How SABR Calibration Frequency Varies by Parameter

Summary

Calibration frequency for stochastic volatility models depends on the model and market, rather than following one fixed schedule. The answer illustrates this with SABR, separating parameters according to how quickly they are typically updated.

The initial volatility parameter α may be recalibrated frequently, even intraday, because at-the-money options are especially important to it. The backbone parameter β is often fixed in advance. The correlation ρ and volatility-of-volatility ν are described as changing more slowly, with recalibration perhaps weekly or monthly depending on the market. These are practical tendencies, not universal rules: the document gives no market-specific evidence or calibration procedure, and the appropriate schedule depends on the model and market conditions.

Key ideas

• SABR calibration frequency varies by parameter rather than following one uniform schedule.
• The initial volatility parameter α may be updated frequently, including intraday, because of its connection to at-the-money options.
• The backbone parameter β is often selected in advance instead of being regularly calibrated.
• The correlation ρ and volatility-of-volatility ν may be recalibrated weekly or monthly, depending on the market.

Tags

# When a stochastic volatility model is calibrated?


# When a stochastic volatility model is calibrated?












In an Investment Bank, how often a stochastic volatility model is calibrated ? Is it calibrated daily ? Is it calibrated whenever a pricing is required ?

Thanks.

## Answer by AdB (score 5)

https://quant.stackexchange.com/a/43162

This depends largely on the model as well as the market, so there is no one-size-fits-all approach. Let us take the Stochastic Alpha Beta Rho (SABR) model, which has four parameters, as an example:

- $\alpha$, the initial instantaneous volatility of an ATM option. This is calibrated frequently and often intra-day due to the importance of ATM options.

- $\beta$, which describes the backbone of volatility. This is often chosen a priori and not calibrated.

- $\rho$, which describes the correlation betweent the Brownian motions driving the forward process and its vol proces, respectivelty. This is not calibrated as frequently - perhaps once a week/month depending on the market.

- $\nu$, which is the vol of vol. This is not calibrated as frequently - perhaps once a week/month depending of the market.

Have a look at the initial paper by Pat Hagan et al. for more information!

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.