Credit spreads (as a market signal)

Definition

Credit spreads — the yield premium of corporate bonds over comparable Treasuries, tracked via investment-grade and high-yield OAS indices — are the bond market’s continuously-priced estimate of default and liquidity risk. As a cross-asset SIGNAL they have unusual academic standing: Gilchrist-Zakrajsek (2012) showed that a purified spread measure (the “GZ spread,” and especially its excess-bond-premium component stripped of mechanical default risk) predicts economic activity better than most financial variables. For equity traders the practical form is simpler: spreads WIDENING while equities rise is among the best- documented divergence warnings; spreads confirming is regime information.

How it works / structure

• What the spread contains: expected default loss + a RISK PREMIUM for bearing it + liquidity premium; the GZ decomposition matters because the excess-bond-premium (the risk-appetite component) is what carries the predictive power — spreads forecast because they price credit-supply conditions, not just defaults.
• The standard gauges (engine-relevant): IG OAS (~1-1.5% calm, 2%+ stressed), HY OAS (~3-4% calm, ~6% recession-warning, 10%+ crisis — 2008 peaked near 20%), and their VELOCITY (100bp of HY widening in a month is regime information whatever the level); ETF proxies (HYG/LQD vs Treasuries) give the intraday read.
• The divergence pattern: credit leads or confirms equity stress more often than it lags — 2007’s spread widening preceded the equity peak (episode-gfc-2008); the documented mechanism is credit’s direct exposure to refinancing conditions, which bite before earnings do.
• Cross-checks: spreads vs regime-volatility (both price stress — disagreement is signal), vs macro-yield-curve (curve leads, spreads confirm), vs equity breadth (indicator-breadth-advance-decline deterioration plus spread widening is the compound warning).

When it applies

Regime dashboards (a spread level+velocity gauge belongs in any risk-state vector); equity-drawdown early warning (the divergence check before adding risk); recession-probability inputs (macro-business-cycle — the GZ evidence is the academic license); credit-sensitive sector analysis (banks, high-leverage cyclicals trade on the same spread).

Risk profile & failure modes

• Level-anchoring across eras: “normal” spread levels drift with index composition (HY quality improved materially post-2015, documented) — decade- old thresholds mislabel today’s regimes; use percentiles within era.
• Policy-backstop distortion: 2020’s facilities compressed spreads by announcement (macro-fed-balance-sheet) — spread signals under active backstops read policy, not credit.
• Liquidity artifacts: OAS in stressed markets partly prices the inability to trade — the signal’s crisis extremes overstate pure default expectations (the GZ liquidity caveat).
• False positives: 2011, 2015-16, and 2018 all saw recession-grade spread scares without US recessions — the signal’s base rate is asymmetric (few misses, meaningful false alarms), and sizing must reflect it.

Evidence & limits

Gilchrist-Zakrajsek (2012, AER) is the anchor evidence with a Fed-maintained live series (EBP); index histories are public. The predictive content is strongest at business-cycle horizons; day-trading granularity inherits ETF-proxy noise.

Falsifiable-thesis examples

Illustrations only, not signals:

• “HY OAS widening >150bp over 2 months while the S&P makes new highs precedes a 10%+ equity drawdown within 2 quarters (divergence thesis)” — falsified by the joint series.
• “Excess-bond-premium above its 90th percentile marks windows where forward 12-month equity returns are below unconditional average (GZ replication)” — falsified by the conditional return spread.

Cross-references

• The academic siblings: macro-yield-curve, macro-business-cycle
• The stress cross-checks: regime-volatility, opt-term-structure
• The instrument mechanics: ext-bonds-rates
• The episode record: episode-gfc-2008