Choosing Sampling Frequency for GARCH Volatility Estimates

Summary

The document asks how to estimate monthly volatility from a year of five-minute stock prices, including whether to fit a GARCH-family model at the intraday or monthly frequency and how to scale the resulting volatility. Its only suggested procedure is to try several sampling frequencies and select the one with the lowest fitting error, giving root mean squared error as an example. This treats frequency choice as a model-selection question rather than prescribing one universal interval.

The reply does not explain how to define the target used to calculate fitting error, how to avoid evaluating on the same data used to choose the frequency, or how to convert intraday estimates into monthly volatility. It gives no evidence that the proposed selection rule works in this setting. Results may depend on the return construction, market microstructure noise, model specification, and whether volatility is stable over time; square-root-of-time scaling also relies on assumptions about returns.

Key ideas

• The question concerns estimating monthly volatility from intraday stock prices using GARCH or TGARCH.
• The reply suggests comparing several sampling frequencies using a fitting-error measure such as root mean squared error.
• The document does not specify the target, validation procedure, or scaling method needed to apply that suggestion.

Tags

# Volatility estimation: sampling frequency and scaling


# Volatility estimation: sampling frequency and scaling












I have a year long stock data sampled at 5 min frequency and would like to estimate monthly volatility using it. I am thinking using GARCH or TGARCH for volatility estimation. However, I am not sure if at frequency I should estimating volatility. After that should I scale volatility by square root of appropriate period?

## Answer by emcor (score -1)

https://quant.stackexchange.com/a/14981

You can try out a range of frequencies and then pick the one with the lowest fitting error (e.g. Root Mean Squared Error).

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.