Corporate bonds do not trade in isolation. The credit risk embedded in a bond OAS is simultaneously priced in the CDS market, partially reflected in sector equity performance, and influenced by the broader interest rate environment through swap spreads and Treasury curve dynamics. When the bond itself has not traded recently, these other markets continue pricing the same underlying risk exposure. The question for a fair value model is which of those cross-asset signals carry enough incremental, non-redundant information to improve the fair value estimate, and at what cost in terms of model complexity and data reliability.
CDS Spreads: The Primary Cross-Asset Input
Single-name credit default swaps are the most direct cross-asset signal for corporate bond fair value. Both instruments price the same fundamental credit risk: the probability of default and expected recovery for a specific obligor. The theoretical relationship between CDS spreads and bond OAS is well established, though the observed basis between them is non-trivial and time-varying.
The CDS-bond basis (typically defined as CDS spread minus bond OAS for the same obligor at comparable maturity) has historically run in the range of 10 to 40bp in normal conditions, with the sign and magnitude depending on funding conditions, supply-demand dynamics in the bond market, and the cheapest-to-deliver option embedded in the CDS contract. During stress periods, the basis can widen substantially or even flip sign as different parts of the credit market behave differently under pressure.
In our model, we use single-name CDS spreads as a continuous reference for bonds where a liquid CDS market exists for the same reference entity. We do not assume a fixed basis relationship. Instead, we estimate a time-varying basis using the most recent available bond prints and CDS observations, updating the basis estimate as new data arrives. When no recent bond prints exist, the basis estimate relies on a sector-level median basis adjusted for the specific obligor's rating and maturity profile.
For bond universes where single-name CDS is unavailable (many mid-sized IG issuers do not have active CDS markets), we substitute index-level CDS signals from CDX IG sub-index constituents at the relevant rating and sector, with an additional basis adjustment for the non-CDS issuer's historical spread relationship to the index.
Index ETF Flows: A Technicals Signal
The growth of passive fixed-income investing has created a substantial channel through which index rebalancing and ETF creation-redemption activity affects individual bond spreads. When large ETF flows into IG credit push authorized participants to create units, they need to acquire bonds from the underlying index. The buying pressure concentrates on index-eligible bonds across all included CUSIPs, which can tighten spreads across the index even absent any fundamental credit news.
We track estimated ETF flow data for the major IG and HY credit ETFs as a technicals adjustment factor. Large positive flows signal buying pressure that may tighten spreads beyond fundamental fair value; large outflows signal the reverse. The signal is most relevant for bonds that are included in major indices, and it has a short-horizon persistence of roughly 1 to 3 days before price impact dissipates as the market absorbs the flow.
The ETF flow signal has meaningful incremental content, but it is genuinely a technicals signal rather than a fundamentals signal. A bond that tightens because of ETF inflows is priced beyond fundamental fair value during the flow period. We incorporate this as a spread adjustment with a specific decay schedule rather than as a permanent fair value component. The practical implication: when strong ETF inflows are pushing spreads tighter across the index, our fair value estimates will be slightly wider than the observed market, reflecting the view that the technicals-driven tightening is temporary.
Interest Rate Dynamics and Swap Spreads
The OAS calculation by definition strips out the Treasury rate contribution to bond yield, so the interest rate level itself does not directly enter OAS-denominated fair value. But interest rate volatility and swap spread dynamics do affect credit spreads through several indirect channels.
First, elevated interest rate volatility raises the cost of carry on fixed-income inventories, which reduces dealer willingness to hold positions and widens bid-ask spreads in the bond market. Higher rate vol is associated with wider credit spreads not through any fundamental credit quality change but through the mechanics of how the market is made. We include a rate volatility term (proxied by MOVE index level relative to trailing average) in our regime score calculation.
Second, swap spreads have a direct relationship with investment-grade bond OAS through the asset swap mechanism. A bond that looks attractive relative to LIBOR/SOFR plus spread will attract asset swap demand; a bond that looks expensive on the same basis will see that trade unwind. The swap spread level therefore anchors the OAS level of IG bonds through arbitrage. When Treasury supply increases and swap spreads tighten (as frequently happens during periods of heavy government issuance), IG OAS tends to widen to maintain the absolute yield pickup over SOFR, even absent any credit deterioration.
Sector Equity as a Direction Indicator
For non-financial IG credits, equity performance of the same sector provides a coarse but real-time signal about the market's current assessment of sector-level credit quality. When sector equities are selling off, it reflects a broad deterioration in investor sentiment toward that sector, which typically precedes or accompanies credit spread widening. The relationship is not tight at the individual issuer level, but at the sector median level it is consistent enough to contribute incremental signal.
We use rolling 5-day sector equity return, compared against the broad equity market return over the same window, as a relative performance measure. A sector underperforming the market on equity suggests credit spread widening pressure specific to that sector. We weight this signal quite lightly for any individual bond because the bond-equity correlation at the single-issuer level is noisy, but it contributes meaningfully at the sector-level input to our cross-sectional comparables weighting.
For financial sector issuers (banks, insurers, broker-dealers), the relationship between equity performance and bond spreads is more direct and better documented. Bank equity performance correlates more tightly with bank subordinated debt spreads than industrial equity performance correlates with industrial senior debt spreads. We treat financials as a separate calibration group and apply higher sector equity weights for that sub-universe.
What Cross-Asset Signals Cannot Replace
The breadth of cross-asset inputs in the model can create an impression that the bond market itself is almost optional for fair value estimation. That impression is wrong. The CDS market, ETF flows, swap spreads, and sector equities all provide continuous pricing information, but they price credit risk in the aggregate. They do not capture the specific supply-demand dynamics of a specific CUSIP on a specific date.
A bond that is cheap on cross-asset comparisons may be cheap for a reason specific to that issue: an overhang from a large upcoming call, a position held concentrated in few hands with a motivated seller, a covenant structure that makes it less attractive as collateral. These idiosyncratic factors will show up in the bond market's trade prints and dealer axes, not in the cross-asset signals. The cross-asset layer provides the macro and sector-level anchor; the bond-specific inputs provide the idiosyncratic adjustment.
In our internal validation framework, removing cross-asset signals and running the model on bond-specific data alone degraded performance particularly for bonds with no recent prints, as expected. But the reverse, running only cross-asset signals without any bond-specific inputs, showed comparable or larger degradation. Both legs contribute; neither is sufficient alone. That is precisely the design intent: a model that uses all available information rather than privileging any one source.