Applying Realized Kernels to Short-Horizon Volatility Estimation

Summary

The document asks whether realized-volatility methods developed for daily estimates can be applied to a much shorter interval, using finer observations such as one-second returns to estimate volatility over ten minutes. Its specific concern is whether realized kernels, which address market microstructure noise, remain valid at that horizon.

The text poses a research question but gives no proposed estimator, derivation, evidence, or answer. It therefore offers a useful prompt about the interaction between sampling frequency, estimation horizon, and noise correction, rather than a validated method. Any conclusion about kernel validity would need to account for the assumptions behind the estimator and the amount and structure of data available within the short interval.

Key ideas

• The document asks whether daily realized-volatility methods can estimate volatility over much shorter intervals.
• It raises the use of fine-grained returns to estimate volatility across a short window.
• It specifically questions whether realized kernels still correct microstructure noise at that scale.
• The document provides no derivation or empirical evidence resolving the question.

Tags

# Intraday Volatility using Realized Kernels


# Intraday Volatility using Realized Kernels












Since the papers about realized volatility calculate daily volatility out of intraday data, is it also possible to apply same methods to calculate e.g. 10 minutes volatility by smaller sampled data e.g. by 1 second returns? I could not find a single paper about that. Specifically, I'm interest if methods which remove microstructure noise, such as kernels, are still valid.

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.