2
$\begingroup$

The Bond OAS computation model used in our bank (The model was created in the 90s and the people who worked on it then are no longer part of the company) uses a fallback interest rate volatility of 27.5%. I am unable to understand the rationale behind this assumption. All I know is the model was created to align with Bloomberg’s OAS model. Why is the fallback volatility of 27.5% used?

$\endgroup$
2
$\begingroup$

Term structure models created in that era frequently used fixed volatility assumptions. These assumptions were usually based on historical realized vols (instead of implied vols from options). A paper published by Salomon Brothers in 1997 reported that the realized volatility for 3-month Treasury rate from 1977 to 1997 to be 27.3%, which might be what the modeler at your bank saw and chose.

$\endgroup$
  • $\begingroup$ Can you please share a link to that paper if you have it $\endgroup$ – Bhaskar Gudimetla Jul 18 '18 at 17:25
  • $\begingroup$ Unfortunately I do not have a digital copy of the paper. Its title is "A term structure model and the pricing of fixed-income securities." It's also available as a chapter in the book "Salmon Smith Barney Guide to Mortgage-backed and Asset-Backed Securities." I just checked the book and it reports the same values. $\endgroup$ – Helin Jul 18 '18 at 17:29
  • $\begingroup$ We are now supposed to justify the fallback 27.5% volatility and the log-normal assumption for interest rate tree construction for this OAS model for getting a model validation pass. Thanks for the help $\endgroup$ – Bhaskar Gudimetla Jul 19 '18 at 4:45
  • $\begingroup$ Is there a reason why it's still in production? Both the vol assumption and lognormal assumption are inappropriate for today's markets. $\endgroup$ – Helin Jul 19 '18 at 5:36
  • $\begingroup$ Hey, do you know any sources where I can fund better interest rate models and the methods of constructing binomial trees for OAS computations using these models. $\endgroup$ – Bhaskar Gudimetla Jul 19 '18 at 7:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.