Using GARCH to Model Forecast Errors Alongside Time-Series Forecasts

Summary

The document clarifies the role of GARCH in forecasting a time series such as a gas price. GARCH models changing conditional variance, so it is useful for describing and forecasting volatility rather than directly supplying the expected level of a series. A separate forecasting model can produce the central price forecast.

One common setup is to fit an ARIMA model to the series and then apply GARCH to its errors. The first model forecasts the series, while the second describes how forecast error variance changes over time. The exchange gives no performance comparison or detailed model specification, and it does not explain how the paper combines its methods. The takeaway is therefore about division of labor between mean and variance models, not evidence that this specific combination will improve every forecast.

Key ideas

• GARCH models time-varying conditional variance rather than the series’ expected level.
• A mean-forecasting model such as ARIMA can be paired with GARCH for its residuals.
• The GARCH component can help describe changing forecast uncertainty.
• The exchange does not provide evidence that the combination improves forecasts in all settings.

Tags

# Can I do a GARCH model to forecast a time series?


# Can I do a GARCH model to forecast a time series?












I read this paper

https://research.aston.ac.uk/portal/files/240393/AURA_2_unmarked_Energy_demand_and_price_forecasting_using_wavelet_transform_and_adaptive_forecasting_models.pdf

the two authors forecasts one day ahead gas price using, between the others, a GARCH model. How does this model works? Isn't a GARCH model useful just to forecast volatility? thank you!

## Answer by Richi Wa (score 1, accepted)

https://quant.stackexchange.com/a/17130

You are right - GARCH model models volatility. They write: " The GARCH [27] can be used to model changes in the variance of the errors as a function of time."

What people often do is to fit an ARIMA model (that can be used to forecast a time series) and apply a GARCH model to the errors (which gives you a feeling for the forecast error). See Hyndman and Athana­sopou­los for a good, free online book on forecasting.

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.