Validating Bermudan Swaption Prices and Greeks

Summary

The document raises a practical model validation question for Bermudan swaptions: how to determine whether computed prices and Greeks are correct. It considers comparing numerical methods such as Monte Carlo, PDEs, and trees under the same model, comparing outputs across models, and checking special cases such as zero strike. It does not include an answer or a recommended validation procedure.

These questions point to useful layers of validation: independent numerical methods can reveal implementation or discretization errors, while analytical limits and degenerate cases can test whether results behave as expected. Agreement between methods is evidence, but not proof, since methods may share assumptions or bugs; agreement across different models is even less conclusive because the models need not produce identical values. The document supplies no benchmark results, Greek-specific checks, or literature references, so it serves as a prompt for validation work rather than a completed guide.

Key ideas

• The document asks how to validate Bermudan swaption prices and sensitivities from a model implementation.
• It proposes cross-checking Monte Carlo, PDE, and tree methods under the same model.
• It also considers comparing models and testing special cases such as a zero-strike swaption.
• Agreement among methods can provide evidence of correctness but cannot establish it by itself.
• The document offers no answer, benchmark, or recommended reading list.

Tags

# Validity of Bermudan Swaption's Price/Greeks


# Validity of Bermudan Swaption's Price/Greeks












I'm implementing a lot of stochastic models on my own for fun and I'm quite puzzled about the usual procedure concerning the correctnes of Bermudan swaptions prices and greeks ? How can you tell that the price that you get from the implementation/model is correct?

Should I price the same instrument using the same model but different methods (MC, PDE, Trees) ?

Should I price the same instrument using different models and if they look similar it means that the price is correct ?

Should I try to price degenerate cases of Bermudan swaptions like bermudan swaptions with strike 0 and just see if the model behaves as expected ? I would usually go for this one but I feel that it's not enough, i may be wrong.

I tried to do some research on the case, but didn't found much even when reading some Model Risk Management books. If possible could you also please provide some litterature ?

Thank you

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.