st: IVTobit with three endogenous variables

[Phil1899@gmx.de]

Hello,

I want to estimate the following model:

Y = bx1 + bx1^2 + bx2 + bx3 +bx1*x2+ e

where x1 is potentially an endogeneous variable (reverse-causality). Due to the squared term and the interaction I thus have three endogenous variables. I estimated this equation using 2SLS IV where I used two instruments for x1 (and calculated the squared term and interaction term accordingly). Several tests reveal that the instruments fulfill both required criteria for IV estimation.

The problem is that my dependent variable has a (rather small) cluster at zero and a few negative observations. Theoretically, it might be possible to argue that the negative values can be recoded to zero so that Tobit estimations could be applied. That is also what I tried. Thus I want to use the IVTobit command. In particular, I want to use the ML estimator of IVTobit as this allows robust estimations. However, Stata goes through many iterations not coming to an end. I read on the list that others had similar problems when they used ivtobit with more than one endogenous variable. It was suggested to use the two-step option. This also works in my case, but as mentioned, there is no robust option in that case.  

(see http://www.stata.com/statalist/archive/2008-04/msg00197.html for the previous discussion)

I would be interested in the reason of the problem. Is this really something that always happens when there are more than one endogenous variable? 

I would appreciate any help!
Best
Phil    

-- 
GRATIS für alle GMX-Mitglieder: Die maxdome Movie-FLAT!
Jetzt freischalten unter http://portal.gmx.net/de/go/maxdome01
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/