RE: st: Question about binary endogenous variable

["Santos Silva, J.M.C." <jmcss@essex.ac.uk>]

Dear Anna,

You may also want to consider a variant of the second method described in 
the post Alfonso mentioned: First, estimate a binary choice model where your
binary endogenous variable is regressed on the instruments (including the
exogenous regressors), and save the predicted probabilities. Second, do a 
standard 2SLS estimation with ivregress (or ivreg2) using as instruments for 
your binary endogenous variable the instruments used in the binary choice 
model as well as the fitted probabilities obtained from it.

This approach was used successfully in (see especially the first 2 lines on page
291):

Windmeijer, F.A.G. and Santos Silva, J.M.C. (1997), Estimation of Count Data 
Models with Endogenous regressors; An Application to Demand for Health Care, 
Journal of Applied Econometrics , 12(3), pp. 281-294.

All the best,

Joao

> Hello fellow Statalisters,
> 
> I would like to fit a regression model with a binary endogenous variable 
> (takes on 0 or 1 values). I looked at the examples for ivregress and they 
> only use
> 
> continuous  endogenous variables. What is the best way to solve this problem? 
> Would it be sensible solution just to use ivregress? Maybe I should use 
> probit for
> 
> the first stage and then use regression with the fitted values for endogenous 
> variable and then manually adjust standard errors? Any thoughts are highly
> 
> appreciated.
> 
> Anna Veksler
> *
> *   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/

*
*   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/