Re: st: limiting instruments in an ivprobit

[Andy Ratto <andyratto@gmail.com>]

To clarify, in the probit analysis, I use an interaction term between
a dummy variable and the potentially endogenous variable.  When Stata
includes this in the first stage, it is attempting to predict the
endogenous variable with the interaction term, which I would not like
to do.

My attempt to limit which variables are used as instruments was done
with this part of the ivprobit notation, which the example in
stackoverflow also contains: (varlist2 = varlist_iv)
Even when I attempt to specify a specific variable as the instrument,
State still includes all independent variables as instruments in the
probit analysis.  Is there some way to specify what will serve as the
instrument that I am missing?

Here is the code I am using:
ivprobit failure rebpolwingnew logpop loggdpc democracy
c.HumanitarianAid#c.rebpolwingnew tim tim2 tim3 (HumanitarianAid=
drought) , first

The output lists the interaction term as an instrument, and I would
prefer to remove it as an instrument:
rebpolwingnew logpop democracy c.HumanitarianAid#c.rebpolwingnew tim
tim2 tim3 drought1

thanks,
andy



On Tue, Mar 26, 2013 at 3:14 PM, Nick Cox <njcoxstata@gmail.com> wrote:
> This is (emphatically) not my field, so this is more or less Esperanto to me.
>
> Two comments nevertheless:
>
> 1. The question on Stack Overflow did receive an answer.
>
> "In general all exogenous variables are always included as
> instruments. Usually instruments are picked for variables which are
> endogenous, but we can think (it follows from the mathematical
> derivation of instrumental variable estimation) that we need to choose
> the instruments for all the variables. Instruments for exogenous
> variables then are naturally themselves."
>
> I think you should explain why that is wrong or irrelevant, which
> could be the only grounds for not mentioning it.
>
> 2. The syntax for -ivprobit- makes explicit how to list instrumental variables
>
>  Maximum likelihood estimator
>
>         ivprobit depvar [varlist1] (varlist2 = varlist_iv) [if] [in] [weight]
>                  [, mle_options]
>
>
>     Two-step estimator
>
>         ivprobit depvar [varlist1] (varlist2 = varlist_iv) [if] [in] [weight] ,
>                  twostep [tse_options]
>
> The questioner on SO did not use that syntax, which may have been
> behind their puzzlement.
>
> (I edited "STATA" to "Stata" below.)
>
> Nick
>
> On Tue, Mar 26, 2013 at 6:57 PM, Andy Ratto <andyratto@gmail.com> wrote:
>
>> I am using ivprobit for an instrumental variable analysis.  By
>> default, ivprobit includes all independent variables as instruments.
>> Here is a post from two years ago that described the issue with
>> ivprobit and displays the output:
>> http://stats.stackexchange.com/questions/8082/why-is-stata-automatically-converting-regressors-to-instrumental-variables-in-iv
>>  There was not a solution to that post, and I was wondering if
>> anything has changed in two years in the options for ivprobit, or if
>> there are other methods in Stata to perform an instrumental variable
>> probit which limits which variables are used as instruments.
>>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/



-- 
=================================================
Andy Ratto
(678) 561-7248
andyratto@gmail.com
=================================================
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/