Why Realized Volatility Data Can Improve on Basic GARCH

Summary

The document addresses why a basic GARCH model may be outperformed by realized GARCH or other realized-volatility models, especially when modeling intraday volatility. Its explanation focuses on information inputs rather than a strict difference between daily and intraday time scales. Basic GARCH uses the observed price or return series, while realized-volatility models can incorporate additional volatility measurements derived from finer-grained data.

Using more information can improve a model’s estimates or forecasts relative to a model that ignores those inputs. The answer is brief and does not give equations, a dataset, comparative results, or implementation guidance. It therefore offers a useful conceptual distinction, but does not show that GARCH is inherently unsuitable for intraday use or that realized measures will always improve performance; those outcomes depend on available data and model evaluation.

Key ideas

• The distinction between basic and realized GARCH is framed around available inputs, not solely the sampling interval.
• Basic GARCH uses the observed price or return series.
• Realized-volatility models can incorporate additional measurements of volatility.
• Additional information may improve performance, but the document gives no empirical comparison or universal guarantee.

Tags

# Why a model like GARCH is only good for daily volatility and not for intraday volatilities?


# Why a model like GARCH is only good for daily volatility and not for intraday volatilities?












I´m currently looking to implement an intraday volatility model and I´m new at the quant world and I learned how superior is GARCH family is for daily volatilities, but in the research stage I found out that GARCH daily model is not good for intraday volatilities. Instead, they present a realized GARCH or use realized volatilities models, so I was wondering if someone can give a simple explanation of why is that?

I would be really thankful for your answer.

## Answer by Richard Hardy (score 2)

https://quant.stackexchange.com/a/58360

The main difference between GARCH and realized GARCH models applied on daily or intradily data might be not the time period but rather the data availability and its use. A GARCH model uses very little information, namely, only the observed price or return series. Often it squeezes out quite good results from it. When additional information such as data on realized volatility is available, other models such as realized GARCH can be built yielding better results. Thus, models that incorporate this additional information can beat GARCH models that do not make use of it.

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.