Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: ivprobit with plot level data and household endogenous variable


From   "Camille Saint-Macary" <Camille.Saint-Macary@uni-hohenheim.de>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: ivprobit with plot level data and household endogenous variable
Date   Mon, 30 Jun 2008 10:21:29 +0200

Hi all,

I have a question concerning the use of ivprobit.

I'm modeling the adoption of certain agricultural technology on land at the
plot level. One of my regressor is a credit constrained dummy variable which
I suspect to be endogenous and I want to instrument. 

1. if I use ivprobit, --> ivprobit Y X (D=Z)
It seems that the instrument equation is estimated with OLS--> which gives
me a predicted probability to adopt (2nd stage) ranking from -4.12 to 2.87,
I guess this is not good, the rest of the model however is fine...

2. As suggested in previous posts, I then use ivprobit and use the predicted
variable as an instrument:
	1. probit D Z X 
	2. predict D*
	3. ivprobit Y X (D=D*)
The results are quite different from the first regression and Wald test
rejects endogeneity.

3. One problem I suspect is that I'm running all at the plot level
(household have different number of plots) but my endogenous variable is a
household level variable as well as Z my instrument--> households with high
number of plots are overweighted... But there are plot specific variables in
X --> so for identification condition I need to include all X in the first
stage right? Or should I run the first regression at the household level and
include the predicted probability as a regressor?

Many thanks for your suggestions!

Camille

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index