Mark and Carola,
You can't do an overID test to check if the excluded instrument Z
violates the exclusion restriction is violated if there is only one
variable in Z and one endogenous regressor--but if you have other
exogenous included variables in a matrix X then it natural to assume
that the variable in Z interacted with variables in X can also be
considered as excluded instruments, so you can try putting them in and
running an overID test. Start with a small subset (even just one) of
the other exogenous included variables interacted with Z, where your
prior is strongest that the other exogenous included variables don't
have any neglected nonlinearity etc. If Z is continuous, it is often
natural to assume not only that E(Z*e) is zero but that E(Z^2 *e) is
zero etc. so you may have additional excluded instruments there as
well.
On Thu, Jan 1, 2009 at 7:27 PM, Schaffer, Mark E <[email protected]> wrote:
> Carola,
>
>> -----Original Message-----
>> From: [email protected]
>> [mailto:[email protected]] On Behalf Of
>> Carola Herrera
>> Sent: 28 December 2008 22:51
>> To: [email protected]
>> Subject: st: Sargan and Hansen tests
>>
>> Dear all,
>>
>> Do the Sargan and Hansen J tests only apply if one has more
>> IVs than endogenous regressors?
>
> Yes.
>
>> These are overidentification
>> tests, right?
>
> Yes.
>
>> How can I check if the IV violates the exclusion restriction
>> is violated if there is only one IV and one endogenous
>> regressor? No way I can show this is the case, right?
>
> Correct.
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
