Forecast of variance by GARCH model

[Lyubov]

I think, I understand the situation.

For dynamic foresast EViews uses values of forecast S.E. as epsilon in equation for variance. So, we can do forecast for many steps, even if we don't have real values of y for forecast's period.
For static forecast it uses series RESID (= real y - forecast y) as epsilon. And if we don't have real values for forecast's period, we can make static forecast only for one step.

Am I right?

O, no... I really don't know... What values EViews use as epsilon in equation for variance, then it make dynamic forecast for period on which we don't have real values of Y (and there are no RESID for this period)...

Can anybody help me, please?

[EViews Glenn]

As in all AR dynamic forecasting, the conditional mean of the residual, which is zero. This is why dynamic AR forecasts have the original bump from the residual(s) in the presample period, then have effects which die out at the AR rate...

[Lyubov]

As in all AR dynamic forecasting, the conditional mean of the residual, which is zero. This is why dynamic AR forecasts have the original bump from the residual(s) in the presample period, then have effects which die out at the AR rate...

As i understand, for forecast's period (for which real values of Y are absent) EViews use epsilon = E(epsilon) = 0 in equation:


But then i compute difference between left and right parts of this equation (i use coefficients, estimated by model; garch terms from series GARCH(optional); epsilon=0) i don't get 0 (or values near 0).

[EViews Glenn]

I'm not entirely sure what you are doing.

All I was saying is that in your expression if you are starting dynamic forecasting at t=1, e_0 is the actual residual, and e_1, ..., e_s are all set to 0.

[Lyubov]

And for static forecasting (when we have real values of Y for forecast's period) EViews use epsilon = actual residuals ?

[joannytan]

Hi,

I think I have the same issue as you mentioned. I use EGARCH (3,3) to estimate my forecast (by clicking on forecast tab) and Eviews returns zero for my one step ahead forecast. When I use the estimation equation and calculate using Excel I have different result which is not zero. I have tried on many data and the same thing happen. However, when I try EGARCH(1,1) the problem does not exist. Anyone has any idea?

My solution now is to store the garch variance through makegarch function and calculate using the estimation equation. It works fine with the same value as I have in Excel. But I do really want to know why this would happen or the forecast tab function is only for the forecast using lag 1? (eg: resid(-1) , garch(-1)).

Thanks.