HY OAS vs VIX: why this unusual divergence challenges the reading of market stress in 2024-2026

HY OAS and VIX showed a high rolling correlation — between 0.7 and 0.85 depending on the window — over 2003-2022. Since late 2023, that correlation has collapsed. Two competing interpretations coexist; no publicly available data lets us arbitrate between them.

The statistical anomaly deserves to be treated as an object in itself, without prematurely converting it into an operational signal. The nature of the divergence interrogates the joint reading of the two canonical market-stress indicators.

1. A Historically High and Stable Correlation

HY OAS and VIX measure, in first approximation, the same object for different purposes: the risk premium investors demand to hold risky assets. This dynamic is mapped in the current regime analysis. HY OAS captures that premium on the high-yield corporate credit segment; VIX captures it on 30-day S&P 500 options. Although the two instruments are constructed differently — one is a yield spread, the other an implied volatility — their sensitivities to the same fundamental macro factors make them historically correlated.

Rolling correlation calculations over 2003-2022 (daily FRED BAMLH0A0HYM2 and daily CBOE VIX data) yield a median coefficient around 0.75 on a 90-day window, with a typical range of 0.65 to 0.85 depending on cycle phase. This correlation stability over nearly two decades is not accidental: it reflects a market structure in which investor flows and hedging decisions treated risky assets (equities, HY credit) as a single block, with continuous arbitrage between the two segments via equity-credit positions and collar hedges.

During every major stress episode 1997-2022, this correlation even strengthened. During the GFC, the 90-day rolling correlation moved above 0.9 between August 2008 and March 2009. During the March 2020 COVID shock, it reached 0.92 on certain windows. During the 2022 inflation cycle, it stayed above 0.75 across the entire episode. The HY-VIX alignment in 2008 as contrast is the historical reference signature: both risk premia move together because flows and arbitrages link them mechanically.

2. The Correlation Collapse Since Late 2023

The correlation breakpoint sits in the autumn of 2023. From November 2023, the 90-day rolling correlation begins to decline atypically, with no identifiable shock. In January 2024, it crosses below 0.5 for the first time since 2003. Across 2024-2025, it oscillates between 0.2 and 0.5, with windows where it even turns slightly negative. This dynamic is unprecedented on the modern series.

The quantitative picture since the breakpoint is singular. HY OAS has compressed durably below 300 basis points over 2024-2026, at a level only observed across three short prior windows (1997, 2006-2007, 2017-2018). The VIX, by contrast, has not followed the same compression trajectory: after touching a low near 12 in 2024, it has oscillated episodically between 15 and 25, with occasional spikes near 30 on certain events (April 2024 tariff shock, November 2024 election, summer 2025 geopolitical tensions). Against those VIX spikes, HY OAS remained anchored in a narrow 280-320-bp range, without proportional reaction.

This divergence is analytically striking. Over the previous twenty years, a VIX at 25 typically accompanied an HY OAS of at least 450-500 bps; observing it today with HY at 300 bps is statistically aberrant relative to historical regularity. The canonical signal said: “if VIX > 20, expect HY ≥ 450.” That signal no longer holds in 2024-2026. The underlying HY-VIX dataset documents the daily series for any reader who wants to reproduce the rolling-correlation analysis independently.

3. First Interpretation — HY OAS Understates Credit Risk

The first reading, widely echoed in credit-strategy notes published since 2024, argues that HY OAS is abnormally compressed relative to what fundamentals would justify. The cause invoked is structural: the deep transformation of the HY investor base over the past fifteen years.

The CLO (Collateralized Loan Obligations) market reaches approximately $1.1 trillion in outstanding volume in 2025, up from $350 billion in 2014 (SIFMA + JP Morgan CLO Research data). CLOs primarily buy leveraged loans rather than HY bonds, but their buying pressure transmits to HY spreads via indirect arbitrage: issuers who can choose between HY issuance and loans tilt toward the cheaper instrument, aligning spreads. On the retail side, combined assets in the two main HY ETFs (HYG and JNK) exceed $35 billion in 2025, with a growing share held by retail investors via consumer platforms. These ETF flows are procyclical by construction: they grow as markets rise, compressing spreads through mechanical bond buying.

