EViews.com

Hi,

Previous posts in this forum mention that it is not possible to estimate VAR/VECM models with panel data however, the posts are relatively old.
I was wondering whether anyone could tell if it possible to set up a VAR/VECM model with panel data in Eviews 9 and observe individual results for panels.

I have already set-up a VAR model with my panel data though I am not sure whether Eviews understands that it is dealing with panel data and maybe it is producing results as if my data was time-series.

Any kind of help is welcome.

Thanks.

Nothing has changed - it still estimates on the stacked data, but lags do not cross seams.

EViews Gareth wrote:Nothing has changed - it still estimates on the stacked data, but lags do not cross seams.

Hi Gareth,

Thanks for your answer. Just to get this right. Does that mean that if I set up a VAR or VECM with panel data, Eviews will simply calculate results for the whole data set and ignore the fact that I have included different panels within my data?

Yes. But lags won't cross seams.

EViews Gareth wrote:Yes. But lags won't cross seams.

But since I am looking to see whether variables Y or Z affect X, do you think that setting up VAR/VECM models for two different panels (ex. Panel A: European countries, Panel B: Asian countries) and then looking at the different coefficients of the two panels is incorrect? (i am not interested in specific country results within panels-just differences between panels as a whole)

Sorry for the multiple questions-beginner here...

I've written two different (and rather user friendly) Panel VAR programs:

> using Arellano-Bond equations: https://github.com/omercadopopular/cgoe ... eneralized).prg

> using Pedroni's heterogeneous Panel VARs: https://github.com/omercadopopular/cgoe ... edroni.prg

To cgoes:

Very interesting and useful programs! Thanks for these links.
Unfortunately, in the last code (panel VAR) bootstrapping does not work in Eviews 9.5 and 10 (when !boot = 1, Eviews crashes ).

Which one? GMM/IV or Pedroni? I'm still working with Eviews 7, but reportedly it works with Eviews 9. Maybe the new update made it incompatible. Would you care to report the error message?

Pedroni. I have tried only this one on Eviews 9.5 and 10. Program makes all calculation and produce three graphs with impulse responses but when starting bootstraping simply stop working without any specific message conserning code error. I suposse that something have change a little bit in Eviews commands after versions 7 and 8.

I'll get access to the new eviews, work on it in the coming weeks and post it back here.

Just one suggestion: you can easily make Eviews add-ins based on your codes. It will inrease the usage of them and citation of your papers.

Gareth is helping me with that. Thanks!

igor wrote:Just one suggestion: you can easily make Eviews add-ins based on your codes. It will inrease the usage of them and citation of your papers.

Just an update. I've corrected the glitch and now it's compatible with Eviews 10:
https://github.com/omercadopopular/cgoe ... views%2010).prg