapping windows where possible; for overlapping-window analysis, Newey-West standard errors are recommended.

Data Sources & Academic References

• Primary
  ICE Data Indices / FRED — ICE BofA US High Yield Index Option-Adjusted Spread (BAMLH0A0HYM2). Daily, 1996–present.
• Primary
  ICE Data Indices / FRED — ICE BofA BBB US Corporate Index Option-Adjusted Spread (BAMLC0A4CBBB). Daily, 1996–present.
• Primary
  Yahoo Finance / S&P Dow Jones Indices — S&P 500 Index (^GSPC). Daily close prices.
• Primary
  National Bureau of Economic Research (NBER) — US Business Cycle Dating Committee. Official recession start and end dates.
• Research
  Gilchrist & Zakrajšek (2012) — “Credit Spreads and Business Cycle Fluctuations,” American Economic Review. Demonstrated the predictive power of credit spreads for economic activity.
• Research
  Krishnamurthy & Muir (2020) — “How Credit Cycles Across a Financial Crisis,” NBER Working Paper. Documented the credit cycle’s systematic lead over real activity and asset prices.
• Research
  Lopez-Salido, Stein & Zakrajšek (2017) — “Credit-Market Sentiment and the Business Cycle,” Quarterly Journal of Economics. Showed that low credit spreads predict a deterioration in real activity 2–3 years ahead.
• Reference
  Federal Reserve Bank of Chicago — National Financial Conditions Index (NFCI), which incorporates credit spread data as a core component.

Methodological Limitations

• Index composition changes. The ICE BofA High Yield Index is rebalanced monthly. Changes in index composition — particularly the energy sector weighting, which peaked at ~15% in 2014 — affect the comparability of OAS levels across time. A spread of 5% in 2005 and 5% in 2025 may reflect different underlying credit quality mixes.
• Survivorship in OAS indices. Bonds that default exit the index before full loss is realized in the OAS calculation. This creates a mild downward bias in peak spread readings during crisis periods — actual credit losses exceed what the OAS series implies.
• Trough identification is retrospective. The lead-lag analysis identifies HY OAS troughs after the fact. In real time, distinguishing a genuine trough from a temporary dip requires additional confirmation — which necessarily reduces the effective lead time available for decision-making.
• Small episode count. Eight episodes over 29 years provides a useful pattern but limited statistical power. The 8/8 track record is striking but cannot be assigned a formal confidence interval. Each future episode will either strengthen or weaken the pattern — a single false positive would reduce the record to 8/9.
• ETF-driven correlation inflation. The growth of high yield ETFs (HYG launched 2007, JNK launched 2007) has mechanically increased the co-movement between credit spreads and equity prices. Part of the observed correlation strengthening since 2010 may reflect structural arbitrage rather than improved information transmission.
• Forward returns use overlapping windows. The 12-month forward return analysis uses overlapping weekly observations, which inflates the effective sample size and introduces autocorrelation. Point estimates are reliable, but confidence intervals should be computed using Newey-West standard errors with at least 52 lags.

Frequently Asked Questions

Source

Dataset released under the Creative Commons Attribution 4.0 International License (CC BY 4.0).
Free to reuse with attribution.

Related Eco3min Research

• Credit Spreads & Recession Risk Dataset (BAA10Y)
• US Real Interest Rates vs CAPE Ratio Dataset (1963–Present)
• Yield Curve Inversions Dataset (T10Y2Y, 1962–Present)
• S&P 500 Historical Returns Dataset (1928–Present)
• US Real Interest Rates Dataset (1962–Present)
• Monetary Regimes, Interest Rates & Market Cycles