Page **2** of **2**

### Re: GARCH(1,1) Forecast Series

Posted: **Thu Apr 12, 2018 2:48 pm**

by **Hanabi**

Hello everyone! I am currently trying to make an out of sample forecast using 10 years of daily data. I have divided my entire sample in two sub-samples the first ranging from the 27th of march 2008 and ending on the 29th of December 2017 and the second one from the 29th of december 2017 to the 27th of March 2018. The latter sample is the one that I am trying to forecast to then compare to the actual forecasted values. I have fitted and AR(1) EGARCH (1,1,1) model in my first sample. However, when I hit the forecast button and I click on my forecasted period for the mean, I get the same value for the entire forecasted period. That doesn't seem right to me. Moreover, when I try to manually calculate the way those numbers have been produced in eviews I get a completely different result. Any ideas regarding what I might be doing wrong?

### Re: GARCH(1,1) Forecast Series

Posted: **Thu Apr 12, 2018 3:26 pm**

by **EViews Gareth**

If you don't have anything in the mean equation other than a constant, and you're doing a dynamic forecast, the forecast of the mean will be (wait for it....) a constant!

### Re: GARCH(1,1) Forecast Series

Posted: **Fri Apr 13, 2018 6:12 am**

by **Hanabi**

Thanks Gareth!

I appreciate the reassurance! However, I still have an issue with the discrepancy between the values that I have calculated manually and the ones that eviews provide. I am pretty sure that the specification I am using to calculate the values manually is correct, any ideas on why that might be the case?

### Re: GARCH(1,1) Forecast Series

Posted: **Fri Apr 13, 2018 7:22 am**

by **EViews Gareth**

The long term forecast is just the constant from the equation output, so calculating it is really easy.

### Re: GARCH(1,1) Forecast Series

Posted: **Fri Apr 13, 2018 8:15 am**

by **startz**

The initial post says there is an AR(1) component. Maybe the OP could clear up exactly what the model is and how the forecast was done in EViews.