Volatility Models for High-Frequency Returns and Order-Book Data

Summary

The document considers forecasting returns and volatility from data sampled at intervals of one second or less. It notes a limitation of relying on ARMA and conventional GARCH alone: these time-series models do not directly represent order imbalance or order flow, which the questioner wants to incorporate alongside price history. The response points to long-memory alternatives, including ARFIMA for returns and FIGARCH or fractionally integrated GARCH approaches for volatility. It also mentions a multiplicative component GARCH specification, which can combine a mean model such as ARMA with volatility components for high-frequency applications.

The answer offers candidate model families, not a comparison or recommendation supported by reported tests. It refers to an implementation in an R package, but gives no performance results, data description, forecasting horizon, or guidance on how to add order-book variables. Model choice therefore remains an empirical question: the cited suggestions do not show which method forecasts best for a particular market, instrument, sampling scheme, or latency constraint. The discussion is a starting point for research rather than a validated trading approach.

Key ideas

• Very high-frequency returns and volatility may exhibit persistence that standard models do not capture well.
• ARFIMA and fractionally integrated GARCH variants are suggested for long-memory behavior.
• A multiplicative component GARCH model can pair volatility components with a mean model such as ARMA.
• The answer does not demonstrate how to incorporate order flow or compare forecast performance.

Tags

# High frequency price forecast model ARMA GARCH or another?


# High frequency price forecast model ARMA GARCH or another?












Can you reccomend model for high frequency data (1 second and less) (returns and volatility forecasting)? Most papers use ARMA, GARCH etc in 1 minute and lower time frame. PROBLEM ARMA does not know nothing about order imbalance and order flow correlation so i looking for model which can combine order book and time series forecasting model. Efficient estimation of volatility Zumbach 2002

Order Book Slope and Price Volatility , Petko Kalev EGARCH aproach. Anybody used it ?Or know something better? What do you use?

I know there are a lot or papers and question already about volatility/But just want to know expert opinion what to choose for high frequency?

## Answer by Alejandro Andrade (score 1)

https://quant.stackexchange.com/a/21947

HF data have a lot of auto correlation so common models to deal with this problems are ARFIMA, FIGARCH, Fractional Integrated GARCH. Engle recently propose the multiplicative components GARCH for high frequency data, which can include a mean model like and ARMA. In this post they explain how to implement it in R with the `rugarch` package, it takes some time to run but it works. http://unstarched.net/2013/03/20/high-frequency-garch-the-multiplicative-component-garch-mcsgarch-model/#comment-8309

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.