Interpreting Model-Implied Corporate Bond Spreads and Default Probabilities

Summary

The note explains why a structural credit model’s implied spread may differ from the observed corporate bond spread. Observed spreads include compensation beyond expected default loss, such as liquidity effects, so a model focused on default risk can produce a narrower spread than the market quote. It cites a study reporting a range for the default component of corporate spreads, but gives no calibration details for a particular bond or model.

Default probabilities inferred from credit spreads or CDS prices are risk-neutral probabilities. They incorporate market pricing of risk and should not be read as objective forecasts of default frequency. Separating the expected-loss component from the credit risk premium is difficult. If a model produces a spread above the observed spread, the response treats calibration or model setup as a likely issue, but does not diagnose a specific case.

Key ideas

• Observed corporate spreads can include liquidity and other components beyond default compensation.
• A model focused on default risk may therefore imply a smaller spread than the market spread.
• CDS spreads can be used to infer default probabilities, but the result is risk-neutral.
• Risk-neutral implied default probabilities are not the same as objective default probabilities.
• Separating default loss compensation from the credit risk premium is difficult.

Tags

# Model-implied yield spread on corporate bonds


# Model-implied yield spread on corporate bonds












While using Merton (or any other) model, is the model-implied yield spread on bonds greater than actual yield spread? And is it possible to estimate actual probablities of default?

## Answer by Quartz (score 1)

https://quant.stackexchange.com/a/4728

First some remarks in a model agnostic sense. Credit spread is actually thought to be significantly smaller than the actual one (some even say about 1/3 only, for corporates Longstaff 2005 reports the range 5%-25% for the default component), as there are many other factors such as liquidity. For PDs it might be slightly better to use CDS spreads. In any case what you'll get are risk neutral implied PDs, not objective ones; disentangling the "actuarial" credit spread from the proper credit premium is not easy (and beware that many improperly call "premium" the full risk neutral credit spread).

If a model returns a higher spread than the actual one, something has probably gone wrong. What are you calibrating to?

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.