> Salvati, Jean wrote:
>
> >>2) If x(t) is predetermined instead of endog in my model,
> there is no
> >>need to instrument for x(t-1) - x(t-2). Will Stata know this?
> >
> > It's your responsibility to tell xtabond whether x is
> predetermined,
> > endogenous, or exogenous. How to do this is explained in the doc.
> >
> > Exogenous regressors can (and should) be used as instruments (for
> > themselves). xtabond automatically does that.
>
> Sorry, I should have made that question more clear. Actually
> I meant does Stata know that there is no need to instrument
> for x(t-1) - x(t-2).
> Are you saying that Stata will instrument for all regressors?
There are 3 basic rules for instrumental variables estimation:
1) The number of independent instruments should be at least equal to the
number of parameters to estimate.
In that sense, all regressors must be instrumented--even though, in the
estimation procedure, an instrument is not tied to a particular
regressor.
2) Instruments are orthogonal to the error process.
An implication of this is that, if x is exogenous, then x is a valid
instrument.
3) The instruments should be as strongly correlated with the regressors
as possible.
An implication of this is that, if x is exogenous, then there's no
better instrument for x than x itself.
If you're not using your exogenous regressors as instruments, you're
throwing away very useful information.
Consider the following command:
xtabond y x1, lags(1) inst(z1 z2) pre(x2) pre(x3, endog)
x1 is exogenous and will automatically be used as an instrument. x2 and
x3 are respectively predetermined and endogenous, and they will be
instrumented using their own lags. z1 z2 are additional instruments that
you provide.
Jean Salvati
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/