Under this reading, the CLO + retail-ETF combination creates a structural demand floor that keeps HY OAS low independent of credit fundamentals. The risk premium demanded by these structural buyers is lower than that demanded by purely active investors, because the former buy according to portfolio rules or savings flows, not according to a fundamental reading of default risk. The consequence is that published HY OAS would be artificially disconnected from the “fundamental” premium that a market of purely active investors would imply. The VIX, by contrast, does not suffer the same distortion and continues to capture the “unabsorbed” equity risk premium.

4. Second Interpretation — VIX Overstates Equity Volatility

The second reading, defended notably by certain equity strategists, argues the opposite: it is the VIX that is abnormally elevated relative to realized equity volatility. The cause invoked is different: the massive growth of the ultra-short options market (0DTE — Zero Days To Expiry) and the systematic vol selling operated by certain asset-management structures since 2020.

The 0DTE options market on indices exploded from 2022, reaching in 2024-2025 between 40% and 50% of total S&P 500 option volume according to CBOE Markets statistics. These very-short-term options have a particular behavior in terms of impact on implied volatility: the associated short-gamma flows (option-selling by market-makers, buying by end-users) inflate spot implied volatility without necessarily reflecting a corresponding increase in realized volatility of the underlying. In parallel, systematic vol selling implemented by certain strategies (covered call ETFs, defined outcome ETFs) creates a structural supply flow of volatility that reacts to spot variations with a lag, accentuating the “stop-and-go” effect of the VIX.

Under this reading, an average VIX at 17-20 in 2024-2026 does not reflect a corresponding increase in fundamental equity stress, but microstructural distortions that keep it elevated. HY OAS, by contrast, does not suffer this distortion (the HY market has no 0DTE equivalent) and continues to capture the “pure” credit premium. Under this reading, the VIX is mis-calibrated, not HY OAS.

5. Partial Test Through Observation

The two readings are by construction difficult to arbitrate ex ante. They make partially opposite predictions on the comparative behavior of the two series during point-in-time shocks — a partial but useful test.

If the structural HY compression hypothesis is correct, one should observe that point-in-time shocks move VIX but not HY OAS (the structural demand floor absorbs marginal shocks). That is broadly the pattern observed across 2024-2026: on the episodic VIX spikes cited above, HY OAS practically did not budge. This constitutes an empirical argument in favor of interpretation 1 — without definitively validating it, because the alternative scenario (VIX overstates) also predicts HY OAS stability, simply for a different reason.

The observation that would potentially arbitrate would be a shock of sufficient amplitude to simultaneously activate both markets. If HY OAS widens decisively at the same time as VIX, that would confirm the two series remain fundamentally connected and that the divergence is cyclical (structural and microstructural effects partially canceling in stress). If HY OAS stays compressed even under a significant shock, that would validate interpretation 1. This observation has not occurred in 2024-2026 — the shocks have been of limited amplitude and have not activated HY OAS.

The current inability to arbitrate between the two readings is itself a data point. It means that the joint HY-VIX reading no longer provides the same informational content as before 2023, and that a user relying on historical regularity must explicitly integrate this signal degradation. The general reading framework for HY OAS integrates this new dimension as a qualification, not as an invalidation.

6. Consequences for the Contemporary Regime Reading

The HY-VIX divergence is not an isolated phenomenon. It belongs to a broader transformation of the reaction functions of both markets, of which the absolute HY OAS compression post-2022 is the most visible expression. Reading one without the other misses the context; the two analytical objects are complementary and belong to the family of correlations across stress indicators identified by Eco3min.

The open question this divergence poses to the reader is the following: if the two canonical market-stress indicators are no longer correlated, which one to favor? The honest answer is that neither taken in isolation delivers a sufficient signal. The analytical reading requires observing both simultaneously, comparing their reactions to point-in-time shocks, and integrating the divergence as information about the state of the market — not as a nuisance to correct by picking one over the other.

Last updated — 28 May 2026