Page 1 of 1

Panel estimation in system window

Posted: Sat Mar 13, 2010 5:56 pm
by Bigbrotherjx
Am I correct in thinking that if I constrain the coefficients (apart from the intercept) across each equation in a SUR framework to be identical, I effectively have a panel?

What I'm really trying to achieve is to see if my SUR (with lagged dependent variable) can be restricted to a dynamic panel structure (my data has a time dimension much larger than the cross-section).

Also, am I correct in thinking that the point of GMM estimators like Anderson-Hsiao and Arellano-Bond is to avoid the bias that arises from the correlation between fixed effects on the lagged dependent variable and the idiosyncratic error?

However, if a SUR with cross-equation coefficient restrictions becomes a dynamic panel, I wouldn't even need fixed effects right? Surely I can account for the fixed effects by letting the intercept term differ across equations, so there is no Nickell bias?