Page 1 of 1
Panel IV inputting instruments
Posted: Sun Mar 15, 2009 4:23 am
by kian
Hi all,
I'm running a panel two-stage-least-square and I wanted to instrument for two different variables that I think are endogenous. However, I do not want to use the same set of instruments for both variables...How do you input the instruments in the Eviews TSLS so as to define which IVs you want to use to instrument for each particular endogenous variable?
Thanks :)
Re: Panel IV inputting instruments
Posted: Mon Mar 16, 2009 3:33 am
by kian
Does anyone have any ideas? Please???
Re: Panel IV inputting instruments
Posted: Mon Mar 16, 2009 9:01 am
by EViews Glenn
Instrumental variables doesn't really work that way.
Instruments have to be valid for all variables and they cannot be restricted to apply to particular endogenous variables.
Re: Panel IV inputting instruments
Posted: Mon Mar 16, 2009 10:15 am
by startz
Instrumental variables doesn't really work that way.
Instruments have to be valid for all variables and they cannot be restricted to apply to particular endogenous variables.
Sometimes people use subsets of valid instruments, essentially using a different first-stage for particular endogenous variables. Or at least that's my memory.
Re: Panel IV inputting instruments
Posted: Mon Mar 16, 2009 12:12 pm
by EViews Glenn
In general, it's not recommended since you don't get the right orthogonality conditions unless there are very specific conditions on the system (I believe you need triangularity of the system). So I guess I should have more precisely said that it's not recommended unless you have very specific reasons for doing so. Apologies.
In any event, it's not something that we do not support as a in-built feature. You could do it yourself by performing the first-stage regressions, but your second-stage standard errors will require adjustment.