Hi,
Thank you for the hints, but my problem is neither a question of ivreg nor one of bi-probit, and let me explain it more clearly:
1. I have longitudinal data ( an individual survey in which the date of birth for each child is recorded) and a community counterpart for which the year when each health facility offering family planning services surveyed was opened. For each woman, dates of birth are observed (noted) for each year since 1967 when expansion of health facilities began, including 1984 when the population policy was introduced, ending in 1998 when the survey was conducted. I merge the two datasets to obtain a panel as indicated above.
2. My problem is to estimate the effects (impact) of various factors (such as whether the health facility providing FP services lives within 5 kms of the community, as well as the introduction of the 1984 population policy) on fertility (whether a woman had a birth during each year of observation).
3. I therefore set up the following two equations to handle unobserved heterogeneity (fixed effects and random effects) at the level of the community and at the individual level:
Equation 1 (Binary logit): P(having an annual birth)= effects of individual factors + effects of community factors (e.g. clinic is located within 5 Kms of the community)+ effect of introduction of 1984 population policy (which is the interaction of year dummy and whether FP facility was open at that time) + .....
Equation 2: (Binary logit): P(FP clinic is located within 5 Kms of the community)= effects of community characteristics + ........
4. Since the two dependent variables are non-linear, I think that I should not use the instrumental variables technique which is for continuous dependent variables.
5. In the final analysis, I think that it is a question of solving (integrating out the random unobserved factors) the two equations with (full information?) maximum likelihood.
How to do it in stata?
________________________________
De: [email protected] de la part de Partha Deb
Date: jeu. 2007-09-20 23:55
�: [email protected]
Objet : Re: st: solving for unobserved hetereogeneity with two simultaneous equations
David,
I'm assuming that "FP clinic is located within 5 Kms of the community" is in
equation 1, i.e., that is your endogenous treatment. If so, you have 2
choices that capture the spirit of what you want to do
1. Estimate a bivariate probit (-biprobit-). The advantage is that the model
respects the binary nature of your dependent variables. The disadvantage is
that it makes a parametric assumption about the distribution (which, of
course, you did with the logit).
2. Estimate a linear simultaneous equations system (-ivregress-, -ivreg2-).
It is not designed explicitly for binary variables, but is less parametric and
is not inappropriate.
If you aren't thinking about endogeneity, you can use -biprobit- or -sur-.
Good luck.
Partha
Ojakaa David wrote:
> I have two equations that try to control for unobserved heterogeneity.
>
> Equation 1 (Binary logit): P(having an annual birth)= effects of individual factors + effects of community factors (for example access to FP clinics)+.....
>
> Equation 2: (Binary logit): P(FP clinic is located within 5 Kms of the community)= effects of community characteristics + ........
>
> What is the stata code that I should use to solve for the two equations simultaneously?
>
> *
> * 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/
--
Partha Deb
Department of Economics
Hunter College
ph: (212) 772-5435
fax: (212) 772-5398
http://urban.hunter.cuny.edu/~deb/
Emancipate yourselves from mental slavery
None but ourselves can free our minds.
- Bob Marley
*
* 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/
<<winmail.dat>>
