|
|  |
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: Re: instrumental variable panel probit
Well, with 2 periods I don't think that you could identify the unobservable
(FE or RE) in the nonlinear framework (probit). Maybe, there are estimations
with T=2 and FE using Conditional Logit, but the endogeneity of your RHS is
a problem to define the two step "selection" argument. Moreover, if the
model is dynamic (as the lag of your dependent variables is the right hand
side) it seems that the problem is even harder. For a large T, there are
suggestions for the bias generated in the dynamic model. Greene (2002):
http://www.stern.nyu.edu/~wgreene/nonlinearfixedeffects.pdf, Arellano and
Hahn (2005): http://www.cemfi.es/~arellano/ah-r3.pdf, Arellano and
Hahn(2006): http://www.cemfi.es/~arellano/arellano-hahn-paper2006.pdf with
appendix: http://www.cemfi.es/~arellano/arellano-hahn-appendix2006.pdf and
Fernandez-Val (2007)
http://people.bu.edu/ivanf/wp_files/panelprobit_feb10_2007.pdf. But T needs
to be at least 4.
R.
----- Original Message -----
From: "marco stampini" <[email protected]>
To: <[email protected]>
Sent: Monday, March 12, 2007 2:27 PM
Subject: Re: st: Re: instrumental variable panel probit
Thanks Rodrigo.
I am sorry but I do not see how treatreg takes care of random (or fixed)
effects.
Furthermore, it seems that it requires a continuous variable, while both
my dependent and endogenous variables are dummy.
Marco
----- Original Message ----
From: Rodrigo A. Alfaro <[email protected]>
To: [email protected]
Sent: Monday, March 12, 2007 1:43:39 PM
Subject: st: Re: instrumental variable panel probit
see -treatreg-
R.
----- Original Message -----
From: "marco stampini" <[email protected]>
To: <[email protected]>
Sent: Monday, March 12, 2007 12:55 PM
Subject: st: instrumental variable panel probit
Hello.
I need to estimate the effect of a program on the likelihood to grow a
certain kind of crop. As participation in the program is endogenous, I
would like to use instrumental variables
ivprobit crop (program=z) x
where x are other household characteristics, and z are instruments.
As I have two points in time, I would like to consider some kind of
household effect. I realize I can't use fixed effect, so I could use
xtprobit, pa
However, if I understand correctly, this does not allow instrumenting.
Question 1: Does anybody know a way out of this problem?
Question 2:
What if estimated
ivprobit crop2 (delta_program=z) crop1 x1
where variables with suffix 2 refer to the second period, and those with
1
to the first period (initial characteristics) and
delta_program=program2-program1
I suspect the estimation would be biased, because of the inclusion of the
lagged dependent variable... Can anybody confirm?
Thank you very much in advance
Marco
____________________________________________________________________________________
No need to miss a message. Get email on-the-go
with Yahoo! Mail for Mobile. Get started.
http://mobile.yahoo.com/mail
*
* 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/
*
* 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/
____________________________________________________________________________________
8:00? 8:25? 8:40? Find a flick in no time
with the Yahoo! Search movie showtime shortcut.
http://tools.search.yahoo.com/shortcuts/#news
*
* 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/
*
* 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/
