EViews does not use any panel estimation techniques for VARs. It simply estimates a standard VAR on the stacked data, with the exception that lags behave properly (they do not cross-across cross-sections).
Thank you for your reply. If I understand well Eviews ignores the panel structure doing a pooled estimation. In terms of algebra the parameters are computed as
beta= (X'X)^(-1)X'Y
where X is an nTxK matrix and Y is nTx1
So fixed effects are ignored and estimates of beta can be biased.
However, one could demean the data taking the deviation from means for different units. This would be (almost) equivalent to make a panel VAR with fixed effects. I say almost because one would ignore the possibility of cross-section heteroschedasticity. But one could deal with this issue expressing the demeaned data in terms of common factors.
So working with demeaned variables (ie x(i,t)-x(i) where x(i) is the individual specific mean ) purged from common factors would allow to interpret the IRFs "correrctly" as in a proper panel VAR which controls for individual specific heterogeneity possible cross untis correlation.
Is this interpretation correct? If not how one can I interpet the IRFs
It would be nice to see the algebra of what Eviews does. Ihave not found an explanation in the manuals; but please correct me if I'm wrong
Thank you again for your help on this important topic
Alfonso