Make System from PanelData

[backhaum]

Dear all,

I am not quick sure, whether this posts relate to rather to Data Manipulation or to Estimation.

I am trying to estimate a 3SLS from panel data (EVIEWS 7) where my cross-sections (firms) are divided into different groups (industries).
The idea is to estimate a 3SLS with a general variance-covariance structure of the error terms of firms from the same industry (group).

I am struggling though to create a system from my panel data with the brand specific X variable for one group like this:

System1 (industry 1);

X(brand a) = c + beta * y….
X(brand b) = c + beta * k…

System 2 (industry 2);

X(brand c) = c + beta * f…
….

Any suggestions on how to build / generate this data structure?

Thank you very much for your help!

[EViews Glenn]

The problem that you are going to have is that the system object does not provide a way to define general correlation structures amongst subsets of the equations in the system. All that is allowed is a general correlation between all of the equations.

The best you can do (using built-in EViews) is to estimate industry specific systems. This would certainly produce consistent results, though does not attain the efficiency that you would get by imposing cross-equation restrictions in the general model, and you would have to handle the multiple sets of coefficients estimates for restricted coefficients in your specification.

[backhaum]

Thank you for the reply. This is also what I was planing to do.
I am struggling though to create exactly this System with my panel data structure?
Do I have to restructure my Data into a pooled form?

Thank you again.

[EViews Glenn]

I think so.