[Openspace] Allowing for exogenous changes in the W matrix in a panel setting?

[Mark Roberts]

Hello,

Sorry that this question isn't specifically GeoDa related, nevertheless I 
would be grateful if anyone could provide advice (it is spatial 
econometrics related!)...

In particular, I was wondering about the feasibility of estimating a 
spatial (lag) panel data (fixed effects) model in which the W matrix is 
subject to *exogenous* changes between time periods. Estimation of the 
standard spatial lag FE model with a time invariant W matrix is fine. 
However, from an identification viewpoint, is it ok to allow the W matrix 
to change between periods by creating a NT*NT matrix with period specific W 
matrices on the diagonal? In this context, I would be thinking of keeping 
the spatial lag parameter constrained to be the same across the time 
periods...

Thanks, Mark