Calibrating Stochastic Intensity Models to Credit Spreads

Summary

The document raises a credit-risk modeling question about calibrating a stochastic default-intensity model. The author considers the Cox–Ingersoll–Ross process because its intensity is nonnegative when the model parameters satisfy the Feller condition. They say they understand calibration to credit default swap quotes and ask whether risky-bond credit spreads can also be used, seeking practical guidance or references.

The text does not provide a calibration procedure, equations linking bond spreads to intensity parameters, or evidence from a worked example. It therefore frames the modeling problem rather than answering it. Any bond-based calibration would need to account for how observed spreads relate to default risk and for other influences on bond prices; those details are not addressed here. The document is useful as a prompt on extending intensity-model calibration beyond CDS quotes, but it cannot establish that the proposed process or a particular fitting method is suitable in a given application.

Key ideas

• The author proposes a CIR process for stochastic default intensity because it can remain nonnegative under the Feller condition.
• The stated calibration target is CDS quotes, with risky-bond credit spreads raised as an additional possibility.
• The document offers no bond calibration method, pricing equations, or empirical results.
• A practical bond calibration would require examining how credit spreads relate to default intensity, a topic left open here.

Tags

# Calibration of intensity model


# Calibration of intensity model












I could use some advice on calibration of stochastic intensity models. I am thinking that the CIR model is most suitable, as it can not take negative values (when feller condition is satisfied).

I am fairly sure I understand how to calibrate the model to CDS quotes. Is it possible to calibrate to for instance credit spreads on risky bonds? If someone could give a couple of hints on how this can be achieved, or has a good article where they do, so please share :-).

Thanks in advance.

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.