Why SABR Beta Is Often Fixed Before Calibrating the Smile

Summary

The note explains why practitioners often treat the SABR beta parameter as fixed rather than estimating all model parameters together. Beta shapes the underlying forward-rate dynamics: beta equal to one corresponds to a lognormal specification, zero to a normal specification, and one half to a square-root-style process. Fixing beta first reduces the calibration problem to fitting the remaining parameters and can reflect assumptions chosen before calibration.

The document gives one half as a common convention and notes that zero may be a more natural choice for yen rates in a negative-rate setting. It says beta can instead be selected by fitting calibration instruments or informed by historical observations. These choices embody modeling judgment, so the note does not establish one universally correct beta. It raises potential implications for independent model review and price verification, but does not explain specific controls or quantify the effects of alternative beta choices.

Key ideas

• Beta determines whether SABR uses lognormal, normal, or an intermediate underlying-rate dynamic.
• Practitioners often fix beta before calibrating the other SABR parameters.
• A value of one half is a common convention, while zero may suit yen rates with negative values.
• Beta may also be chosen through instrument fit or historical observations.
• The appropriate choice depends on modeling assumptions and can affect review and valuation processes.

Tags

# SABR Question: Why does the market take the beta parameter as a constant?


# SABR Question: Why does the market take the beta parameter as a constant?












SABR Question

Why does the market take the $\beta$ parameter as a "constant"?

- I see most brokers quoting SABR parameters nowadays.

- I've seen many banks use $\beta$=0.5 as a rule.

- I've seen quants select a $\beta$ based on best fit to calibration instruments.

What is most correct to the spirit of the paper, and explain any issues to anticipate with IMR/IPV processes.

## Answer by AKdemy (score 3, accepted)

https://quant.stackexchange.com/a/63749

Managing Smile risk from Hagan et. al.

Generally if you pre-select $\beta$, it is from a priori considerations.

- $\beta = 1$ corresponds to stochastic lognormal

- $\beta = 0$ is stochastic normal

- $\beta=1/2$ CIR

In the SABR model, beta is usually calibrated first, followed by the other 3 parameters. Frequently, instead of calibrating beta, it is simply assumed to have $\beta=1/2$ (since CIR is widely used). That said, JPY it is also natural to select 0 for JPY due to negative rates. The paper explains both. "Aesthetic" consideration (a priori - fixed) or determined form historical observations.

Click here for an intuitive explanation of the SABR model. How to estimate $\beta$ can be seen in this answer.

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.