st: simultaneously estimate MNL and Logit

[Igor Ramzak <igor.ramzak@gmail.com>]

Hi,

I want to estimate a recursive model. The dependent variable Y1 is a
dummy variable that depends on a vector of exogenous variables and a
categorical variable, represented by 4 dummies to take into account 5
distinct groups. My main interest goes to the effect of each of these
categories on Y1.

But… the categorical variable might be endogenous. For this reason, I
would like to estimate a multinomial logit to explain the categorical
variable as a function of the exogenous variables that explain Y1 and
some instruments that are not correlated with Y1 but explain the
categorical variables.

Is there a command in Stata that makes this possible? I guess I could
use -biprobit- if only my endogenous variable didn’t contain 4
categories instead of 2. I know the first step can be estimated by
-mlogit- and the second step with -logit- but how to estimate both
simultaneously? I guess this will be necessary in order to control for
the covariance of the error terms of both equations.

Any ideas how to estimate this? Could the -sem- command be of any assistance?

Thanks for your consideration.

Igor Ramzak

*
*   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/