Hi,
I am performing fixed effects and gmm panel regressions with xtreg and
xtabond. One of the right-hand variables is calculated with another
model (a simple linear regression model). Is the fixed effects model and
the gmm model still valid by using such a proxy variable, or do I have
to take into account that these models are not valid anymore. Possible
evidence for non-validity may be the fact that simple ols regressions
perform as well as the more complicated procedures, which means that
there is no efficiency gain in using the latter. If this is the case,
what can I do instead of these models, or is there a way to account for
this kind of proxy variables? Is there any literature on proxy variables
and their problems in general?
Thanks,
Kai
