Hi everyone!

I'm estimating a dynamic panel model using the xtdpdsys command in Stata 11. The problem I find is that if you include a dummy for a panel variable (country, factory, individual) xtdpdsys assigns explanatory power to this dummy.

According to the idea of the estimator it should obviously not do this as observation entity specific fixed effects are an integral part of the estimator - and they should absorb the effect, thereby making an additional dummy superfluous and perfectly collinear to the fixed effect. To make this point clear I have this example using Stata data:

use

http://www.stata-press.com/data/r11/abdataxtset id year

gen seven=0

replace seven=1 if id==7

xtabond n L(0/2).(w k) yr1980-yr1984 year seven, vce(robust)

xtdpdsys n L(0/2).(w k) yr1980-yr1984 year seven, vce(robust)

xtabond is doing what it should do and omits the entity specific 'seven'

xtdpdsys estimates some coefficient it shouldn't.

Is this already a known problem? Or do I misunderstand something?

(My application is a model in which I have dummies for country groups (a b c). These dummies are interacted with a variable x.

The problem I find is that if I estimate a subsample for which a==1. I find that a*x has a negative coefficient. But if I estimate the whole sample the coefficient of a*x is (falsely) positive.)

Cheers Lara