Fixed Effects

[nicojaq]

Hi all,

I am using Eviews 7 and have an issue with a panel fixed-effect regression. I have used cross-section fixed effects since they seemed intuitive in my setting and the tests were highly significant. However, when I compute forecasts from this regression, for a few of my cross-sections (4 out of 20), the forecasts are computed without the fixed effects, resulting in the forecast showing an important bias against actual data. I don't understand why Eviews "ignores" these fixed effects (I've checked and they are non-zero) for these cross-sections.
Any help would be really appreciated.

Regards.

[EViews Gareth]

Could you post your workfile?

[nicojaq]

Hi Gareth,

Here is the file. I recommend you look at eq1 on the panel_usd page of the workfile and at the graph_forecast and you'll see what I mean. At first I thought it was technical issue, but now I am wondering if it's not an econometric issue... am I pooling cross-sections that are too heterogeneous maybe?

Thanks for any opinion on the issue.

[EViews Gareth]

Which cross-sections do you believe are not using the fixed effects?

[nicojaq]

I think in this setting THB, TWD, ZAR, PLN, RON, HUF, RUB and ILS don't use the fixed effects.

[EViews Gareth]

I see what you mean. We'll investigate.

[nicojaq]

Hi Gareth,

Did you find anything during your investigations?

Thanks.

Regards

[EViews Gareth]

Yep, this issue was fixed in a patch a while back.

[nicojaq]

a patch? sorry I don't understand.... what do I have to do? Thanks

[startz]

Update your copy of EViews using the command on the Help menu.

[nicojaq]

Hi Startz,

I have updated my version but I still get this issue. Any idea?

Thanks.

[EViews Gareth]

If you click on Help->About EViews, what is the build date of your copy?

[nicojaq]

May 122 2010. Does it mean it hasn't updated?

[startz]

I believe the current version is October 14, 2011

[EViews Gareth]

Download the upgrade installer:
http://eviews.com/download/download.html