Re: st: partial observability biprobit + IV ???

[Patrik Morgetz <pmorgetz@gmail.com>]

Hello everyone

Does anybody has any idea on this topic?

http://www.stata.com/statalist/archive/2009-09/msg00814.html

I found it on stata list archives but without any reply.

Well, my colleague and I have a similar problem. We are working on a
partial observability bivariate probit model that we want to use to
understand the determinants of individual's choice to participate in a
cards game that requires exactly two people to start (we observe who
participates and who does not, that is, the joint decision but not the
individual decision of every player). However, we have quite a similar
problem, the wealth of every individual enters both decision
equations, however, there might be endogeneity due to simultaneity in
this variable, since the game may affect the individual's wealth (if
they win, they earn money).

So, what we are doing until now is similar to what was asked a couple
of years before here in the stata list, a two-stage procedure
regressing wealth on a couple of available instrumental variables and
then including the fitted values in the bivariate probit model. To
deal with the variability of the first stage, we are trying to
estimate standard errors using bootstrapping methods. However, we are
not sure about this procedure, since we have not found any reference
in the literature that support it for this kind of models.

Thanks in advance for any idea or literature reference that might be
useful but I have missed so far.

Patrik
*
*   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/