Choosing and Calibrating an Interest Rate Model

Summary

The answer outlines the main choices involved in building an interest rate model. A model may describe short rates, forward rates, or more complex rates such as swap rates. The appropriate rate representation and model depend on the intended application, the asset class, and assumptions that must remain compatible with other parts of the framework. Models commonly specify rate dynamics with stochastic differential equations and parameters that must be calibrated to observed data.

Calibration is identified as a particularly difficult step because it requires both theoretical understanding and practical computation. The response says there is no single regulation prescribing one model, but models should satisfy properties suited to their use. Positivity is given as an example of a desired property that became less compelling when negative rates were observed. The answer is intentionally broad and does not provide a calibration recipe or a definitive list of required model properties.

Key ideas

• Choose whether to model short rates, forward rates, or another rate relevant to the application.
• Model choice depends on the use case, asset class, and compatible assumptions.
• Interest rate dynamics are often expressed with stochastic differential equations.
• Calibrating model parameters to real market data can be theoretically and computationally demanding.
• Desired model properties depend on application, and observed negative rates challenge a universal positivity requirement.

Tags

# How to develop your own interest rate model?


# How to develop your own interest rate model?












How can you develop your own interest rate model? What must be take cautiously before making one? Also what is the regulation that one mustfollow? Also what is some common properties that models would be satisfying?

## Answer by Lucas Morin (score 4, accepted)

https://quant.stackexchange.com/a/14491

There is a lot of different kind of IR models. You can modelize short rates, forward rates or more complex rates like Swap rates. Once you chose a rate and a model that fit your needs (some are easier to use for a given application or asset class, some models are incompatible with some hypothesis or other models). The model will usually take the form of a Stochastic differential equation, wich give the dynamic of your rate(s). The SDE will contain some parameters. You then have to find the parameters that represent the real life (calibration process). This is probably the hardest part as it require both theoretical and practical computations.

There is no regulation on that, just properties that your model should respect or not based on what you need. But this is not written in stone. For exemple some models were heavily criticised because they could become negative, people preferred positive models wich would be more reallistic because everybody know rates could not be negative. Some rates recently become negative...

This is a very broad answer as I can't sum up decades of works of thousands people in one post. You should really read a courses on IR models. Look for the slides of Brigo and/or Mercurio courses.

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.