Panel Estimation - Cross sections

[LJG]

If you have an unbalanced panel regression, is there a way to include future cross sections for a forecast (cross sections are vintages in this example)? The workfile is set up to include future cross sections (vintage) but the forecast isn't currently using them.

Thanks for your help :)

[EViews Gareth]

Sure. The forecast will occur over whatever sample you tell it to.

[LJG]

The forecast being produced over the correct sample but is assuming only the existing vintages in the model estimation. The workfile is set up to include future vintages but these are not populated in the forecast. Is this because we don't have any fixed/random effects for these vintages? If so, is there a way of extending these effects for future vintages without re-estimating the mdoel?

Thanks

[EViews Gareth]

I think you'll need to provide your workfile

[EViews Glenn]

There is no estimate of the fixed or random effect for a cross-section that wasn't there in original estimation so those terms are set to zero in forecasts. Out-of-sample forecasts for existing cross-sections will use the estimated effects.

[EViews Gareth]

But, further to Glenn's point - you won't receive an estimate of the fixed or random effects, but you'll still receive a forecast value due to the other terms.

[LJG]

Is there a way we can 'force' E-Views to use future cross-sections in the forecast output? The workfile contains values for the drivers of the model for future vintages but E-Views doesn't produce a forecast on these cross-sections. Thanks

[EViews Gareth]

Yes, it does. There is no need to force it - it does it every time.